drm/amdgpu: Add support for CIK parts

This patch adds support for CIK parts.  These parts
are also supported by radeon which is the preferred
option, so there is a config option to enable support
for CIK parts in amdgpu for testing.
Acked-by: NChristian König <christian.koenig@amd.com>
Acked-by: NJammy Zhou <Jammy.Zhou@amd.com>
Signed-off-by: NAlex Deucher <alexander.deucher@amd.com>

drivers/gpu/drm/amd/amdgpu/Makefile

@@ -18,6 +18,9 @@ amdgpu-y += amdgpu_device.o amdgpu_kms.o \
 	amdgpu_prime.o amdgpu_vm.o amdgpu_ib.o amdgpu_pll.o \
 	amdgpu_ucode.o amdgpu_bo_list.o amdgpu_ctx.o amdgpu_sync.o
 
+amdgpu-$(CONFIG_DRM_AMDGPU_CIK)+= cik.o gmc_v7_0.o cik_ih.o kv_smc.o kv_dpm.o \
+	ci_smc.o ci_dpm.o dce_v8_0.o gfx_v7_0.o cik_sdma.o uvd_v4_2.o vce_v2_0.o
+
 # add IH block
 amdgpu-y += \
 	amdgpu_irq.o \

drivers/gpu/drm/amd/amdgpu/amdgpu_device.c

@@ -38,6 +38,9 @@
 #include "amdgpu_i2c.h"
 #include "atom.h"
 #include "amdgpu_atombios.h"
+#ifdef CONFIG_DRM_AMDGPU_CIK
+#include "cik.h"
+#endif
 #include "bif/bif_4_1_d.h"
 
 static int amdgpu_debugfs_regs_init(struct amdgpu_device *adev);
@@ -1154,6 +1157,22 @@ static int amdgpu_early_init(struct amdgpu_device *adev)
 	int i, r = -EINVAL;
 
 	switch (adev->asic_type) {
+#ifdef CONFIG_DRM_AMDGPU_CIK
+	case CHIP_BONAIRE:
+	case CHIP_HAWAII:
+	case CHIP_KAVERI:
+	case CHIP_KABINI:
+	case CHIP_MULLINS:
+		if ((adev->asic_type == CHIP_BONAIRE) || (adev->asic_type == CHIP_HAWAII))
+			adev->family = AMDGPU_FAMILY_CI;
+		else
+			adev->family = AMDGPU_FAMILY_KV;
+
+		r = cik_set_ip_blocks(adev);
+		if (r)
+			return r;
+		break;
+#endif
 	default:
 		/* FIXME: not supported yet */
 		return -EINVAL;

drivers/gpu/drm/amd/amdgpu/ci_dpm.h

+/*
+ * Copyright 2013 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+#ifndef __CI_DPM_H__
+#define __CI_DPM_H__
+
+#include "amdgpu_atombios.h"
+#include "ppsmc.h"
+
+#define SMU__NUM_SCLK_DPM_STATE  8
+#define SMU__NUM_MCLK_DPM_LEVELS 6
+#define SMU__NUM_LCLK_DPM_LEVELS 8
+#define SMU__NUM_PCIE_DPM_LEVELS 8
+#include "smu7_discrete.h"
+
+#define CISLANDS_MAX_HARDWARE_POWERLEVELS 2
+
+#define CISLANDS_UNUSED_GPIO_PIN 0x7F
+
+struct ci_pl {
+	u32 mclk;
+	u32 sclk;
+	enum amdgpu_pcie_gen pcie_gen;
+	u16 pcie_lane;
+};
+
+struct ci_ps {
+	u16 performance_level_count;
+	bool dc_compatible;
+	u32 sclk_t;
+	struct ci_pl performance_levels[CISLANDS_MAX_HARDWARE_POWERLEVELS];
+};
+
+struct ci_dpm_level {
+	bool enabled;
+	u32 value;
+	u32 param1;
+};
+
+#define CISLAND_MAX_DEEPSLEEP_DIVIDER_ID 5
+#define MAX_REGULAR_DPM_NUMBER 8
+#define CISLAND_MINIMUM_ENGINE_CLOCK 800
+
+struct ci_single_dpm_table {
+	u32 count;
+	struct ci_dpm_level dpm_levels[MAX_REGULAR_DPM_NUMBER];
+};
+
+struct ci_dpm_table {
+	struct ci_single_dpm_table sclk_table;
+	struct ci_single_dpm_table mclk_table;
+	struct ci_single_dpm_table pcie_speed_table;
+	struct ci_single_dpm_table vddc_table;
+	struct ci_single_dpm_table vddci_table;
+	struct ci_single_dpm_table mvdd_table;
+};
+
+struct ci_mc_reg_entry {
+	u32 mclk_max;
+	u32 mc_data[SMU7_DISCRETE_MC_REGISTER_ARRAY_SIZE];
+};
+
+struct ci_mc_reg_table {
+	u8 last;
+	u8 num_entries;
+	u16 valid_flag;
+	struct ci_mc_reg_entry mc_reg_table_entry[MAX_AC_TIMING_ENTRIES];
+	SMU7_Discrete_MCRegisterAddress mc_reg_address[SMU7_DISCRETE_MC_REGISTER_ARRAY_SIZE];
+};
+
+struct ci_ulv_parm
+{
+	bool supported;
+	u32 cg_ulv_parameter;
+	u32 volt_change_delay;
+	struct ci_pl pl;
+};
+
+#define CISLANDS_MAX_LEAKAGE_COUNT  8
+
+struct ci_leakage_voltage {
+	u16 count;
+	u16 leakage_id[CISLANDS_MAX_LEAKAGE_COUNT];
+	u16 actual_voltage[CISLANDS_MAX_LEAKAGE_COUNT];
+};
+
+struct ci_dpm_level_enable_mask {
+	u32 uvd_dpm_enable_mask;
+	u32 vce_dpm_enable_mask;
+	u32 acp_dpm_enable_mask;
+	u32 samu_dpm_enable_mask;
+	u32 sclk_dpm_enable_mask;
+	u32 mclk_dpm_enable_mask;
+	u32 pcie_dpm_enable_mask;
+};
+
+struct ci_vbios_boot_state
+{
+	u16 mvdd_bootup_value;
+	u16 vddc_bootup_value;
+	u16 vddci_bootup_value;
+	u32 sclk_bootup_value;
+	u32 mclk_bootup_value;
+	u16 pcie_gen_bootup_value;
+	u16 pcie_lane_bootup_value;
+};
+
+struct ci_clock_registers {
+	u32 cg_spll_func_cntl;
+	u32 cg_spll_func_cntl_2;
+	u32 cg_spll_func_cntl_3;
+	u32 cg_spll_func_cntl_4;
+	u32 cg_spll_spread_spectrum;
+	u32 cg_spll_spread_spectrum_2;
+	u32 dll_cntl;
+	u32 mclk_pwrmgt_cntl;
+	u32 mpll_ad_func_cntl;
+	u32 mpll_dq_func_cntl;
+	u32 mpll_func_cntl;
+	u32 mpll_func_cntl_1;
+	u32 mpll_func_cntl_2;
+	u32 mpll_ss1;
+	u32 mpll_ss2;
+};
+
+struct ci_thermal_temperature_setting {
+	s32 temperature_low;
+	s32 temperature_high;
+	s32 temperature_shutdown;
+};
+
+struct ci_pcie_perf_range {
+	u16 max;
+	u16 min;
+};
+
+enum ci_pt_config_reg_type {
+	CISLANDS_CONFIGREG_MMR = 0,
+	CISLANDS_CONFIGREG_SMC_IND,
+	CISLANDS_CONFIGREG_DIDT_IND,
+	CISLANDS_CONFIGREG_CACHE,
+	CISLANDS_CONFIGREG_MAX
+};
+
+#define POWERCONTAINMENT_FEATURE_BAPM            0x00000001
+#define POWERCONTAINMENT_FEATURE_TDCLimit        0x00000002
+#define POWERCONTAINMENT_FEATURE_PkgPwrLimit     0x00000004
+
+struct ci_pt_config_reg {
+	u32 offset;
+	u32 mask;
+	u32 shift;
+	u32 value;
+	enum ci_pt_config_reg_type type;
+};
+
+struct ci_pt_defaults {
+	u8 svi_load_line_en;
+	u8 svi_load_line_vddc;
+	u8 tdc_vddc_throttle_release_limit_perc;
+	u8 tdc_mawt;
+	u8 tdc_waterfall_ctl;
+	u8 dte_ambient_temp_base;
+	u32 display_cac;
+	u32 bapm_temp_gradient;
+	u16 bapmti_r[SMU7_DTE_ITERATIONS * SMU7_DTE_SOURCES * SMU7_DTE_SINKS];
+	u16 bapmti_rc[SMU7_DTE_ITERATIONS * SMU7_DTE_SOURCES * SMU7_DTE_SINKS];
+};
+
+#define DPMTABLE_OD_UPDATE_SCLK     0x00000001
+#define DPMTABLE_OD_UPDATE_MCLK     0x00000002
+#define DPMTABLE_UPDATE_SCLK        0x00000004
+#define DPMTABLE_UPDATE_MCLK        0x00000008
+
+struct ci_power_info {
+	struct ci_dpm_table dpm_table;
+	u32 voltage_control;
+	u32 mvdd_control;
+	u32 vddci_control;
+	u32 active_auto_throttle_sources;
+	struct ci_clock_registers clock_registers;
+	u16 acpi_vddc;
+	u16 acpi_vddci;
+	enum amdgpu_pcie_gen force_pcie_gen;
+	enum amdgpu_pcie_gen acpi_pcie_gen;
+	struct ci_leakage_voltage vddc_leakage;
+	struct ci_leakage_voltage vddci_leakage;
+	u16 max_vddc_in_pp_table;
+	u16 min_vddc_in_pp_table;
+	u16 max_vddci_in_pp_table;
+	u16 min_vddci_in_pp_table;
+	u32 mclk_strobe_mode_threshold;
+	u32 mclk_stutter_mode_threshold;
+	u32 mclk_edc_enable_threshold;
+	u32 mclk_edc_wr_enable_threshold;
+	struct ci_vbios_boot_state vbios_boot_state;
+	/* smc offsets */
+	u32 sram_end;
+	u32 dpm_table_start;
+	u32 soft_regs_start;
+	u32 mc_reg_table_start;
+	u32 fan_table_start;
+	u32 arb_table_start;
+	/* smc tables */
+	SMU7_Discrete_DpmTable smc_state_table;
+	SMU7_Discrete_MCRegisters smc_mc_reg_table;
+	SMU7_Discrete_PmFuses smc_powertune_table;
+	/* other stuff */
+	struct ci_mc_reg_table mc_reg_table;
+	struct atom_voltage_table vddc_voltage_table;
+	struct atom_voltage_table vddci_voltage_table;
+	struct atom_voltage_table mvdd_voltage_table;
+	struct ci_ulv_parm ulv;
+	u32 power_containment_features;
+	const struct ci_pt_defaults *powertune_defaults;
+	u32 dte_tj_offset;
+	bool vddc_phase_shed_control;
+	struct ci_thermal_temperature_setting thermal_temp_setting;
+	struct ci_dpm_level_enable_mask dpm_level_enable_mask;
+	u32 need_update_smu7_dpm_table;
+	u32 sclk_dpm_key_disabled;
+	u32 mclk_dpm_key_disabled;
+	u32 pcie_dpm_key_disabled;
+	u32 thermal_sclk_dpm_enabled;
+	struct ci_pcie_perf_range pcie_gen_performance;
+	struct ci_pcie_perf_range pcie_lane_performance;
+	struct ci_pcie_perf_range pcie_gen_powersaving;
+	struct ci_pcie_perf_range pcie_lane_powersaving;
+	u32 activity_target[SMU7_MAX_LEVELS_GRAPHICS];
+	u32 mclk_activity_target;
+	u32 low_sclk_interrupt_t;
+	u32 last_mclk_dpm_enable_mask;
+	u32 sys_pcie_mask;
+	/* caps */
+	bool caps_power_containment;
+	bool caps_cac;
+	bool caps_sq_ramping;
+	bool caps_db_ramping;
+	bool caps_td_ramping;
+	bool caps_tcp_ramping;
+	bool caps_fps;
+	bool caps_sclk_ds;
+	bool caps_sclk_ss_support;
+	bool caps_mclk_ss_support;
+	bool caps_uvd_dpm;
+	bool caps_vce_dpm;
+	bool caps_samu_dpm;
+	bool caps_acp_dpm;
+	bool caps_automatic_dc_transition;
+	bool caps_sclk_throttle_low_notification;
+	bool caps_dynamic_ac_timing;
+	bool caps_od_fuzzy_fan_control_support;
+	/* flags */
+	bool thermal_protection;
+	bool pcie_performance_request;
+	bool dynamic_ss;
+	bool dll_default_on;
+	bool cac_enabled;
+	bool uvd_enabled;
+	bool battery_state;
+	bool pspp_notify_required;
+	bool enable_bapm_feature;
+	bool enable_tdc_limit_feature;
+	bool enable_pkg_pwr_tracking_feature;
+	bool use_pcie_performance_levels;
+	bool use_pcie_powersaving_levels;
+	bool uvd_power_gated;
+	/* driver states */
+	struct amdgpu_ps current_rps;
+	struct ci_ps current_ps;
+	struct amdgpu_ps requested_rps;
+	struct ci_ps requested_ps;
+	/* fan control */
+	bool fan_ctrl_is_in_default_mode;
+	bool fan_is_controlled_by_smc;
+	u32 t_min;
+	u32 fan_ctrl_default_mode;
+};
+
+#define CISLANDS_VOLTAGE_CONTROL_NONE                   0x0
+#define CISLANDS_VOLTAGE_CONTROL_BY_GPIO                0x1
+#define CISLANDS_VOLTAGE_CONTROL_BY_SVID2               0x2
+
+#define CISLANDS_Q88_FORMAT_CONVERSION_UNIT             256
+
+#define CISLANDS_VRC_DFLT0                              0x3FFFC000
+#define CISLANDS_VRC_DFLT1                              0x000400
+#define CISLANDS_VRC_DFLT2                              0xC00080
+#define CISLANDS_VRC_DFLT3                              0xC00200
+#define CISLANDS_VRC_DFLT4                              0xC01680
+#define CISLANDS_VRC_DFLT5                              0xC00033
+#define CISLANDS_VRC_DFLT6                              0xC00033
+#define CISLANDS_VRC_DFLT7                              0x3FFFC000
+
+#define CISLANDS_CGULVPARAMETER_DFLT                    0x00040035
+#define CISLAND_TARGETACTIVITY_DFLT                     30
+#define CISLAND_MCLK_TARGETACTIVITY_DFLT                10
+
+#define PCIE_PERF_REQ_REMOVE_REGISTRY   0
+#define PCIE_PERF_REQ_FORCE_LOWPOWER    1
+#define PCIE_PERF_REQ_PECI_GEN1         2
+#define PCIE_PERF_REQ_PECI_GEN2         3
+#define PCIE_PERF_REQ_PECI_GEN3         4
+
+#define CISLANDS_SSTU_DFLT                               0
+#define CISLANDS_SST_DFLT                                0x00C8
+
+/* XXX are these ok? */
+#define CISLANDS_TEMP_RANGE_MIN (90 * 1000)
+#define CISLANDS_TEMP_RANGE_MAX (120 * 1000)
+
+int amdgpu_ci_copy_bytes_to_smc(struct amdgpu_device *adev,
+			 u32 smc_start_address,
+			 const u8 *src, u32 byte_count, u32 limit);
+void amdgpu_ci_start_smc(struct amdgpu_device *adev);
+void amdgpu_ci_reset_smc(struct amdgpu_device *adev);
+int amdgpu_ci_program_jump_on_start(struct amdgpu_device *adev);
+void amdgpu_ci_stop_smc_clock(struct amdgpu_device *adev);
+void amdgpu_ci_start_smc_clock(struct amdgpu_device *adev);
+bool amdgpu_ci_is_smc_running(struct amdgpu_device *adev);
+PPSMC_Result amdgpu_ci_send_msg_to_smc(struct amdgpu_device *adev, PPSMC_Msg msg);
+PPSMC_Result amdgpu_ci_wait_for_smc_inactive(struct amdgpu_device *adev);
+int amdgpu_ci_load_smc_ucode(struct amdgpu_device *adev, u32 limit);
+int amdgpu_ci_read_smc_sram_dword(struct amdgpu_device *adev,
+			   u32 smc_address, u32 *value, u32 limit);
+int amdgpu_ci_write_smc_sram_dword(struct amdgpu_device *adev,
+			    u32 smc_address, u32 value, u32 limit);
+
+#endif

drivers/gpu/drm/amd/amdgpu/ci_smc.c

+/*
+ * Copyright 2011 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Alex Deucher
+ */
+
+#include <linux/firmware.h>
+#include "drmP.h"
+#include "amdgpu.h"
+#include "cikd.h"
+#include "ppsmc.h"
+#include "amdgpu_ucode.h"
+#include "ci_dpm.h"
+
+#include "smu/smu_7_0_1_d.h"
+#include "smu/smu_7_0_1_sh_mask.h"
+
+static int ci_set_smc_sram_address(struct amdgpu_device *adev,
+				   u32 smc_address, u32 limit)
+{
+	if (smc_address & 3)
+		return -EINVAL;
+	if ((smc_address + 3) > limit)
+		return -EINVAL;
+
+	WREG32(mmSMC_IND_INDEX_0, smc_address);
+	WREG32_P(mmSMC_IND_ACCESS_CNTL, 0, ~SMC_IND_ACCESS_CNTL__AUTO_INCREMENT_IND_0_MASK);
+
+	return 0;
+}
+
+int amdgpu_ci_copy_bytes_to_smc(struct amdgpu_device *adev,
+			 u32 smc_start_address,
+			 const u8 *src, u32 byte_count, u32 limit)
+{
+	unsigned long flags;
+	u32 data, original_data;
+	u32 addr;
+	u32 extra_shift;
+	int ret = 0;
+
+	if (smc_start_address & 3)
+		return -EINVAL;
+	if ((smc_start_address + byte_count) > limit)
+		return -EINVAL;
+
+	addr = smc_start_address;
+
+	spin_lock_irqsave(&adev->smc_idx_lock, flags);
+	while (byte_count >= 4) {
+		/* SMC address space is BE */
+		data = (src[0] << 24) | (src[1] << 16) | (src[2] << 8) | src[3];
+
+		ret = ci_set_smc_sram_address(adev, addr, limit);
+		if (ret)
+			goto done;
+
+		WREG32(mmSMC_IND_DATA_0, data);
+
+		src += 4;
+		byte_count -= 4;
+		addr += 4;
+	}
+
+	/* RMW for the final bytes */
+	if (byte_count > 0) {
+		data = 0;
+
+		ret = ci_set_smc_sram_address(adev, addr, limit);
+		if (ret)
+			goto done;
+
+		original_data = RREG32(mmSMC_IND_DATA_0);
+
+		extra_shift = 8 * (4 - byte_count);
+
+		while (byte_count > 0) {
+			data = (data << 8) + *src++;
+			byte_count--;
+		}
+
+		data <<= extra_shift;
+
+		data |= (original_data & ~((~0UL) << extra_shift));
+
+		ret = ci_set_smc_sram_address(adev, addr, limit);
+		if (ret)
+			goto done;
+
+		WREG32(mmSMC_IND_DATA_0, data);
+	}
+
+done:
+	spin_unlock_irqrestore(&adev->smc_idx_lock, flags);
+
+	return ret;
+}
+
+void amdgpu_ci_start_smc(struct amdgpu_device *adev)
+{
+	u32 tmp = RREG32_SMC(ixSMC_SYSCON_RESET_CNTL);
+
+	tmp &= ~SMC_SYSCON_RESET_CNTL__rst_reg_MASK;
+	WREG32_SMC(ixSMC_SYSCON_RESET_CNTL, tmp);
+}
+
+void amdgpu_ci_reset_smc(struct amdgpu_device *adev)
+{
+	u32 tmp = RREG32_SMC(ixSMC_SYSCON_RESET_CNTL);
+
+	tmp |= SMC_SYSCON_RESET_CNTL__rst_reg_MASK;
+	WREG32_SMC(ixSMC_SYSCON_RESET_CNTL, tmp);
+}
+
+int amdgpu_ci_program_jump_on_start(struct amdgpu_device *adev)
+{
+	static u8 data[] = { 0xE0, 0x00, 0x80, 0x40 };
+
+	return amdgpu_ci_copy_bytes_to_smc(adev, 0x0, data, 4, sizeof(data)+1);
+}
+
+void amdgpu_ci_stop_smc_clock(struct amdgpu_device *adev)
+{
+	u32 tmp = RREG32_SMC(ixSMC_SYSCON_CLOCK_CNTL_0);
+
+	tmp |= SMC_SYSCON_CLOCK_CNTL_0__ck_disable_MASK;
+
+	WREG32_SMC(ixSMC_SYSCON_CLOCK_CNTL_0, tmp);
+}
+
+void amdgpu_ci_start_smc_clock(struct amdgpu_device *adev)
+{
+	u32 tmp = RREG32_SMC(ixSMC_SYSCON_CLOCK_CNTL_0);
+
+	tmp &= ~SMC_SYSCON_CLOCK_CNTL_0__ck_disable_MASK;
+
+	WREG32_SMC(ixSMC_SYSCON_CLOCK_CNTL_0, tmp);
+}
+
+bool amdgpu_ci_is_smc_running(struct amdgpu_device *adev)
+{
+	u32 clk = RREG32_SMC(ixSMC_SYSCON_CLOCK_CNTL_0);
+	u32 pc_c = RREG32_SMC(ixSMC_PC_C);
+
+	if (!(clk & SMC_SYSCON_CLOCK_CNTL_0__ck_disable_MASK) && (0x20100 <= pc_c))
+		return true;
+
+	return false;
+}
+
+PPSMC_Result amdgpu_ci_send_msg_to_smc(struct amdgpu_device *adev, PPSMC_Msg msg)
+{
+	u32 tmp;
+	int i;
+
+	if (!amdgpu_ci_is_smc_running(adev))
+		return PPSMC_Result_Failed;
+
+	WREG32(mmSMC_MESSAGE_0, msg);
+
+	for (i = 0; i < adev->usec_timeout; i++) {
+		tmp = RREG32(mmSMC_RESP_0);
+		if (tmp != 0)
+			break;
+		udelay(1);
+	}
+	tmp = RREG32(mmSMC_RESP_0);
+
+	return (PPSMC_Result)tmp;
+}
+
+PPSMC_Result amdgpu_ci_wait_for_smc_inactive(struct amdgpu_device *adev)
+{
+	u32 tmp;
+	int i;
+
+	if (!amdgpu_ci_is_smc_running(adev))
+		return PPSMC_Result_OK;
+
+	for (i = 0; i < adev->usec_timeout; i++) {
+		tmp = RREG32_SMC(ixSMC_SYSCON_CLOCK_CNTL_0);
+		if ((tmp & SMC_SYSCON_CLOCK_CNTL_0__cken_MASK) == 0)
+			break;
+		udelay(1);
+	}
+
+	return PPSMC_Result_OK;
+}
+
+int amdgpu_ci_load_smc_ucode(struct amdgpu_device *adev, u32 limit)
+{
+	const struct smc_firmware_header_v1_0 *hdr;
+	unsigned long flags;
+	u32 ucode_start_address;
+	u32 ucode_size;
+	const u8 *src;
+	u32 data;
+
+	if (!adev->pm.fw)
+		return -EINVAL;
+
+	hdr = (const struct smc_firmware_header_v1_0 *)adev->pm.fw->data;
+	amdgpu_ucode_print_smc_hdr(&hdr->header);
+
+	adev->pm.fw_version = le32_to_cpu(hdr->header.ucode_version);
+	ucode_start_address = le32_to_cpu(hdr->ucode_start_addr);
+	ucode_size = le32_to_cpu(hdr->header.ucode_size_bytes);
+	src = (const u8 *)
+		(adev->pm.fw->data + le32_to_cpu(hdr->header.ucode_array_offset_bytes));
+
+	if (ucode_size & 3)
+		return -EINVAL;
+
+	spin_lock_irqsave(&adev->smc_idx_lock, flags);
+	WREG32(mmSMC_IND_INDEX_0, ucode_start_address);
+	WREG32_P(mmSMC_IND_ACCESS_CNTL, SMC_IND_ACCESS_CNTL__AUTO_INCREMENT_IND_0_MASK,
+		~SMC_IND_ACCESS_CNTL__AUTO_INCREMENT_IND_0_MASK);
+	while (ucode_size >= 4) {
+		/* SMC address space is BE */
+		data = (src[0] << 24) | (src[1] << 16) | (src[2] << 8) | src[3];
+
+		WREG32(mmSMC_IND_DATA_0, data);
+
+		src += 4;
+		ucode_size -= 4;
+	}
+	WREG32_P(mmSMC_IND_ACCESS_CNTL, 0, ~SMC_IND_ACCESS_CNTL__AUTO_INCREMENT_IND_0_MASK);
+	spin_unlock_irqrestore(&adev->smc_idx_lock, flags);
+
+	return 0;
+}
+
+int amdgpu_ci_read_smc_sram_dword(struct amdgpu_device *adev,
+			   u32 smc_address, u32 *value, u32 limit)
+{
+	unsigned long flags;
+	int ret;
+
+	spin_lock_irqsave(&adev->smc_idx_lock, flags);
+	ret = ci_set_smc_sram_address(adev, smc_address, limit);
+	if (ret == 0)
+		*value = RREG32(mmSMC_IND_DATA_0);
+	spin_unlock_irqrestore(&adev->smc_idx_lock, flags);
+
+	return ret;
+}
+
+int amdgpu_ci_write_smc_sram_dword(struct amdgpu_device *adev,
+			    u32 smc_address, u32 value, u32 limit)
+{
+	unsigned long flags;
+	int ret;
+
+	spin_lock_irqsave(&adev->smc_idx_lock, flags);
+	ret = ci_set_smc_sram_address(adev, smc_address, limit);
+	if (ret == 0)
+		WREG32(mmSMC_IND_DATA_0, value);
+	spin_unlock_irqrestore(&adev->smc_idx_lock, flags);
+
+	return ret;
+}

drivers/gpu/drm/amd/amdgpu/cik.c

+/*
+ * Copyright 2012 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Alex Deucher
+ */
+#include <linux/firmware.h>
+#include <linux/slab.h>
+#include <linux/module.h>
+#include "drmP.h"
+#include "amdgpu.h"
+#include "amdgpu_atombios.h"
+#include "amdgpu_ih.h"
+#include "amdgpu_uvd.h"
+#include "amdgpu_vce.h"
+#include "cikd.h"
+#include "atom.h"
+
+#include "cik.h"
+#include "gmc_v7_0.h"
+#include "cik_ih.h"
+#include "dce_v8_0.h"
+#include "gfx_v7_0.h"
+#include "cik_sdma.h"
+#include "uvd_v4_2.h"
+#include "vce_v2_0.h"
+#include "cik_dpm.h"
+
+#include "uvd/uvd_4_2_d.h"
+
+#include "smu/smu_7_0_1_d.h"
+#include "smu/smu_7_0_1_sh_mask.h"
+
+#include "dce/dce_8_0_d.h"
+#include "dce/dce_8_0_sh_mask.h"
+
+#include "bif/bif_4_1_d.h"
+#include "bif/bif_4_1_sh_mask.h"
+
+#include "gca/gfx_7_2_d.h"
+#include "gca/gfx_7_2_enum.h"
+#include "gca/gfx_7_2_sh_mask.h"
+
+#include "gmc/gmc_7_1_d.h"
+#include "gmc/gmc_7_1_sh_mask.h"
+
+#include "oss/oss_2_0_d.h"
+#include "oss/oss_2_0_sh_mask.h"
+
+/*
+ * Indirect registers accessor
+ */
+static u32 cik_pcie_rreg(struct amdgpu_device *adev, u32 reg)
+{
+	unsigned long flags;
+	u32 r;
+
+	spin_lock_irqsave(&adev->pcie_idx_lock, flags);
+	WREG32(mmPCIE_INDEX, reg);
+	(void)RREG32(mmPCIE_INDEX);
+	r = RREG32(mmPCIE_DATA);
+	spin_unlock_irqrestore(&adev->pcie_idx_lock, flags);
+	return r;
+}
+
+static void cik_pcie_wreg(struct amdgpu_device *adev, u32 reg, u32 v)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&adev->pcie_idx_lock, flags);
+	WREG32(mmPCIE_INDEX, reg);
+	(void)RREG32(mmPCIE_INDEX);
+	WREG32(mmPCIE_DATA, v);
+	(void)RREG32(mmPCIE_DATA);
+	spin_unlock_irqrestore(&adev->pcie_idx_lock, flags);
+}
+
+static u32 cik_smc_rreg(struct amdgpu_device *adev, u32 reg)
+{
+	unsigned long flags;
+	u32 r;
+
+	spin_lock_irqsave(&adev->smc_idx_lock, flags);
+	WREG32(mmSMC_IND_INDEX_0, (reg));
+	r = RREG32(mmSMC_IND_DATA_0);
+	spin_unlock_irqrestore(&adev->smc_idx_lock, flags);
+	return r;
+}
+
+static void cik_smc_wreg(struct amdgpu_device *adev, u32 reg, u32 v)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&adev->smc_idx_lock, flags);
+	WREG32(mmSMC_IND_INDEX_0, (reg));
+	WREG32(mmSMC_IND_DATA_0, (v));
+	spin_unlock_irqrestore(&adev->smc_idx_lock, flags);
+}
+
+static u32 cik_uvd_ctx_rreg(struct amdgpu_device *adev, u32 reg)
+{
+	unsigned long flags;
+	u32 r;
+
+	spin_lock_irqsave(&adev->uvd_ctx_idx_lock, flags);
+	WREG32(mmUVD_CTX_INDEX, ((reg) & 0x1ff));
+	r = RREG32(mmUVD_CTX_DATA);
+	spin_unlock_irqrestore(&adev->uvd_ctx_idx_lock, flags);
+	return r;
+}
+
+static void cik_uvd_ctx_wreg(struct amdgpu_device *adev, u32 reg, u32 v)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&adev->uvd_ctx_idx_lock, flags);
+	WREG32(mmUVD_CTX_INDEX, ((reg) & 0x1ff));
+	WREG32(mmUVD_CTX_DATA, (v));
+	spin_unlock_irqrestore(&adev->uvd_ctx_idx_lock, flags);
+}
+
+static u32 cik_didt_rreg(struct amdgpu_device *adev, u32 reg)
+{
+	unsigned long flags;
+	u32 r;
+
+	spin_lock_irqsave(&adev->didt_idx_lock, flags);
+	WREG32(mmDIDT_IND_INDEX, (reg));
+	r = RREG32(mmDIDT_IND_DATA);
+	spin_unlock_irqrestore(&adev->didt_idx_lock, flags);
+	return r;
+}
+
+static void cik_didt_wreg(struct amdgpu_device *adev, u32 reg, u32 v)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&adev->didt_idx_lock, flags);
+	WREG32(mmDIDT_IND_INDEX, (reg));
+	WREG32(mmDIDT_IND_DATA, (v));
+	spin_unlock_irqrestore(&adev->didt_idx_lock, flags);
+}
+
+static const u32 bonaire_golden_spm_registers[] =
+{
+	0xc200, 0xe0ffffff, 0xe0000000
+};
+
+static const u32 bonaire_golden_common_registers[] =
+{
+	0x31dc, 0xffffffff, 0x00000800,
+	0x31dd, 0xffffffff, 0x00000800,
+	0x31e6, 0xffffffff, 0x00007fbf,
+	0x31e7, 0xffffffff, 0x00007faf
+};
+
+static const u32 bonaire_golden_registers[] =
+{
+	0xcd5, 0x00000333, 0x00000333,
+	0xcd4, 0x000c0fc0, 0x00040200,
+	0x2684, 0x00010000, 0x00058208,
+	0xf000, 0xffff1fff, 0x00140000,
+	0xf080, 0xfdfc0fff, 0x00000100,
+	0xf08d, 0x40000000, 0x40000200,
+	0x260c, 0xffffffff, 0x00000000,
+	0x260d, 0xf00fffff, 0x00000400,
+	0x260e, 0x0002021c, 0x00020200,
+	0x31e, 0x00000080, 0x00000000,
+	0x16ec, 0x000000f0, 0x00000070,
+	0x16f0, 0xf0311fff, 0x80300000,
+	0x263e, 0x73773777, 0x12010001,
+	0xd43, 0x00810000, 0x408af000,
+	0x1c0c, 0x31000111, 0x00000011,
+	0xbd2, 0x73773777, 0x12010001,
+	0x883, 0x00007fb6, 0x0021a1b1,
+	0x884, 0x00007fb6, 0x002021b1,
+	0x860, 0x00007fb6, 0x00002191,
+	0x886, 0x00007fb6, 0x002121b1,
+	0x887, 0x00007fb6, 0x002021b1,
+	0x877, 0x00007fb6, 0x00002191,
+	0x878, 0x00007fb6, 0x00002191,
+	0xd8a, 0x0000003f, 0x0000000a,
+	0xd8b, 0x0000003f, 0x0000000a,
+	0xab9, 0x00073ffe, 0x000022a2,
+	0x903, 0x000007ff, 0x00000000,
+	0x2285, 0xf000003f, 0x00000007,
+	0x22fc, 0x00002001, 0x00000001,
+	0x22c9, 0xffffffff, 0x00ffffff,
+	0xc281, 0x0000ff0f, 0x00000000,
+	0xa293, 0x07ffffff, 0x06000000,
+	0x136, 0x00000fff, 0x00000100,
+	0xf9e, 0x00000001, 0x00000002,
+	0x2440, 0x03000000, 0x0362c688,
+	0x2300, 0x000000ff, 0x00000001,
+	0x390, 0x00001fff, 0x00001fff,
+	0x2418, 0x0000007f, 0x00000020,
+	0x2542, 0x00010000, 0x00010000,
+	0x2b05, 0x000003ff, 0x000000f3,
+	0x2b03, 0xffffffff, 0x00001032
+};
+
+static const u32 bonaire_mgcg_cgcg_init[] =
+{
+	0x3108, 0xffffffff, 0xfffffffc,
+	0xc200, 0xffffffff, 0xe0000000,
+	0xf0a8, 0xffffffff, 0x00000100,
+	0xf082, 0xffffffff, 0x00000100,
+	0xf0b0, 0xffffffff, 0xc0000100,
+	0xf0b2, 0xffffffff, 0xc0000100,
+	0xf0b1, 0xffffffff, 0xc0000100,
+	0x1579, 0xffffffff, 0x00600100,
+	0xf0a0, 0xffffffff, 0x00000100,
+	0xf085, 0xffffffff, 0x06000100,
+	0xf088, 0xffffffff, 0x00000100,
+	0xf086, 0xffffffff, 0x06000100,
+	0xf081, 0xffffffff, 0x00000100,
+	0xf0b8, 0xffffffff, 0x00000100,
+	0xf089, 0xffffffff, 0x00000100,
+	0xf080, 0xffffffff, 0x00000100,
+	0xf08c, 0xffffffff, 0x00000100,
+	0xf08d, 0xffffffff, 0x00000100,
+	0xf094, 0xffffffff, 0x00000100,
+	0xf095, 0xffffffff, 0x00000100,
+	0xf096, 0xffffffff, 0x00000100,
+	0xf097, 0xffffffff, 0x00000100,
+	0xf098, 0xffffffff, 0x00000100,
+	0xf09f, 0xffffffff, 0x00000100,
+	0xf09e, 0xffffffff, 0x00000100,
+	0xf084, 0xffffffff, 0x06000100,
+	0xf0a4, 0xffffffff, 0x00000100,
+	0xf09d, 0xffffffff, 0x00000100,
+	0xf0ad, 0xffffffff, 0x00000100,
+	0xf0ac, 0xffffffff, 0x00000100,
+	0xf09c, 0xffffffff, 0x00000100,
+	0xc200, 0xffffffff, 0xe0000000,
+	0xf008, 0xffffffff, 0x00010000,
+	0xf009, 0xffffffff, 0x00030002,
+	0xf00a, 0xffffffff, 0x00040007,
+	0xf00b, 0xffffffff, 0x00060005,
+	0xf00c, 0xffffffff, 0x00090008,
+	0xf00d, 0xffffffff, 0x00010000,
+	0xf00e, 0xffffffff, 0x00030002,
+	0xf00f, 0xffffffff, 0x00040007,
+	0xf010, 0xffffffff, 0x00060005,
+	0xf011, 0xffffffff, 0x00090008,
+	0xf012, 0xffffffff, 0x00010000,
+	0xf013, 0xffffffff, 0x00030002,
+	0xf014, 0xffffffff, 0x00040007,
+	0xf015, 0xffffffff, 0x00060005,
+	0xf016, 0xffffffff, 0x00090008,
+	0xf017, 0xffffffff, 0x00010000,
+	0xf018, 0xffffffff, 0x00030002,
+	0xf019, 0xffffffff, 0x00040007,
+	0xf01a, 0xffffffff, 0x00060005,
+	0xf01b, 0xffffffff, 0x00090008,
+	0xf01c, 0xffffffff, 0x00010000,
+	0xf01d, 0xffffffff, 0x00030002,
+	0xf01e, 0xffffffff, 0x00040007,
+	0xf01f, 0xffffffff, 0x00060005,
+	0xf020, 0xffffffff, 0x00090008,
+	0xf021, 0xffffffff, 0x00010000,
+	0xf022, 0xffffffff, 0x00030002,
+	0xf023, 0xffffffff, 0x00040007,
+	0xf024, 0xffffffff, 0x00060005,
+	0xf025, 0xffffffff, 0x00090008,
+	0xf026, 0xffffffff, 0x00010000,
+	0xf027, 0xffffffff, 0x00030002,
+	0xf028, 0xffffffff, 0x00040007,
+	0xf029, 0xffffffff, 0x00060005,
+	0xf02a, 0xffffffff, 0x00090008,
+	0xf000, 0xffffffff, 0x96e00200,
+	0x21c2, 0xffffffff, 0x00900100,
+	0x3109, 0xffffffff, 0x0020003f,
+	0xe, 0xffffffff, 0x0140001c,
+	0xf, 0x000f0000, 0x000f0000,
+	0x88, 0xffffffff, 0xc060000c,
+	0x89, 0xc0000fff, 0x00000100,
+	0x3e4, 0xffffffff, 0x00000100,
+	0x3e6, 0x00000101, 0x00000000,
+	0x82a, 0xffffffff, 0x00000104,
+	0x1579, 0xff000fff, 0x00000100,
+	0xc33, 0xc0000fff, 0x00000104,
+	0x3079, 0x00000001, 0x00000001,
+	0x3403, 0xff000ff0, 0x00000100,
+	0x3603, 0xff000ff0, 0x00000100
+};
+
+static const u32 spectre_golden_spm_registers[] =
+{
+	0xc200, 0xe0ffffff, 0xe0000000
+};
+
+static const u32 spectre_golden_common_registers[] =
+{
+	0x31dc, 0xffffffff, 0x00000800,
+	0x31dd, 0xffffffff, 0x00000800,
+	0x31e6, 0xffffffff, 0x00007fbf,
+	0x31e7, 0xffffffff, 0x00007faf
+};
+
+static const u32 spectre_golden_registers[] =
+{
+	0xf000, 0xffff1fff, 0x96940200,
+	0xf003, 0xffff0001, 0xff000000,
+	0xf080, 0xfffc0fff, 0x00000100,
+	0x1bb6, 0x00010101, 0x00010000,
+	0x260d, 0xf00fffff, 0x00000400,
+	0x260e, 0xfffffffc, 0x00020200,
+	0x16ec, 0x000000f0, 0x00000070,
+	0x16f0, 0xf0311fff, 0x80300000,
+	0x263e, 0x73773777, 0x12010001,
+	0x26df, 0x00ff0000, 0x00fc0000,
+	0xbd2, 0x73773777, 0x12010001,
+	0x2285, 0xf000003f, 0x00000007,
+	0x22c9, 0xffffffff, 0x00ffffff,
+	0xa0d4, 0x3f3f3fff, 0x00000082,
+	0xa0d5, 0x0000003f, 0x00000000,
+	0xf9e, 0x00000001, 0x00000002,
+	0x244f, 0xffff03df, 0x00000004,
+	0x31da, 0x00000008, 0x00000008,
+	0x2300, 0x000008ff, 0x00000800,
+	0x2542, 0x00010000, 0x00010000,
+	0x2b03, 0xffffffff, 0x54763210,
+	0x853e, 0x01ff01ff, 0x00000002,
+	0x8526, 0x007ff800, 0x00200000,
+	0x8057, 0xffffffff, 0x00000f40,
+	0xc24d, 0xffffffff, 0x00000001
+};
+
+static const u32 spectre_mgcg_cgcg_init[] =
+{
+	0x3108, 0xffffffff, 0xfffffffc,
+	0xc200, 0xffffffff, 0xe0000000,
+	0xf0a8, 0xffffffff, 0x00000100,
+	0xf082, 0xffffffff, 0x00000100,
+	0xf0b0, 0xffffffff, 0x00000100,
+	0xf0b2, 0xffffffff, 0x00000100,
+	0xf0b1, 0xffffffff, 0x00000100,
+	0x1579, 0xffffffff, 0x00600100,
+	0xf0a0, 0xffffffff, 0x00000100,
+	0xf085, 0xffffffff, 0x06000100,
+	0xf088, 0xffffffff, 0x00000100,
+	0xf086, 0xffffffff, 0x06000100,
+	0xf081, 0xffffffff, 0x00000100,
+	0xf0b8, 0xffffffff, 0x00000100,
+	0xf089, 0xffffffff, 0x00000100,
+	0xf080, 0xffffffff, 0x00000100,
+	0xf08c, 0xffffffff, 0x00000100,
+	0xf08d, 0xffffffff, 0x00000100,
+	0xf094, 0xffffffff, 0x00000100,
+	0xf095, 0xffffffff, 0x00000100,
+	0xf096, 0xffffffff, 0x00000100,
+	0xf097, 0xffffffff, 0x00000100,
+	0xf098, 0xffffffff, 0x00000100,
+	0xf09f, 0xffffffff, 0x00000100,
+	0xf09e, 0xffffffff, 0x00000100,
+	0xf084, 0xffffffff, 0x06000100,
+	0xf0a4, 0xffffffff, 0x00000100,
+	0xf09d, 0xffffffff, 0x00000100,
+	0xf0ad, 0xffffffff, 0x00000100,
+	0xf0ac, 0xffffffff, 0x00000100,
+	0xf09c, 0xffffffff, 0x00000100,
+	0xc200, 0xffffffff, 0xe0000000,
+	0xf008, 0xffffffff, 0x00010000,
+	0xf009, 0xffffffff, 0x00030002,
+	0xf00a, 0xffffffff, 0x00040007,
+	0xf00b, 0xffffffff, 0x00060005,
+	0xf00c, 0xffffffff, 0x00090008,
+	0xf00d, 0xffffffff, 0x00010000,
+	0xf00e, 0xffffffff, 0x00030002,
+	0xf00f, 0xffffffff, 0x00040007,
+	0xf010, 0xffffffff, 0x00060005,
+	0xf011, 0xffffffff, 0x00090008,
+	0xf012, 0xffffffff, 0x00010000,
+	0xf013, 0xffffffff, 0x00030002,
+	0xf014, 0xffffffff, 0x00040007,
+	0xf015, 0xffffffff, 0x00060005,
+	0xf016, 0xffffffff, 0x00090008,
+	0xf017, 0xffffffff, 0x00010000,
+	0xf018, 0xffffffff, 0x00030002,
+	0xf019, 0xffffffff, 0x00040007,
+	0xf01a, 0xffffffff, 0x00060005,
+	0xf01b, 0xffffffff, 0x00090008,
+	0xf01c, 0xffffffff, 0x00010000,
+	0xf01d, 0xffffffff, 0x00030002,
+	0xf01e, 0xffffffff, 0x00040007,
+	0xf01f, 0xffffffff, 0x00060005,
+	0xf020, 0xffffffff, 0x00090008,
+	0xf021, 0xffffffff, 0x00010000,
+	0xf022, 0xffffffff, 0x00030002,
+	0xf023, 0xffffffff, 0x00040007,
+	0xf024, 0xffffffff, 0x00060005,
+	0xf025, 0xffffffff, 0x00090008,
+	0xf026, 0xffffffff, 0x00010000,
+	0xf027, 0xffffffff, 0x00030002,
+	0xf028, 0xffffffff, 0x00040007,
+	0xf029, 0xffffffff, 0x00060005,
+	0xf02a, 0xffffffff, 0x00090008,
+	0xf02b, 0xffffffff, 0x00010000,
+	0xf02c, 0xffffffff, 0x00030002,
+	0xf02d, 0xffffffff, 0x00040007,
+	0xf02e, 0xffffffff, 0x00060005,
+	0xf02f, 0xffffffff, 0x00090008,
+	0xf000, 0xffffffff, 0x96e00200,
+	0x21c2, 0xffffffff, 0x00900100,
+	0x3109, 0xffffffff, 0x0020003f,
+	0xe, 0xffffffff, 0x0140001c,
+	0xf, 0x000f0000, 0x000f0000,
+	0x88, 0xffffffff, 0xc060000c,
+	0x89, 0xc0000fff, 0x00000100,
+	0x3e4, 0xffffffff, 0x00000100,
+	0x3e6, 0x00000101, 0x00000000,
+	0x82a, 0xffffffff, 0x00000104,
+	0x1579, 0xff000fff, 0x00000100,
+	0xc33, 0xc0000fff, 0x00000104,
+	0x3079, 0x00000001, 0x00000001,
+	0x3403, 0xff000ff0, 0x00000100,
+	0x3603, 0xff000ff0, 0x00000100
+};
+
+static const u32 kalindi_golden_spm_registers[] =
+{
+	0xc200, 0xe0ffffff, 0xe0000000
+};
+
+static const u32 kalindi_golden_common_registers[] =
+{
+	0x31dc, 0xffffffff, 0x00000800,
+	0x31dd, 0xffffffff, 0x00000800,
+	0x31e6, 0xffffffff, 0x00007fbf,
+	0x31e7, 0xffffffff, 0x00007faf
+};
+
+static const u32 kalindi_golden_registers[] =
+{
+	0xf000, 0xffffdfff, 0x6e944040,
+	0x1579, 0xff607fff, 0xfc000100,
+	0xf088, 0xff000fff, 0x00000100,
+	0xf089, 0xff000fff, 0x00000100,
+	0xf080, 0xfffc0fff, 0x00000100,
+	0x1bb6, 0x00010101, 0x00010000,
+	0x260c, 0xffffffff, 0x00000000,
+	0x260d, 0xf00fffff, 0x00000400,
+	0x16ec, 0x000000f0, 0x00000070,
+	0x16f0, 0xf0311fff, 0x80300000,
+	0x263e, 0x73773777, 0x12010001,
+	0x263f, 0xffffffff, 0x00000010,
+	0x26df, 0x00ff0000, 0x00fc0000,
+	0x200c, 0x00001f0f, 0x0000100a,
+	0xbd2, 0x73773777, 0x12010001,
+	0x902, 0x000fffff, 0x000c007f,
+	0x2285, 0xf000003f, 0x00000007,
+	0x22c9, 0x3fff3fff, 0x00ffcfff,
+	0xc281, 0x0000ff0f, 0x00000000,
+	0xa293, 0x07ffffff, 0x06000000,
+	0x136, 0x00000fff, 0x00000100,
+	0xf9e, 0x00000001, 0x00000002,
+	0x31da, 0x00000008, 0x00000008,
+	0x2300, 0x000000ff, 0x00000003,
+	0x853e, 0x01ff01ff, 0x00000002,
+	0x8526, 0x007ff800, 0x00200000,
+	0x8057, 0xffffffff, 0x00000f40,
+	0x2231, 0x001f3ae3, 0x00000082,
+	0x2235, 0x0000001f, 0x00000010,
+	0xc24d, 0xffffffff, 0x00000000
+};
+
+static const u32 kalindi_mgcg_cgcg_init[] =
+{
+	0x3108, 0xffffffff, 0xfffffffc,
+	0xc200, 0xffffffff, 0xe0000000,
+	0xf0a8, 0xffffffff, 0x00000100,
+	0xf082, 0xffffffff, 0x00000100,
+	0xf0b0, 0xffffffff, 0x00000100,
+	0xf0b2, 0xffffffff, 0x00000100,
+	0xf0b1, 0xffffffff, 0x00000100,
+	0x1579, 0xffffffff, 0x00600100,
+	0xf0a0, 0xffffffff, 0x00000100,
+	0xf085, 0xffffffff, 0x06000100,
+	0xf088, 0xffffffff, 0x00000100,
+	0xf086, 0xffffffff, 0x06000100,
+	0xf081, 0xffffffff, 0x00000100,
+	0xf0b8, 0xffffffff, 0x00000100,
+	0xf089, 0xffffffff, 0x00000100,
+	0xf080, 0xffffffff, 0x00000100,
+	0xf08c, 0xffffffff, 0x00000100,
+	0xf08d, 0xffffffff, 0x00000100,
+	0xf094, 0xffffffff, 0x00000100,
+	0xf095, 0xffffffff, 0x00000100,
+	0xf096, 0xffffffff, 0x00000100,
+	0xf097, 0xffffffff, 0x00000100,
+	0xf098, 0xffffffff, 0x00000100,
+	0xf09f, 0xffffffff, 0x00000100,
+	0xf09e, 0xffffffff, 0x00000100,
+	0xf084, 0xffffffff, 0x06000100,
+	0xf0a4, 0xffffffff, 0x00000100,
+	0xf09d, 0xffffffff, 0x00000100,
+	0xf0ad, 0xffffffff, 0x00000100,
+	0xf0ac, 0xffffffff, 0x00000100,
+	0xf09c, 0xffffffff, 0x00000100,
+	0xc200, 0xffffffff, 0xe0000000,
+	0xf008, 0xffffffff, 0x00010000,
+	0xf009, 0xffffffff, 0x00030002,
+	0xf00a, 0xffffffff, 0x00040007,
+	0xf00b, 0xffffffff, 0x00060005,
+	0xf00c, 0xffffffff, 0x00090008,
+	0xf00d, 0xffffffff, 0x00010000,
+	0xf00e, 0xffffffff, 0x00030002,
+	0xf00f, 0xffffffff, 0x00040007,
+	0xf010, 0xffffffff, 0x00060005,
+	0xf011, 0xffffffff, 0x00090008,
+	0xf000, 0xffffffff, 0x96e00200,
+	0x21c2, 0xffffffff, 0x00900100,
+	0x3109, 0xffffffff, 0x0020003f,
+	0xe, 0xffffffff, 0x0140001c,
+	0xf, 0x000f0000, 0x000f0000,
+	0x88, 0xffffffff, 0xc060000c,
+	0x89, 0xc0000fff, 0x00000100,
+	0x82a, 0xffffffff, 0x00000104,
+	0x1579, 0xff000fff, 0x00000100,
+	0xc33, 0xc0000fff, 0x00000104,
+	0x3079, 0x00000001, 0x00000001,
+	0x3403, 0xff000ff0, 0x00000100,
+	0x3603, 0xff000ff0, 0x00000100
+};
+
+static const u32 hawaii_golden_spm_registers[] =
+{
+	0xc200, 0xe0ffffff, 0xe0000000
+};
+
+static const u32 hawaii_golden_common_registers[] =
+{
+	0xc200, 0xffffffff, 0xe0000000,
+	0xa0d4, 0xffffffff, 0x3a00161a,
+	0xa0d5, 0xffffffff, 0x0000002e,
+	0x2684, 0xffffffff, 0x00018208,
+	0x263e, 0xffffffff, 0x12011003
+};
+
+static const u32 hawaii_golden_registers[] =
+{
+	0xcd5, 0x00000333, 0x00000333,
+	0x2684, 0x00010000, 0x00058208,
+	0x260c, 0xffffffff, 0x00000000,
+	0x260d, 0xf00fffff, 0x00000400,
+	0x260e, 0x0002021c, 0x00020200,
+	0x31e, 0x00000080, 0x00000000,
+	0x16ec, 0x000000f0, 0x00000070,
+	0x16f0, 0xf0311fff, 0x80300000,
+	0xd43, 0x00810000, 0x408af000,
+	0x1c0c, 0x31000111, 0x00000011,
+	0xbd2, 0x73773777, 0x12010001,
+	0x848, 0x0000007f, 0x0000001b,
+	0x877, 0x00007fb6, 0x00002191,
+	0xd8a, 0x0000003f, 0x0000000a,
+	0xd8b, 0x0000003f, 0x0000000a,
+	0xab9, 0x00073ffe, 0x000022a2,
+	0x903, 0x000007ff, 0x00000000,
+	0x22fc, 0x00002001, 0x00000001,
+	0x22c9, 0xffffffff, 0x00ffffff,
+	0xc281, 0x0000ff0f, 0x00000000,
+	0xa293, 0x07ffffff, 0x06000000,
+	0xf9e, 0x00000001, 0x00000002,
+	0x31da, 0x00000008, 0x00000008,
+	0x31dc, 0x00000f00, 0x00000800,
+	0x31dd, 0x00000f00, 0x00000800,
+	0x31e6, 0x00ffffff, 0x00ff7fbf,
+	0x31e7, 0x00ffffff, 0x00ff7faf,
+	0x2300, 0x000000ff, 0x00000800,
+	0x390, 0x00001fff, 0x00001fff,
+	0x2418, 0x0000007f, 0x00000020,
+	0x2542, 0x00010000, 0x00010000,
+	0x2b80, 0x00100000, 0x000ff07c,
+	0x2b05, 0x000003ff, 0x0000000f,
+	0x2b04, 0xffffffff, 0x7564fdec,
+	0x2b03, 0xffffffff, 0x3120b9a8,
+	0x2b02, 0x20000000, 0x0f9c0000
+};
+
+static const u32 hawaii_mgcg_cgcg_init[] =
+{
+	0x3108, 0xffffffff, 0xfffffffd,
+	0xc200, 0xffffffff, 0xe0000000,
+	0xf0a8, 0xffffffff, 0x00000100,
+	0xf082, 0xffffffff, 0x00000100,
+	0xf0b0, 0xffffffff, 0x00000100,
+	0xf0b2, 0xffffffff, 0x00000100,
+	0xf0b1, 0xffffffff, 0x00000100,
+	0x1579, 0xffffffff, 0x00200100,
+	0xf0a0, 0xffffffff, 0x00000100,
+	0xf085, 0xffffffff, 0x06000100,
+	0xf088, 0xffffffff, 0x00000100,
+	0xf086, 0xffffffff, 0x06000100,
+	0xf081, 0xffffffff, 0x00000100,
+	0xf0b8, 0xffffffff, 0x00000100,
+	0xf089, 0xffffffff, 0x00000100,
+	0xf080, 0xffffffff, 0x00000100,
+	0xf08c, 0xffffffff, 0x00000100,
+	0xf08d, 0xffffffff, 0x00000100,
+	0xf094, 0xffffffff, 0x00000100,
+	0xf095, 0xffffffff, 0x00000100,
+	0xf096, 0xffffffff, 0x00000100,
+	0xf097, 0xffffffff, 0x00000100,
+	0xf098, 0xffffffff, 0x00000100,
+	0xf09f, 0xffffffff, 0x00000100,
+	0xf09e, 0xffffffff, 0x00000100,
+	0xf084, 0xffffffff, 0x06000100,
+	0xf0a4, 0xffffffff, 0x00000100,
+	0xf09d, 0xffffffff, 0x00000100,
+	0xf0ad, 0xffffffff, 0x00000100,
+	0xf0ac, 0xffffffff, 0x00000100,
+	0xf09c, 0xffffffff, 0x00000100,
+	0xc200, 0xffffffff, 0xe0000000,
+	0xf008, 0xffffffff, 0x00010000,
+	0xf009, 0xffffffff, 0x00030002,
+	0xf00a, 0xffffffff, 0x00040007,
+	0xf00b, 0xffffffff, 0x00060005,
+	0xf00c, 0xffffffff, 0x00090008,
+	0xf00d, 0xffffffff, 0x00010000,
+	0xf00e, 0xffffffff, 0x00030002,
+	0xf00f, 0xffffffff, 0x00040007,
+	0xf010, 0xffffffff, 0x00060005,
+	0xf011, 0xffffffff, 0x00090008,
+	0xf012, 0xffffffff, 0x00010000,
+	0xf013, 0xffffffff, 0x00030002,
+	0xf014, 0xffffffff, 0x00040007,
+	0xf015, 0xffffffff, 0x00060005,
+	0xf016, 0xffffffff, 0x00090008,
+	0xf017, 0xffffffff, 0x00010000,
+	0xf018, 0xffffffff, 0x00030002,
+	0xf019, 0xffffffff, 0x00040007,
+	0xf01a, 0xffffffff, 0x00060005,
+	0xf01b, 0xffffffff, 0x00090008,
+	0xf01c, 0xffffffff, 0x00010000,
+	0xf01d, 0xffffffff, 0x00030002,
+	0xf01e, 0xffffffff, 0x00040007,
+	0xf01f, 0xffffffff, 0x00060005,
+	0xf020, 0xffffffff, 0x00090008,
+	0xf021, 0xffffffff, 0x00010000,
+	0xf022, 0xffffffff, 0x00030002,
+	0xf023, 0xffffffff, 0x00040007,
+	0xf024, 0xffffffff, 0x00060005,
+	0xf025, 0xffffffff, 0x00090008,
+	0xf026, 0xffffffff, 0x00010000,
+	0xf027, 0xffffffff, 0x00030002,
+	0xf028, 0xffffffff, 0x00040007,
+	0xf029, 0xffffffff, 0x00060005,
+	0xf02a, 0xffffffff, 0x00090008,
+	0xf02b, 0xffffffff, 0x00010000,
+	0xf02c, 0xffffffff, 0x00030002,
+	0xf02d, 0xffffffff, 0x00040007,
+	0xf02e, 0xffffffff, 0x00060005,
+	0xf02f, 0xffffffff, 0x00090008,
+	0xf030, 0xffffffff, 0x00010000,
+	0xf031, 0xffffffff, 0x00030002,
+	0xf032, 0xffffffff, 0x00040007,
+	0xf033, 0xffffffff, 0x00060005,
+	0xf034, 0xffffffff, 0x00090008,
+	0xf035, 0xffffffff, 0x00010000,
+	0xf036, 0xffffffff, 0x00030002,
+	0xf037, 0xffffffff, 0x00040007,
+	0xf038, 0xffffffff, 0x00060005,
+	0xf039, 0xffffffff, 0x00090008,
+	0xf03a, 0xffffffff, 0x00010000,
+	0xf03b, 0xffffffff, 0x00030002,
+	0xf03c, 0xffffffff, 0x00040007,
+	0xf03d, 0xffffffff, 0x00060005,
+	0xf03e, 0xffffffff, 0x00090008,
+	0x30c6, 0xffffffff, 0x00020200,
+	0xcd4, 0xffffffff, 0x00000200,
+	0x570, 0xffffffff, 0x00000400,
+	0x157a, 0xffffffff, 0x00000000,
+	0xbd4, 0xffffffff, 0x00000902,
+	0xf000, 0xffffffff, 0x96940200,
+	0x21c2, 0xffffffff, 0x00900100,
+	0x3109, 0xffffffff, 0x0020003f,
+	0xe, 0xffffffff, 0x0140001c,
+	0xf, 0x000f0000, 0x000f0000,
+	0x88, 0xffffffff, 0xc060000c,
+	0x89, 0xc0000fff, 0x00000100,
+	0x3e4, 0xffffffff, 0x00000100,
+	0x3e6, 0x00000101, 0x00000000,
+	0x82a, 0xffffffff, 0x00000104,
+	0x1579, 0xff000fff, 0x00000100,
+	0xc33, 0xc0000fff, 0x00000104,
+	0x3079, 0x00000001, 0x00000001,
+	0x3403, 0xff000ff0, 0x00000100,
+	0x3603, 0xff000ff0, 0x00000100
+};
+
+static const u32 godavari_golden_registers[] =
+{
+	0x1579, 0xff607fff, 0xfc000100,
+	0x1bb6, 0x00010101, 0x00010000,
+	0x260c, 0xffffffff, 0x00000000,
+	0x260c0, 0xf00fffff, 0x00000400,
+	0x184c, 0xffffffff, 0x00010000,
+	0x16ec, 0x000000f0, 0x00000070,
+	0x16f0, 0xf0311fff, 0x80300000,
+	0x263e, 0x73773777, 0x12010001,
+	0x263f, 0xffffffff, 0x00000010,
+	0x200c, 0x00001f0f, 0x0000100a,
+	0xbd2, 0x73773777, 0x12010001,
+	0x902, 0x000fffff, 0x000c007f,
+	0x2285, 0xf000003f, 0x00000007,
+	0x22c9, 0xffffffff, 0x00ff0fff,
+	0xc281, 0x0000ff0f, 0x00000000,
+	0xa293, 0x07ffffff, 0x06000000,
+	0x136, 0x00000fff, 0x00000100,
+	0x3405, 0x00010000, 0x00810001,
+	0x3605, 0x00010000, 0x00810001,
+	0xf9e, 0x00000001, 0x00000002,
+	0x31da, 0x00000008, 0x00000008,
+	0x31dc, 0x00000f00, 0x00000800,
+	0x31dd, 0x00000f00, 0x00000800,
+	0x31e6, 0x00ffffff, 0x00ff7fbf,
+	0x31e7, 0x00ffffff, 0x00ff7faf,
+	0x2300, 0x000000ff, 0x00000001,
+	0x853e, 0x01ff01ff, 0x00000002,
+	0x8526, 0x007ff800, 0x00200000,
+	0x8057, 0xffffffff, 0x00000f40,
+	0x2231, 0x001f3ae3, 0x00000082,
+	0x2235, 0x0000001f, 0x00000010,
+	0xc24d, 0xffffffff, 0x00000000
+};
+
+static void cik_init_golden_registers(struct amdgpu_device *adev)
+{
+	/* Some of the registers might be dependent on GRBM_GFX_INDEX */
+	mutex_lock(&adev->grbm_idx_mutex);
+
+	switch (adev->asic_type) {
+	case CHIP_BONAIRE:
+		amdgpu_program_register_sequence(adev,
+						 bonaire_mgcg_cgcg_init,
+						 (const u32)ARRAY_SIZE(bonaire_mgcg_cgcg_init));
+		amdgpu_program_register_sequence(adev,
+						 bonaire_golden_registers,
+						 (const u32)ARRAY_SIZE(bonaire_golden_registers));
+		amdgpu_program_register_sequence(adev,
+						 bonaire_golden_common_registers,
+						 (const u32)ARRAY_SIZE(bonaire_golden_common_registers));
+		amdgpu_program_register_sequence(adev,
+						 bonaire_golden_spm_registers,
+						 (const u32)ARRAY_SIZE(bonaire_golden_spm_registers));
+		break;
+	case CHIP_KABINI:
+		amdgpu_program_register_sequence(adev,
+						 kalindi_mgcg_cgcg_init,
+						 (const u32)ARRAY_SIZE(kalindi_mgcg_cgcg_init));
+		amdgpu_program_register_sequence(adev,
+						 kalindi_golden_registers,
+						 (const u32)ARRAY_SIZE(kalindi_golden_registers));
+		amdgpu_program_register_sequence(adev,
+						 kalindi_golden_common_registers,
+						 (const u32)ARRAY_SIZE(kalindi_golden_common_registers));
+		amdgpu_program_register_sequence(adev,
+						 kalindi_golden_spm_registers,
+						 (const u32)ARRAY_SIZE(kalindi_golden_spm_registers));
+		break;
+	case CHIP_MULLINS:
+		amdgpu_program_register_sequence(adev,
+						 kalindi_mgcg_cgcg_init,
+						 (const u32)ARRAY_SIZE(kalindi_mgcg_cgcg_init));
+		amdgpu_program_register_sequence(adev,
+						 godavari_golden_registers,
+						 (const u32)ARRAY_SIZE(godavari_golden_registers));
+		amdgpu_program_register_sequence(adev,
+						 kalindi_golden_common_registers,
+						 (const u32)ARRAY_SIZE(kalindi_golden_common_registers));
+		amdgpu_program_register_sequence(adev,
+						 kalindi_golden_spm_registers,
+						 (const u32)ARRAY_SIZE(kalindi_golden_spm_registers));
+		break;
+	case CHIP_KAVERI:
+		amdgpu_program_register_sequence(adev,
+						 spectre_mgcg_cgcg_init,
+						 (const u32)ARRAY_SIZE(spectre_mgcg_cgcg_init));
+		amdgpu_program_register_sequence(adev,
+						 spectre_golden_registers,
+						 (const u32)ARRAY_SIZE(spectre_golden_registers));
+		amdgpu_program_register_sequence(adev,
+						 spectre_golden_common_registers,
+						 (const u32)ARRAY_SIZE(spectre_golden_common_registers));
+		amdgpu_program_register_sequence(adev,
+						 spectre_golden_spm_registers,
+						 (const u32)ARRAY_SIZE(spectre_golden_spm_registers));
+		break;
+	case CHIP_HAWAII:
+		amdgpu_program_register_sequence(adev,
+						 hawaii_mgcg_cgcg_init,
+						 (const u32)ARRAY_SIZE(hawaii_mgcg_cgcg_init));
+		amdgpu_program_register_sequence(adev,
+						 hawaii_golden_registers,
+						 (const u32)ARRAY_SIZE(hawaii_golden_registers));
+		amdgpu_program_register_sequence(adev,
+						 hawaii_golden_common_registers,
+						 (const u32)ARRAY_SIZE(hawaii_golden_common_registers));
+		amdgpu_program_register_sequence(adev,
+						 hawaii_golden_spm_registers,
+						 (const u32)ARRAY_SIZE(hawaii_golden_spm_registers));
+		break;
+	default:
+		break;
+	}
+	mutex_unlock(&adev->grbm_idx_mutex);
+}
+
+/**
+ * cik_get_xclk - get the xclk
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Returns the reference clock used by the gfx engine
+ * (CIK).
+ */
+static u32 cik_get_xclk(struct amdgpu_device *adev)
+{
+	u32 reference_clock = adev->clock.spll.reference_freq;
+
+	if (adev->flags & AMDGPU_IS_APU) {
+		if (RREG32_SMC(ixGENERAL_PWRMGT) & GENERAL_PWRMGT__GPU_COUNTER_CLK_MASK)
+			return reference_clock / 2;
+	} else {
+		if (RREG32_SMC(ixCG_CLKPIN_CNTL) & CG_CLKPIN_CNTL__XTALIN_DIVIDE_MASK)
+			return reference_clock / 4;
+	}
+	return reference_clock;
+}
+
+/**
+ * cik_srbm_select - select specific register instances
+ *
+ * @adev: amdgpu_device pointer
+ * @me: selected ME (micro engine)
+ * @pipe: pipe
+ * @queue: queue
+ * @vmid: VMID
+ *
+ * Switches the currently active registers instances.  Some
+ * registers are instanced per VMID, others are instanced per
+ * me/pipe/queue combination.
+ */
+void cik_srbm_select(struct amdgpu_device *adev,
+		     u32 me, u32 pipe, u32 queue, u32 vmid)
+{
+	u32 srbm_gfx_cntl =
+		(((pipe << SRBM_GFX_CNTL__PIPEID__SHIFT) & SRBM_GFX_CNTL__PIPEID_MASK)|
+		((me << SRBM_GFX_CNTL__MEID__SHIFT) & SRBM_GFX_CNTL__MEID_MASK)|
+		((vmid << SRBM_GFX_CNTL__VMID__SHIFT) & SRBM_GFX_CNTL__VMID_MASK)|
+		((queue << SRBM_GFX_CNTL__QUEUEID__SHIFT) & SRBM_GFX_CNTL__QUEUEID_MASK));
+	WREG32(mmSRBM_GFX_CNTL, srbm_gfx_cntl);
+}
+
+static void cik_vga_set_state(struct amdgpu_device *adev, bool state)
+{
+	uint32_t tmp;
+
+	tmp = RREG32(mmCONFIG_CNTL);
+	if (state == false)
+		tmp |= CONFIG_CNTL__VGA_DIS_MASK;
+	else
+		tmp &= ~CONFIG_CNTL__VGA_DIS_MASK;
+	WREG32(mmCONFIG_CNTL, tmp);
+}
+
+static bool cik_read_disabled_bios(struct amdgpu_device *adev)
+{
+	u32 bus_cntl;
+	u32 d1vga_control = 0;
+	u32 d2vga_control = 0;
+	u32 vga_render_control = 0;
+	u32 rom_cntl;
+	bool r;
+
+	bus_cntl = RREG32(mmBUS_CNTL);
+	if (adev->mode_info.num_crtc) {
+		d1vga_control = RREG32(mmD1VGA_CONTROL);
+		d2vga_control = RREG32(mmD2VGA_CONTROL);
+		vga_render_control = RREG32(mmVGA_RENDER_CONTROL);
+	}
+	rom_cntl = RREG32_SMC(ixROM_CNTL);
+
+	/* enable the rom */
+	WREG32(mmBUS_CNTL, (bus_cntl & ~BUS_CNTL__BIOS_ROM_DIS_MASK));
+	if (adev->mode_info.num_crtc) {
+		/* Disable VGA mode */
+		WREG32(mmD1VGA_CONTROL,
+		       (d1vga_control & ~(D1VGA_CONTROL__D1VGA_MODE_ENABLE_MASK |
+					  D1VGA_CONTROL__D1VGA_TIMING_SELECT_MASK)));
+		WREG32(mmD2VGA_CONTROL,
+		       (d2vga_control & ~(D1VGA_CONTROL__D1VGA_MODE_ENABLE_MASK |
+					  D1VGA_CONTROL__D1VGA_TIMING_SELECT_MASK)));
+		WREG32(mmVGA_RENDER_CONTROL,
+		       (vga_render_control & ~VGA_RENDER_CONTROL__VGA_VSTATUS_CNTL_MASK));
+	}
+	WREG32_SMC(ixROM_CNTL, rom_cntl | ROM_CNTL__SCK_OVERWRITE_MASK);
+
+	r = amdgpu_read_bios(adev);
+
+	/* restore regs */
+	WREG32(mmBUS_CNTL, bus_cntl);
+	if (adev->mode_info.num_crtc) {
+		WREG32(mmD1VGA_CONTROL, d1vga_control);
+		WREG32(mmD2VGA_CONTROL, d2vga_control);
+		WREG32(mmVGA_RENDER_CONTROL, vga_render_control);
+	}
+	WREG32_SMC(ixROM_CNTL, rom_cntl);
+	return r;
+}
+
+static struct amdgpu_allowed_register_entry cik_allowed_read_registers[] = {
+	{mmGRBM_STATUS, false},
+	{mmGB_ADDR_CONFIG, false},
+	{mmMC_ARB_RAMCFG, false},
+	{mmGB_TILE_MODE0, false},
+	{mmGB_TILE_MODE1, false},
+	{mmGB_TILE_MODE2, false},
+	{mmGB_TILE_MODE3, false},
+	{mmGB_TILE_MODE4, false},
+	{mmGB_TILE_MODE5, false},
+	{mmGB_TILE_MODE6, false},
+	{mmGB_TILE_MODE7, false},
+	{mmGB_TILE_MODE8, false},
+	{mmGB_TILE_MODE9, false},
+	{mmGB_TILE_MODE10, false},
+	{mmGB_TILE_MODE11, false},
+	{mmGB_TILE_MODE12, false},
+	{mmGB_TILE_MODE13, false},
+	{mmGB_TILE_MODE14, false},
+	{mmGB_TILE_MODE15, false},
+	{mmGB_TILE_MODE16, false},
+	{mmGB_TILE_MODE17, false},
+	{mmGB_TILE_MODE18, false},
+	{mmGB_TILE_MODE19, false},
+	{mmGB_TILE_MODE20, false},
+	{mmGB_TILE_MODE21, false},
+	{mmGB_TILE_MODE22, false},
+	{mmGB_TILE_MODE23, false},
+	{mmGB_TILE_MODE24, false},
+	{mmGB_TILE_MODE25, false},
+	{mmGB_TILE_MODE26, false},
+	{mmGB_TILE_MODE27, false},
+	{mmGB_TILE_MODE28, false},
+	{mmGB_TILE_MODE29, false},
+	{mmGB_TILE_MODE30, false},
+	{mmGB_TILE_MODE31, false},
+	{mmGB_MACROTILE_MODE0, false},
+	{mmGB_MACROTILE_MODE1, false},
+	{mmGB_MACROTILE_MODE2, false},
+	{mmGB_MACROTILE_MODE3, false},
+	{mmGB_MACROTILE_MODE4, false},
+	{mmGB_MACROTILE_MODE5, false},
+	{mmGB_MACROTILE_MODE6, false},
+	{mmGB_MACROTILE_MODE7, false},
+	{mmGB_MACROTILE_MODE8, false},
+	{mmGB_MACROTILE_MODE9, false},
+	{mmGB_MACROTILE_MODE10, false},
+	{mmGB_MACROTILE_MODE11, false},
+	{mmGB_MACROTILE_MODE12, false},
+	{mmGB_MACROTILE_MODE13, false},
+	{mmGB_MACROTILE_MODE14, false},
+	{mmGB_MACROTILE_MODE15, false},
+	{mmCC_RB_BACKEND_DISABLE, false, true},
+	{mmGC_USER_RB_BACKEND_DISABLE, false, true},
+	{mmGB_BACKEND_MAP, false, false},
+	{mmPA_SC_RASTER_CONFIG, false, true},
+	{mmPA_SC_RASTER_CONFIG_1, false, true},
+};
+
+static uint32_t cik_read_indexed_register(struct amdgpu_device *adev,
+					  u32 se_num, u32 sh_num,
+					  u32 reg_offset)
+{
+	uint32_t val;
+
+	mutex_lock(&adev->grbm_idx_mutex);
+	if (se_num != 0xffffffff || sh_num != 0xffffffff)
+		gfx_v7_0_select_se_sh(adev, se_num, sh_num);
+
+	val = RREG32(reg_offset);
+
+	if (se_num != 0xffffffff || sh_num != 0xffffffff)
+		gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff);
+	mutex_unlock(&adev->grbm_idx_mutex);
+	return val;
+}
+
+static int cik_read_register(struct amdgpu_device *adev, u32 se_num,
+			     u32 sh_num, u32 reg_offset, u32 *value)
+{
+	uint32_t i;
+
+	*value = 0;
+	for (i = 0; i < ARRAY_SIZE(cik_allowed_read_registers); i++) {
+		if (reg_offset != cik_allowed_read_registers[i].reg_offset)
+			continue;
+
+		if (!cik_allowed_read_registers[i].untouched)
+			*value = cik_allowed_read_registers[i].grbm_indexed ?
+				 cik_read_indexed_register(adev, se_num,
+							   sh_num, reg_offset) :
+				 RREG32(reg_offset);
+		return 0;
+	}
+	return -EINVAL;
+}
+
+static void cik_print_gpu_status_regs(struct amdgpu_device *adev)
+{
+	dev_info(adev->dev, "  GRBM_STATUS=0x%08X\n",
+		RREG32(mmGRBM_STATUS));
+	dev_info(adev->dev, "  GRBM_STATUS2=0x%08X\n",
+		RREG32(mmGRBM_STATUS2));
+	dev_info(adev->dev, "  GRBM_STATUS_SE0=0x%08X\n",
+		RREG32(mmGRBM_STATUS_SE0));
+	dev_info(adev->dev, "  GRBM_STATUS_SE1=0x%08X\n",
+		RREG32(mmGRBM_STATUS_SE1));
+	dev_info(adev->dev, "  GRBM_STATUS_SE2=0x%08X\n",
+		RREG32(mmGRBM_STATUS_SE2));
+	dev_info(adev->dev, "  GRBM_STATUS_SE3=0x%08X\n",
+		RREG32(mmGRBM_STATUS_SE3));
+	dev_info(adev->dev, "  SRBM_STATUS=0x%08X\n",
+		RREG32(mmSRBM_STATUS));
+	dev_info(adev->dev, "  SRBM_STATUS2=0x%08X\n",
+		RREG32(mmSRBM_STATUS2));
+	dev_info(adev->dev, "  SDMA0_STATUS_REG   = 0x%08X\n",
+		RREG32(mmSDMA0_STATUS_REG + SDMA0_REGISTER_OFFSET));
+	dev_info(adev->dev, "  SDMA1_STATUS_REG   = 0x%08X\n",
+		 RREG32(mmSDMA0_STATUS_REG + SDMA1_REGISTER_OFFSET));
+	dev_info(adev->dev, "  CP_STAT = 0x%08x\n", RREG32(mmCP_STAT));
+	dev_info(adev->dev, "  CP_STALLED_STAT1 = 0x%08x\n",
+		 RREG32(mmCP_STALLED_STAT1));
+	dev_info(adev->dev, "  CP_STALLED_STAT2 = 0x%08x\n",
+		 RREG32(mmCP_STALLED_STAT2));
+	dev_info(adev->dev, "  CP_STALLED_STAT3 = 0x%08x\n",
+		 RREG32(mmCP_STALLED_STAT3));
+	dev_info(adev->dev, "  CP_CPF_BUSY_STAT = 0x%08x\n",
+		 RREG32(mmCP_CPF_BUSY_STAT));
+	dev_info(adev->dev, "  CP_CPF_STALLED_STAT1 = 0x%08x\n",
+		 RREG32(mmCP_CPF_STALLED_STAT1));
+	dev_info(adev->dev, "  CP_CPF_STATUS = 0x%08x\n", RREG32(mmCP_CPF_STATUS));
+	dev_info(adev->dev, "  CP_CPC_BUSY_STAT = 0x%08x\n", RREG32(mmCP_CPC_BUSY_STAT));
+	dev_info(adev->dev, "  CP_CPC_STALLED_STAT1 = 0x%08x\n",
+		 RREG32(mmCP_CPC_STALLED_STAT1));
+	dev_info(adev->dev, "  CP_CPC_STATUS = 0x%08x\n", RREG32(mmCP_CPC_STATUS));
+}
+
+/**
+ * cik_gpu_check_soft_reset - check which blocks are busy
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Check which blocks are busy and return the relevant reset
+ * mask to be used by cik_gpu_soft_reset().
+ * Returns a mask of the blocks to be reset.
+ */
+u32 amdgpu_cik_gpu_check_soft_reset(struct amdgpu_device *adev)
+{
+	u32 reset_mask = 0;
+	u32 tmp;
+
+	/* GRBM_STATUS */
+	tmp = RREG32(mmGRBM_STATUS);
+	if (tmp & (GRBM_STATUS__PA_BUSY_MASK | GRBM_STATUS__SC_BUSY_MASK |
+		   GRBM_STATUS__BCI_BUSY_MASK | GRBM_STATUS__SX_BUSY_MASK |
+		   GRBM_STATUS__TA_BUSY_MASK | GRBM_STATUS__VGT_BUSY_MASK |
+		   GRBM_STATUS__DB_BUSY_MASK | GRBM_STATUS__CB_BUSY_MASK |
+		   GRBM_STATUS__GDS_BUSY_MASK | GRBM_STATUS__SPI_BUSY_MASK |
+		   GRBM_STATUS__IA_BUSY_MASK | GRBM_STATUS__IA_BUSY_NO_DMA_MASK))
+		reset_mask |= AMDGPU_RESET_GFX;
+
+	if (tmp & (GRBM_STATUS__CP_BUSY_MASK | GRBM_STATUS__CP_COHERENCY_BUSY_MASK))
+		reset_mask |= AMDGPU_RESET_CP;
+
+	/* GRBM_STATUS2 */
+	tmp = RREG32(mmGRBM_STATUS2);
+	if (tmp & GRBM_STATUS2__RLC_BUSY_MASK)
+		reset_mask |= AMDGPU_RESET_RLC;
+
+	/* SDMA0_STATUS_REG */
+	tmp = RREG32(mmSDMA0_STATUS_REG + SDMA0_REGISTER_OFFSET);
+	if (!(tmp & SDMA0_STATUS_REG__IDLE_MASK))
+		reset_mask |= AMDGPU_RESET_DMA;
+
+	/* SDMA1_STATUS_REG */
+	tmp = RREG32(mmSDMA0_STATUS_REG + SDMA1_REGISTER_OFFSET);
+	if (!(tmp & SDMA0_STATUS_REG__IDLE_MASK))
+		reset_mask |= AMDGPU_RESET_DMA1;
+
+	/* SRBM_STATUS2 */
+	tmp = RREG32(mmSRBM_STATUS2);
+	if (tmp & SRBM_STATUS2__SDMA_BUSY_MASK)
+		reset_mask |= AMDGPU_RESET_DMA;
+
+	if (tmp & SRBM_STATUS2__SDMA1_BUSY_MASK)
+		reset_mask |= AMDGPU_RESET_DMA1;
+
+	/* SRBM_STATUS */
+	tmp = RREG32(mmSRBM_STATUS);
+
+	if (tmp & SRBM_STATUS__IH_BUSY_MASK)
+		reset_mask |= AMDGPU_RESET_IH;
+
+	if (tmp & SRBM_STATUS__SEM_BUSY_MASK)
+		reset_mask |= AMDGPU_RESET_SEM;
+
+	if (tmp & SRBM_STATUS__GRBM_RQ_PENDING_MASK)
+		reset_mask |= AMDGPU_RESET_GRBM;
+
+	if (tmp & SRBM_STATUS__VMC_BUSY_MASK)
+		reset_mask |= AMDGPU_RESET_VMC;
+
+	if (tmp & (SRBM_STATUS__MCB_BUSY_MASK | SRBM_STATUS__MCB_NON_DISPLAY_BUSY_MASK |
+		   SRBM_STATUS__MCC_BUSY_MASK | SRBM_STATUS__MCD_BUSY_MASK))
+		reset_mask |= AMDGPU_RESET_MC;
+
+	if (amdgpu_display_is_display_hung(adev))
+		reset_mask |= AMDGPU_RESET_DISPLAY;
+
+	/* Skip MC reset as it's mostly likely not hung, just busy */
+	if (reset_mask & AMDGPU_RESET_MC) {
+		DRM_DEBUG("MC busy: 0x%08X, clearing.\n", reset_mask);
+		reset_mask &= ~AMDGPU_RESET_MC;
+	}
+
+	return reset_mask;
+}
+
+/**
+ * cik_gpu_soft_reset - soft reset GPU
+ *
+ * @adev: amdgpu_device pointer
+ * @reset_mask: mask of which blocks to reset
+ *
+ * Soft reset the blocks specified in @reset_mask.
+ */
+static void cik_gpu_soft_reset(struct amdgpu_device *adev, u32 reset_mask)
+{
+	struct amdgpu_mode_mc_save save;
+	u32 grbm_soft_reset = 0, srbm_soft_reset = 0;
+	u32 tmp;
+
+	if (reset_mask == 0)
+		return;
+
+	dev_info(adev->dev, "GPU softreset: 0x%08X\n", reset_mask);
+
+	cik_print_gpu_status_regs(adev);
+	dev_info(adev->dev, "  VM_CONTEXT1_PROTECTION_FAULT_ADDR   0x%08X\n",
+		 RREG32(mmVM_CONTEXT1_PROTECTION_FAULT_ADDR));
+	dev_info(adev->dev, "  VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x%08X\n",
+		 RREG32(mmVM_CONTEXT1_PROTECTION_FAULT_STATUS));
+
+	/* disable CG/PG */
+
+	/* stop the rlc */
+	gfx_v7_0_rlc_stop(adev);
+
+	/* Disable GFX parsing/prefetching */
+	WREG32(mmCP_ME_CNTL, CP_ME_CNTL__ME_HALT_MASK | CP_ME_CNTL__PFP_HALT_MASK | CP_ME_CNTL__CE_HALT_MASK);
+
+	/* Disable MEC parsing/prefetching */
+	WREG32(mmCP_MEC_CNTL, CP_MEC_CNTL__MEC_ME1_HALT_MASK | CP_MEC_CNTL__MEC_ME2_HALT_MASK);
+
+	if (reset_mask & AMDGPU_RESET_DMA) {
+		/* sdma0 */
+		tmp = RREG32(mmSDMA0_F32_CNTL + SDMA0_REGISTER_OFFSET);
+		tmp |= SDMA0_F32_CNTL__HALT_MASK;
+		WREG32(mmSDMA0_F32_CNTL + SDMA0_REGISTER_OFFSET, tmp);
+	}
+	if (reset_mask & AMDGPU_RESET_DMA1) {
+		/* sdma1 */
+		tmp = RREG32(mmSDMA0_F32_CNTL + SDMA1_REGISTER_OFFSET);
+		tmp |= SDMA0_F32_CNTL__HALT_MASK;
+		WREG32(mmSDMA0_F32_CNTL + SDMA1_REGISTER_OFFSET, tmp);
+	}
+
+	gmc_v7_0_mc_stop(adev, &save);
+	if (amdgpu_asic_wait_for_mc_idle(adev)) {
+		dev_warn(adev->dev, "Wait for MC idle timedout !\n");
+	}
+
+	if (reset_mask & (AMDGPU_RESET_GFX | AMDGPU_RESET_COMPUTE | AMDGPU_RESET_CP))
+		grbm_soft_reset = GRBM_SOFT_RESET__SOFT_RESET_CP_MASK |
+			GRBM_SOFT_RESET__SOFT_RESET_GFX_MASK;
+
+	if (reset_mask & AMDGPU_RESET_CP) {
+		grbm_soft_reset |= GRBM_SOFT_RESET__SOFT_RESET_CP_MASK;
+
+		srbm_soft_reset |= SRBM_SOFT_RESET__SOFT_RESET_GRBM_MASK;
+	}
+
+	if (reset_mask & AMDGPU_RESET_DMA)
+		srbm_soft_reset |= SRBM_SOFT_RESET__SOFT_RESET_SDMA_MASK;
+
+	if (reset_mask & AMDGPU_RESET_DMA1)
+		srbm_soft_reset |= SRBM_SOFT_RESET__SOFT_RESET_SDMA1_MASK;
+
+	if (reset_mask & AMDGPU_RESET_DISPLAY)
+		srbm_soft_reset |= SRBM_SOFT_RESET__SOFT_RESET_DC_MASK;
+
+	if (reset_mask & AMDGPU_RESET_RLC)
+		grbm_soft_reset |= GRBM_SOFT_RESET__SOFT_RESET_RLC_MASK;
+
+	if (reset_mask & AMDGPU_RESET_SEM)
+		srbm_soft_reset |= SRBM_SOFT_RESET__SOFT_RESET_SEM_MASK;
+
+	if (reset_mask & AMDGPU_RESET_IH)
+		srbm_soft_reset |= SRBM_SOFT_RESET__SOFT_RESET_IH_MASK;
+
+	if (reset_mask & AMDGPU_RESET_GRBM)
+		srbm_soft_reset |= SRBM_SOFT_RESET__SOFT_RESET_GRBM_MASK;
+
+	if (reset_mask & AMDGPU_RESET_VMC)
+		srbm_soft_reset |= SRBM_SOFT_RESET__SOFT_RESET_VMC_MASK;
+
+	if (!(adev->flags & AMDGPU_IS_APU)) {
+		if (reset_mask & AMDGPU_RESET_MC)
+			srbm_soft_reset |= SRBM_SOFT_RESET__SOFT_RESET_MC_MASK;
+	}
+
+	if (grbm_soft_reset) {
+		tmp = RREG32(mmGRBM_SOFT_RESET);
+		tmp |= grbm_soft_reset;
+		dev_info(adev->dev, "GRBM_SOFT_RESET=0x%08X\n", tmp);
+		WREG32(mmGRBM_SOFT_RESET, tmp);
+		tmp = RREG32(mmGRBM_SOFT_RESET);
+
+		udelay(50);
+
+		tmp &= ~grbm_soft_reset;
+		WREG32(mmGRBM_SOFT_RESET, tmp);
+		tmp = RREG32(mmGRBM_SOFT_RESET);
+	}
+
+	if (srbm_soft_reset) {
+		tmp = RREG32(mmSRBM_SOFT_RESET);
+		tmp |= srbm_soft_reset;
+		dev_info(adev->dev, "SRBM_SOFT_RESET=0x%08X\n", tmp);
+		WREG32(mmSRBM_SOFT_RESET, tmp);
+		tmp = RREG32(mmSRBM_SOFT_RESET);
+
+		udelay(50);
+
+		tmp &= ~srbm_soft_reset;
+		WREG32(mmSRBM_SOFT_RESET, tmp);
+		tmp = RREG32(mmSRBM_SOFT_RESET);
+	}
+
+	/* Wait a little for things to settle down */
+	udelay(50);
+
+	gmc_v7_0_mc_resume(adev, &save);
+	udelay(50);
+
+	cik_print_gpu_status_regs(adev);
+}
+
+struct kv_reset_save_regs {
+	u32 gmcon_reng_execute;
+	u32 gmcon_misc;
+	u32 gmcon_misc3;
+};
+
+static void kv_save_regs_for_reset(struct amdgpu_device *adev,
+				   struct kv_reset_save_regs *save)
+{
+	save->gmcon_reng_execute = RREG32(mmGMCON_RENG_EXECUTE);
+	save->gmcon_misc = RREG32(mmGMCON_MISC);
+	save->gmcon_misc3 = RREG32(mmGMCON_MISC3);
+
+	WREG32(mmGMCON_RENG_EXECUTE, save->gmcon_reng_execute &
+		~GMCON_RENG_EXECUTE__RENG_EXECUTE_ON_PWR_UP_MASK);
+	WREG32(mmGMCON_MISC, save->gmcon_misc &
+		~(GMCON_MISC__RENG_EXECUTE_ON_REG_UPDATE_MASK |
+			GMCON_MISC__STCTRL_STUTTER_EN_MASK));
+}
+
+static void kv_restore_regs_for_reset(struct amdgpu_device *adev,
+				      struct kv_reset_save_regs *save)
+{
+	int i;
+
+	WREG32(mmGMCON_PGFSM_WRITE, 0);
+	WREG32(mmGMCON_PGFSM_CONFIG, 0x200010ff);
+
+	for (i = 0; i < 5; i++)
+		WREG32(mmGMCON_PGFSM_WRITE, 0);
+
+	WREG32(mmGMCON_PGFSM_WRITE, 0);
+	WREG32(mmGMCON_PGFSM_CONFIG, 0x300010ff);
+
+	for (i = 0; i < 5; i++)
+		WREG32(mmGMCON_PGFSM_WRITE, 0);
+
+	WREG32(mmGMCON_PGFSM_WRITE, 0x210000);
+	WREG32(mmGMCON_PGFSM_CONFIG, 0xa00010ff);
+
+	for (i = 0; i < 5; i++)
+		WREG32(mmGMCON_PGFSM_WRITE, 0);
+
+	WREG32(mmGMCON_PGFSM_WRITE, 0x21003);
+	WREG32(mmGMCON_PGFSM_CONFIG, 0xb00010ff);
+
+	for (i = 0; i < 5; i++)
+		WREG32(mmGMCON_PGFSM_WRITE, 0);
+
+	WREG32(mmGMCON_PGFSM_WRITE, 0x2b00);
+	WREG32(mmGMCON_PGFSM_CONFIG, 0xc00010ff);
+
+	for (i = 0; i < 5; i++)
+		WREG32(mmGMCON_PGFSM_WRITE, 0);
+
+	WREG32(mmGMCON_PGFSM_WRITE, 0);
+	WREG32(mmGMCON_PGFSM_CONFIG, 0xd00010ff);
+
+	for (i = 0; i < 5; i++)
+		WREG32(mmGMCON_PGFSM_WRITE, 0);
+
+	WREG32(mmGMCON_PGFSM_WRITE, 0x420000);
+	WREG32(mmGMCON_PGFSM_CONFIG, 0x100010ff);
+
+	for (i = 0; i < 5; i++)
+		WREG32(mmGMCON_PGFSM_WRITE, 0);
+
+	WREG32(mmGMCON_PGFSM_WRITE, 0x120202);
+	WREG32(mmGMCON_PGFSM_CONFIG, 0x500010ff);
+
+	for (i = 0; i < 5; i++)
+		WREG32(mmGMCON_PGFSM_WRITE, 0);
+
+	WREG32(mmGMCON_PGFSM_WRITE, 0x3e3e36);
+	WREG32(mmGMCON_PGFSM_CONFIG, 0x600010ff);
+
+	for (i = 0; i < 5; i++)
+		WREG32(mmGMCON_PGFSM_WRITE, 0);
+
+	WREG32(mmGMCON_PGFSM_WRITE, 0x373f3e);
+	WREG32(mmGMCON_PGFSM_CONFIG, 0x700010ff);
+
+	for (i = 0; i < 5; i++)
+		WREG32(mmGMCON_PGFSM_WRITE, 0);
+
+	WREG32(mmGMCON_PGFSM_WRITE, 0x3e1332);
+	WREG32(mmGMCON_PGFSM_CONFIG, 0xe00010ff);
+
+	WREG32(mmGMCON_MISC3, save->gmcon_misc3);
+	WREG32(mmGMCON_MISC, save->gmcon_misc);
+	WREG32(mmGMCON_RENG_EXECUTE, save->gmcon_reng_execute);
+}
+
+static void cik_gpu_pci_config_reset(struct amdgpu_device *adev)
+{
+	struct amdgpu_mode_mc_save save;
+	struct kv_reset_save_regs kv_save = { 0 };
+	u32 tmp, i;
+
+	dev_info(adev->dev, "GPU pci config reset\n");
+
+	/* disable dpm? */
+
+	/* disable cg/pg */
+
+	/* Disable GFX parsing/prefetching */
+	WREG32(mmCP_ME_CNTL, CP_ME_CNTL__ME_HALT_MASK |
+		CP_ME_CNTL__PFP_HALT_MASK | CP_ME_CNTL__CE_HALT_MASK);
+
+	/* Disable MEC parsing/prefetching */
+	WREG32(mmCP_MEC_CNTL,
+			CP_MEC_CNTL__MEC_ME1_HALT_MASK | CP_MEC_CNTL__MEC_ME2_HALT_MASK);
+
+	/* sdma0 */
+	tmp = RREG32(mmSDMA0_F32_CNTL + SDMA0_REGISTER_OFFSET);
+	tmp |= SDMA0_F32_CNTL__HALT_MASK;
+	WREG32(mmSDMA0_F32_CNTL + SDMA0_REGISTER_OFFSET, tmp);
+	/* sdma1 */
+	tmp = RREG32(mmSDMA0_F32_CNTL + SDMA1_REGISTER_OFFSET);
+	tmp |= SDMA0_F32_CNTL__HALT_MASK;
+	WREG32(mmSDMA0_F32_CNTL + SDMA1_REGISTER_OFFSET, tmp);
+	/* XXX other engines? */
+
+	/* halt the rlc, disable cp internal ints */
+	gfx_v7_0_rlc_stop(adev);
+
+	udelay(50);
+
+	/* disable mem access */
+	gmc_v7_0_mc_stop(adev, &save);
+	if (amdgpu_asic_wait_for_mc_idle(adev)) {
+		dev_warn(adev->dev, "Wait for MC idle timed out !\n");
+	}
+
+	if (adev->flags & AMDGPU_IS_APU)
+		kv_save_regs_for_reset(adev, &kv_save);
+
+	/* disable BM */
+	pci_clear_master(adev->pdev);
+	/* reset */
+	amdgpu_pci_config_reset(adev);
+
+	udelay(100);
+
+	/* wait for asic to come out of reset */
+	for (i = 0; i < adev->usec_timeout; i++) {
+		if (RREG32(mmCONFIG_MEMSIZE) != 0xffffffff)
+			break;
+		udelay(1);
+	}
+
+	/* does asic init need to be run first??? */
+	if (adev->flags & AMDGPU_IS_APU)
+		kv_restore_regs_for_reset(adev, &kv_save);
+}
+
+static void cik_set_bios_scratch_engine_hung(struct amdgpu_device *adev, bool hung)
+{
+	u32 tmp = RREG32(mmBIOS_SCRATCH_3);
+
+	if (hung)
+		tmp |= ATOM_S3_ASIC_GUI_ENGINE_HUNG;
+	else
+		tmp &= ~ATOM_S3_ASIC_GUI_ENGINE_HUNG;
+
+	WREG32(mmBIOS_SCRATCH_3, tmp);
+}
+
+/**
+ * cik_asic_reset - soft reset GPU
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Look up which blocks are hung and attempt
+ * to reset them.
+ * Returns 0 for success.
+ */
+static int cik_asic_reset(struct amdgpu_device *adev)
+{
+	u32 reset_mask;
+
+	reset_mask = amdgpu_cik_gpu_check_soft_reset(adev);
+
+	if (reset_mask)
+		cik_set_bios_scratch_engine_hung(adev, true);
+
+	/* try soft reset */
+	cik_gpu_soft_reset(adev, reset_mask);
+
+	reset_mask = amdgpu_cik_gpu_check_soft_reset(adev);
+
+	/* try pci config reset */
+	if (reset_mask && amdgpu_hard_reset)
+		cik_gpu_pci_config_reset(adev);
+
+	reset_mask = amdgpu_cik_gpu_check_soft_reset(adev);
+
+	if (!reset_mask)
+		cik_set_bios_scratch_engine_hung(adev, false);
+
+	return 0;
+}
+
+static int cik_set_uvd_clock(struct amdgpu_device *adev, u32 clock,
+			      u32 cntl_reg, u32 status_reg)
+{
+	int r, i;
+	struct atom_clock_dividers dividers;
+	uint32_t tmp;
+
+	r = amdgpu_atombios_get_clock_dividers(adev,
+					       COMPUTE_GPUCLK_INPUT_FLAG_DEFAULT_GPUCLK,
+					       clock, false, &dividers);
+	if (r)
+		return r;
+
+	tmp = RREG32_SMC(cntl_reg);
+	tmp &= ~(CG_DCLK_CNTL__DCLK_DIR_CNTL_EN_MASK |
+		CG_DCLK_CNTL__DCLK_DIVIDER_MASK);
+	tmp |= dividers.post_divider;
+	WREG32_SMC(cntl_reg, tmp);
+
+	for (i = 0; i < 100; i++) {
+		if (RREG32_SMC(status_reg) & CG_DCLK_STATUS__DCLK_STATUS_MASK)
+			break;
+		mdelay(10);
+	}
+	if (i == 100)
+		return -ETIMEDOUT;
+
+	return 0;
+}
+
+static int cik_set_uvd_clocks(struct amdgpu_device *adev, u32 vclk, u32 dclk)
+{
+	int r = 0;
+
+	r = cik_set_uvd_clock(adev, vclk, ixCG_VCLK_CNTL, ixCG_VCLK_STATUS);
+	if (r)
+		return r;
+
+	r = cik_set_uvd_clock(adev, dclk, ixCG_DCLK_CNTL, ixCG_DCLK_STATUS);
+	return r;
+}
+
+static int cik_set_vce_clocks(struct amdgpu_device *adev, u32 evclk, u32 ecclk)
+{
+	int r, i;
+	struct atom_clock_dividers dividers;
+	u32 tmp;
+
+	r = amdgpu_atombios_get_clock_dividers(adev,
+					       COMPUTE_GPUCLK_INPUT_FLAG_DEFAULT_GPUCLK,
+					       ecclk, false, &dividers);
+	if (r)
+		return r;
+
+	for (i = 0; i < 100; i++) {
+		if (RREG32_SMC(ixCG_ECLK_STATUS) & CG_ECLK_STATUS__ECLK_STATUS_MASK)
+			break;
+		mdelay(10);
+	}
+	if (i == 100)
+		return -ETIMEDOUT;
+
+	tmp = RREG32_SMC(ixCG_ECLK_CNTL);
+	tmp &= ~(CG_ECLK_CNTL__ECLK_DIR_CNTL_EN_MASK |
+		CG_ECLK_CNTL__ECLK_DIVIDER_MASK);
+	tmp |= dividers.post_divider;
+	WREG32_SMC(ixCG_ECLK_CNTL, tmp);
+
+	for (i = 0; i < 100; i++) {
+		if (RREG32_SMC(ixCG_ECLK_STATUS) & CG_ECLK_STATUS__ECLK_STATUS_MASK)
+			break;
+		mdelay(10);
+	}
+	if (i == 100)
+		return -ETIMEDOUT;
+
+	return 0;
+}
+
+static void cik_pcie_gen3_enable(struct amdgpu_device *adev)
+{
+	struct pci_dev *root = adev->pdev->bus->self;
+	int bridge_pos, gpu_pos;
+	u32 speed_cntl, mask, current_data_rate;
+	int ret, i;
+	u16 tmp16;
+
+	if (amdgpu_pcie_gen2 == 0)
+		return;
+
+	if (adev->flags & AMDGPU_IS_APU)
+		return;
+
+	ret = drm_pcie_get_speed_cap_mask(adev->ddev, &mask);
+	if (ret != 0)
+		return;
+
+	if (!(mask & (DRM_PCIE_SPEED_50 | DRM_PCIE_SPEED_80)))
+		return;
+
+	speed_cntl = RREG32_PCIE(ixPCIE_LC_SPEED_CNTL);
+	current_data_rate = (speed_cntl & PCIE_LC_SPEED_CNTL__LC_CURRENT_DATA_RATE_MASK) >>
+		PCIE_LC_SPEED_CNTL__LC_CURRENT_DATA_RATE__SHIFT;
+	if (mask & DRM_PCIE_SPEED_80) {
+		if (current_data_rate == 2) {
+			DRM_INFO("PCIE gen 3 link speeds already enabled\n");
+			return;
+		}
+		DRM_INFO("enabling PCIE gen 3 link speeds, disable with amdgpu.pcie_gen2=0\n");
+	} else if (mask & DRM_PCIE_SPEED_50) {
+		if (current_data_rate == 1) {
+			DRM_INFO("PCIE gen 2 link speeds already enabled\n");
+			return;
+		}
+		DRM_INFO("enabling PCIE gen 2 link speeds, disable with amdgpu.pcie_gen2=0\n");
+	}
+
+	bridge_pos = pci_pcie_cap(root);
+	if (!bridge_pos)
+		return;
+
+	gpu_pos = pci_pcie_cap(adev->pdev);
+	if (!gpu_pos)
+		return;
+
+	if (mask & DRM_PCIE_SPEED_80) {
+		/* re-try equalization if gen3 is not already enabled */
+		if (current_data_rate != 2) {
+			u16 bridge_cfg, gpu_cfg;
+			u16 bridge_cfg2, gpu_cfg2;
+			u32 max_lw, current_lw, tmp;
+
+			pci_read_config_word(root, bridge_pos + PCI_EXP_LNKCTL, &bridge_cfg);
+			pci_read_config_word(adev->pdev, gpu_pos + PCI_EXP_LNKCTL, &gpu_cfg);
+
+			tmp16 = bridge_cfg | PCI_EXP_LNKCTL_HAWD;
+			pci_write_config_word(root, bridge_pos + PCI_EXP_LNKCTL, tmp16);
+
+			tmp16 = gpu_cfg | PCI_EXP_LNKCTL_HAWD;
+			pci_write_config_word(adev->pdev, gpu_pos + PCI_EXP_LNKCTL, tmp16);
+
+			tmp = RREG32_PCIE(ixPCIE_LC_STATUS1);
+			max_lw = (tmp & PCIE_LC_STATUS1__LC_DETECTED_LINK_WIDTH_MASK) >>
+				PCIE_LC_STATUS1__LC_DETECTED_LINK_WIDTH__SHIFT;
+			current_lw = (tmp & PCIE_LC_STATUS1__LC_OPERATING_LINK_WIDTH_MASK)
+				>> PCIE_LC_STATUS1__LC_OPERATING_LINK_WIDTH__SHIFT;
+
+			if (current_lw < max_lw) {
+				tmp = RREG32_PCIE(ixPCIE_LC_LINK_WIDTH_CNTL);
+				if (tmp & PCIE_LC_LINK_WIDTH_CNTL__LC_RENEGOTIATION_SUPPORT_MASK) {
+					tmp &= ~(PCIE_LC_LINK_WIDTH_CNTL__LC_LINK_WIDTH_MASK |
+						PCIE_LC_LINK_WIDTH_CNTL__LC_UPCONFIGURE_DIS_MASK);
+					tmp |= (max_lw <<
+						PCIE_LC_LINK_WIDTH_CNTL__LC_LINK_WIDTH__SHIFT);
+					tmp |= PCIE_LC_LINK_WIDTH_CNTL__LC_UPCONFIGURE_SUPPORT_MASK |
+					PCIE_LC_LINK_WIDTH_CNTL__LC_RENEGOTIATE_EN_MASK |
+					PCIE_LC_LINK_WIDTH_CNTL__LC_RECONFIG_NOW_MASK;
+					WREG32_PCIE(ixPCIE_LC_LINK_WIDTH_CNTL, tmp);
+				}
+			}
+
+			for (i = 0; i < 10; i++) {
+				/* check status */
+				pci_read_config_word(adev->pdev, gpu_pos + PCI_EXP_DEVSTA, &tmp16);
+				if (tmp16 & PCI_EXP_DEVSTA_TRPND)
+					break;
+
+				pci_read_config_word(root, bridge_pos + PCI_EXP_LNKCTL, &bridge_cfg);
+				pci_read_config_word(adev->pdev, gpu_pos + PCI_EXP_LNKCTL, &gpu_cfg);
+
+				pci_read_config_word(root, bridge_pos + PCI_EXP_LNKCTL2, &bridge_cfg2);
+				pci_read_config_word(adev->pdev, gpu_pos + PCI_EXP_LNKCTL2, &gpu_cfg2);
+
+				tmp = RREG32_PCIE(ixPCIE_LC_CNTL4);
+				tmp |= PCIE_LC_CNTL4__LC_SET_QUIESCE_MASK;
+				WREG32_PCIE(ixPCIE_LC_CNTL4, tmp);
+
+				tmp = RREG32_PCIE(ixPCIE_LC_CNTL4);
+				tmp |= PCIE_LC_CNTL4__LC_REDO_EQ_MASK;
+				WREG32_PCIE(ixPCIE_LC_CNTL4, tmp);
+
+				mdelay(100);
+
+				/* linkctl */
+				pci_read_config_word(root, bridge_pos + PCI_EXP_LNKCTL, &tmp16);
+				tmp16 &= ~PCI_EXP_LNKCTL_HAWD;
+				tmp16 |= (bridge_cfg & PCI_EXP_LNKCTL_HAWD);
+				pci_write_config_word(root, bridge_pos + PCI_EXP_LNKCTL, tmp16);
+
+				pci_read_config_word(adev->pdev, gpu_pos + PCI_EXP_LNKCTL, &tmp16);
+				tmp16 &= ~PCI_EXP_LNKCTL_HAWD;
+				tmp16 |= (gpu_cfg & PCI_EXP_LNKCTL_HAWD);
+				pci_write_config_word(adev->pdev, gpu_pos + PCI_EXP_LNKCTL, tmp16);
+
+				/* linkctl2 */
+				pci_read_config_word(root, bridge_pos + PCI_EXP_LNKCTL2, &tmp16);
+				tmp16 &= ~((1 << 4) | (7 << 9));
+				tmp16 |= (bridge_cfg2 & ((1 << 4) | (7 << 9)));
+				pci_write_config_word(root, bridge_pos + PCI_EXP_LNKCTL2, tmp16);
+
+				pci_read_config_word(adev->pdev, gpu_pos + PCI_EXP_LNKCTL2, &tmp16);
+				tmp16 &= ~((1 << 4) | (7 << 9));
+				tmp16 |= (gpu_cfg2 & ((1 << 4) | (7 << 9)));
+				pci_write_config_word(adev->pdev, gpu_pos + PCI_EXP_LNKCTL2, tmp16);
+
+				tmp = RREG32_PCIE(ixPCIE_LC_CNTL4);
+				tmp &= ~PCIE_LC_CNTL4__LC_SET_QUIESCE_MASK;
+				WREG32_PCIE(ixPCIE_LC_CNTL4, tmp);
+			}
+		}
+	}
+
+	/* set the link speed */
+	speed_cntl |= PCIE_LC_SPEED_CNTL__LC_FORCE_EN_SW_SPEED_CHANGE_MASK |
+		PCIE_LC_SPEED_CNTL__LC_FORCE_DIS_HW_SPEED_CHANGE_MASK;
+	speed_cntl &= ~PCIE_LC_SPEED_CNTL__LC_FORCE_DIS_SW_SPEED_CHANGE_MASK;
+	WREG32_PCIE(ixPCIE_LC_SPEED_CNTL, speed_cntl);
+
+	pci_read_config_word(adev->pdev, gpu_pos + PCI_EXP_LNKCTL2, &tmp16);
+	tmp16 &= ~0xf;
+	if (mask & DRM_PCIE_SPEED_80)
+		tmp16 |= 3; /* gen3 */
+	else if (mask & DRM_PCIE_SPEED_50)
+		tmp16 |= 2; /* gen2 */
+	else
+		tmp16 |= 1; /* gen1 */
+	pci_write_config_word(adev->pdev, gpu_pos + PCI_EXP_LNKCTL2, tmp16);
+
+	speed_cntl = RREG32_PCIE(ixPCIE_LC_SPEED_CNTL);
+	speed_cntl |= PCIE_LC_SPEED_CNTL__LC_INITIATE_LINK_SPEED_CHANGE_MASK;
+	WREG32_PCIE(ixPCIE_LC_SPEED_CNTL, speed_cntl);
+
+	for (i = 0; i < adev->usec_timeout; i++) {
+		speed_cntl = RREG32_PCIE(ixPCIE_LC_SPEED_CNTL);
+		if ((speed_cntl & PCIE_LC_SPEED_CNTL__LC_INITIATE_LINK_SPEED_CHANGE_MASK) == 0)
+			break;
+		udelay(1);
+	}
+}
+
+static void cik_program_aspm(struct amdgpu_device *adev)
+{
+	u32 data, orig;
+	bool disable_l0s = false, disable_l1 = false, disable_plloff_in_l1 = false;
+	bool disable_clkreq = false;
+
+	if (amdgpu_aspm == 0)
+		return;
+
+	/* XXX double check APUs */
+	if (adev->flags & AMDGPU_IS_APU)
+		return;
+
+	orig = data = RREG32_PCIE(ixPCIE_LC_N_FTS_CNTL);
+	data &= ~PCIE_LC_N_FTS_CNTL__LC_XMIT_N_FTS_MASK;
+	data |= (0x24 << PCIE_LC_N_FTS_CNTL__LC_XMIT_N_FTS__SHIFT) |
+		PCIE_LC_N_FTS_CNTL__LC_XMIT_N_FTS_OVERRIDE_EN_MASK;
+	if (orig != data)
+		WREG32_PCIE(ixPCIE_LC_N_FTS_CNTL, data);
+
+	orig = data = RREG32_PCIE(ixPCIE_LC_CNTL3);
+	data |= PCIE_LC_CNTL3__LC_GO_TO_RECOVERY_MASK;
+	if (orig != data)
+		WREG32_PCIE(ixPCIE_LC_CNTL3, data);
+
+	orig = data = RREG32_PCIE(ixPCIE_P_CNTL);
+	data |= PCIE_P_CNTL__P_IGNORE_EDB_ERR_MASK;
+	if (orig != data)
+		WREG32_PCIE(ixPCIE_P_CNTL, data);
+
+	orig = data = RREG32_PCIE(ixPCIE_LC_CNTL);
+	data &= ~(PCIE_LC_CNTL__LC_L0S_INACTIVITY_MASK |
+		PCIE_LC_CNTL__LC_L1_INACTIVITY_MASK);
+	data |= PCIE_LC_CNTL__LC_PMI_TO_L1_DIS_MASK;
+	if (!disable_l0s)
+		data |= (7 << PCIE_LC_CNTL__LC_L0S_INACTIVITY__SHIFT);
+
+	if (!disable_l1) {
+		data |= (7 << PCIE_LC_CNTL__LC_L1_INACTIVITY__SHIFT);
+		data &= ~PCIE_LC_CNTL__LC_PMI_TO_L1_DIS_MASK;
+		if (orig != data)
+			WREG32_PCIE(ixPCIE_LC_CNTL, data);
+
+		if (!disable_plloff_in_l1) {
+			bool clk_req_support;
+
+			orig = data = RREG32_PCIE(ixPB0_PIF_PWRDOWN_0);
+			data &= ~(PB0_PIF_PWRDOWN_0__PLL_POWER_STATE_IN_OFF_0_MASK |
+				PB0_PIF_PWRDOWN_0__PLL_POWER_STATE_IN_TXS2_0_MASK);
+			data |= (7 << PB0_PIF_PWRDOWN_0__PLL_POWER_STATE_IN_OFF_0__SHIFT) |
+				(7 << PB0_PIF_PWRDOWN_0__PLL_POWER_STATE_IN_TXS2_0__SHIFT);
+			if (orig != data)
+				WREG32_PCIE(ixPB0_PIF_PWRDOWN_0, data);
+
+			orig = data = RREG32_PCIE(ixPB0_PIF_PWRDOWN_1);
+			data &= ~(PB0_PIF_PWRDOWN_1__PLL_POWER_STATE_IN_OFF_1_MASK |
+				PB0_PIF_PWRDOWN_1__PLL_POWER_STATE_IN_TXS2_1_MASK);
+			data |= (7 << PB0_PIF_PWRDOWN_1__PLL_POWER_STATE_IN_OFF_1__SHIFT) |
+				(7 << PB0_PIF_PWRDOWN_1__PLL_POWER_STATE_IN_TXS2_1__SHIFT);
+			if (orig != data)
+				WREG32_PCIE(ixPB0_PIF_PWRDOWN_1, data);
+
+			orig = data = RREG32_PCIE(ixPB1_PIF_PWRDOWN_0);
+			data &= ~(PB1_PIF_PWRDOWN_0__PLL_POWER_STATE_IN_OFF_0_MASK |
+				PB1_PIF_PWRDOWN_0__PLL_POWER_STATE_IN_TXS2_0_MASK);
+			data |= (7 << PB1_PIF_PWRDOWN_0__PLL_POWER_STATE_IN_OFF_0__SHIFT) |
+				(7 << PB1_PIF_PWRDOWN_0__PLL_POWER_STATE_IN_TXS2_0__SHIFT);
+			if (orig != data)
+				WREG32_PCIE(ixPB1_PIF_PWRDOWN_0, data);
+
+			orig = data = RREG32_PCIE(ixPB1_PIF_PWRDOWN_1);
+			data &= ~(PB1_PIF_PWRDOWN_1__PLL_POWER_STATE_IN_OFF_1_MASK |
+				PB1_PIF_PWRDOWN_1__PLL_POWER_STATE_IN_TXS2_1_MASK);
+			data |= (7 << PB1_PIF_PWRDOWN_1__PLL_POWER_STATE_IN_OFF_1__SHIFT) |
+				(7 << PB1_PIF_PWRDOWN_1__PLL_POWER_STATE_IN_TXS2_1__SHIFT);
+			if (orig != data)
+				WREG32_PCIE(ixPB1_PIF_PWRDOWN_1, data);
+
+			orig = data = RREG32_PCIE(ixPCIE_LC_LINK_WIDTH_CNTL);
+			data &= ~PCIE_LC_LINK_WIDTH_CNTL__LC_DYN_LANES_PWR_STATE_MASK;
+			data |= ~(3 << PCIE_LC_LINK_WIDTH_CNTL__LC_DYN_LANES_PWR_STATE__SHIFT);
+			if (orig != data)
+				WREG32_PCIE(ixPCIE_LC_LINK_WIDTH_CNTL, data);
+
+			if (!disable_clkreq) {
+				struct pci_dev *root = adev->pdev->bus->self;
+				u32 lnkcap;
+
+				clk_req_support = false;
+				pcie_capability_read_dword(root, PCI_EXP_LNKCAP, &lnkcap);
+				if (lnkcap & PCI_EXP_LNKCAP_CLKPM)
+					clk_req_support = true;
+			} else {
+				clk_req_support = false;
+			}
+
+			if (clk_req_support) {
+				orig = data = RREG32_PCIE(ixPCIE_LC_CNTL2);
+				data |= PCIE_LC_CNTL2__LC_ALLOW_PDWN_IN_L1_MASK |
+					PCIE_LC_CNTL2__LC_ALLOW_PDWN_IN_L23_MASK;
+				if (orig != data)
+					WREG32_PCIE(ixPCIE_LC_CNTL2, data);
+
+				orig = data = RREG32_SMC(ixTHM_CLK_CNTL);
+				data &= ~(THM_CLK_CNTL__CMON_CLK_SEL_MASK |
+					THM_CLK_CNTL__TMON_CLK_SEL_MASK);
+				data |= (1 << THM_CLK_CNTL__CMON_CLK_SEL__SHIFT) |
+					(1 << THM_CLK_CNTL__TMON_CLK_SEL__SHIFT);
+				if (orig != data)
+					WREG32_SMC(ixTHM_CLK_CNTL, data);
+
+				orig = data = RREG32_SMC(ixMISC_CLK_CTRL);
+				data &= ~(MISC_CLK_CTRL__DEEP_SLEEP_CLK_SEL_MASK |
+					MISC_CLK_CTRL__ZCLK_SEL_MASK);
+				data |= (1 << MISC_CLK_CTRL__DEEP_SLEEP_CLK_SEL__SHIFT) |
+					(1 << MISC_CLK_CTRL__ZCLK_SEL__SHIFT);
+				if (orig != data)
+					WREG32_SMC(ixMISC_CLK_CTRL, data);
+
+				orig = data = RREG32_SMC(ixCG_CLKPIN_CNTL);
+				data &= ~CG_CLKPIN_CNTL__BCLK_AS_XCLK_MASK;
+				if (orig != data)
+					WREG32_SMC(ixCG_CLKPIN_CNTL, data);
+
+				orig = data = RREG32_SMC(ixCG_CLKPIN_CNTL_2);
+				data &= ~CG_CLKPIN_CNTL_2__FORCE_BIF_REFCLK_EN_MASK;
+				if (orig != data)
+					WREG32_SMC(ixCG_CLKPIN_CNTL_2, data);
+
+				orig = data = RREG32_SMC(ixMPLL_BYPASSCLK_SEL);
+				data &= ~MPLL_BYPASSCLK_SEL__MPLL_CLKOUT_SEL_MASK;
+				data |= (4 << MPLL_BYPASSCLK_SEL__MPLL_CLKOUT_SEL__SHIFT);
+				if (orig != data)
+					WREG32_SMC(ixMPLL_BYPASSCLK_SEL, data);
+			}
+		}
+	} else {
+		if (orig != data)
+			WREG32_PCIE(ixPCIE_LC_CNTL, data);
+	}
+
+	orig = data = RREG32_PCIE(ixPCIE_CNTL2);
+	data |= PCIE_CNTL2__SLV_MEM_LS_EN_MASK |
+		PCIE_CNTL2__MST_MEM_LS_EN_MASK |
+		PCIE_CNTL2__REPLAY_MEM_LS_EN_MASK;
+	if (orig != data)
+		WREG32_PCIE(ixPCIE_CNTL2, data);
+
+	if (!disable_l0s) {
+		data = RREG32_PCIE(ixPCIE_LC_N_FTS_CNTL);
+		if ((data & PCIE_LC_N_FTS_CNTL__LC_N_FTS_MASK) ==
+				PCIE_LC_N_FTS_CNTL__LC_N_FTS_MASK) {
+			data = RREG32_PCIE(ixPCIE_LC_STATUS1);
+			if ((data & PCIE_LC_STATUS1__LC_REVERSE_XMIT_MASK) &&
+			(data & PCIE_LC_STATUS1__LC_REVERSE_RCVR_MASK)) {
+				orig = data = RREG32_PCIE(ixPCIE_LC_CNTL);
+				data &= ~PCIE_LC_CNTL__LC_L0S_INACTIVITY_MASK;
+				if (orig != data)
+					WREG32_PCIE(ixPCIE_LC_CNTL, data);
+			}
+		}
+	}
+}
+
+static uint32_t cik_get_rev_id(struct amdgpu_device *adev)
+{
+	return (RREG32(mmCC_DRM_ID_STRAPS) & CC_DRM_ID_STRAPS__ATI_REV_ID_MASK)
+		>> CC_DRM_ID_STRAPS__ATI_REV_ID__SHIFT;
+}
+
+static const struct amdgpu_ip_block_version bonaire_ip_blocks[] =
+{
+	/* ORDER MATTERS! */
+	{
+		.type = AMDGPU_IP_BLOCK_TYPE_COMMON,
+		.major = 1,
+		.minor = 0,
+		.rev = 0,
+		.funcs = &cik_common_ip_funcs,
+	},
+	{
+		.type = AMDGPU_IP_BLOCK_TYPE_GMC,
+		.major = 7,
+		.minor = 0,
+		.rev = 0,
+		.funcs = &gmc_v7_0_ip_funcs,
+	},
+	{
+		.type = AMDGPU_IP_BLOCK_TYPE_IH,
+		.major = 2,
+		.minor = 0,
+		.rev = 0,
+		.funcs = &cik_ih_ip_funcs,
+	},
+	{
+		.type = AMDGPU_IP_BLOCK_TYPE_SMC,
+		.major = 7,
+		.minor = 0,
+		.rev = 0,
+		.funcs = &ci_dpm_ip_funcs,
+	},
+	{
+		.type = AMDGPU_IP_BLOCK_TYPE_DCE,
+		.major = 8,
+		.minor = 2,
+		.rev = 0,
+		.funcs = &dce_v8_0_ip_funcs,
+	},
+	{
+		.type = AMDGPU_IP_BLOCK_TYPE_GFX,
+		.major = 7,
+		.minor = 2,
+		.rev = 0,
+		.funcs = &gfx_v7_0_ip_funcs,
+	},
+	{
+		.type = AMDGPU_IP_BLOCK_TYPE_SDMA,
+		.major = 2,
+		.minor = 0,
+		.rev = 0,
+		.funcs = &cik_sdma_ip_funcs,
+	},
+	{
+		.type = AMDGPU_IP_BLOCK_TYPE_UVD,
+		.major = 4,
+		.minor = 2,
+		.rev = 0,
+		.funcs = &uvd_v4_2_ip_funcs,
+	},
+	{
+		.type = AMDGPU_IP_BLOCK_TYPE_VCE,
+		.major = 2,
+		.minor = 0,
+		.rev = 0,
+		.funcs = &vce_v2_0_ip_funcs,
+	},
+};
+
+static const struct amdgpu_ip_block_version hawaii_ip_blocks[] =
+{
+	/* ORDER MATTERS! */
+	{
+		.type = AMDGPU_IP_BLOCK_TYPE_COMMON,
+		.major = 1,
+		.minor = 0,
+		.rev = 0,
+		.funcs = &cik_common_ip_funcs,
+	},
+	{
+		.type = AMDGPU_IP_BLOCK_TYPE_GMC,
+		.major = 7,
+		.minor = 0,
+		.rev = 0,
+		.funcs = &gmc_v7_0_ip_funcs,
+	},
+	{
+		.type = AMDGPU_IP_BLOCK_TYPE_IH,
+		.major = 2,
+		.minor = 0,
+		.rev = 0,
+		.funcs = &cik_ih_ip_funcs,
+	},
+	{
+		.type = AMDGPU_IP_BLOCK_TYPE_SMC,
+		.major = 7,
+		.minor = 0,
+		.rev = 0,
+		.funcs = &ci_dpm_ip_funcs,
+	},
+	{
+		.type = AMDGPU_IP_BLOCK_TYPE_DCE,
+		.major = 8,
+		.minor = 5,
+		.rev = 0,
+		.funcs = &dce_v8_0_ip_funcs,
+	},
+	{
+		.type = AMDGPU_IP_BLOCK_TYPE_GFX,
+		.major = 7,
+		.minor = 3,
+		.rev = 0,
+		.funcs = &gfx_v7_0_ip_funcs,
+	},
+	{
+		.type = AMDGPU_IP_BLOCK_TYPE_SDMA,
+		.major = 2,
+		.minor = 0,
+		.rev = 0,
+		.funcs = &cik_sdma_ip_funcs,
+	},
+	{
+		.type = AMDGPU_IP_BLOCK_TYPE_UVD,
+		.major = 4,
+		.minor = 2,
+		.rev = 0,
+		.funcs = &uvd_v4_2_ip_funcs,
+	},
+	{
+		.type = AMDGPU_IP_BLOCK_TYPE_VCE,
+		.major = 2,
+		.minor = 0,
+		.rev = 0,
+		.funcs = &vce_v2_0_ip_funcs,
+	},
+};
+
+static const struct amdgpu_ip_block_version kabini_ip_blocks[] =
+{
+	/* ORDER MATTERS! */
+	{
+		.type = AMDGPU_IP_BLOCK_TYPE_COMMON,
+		.major = 1,
+		.minor = 0,
+		.rev = 0,
+		.funcs = &cik_common_ip_funcs,
+	},
+	{
+		.type = AMDGPU_IP_BLOCK_TYPE_GMC,
+		.major = 7,
+		.minor = 0,
+		.rev = 0,
+		.funcs = &gmc_v7_0_ip_funcs,
+	},
+	{
+		.type = AMDGPU_IP_BLOCK_TYPE_IH,
+		.major = 2,
+		.minor = 0,
+		.rev = 0,
+		.funcs = &cik_ih_ip_funcs,
+	},
+	{
+		.type = AMDGPU_IP_BLOCK_TYPE_SMC,
+		.major = 7,
+		.minor = 0,
+		.rev = 0,
+		.funcs = &kv_dpm_ip_funcs,
+	},
+	{
+		.type = AMDGPU_IP_BLOCK_TYPE_DCE,
+		.major = 8,
+		.minor = 3,
+		.rev = 0,
+		.funcs = &dce_v8_0_ip_funcs,
+	},
+	{
+		.type = AMDGPU_IP_BLOCK_TYPE_GFX,
+		.major = 7,
+		.minor = 2,
+		.rev = 0,
+		.funcs = &gfx_v7_0_ip_funcs,
+	},
+	{
+		.type = AMDGPU_IP_BLOCK_TYPE_SDMA,
+		.major = 2,
+		.minor = 0,
+		.rev = 0,
+		.funcs = &cik_sdma_ip_funcs,
+	},
+	{
+		.type = AMDGPU_IP_BLOCK_TYPE_UVD,
+		.major = 4,
+		.minor = 2,
+		.rev = 0,
+		.funcs = &uvd_v4_2_ip_funcs,
+	},
+	{
+		.type = AMDGPU_IP_BLOCK_TYPE_VCE,
+		.major = 2,
+		.minor = 0,
+		.rev = 0,
+		.funcs = &vce_v2_0_ip_funcs,
+	},
+};
+
+static const struct amdgpu_ip_block_version mullins_ip_blocks[] =
+{
+	/* ORDER MATTERS! */
+	{
+		.type = AMDGPU_IP_BLOCK_TYPE_COMMON,
+		.major = 1,
+		.minor = 0,
+		.rev = 0,
+		.funcs = &cik_common_ip_funcs,
+	},
+	{
+		.type = AMDGPU_IP_BLOCK_TYPE_GMC,
+		.major = 7,
+		.minor = 0,
+		.rev = 0,
+		.funcs = &gmc_v7_0_ip_funcs,
+	},
+	{
+		.type = AMDGPU_IP_BLOCK_TYPE_IH,
+		.major = 2,
+		.minor = 0,
+		.rev = 0,
+		.funcs = &cik_ih_ip_funcs,
+	},
+	{
+		.type = AMDGPU_IP_BLOCK_TYPE_SMC,
+		.major = 7,
+		.minor = 0,
+		.rev = 0,
+		.funcs = &kv_dpm_ip_funcs,
+	},
+	{
+		.type = AMDGPU_IP_BLOCK_TYPE_DCE,
+		.major = 8,
+		.minor = 3,
+		.rev = 0,
+		.funcs = &dce_v8_0_ip_funcs,
+	},
+	{
+		.type = AMDGPU_IP_BLOCK_TYPE_GFX,
+		.major = 7,
+		.minor = 2,
+		.rev = 0,
+		.funcs = &gfx_v7_0_ip_funcs,
+	},
+	{
+		.type = AMDGPU_IP_BLOCK_TYPE_SDMA,
+		.major = 2,
+		.minor = 0,
+		.rev = 0,
+		.funcs = &cik_sdma_ip_funcs,
+	},
+	{
+		.type = AMDGPU_IP_BLOCK_TYPE_UVD,
+		.major = 4,
+		.minor = 2,
+		.rev = 0,
+		.funcs = &uvd_v4_2_ip_funcs,
+	},
+	{
+		.type = AMDGPU_IP_BLOCK_TYPE_VCE,
+		.major = 2,
+		.minor = 0,
+		.rev = 0,
+		.funcs = &vce_v2_0_ip_funcs,
+	},
+};
+
+static const struct amdgpu_ip_block_version kaveri_ip_blocks[] =
+{
+	/* ORDER MATTERS! */
+	{
+		.type = AMDGPU_IP_BLOCK_TYPE_COMMON,
+		.major = 1,
+		.minor = 0,
+		.rev = 0,
+		.funcs = &cik_common_ip_funcs,
+	},
+	{
+		.type = AMDGPU_IP_BLOCK_TYPE_GMC,
+		.major = 7,
+		.minor = 0,
+		.rev = 0,
+		.funcs = &gmc_v7_0_ip_funcs,
+	},
+	{
+		.type = AMDGPU_IP_BLOCK_TYPE_IH,
+		.major = 2,
+		.minor = 0,
+		.rev = 0,
+		.funcs = &cik_ih_ip_funcs,
+	},
+	{
+		.type = AMDGPU_IP_BLOCK_TYPE_SMC,
+		.major = 7,
+		.minor = 0,
+		.rev = 0,
+		.funcs = &kv_dpm_ip_funcs,
+	},
+	{
+		.type = AMDGPU_IP_BLOCK_TYPE_DCE,
+		.major = 8,
+		.minor = 1,
+		.rev = 0,
+		.funcs = &dce_v8_0_ip_funcs,
+	},
+	{
+		.type = AMDGPU_IP_BLOCK_TYPE_GFX,
+		.major = 7,
+		.minor = 1,
+		.rev = 0,
+		.funcs = &gfx_v7_0_ip_funcs,
+	},
+	{
+		.type = AMDGPU_IP_BLOCK_TYPE_SDMA,
+		.major = 2,
+		.minor = 0,
+		.rev = 0,
+		.funcs = &cik_sdma_ip_funcs,
+	},
+	{
+		.type = AMDGPU_IP_BLOCK_TYPE_UVD,
+		.major = 4,
+		.minor = 2,
+		.rev = 0,
+		.funcs = &uvd_v4_2_ip_funcs,
+	},
+	{
+		.type = AMDGPU_IP_BLOCK_TYPE_VCE,
+		.major = 2,
+		.minor = 0,
+		.rev = 0,
+		.funcs = &vce_v2_0_ip_funcs,
+	},
+};
+
+int cik_set_ip_blocks(struct amdgpu_device *adev)
+{
+	switch (adev->asic_type) {
+	case CHIP_BONAIRE:
+		adev->ip_blocks = bonaire_ip_blocks;
+		adev->num_ip_blocks = ARRAY_SIZE(bonaire_ip_blocks);
+		break;
+	case CHIP_HAWAII:
+		adev->ip_blocks = hawaii_ip_blocks;
+		adev->num_ip_blocks = ARRAY_SIZE(hawaii_ip_blocks);
+		break;
+	case CHIP_KAVERI:
+		adev->ip_blocks = kaveri_ip_blocks;
+		adev->num_ip_blocks = ARRAY_SIZE(kaveri_ip_blocks);
+		break;
+	case CHIP_KABINI:
+		adev->ip_blocks = kabini_ip_blocks;
+		adev->num_ip_blocks = ARRAY_SIZE(kabini_ip_blocks);
+		break;
+	case CHIP_MULLINS:
+		adev->ip_blocks = mullins_ip_blocks;
+		adev->num_ip_blocks = ARRAY_SIZE(mullins_ip_blocks);
+		break;
+	default:
+		/* FIXME: not supported yet */
+		return -EINVAL;
+	}
+
+	adev->ip_block_enabled = kcalloc(adev->num_ip_blocks, sizeof(bool), GFP_KERNEL);
+	if (adev->ip_block_enabled == NULL)
+		return -ENOMEM;
+
+	return 0;
+}
+
+static const struct amdgpu_asic_funcs cik_asic_funcs =
+{
+	.read_disabled_bios = &cik_read_disabled_bios,
+	.read_register = &cik_read_register,
+	.reset = &cik_asic_reset,
+	.set_vga_state = &cik_vga_set_state,
+	.get_xclk = &cik_get_xclk,
+	.set_uvd_clocks = &cik_set_uvd_clocks,
+	.set_vce_clocks = &cik_set_vce_clocks,
+	.get_cu_info = &gfx_v7_0_get_cu_info,
+	/* these should be moved to their own ip modules */
+	.get_gpu_clock_counter = &gfx_v7_0_get_gpu_clock_counter,
+	.wait_for_mc_idle = &gmc_v7_0_mc_wait_for_idle,
+};
+
+static int cik_common_early_init(struct amdgpu_device *adev)
+{
+	adev->smc_rreg = &cik_smc_rreg;
+	adev->smc_wreg = &cik_smc_wreg;
+	adev->pcie_rreg = &cik_pcie_rreg;
+	adev->pcie_wreg = &cik_pcie_wreg;
+	adev->uvd_ctx_rreg = &cik_uvd_ctx_rreg;
+	adev->uvd_ctx_wreg = &cik_uvd_ctx_wreg;
+	adev->didt_rreg = &cik_didt_rreg;
+	adev->didt_wreg = &cik_didt_wreg;
+
+	adev->asic_funcs = &cik_asic_funcs;
+
+	adev->has_uvd = true;
+
+	adev->rev_id = cik_get_rev_id(adev);
+	adev->external_rev_id = 0xFF;
+	switch (adev->asic_type) {
+	case CHIP_BONAIRE:
+		adev->cg_flags =
+			AMDGPU_CG_SUPPORT_GFX_MGCG |
+			AMDGPU_CG_SUPPORT_GFX_MGLS |
+			/*AMDGPU_CG_SUPPORT_GFX_CGCG |*/
+			AMDGPU_CG_SUPPORT_GFX_CGLS |
+			AMDGPU_CG_SUPPORT_GFX_CGTS |
+			AMDGPU_CG_SUPPORT_GFX_CGTS_LS |
+			AMDGPU_CG_SUPPORT_GFX_CP_LS |
+			AMDGPU_CG_SUPPORT_MC_LS |
+			AMDGPU_CG_SUPPORT_MC_MGCG |
+			AMDGPU_CG_SUPPORT_SDMA_MGCG |
+			AMDGPU_CG_SUPPORT_SDMA_LS |
+			AMDGPU_CG_SUPPORT_BIF_LS |
+			AMDGPU_CG_SUPPORT_VCE_MGCG |
+			AMDGPU_CG_SUPPORT_UVD_MGCG |
+			AMDGPU_CG_SUPPORT_HDP_LS |
+			AMDGPU_CG_SUPPORT_HDP_MGCG;
+		adev->pg_flags = 0;
+		adev->external_rev_id = adev->rev_id + 0x14;
+		break;
+	case CHIP_HAWAII:
+		adev->cg_flags =
+			AMDGPU_CG_SUPPORT_GFX_MGCG |
+			AMDGPU_CG_SUPPORT_GFX_MGLS |
+			/*AMDGPU_CG_SUPPORT_GFX_CGCG |*/
+			AMDGPU_CG_SUPPORT_GFX_CGLS |
+			AMDGPU_CG_SUPPORT_GFX_CGTS |
+			AMDGPU_CG_SUPPORT_GFX_CP_LS |
+			AMDGPU_CG_SUPPORT_MC_LS |
+			AMDGPU_CG_SUPPORT_MC_MGCG |
+			AMDGPU_CG_SUPPORT_SDMA_MGCG |
+			AMDGPU_CG_SUPPORT_SDMA_LS |
+			AMDGPU_CG_SUPPORT_BIF_LS |
+			AMDGPU_CG_SUPPORT_VCE_MGCG |
+			AMDGPU_CG_SUPPORT_UVD_MGCG |
+			AMDGPU_CG_SUPPORT_HDP_LS |
+			AMDGPU_CG_SUPPORT_HDP_MGCG;
+		adev->pg_flags = 0;
+		adev->external_rev_id = 0x28;
+		break;
+	case CHIP_KAVERI:
+		adev->cg_flags =
+			AMDGPU_CG_SUPPORT_GFX_MGCG |
+			AMDGPU_CG_SUPPORT_GFX_MGLS |
+			/*AMDGPU_CG_SUPPORT_GFX_CGCG |*/
+			AMDGPU_CG_SUPPORT_GFX_CGLS |
+			AMDGPU_CG_SUPPORT_GFX_CGTS |
+			AMDGPU_CG_SUPPORT_GFX_CGTS_LS |
+			AMDGPU_CG_SUPPORT_GFX_CP_LS |
+			AMDGPU_CG_SUPPORT_SDMA_MGCG |
+			AMDGPU_CG_SUPPORT_SDMA_LS |
+			AMDGPU_CG_SUPPORT_BIF_LS |
+			AMDGPU_CG_SUPPORT_VCE_MGCG |
+			AMDGPU_CG_SUPPORT_UVD_MGCG |
+			AMDGPU_CG_SUPPORT_HDP_LS |
+			AMDGPU_CG_SUPPORT_HDP_MGCG;
+		adev->pg_flags =
+			/*AMDGPU_PG_SUPPORT_GFX_PG |
+			  AMDGPU_PG_SUPPORT_GFX_SMG |
+			  AMDGPU_PG_SUPPORT_GFX_DMG |*/
+			AMDGPU_PG_SUPPORT_UVD |
+			/*AMDGPU_PG_SUPPORT_VCE |
+			  AMDGPU_PG_SUPPORT_CP |
+			  AMDGPU_PG_SUPPORT_GDS |
+			  AMDGPU_PG_SUPPORT_RLC_SMU_HS |
+			  AMDGPU_PG_SUPPORT_ACP |
+			  AMDGPU_PG_SUPPORT_SAMU |*/
+			0;
+		if (adev->pdev->device == 0x1312 ||
+			adev->pdev->device == 0x1316 ||
+			adev->pdev->device == 0x1317)
+			adev->external_rev_id = 0x41;
+		else
+			adev->external_rev_id = 0x1;
+		break;
+	case CHIP_KABINI:
+	case CHIP_MULLINS:
+		adev->cg_flags =
+			AMDGPU_CG_SUPPORT_GFX_MGCG |
+			AMDGPU_CG_SUPPORT_GFX_MGLS |
+			/*AMDGPU_CG_SUPPORT_GFX_CGCG |*/
+			AMDGPU_CG_SUPPORT_GFX_CGLS |
+			AMDGPU_CG_SUPPORT_GFX_CGTS |
+			AMDGPU_CG_SUPPORT_GFX_CGTS_LS |
+			AMDGPU_CG_SUPPORT_GFX_CP_LS |
+			AMDGPU_CG_SUPPORT_SDMA_MGCG |
+			AMDGPU_CG_SUPPORT_SDMA_LS |
+			AMDGPU_CG_SUPPORT_BIF_LS |
+			AMDGPU_CG_SUPPORT_VCE_MGCG |
+			AMDGPU_CG_SUPPORT_UVD_MGCG |
+			AMDGPU_CG_SUPPORT_HDP_LS |
+			AMDGPU_CG_SUPPORT_HDP_MGCG;
+		adev->pg_flags =
+			/*AMDGPU_PG_SUPPORT_GFX_PG |
+			  AMDGPU_PG_SUPPORT_GFX_SMG | */
+			AMDGPU_PG_SUPPORT_UVD |
+			/*AMDGPU_PG_SUPPORT_VCE |
+			  AMDGPU_PG_SUPPORT_CP |
+			  AMDGPU_PG_SUPPORT_GDS |
+			  AMDGPU_PG_SUPPORT_RLC_SMU_HS |
+			  AMDGPU_PG_SUPPORT_SAMU |*/
+			0;
+		if (adev->asic_type == CHIP_KABINI) {
+			if (adev->rev_id == 0)
+				adev->external_rev_id = 0x81;
+			else if (adev->rev_id == 1)
+				adev->external_rev_id = 0x82;
+			else if (adev->rev_id == 2)
+				adev->external_rev_id = 0x85;
+		} else
+			adev->external_rev_id = adev->rev_id + 0xa1;
+		break;
+	default:
+		/* FIXME: not supported yet */
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+static int cik_common_sw_init(struct amdgpu_device *adev)
+{
+	return 0;
+}
+
+static int cik_common_sw_fini(struct amdgpu_device *adev)
+{
+	return 0;
+}
+
+static int cik_common_hw_init(struct amdgpu_device *adev)
+{
+	/* move the golden regs per IP block */
+	cik_init_golden_registers(adev);
+	/* enable pcie gen2/3 link */
+	cik_pcie_gen3_enable(adev);
+	/* enable aspm */
+	cik_program_aspm(adev);
+
+	return 0;
+}
+
+static int cik_common_hw_fini(struct amdgpu_device *adev)
+{
+	return 0;
+}
+
+static int cik_common_suspend(struct amdgpu_device *adev)
+{
+	return cik_common_hw_fini(adev);
+}
+
+static int cik_common_resume(struct amdgpu_device *adev)
+{
+	return cik_common_hw_init(adev);
+}
+
+static bool cik_common_is_idle(struct amdgpu_device *adev)
+{
+	return true;
+}
+
+static int cik_common_wait_for_idle(struct amdgpu_device *adev)
+{
+	return 0;
+}
+
+static void cik_common_print_status(struct amdgpu_device *adev)
+{
+
+}
+
+static int cik_common_soft_reset(struct amdgpu_device *adev)
+{
+	/* XXX hard reset?? */
+	return 0;
+}
+
+static int cik_common_set_clockgating_state(struct amdgpu_device *adev,
+					    enum amdgpu_clockgating_state state)
+{
+	return 0;
+}
+
+static int cik_common_set_powergating_state(struct amdgpu_device *adev,
+					    enum amdgpu_powergating_state state)
+{
+	return 0;
+}
+
+const struct amdgpu_ip_funcs cik_common_ip_funcs = {
+	.early_init = cik_common_early_init,
+	.late_init = NULL,
+	.sw_init = cik_common_sw_init,
+	.sw_fini = cik_common_sw_fini,
+	.hw_init = cik_common_hw_init,
+	.hw_fini = cik_common_hw_fini,
+	.suspend = cik_common_suspend,
+	.resume = cik_common_resume,
+	.is_idle = cik_common_is_idle,
+	.wait_for_idle = cik_common_wait_for_idle,
+	.soft_reset = cik_common_soft_reset,
+	.print_status = cik_common_print_status,
+	.set_clockgating_state = cik_common_set_clockgating_state,
+	.set_powergating_state = cik_common_set_powergating_state,
+};

drivers/gpu/drm/amd/amdgpu/cik.h

+/*
+ * Copyright 2014 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __CIK_H__
+#define __CIK_H__
+
+extern const struct amdgpu_ip_funcs cik_common_ip_funcs;
+
+void cik_srbm_select(struct amdgpu_device *adev,
+		     u32 me, u32 pipe, u32 queue, u32 vmid);
+int cik_set_ip_blocks(struct amdgpu_device *adev);
+
+#endif

drivers/gpu/drm/amd/amdgpu/cik_dpm.h

+/*
+ * Copyright 2014 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __CIK_DPM_H__
+#define __CIK_DPM_H__
+
+extern const struct amdgpu_ip_funcs ci_dpm_ip_funcs;
+extern const struct amdgpu_ip_funcs kv_dpm_ip_funcs;
+
+#endif

drivers/gpu/drm/amd/amdgpu/cik_ih.c

+/*
+ * Copyright 2012 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+#include "drmP.h"
+#include "amdgpu.h"
+#include "amdgpu_ih.h"
+#include "cikd.h"
+
+#include "bif/bif_4_1_d.h"
+#include "bif/bif_4_1_sh_mask.h"
+
+#include "oss/oss_2_0_d.h"
+#include "oss/oss_2_0_sh_mask.h"
+
+/*
+ * Interrupts
+ * Starting with r6xx, interrupts are handled via a ring buffer.
+ * Ring buffers are areas of GPU accessible memory that the GPU
+ * writes interrupt vectors into and the host reads vectors out of.
+ * There is a rptr (read pointer) that determines where the
+ * host is currently reading, and a wptr (write pointer)
+ * which determines where the GPU has written.  When the
+ * pointers are equal, the ring is idle.  When the GPU
+ * writes vectors to the ring buffer, it increments the
+ * wptr.  When there is an interrupt, the host then starts
+ * fetching commands and processing them until the pointers are
+ * equal again at which point it updates the rptr.
+ */
+
+static void cik_ih_set_interrupt_funcs(struct amdgpu_device *adev);
+
+/**
+ * cik_ih_enable_interrupts - Enable the interrupt ring buffer
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Enable the interrupt ring buffer (CIK).
+ */
+static void cik_ih_enable_interrupts(struct amdgpu_device *adev)
+{
+	u32 ih_cntl = RREG32(mmIH_CNTL);
+	u32 ih_rb_cntl = RREG32(mmIH_RB_CNTL);
+
+	ih_cntl |= IH_CNTL__ENABLE_INTR_MASK;
+	ih_rb_cntl |= IH_RB_CNTL__RB_ENABLE_MASK;
+	WREG32(mmIH_CNTL, ih_cntl);
+	WREG32(mmIH_RB_CNTL, ih_rb_cntl);
+	adev->irq.ih.enabled = true;
+}
+
+/**
+ * cik_ih_disable_interrupts - Disable the interrupt ring buffer
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Disable the interrupt ring buffer (CIK).
+ */
+static void cik_ih_disable_interrupts(struct amdgpu_device *adev)
+{
+	u32 ih_rb_cntl = RREG32(mmIH_RB_CNTL);
+	u32 ih_cntl = RREG32(mmIH_CNTL);
+
+	ih_rb_cntl &= ~IH_RB_CNTL__RB_ENABLE_MASK;
+	ih_cntl &= ~IH_CNTL__ENABLE_INTR_MASK;
+	WREG32(mmIH_RB_CNTL, ih_rb_cntl);
+	WREG32(mmIH_CNTL, ih_cntl);
+	/* set rptr, wptr to 0 */
+	WREG32(mmIH_RB_RPTR, 0);
+	WREG32(mmIH_RB_WPTR, 0);
+	adev->irq.ih.enabled = false;
+	adev->irq.ih.rptr = 0;
+}
+
+/**
+ * cik_ih_irq_init - init and enable the interrupt ring
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Allocate a ring buffer for the interrupt controller,
+ * enable the RLC, disable interrupts, enable the IH
+ * ring buffer and enable it (CIK).
+ * Called at device load and reume.
+ * Returns 0 for success, errors for failure.
+ */
+static int cik_ih_irq_init(struct amdgpu_device *adev)
+{
+	int ret = 0;
+	int rb_bufsz;
+	u32 interrupt_cntl, ih_cntl, ih_rb_cntl;
+	u64 wptr_off;
+
+	/* disable irqs */
+	cik_ih_disable_interrupts(adev);
+
+	/* setup interrupt control */
+	WREG32(mmINTERRUPT_CNTL2, adev->dummy_page.addr >> 8);
+	interrupt_cntl = RREG32(mmINTERRUPT_CNTL);
+	/* INTERRUPT_CNTL__IH_DUMMY_RD_OVERRIDE_MASK=0 - dummy read disabled with msi, enabled without msi
+	 * INTERRUPT_CNTL__IH_DUMMY_RD_OVERRIDE_MASK=1 - dummy read controlled by IH_DUMMY_RD_EN
+	 */
+	interrupt_cntl &= ~INTERRUPT_CNTL__IH_DUMMY_RD_OVERRIDE_MASK;
+	/* INTERRUPT_CNTL__IH_REQ_NONSNOOP_EN_MASK=1 if ring is in non-cacheable memory, e.g., vram */
+	interrupt_cntl &= ~INTERRUPT_CNTL__IH_REQ_NONSNOOP_EN_MASK;
+	WREG32(mmINTERRUPT_CNTL, interrupt_cntl);
+
+	WREG32(mmIH_RB_BASE, adev->irq.ih.gpu_addr >> 8);
+	rb_bufsz = order_base_2(adev->irq.ih.ring_size / 4);
+
+	ih_rb_cntl = (IH_RB_CNTL__WPTR_OVERFLOW_ENABLE_MASK |
+		      IH_RB_CNTL__WPTR_OVERFLOW_CLEAR_MASK |
+		      (rb_bufsz << 1));
+
+	ih_rb_cntl |= IH_RB_CNTL__WPTR_WRITEBACK_ENABLE_MASK;
+
+	/* set the writeback address whether it's enabled or not */
+	wptr_off = adev->wb.gpu_addr + (adev->irq.ih.wptr_offs * 4);
+	WREG32(mmIH_RB_WPTR_ADDR_LO, lower_32_bits(wptr_off));
+	WREG32(mmIH_RB_WPTR_ADDR_HI, upper_32_bits(wptr_off) & 0xFF);
+
+	WREG32(mmIH_RB_CNTL, ih_rb_cntl);
+
+	/* set rptr, wptr to 0 */
+	WREG32(mmIH_RB_RPTR, 0);
+	WREG32(mmIH_RB_WPTR, 0);
+
+	/* Default settings for IH_CNTL (disabled at first) */
+	ih_cntl = (0x10 << IH_CNTL__MC_WRREQ_CREDIT__SHIFT) |
+		(0x10 << IH_CNTL__MC_WR_CLEAN_CNT__SHIFT) |
+		(0 << IH_CNTL__MC_VMID__SHIFT);
+	/* IH_CNTL__RPTR_REARM_MASK only works if msi's are enabled */
+	if (adev->irq.msi_enabled)
+		ih_cntl |= IH_CNTL__RPTR_REARM_MASK;
+	WREG32(mmIH_CNTL, ih_cntl);
+
+	pci_set_master(adev->pdev);
+
+	/* enable irqs */
+	cik_ih_enable_interrupts(adev);
+
+	return ret;
+}
+
+/**
+ * cik_ih_irq_disable - disable interrupts
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Disable interrupts on the hw (CIK).
+ */
+static void cik_ih_irq_disable(struct amdgpu_device *adev)
+{
+	cik_ih_disable_interrupts(adev);
+	/* Wait and acknowledge irq */
+	mdelay(1);
+}
+
+/**
+ * cik_ih_get_wptr - get the IH ring buffer wptr
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Get the IH ring buffer wptr from either the register
+ * or the writeback memory buffer (CIK).  Also check for
+ * ring buffer overflow and deal with it.
+ * Used by cik_irq_process().
+ * Returns the value of the wptr.
+ */
+static u32 cik_ih_get_wptr(struct amdgpu_device *adev)
+{
+	u32 wptr, tmp;
+
+	wptr = le32_to_cpu(adev->wb.wb[adev->irq.ih.wptr_offs]);
+
+	if (wptr & IH_RB_WPTR__RB_OVERFLOW_MASK) {
+		wptr &= ~IH_RB_WPTR__RB_OVERFLOW_MASK;
+		/* When a ring buffer overflow happen start parsing interrupt
+		 * from the last not overwritten vector (wptr + 16). Hopefully
+		 * this should allow us to catchup.
+		 */
+		dev_warn(adev->dev, "IH ring buffer overflow (0x%08X, 0x%08X, 0x%08X)\n",
+			wptr, adev->irq.ih.rptr, (wptr + 16) & adev->irq.ih.ptr_mask);
+		adev->irq.ih.rptr = (wptr + 16) & adev->irq.ih.ptr_mask;
+		tmp = RREG32(mmIH_RB_CNTL);
+		tmp |= IH_RB_CNTL__WPTR_OVERFLOW_CLEAR_MASK;
+		WREG32(mmIH_RB_CNTL, tmp);
+	}
+	return (wptr & adev->irq.ih.ptr_mask);
+}
+
+/*        CIK IV Ring
+ * Each IV ring entry is 128 bits:
+ * [7:0]    - interrupt source id
+ * [31:8]   - reserved
+ * [59:32]  - interrupt source data
+ * [63:60]  - reserved
+ * [71:64]  - RINGID
+ *            CP:
+ *            ME_ID [1:0], PIPE_ID[1:0], QUEUE_ID[2:0]
+ *            QUEUE_ID - for compute, which of the 8 queues owned by the dispatcher
+ *                     - for gfx, hw shader state (0=PS...5=LS, 6=CS)
+ *            ME_ID - 0 = gfx, 1 = first 4 CS pipes, 2 = second 4 CS pipes
+ *            PIPE_ID - ME0 0=3D
+ *                    - ME1&2 compute dispatcher (4 pipes each)
+ *            SDMA:
+ *            INSTANCE_ID [1:0], QUEUE_ID[1:0]
+ *            INSTANCE_ID - 0 = sdma0, 1 = sdma1
+ *            QUEUE_ID - 0 = gfx, 1 = rlc0, 2 = rlc1
+ * [79:72]  - VMID
+ * [95:80]  - PASID
+ * [127:96] - reserved
+ */
+
+ /**
+ * cik_ih_decode_iv - decode an interrupt vector
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Decodes the interrupt vector at the current rptr
+ * position and also advance the position.
+ */
+static void cik_ih_decode_iv(struct amdgpu_device *adev,
+			     struct amdgpu_iv_entry *entry)
+{
+	/* wptr/rptr are in bytes! */
+	u32 ring_index = adev->irq.ih.rptr >> 2;
+	uint32_t dw[4];
+	
+	dw[0] = le32_to_cpu(adev->irq.ih.ring[ring_index + 0]);
+	dw[1] = le32_to_cpu(adev->irq.ih.ring[ring_index + 1]);
+	dw[2] = le32_to_cpu(adev->irq.ih.ring[ring_index + 2]);
+	dw[3] = le32_to_cpu(adev->irq.ih.ring[ring_index + 3]);
+
+	entry->src_id = dw[0] & 0xff;
+	entry->src_data = dw[1] & 0xfffffff;
+	entry->ring_id = dw[2] & 0xff;
+	entry->vm_id = (dw[2] >> 8) & 0xff;
+	entry->pas_id = (dw[2] >> 16) & 0xffff;
+
+	/* wptr/rptr are in bytes! */
+	adev->irq.ih.rptr += 16;
+}
+
+/**
+ * cik_ih_set_rptr - set the IH ring buffer rptr
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Set the IH ring buffer rptr.
+ */
+static void cik_ih_set_rptr(struct amdgpu_device *adev)
+{
+	WREG32(mmIH_RB_RPTR, adev->irq.ih.rptr);
+}
+
+static int cik_ih_early_init(struct amdgpu_device *adev)
+{
+	cik_ih_set_interrupt_funcs(adev);
+
+	return 0;
+}
+
+static int cik_ih_sw_init(struct amdgpu_device *adev)
+{
+	int r;
+
+	r = amdgpu_ih_ring_init(adev, 64 * 1024, false);
+	if (r)
+		return r;
+
+	r = amdgpu_irq_init(adev);
+
+	return r;
+}
+
+static int cik_ih_sw_fini(struct amdgpu_device *adev)
+{
+	amdgpu_irq_fini(adev);
+	amdgpu_ih_ring_fini(adev);
+
+	return 0;
+}
+
+static int cik_ih_hw_init(struct amdgpu_device *adev)
+{
+	int r;
+
+	r = cik_ih_irq_init(adev);
+	if (r)
+		return r;
+
+	return 0;
+}
+
+static int cik_ih_hw_fini(struct amdgpu_device *adev)
+{
+	cik_ih_irq_disable(adev);
+
+	return 0;
+}
+
+static int cik_ih_suspend(struct amdgpu_device *adev)
+{
+	return cik_ih_hw_fini(adev);
+}
+
+static int cik_ih_resume(struct amdgpu_device *adev)
+{
+	return cik_ih_hw_init(adev);
+}
+
+static bool cik_ih_is_idle(struct amdgpu_device *adev)
+{
+	u32 tmp = RREG32(mmSRBM_STATUS);
+
+	if (tmp & SRBM_STATUS__IH_BUSY_MASK)
+		return false;
+
+	return true;
+}
+
+static int cik_ih_wait_for_idle(struct amdgpu_device *adev)
+{
+	unsigned i;
+	u32 tmp;
+
+	for (i = 0; i < adev->usec_timeout; i++) {
+		/* read MC_STATUS */
+		tmp = RREG32(mmSRBM_STATUS) & SRBM_STATUS__IH_BUSY_MASK;
+		if (!tmp)
+			return 0;
+		udelay(1);
+	}
+	return -ETIMEDOUT;
+}
+
+static void cik_ih_print_status(struct amdgpu_device *adev)
+{
+	dev_info(adev->dev, "CIK IH registers\n");
+	dev_info(adev->dev, "  SRBM_STATUS=0x%08X\n",
+		RREG32(mmSRBM_STATUS));
+	dev_info(adev->dev, "  SRBM_STATUS2=0x%08X\n",
+		RREG32(mmSRBM_STATUS2));
+	dev_info(adev->dev, "  INTERRUPT_CNTL=0x%08X\n",
+		 RREG32(mmINTERRUPT_CNTL));
+	dev_info(adev->dev, "  INTERRUPT_CNTL2=0x%08X\n",
+		 RREG32(mmINTERRUPT_CNTL2));
+	dev_info(adev->dev, "  IH_CNTL=0x%08X\n",
+		 RREG32(mmIH_CNTL));
+	dev_info(adev->dev, "  IH_RB_CNTL=0x%08X\n",
+		 RREG32(mmIH_RB_CNTL));
+	dev_info(adev->dev, "  IH_RB_BASE=0x%08X\n",
+		 RREG32(mmIH_RB_BASE));
+	dev_info(adev->dev, "  IH_RB_WPTR_ADDR_LO=0x%08X\n",
+		 RREG32(mmIH_RB_WPTR_ADDR_LO));
+	dev_info(adev->dev, "  IH_RB_WPTR_ADDR_HI=0x%08X\n",
+		 RREG32(mmIH_RB_WPTR_ADDR_HI));
+	dev_info(adev->dev, "  IH_RB_RPTR=0x%08X\n",
+		 RREG32(mmIH_RB_RPTR));
+	dev_info(adev->dev, "  IH_RB_WPTR=0x%08X\n",
+		 RREG32(mmIH_RB_WPTR));
+}
+
+static int cik_ih_soft_reset(struct amdgpu_device *adev)
+{
+	u32 srbm_soft_reset = 0;
+	u32 tmp = RREG32(mmSRBM_STATUS);
+
+	if (tmp & SRBM_STATUS__IH_BUSY_MASK)
+		srbm_soft_reset |= SRBM_SOFT_RESET__SOFT_RESET_IH_MASK;
+
+	if (srbm_soft_reset) {
+		cik_ih_print_status(adev);
+
+		tmp = RREG32(mmSRBM_SOFT_RESET);
+		tmp |= srbm_soft_reset;
+		dev_info(adev->dev, "SRBM_SOFT_RESET=0x%08X\n", tmp);
+		WREG32(mmSRBM_SOFT_RESET, tmp);
+		tmp = RREG32(mmSRBM_SOFT_RESET);
+
+		udelay(50);
+
+		tmp &= ~srbm_soft_reset;
+		WREG32(mmSRBM_SOFT_RESET, tmp);
+		tmp = RREG32(mmSRBM_SOFT_RESET);
+
+		/* Wait a little for things to settle down */
+		udelay(50);
+
+		cik_ih_print_status(adev);
+	}
+
+	return 0;
+}
+
+static int cik_ih_set_clockgating_state(struct amdgpu_device *adev,
+					  enum amdgpu_clockgating_state state)
+{
+	return 0;
+}
+
+static int cik_ih_set_powergating_state(struct amdgpu_device *adev,
+					  enum amdgpu_powergating_state state)
+{
+	return 0;
+}
+
+const struct amdgpu_ip_funcs cik_ih_ip_funcs = {
+	.early_init = cik_ih_early_init,
+	.late_init = NULL,
+	.sw_init = cik_ih_sw_init,
+	.sw_fini = cik_ih_sw_fini,
+	.hw_init = cik_ih_hw_init,
+	.hw_fini = cik_ih_hw_fini,
+	.suspend = cik_ih_suspend,
+	.resume = cik_ih_resume,
+	.is_idle = cik_ih_is_idle,
+	.wait_for_idle = cik_ih_wait_for_idle,
+	.soft_reset = cik_ih_soft_reset,
+	.print_status = cik_ih_print_status,
+	.set_clockgating_state = cik_ih_set_clockgating_state,
+	.set_powergating_state = cik_ih_set_powergating_state,
+};
+
+static const struct amdgpu_ih_funcs cik_ih_funcs = {
+	.get_wptr = cik_ih_get_wptr,
+	.decode_iv = cik_ih_decode_iv,
+	.set_rptr = cik_ih_set_rptr
+};
+
+static void cik_ih_set_interrupt_funcs(struct amdgpu_device *adev)
+{
+	if (adev->irq.ih_funcs == NULL)
+		adev->irq.ih_funcs = &cik_ih_funcs;
+}

drivers/gpu/drm/amd/amdgpu/cik_ih.h

+/*
+ * Copyright 2014 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __CIK_IH_H__
+#define __CIK_IH_H__
+
+extern const struct amdgpu_ip_funcs cik_ih_ip_funcs;
+
+#endif

drivers/gpu/drm/amd/amdgpu/cik_sdma.c

+/*
+ * Copyright 2013 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Alex Deucher
+ */
+#include <linux/firmware.h>
+#include <drm/drmP.h>
+#include "amdgpu.h"
+#include "amdgpu_ucode.h"
+#include "amdgpu_trace.h"
+#include "cikd.h"
+#include "cik.h"
+
+#include "bif/bif_4_1_d.h"
+#include "bif/bif_4_1_sh_mask.h"
+
+#include "gca/gfx_7_2_d.h"
+
+#include "gmc/gmc_7_1_d.h"
+#include "gmc/gmc_7_1_sh_mask.h"
+
+#include "oss/oss_2_0_d.h"
+#include "oss/oss_2_0_sh_mask.h"
+
+static const u32 sdma_offsets[SDMA_MAX_INSTANCE] =
+{
+	SDMA0_REGISTER_OFFSET,
+	SDMA1_REGISTER_OFFSET
+};
+
+static void cik_sdma_set_ring_funcs(struct amdgpu_device *adev);
+static void cik_sdma_set_irq_funcs(struct amdgpu_device *adev);
+static void cik_sdma_set_buffer_funcs(struct amdgpu_device *adev);
+static void cik_sdma_set_vm_pte_funcs(struct amdgpu_device *adev);
+
+MODULE_FIRMWARE("radeon/bonaire_sdma.bin");
+MODULE_FIRMWARE("radeon/bonaire_sdma1.bin");
+MODULE_FIRMWARE("radeon/hawaii_sdma.bin");
+MODULE_FIRMWARE("radeon/hawaii_sdma1.bin");
+MODULE_FIRMWARE("radeon/kaveri_sdma.bin");
+MODULE_FIRMWARE("radeon/kaveri_sdma1.bin");
+MODULE_FIRMWARE("radeon/kabini_sdma.bin");
+MODULE_FIRMWARE("radeon/kabini_sdma1.bin");
+MODULE_FIRMWARE("radeon/mullins_sdma.bin");
+MODULE_FIRMWARE("radeon/mullins_sdma1.bin");
+
+u32 amdgpu_cik_gpu_check_soft_reset(struct amdgpu_device *adev);
+
+/*
+ * sDMA - System DMA
+ * Starting with CIK, the GPU has new asynchronous
+ * DMA engines.  These engines are used for compute
+ * and gfx.  There are two DMA engines (SDMA0, SDMA1)
+ * and each one supports 1 ring buffer used for gfx
+ * and 2 queues used for compute.
+ *
+ * The programming model is very similar to the CP
+ * (ring buffer, IBs, etc.), but sDMA has it's own
+ * packet format that is different from the PM4 format
+ * used by the CP. sDMA supports copying data, writing
+ * embedded data, solid fills, and a number of other
+ * things.  It also has support for tiling/detiling of
+ * buffers.
+ */
+
+/**
+ * cik_sdma_init_microcode - load ucode images from disk
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Use the firmware interface to load the ucode images into
+ * the driver (not loaded into hw).
+ * Returns 0 on success, error on failure.
+ */
+static int cik_sdma_init_microcode(struct amdgpu_device *adev)
+{
+	const char *chip_name;
+	char fw_name[30];
+	int err, i;
+
+	DRM_DEBUG("\n");
+
+	switch (adev->asic_type) {
+	case CHIP_BONAIRE:
+		chip_name = "bonaire";
+		break;
+	case CHIP_HAWAII:
+		chip_name = "hawaii";
+		break;
+	case CHIP_KAVERI:
+		chip_name = "kaveri";
+		break;
+	case CHIP_KABINI:
+		chip_name = "kabini";
+		break;
+	case CHIP_MULLINS:
+		chip_name = "mullins";
+		break;
+	default: BUG();
+	}
+
+	for (i = 0; i < SDMA_MAX_INSTANCE; i++) {
+		if (i == 0)
+			snprintf(fw_name, sizeof(fw_name), "radeon/%s_sdma.bin", chip_name);
+		else
+			snprintf(fw_name, sizeof(fw_name), "radeon/%s_sdma1.bin", chip_name);
+		err = request_firmware(&adev->sdma[i].fw, fw_name, adev->dev);
+		if (err)
+			goto out;
+		err = amdgpu_ucode_validate(adev->sdma[i].fw);
+	}
+out:
+	if (err) {
+		printk(KERN_ERR
+		       "cik_sdma: Failed to load firmware \"%s\"\n",
+		       fw_name);
+		for (i = 0; i < SDMA_MAX_INSTANCE; i++) {
+			release_firmware(adev->sdma[i].fw);
+			adev->sdma[i].fw = NULL;
+		}
+	}
+	return err;
+}
+
+/**
+ * cik_sdma_ring_get_rptr - get the current read pointer
+ *
+ * @ring: amdgpu ring pointer
+ *
+ * Get the current rptr from the hardware (CIK+).
+ */
+static uint32_t cik_sdma_ring_get_rptr(struct amdgpu_ring *ring)
+{
+	u32 rptr;
+
+	rptr = ring->adev->wb.wb[ring->rptr_offs];
+
+	return (rptr & 0x3fffc) >> 2;
+}
+
+/**
+ * cik_sdma_ring_get_wptr - get the current write pointer
+ *
+ * @ring: amdgpu ring pointer
+ *
+ * Get the current wptr from the hardware (CIK+).
+ */
+static uint32_t cik_sdma_ring_get_wptr(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	u32 me = (ring == &adev->sdma[0].ring) ? 0 : 1;
+
+	return (RREG32(mmSDMA0_GFX_RB_WPTR + sdma_offsets[me]) & 0x3fffc) >> 2;
+}
+
+/**
+ * cik_sdma_ring_set_wptr - commit the write pointer
+ *
+ * @ring: amdgpu ring pointer
+ *
+ * Write the wptr back to the hardware (CIK+).
+ */
+static void cik_sdma_ring_set_wptr(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	u32 me = (ring == &adev->sdma[0].ring) ? 0 : 1;
+
+	WREG32(mmSDMA0_GFX_RB_WPTR + sdma_offsets[me], (ring->wptr << 2) & 0x3fffc);
+}
+
+static void cik_sdma_hdp_flush_ring_emit(struct amdgpu_ring *);
+
+/**
+ * cik_sdma_ring_emit_ib - Schedule an IB on the DMA engine
+ *
+ * @ring: amdgpu ring pointer
+ * @ib: IB object to schedule
+ *
+ * Schedule an IB in the DMA ring (CIK).
+ */
+static void cik_sdma_ring_emit_ib(struct amdgpu_ring *ring,
+			   struct amdgpu_ib *ib)
+{
+	u32 extra_bits = (ib->vm ? ib->vm->ids[ring->idx].id : 0) & 0xf;
+	u32 next_rptr = ring->wptr + 5;
+
+	if (ib->flush_hdp_writefifo)
+		next_rptr += 6;
+
+	while ((next_rptr & 7) != 4)
+		next_rptr++;
+
+	next_rptr += 4;
+	amdgpu_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_WRITE, SDMA_WRITE_SUB_OPCODE_LINEAR, 0));
+	amdgpu_ring_write(ring, ring->next_rptr_gpu_addr & 0xfffffffc);
+	amdgpu_ring_write(ring, upper_32_bits(ring->next_rptr_gpu_addr) & 0xffffffff);
+	amdgpu_ring_write(ring, 1); /* number of DWs to follow */
+	amdgpu_ring_write(ring, next_rptr);
+
+	if (ib->flush_hdp_writefifo) {
+		/* flush HDP */
+		cik_sdma_hdp_flush_ring_emit(ring);
+	}
+
+	/* IB packet must end on a 8 DW boundary */
+	while ((ring->wptr & 7) != 4)
+		amdgpu_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_NOP, 0, 0));
+	amdgpu_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_INDIRECT_BUFFER, 0, extra_bits));
+	amdgpu_ring_write(ring, ib->gpu_addr & 0xffffffe0); /* base must be 32 byte aligned */
+	amdgpu_ring_write(ring, upper_32_bits(ib->gpu_addr) & 0xffffffff);
+	amdgpu_ring_write(ring, ib->length_dw);
+
+}
+
+/**
+ * cik_sdma_hdp_flush_ring_emit - emit an hdp flush on the DMA ring
+ *
+ * @ring: amdgpu ring pointer
+ *
+ * Emit an hdp flush packet on the requested DMA ring.
+ */
+static void cik_sdma_hdp_flush_ring_emit(struct amdgpu_ring *ring)
+{
+	u32 extra_bits = (SDMA_POLL_REG_MEM_EXTRA_OP(1) |
+			  SDMA_POLL_REG_MEM_EXTRA_FUNC(3)); /* == */
+	u32 ref_and_mask;
+
+	if (ring == &ring->adev->sdma[0].ring)
+		ref_and_mask = GPU_HDP_FLUSH_DONE__SDMA0_MASK;
+	else
+		ref_and_mask = GPU_HDP_FLUSH_DONE__SDMA1_MASK;
+
+	amdgpu_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_POLL_REG_MEM, 0, extra_bits));
+	amdgpu_ring_write(ring, mmGPU_HDP_FLUSH_DONE << 2);
+	amdgpu_ring_write(ring, mmGPU_HDP_FLUSH_REQ << 2);
+	amdgpu_ring_write(ring, ref_and_mask); /* reference */
+	amdgpu_ring_write(ring, ref_and_mask); /* mask */
+	amdgpu_ring_write(ring, (0xfff << 16) | 10); /* retry count, poll interval */
+}
+
+/**
+ * cik_sdma_ring_emit_fence - emit a fence on the DMA ring
+ *
+ * @ring: amdgpu ring pointer
+ * @fence: amdgpu fence object
+ *
+ * Add a DMA fence packet to the ring to write
+ * the fence seq number and DMA trap packet to generate
+ * an interrupt if needed (CIK).
+ */
+static void cik_sdma_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, u64 seq,
+				     bool write64bit)
+{
+	/* write the fence */
+	amdgpu_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_FENCE, 0, 0));
+	amdgpu_ring_write(ring, lower_32_bits(addr));
+	amdgpu_ring_write(ring, upper_32_bits(addr));
+	amdgpu_ring_write(ring, lower_32_bits(seq));
+
+	/* optionally write high bits as well */
+	if (write64bit) {
+		addr += 4;
+		amdgpu_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_FENCE, 0, 0));
+		amdgpu_ring_write(ring, lower_32_bits(addr));
+		amdgpu_ring_write(ring, upper_32_bits(addr));
+		amdgpu_ring_write(ring, upper_32_bits(seq));
+	}
+
+	/* generate an interrupt */
+	amdgpu_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_TRAP, 0, 0));
+}
+
+/**
+ * cik_sdma_ring_emit_semaphore - emit a semaphore on the dma ring
+ *
+ * @ring: amdgpu_ring structure holding ring information
+ * @semaphore: amdgpu semaphore object
+ * @emit_wait: wait or signal semaphore
+ *
+ * Add a DMA semaphore packet to the ring wait on or signal
+ * other rings (CIK).
+ */
+static bool cik_sdma_ring_emit_semaphore(struct amdgpu_ring *ring,
+					 struct amdgpu_semaphore *semaphore,
+					 bool emit_wait)
+{
+	u64 addr = semaphore->gpu_addr;
+	u32 extra_bits = emit_wait ? 0 : SDMA_SEMAPHORE_EXTRA_S;
+
+	amdgpu_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_SEMAPHORE, 0, extra_bits));
+	amdgpu_ring_write(ring, addr & 0xfffffff8);
+	amdgpu_ring_write(ring, upper_32_bits(addr) & 0xffffffff);
+
+	return true;
+}
+
+/**
+ * cik_sdma_gfx_stop - stop the gfx async dma engines
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Stop the gfx async dma ring buffers (CIK).
+ */
+static void cik_sdma_gfx_stop(struct amdgpu_device *adev)
+{
+	struct amdgpu_ring *sdma0 = &adev->sdma[0].ring;
+	struct amdgpu_ring *sdma1 = &adev->sdma[1].ring;
+	u32 rb_cntl;
+	int i;
+
+	if ((adev->mman.buffer_funcs_ring == sdma0) ||
+	    (adev->mman.buffer_funcs_ring == sdma1))
+		amdgpu_ttm_set_active_vram_size(adev, adev->mc.visible_vram_size);
+
+	for (i = 0; i < SDMA_MAX_INSTANCE; i++) {
+		rb_cntl = RREG32(mmSDMA0_GFX_RB_CNTL + sdma_offsets[i]);
+		rb_cntl &= ~SDMA0_GFX_RB_CNTL__RB_ENABLE_MASK;
+		WREG32(mmSDMA0_GFX_RB_CNTL + sdma_offsets[i], rb_cntl);
+		WREG32(mmSDMA0_GFX_IB_CNTL + sdma_offsets[i], 0);
+	}
+	sdma0->ready = false;
+	sdma1->ready = false;
+}
+
+/**
+ * cik_sdma_rlc_stop - stop the compute async dma engines
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Stop the compute async dma queues (CIK).
+ */
+static void cik_sdma_rlc_stop(struct amdgpu_device *adev)
+{
+	/* XXX todo */
+}
+
+/**
+ * cik_sdma_enable - stop the async dma engines
+ *
+ * @adev: amdgpu_device pointer
+ * @enable: enable/disable the DMA MEs.
+ *
+ * Halt or unhalt the async dma engines (CIK).
+ */
+static void cik_sdma_enable(struct amdgpu_device *adev, bool enable)
+{
+	u32 me_cntl;
+	int i;
+
+	if (enable == false) {
+		cik_sdma_gfx_stop(adev);
+		cik_sdma_rlc_stop(adev);
+	}
+
+	for (i = 0; i < SDMA_MAX_INSTANCE; i++) {
+		me_cntl = RREG32(mmSDMA0_F32_CNTL + sdma_offsets[i]);
+		if (enable)
+			me_cntl &= ~SDMA0_F32_CNTL__HALT_MASK;
+		else
+			me_cntl |= SDMA0_F32_CNTL__HALT_MASK;
+		WREG32(mmSDMA0_F32_CNTL + sdma_offsets[i], me_cntl);
+	}
+}
+
+/**
+ * cik_sdma_gfx_resume - setup and start the async dma engines
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Set up the gfx DMA ring buffers and enable them (CIK).
+ * Returns 0 for success, error for failure.
+ */
+static int cik_sdma_gfx_resume(struct amdgpu_device *adev)
+{
+	struct amdgpu_ring *ring;
+	u32 rb_cntl, ib_cntl;
+	u32 rb_bufsz;
+	u32 wb_offset;
+	int i, j, r;
+
+	for (i = 0; i < SDMA_MAX_INSTANCE; i++) {
+		ring = &adev->sdma[i].ring;
+		wb_offset = (ring->rptr_offs * 4);
+
+		mutex_lock(&adev->srbm_mutex);
+		for (j = 0; j < 16; j++) {
+			cik_srbm_select(adev, 0, 0, 0, j);
+			/* SDMA GFX */
+			WREG32(mmSDMA0_GFX_VIRTUAL_ADDR + sdma_offsets[i], 0);
+			WREG32(mmSDMA0_GFX_APE1_CNTL + sdma_offsets[i], 0);
+			/* XXX SDMA RLC - todo */
+		}
+		cik_srbm_select(adev, 0, 0, 0, 0);
+		mutex_unlock(&adev->srbm_mutex);
+
+		WREG32(mmSDMA0_SEM_INCOMPLETE_TIMER_CNTL + sdma_offsets[i], 0);
+		WREG32(mmSDMA0_SEM_WAIT_FAIL_TIMER_CNTL + sdma_offsets[i], 0);
+
+		/* Set ring buffer size in dwords */
+		rb_bufsz = order_base_2(ring->ring_size / 4);
+		rb_cntl = rb_bufsz << 1;
+#ifdef __BIG_ENDIAN
+		rb_cntl |= SDMA_RB_SWAP_ENABLE | SDMA_RPTR_WRITEBACK_SWAP_ENABLE;
+#endif
+		WREG32(mmSDMA0_GFX_RB_CNTL + sdma_offsets[i], rb_cntl);
+
+		/* Initialize the ring buffer's read and write pointers */
+		WREG32(mmSDMA0_GFX_RB_RPTR + sdma_offsets[i], 0);
+		WREG32(mmSDMA0_GFX_RB_WPTR + sdma_offsets[i], 0);
+
+		/* set the wb address whether it's enabled or not */
+		WREG32(mmSDMA0_GFX_RB_RPTR_ADDR_HI + sdma_offsets[i],
+		       upper_32_bits(adev->wb.gpu_addr + wb_offset) & 0xFFFFFFFF);
+		WREG32(mmSDMA0_GFX_RB_RPTR_ADDR_LO + sdma_offsets[i],
+		       ((adev->wb.gpu_addr + wb_offset) & 0xFFFFFFFC));
+
+		rb_cntl |= SDMA0_GFX_RB_CNTL__RPTR_WRITEBACK_ENABLE_MASK;
+
+		WREG32(mmSDMA0_GFX_RB_BASE + sdma_offsets[i], ring->gpu_addr >> 8);
+		WREG32(mmSDMA0_GFX_RB_BASE_HI + sdma_offsets[i], ring->gpu_addr >> 40);
+
+		ring->wptr = 0;
+		WREG32(mmSDMA0_GFX_RB_WPTR + sdma_offsets[i], ring->wptr << 2);
+
+		/* enable DMA RB */
+		WREG32(mmSDMA0_GFX_RB_CNTL + sdma_offsets[i],
+		       rb_cntl | SDMA0_GFX_RB_CNTL__RB_ENABLE_MASK);
+
+		ib_cntl = SDMA0_GFX_IB_CNTL__IB_ENABLE_MASK;
+#ifdef __BIG_ENDIAN
+		ib_cntl |= SDMA0_GFX_IB_CNTL__IB_SWAP_ENABLE_MASK;
+#endif
+		/* enable DMA IBs */
+		WREG32(mmSDMA0_GFX_IB_CNTL + sdma_offsets[i], ib_cntl);
+
+		ring->ready = true;
+
+		r = amdgpu_ring_test_ring(ring);
+		if (r) {
+			ring->ready = false;
+			return r;
+		}
+
+		if (adev->mman.buffer_funcs_ring == ring)
+			amdgpu_ttm_set_active_vram_size(adev, adev->mc.real_vram_size);
+	}
+
+	return 0;
+}
+
+/**
+ * cik_sdma_rlc_resume - setup and start the async dma engines
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Set up the compute DMA queues and enable them (CIK).
+ * Returns 0 for success, error for failure.
+ */
+static int cik_sdma_rlc_resume(struct amdgpu_device *adev)
+{
+	/* XXX todo */
+	return 0;
+}
+
+/**
+ * cik_sdma_load_microcode - load the sDMA ME ucode
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Loads the sDMA0/1 ucode.
+ * Returns 0 for success, -EINVAL if the ucode is not available.
+ */
+static int cik_sdma_load_microcode(struct amdgpu_device *adev)
+{
+	const struct sdma_firmware_header_v1_0 *hdr;
+	const __le32 *fw_data;
+	u32 fw_size;
+	int i, j;
+
+	if (!adev->sdma[0].fw || !adev->sdma[1].fw)
+		return -EINVAL;
+
+	/* halt the MEs */
+	cik_sdma_enable(adev, false);
+
+	for (i = 0; i < SDMA_MAX_INSTANCE; i++) {
+		hdr = (const struct sdma_firmware_header_v1_0 *)adev->sdma[i].fw->data;
+		amdgpu_ucode_print_sdma_hdr(&hdr->header);
+		fw_size = le32_to_cpu(hdr->header.ucode_size_bytes) / 4;
+		adev->sdma[i].fw_version = le32_to_cpu(hdr->header.ucode_version);
+		fw_data = (const __le32 *)
+			(adev->sdma[i].fw->data + le32_to_cpu(hdr->header.ucode_array_offset_bytes));
+		WREG32(mmSDMA0_UCODE_ADDR + sdma_offsets[i], 0);
+		for (j = 0; j < fw_size; j++)
+			WREG32(mmSDMA0_UCODE_DATA + sdma_offsets[i], le32_to_cpup(fw_data++));
+		WREG32(mmSDMA0_UCODE_ADDR + sdma_offsets[i], adev->sdma[i].fw_version);
+	}
+
+	return 0;
+}
+
+/**
+ * cik_sdma_start - setup and start the async dma engines
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Set up the DMA engines and enable them (CIK).
+ * Returns 0 for success, error for failure.
+ */
+static int cik_sdma_start(struct amdgpu_device *adev)
+{
+	int r;
+
+	r = cik_sdma_load_microcode(adev);
+	if (r)
+		return r;
+
+	/* unhalt the MEs */
+	cik_sdma_enable(adev, true);
+
+	/* start the gfx rings and rlc compute queues */
+	r = cik_sdma_gfx_resume(adev);
+	if (r)
+		return r;
+	r = cik_sdma_rlc_resume(adev);
+	if (r)
+		return r;
+
+	return 0;
+}
+
+/**
+ * cik_sdma_ring_test_ring - simple async dma engine test
+ *
+ * @ring: amdgpu_ring structure holding ring information
+ *
+ * Test the DMA engine by writing using it to write an
+ * value to memory. (CIK).
+ * Returns 0 for success, error for failure.
+ */
+static int cik_sdma_ring_test_ring(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	unsigned i;
+	unsigned index;
+	int r;
+	u32 tmp;
+	u64 gpu_addr;
+
+	r = amdgpu_wb_get(adev, &index);
+	if (r) {
+		dev_err(adev->dev, "(%d) failed to allocate wb slot\n", r);
+		return r;
+	}
+
+	gpu_addr = adev->wb.gpu_addr + (index * 4);
+	tmp = 0xCAFEDEAD;
+	adev->wb.wb[index] = cpu_to_le32(tmp);
+
+	r = amdgpu_ring_lock(ring, 5);
+	if (r) {
+		DRM_ERROR("amdgpu: dma failed to lock ring %d (%d).\n", ring->idx, r);
+		amdgpu_wb_free(adev, index);
+		return r;
+	}
+	amdgpu_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_WRITE, SDMA_WRITE_SUB_OPCODE_LINEAR, 0));
+	amdgpu_ring_write(ring, lower_32_bits(gpu_addr));
+	amdgpu_ring_write(ring, upper_32_bits(gpu_addr));
+	amdgpu_ring_write(ring, 1); /* number of DWs to follow */
+	amdgpu_ring_write(ring, 0xDEADBEEF);
+	amdgpu_ring_unlock_commit(ring);
+
+	for (i = 0; i < adev->usec_timeout; i++) {
+		tmp = le32_to_cpu(adev->wb.wb[index]);
+		if (tmp == 0xDEADBEEF)
+			break;
+		DRM_UDELAY(1);
+	}
+
+	if (i < adev->usec_timeout) {
+		DRM_INFO("ring test on %d succeeded in %d usecs\n", ring->idx, i);
+	} else {
+		DRM_ERROR("amdgpu: ring %d test failed (0x%08X)\n",
+			  ring->idx, tmp);
+		r = -EINVAL;
+	}
+	amdgpu_wb_free(adev, index);
+
+	return r;
+}
+
+/**
+ * cik_sdma_ring_test_ib - test an IB on the DMA engine
+ *
+ * @ring: amdgpu_ring structure holding ring information
+ *
+ * Test a simple IB in the DMA ring (CIK).
+ * Returns 0 on success, error on failure.
+ */
+static int cik_sdma_ring_test_ib(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	struct amdgpu_ib ib;
+	unsigned i;
+	unsigned index;
+	int r;
+	u32 tmp = 0;
+	u64 gpu_addr;
+
+	r = amdgpu_wb_get(adev, &index);
+	if (r) {
+		dev_err(adev->dev, "(%d) failed to allocate wb slot\n", r);
+		return r;
+	}
+
+	gpu_addr = adev->wb.gpu_addr + (index * 4);
+	tmp = 0xCAFEDEAD;
+	adev->wb.wb[index] = cpu_to_le32(tmp);
+
+	r = amdgpu_ib_get(ring, NULL, 256, &ib);
+	if (r) {
+		amdgpu_wb_free(adev, index);
+		DRM_ERROR("amdgpu: failed to get ib (%d).\n", r);
+		return r;
+	}
+
+	ib.ptr[0] = SDMA_PACKET(SDMA_OPCODE_WRITE, SDMA_WRITE_SUB_OPCODE_LINEAR, 0);
+	ib.ptr[1] = lower_32_bits(gpu_addr);
+	ib.ptr[2] = upper_32_bits(gpu_addr);
+	ib.ptr[3] = 1;
+	ib.ptr[4] = 0xDEADBEEF;
+	ib.length_dw = 5;
+
+	r = amdgpu_ib_schedule(adev, 1, &ib, AMDGPU_FENCE_OWNER_UNDEFINED);
+	if (r) {
+		amdgpu_ib_free(adev, &ib);
+		amdgpu_wb_free(adev, index);
+		DRM_ERROR("amdgpu: failed to schedule ib (%d).\n", r);
+		return r;
+	}
+	r = amdgpu_fence_wait(ib.fence, false);
+	if (r) {
+		amdgpu_ib_free(adev, &ib);
+		amdgpu_wb_free(adev, index);
+		DRM_ERROR("amdgpu: fence wait failed (%d).\n", r);
+		return r;
+	}
+	for (i = 0; i < adev->usec_timeout; i++) {
+		tmp = le32_to_cpu(adev->wb.wb[index]);
+		if (tmp == 0xDEADBEEF)
+			break;
+		DRM_UDELAY(1);
+	}
+	if (i < adev->usec_timeout) {
+		DRM_INFO("ib test on ring %d succeeded in %u usecs\n",
+			 ib.fence->ring->idx, i);
+	} else {
+		DRM_ERROR("amdgpu: ib test failed (0x%08X)\n", tmp);
+		r = -EINVAL;
+	}
+	amdgpu_ib_free(adev, &ib);
+	amdgpu_wb_free(adev, index);
+	return r;
+}
+
+/**
+ * cik_sdma_vm_copy_pages - update PTEs by copying them from the GART
+ *
+ * @ib: indirect buffer to fill with commands
+ * @pe: addr of the page entry
+ * @src: src addr to copy from
+ * @count: number of page entries to update
+ *
+ * Update PTEs by copying them from the GART using sDMA (CIK).
+ */
+static void cik_sdma_vm_copy_pte(struct amdgpu_ib *ib,
+				 uint64_t pe, uint64_t src,
+				 unsigned count)
+{
+	while (count) {
+		unsigned bytes = count * 8;
+		if (bytes > 0x1FFFF8)
+			bytes = 0x1FFFF8;
+
+		ib->ptr[ib->length_dw++] = SDMA_PACKET(SDMA_OPCODE_COPY,
+			SDMA_WRITE_SUB_OPCODE_LINEAR, 0);
+		ib->ptr[ib->length_dw++] = bytes;
+		ib->ptr[ib->length_dw++] = 0; /* src/dst endian swap */
+		ib->ptr[ib->length_dw++] = lower_32_bits(src);
+		ib->ptr[ib->length_dw++] = upper_32_bits(src);
+		ib->ptr[ib->length_dw++] = lower_32_bits(pe);
+		ib->ptr[ib->length_dw++] = upper_32_bits(pe);
+
+		pe += bytes;
+		src += bytes;
+		count -= bytes / 8;
+	}
+}
+
+/**
+ * cik_sdma_vm_write_pages - update PTEs by writing them manually
+ *
+ * @ib: indirect buffer to fill with commands
+ * @pe: addr of the page entry
+ * @addr: dst addr to write into pe
+ * @count: number of page entries to update
+ * @incr: increase next addr by incr bytes
+ * @flags: access flags
+ *
+ * Update PTEs by writing them manually using sDMA (CIK).
+ */
+static void cik_sdma_vm_write_pte(struct amdgpu_ib *ib,
+				  uint64_t pe,
+				  uint64_t addr, unsigned count,
+				  uint32_t incr, uint32_t flags)
+{
+	uint64_t value;
+	unsigned ndw;
+
+	while (count) {
+		ndw = count * 2;
+		if (ndw > 0xFFFFE)
+			ndw = 0xFFFFE;
+
+		/* for non-physically contiguous pages (system) */
+		ib->ptr[ib->length_dw++] = SDMA_PACKET(SDMA_OPCODE_WRITE,
+			SDMA_WRITE_SUB_OPCODE_LINEAR, 0);
+		ib->ptr[ib->length_dw++] = pe;
+		ib->ptr[ib->length_dw++] = upper_32_bits(pe);
+		ib->ptr[ib->length_dw++] = ndw;
+		for (; ndw > 0; ndw -= 2, --count, pe += 8) {
+			if (flags & AMDGPU_PTE_SYSTEM) {
+				value = amdgpu_vm_map_gart(ib->ring->adev, addr);
+				value &= 0xFFFFFFFFFFFFF000ULL;
+			} else if (flags & AMDGPU_PTE_VALID) {
+				value = addr;
+			} else {
+				value = 0;
+			}
+			addr += incr;
+			value |= flags;
+			ib->ptr[ib->length_dw++] = value;
+			ib->ptr[ib->length_dw++] = upper_32_bits(value);
+		}
+	}
+}
+
+/**
+ * cik_sdma_vm_set_pages - update the page tables using sDMA
+ *
+ * @ib: indirect buffer to fill with commands
+ * @pe: addr of the page entry
+ * @addr: dst addr to write into pe
+ * @count: number of page entries to update
+ * @incr: increase next addr by incr bytes
+ * @flags: access flags
+ *
+ * Update the page tables using sDMA (CIK).
+ */
+static void cik_sdma_vm_set_pte_pde(struct amdgpu_ib *ib,
+				    uint64_t pe,
+				    uint64_t addr, unsigned count,
+				    uint32_t incr, uint32_t flags)
+{
+	uint64_t value;
+	unsigned ndw;
+
+	while (count) {
+		ndw = count;
+		if (ndw > 0x7FFFF)
+			ndw = 0x7FFFF;
+
+		if (flags & AMDGPU_PTE_VALID)
+			value = addr;
+		else
+			value = 0;
+
+		/* for physically contiguous pages (vram) */
+		ib->ptr[ib->length_dw++] = SDMA_PACKET(SDMA_OPCODE_GENERATE_PTE_PDE, 0, 0);
+		ib->ptr[ib->length_dw++] = pe; /* dst addr */
+		ib->ptr[ib->length_dw++] = upper_32_bits(pe);
+		ib->ptr[ib->length_dw++] = flags; /* mask */
+		ib->ptr[ib->length_dw++] = 0;
+		ib->ptr[ib->length_dw++] = value; /* value */
+		ib->ptr[ib->length_dw++] = upper_32_bits(value);
+		ib->ptr[ib->length_dw++] = incr; /* increment size */
+		ib->ptr[ib->length_dw++] = 0;
+		ib->ptr[ib->length_dw++] = ndw; /* number of entries */
+
+		pe += ndw * 8;
+		addr += ndw * incr;
+		count -= ndw;
+	}
+}
+
+/**
+ * cik_sdma_vm_pad_ib - pad the IB to the required number of dw
+ *
+ * @ib: indirect buffer to fill with padding
+ *
+ */
+static void cik_sdma_vm_pad_ib(struct amdgpu_ib *ib)
+{
+	while (ib->length_dw & 0x7)
+		ib->ptr[ib->length_dw++] = SDMA_PACKET(SDMA_OPCODE_NOP, 0, 0);
+}
+
+/**
+ * cik_sdma_ring_emit_vm_flush - cik vm flush using sDMA
+ *
+ * @ring: amdgpu_ring pointer
+ * @vm: amdgpu_vm pointer
+ *
+ * Update the page table base and flush the VM TLB
+ * using sDMA (CIK).
+ */
+static void cik_sdma_ring_emit_vm_flush(struct amdgpu_ring *ring,
+					unsigned vm_id, uint64_t pd_addr)
+{
+	u32 extra_bits = (SDMA_POLL_REG_MEM_EXTRA_OP(0) |
+			  SDMA_POLL_REG_MEM_EXTRA_FUNC(0)); /* always */
+
+	amdgpu_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_SRBM_WRITE, 0, 0xf000));
+	if (vm_id < 8) {
+		amdgpu_ring_write(ring, (mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR + vm_id));
+	} else {
+		amdgpu_ring_write(ring, (mmVM_CONTEXT8_PAGE_TABLE_BASE_ADDR + vm_id - 8));
+	}
+	amdgpu_ring_write(ring, pd_addr >> 12);
+
+	/* update SH_MEM_* regs */
+	amdgpu_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_SRBM_WRITE, 0, 0xf000));
+	amdgpu_ring_write(ring, mmSRBM_GFX_CNTL);
+	amdgpu_ring_write(ring, VMID(vm_id));
+
+	amdgpu_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_SRBM_WRITE, 0, 0xf000));
+	amdgpu_ring_write(ring, mmSH_MEM_BASES);
+	amdgpu_ring_write(ring, 0);
+
+	amdgpu_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_SRBM_WRITE, 0, 0xf000));
+	amdgpu_ring_write(ring, mmSH_MEM_CONFIG);
+	amdgpu_ring_write(ring, 0);
+
+	amdgpu_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_SRBM_WRITE, 0, 0xf000));
+	amdgpu_ring_write(ring, mmSH_MEM_APE1_BASE);
+	amdgpu_ring_write(ring, 1);
+
+	amdgpu_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_SRBM_WRITE, 0, 0xf000));
+	amdgpu_ring_write(ring, mmSH_MEM_APE1_LIMIT);
+	amdgpu_ring_write(ring, 0);
+
+	amdgpu_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_SRBM_WRITE, 0, 0xf000));
+	amdgpu_ring_write(ring, mmSRBM_GFX_CNTL);
+	amdgpu_ring_write(ring, VMID(0));
+
+	/* flush TLB */
+	amdgpu_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_SRBM_WRITE, 0, 0xf000));
+	amdgpu_ring_write(ring, mmVM_INVALIDATE_REQUEST);
+	amdgpu_ring_write(ring, 1 << vm_id);
+
+	amdgpu_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_POLL_REG_MEM, 0, extra_bits));
+	amdgpu_ring_write(ring, mmVM_INVALIDATE_REQUEST << 2);
+	amdgpu_ring_write(ring, 0);
+	amdgpu_ring_write(ring, 0); /* reference */
+	amdgpu_ring_write(ring, 0); /* mask */
+	amdgpu_ring_write(ring, (0xfff << 16) | 10); /* retry count, poll interval */
+}
+
+static void cik_enable_sdma_mgcg(struct amdgpu_device *adev,
+				 bool enable)
+{
+	u32 orig, data;
+
+	if (enable && (adev->cg_flags & AMDGPU_CG_SUPPORT_SDMA_MGCG)) {
+		WREG32(mmSDMA0_CLK_CTRL + SDMA0_REGISTER_OFFSET, 0x00000100);
+		WREG32(mmSDMA0_CLK_CTRL + SDMA1_REGISTER_OFFSET, 0x00000100);
+	} else {
+		orig = data = RREG32(mmSDMA0_CLK_CTRL + SDMA0_REGISTER_OFFSET);
+		data |= 0xff000000;
+		if (data != orig)
+			WREG32(mmSDMA0_CLK_CTRL + SDMA0_REGISTER_OFFSET, data);
+
+		orig = data = RREG32(mmSDMA0_CLK_CTRL + SDMA1_REGISTER_OFFSET);
+		data |= 0xff000000;
+		if (data != orig)
+			WREG32(mmSDMA0_CLK_CTRL + SDMA1_REGISTER_OFFSET, data);
+	}
+}
+
+static void cik_enable_sdma_mgls(struct amdgpu_device *adev,
+				 bool enable)
+{
+	u32 orig, data;
+
+	if (enable && (adev->cg_flags & AMDGPU_CG_SUPPORT_SDMA_LS)) {
+		orig = data = RREG32(mmSDMA0_POWER_CNTL + SDMA0_REGISTER_OFFSET);
+		data |= 0x100;
+		if (orig != data)
+			WREG32(mmSDMA0_POWER_CNTL + SDMA0_REGISTER_OFFSET, data);
+
+		orig = data = RREG32(mmSDMA0_POWER_CNTL + SDMA1_REGISTER_OFFSET);
+		data |= 0x100;
+		if (orig != data)
+			WREG32(mmSDMA0_POWER_CNTL + SDMA1_REGISTER_OFFSET, data);
+	} else {
+		orig = data = RREG32(mmSDMA0_POWER_CNTL + SDMA0_REGISTER_OFFSET);
+		data &= ~0x100;
+		if (orig != data)
+			WREG32(mmSDMA0_POWER_CNTL + SDMA0_REGISTER_OFFSET, data);
+
+		orig = data = RREG32(mmSDMA0_POWER_CNTL + SDMA1_REGISTER_OFFSET);
+		data &= ~0x100;
+		if (orig != data)
+			WREG32(mmSDMA0_POWER_CNTL + SDMA1_REGISTER_OFFSET, data);
+	}
+}
+
+static int cik_sdma_early_init(struct amdgpu_device *adev)
+{
+	cik_sdma_set_ring_funcs(adev);
+	cik_sdma_set_irq_funcs(adev);
+	cik_sdma_set_buffer_funcs(adev);
+	cik_sdma_set_vm_pte_funcs(adev);
+
+	return 0;
+}
+
+static int cik_sdma_sw_init(struct amdgpu_device *adev)
+{
+	struct amdgpu_ring *ring;
+	int r;
+
+	r = cik_sdma_init_microcode(adev);
+	if (r) {
+		DRM_ERROR("Failed to load sdma firmware!\n");
+		return r;
+	}
+
+	/* SDMA trap event */
+	r = amdgpu_irq_add_id(adev, 224, &adev->sdma_trap_irq);
+	if (r)
+		return r;
+
+	/* SDMA Privileged inst */
+	r = amdgpu_irq_add_id(adev, 241, &adev->sdma_illegal_inst_irq);
+	if (r)
+		return r;
+
+	/* SDMA Privileged inst */
+	r = amdgpu_irq_add_id(adev, 247, &adev->sdma_illegal_inst_irq);
+	if (r)
+		return r;
+
+	ring = &adev->sdma[0].ring;
+	ring->ring_obj = NULL;
+
+	ring = &adev->sdma[1].ring;
+	ring->ring_obj = NULL;
+
+	ring = &adev->sdma[0].ring;
+	sprintf(ring->name, "sdma0");
+	r = amdgpu_ring_init(adev, ring, 256 * 1024,
+			     SDMA_PACKET(SDMA_OPCODE_NOP, 0, 0), 0xf,
+			     &adev->sdma_trap_irq, AMDGPU_SDMA_IRQ_TRAP0,
+			     AMDGPU_RING_TYPE_SDMA);
+	if (r)
+		return r;
+
+	ring = &adev->sdma[1].ring;
+	sprintf(ring->name, "sdma1");
+	r = amdgpu_ring_init(adev, ring, 256 * 1024,
+			     SDMA_PACKET(SDMA_OPCODE_NOP, 0, 0), 0xf,
+			     &adev->sdma_trap_irq, AMDGPU_SDMA_IRQ_TRAP1,
+			     AMDGPU_RING_TYPE_SDMA);
+	if (r)
+		return r;
+
+	return r;
+}
+
+static int cik_sdma_sw_fini(struct amdgpu_device *adev)
+{
+	amdgpu_ring_fini(&adev->sdma[0].ring);
+	amdgpu_ring_fini(&adev->sdma[1].ring);
+
+	return 0;
+}
+
+static int cik_sdma_hw_init(struct amdgpu_device *adev)
+{
+	int r;
+
+	r = cik_sdma_start(adev);
+	if (r)
+		return r;
+
+	return r;
+}
+
+static int cik_sdma_hw_fini(struct amdgpu_device *adev)
+{
+	cik_sdma_enable(adev, false);
+
+	return 0;
+}
+
+static int cik_sdma_suspend(struct amdgpu_device *adev)
+{
+
+	return cik_sdma_hw_fini(adev);
+}
+
+static int cik_sdma_resume(struct amdgpu_device *adev)
+{
+
+	return cik_sdma_hw_init(adev);
+}
+
+static bool cik_sdma_is_idle(struct amdgpu_device *adev)
+{
+	u32 tmp = RREG32(mmSRBM_STATUS2);
+
+	if (tmp & (SRBM_STATUS2__SDMA_BUSY_MASK |
+				SRBM_STATUS2__SDMA1_BUSY_MASK))
+	    return false;
+
+	return true;
+}
+
+static int cik_sdma_wait_for_idle(struct amdgpu_device *adev)
+{
+	unsigned i;
+	u32 tmp;
+
+	for (i = 0; i < adev->usec_timeout; i++) {
+		tmp = RREG32(mmSRBM_STATUS2) & (SRBM_STATUS2__SDMA_BUSY_MASK |
+				SRBM_STATUS2__SDMA1_BUSY_MASK);
+
+		if (!tmp)
+			return 0;
+		udelay(1);
+	}
+	return -ETIMEDOUT;
+}
+
+static void cik_sdma_print_status(struct amdgpu_device *adev)
+{
+	int i, j;
+
+	dev_info(adev->dev, "CIK SDMA registers\n");
+	dev_info(adev->dev, "  SRBM_STATUS2=0x%08X\n",
+		 RREG32(mmSRBM_STATUS2));
+	for (i = 0; i < SDMA_MAX_INSTANCE; i++) {
+		dev_info(adev->dev, "  SDMA%d_STATUS_REG=0x%08X\n",
+			 i, RREG32(mmSDMA0_STATUS_REG + sdma_offsets[i]));
+		dev_info(adev->dev, "  SDMA%d_ME_CNTL=0x%08X\n",
+			 i, RREG32(mmSDMA0_F32_CNTL + sdma_offsets[i]));
+		dev_info(adev->dev, "  SDMA%d_CNTL=0x%08X\n",
+			 i, RREG32(mmSDMA0_CNTL + sdma_offsets[i]));
+		dev_info(adev->dev, "  SDMA%d_SEM_INCOMPLETE_TIMER_CNTL=0x%08X\n",
+			 i, RREG32(mmSDMA0_SEM_INCOMPLETE_TIMER_CNTL + sdma_offsets[i]));
+		dev_info(adev->dev, "  SDMA%d_SEM_WAIT_FAIL_TIMER_CNTL=0x%08X\n",
+			 i, RREG32(mmSDMA0_SEM_WAIT_FAIL_TIMER_CNTL + sdma_offsets[i]));
+		dev_info(adev->dev, "  SDMA%d_GFX_IB_CNTL=0x%08X\n",
+			 i, RREG32(mmSDMA0_GFX_IB_CNTL + sdma_offsets[i]));
+		dev_info(adev->dev, "  SDMA%d_GFX_RB_CNTL=0x%08X\n",
+			 i, RREG32(mmSDMA0_GFX_RB_CNTL + sdma_offsets[i]));
+		dev_info(adev->dev, "  SDMA%d_GFX_RB_RPTR=0x%08X\n",
+			 i, RREG32(mmSDMA0_GFX_RB_RPTR + sdma_offsets[i]));
+		dev_info(adev->dev, "  SDMA%d_GFX_RB_WPTR=0x%08X\n",
+			 i, RREG32(mmSDMA0_GFX_RB_WPTR + sdma_offsets[i]));
+		dev_info(adev->dev, "  SDMA%d_GFX_RB_RPTR_ADDR_HI=0x%08X\n",
+			 i, RREG32(mmSDMA0_GFX_RB_RPTR_ADDR_HI + sdma_offsets[i]));
+		dev_info(adev->dev, "  SDMA%d_GFX_RB_RPTR_ADDR_LO=0x%08X\n",
+			 i, RREG32(mmSDMA0_GFX_RB_RPTR_ADDR_LO + sdma_offsets[i]));
+		dev_info(adev->dev, "  SDMA%d_GFX_RB_BASE=0x%08X\n",
+			 i, RREG32(mmSDMA0_GFX_RB_BASE + sdma_offsets[i]));
+		dev_info(adev->dev, "  SDMA%d_GFX_RB_BASE_HI=0x%08X\n",
+			 i, RREG32(mmSDMA0_GFX_RB_BASE_HI + sdma_offsets[i]));
+		mutex_lock(&adev->srbm_mutex);
+		for (j = 0; j < 16; j++) {
+			cik_srbm_select(adev, 0, 0, 0, j);
+			dev_info(adev->dev, "  VM %d:\n", j);
+			dev_info(adev->dev, "  SDMA0_GFX_VIRTUAL_ADDR=0x%08X\n",
+				 RREG32(mmSDMA0_GFX_VIRTUAL_ADDR + sdma_offsets[i]));
+			dev_info(adev->dev, "  SDMA0_GFX_APE1_CNTL=0x%08X\n",
+				 RREG32(mmSDMA0_GFX_APE1_CNTL + sdma_offsets[i]));
+		}
+		cik_srbm_select(adev, 0, 0, 0, 0);
+		mutex_unlock(&adev->srbm_mutex);
+	}
+}
+
+static int cik_sdma_soft_reset(struct amdgpu_device *adev)
+{
+	u32 srbm_soft_reset = 0;
+	u32 tmp = RREG32(mmSRBM_STATUS2);
+
+	if (tmp & SRBM_STATUS2__SDMA_BUSY_MASK) {
+		/* sdma0 */
+		tmp = RREG32(mmSDMA0_F32_CNTL + SDMA0_REGISTER_OFFSET);
+		tmp |= SDMA0_F32_CNTL__HALT_MASK;
+		WREG32(mmSDMA0_F32_CNTL + SDMA0_REGISTER_OFFSET, tmp);
+		srbm_soft_reset |= SRBM_SOFT_RESET__SOFT_RESET_SDMA_MASK;
+	}
+	if (tmp & SRBM_STATUS2__SDMA1_BUSY_MASK) {
+		/* sdma1 */
+		tmp = RREG32(mmSDMA0_F32_CNTL + SDMA1_REGISTER_OFFSET);
+		tmp |= SDMA0_F32_CNTL__HALT_MASK;
+		WREG32(mmSDMA0_F32_CNTL + SDMA1_REGISTER_OFFSET, tmp);
+		srbm_soft_reset |= SRBM_SOFT_RESET__SOFT_RESET_SDMA1_MASK;
+	}
+
+	if (srbm_soft_reset) {
+		cik_sdma_print_status(adev);
+
+		tmp = RREG32(mmSRBM_SOFT_RESET);
+		tmp |= srbm_soft_reset;
+		dev_info(adev->dev, "SRBM_SOFT_RESET=0x%08X\n", tmp);
+		WREG32(mmSRBM_SOFT_RESET, tmp);
+		tmp = RREG32(mmSRBM_SOFT_RESET);
+
+		udelay(50);
+
+		tmp &= ~srbm_soft_reset;
+		WREG32(mmSRBM_SOFT_RESET, tmp);
+		tmp = RREG32(mmSRBM_SOFT_RESET);
+
+		/* Wait a little for things to settle down */
+		udelay(50);
+
+		cik_sdma_print_status(adev);
+	}
+
+	return 0;
+}
+
+static int cik_sdma_set_trap_irq_state(struct amdgpu_device *adev,
+				       struct amdgpu_irq_src *src,
+				       unsigned type,
+				       enum amdgpu_interrupt_state state)
+{
+	u32 sdma_cntl;
+
+	switch (type) {
+	case AMDGPU_SDMA_IRQ_TRAP0:
+		switch (state) {
+		case AMDGPU_IRQ_STATE_DISABLE:
+			sdma_cntl = RREG32(mmSDMA0_CNTL + SDMA0_REGISTER_OFFSET);
+			sdma_cntl &= ~SDMA0_CNTL__TRAP_ENABLE_MASK;
+			WREG32(mmSDMA0_CNTL + SDMA0_REGISTER_OFFSET, sdma_cntl);
+			break;
+		case AMDGPU_IRQ_STATE_ENABLE:
+			sdma_cntl = RREG32(mmSDMA0_CNTL + SDMA0_REGISTER_OFFSET);
+			sdma_cntl |= SDMA0_CNTL__TRAP_ENABLE_MASK;
+			WREG32(mmSDMA0_CNTL + SDMA0_REGISTER_OFFSET, sdma_cntl);
+			break;
+		default:
+			break;
+		}
+		break;
+	case AMDGPU_SDMA_IRQ_TRAP1:
+		switch (state) {
+		case AMDGPU_IRQ_STATE_DISABLE:
+			sdma_cntl = RREG32(mmSDMA0_CNTL + SDMA1_REGISTER_OFFSET);
+			sdma_cntl &= ~SDMA0_CNTL__TRAP_ENABLE_MASK;
+			WREG32(mmSDMA0_CNTL + SDMA1_REGISTER_OFFSET, sdma_cntl);
+			break;
+		case AMDGPU_IRQ_STATE_ENABLE:
+			sdma_cntl = RREG32(mmSDMA0_CNTL + SDMA1_REGISTER_OFFSET);
+			sdma_cntl |= SDMA0_CNTL__TRAP_ENABLE_MASK;
+			WREG32(mmSDMA0_CNTL + SDMA1_REGISTER_OFFSET, sdma_cntl);
+			break;
+		default:
+			break;
+		}
+		break;
+	default:
+		break;
+	}
+	return 0;
+}
+
+static int cik_sdma_process_trap_irq(struct amdgpu_device *adev,
+				     struct amdgpu_irq_src *source,
+				     struct amdgpu_iv_entry *entry)
+{
+	u8 instance_id, queue_id;
+
+	instance_id = (entry->ring_id & 0x3) >> 0;
+	queue_id = (entry->ring_id & 0xc) >> 2;
+	DRM_DEBUG("IH: SDMA trap\n");
+	switch (instance_id) {
+	case 0:
+		switch (queue_id) {
+		case 0:
+			amdgpu_fence_process(&adev->sdma[0].ring);
+			break;
+		case 1:
+			/* XXX compute */
+			break;
+		case 2:
+			/* XXX compute */
+			break;
+		}
+		break;
+	case 1:
+		switch (queue_id) {
+		case 0:
+			amdgpu_fence_process(&adev->sdma[1].ring);
+			break;
+		case 1:
+			/* XXX compute */
+			break;
+		case 2:
+			/* XXX compute */
+			break;
+		}
+		break;
+	}
+
+	return 0;
+}
+
+static int cik_sdma_process_illegal_inst_irq(struct amdgpu_device *adev,
+					     struct amdgpu_irq_src *source,
+					     struct amdgpu_iv_entry *entry)
+{
+	DRM_ERROR("Illegal instruction in SDMA command stream\n");
+	schedule_work(&adev->reset_work);
+	return 0;
+}
+
+static int cik_sdma_set_clockgating_state(struct amdgpu_device *adev,
+					  enum amdgpu_clockgating_state state)
+{
+	bool gate = false;
+
+	if (state == AMDGPU_CG_STATE_GATE)
+		gate = true;
+
+	cik_enable_sdma_mgcg(adev, gate);
+	cik_enable_sdma_mgls(adev, gate);
+
+	return 0;
+}
+
+static int cik_sdma_set_powergating_state(struct amdgpu_device *adev,
+					  enum amdgpu_powergating_state state)
+{
+	return 0;
+}
+
+const struct amdgpu_ip_funcs cik_sdma_ip_funcs = {
+	.early_init = cik_sdma_early_init,
+	.late_init = NULL,
+	.sw_init = cik_sdma_sw_init,
+	.sw_fini = cik_sdma_sw_fini,
+	.hw_init = cik_sdma_hw_init,
+	.hw_fini = cik_sdma_hw_fini,
+	.suspend = cik_sdma_suspend,
+	.resume = cik_sdma_resume,
+	.is_idle = cik_sdma_is_idle,
+	.wait_for_idle = cik_sdma_wait_for_idle,
+	.soft_reset = cik_sdma_soft_reset,
+	.print_status = cik_sdma_print_status,
+	.set_clockgating_state = cik_sdma_set_clockgating_state,
+	.set_powergating_state = cik_sdma_set_powergating_state,
+};
+
+/**
+ * cik_sdma_ring_is_lockup - Check if the DMA engine is locked up
+ *
+ * @ring: amdgpu_ring structure holding ring information
+ *
+ * Check if the async DMA engine is locked up (CIK).
+ * Returns true if the engine appears to be locked up, false if not.
+ */
+static bool cik_sdma_ring_is_lockup(struct amdgpu_ring *ring)
+{
+
+	if (cik_sdma_is_idle(ring->adev)) {
+		amdgpu_ring_lockup_update(ring);
+		return false;
+	}
+	return amdgpu_ring_test_lockup(ring);
+}
+
+static const struct amdgpu_ring_funcs cik_sdma_ring_funcs = {
+	.get_rptr = cik_sdma_ring_get_rptr,
+	.get_wptr = cik_sdma_ring_get_wptr,
+	.set_wptr = cik_sdma_ring_set_wptr,
+	.parse_cs = NULL,
+	.emit_ib = cik_sdma_ring_emit_ib,
+	.emit_fence = cik_sdma_ring_emit_fence,
+	.emit_semaphore = cik_sdma_ring_emit_semaphore,
+	.emit_vm_flush = cik_sdma_ring_emit_vm_flush,
+	.test_ring = cik_sdma_ring_test_ring,
+	.test_ib = cik_sdma_ring_test_ib,
+	.is_lockup = cik_sdma_ring_is_lockup,
+};
+
+static void cik_sdma_set_ring_funcs(struct amdgpu_device *adev)
+{
+	adev->sdma[0].ring.funcs = &cik_sdma_ring_funcs;
+	adev->sdma[1].ring.funcs = &cik_sdma_ring_funcs;
+}
+
+static const struct amdgpu_irq_src_funcs cik_sdma_trap_irq_funcs = {
+	.set = cik_sdma_set_trap_irq_state,
+	.process = cik_sdma_process_trap_irq,
+};
+
+static const struct amdgpu_irq_src_funcs cik_sdma_illegal_inst_irq_funcs = {
+	.process = cik_sdma_process_illegal_inst_irq,
+};
+
+static void cik_sdma_set_irq_funcs(struct amdgpu_device *adev)
+{
+	adev->sdma_trap_irq.num_types = AMDGPU_SDMA_IRQ_LAST;
+	adev->sdma_trap_irq.funcs = &cik_sdma_trap_irq_funcs;
+	adev->sdma_illegal_inst_irq.funcs = &cik_sdma_illegal_inst_irq_funcs;
+}
+
+/**
+ * cik_sdma_emit_copy_buffer - copy buffer using the sDMA engine
+ *
+ * @ring: amdgpu_ring structure holding ring information
+ * @src_offset: src GPU address
+ * @dst_offset: dst GPU address
+ * @byte_count: number of bytes to xfer
+ *
+ * Copy GPU buffers using the DMA engine (CIK).
+ * Used by the amdgpu ttm implementation to move pages if
+ * registered as the asic copy callback.
+ */
+static void cik_sdma_emit_copy_buffer(struct amdgpu_ring *ring,
+				      uint64_t src_offset,
+				      uint64_t dst_offset,
+				      uint32_t byte_count)
+{
+	amdgpu_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_COPY, SDMA_COPY_SUB_OPCODE_LINEAR, 0));
+	amdgpu_ring_write(ring, byte_count);
+	amdgpu_ring_write(ring, 0); /* src/dst endian swap */
+	amdgpu_ring_write(ring, lower_32_bits(src_offset));
+	amdgpu_ring_write(ring, upper_32_bits(src_offset));
+	amdgpu_ring_write(ring, lower_32_bits(dst_offset));
+	amdgpu_ring_write(ring, upper_32_bits(dst_offset));
+}
+
+/**
+ * cik_sdma_emit_fill_buffer - fill buffer using the sDMA engine
+ *
+ * @ring: amdgpu_ring structure holding ring information
+ * @src_data: value to write to buffer
+ * @dst_offset: dst GPU address
+ * @byte_count: number of bytes to xfer
+ *
+ * Fill GPU buffers using the DMA engine (CIK).
+ */
+static void cik_sdma_emit_fill_buffer(struct amdgpu_ring *ring,
+				      uint32_t src_data,
+				      uint64_t dst_offset,
+				      uint32_t byte_count)
+{
+	amdgpu_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_CONSTANT_FILL, 0, 0));
+	amdgpu_ring_write(ring, lower_32_bits(dst_offset));
+	amdgpu_ring_write(ring, upper_32_bits(dst_offset));
+	amdgpu_ring_write(ring, src_data);
+	amdgpu_ring_write(ring, byte_count);
+}
+
+static const struct amdgpu_buffer_funcs cik_sdma_buffer_funcs = {
+	.copy_max_bytes = 0x1fffff,
+	.copy_num_dw = 7,
+	.emit_copy_buffer = cik_sdma_emit_copy_buffer,
+
+	.fill_max_bytes = 0x1fffff,
+	.fill_num_dw = 5,
+	.emit_fill_buffer = cik_sdma_emit_fill_buffer,
+};
+
+static void cik_sdma_set_buffer_funcs(struct amdgpu_device *adev)
+{
+	if (adev->mman.buffer_funcs == NULL) {
+		adev->mman.buffer_funcs = &cik_sdma_buffer_funcs;
+		adev->mman.buffer_funcs_ring = &adev->sdma[0].ring;
+	}
+}
+
+static const struct amdgpu_vm_pte_funcs cik_sdma_vm_pte_funcs = {
+	.copy_pte = cik_sdma_vm_copy_pte,
+	.write_pte = cik_sdma_vm_write_pte,
+	.set_pte_pde = cik_sdma_vm_set_pte_pde,
+	.pad_ib = cik_sdma_vm_pad_ib,
+};
+
+static void cik_sdma_set_vm_pte_funcs(struct amdgpu_device *adev)
+{
+	if (adev->vm_manager.vm_pte_funcs == NULL) {
+		adev->vm_manager.vm_pte_funcs = &cik_sdma_vm_pte_funcs;
+		adev->vm_manager.vm_pte_funcs_ring = &adev->sdma[0].ring;
+	}
+}

drivers/gpu/drm/amd/amdgpu/cik_sdma.h

+/*
+ * Copyright 2014 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __CIK_SDMA_H__
+#define __CIK_SDMA_H__
+
+extern const struct amdgpu_ip_funcs cik_sdma_ip_funcs;
+
+#endif

drivers/gpu/drm/amd/amdgpu/clearstate_ci.h

+/*
+ * Copyright 2013 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+static const unsigned int ci_SECT_CONTEXT_def_1[] =
+{
+    0x00000000, // DB_RENDER_CONTROL
+    0x00000000, // DB_COUNT_CONTROL
+    0x00000000, // DB_DEPTH_VIEW
+    0x00000000, // DB_RENDER_OVERRIDE
+    0x00000000, // DB_RENDER_OVERRIDE2
+    0x00000000, // DB_HTILE_DATA_BASE
+    0, // HOLE
+    0, // HOLE
+    0x00000000, // DB_DEPTH_BOUNDS_MIN
+    0x00000000, // DB_DEPTH_BOUNDS_MAX
+    0x00000000, // DB_STENCIL_CLEAR
+    0x00000000, // DB_DEPTH_CLEAR
+    0x00000000, // PA_SC_SCREEN_SCISSOR_TL
+    0x40004000, // PA_SC_SCREEN_SCISSOR_BR
+    0, // HOLE
+    0x00000000, // DB_DEPTH_INFO
+    0x00000000, // DB_Z_INFO
+    0x00000000, // DB_STENCIL_INFO
+    0x00000000, // DB_Z_READ_BASE
+    0x00000000, // DB_STENCIL_READ_BASE
+    0x00000000, // DB_Z_WRITE_BASE
+    0x00000000, // DB_STENCIL_WRITE_BASE
+    0x00000000, // DB_DEPTH_SIZE
+    0x00000000, // DB_DEPTH_SLICE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0x00000000, // TA_BC_BASE_ADDR
+    0x00000000, // TA_BC_BASE_ADDR_HI
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0x00000000, // COHER_DEST_BASE_HI_0
+    0x00000000, // COHER_DEST_BASE_HI_1
+    0x00000000, // COHER_DEST_BASE_HI_2
+    0x00000000, // COHER_DEST_BASE_HI_3
+    0x00000000, // COHER_DEST_BASE_2
+    0x00000000, // COHER_DEST_BASE_3
+    0x00000000, // PA_SC_WINDOW_OFFSET
+    0x80000000, // PA_SC_WINDOW_SCISSOR_TL
+    0x40004000, // PA_SC_WINDOW_SCISSOR_BR
+    0x0000ffff, // PA_SC_CLIPRECT_RULE
+    0x00000000, // PA_SC_CLIPRECT_0_TL
+    0x40004000, // PA_SC_CLIPRECT_0_BR
+    0x00000000, // PA_SC_CLIPRECT_1_TL
+    0x40004000, // PA_SC_CLIPRECT_1_BR
+    0x00000000, // PA_SC_CLIPRECT_2_TL
+    0x40004000, // PA_SC_CLIPRECT_2_BR
+    0x00000000, // PA_SC_CLIPRECT_3_TL
+    0x40004000, // PA_SC_CLIPRECT_3_BR
+    0xaa99aaaa, // PA_SC_EDGERULE
+    0x00000000, // PA_SU_HARDWARE_SCREEN_OFFSET
+    0xffffffff, // CB_TARGET_MASK
+    0xffffffff, // CB_SHADER_MASK
+    0x80000000, // PA_SC_GENERIC_SCISSOR_TL
+    0x40004000, // PA_SC_GENERIC_SCISSOR_BR
+    0x00000000, // COHER_DEST_BASE_0
+    0x00000000, // COHER_DEST_BASE_1
+    0x80000000, // PA_SC_VPORT_SCISSOR_0_TL
+    0x40004000, // PA_SC_VPORT_SCISSOR_0_BR
+    0x80000000, // PA_SC_VPORT_SCISSOR_1_TL
+    0x40004000, // PA_SC_VPORT_SCISSOR_1_BR
+    0x80000000, // PA_SC_VPORT_SCISSOR_2_TL
+    0x40004000, // PA_SC_VPORT_SCISSOR_2_BR
+    0x80000000, // PA_SC_VPORT_SCISSOR_3_TL
+    0x40004000, // PA_SC_VPORT_SCISSOR_3_BR
+    0x80000000, // PA_SC_VPORT_SCISSOR_4_TL
+    0x40004000, // PA_SC_VPORT_SCISSOR_4_BR
+    0x80000000, // PA_SC_VPORT_SCISSOR_5_TL
+    0x40004000, // PA_SC_VPORT_SCISSOR_5_BR
+    0x80000000, // PA_SC_VPORT_SCISSOR_6_TL
+    0x40004000, // PA_SC_VPORT_SCISSOR_6_BR
+    0x80000000, // PA_SC_VPORT_SCISSOR_7_TL
+    0x40004000, // PA_SC_VPORT_SCISSOR_7_BR
+    0x80000000, // PA_SC_VPORT_SCISSOR_8_TL
+    0x40004000, // PA_SC_VPORT_SCISSOR_8_BR
+    0x80000000, // PA_SC_VPORT_SCISSOR_9_TL
+    0x40004000, // PA_SC_VPORT_SCISSOR_9_BR
+    0x80000000, // PA_SC_VPORT_SCISSOR_10_TL
+    0x40004000, // PA_SC_VPORT_SCISSOR_10_BR
+    0x80000000, // PA_SC_VPORT_SCISSOR_11_TL
+    0x40004000, // PA_SC_VPORT_SCISSOR_11_BR
+    0x80000000, // PA_SC_VPORT_SCISSOR_12_TL
+    0x40004000, // PA_SC_VPORT_SCISSOR_12_BR
+    0x80000000, // PA_SC_VPORT_SCISSOR_13_TL
+    0x40004000, // PA_SC_VPORT_SCISSOR_13_BR
+    0x80000000, // PA_SC_VPORT_SCISSOR_14_TL
+    0x40004000, // PA_SC_VPORT_SCISSOR_14_BR
+    0x80000000, // PA_SC_VPORT_SCISSOR_15_TL
+    0x40004000, // PA_SC_VPORT_SCISSOR_15_BR
+    0x00000000, // PA_SC_VPORT_ZMIN_0
+    0x3f800000, // PA_SC_VPORT_ZMAX_0
+    0x00000000, // PA_SC_VPORT_ZMIN_1
+    0x3f800000, // PA_SC_VPORT_ZMAX_1
+    0x00000000, // PA_SC_VPORT_ZMIN_2
+    0x3f800000, // PA_SC_VPORT_ZMAX_2
+    0x00000000, // PA_SC_VPORT_ZMIN_3
+    0x3f800000, // PA_SC_VPORT_ZMAX_3
+    0x00000000, // PA_SC_VPORT_ZMIN_4
+    0x3f800000, // PA_SC_VPORT_ZMAX_4
+    0x00000000, // PA_SC_VPORT_ZMIN_5
+    0x3f800000, // PA_SC_VPORT_ZMAX_5
+    0x00000000, // PA_SC_VPORT_ZMIN_6
+    0x3f800000, // PA_SC_VPORT_ZMAX_6
+    0x00000000, // PA_SC_VPORT_ZMIN_7
+    0x3f800000, // PA_SC_VPORT_ZMAX_7
+    0x00000000, // PA_SC_VPORT_ZMIN_8
+    0x3f800000, // PA_SC_VPORT_ZMAX_8
+    0x00000000, // PA_SC_VPORT_ZMIN_9
+    0x3f800000, // PA_SC_VPORT_ZMAX_9
+    0x00000000, // PA_SC_VPORT_ZMIN_10
+    0x3f800000, // PA_SC_VPORT_ZMAX_10
+    0x00000000, // PA_SC_VPORT_ZMIN_11
+    0x3f800000, // PA_SC_VPORT_ZMAX_11
+    0x00000000, // PA_SC_VPORT_ZMIN_12
+    0x3f800000, // PA_SC_VPORT_ZMAX_12
+    0x00000000, // PA_SC_VPORT_ZMIN_13
+    0x3f800000, // PA_SC_VPORT_ZMAX_13
+    0x00000000, // PA_SC_VPORT_ZMIN_14
+    0x3f800000, // PA_SC_VPORT_ZMAX_14
+    0x00000000, // PA_SC_VPORT_ZMIN_15
+    0x3f800000, // PA_SC_VPORT_ZMAX_15
+};
+static const unsigned int ci_SECT_CONTEXT_def_2[] =
+{
+    0x00000000, // PA_SC_SCREEN_EXTENT_CONTROL
+    0, // HOLE
+    0x00000000, // CP_PERFMON_CNTX_CNTL
+    0x00000000, // CP_RINGID
+    0x00000000, // CP_VMID
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0xffffffff, // VGT_MAX_VTX_INDX
+    0x00000000, // VGT_MIN_VTX_INDX
+    0x00000000, // VGT_INDX_OFFSET
+    0x00000000, // VGT_MULTI_PRIM_IB_RESET_INDX
+    0, // HOLE
+    0x00000000, // CB_BLEND_RED
+    0x00000000, // CB_BLEND_GREEN
+    0x00000000, // CB_BLEND_BLUE
+    0x00000000, // CB_BLEND_ALPHA
+    0, // HOLE
+    0, // HOLE
+    0x00000000, // DB_STENCIL_CONTROL
+    0x00000000, // DB_STENCILREFMASK
+    0x00000000, // DB_STENCILREFMASK_BF
+    0, // HOLE
+    0x00000000, // PA_CL_VPORT_XSCALE
+    0x00000000, // PA_CL_VPORT_XOFFSET
+    0x00000000, // PA_CL_VPORT_YSCALE
+    0x00000000, // PA_CL_VPORT_YOFFSET
+    0x00000000, // PA_CL_VPORT_ZSCALE
+    0x00000000, // PA_CL_VPORT_ZOFFSET
+    0x00000000, // PA_CL_VPORT_XSCALE_1
+    0x00000000, // PA_CL_VPORT_XOFFSET_1
+    0x00000000, // PA_CL_VPORT_YSCALE_1
+    0x00000000, // PA_CL_VPORT_YOFFSET_1
+    0x00000000, // PA_CL_VPORT_ZSCALE_1
+    0x00000000, // PA_CL_VPORT_ZOFFSET_1
+    0x00000000, // PA_CL_VPORT_XSCALE_2
+    0x00000000, // PA_CL_VPORT_XOFFSET_2
+    0x00000000, // PA_CL_VPORT_YSCALE_2
+    0x00000000, // PA_CL_VPORT_YOFFSET_2
+    0x00000000, // PA_CL_VPORT_ZSCALE_2
+    0x00000000, // PA_CL_VPORT_ZOFFSET_2
+    0x00000000, // PA_CL_VPORT_XSCALE_3
+    0x00000000, // PA_CL_VPORT_XOFFSET_3
+    0x00000000, // PA_CL_VPORT_YSCALE_3
+    0x00000000, // PA_CL_VPORT_YOFFSET_3
+    0x00000000, // PA_CL_VPORT_ZSCALE_3
+    0x00000000, // PA_CL_VPORT_ZOFFSET_3
+    0x00000000, // PA_CL_VPORT_XSCALE_4
+    0x00000000, // PA_CL_VPORT_XOFFSET_4
+    0x00000000, // PA_CL_VPORT_YSCALE_4
+    0x00000000, // PA_CL_VPORT_YOFFSET_4
+    0x00000000, // PA_CL_VPORT_ZSCALE_4
+    0x00000000, // PA_CL_VPORT_ZOFFSET_4
+    0x00000000, // PA_CL_VPORT_XSCALE_5
+    0x00000000, // PA_CL_VPORT_XOFFSET_5
+    0x00000000, // PA_CL_VPORT_YSCALE_5
+    0x00000000, // PA_CL_VPORT_YOFFSET_5
+    0x00000000, // PA_CL_VPORT_ZSCALE_5
+    0x00000000, // PA_CL_VPORT_ZOFFSET_5
+    0x00000000, // PA_CL_VPORT_XSCALE_6
+    0x00000000, // PA_CL_VPORT_XOFFSET_6
+    0x00000000, // PA_CL_VPORT_YSCALE_6
+    0x00000000, // PA_CL_VPORT_YOFFSET_6
+    0x00000000, // PA_CL_VPORT_ZSCALE_6
+    0x00000000, // PA_CL_VPORT_ZOFFSET_6
+    0x00000000, // PA_CL_VPORT_XSCALE_7
+    0x00000000, // PA_CL_VPORT_XOFFSET_7
+    0x00000000, // PA_CL_VPORT_YSCALE_7
+    0x00000000, // PA_CL_VPORT_YOFFSET_7
+    0x00000000, // PA_CL_VPORT_ZSCALE_7
+    0x00000000, // PA_CL_VPORT_ZOFFSET_7
+    0x00000000, // PA_CL_VPORT_XSCALE_8
+    0x00000000, // PA_CL_VPORT_XOFFSET_8
+    0x00000000, // PA_CL_VPORT_YSCALE_8
+    0x00000000, // PA_CL_VPORT_YOFFSET_8
+    0x00000000, // PA_CL_VPORT_ZSCALE_8
+    0x00000000, // PA_CL_VPORT_ZOFFSET_8
+    0x00000000, // PA_CL_VPORT_XSCALE_9
+    0x00000000, // PA_CL_VPORT_XOFFSET_9
+    0x00000000, // PA_CL_VPORT_YSCALE_9
+    0x00000000, // PA_CL_VPORT_YOFFSET_9
+    0x00000000, // PA_CL_VPORT_ZSCALE_9
+    0x00000000, // PA_CL_VPORT_ZOFFSET_9
+    0x00000000, // PA_CL_VPORT_XSCALE_10
+    0x00000000, // PA_CL_VPORT_XOFFSET_10
+    0x00000000, // PA_CL_VPORT_YSCALE_10
+    0x00000000, // PA_CL_VPORT_YOFFSET_10
+    0x00000000, // PA_CL_VPORT_ZSCALE_10
+    0x00000000, // PA_CL_VPORT_ZOFFSET_10
+    0x00000000, // PA_CL_VPORT_XSCALE_11
+    0x00000000, // PA_CL_VPORT_XOFFSET_11
+    0x00000000, // PA_CL_VPORT_YSCALE_11
+    0x00000000, // PA_CL_VPORT_YOFFSET_11
+    0x00000000, // PA_CL_VPORT_ZSCALE_11
+    0x00000000, // PA_CL_VPORT_ZOFFSET_11
+    0x00000000, // PA_CL_VPORT_XSCALE_12
+    0x00000000, // PA_CL_VPORT_XOFFSET_12
+    0x00000000, // PA_CL_VPORT_YSCALE_12
+    0x00000000, // PA_CL_VPORT_YOFFSET_12
+    0x00000000, // PA_CL_VPORT_ZSCALE_12
+    0x00000000, // PA_CL_VPORT_ZOFFSET_12
+    0x00000000, // PA_CL_VPORT_XSCALE_13
+    0x00000000, // PA_CL_VPORT_XOFFSET_13
+    0x00000000, // PA_CL_VPORT_YSCALE_13
+    0x00000000, // PA_CL_VPORT_YOFFSET_13
+    0x00000000, // PA_CL_VPORT_ZSCALE_13
+    0x00000000, // PA_CL_VPORT_ZOFFSET_13
+    0x00000000, // PA_CL_VPORT_XSCALE_14
+    0x00000000, // PA_CL_VPORT_XOFFSET_14
+    0x00000000, // PA_CL_VPORT_YSCALE_14
+    0x00000000, // PA_CL_VPORT_YOFFSET_14
+    0x00000000, // PA_CL_VPORT_ZSCALE_14
+    0x00000000, // PA_CL_VPORT_ZOFFSET_14
+    0x00000000, // PA_CL_VPORT_XSCALE_15
+    0x00000000, // PA_CL_VPORT_XOFFSET_15
+    0x00000000, // PA_CL_VPORT_YSCALE_15
+    0x00000000, // PA_CL_VPORT_YOFFSET_15
+    0x00000000, // PA_CL_VPORT_ZSCALE_15
+    0x00000000, // PA_CL_VPORT_ZOFFSET_15
+    0x00000000, // PA_CL_UCP_0_X
+    0x00000000, // PA_CL_UCP_0_Y
+    0x00000000, // PA_CL_UCP_0_Z
+    0x00000000, // PA_CL_UCP_0_W
+    0x00000000, // PA_CL_UCP_1_X
+    0x00000000, // PA_CL_UCP_1_Y
+    0x00000000, // PA_CL_UCP_1_Z
+    0x00000000, // PA_CL_UCP_1_W
+    0x00000000, // PA_CL_UCP_2_X
+    0x00000000, // PA_CL_UCP_2_Y
+    0x00000000, // PA_CL_UCP_2_Z
+    0x00000000, // PA_CL_UCP_2_W
+    0x00000000, // PA_CL_UCP_3_X
+    0x00000000, // PA_CL_UCP_3_Y
+    0x00000000, // PA_CL_UCP_3_Z
+    0x00000000, // PA_CL_UCP_3_W
+    0x00000000, // PA_CL_UCP_4_X
+    0x00000000, // PA_CL_UCP_4_Y
+    0x00000000, // PA_CL_UCP_4_Z
+    0x00000000, // PA_CL_UCP_4_W
+    0x00000000, // PA_CL_UCP_5_X
+    0x00000000, // PA_CL_UCP_5_Y
+    0x00000000, // PA_CL_UCP_5_Z
+    0x00000000, // PA_CL_UCP_5_W
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0x00000000, // SPI_PS_INPUT_CNTL_0
+    0x00000000, // SPI_PS_INPUT_CNTL_1
+    0x00000000, // SPI_PS_INPUT_CNTL_2
+    0x00000000, // SPI_PS_INPUT_CNTL_3
+    0x00000000, // SPI_PS_INPUT_CNTL_4
+    0x00000000, // SPI_PS_INPUT_CNTL_5
+    0x00000000, // SPI_PS_INPUT_CNTL_6
+    0x00000000, // SPI_PS_INPUT_CNTL_7
+    0x00000000, // SPI_PS_INPUT_CNTL_8
+    0x00000000, // SPI_PS_INPUT_CNTL_9
+    0x00000000, // SPI_PS_INPUT_CNTL_10
+    0x00000000, // SPI_PS_INPUT_CNTL_11
+    0x00000000, // SPI_PS_INPUT_CNTL_12
+    0x00000000, // SPI_PS_INPUT_CNTL_13
+    0x00000000, // SPI_PS_INPUT_CNTL_14
+    0x00000000, // SPI_PS_INPUT_CNTL_15
+    0x00000000, // SPI_PS_INPUT_CNTL_16
+    0x00000000, // SPI_PS_INPUT_CNTL_17
+    0x00000000, // SPI_PS_INPUT_CNTL_18
+    0x00000000, // SPI_PS_INPUT_CNTL_19
+    0x00000000, // SPI_PS_INPUT_CNTL_20
+    0x00000000, // SPI_PS_INPUT_CNTL_21
+    0x00000000, // SPI_PS_INPUT_CNTL_22
+    0x00000000, // SPI_PS_INPUT_CNTL_23
+    0x00000000, // SPI_PS_INPUT_CNTL_24
+    0x00000000, // SPI_PS_INPUT_CNTL_25
+    0x00000000, // SPI_PS_INPUT_CNTL_26
+    0x00000000, // SPI_PS_INPUT_CNTL_27
+    0x00000000, // SPI_PS_INPUT_CNTL_28
+    0x00000000, // SPI_PS_INPUT_CNTL_29
+    0x00000000, // SPI_PS_INPUT_CNTL_30
+    0x00000000, // SPI_PS_INPUT_CNTL_31
+    0x00000000, // SPI_VS_OUT_CONFIG
+    0, // HOLE
+    0x00000000, // SPI_PS_INPUT_ENA
+    0x00000000, // SPI_PS_INPUT_ADDR
+    0x00000000, // SPI_INTERP_CONTROL_0
+    0x00000002, // SPI_PS_IN_CONTROL
+    0, // HOLE
+    0x00000000, // SPI_BARYC_CNTL
+    0, // HOLE
+    0x00000000, // SPI_TMPRING_SIZE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0x00000000, // SPI_SHADER_POS_FORMAT
+    0x00000000, // SPI_SHADER_Z_FORMAT
+    0x00000000, // SPI_SHADER_COL_FORMAT
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0x00000000, // CB_BLEND0_CONTROL
+    0x00000000, // CB_BLEND1_CONTROL
+    0x00000000, // CB_BLEND2_CONTROL
+    0x00000000, // CB_BLEND3_CONTROL
+    0x00000000, // CB_BLEND4_CONTROL
+    0x00000000, // CB_BLEND5_CONTROL
+    0x00000000, // CB_BLEND6_CONTROL
+    0x00000000, // CB_BLEND7_CONTROL
+};
+static const unsigned int ci_SECT_CONTEXT_def_3[] =
+{
+    0x00000000, // PA_CL_POINT_X_RAD
+    0x00000000, // PA_CL_POINT_Y_RAD
+    0x00000000, // PA_CL_POINT_SIZE
+    0x00000000, // PA_CL_POINT_CULL_RAD
+    0x00000000, // VGT_DMA_BASE_HI
+    0x00000000, // VGT_DMA_BASE
+};
+static const unsigned int ci_SECT_CONTEXT_def_4[] =
+{
+    0x00000000, // DB_DEPTH_CONTROL
+    0x00000000, // DB_EQAA
+    0x00000000, // CB_COLOR_CONTROL
+    0x00000000, // DB_SHADER_CONTROL
+    0x00090000, // PA_CL_CLIP_CNTL
+    0x00000004, // PA_SU_SC_MODE_CNTL
+    0x00000000, // PA_CL_VTE_CNTL
+    0x00000000, // PA_CL_VS_OUT_CNTL
+    0x00000000, // PA_CL_NANINF_CNTL
+    0x00000000, // PA_SU_LINE_STIPPLE_CNTL
+    0x00000000, // PA_SU_LINE_STIPPLE_SCALE
+    0x00000000, // PA_SU_PRIM_FILTER_CNTL
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0x00000000, // PA_SU_POINT_SIZE
+    0x00000000, // PA_SU_POINT_MINMAX
+    0x00000000, // PA_SU_LINE_CNTL
+    0x00000000, // PA_SC_LINE_STIPPLE
+    0x00000000, // VGT_OUTPUT_PATH_CNTL
+    0x00000000, // VGT_HOS_CNTL
+    0x00000000, // VGT_HOS_MAX_TESS_LEVEL
+    0x00000000, // VGT_HOS_MIN_TESS_LEVEL
+    0x00000000, // VGT_HOS_REUSE_DEPTH
+    0x00000000, // VGT_GROUP_PRIM_TYPE
+    0x00000000, // VGT_GROUP_FIRST_DECR
+    0x00000000, // VGT_GROUP_DECR
+    0x00000000, // VGT_GROUP_VECT_0_CNTL
+    0x00000000, // VGT_GROUP_VECT_1_CNTL
+    0x00000000, // VGT_GROUP_VECT_0_FMT_CNTL
+    0x00000000, // VGT_GROUP_VECT_1_FMT_CNTL
+    0x00000000, // VGT_GS_MODE
+    0x00000000, // VGT_GS_ONCHIP_CNTL
+    0x00000000, // PA_SC_MODE_CNTL_0
+    0x00000000, // PA_SC_MODE_CNTL_1
+    0x00000000, // VGT_ENHANCE
+    0x00000100, // VGT_GS_PER_ES
+    0x00000080, // VGT_ES_PER_GS
+    0x00000002, // VGT_GS_PER_VS
+    0x00000000, // VGT_GSVS_RING_OFFSET_1
+    0x00000000, // VGT_GSVS_RING_OFFSET_2
+    0x00000000, // VGT_GSVS_RING_OFFSET_3
+    0x00000000, // VGT_GS_OUT_PRIM_TYPE
+    0x00000000, // IA_ENHANCE
+};
+static const unsigned int ci_SECT_CONTEXT_def_5[] =
+{
+    0x00000000, // WD_ENHANCE
+    0x00000000, // VGT_PRIMITIVEID_EN
+};
+static const unsigned int ci_SECT_CONTEXT_def_6[] =
+{
+    0x00000000, // VGT_PRIMITIVEID_RESET
+};
+static const unsigned int ci_SECT_CONTEXT_def_7[] =
+{
+    0x00000000, // VGT_MULTI_PRIM_IB_RESET_EN
+    0, // HOLE
+    0, // HOLE
+    0x00000000, // VGT_INSTANCE_STEP_RATE_0
+    0x00000000, // VGT_INSTANCE_STEP_RATE_1
+    0x000000ff, // IA_MULTI_VGT_PARAM
+    0x00000000, // VGT_ESGS_RING_ITEMSIZE
+    0x00000000, // VGT_GSVS_RING_ITEMSIZE
+    0x00000000, // VGT_REUSE_OFF
+    0x00000000, // VGT_VTX_CNT_EN
+    0x00000000, // DB_HTILE_SURFACE
+    0x00000000, // DB_SRESULTS_COMPARE_STATE0
+    0x00000000, // DB_SRESULTS_COMPARE_STATE1
+    0x00000000, // DB_PRELOAD_CONTROL
+    0, // HOLE
+    0x00000000, // VGT_STRMOUT_BUFFER_SIZE_0
+    0x00000000, // VGT_STRMOUT_VTX_STRIDE_0
+    0, // HOLE
+    0x00000000, // VGT_STRMOUT_BUFFER_OFFSET_0
+    0x00000000, // VGT_STRMOUT_BUFFER_SIZE_1
+    0x00000000, // VGT_STRMOUT_VTX_STRIDE_1
+    0, // HOLE
+    0x00000000, // VGT_STRMOUT_BUFFER_OFFSET_1
+    0x00000000, // VGT_STRMOUT_BUFFER_SIZE_2
+    0x00000000, // VGT_STRMOUT_VTX_STRIDE_2
+    0, // HOLE
+    0x00000000, // VGT_STRMOUT_BUFFER_OFFSET_2
+    0x00000000, // VGT_STRMOUT_BUFFER_SIZE_3
+    0x00000000, // VGT_STRMOUT_VTX_STRIDE_3
+    0, // HOLE
+    0x00000000, // VGT_STRMOUT_BUFFER_OFFSET_3
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0x00000000, // VGT_STRMOUT_DRAW_OPAQUE_OFFSET
+    0x00000000, // VGT_STRMOUT_DRAW_OPAQUE_BUFFER_FILLED_SIZE
+    0x00000000, // VGT_STRMOUT_DRAW_OPAQUE_VERTEX_STRIDE
+    0, // HOLE
+    0x00000000, // VGT_GS_MAX_VERT_OUT
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0x00000000, // VGT_SHADER_STAGES_EN
+    0x00000000, // VGT_LS_HS_CONFIG
+    0x00000000, // VGT_GS_VERT_ITEMSIZE
+    0x00000000, // VGT_GS_VERT_ITEMSIZE_1
+    0x00000000, // VGT_GS_VERT_ITEMSIZE_2
+    0x00000000, // VGT_GS_VERT_ITEMSIZE_3
+    0x00000000, // VGT_TF_PARAM
+    0x00000000, // DB_ALPHA_TO_MASK
+    0, // HOLE
+    0x00000000, // PA_SU_POLY_OFFSET_DB_FMT_CNTL
+    0x00000000, // PA_SU_POLY_OFFSET_CLAMP
+    0x00000000, // PA_SU_POLY_OFFSET_FRONT_SCALE
+    0x00000000, // PA_SU_POLY_OFFSET_FRONT_OFFSET
+    0x00000000, // PA_SU_POLY_OFFSET_BACK_SCALE
+    0x00000000, // PA_SU_POLY_OFFSET_BACK_OFFSET
+    0x00000000, // VGT_GS_INSTANCE_CNT
+    0x00000000, // VGT_STRMOUT_CONFIG
+    0x00000000, // VGT_STRMOUT_BUFFER_CONFIG
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0x00000000, // PA_SC_CENTROID_PRIORITY_0
+    0x00000000, // PA_SC_CENTROID_PRIORITY_1
+    0x00001000, // PA_SC_LINE_CNTL
+    0x00000000, // PA_SC_AA_CONFIG
+    0x00000005, // PA_SU_VTX_CNTL
+    0x3f800000, // PA_CL_GB_VERT_CLIP_ADJ
+    0x3f800000, // PA_CL_GB_VERT_DISC_ADJ
+    0x3f800000, // PA_CL_GB_HORZ_CLIP_ADJ
+    0x3f800000, // PA_CL_GB_HORZ_DISC_ADJ
+    0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X0Y0_0
+    0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X0Y0_1
+    0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X0Y0_2
+    0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X0Y0_3
+    0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X1Y0_0
+    0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X1Y0_1
+    0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X1Y0_2
+    0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X1Y0_3
+    0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X0Y1_0
+    0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X0Y1_1
+    0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X0Y1_2
+    0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X0Y1_3
+    0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X1Y1_0
+    0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X1Y1_1
+    0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X1Y1_2
+    0x00000000, // PA_SC_AA_SAMPLE_LOCS_PIXEL_X1Y1_3
+    0xffffffff, // PA_SC_AA_MASK_X0Y0_X1Y0
+    0xffffffff, // PA_SC_AA_MASK_X0Y1_X1Y1
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0, // HOLE
+    0x0000000e, // VGT_VERTEX_REUSE_BLOCK_CNTL
+    0x00000010, // VGT_OUT_DEALLOC_CNTL
+    0x00000000, // CB_COLOR0_BASE
+    0x00000000, // CB_COLOR0_PITCH
+    0x00000000, // CB_COLOR0_SLICE
+    0x00000000, // CB_COLOR0_VIEW
+    0x00000000, // CB_COLOR0_INFO
+    0x00000000, // CB_COLOR0_ATTRIB
+    0, // HOLE
+    0x00000000, // CB_COLOR0_CMASK
+    0x00000000, // CB_COLOR0_CMASK_SLICE
+    0x00000000, // CB_COLOR0_FMASK
+    0x00000000, // CB_COLOR0_FMASK_SLICE
+    0x00000000, // CB_COLOR0_CLEAR_WORD0
+    0x00000000, // CB_COLOR0_CLEAR_WORD1
+    0, // HOLE
+    0, // HOLE
+    0x00000000, // CB_COLOR1_BASE
+    0x00000000, // CB_COLOR1_PITCH
+    0x00000000, // CB_COLOR1_SLICE
+    0x00000000, // CB_COLOR1_VIEW
+    0x00000000, // CB_COLOR1_INFO
+    0x00000000, // CB_COLOR1_ATTRIB
+    0, // HOLE
+    0x00000000, // CB_COLOR1_CMASK
+    0x00000000, // CB_COLOR1_CMASK_SLICE
+    0x00000000, // CB_COLOR1_FMASK
+    0x00000000, // CB_COLOR1_FMASK_SLICE
+    0x00000000, // CB_COLOR1_CLEAR_WORD0
+    0x00000000, // CB_COLOR1_CLEAR_WORD1
+    0, // HOLE
+    0, // HOLE
+    0x00000000, // CB_COLOR2_BASE
+    0x00000000, // CB_COLOR2_PITCH
+    0x00000000, // CB_COLOR2_SLICE
+    0x00000000, // CB_COLOR2_VIEW
+    0x00000000, // CB_COLOR2_INFO
+    0x00000000, // CB_COLOR2_ATTRIB
+    0, // HOLE
+    0x00000000, // CB_COLOR2_CMASK
+    0x00000000, // CB_COLOR2_CMASK_SLICE
+    0x00000000, // CB_COLOR2_FMASK
+    0x00000000, // CB_COLOR2_FMASK_SLICE
+    0x00000000, // CB_COLOR2_CLEAR_WORD0
+    0x00000000, // CB_COLOR2_CLEAR_WORD1
+    0, // HOLE
+    0, // HOLE
+    0x00000000, // CB_COLOR3_BASE
+    0x00000000, // CB_COLOR3_PITCH
+    0x00000000, // CB_COLOR3_SLICE
+    0x00000000, // CB_COLOR3_VIEW
+    0x00000000, // CB_COLOR3_INFO
+    0x00000000, // CB_COLOR3_ATTRIB
+    0, // HOLE
+    0x00000000, // CB_COLOR3_CMASK
+    0x00000000, // CB_COLOR3_CMASK_SLICE
+    0x00000000, // CB_COLOR3_FMASK
+    0x00000000, // CB_COLOR3_FMASK_SLICE
+    0x00000000, // CB_COLOR3_CLEAR_WORD0
+    0x00000000, // CB_COLOR3_CLEAR_WORD1
+    0, // HOLE
+    0, // HOLE
+    0x00000000, // CB_COLOR4_BASE
+    0x00000000, // CB_COLOR4_PITCH
+    0x00000000, // CB_COLOR4_SLICE
+    0x00000000, // CB_COLOR4_VIEW
+    0x00000000, // CB_COLOR4_INFO
+    0x00000000, // CB_COLOR4_ATTRIB
+    0, // HOLE
+    0x00000000, // CB_COLOR4_CMASK
+    0x00000000, // CB_COLOR4_CMASK_SLICE
+    0x00000000, // CB_COLOR4_FMASK
+    0x00000000, // CB_COLOR4_FMASK_SLICE
+    0x00000000, // CB_COLOR4_CLEAR_WORD0
+    0x00000000, // CB_COLOR4_CLEAR_WORD1
+    0, // HOLE
+    0, // HOLE
+    0x00000000, // CB_COLOR5_BASE
+    0x00000000, // CB_COLOR5_PITCH
+    0x00000000, // CB_COLOR5_SLICE
+    0x00000000, // CB_COLOR5_VIEW
+    0x00000000, // CB_COLOR5_INFO
+    0x00000000, // CB_COLOR5_ATTRIB
+    0, // HOLE
+    0x00000000, // CB_COLOR5_CMASK
+    0x00000000, // CB_COLOR5_CMASK_SLICE
+    0x00000000, // CB_COLOR5_FMASK
+    0x00000000, // CB_COLOR5_FMASK_SLICE
+    0x00000000, // CB_COLOR5_CLEAR_WORD0
+    0x00000000, // CB_COLOR5_CLEAR_WORD1
+    0, // HOLE
+    0, // HOLE
+    0x00000000, // CB_COLOR6_BASE
+    0x00000000, // CB_COLOR6_PITCH
+    0x00000000, // CB_COLOR6_SLICE
+    0x00000000, // CB_COLOR6_VIEW
+    0x00000000, // CB_COLOR6_INFO
+    0x00000000, // CB_COLOR6_ATTRIB
+    0, // HOLE
+    0x00000000, // CB_COLOR6_CMASK
+    0x00000000, // CB_COLOR6_CMASK_SLICE
+    0x00000000, // CB_COLOR6_FMASK
+    0x00000000, // CB_COLOR6_FMASK_SLICE
+    0x00000000, // CB_COLOR6_CLEAR_WORD0
+    0x00000000, // CB_COLOR6_CLEAR_WORD1
+    0, // HOLE
+    0, // HOLE
+    0x00000000, // CB_COLOR7_BASE
+    0x00000000, // CB_COLOR7_PITCH
+    0x00000000, // CB_COLOR7_SLICE
+    0x00000000, // CB_COLOR7_VIEW
+    0x00000000, // CB_COLOR7_INFO
+    0x00000000, // CB_COLOR7_ATTRIB
+    0, // HOLE
+    0x00000000, // CB_COLOR7_CMASK
+    0x00000000, // CB_COLOR7_CMASK_SLICE
+    0x00000000, // CB_COLOR7_FMASK
+    0x00000000, // CB_COLOR7_FMASK_SLICE
+    0x00000000, // CB_COLOR7_CLEAR_WORD0
+    0x00000000, // CB_COLOR7_CLEAR_WORD1
+};
+static const struct cs_extent_def ci_SECT_CONTEXT_defs[] =
+{
+    {ci_SECT_CONTEXT_def_1, 0x0000a000, 212 },
+    {ci_SECT_CONTEXT_def_2, 0x0000a0d6, 274 },
+    {ci_SECT_CONTEXT_def_3, 0x0000a1f5, 6 },
+    {ci_SECT_CONTEXT_def_4, 0x0000a200, 157 },
+    {ci_SECT_CONTEXT_def_5, 0x0000a2a0, 2 },
+    {ci_SECT_CONTEXT_def_6, 0x0000a2a3, 1 },
+    {ci_SECT_CONTEXT_def_7, 0x0000a2a5, 233 },
+    { 0, 0, 0 }
+};
+static const struct cs_section_def ci_cs_data[] = {
+    { ci_SECT_CONTEXT_defs, SECT_CONTEXT },
+    { 0, SECT_NONE }
+};

drivers/gpu/drm/amd/amdgpu/dce_v8_0.c

+/*
+ * Copyright 2014 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+#include "drmP.h"
+#include "amdgpu.h"
+#include "amdgpu_pm.h"
+#include "amdgpu_i2c.h"
+#include "cikd.h"
+#include "atom.h"
+#include "amdgpu_atombios.h"
+#include "atombios_crtc.h"
+#include "atombios_encoders.h"
+#include "amdgpu_pll.h"
+#include "amdgpu_connectors.h"
+
+#include "dce/dce_8_0_d.h"
+#include "dce/dce_8_0_sh_mask.h"
+
+#include "gca/gfx_7_2_enum.h"
+
+#include "gmc/gmc_7_1_d.h"
+#include "gmc/gmc_7_1_sh_mask.h"
+
+#include "oss/oss_2_0_d.h"
+#include "oss/oss_2_0_sh_mask.h"
+
+static void dce_v8_0_set_display_funcs(struct amdgpu_device *adev);
+static void dce_v8_0_set_irq_funcs(struct amdgpu_device *adev);
+
+static const u32 crtc_offsets[6] =
+{
+	CRTC0_REGISTER_OFFSET,
+	CRTC1_REGISTER_OFFSET,
+	CRTC2_REGISTER_OFFSET,
+	CRTC3_REGISTER_OFFSET,
+	CRTC4_REGISTER_OFFSET,
+	CRTC5_REGISTER_OFFSET
+};
+
+static const uint32_t dig_offsets[] = {
+	CRTC0_REGISTER_OFFSET,
+	CRTC1_REGISTER_OFFSET,
+	CRTC2_REGISTER_OFFSET,
+	CRTC3_REGISTER_OFFSET,
+	CRTC4_REGISTER_OFFSET,
+	CRTC5_REGISTER_OFFSET,
+	(0x13830 - 0x7030) >> 2,
+};
+
+static const struct {
+	uint32_t	reg;
+	uint32_t	vblank;
+	uint32_t	vline;
+	uint32_t	hpd;
+
+} interrupt_status_offsets[6] = { {
+	.reg = mmDISP_INTERRUPT_STATUS,
+	.vblank = DISP_INTERRUPT_STATUS__LB_D1_VBLANK_INTERRUPT_MASK,
+	.vline = DISP_INTERRUPT_STATUS__LB_D1_VLINE_INTERRUPT_MASK,
+	.hpd = DISP_INTERRUPT_STATUS__DC_HPD1_INTERRUPT_MASK
+}, {
+	.reg = mmDISP_INTERRUPT_STATUS_CONTINUE,
+	.vblank = DISP_INTERRUPT_STATUS_CONTINUE__LB_D2_VBLANK_INTERRUPT_MASK,
+	.vline = DISP_INTERRUPT_STATUS_CONTINUE__LB_D2_VLINE_INTERRUPT_MASK,
+	.hpd = DISP_INTERRUPT_STATUS_CONTINUE__DC_HPD2_INTERRUPT_MASK
+}, {
+	.reg = mmDISP_INTERRUPT_STATUS_CONTINUE2,
+	.vblank = DISP_INTERRUPT_STATUS_CONTINUE2__LB_D3_VBLANK_INTERRUPT_MASK,
+	.vline = DISP_INTERRUPT_STATUS_CONTINUE2__LB_D3_VLINE_INTERRUPT_MASK,
+	.hpd = DISP_INTERRUPT_STATUS_CONTINUE2__DC_HPD3_INTERRUPT_MASK
+}, {
+	.reg = mmDISP_INTERRUPT_STATUS_CONTINUE3,
+	.vblank = DISP_INTERRUPT_STATUS_CONTINUE3__LB_D4_VBLANK_INTERRUPT_MASK,
+	.vline = DISP_INTERRUPT_STATUS_CONTINUE3__LB_D4_VLINE_INTERRUPT_MASK,
+	.hpd = DISP_INTERRUPT_STATUS_CONTINUE3__DC_HPD4_INTERRUPT_MASK
+}, {
+	.reg = mmDISP_INTERRUPT_STATUS_CONTINUE4,
+	.vblank = DISP_INTERRUPT_STATUS_CONTINUE4__LB_D5_VBLANK_INTERRUPT_MASK,
+	.vline = DISP_INTERRUPT_STATUS_CONTINUE4__LB_D5_VLINE_INTERRUPT_MASK,
+	.hpd = DISP_INTERRUPT_STATUS_CONTINUE4__DC_HPD5_INTERRUPT_MASK
+}, {
+	.reg = mmDISP_INTERRUPT_STATUS_CONTINUE5,
+	.vblank = DISP_INTERRUPT_STATUS_CONTINUE5__LB_D6_VBLANK_INTERRUPT_MASK,
+	.vline = DISP_INTERRUPT_STATUS_CONTINUE5__LB_D6_VLINE_INTERRUPT_MASK,
+	.hpd = DISP_INTERRUPT_STATUS_CONTINUE5__DC_HPD6_INTERRUPT_MASK
+} };
+
+static const uint32_t hpd_int_control_offsets[6] = {
+	mmDC_HPD1_INT_CONTROL,
+	mmDC_HPD2_INT_CONTROL,
+	mmDC_HPD3_INT_CONTROL,
+	mmDC_HPD4_INT_CONTROL,
+	mmDC_HPD5_INT_CONTROL,
+	mmDC_HPD6_INT_CONTROL,
+};
+
+static u32 dce_v8_0_audio_endpt_rreg(struct amdgpu_device *adev,
+				     u32 block_offset, u32 reg)
+{
+	unsigned long flags;
+	u32 r;
+
+	spin_lock_irqsave(&adev->audio_endpt_idx_lock, flags);
+	WREG32(mmAZALIA_F0_CODEC_ENDPOINT_INDEX + block_offset, reg);
+	r = RREG32(mmAZALIA_F0_CODEC_ENDPOINT_DATA + block_offset);
+	spin_unlock_irqrestore(&adev->audio_endpt_idx_lock, flags);
+
+	return r;
+}
+
+static void dce_v8_0_audio_endpt_wreg(struct amdgpu_device *adev,
+				      u32 block_offset, u32 reg, u32 v)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&adev->audio_endpt_idx_lock, flags);
+	WREG32(mmAZALIA_F0_CODEC_ENDPOINT_INDEX + block_offset, reg);
+	WREG32(mmAZALIA_F0_CODEC_ENDPOINT_DATA + block_offset, v);
+	spin_unlock_irqrestore(&adev->audio_endpt_idx_lock, flags);
+}
+
+static bool dce_v8_0_is_in_vblank(struct amdgpu_device *adev, int crtc)
+{
+	if (RREG32(mmCRTC_STATUS + crtc_offsets[crtc]) &
+			CRTC_V_BLANK_START_END__CRTC_V_BLANK_START_MASK)
+		return true;
+	else
+		return false;
+}
+
+static bool dce_v8_0_is_counter_moving(struct amdgpu_device *adev, int crtc)
+{
+	u32 pos1, pos2;
+
+	pos1 = RREG32(mmCRTC_STATUS_POSITION + crtc_offsets[crtc]);
+	pos2 = RREG32(mmCRTC_STATUS_POSITION + crtc_offsets[crtc]);
+
+	if (pos1 != pos2)
+		return true;
+	else
+		return false;
+}
+
+/**
+ * dce_v8_0_vblank_wait - vblank wait asic callback.
+ *
+ * @adev: amdgpu_device pointer
+ * @crtc: crtc to wait for vblank on
+ *
+ * Wait for vblank on the requested crtc (evergreen+).
+ */
+static void dce_v8_0_vblank_wait(struct amdgpu_device *adev, int crtc)
+{
+	unsigned i = 0;
+
+	if (crtc >= adev->mode_info.num_crtc)
+		return;
+
+	if (!(RREG32(mmCRTC_CONTROL + crtc_offsets[crtc]) & CRTC_CONTROL__CRTC_MASTER_EN_MASK))
+		return;
+
+	/* depending on when we hit vblank, we may be close to active; if so,
+	 * wait for another frame.
+	 */
+	while (dce_v8_0_is_in_vblank(adev, crtc)) {
+		if (i++ % 100 == 0) {
+			if (!dce_v8_0_is_counter_moving(adev, crtc))
+				break;
+		}
+	}
+
+	while (!dce_v8_0_is_in_vblank(adev, crtc)) {
+		if (i++ % 100 == 0) {
+			if (!dce_v8_0_is_counter_moving(adev, crtc))
+				break;
+		}
+	}
+}
+
+static u32 dce_v8_0_vblank_get_counter(struct amdgpu_device *adev, int crtc)
+{
+	if (crtc >= adev->mode_info.num_crtc)
+		return 0;
+	else
+		return RREG32(mmCRTC_STATUS_FRAME_COUNT + crtc_offsets[crtc]);
+}
+
+/**
+ * dce_v8_0_page_flip - pageflip callback.
+ *
+ * @adev: amdgpu_device pointer
+ * @crtc_id: crtc to cleanup pageflip on
+ * @crtc_base: new address of the crtc (GPU MC address)
+ *
+ * Does the actual pageflip (evergreen+).
+ * During vblank we take the crtc lock and wait for the update_pending
+ * bit to go high, when it does, we release the lock, and allow the
+ * double buffered update to take place.
+ * Returns the current update pending status.
+ */
+static void dce_v8_0_page_flip(struct amdgpu_device *adev,
+			      int crtc_id, u64 crtc_base)
+{
+	struct amdgpu_crtc *amdgpu_crtc = adev->mode_info.crtcs[crtc_id];
+	u32 tmp = RREG32(mmGRPH_UPDATE + amdgpu_crtc->crtc_offset);
+	int i;
+
+	/* Lock the graphics update lock */
+	tmp |= GRPH_UPDATE__GRPH_UPDATE_LOCK_MASK;
+	WREG32(mmGRPH_UPDATE + amdgpu_crtc->crtc_offset, tmp);
+
+	/* update the scanout addresses */
+	WREG32(mmGRPH_SECONDARY_SURFACE_ADDRESS_HIGH + amdgpu_crtc->crtc_offset,
+	       upper_32_bits(crtc_base));
+	WREG32(mmGRPH_SECONDARY_SURFACE_ADDRESS + amdgpu_crtc->crtc_offset,
+	       (u32)crtc_base);
+
+	WREG32(mmGRPH_PRIMARY_SURFACE_ADDRESS_HIGH + amdgpu_crtc->crtc_offset,
+	       upper_32_bits(crtc_base));
+	WREG32(mmGRPH_PRIMARY_SURFACE_ADDRESS + amdgpu_crtc->crtc_offset,
+	       (u32)crtc_base);
+
+	/* Wait for update_pending to go high. */
+	for (i = 0; i < adev->usec_timeout; i++) {
+		if (RREG32(mmGRPH_UPDATE + amdgpu_crtc->crtc_offset) &
+				GRPH_UPDATE__GRPH_SURFACE_UPDATE_PENDING_MASK)
+			break;
+		udelay(1);
+	}
+	DRM_DEBUG("Update pending now high. Unlocking vupdate_lock.\n");
+
+	/* Unlock the lock, so double-buffering can take place inside vblank */
+	tmp &= ~GRPH_UPDATE__GRPH_UPDATE_LOCK_MASK;
+	WREG32(mmGRPH_UPDATE + amdgpu_crtc->crtc_offset, tmp);
+}
+
+static int dce_v8_0_crtc_get_scanoutpos(struct amdgpu_device *adev, int crtc,
+					u32 *vbl, u32 *position)
+{
+	if ((crtc < 0) || (crtc >= adev->mode_info.num_crtc))
+		return -EINVAL;
+
+	*vbl = RREG32(mmCRTC_V_BLANK_START_END + crtc_offsets[crtc]);
+	*position = RREG32(mmCRTC_STATUS_POSITION + crtc_offsets[crtc]);
+
+	return 0;
+}
+
+/**
+ * dce_v8_0_hpd_sense - hpd sense callback.
+ *
+ * @adev: amdgpu_device pointer
+ * @hpd: hpd (hotplug detect) pin
+ *
+ * Checks if a digital monitor is connected (evergreen+).
+ * Returns true if connected, false if not connected.
+ */
+static bool dce_v8_0_hpd_sense(struct amdgpu_device *adev,
+			       enum amdgpu_hpd_id hpd)
+{
+	bool connected = false;
+
+	switch (hpd) {
+	case AMDGPU_HPD_1:
+		if (RREG32(mmDC_HPD1_INT_STATUS) & DC_HPD1_INT_STATUS__DC_HPD1_SENSE_MASK)
+			connected = true;
+		break;
+	case AMDGPU_HPD_2:
+		if (RREG32(mmDC_HPD2_INT_STATUS) & DC_HPD2_INT_STATUS__DC_HPD2_SENSE_MASK)
+			connected = true;
+		break;
+	case AMDGPU_HPD_3:
+		if (RREG32(mmDC_HPD3_INT_STATUS) & DC_HPD3_INT_STATUS__DC_HPD3_SENSE_MASK)
+			connected = true;
+		break;
+	case AMDGPU_HPD_4:
+		if (RREG32(mmDC_HPD4_INT_STATUS) & DC_HPD4_INT_STATUS__DC_HPD4_SENSE_MASK)
+			connected = true;
+		break;
+	case AMDGPU_HPD_5:
+		if (RREG32(mmDC_HPD5_INT_STATUS) & DC_HPD5_INT_STATUS__DC_HPD5_SENSE_MASK)
+			connected = true;
+		break;
+	case AMDGPU_HPD_6:
+		if (RREG32(mmDC_HPD6_INT_STATUS) & DC_HPD6_INT_STATUS__DC_HPD6_SENSE_MASK)
+			connected = true;
+		break;
+	default:
+		break;
+	}
+
+	return connected;
+}
+
+/**
+ * dce_v8_0_hpd_set_polarity - hpd set polarity callback.
+ *
+ * @adev: amdgpu_device pointer
+ * @hpd: hpd (hotplug detect) pin
+ *
+ * Set the polarity of the hpd pin (evergreen+).
+ */
+static void dce_v8_0_hpd_set_polarity(struct amdgpu_device *adev,
+				      enum amdgpu_hpd_id hpd)
+{
+	u32 tmp;
+	bool connected = dce_v8_0_hpd_sense(adev, hpd);
+
+	switch (hpd) {
+	case AMDGPU_HPD_1:
+		tmp = RREG32(mmDC_HPD1_INT_CONTROL);
+		if (connected)
+			tmp &= ~DC_HPD1_INT_CONTROL__DC_HPD1_INT_POLARITY_MASK;
+		else
+			tmp |= DC_HPD1_INT_CONTROL__DC_HPD1_INT_POLARITY_MASK;
+		WREG32(mmDC_HPD1_INT_CONTROL, tmp);
+		break;
+	case AMDGPU_HPD_2:
+		tmp = RREG32(mmDC_HPD2_INT_CONTROL);
+		if (connected)
+			tmp &= ~DC_HPD2_INT_CONTROL__DC_HPD2_INT_POLARITY_MASK;
+		else
+			tmp |= DC_HPD2_INT_CONTROL__DC_HPD2_INT_POLARITY_MASK;
+		WREG32(mmDC_HPD2_INT_CONTROL, tmp);
+		break;
+	case AMDGPU_HPD_3:
+		tmp = RREG32(mmDC_HPD3_INT_CONTROL);
+		if (connected)
+			tmp &= ~DC_HPD3_INT_CONTROL__DC_HPD3_INT_POLARITY_MASK;
+		else
+			tmp |= DC_HPD3_INT_CONTROL__DC_HPD3_INT_POLARITY_MASK;
+		WREG32(mmDC_HPD3_INT_CONTROL, tmp);
+		break;
+	case AMDGPU_HPD_4:
+		tmp = RREG32(mmDC_HPD4_INT_CONTROL);
+		if (connected)
+			tmp &= ~DC_HPD4_INT_CONTROL__DC_HPD4_INT_POLARITY_MASK;
+		else
+			tmp |= DC_HPD4_INT_CONTROL__DC_HPD4_INT_POLARITY_MASK;
+		WREG32(mmDC_HPD4_INT_CONTROL, tmp);
+		break;
+	case AMDGPU_HPD_5:
+		tmp = RREG32(mmDC_HPD5_INT_CONTROL);
+		if (connected)
+			tmp &= ~DC_HPD5_INT_CONTROL__DC_HPD5_INT_POLARITY_MASK;
+		else
+			tmp |= DC_HPD5_INT_CONTROL__DC_HPD5_INT_POLARITY_MASK;
+		WREG32(mmDC_HPD5_INT_CONTROL, tmp);
+			break;
+	case AMDGPU_HPD_6:
+		tmp = RREG32(mmDC_HPD6_INT_CONTROL);
+		if (connected)
+			tmp &= ~DC_HPD6_INT_CONTROL__DC_HPD6_INT_POLARITY_MASK;
+		else
+			tmp |= DC_HPD6_INT_CONTROL__DC_HPD6_INT_POLARITY_MASK;
+		WREG32(mmDC_HPD6_INT_CONTROL, tmp);
+		break;
+	default:
+		break;
+	}
+}
+
+/**
+ * dce_v8_0_hpd_init - hpd setup callback.
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Setup the hpd pins used by the card (evergreen+).
+ * Enable the pin, set the polarity, and enable the hpd interrupts.
+ */
+static void dce_v8_0_hpd_init(struct amdgpu_device *adev)
+{
+	struct drm_device *dev = adev->ddev;
+	struct drm_connector *connector;
+	u32 tmp = (0x9c4 << DC_HPD1_CONTROL__DC_HPD1_CONNECTION_TIMER__SHIFT) |
+		(0xfa << DC_HPD1_CONTROL__DC_HPD1_RX_INT_TIMER__SHIFT) |
+		DC_HPD1_CONTROL__DC_HPD1_EN_MASK;
+
+	list_for_each_entry(connector, &dev->mode_config.connector_list, head) {
+		struct amdgpu_connector *amdgpu_connector = to_amdgpu_connector(connector);
+
+		if (connector->connector_type == DRM_MODE_CONNECTOR_eDP ||
+		    connector->connector_type == DRM_MODE_CONNECTOR_LVDS) {
+			/* don't try to enable hpd on eDP or LVDS avoid breaking the
+			 * aux dp channel on imac and help (but not completely fix)
+			 * https://bugzilla.redhat.com/show_bug.cgi?id=726143
+			 * also avoid interrupt storms during dpms.
+			 */
+			continue;
+		}
+		switch (amdgpu_connector->hpd.hpd) {
+		case AMDGPU_HPD_1:
+			WREG32(mmDC_HPD1_CONTROL, tmp);
+			break;
+		case AMDGPU_HPD_2:
+			WREG32(mmDC_HPD2_CONTROL, tmp);
+			break;
+		case AMDGPU_HPD_3:
+			WREG32(mmDC_HPD3_CONTROL, tmp);
+			break;
+		case AMDGPU_HPD_4:
+			WREG32(mmDC_HPD4_CONTROL, tmp);
+			break;
+		case AMDGPU_HPD_5:
+			WREG32(mmDC_HPD5_CONTROL, tmp);
+			break;
+		case AMDGPU_HPD_6:
+			WREG32(mmDC_HPD6_CONTROL, tmp);
+			break;
+		default:
+			break;
+		}
+		dce_v8_0_hpd_set_polarity(adev, amdgpu_connector->hpd.hpd);
+		amdgpu_irq_get(adev, &adev->hpd_irq, amdgpu_connector->hpd.hpd);
+	}
+}
+
+/**
+ * dce_v8_0_hpd_fini - hpd tear down callback.
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Tear down the hpd pins used by the card (evergreen+).
+ * Disable the hpd interrupts.
+ */
+static void dce_v8_0_hpd_fini(struct amdgpu_device *adev)
+{
+	struct drm_device *dev = adev->ddev;
+	struct drm_connector *connector;
+
+	list_for_each_entry(connector, &dev->mode_config.connector_list, head) {
+		struct amdgpu_connector *amdgpu_connector = to_amdgpu_connector(connector);
+
+		switch (amdgpu_connector->hpd.hpd) {
+		case AMDGPU_HPD_1:
+			WREG32(mmDC_HPD1_CONTROL, 0);
+			break;
+		case AMDGPU_HPD_2:
+			WREG32(mmDC_HPD2_CONTROL, 0);
+			break;
+		case AMDGPU_HPD_3:
+			WREG32(mmDC_HPD3_CONTROL, 0);
+			break;
+		case AMDGPU_HPD_4:
+			WREG32(mmDC_HPD4_CONTROL, 0);
+			break;
+		case AMDGPU_HPD_5:
+			WREG32(mmDC_HPD5_CONTROL, 0);
+			break;
+		case AMDGPU_HPD_6:
+			WREG32(mmDC_HPD6_CONTROL, 0);
+			break;
+		default:
+			break;
+		}
+		amdgpu_irq_put(adev, &adev->hpd_irq, amdgpu_connector->hpd.hpd);
+	}
+}
+
+static u32 dce_v8_0_hpd_get_gpio_reg(struct amdgpu_device *adev)
+{
+	return mmDC_GPIO_HPD_A;
+}
+
+static bool dce_v8_0_is_display_hung(struct amdgpu_device *adev)
+{
+	u32 crtc_hung = 0;
+	u32 crtc_status[6];
+	u32 i, j, tmp;
+
+	for (i = 0; i < adev->mode_info.num_crtc; i++) {
+		if (RREG32(mmCRTC_CONTROL + crtc_offsets[i]) & CRTC_CONTROL__CRTC_MASTER_EN_MASK) {
+			crtc_status[i] = RREG32(mmCRTC_STATUS_HV_COUNT + crtc_offsets[i]);
+			crtc_hung |= (1 << i);
+		}
+	}
+
+	for (j = 0; j < 10; j++) {
+		for (i = 0; i < adev->mode_info.num_crtc; i++) {
+			if (crtc_hung & (1 << i)) {
+				tmp = RREG32(mmCRTC_STATUS_HV_COUNT + crtc_offsets[i]);
+				if (tmp != crtc_status[i])
+					crtc_hung &= ~(1 << i);
+			}
+		}
+		if (crtc_hung == 0)
+			return false;
+		udelay(100);
+	}
+
+	return true;
+}
+
+static void dce_v8_0_stop_mc_access(struct amdgpu_device *adev,
+				    struct amdgpu_mode_mc_save *save)
+{
+	u32 crtc_enabled, tmp;
+	int i;
+
+	save->vga_render_control = RREG32(mmVGA_RENDER_CONTROL);
+	save->vga_hdp_control = RREG32(mmVGA_HDP_CONTROL);
+
+	/* disable VGA render */
+	tmp = RREG32(mmVGA_RENDER_CONTROL);
+	tmp = REG_SET_FIELD(tmp, VGA_RENDER_CONTROL, VGA_VSTATUS_CNTL, 0);
+	WREG32(mmVGA_RENDER_CONTROL, tmp);
+
+	/* blank the display controllers */
+	for (i = 0; i < adev->mode_info.num_crtc; i++) {
+		crtc_enabled = REG_GET_FIELD(RREG32(mmCRTC_CONTROL + crtc_offsets[i]),
+					     CRTC_CONTROL, CRTC_MASTER_EN);
+		if (crtc_enabled) {
+#if 0
+			u32 frame_count;
+			int j;
+
+			save->crtc_enabled[i] = true;
+			tmp = RREG32(mmCRTC_BLANK_CONTROL + crtc_offsets[i]);
+			if (REG_GET_FIELD(tmp, CRTC_BLANK_CONTROL, CRTC_BLANK_DATA_EN) == 0) {
+				amdgpu_display_vblank_wait(adev, i);
+				WREG32(mmCRTC_UPDATE_LOCK + crtc_offsets[i], 1);
+				tmp = REG_SET_FIELD(tmp, CRTC_BLANK_CONTROL, CRTC_BLANK_DATA_EN, 1);
+				WREG32(mmCRTC_BLANK_CONTROL + crtc_offsets[i], tmp);
+				WREG32(mmCRTC_UPDATE_LOCK + crtc_offsets[i], 0);
+			}
+			/* wait for the next frame */
+			frame_count = amdgpu_display_vblank_get_counter(adev, i);
+			for (j = 0; j < adev->usec_timeout; j++) {
+				if (amdgpu_display_vblank_get_counter(adev, i) != frame_count)
+					break;
+				udelay(1);
+			}
+			tmp = RREG32(mmGRPH_UPDATE + crtc_offsets[i]);
+			if (REG_GET_FIELD(tmp, GRPH_UPDATE, GRPH_UPDATE_LOCK) == 0) {
+				tmp = REG_SET_FIELD(tmp, GRPH_UPDATE, GRPH_UPDATE_LOCK, 1);
+				WREG32(mmGRPH_UPDATE + crtc_offsets[i], tmp);
+			}
+			tmp = RREG32(mmMASTER_UPDATE_LOCK + crtc_offsets[i]);
+			if (REG_GET_FIELD(tmp, MASTER_UPDATE_LOCK, MASTER_UPDATE_LOCK) == 0) {
+				tmp = REG_SET_FIELD(tmp, MASTER_UPDATE_LOCK, MASTER_UPDATE_LOCK, 1);
+				WREG32(mmMASTER_UPDATE_LOCK + crtc_offsets[i], tmp);
+			}
+#else
+			/* XXX this is a hack to avoid strange behavior with EFI on certain systems */
+			WREG32(mmCRTC_UPDATE_LOCK + crtc_offsets[i], 1);
+			tmp = RREG32(mmCRTC_CONTROL + crtc_offsets[i]);
+			tmp = REG_SET_FIELD(tmp, CRTC_CONTROL, CRTC_MASTER_EN, 0);
+			WREG32(mmCRTC_CONTROL + crtc_offsets[i], tmp);
+			WREG32(mmCRTC_UPDATE_LOCK + crtc_offsets[i], 0);
+			save->crtc_enabled[i] = false;
+			/* ***** */
+#endif
+		} else {
+			save->crtc_enabled[i] = false;
+		}
+	}
+}
+
+static void dce_v8_0_resume_mc_access(struct amdgpu_device *adev,
+				      struct amdgpu_mode_mc_save *save)
+{
+	u32 tmp, frame_count;
+	int i, j;
+
+	/* update crtc base addresses */
+	for (i = 0; i < adev->mode_info.num_crtc; i++) {
+		WREG32(mmGRPH_PRIMARY_SURFACE_ADDRESS_HIGH + crtc_offsets[i],
+		       upper_32_bits(adev->mc.vram_start));
+		WREG32(mmGRPH_SECONDARY_SURFACE_ADDRESS_HIGH + crtc_offsets[i],
+		       upper_32_bits(adev->mc.vram_start));
+		WREG32(mmGRPH_PRIMARY_SURFACE_ADDRESS + crtc_offsets[i],
+		       (u32)adev->mc.vram_start);
+		WREG32(mmGRPH_SECONDARY_SURFACE_ADDRESS + crtc_offsets[i],
+		       (u32)adev->mc.vram_start);
+
+		if (save->crtc_enabled[i]) {
+			tmp = RREG32(mmMASTER_UPDATE_MODE + crtc_offsets[i]);
+			if (REG_GET_FIELD(tmp, MASTER_UPDATE_MODE, MASTER_UPDATE_MODE) != 3) {
+				tmp = REG_SET_FIELD(tmp, MASTER_UPDATE_MODE, MASTER_UPDATE_MODE, 3);
+				WREG32(mmMASTER_UPDATE_MODE + crtc_offsets[i], tmp);
+			}
+			tmp = RREG32(mmGRPH_UPDATE + crtc_offsets[i]);
+			if (REG_GET_FIELD(tmp, GRPH_UPDATE, GRPH_UPDATE_LOCK)) {
+				tmp = REG_SET_FIELD(tmp, GRPH_UPDATE, GRPH_UPDATE_LOCK, 0);
+				WREG32(mmGRPH_UPDATE + crtc_offsets[i], tmp);
+			}
+			tmp = RREG32(mmMASTER_UPDATE_LOCK + crtc_offsets[i]);
+			if (REG_GET_FIELD(tmp, MASTER_UPDATE_LOCK, MASTER_UPDATE_LOCK)) {
+				tmp = REG_SET_FIELD(tmp, MASTER_UPDATE_LOCK, MASTER_UPDATE_LOCK, 0);
+				WREG32(mmMASTER_UPDATE_LOCK + crtc_offsets[i], tmp);
+			}
+			for (j = 0; j < adev->usec_timeout; j++) {
+				tmp = RREG32(mmGRPH_UPDATE + crtc_offsets[i]);
+				if (REG_GET_FIELD(tmp, GRPH_UPDATE, GRPH_SURFACE_UPDATE_PENDING) == 0)
+					break;
+				udelay(1);
+			}
+			tmp = RREG32(mmCRTC_BLANK_CONTROL + crtc_offsets[i]);
+			tmp = REG_SET_FIELD(tmp, CRTC_BLANK_CONTROL, CRTC_BLANK_DATA_EN, 0);
+			WREG32(mmCRTC_UPDATE_LOCK + crtc_offsets[i], 1);
+			WREG32(mmCRTC_BLANK_CONTROL + crtc_offsets[i], tmp);
+			WREG32(mmCRTC_UPDATE_LOCK + crtc_offsets[i], 0);
+			/* wait for the next frame */
+			frame_count = amdgpu_display_vblank_get_counter(adev, i);
+			for (j = 0; j < adev->usec_timeout; j++) {
+				if (amdgpu_display_vblank_get_counter(adev, i) != frame_count)
+					break;
+				udelay(1);
+			}
+		}
+	}
+
+	WREG32(mmVGA_MEMORY_BASE_ADDRESS_HIGH, upper_32_bits(adev->mc.vram_start));
+	WREG32(mmVGA_MEMORY_BASE_ADDRESS, lower_32_bits(adev->mc.vram_start));
+
+	/* Unlock vga access */
+	WREG32(mmVGA_HDP_CONTROL, save->vga_hdp_control);
+	mdelay(1);
+	WREG32(mmVGA_RENDER_CONTROL, save->vga_render_control);
+}
+
+static void dce_v8_0_set_vga_render_state(struct amdgpu_device *adev,
+					  bool render)
+{
+	u32 tmp;
+
+	/* Lockout access through VGA aperture*/
+	tmp = RREG32(mmVGA_HDP_CONTROL);
+	if (render)
+		tmp = REG_SET_FIELD(tmp, VGA_HDP_CONTROL, VGA_MEMORY_DISABLE, 0);
+	else
+		tmp = REG_SET_FIELD(tmp, VGA_HDP_CONTROL, VGA_MEMORY_DISABLE, 1);
+	WREG32(mmVGA_HDP_CONTROL, tmp);
+
+	/* disable VGA render */
+	tmp = RREG32(mmVGA_RENDER_CONTROL);
+	if (render)
+		tmp = REG_SET_FIELD(tmp, VGA_RENDER_CONTROL, VGA_VSTATUS_CNTL, 1);
+	else
+		tmp = REG_SET_FIELD(tmp, VGA_RENDER_CONTROL, VGA_VSTATUS_CNTL, 0);
+	WREG32(mmVGA_RENDER_CONTROL, tmp);
+}
+
+static void dce_v8_0_program_fmt(struct drm_encoder *encoder)
+{
+	struct drm_device *dev = encoder->dev;
+	struct amdgpu_device *adev = dev->dev_private;
+	struct amdgpu_encoder *amdgpu_encoder = to_amdgpu_encoder(encoder);
+	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(encoder->crtc);
+	struct drm_connector *connector = amdgpu_get_connector_for_encoder(encoder);
+	int bpc = 0;
+	u32 tmp = 0;
+	enum amdgpu_connector_dither dither = AMDGPU_FMT_DITHER_DISABLE;
+
+	if (connector) {
+		struct amdgpu_connector *amdgpu_connector = to_amdgpu_connector(connector);
+		bpc = amdgpu_connector_get_monitor_bpc(connector);
+		dither = amdgpu_connector->dither;
+	}
+
+	/* LVDS/eDP FMT is set up by atom */
+	if (amdgpu_encoder->devices & ATOM_DEVICE_LCD_SUPPORT)
+		return;
+
+	/* not needed for analog */
+	if ((amdgpu_encoder->encoder_id == ENCODER_OBJECT_ID_INTERNAL_KLDSCP_DAC1) ||
+	    (amdgpu_encoder->encoder_id == ENCODER_OBJECT_ID_INTERNAL_KLDSCP_DAC2))
+		return;
+
+	if (bpc == 0)
+		return;
+
+	switch (bpc) {
+	case 6:
+		if (dither == AMDGPU_FMT_DITHER_ENABLE)
+			/* XXX sort out optimal dither settings */
+			tmp |= (FMT_BIT_DEPTH_CONTROL__FMT_FRAME_RANDOM_ENABLE_MASK |
+				FMT_BIT_DEPTH_CONTROL__FMT_HIGHPASS_RANDOM_ENABLE_MASK |
+				FMT_BIT_DEPTH_CONTROL__FMT_SPATIAL_DITHER_EN_MASK |
+				(0 << FMT_BIT_DEPTH_CONTROL__FMT_SPATIAL_DITHER_DEPTH__SHIFT));
+		else
+			tmp |= (FMT_BIT_DEPTH_CONTROL__FMT_TRUNCATE_EN_MASK |
+			(0 << FMT_BIT_DEPTH_CONTROL__FMT_TRUNCATE_DEPTH__SHIFT));
+		break;
+	case 8:
+		if (dither == AMDGPU_FMT_DITHER_ENABLE)
+			/* XXX sort out optimal dither settings */
+			tmp |= (FMT_BIT_DEPTH_CONTROL__FMT_FRAME_RANDOM_ENABLE_MASK |
+				FMT_BIT_DEPTH_CONTROL__FMT_HIGHPASS_RANDOM_ENABLE_MASK |
+				FMT_BIT_DEPTH_CONTROL__FMT_RGB_RANDOM_ENABLE_MASK |
+				FMT_BIT_DEPTH_CONTROL__FMT_SPATIAL_DITHER_EN_MASK |
+				(1 << FMT_BIT_DEPTH_CONTROL__FMT_SPATIAL_DITHER_DEPTH__SHIFT));
+		else
+			tmp |= (FMT_BIT_DEPTH_CONTROL__FMT_TRUNCATE_EN_MASK |
+			(1 << FMT_BIT_DEPTH_CONTROL__FMT_TRUNCATE_DEPTH__SHIFT));
+		break;
+	case 10:
+		if (dither == AMDGPU_FMT_DITHER_ENABLE)
+			/* XXX sort out optimal dither settings */
+			tmp |= (FMT_BIT_DEPTH_CONTROL__FMT_FRAME_RANDOM_ENABLE_MASK |
+				FMT_BIT_DEPTH_CONTROL__FMT_HIGHPASS_RANDOM_ENABLE_MASK |
+				FMT_BIT_DEPTH_CONTROL__FMT_RGB_RANDOM_ENABLE_MASK |
+				FMT_BIT_DEPTH_CONTROL__FMT_SPATIAL_DITHER_EN_MASK |
+				(2 << FMT_BIT_DEPTH_CONTROL__FMT_SPATIAL_DITHER_DEPTH__SHIFT));
+		else
+			tmp |= (FMT_BIT_DEPTH_CONTROL__FMT_TRUNCATE_EN_MASK |
+			(2 << FMT_BIT_DEPTH_CONTROL__FMT_TRUNCATE_DEPTH__SHIFT));
+		break;
+	default:
+		/* not needed */
+		break;
+	}
+
+	WREG32(mmFMT_BIT_DEPTH_CONTROL + amdgpu_crtc->crtc_offset, tmp);
+}
+
+
+/* display watermark setup */
+/**
+ * dce_v8_0_line_buffer_adjust - Set up the line buffer
+ *
+ * @adev: amdgpu_device pointer
+ * @amdgpu_crtc: the selected display controller
+ * @mode: the current display mode on the selected display
+ * controller
+ *
+ * Setup up the line buffer allocation for
+ * the selected display controller (CIK).
+ * Returns the line buffer size in pixels.
+ */
+static u32 dce_v8_0_line_buffer_adjust(struct amdgpu_device *adev,
+				       struct amdgpu_crtc *amdgpu_crtc,
+				       struct drm_display_mode *mode)
+{
+	u32 tmp, buffer_alloc, i;
+	u32 pipe_offset = amdgpu_crtc->crtc_id * 0x8;
+	/*
+	 * Line Buffer Setup
+	 * There are 6 line buffers, one for each display controllers.
+	 * There are 3 partitions per LB. Select the number of partitions
+	 * to enable based on the display width.  For display widths larger
+	 * than 4096, you need use to use 2 display controllers and combine
+	 * them using the stereo blender.
+	 */
+	if (amdgpu_crtc->base.enabled && mode) {
+		if (mode->crtc_hdisplay < 1920) {
+			tmp = 1;
+			buffer_alloc = 2;
+		} else if (mode->crtc_hdisplay < 2560) {
+			tmp = 2;
+			buffer_alloc = 2;
+		} else if (mode->crtc_hdisplay < 4096) {
+			tmp = 0;
+			buffer_alloc = (adev->flags & AMDGPU_IS_APU) ? 2 : 4;
+		} else {
+			DRM_DEBUG_KMS("Mode too big for LB!\n");
+			tmp = 0;
+			buffer_alloc = (adev->flags & AMDGPU_IS_APU) ? 2 : 4;
+		}
+	} else {
+		tmp = 1;
+		buffer_alloc = 0;
+	}
+
+	WREG32(mmLB_MEMORY_CTRL + amdgpu_crtc->crtc_offset,
+	      (tmp << LB_MEMORY_CTRL__LB_MEMORY_CONFIG__SHIFT) |
+	      (0x6B0 << LB_MEMORY_CTRL__LB_MEMORY_SIZE__SHIFT));
+
+	WREG32(mmPIPE0_DMIF_BUFFER_CONTROL + pipe_offset,
+	       (buffer_alloc << PIPE0_DMIF_BUFFER_CONTROL__DMIF_BUFFERS_ALLOCATED__SHIFT));
+	for (i = 0; i < adev->usec_timeout; i++) {
+		if (RREG32(mmPIPE0_DMIF_BUFFER_CONTROL + pipe_offset) &
+		    PIPE0_DMIF_BUFFER_CONTROL__DMIF_BUFFERS_ALLOCATION_COMPLETED_MASK)
+			break;
+		udelay(1);
+	}
+
+	if (amdgpu_crtc->base.enabled && mode) {
+		switch (tmp) {
+		case 0:
+		default:
+			return 4096 * 2;
+		case 1:
+			return 1920 * 2;
+		case 2:
+			return 2560 * 2;
+		}
+	}
+
+	/* controller not enabled, so no lb used */
+	return 0;
+}
+
+/**
+ * cik_get_number_of_dram_channels - get the number of dram channels
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Look up the number of video ram channels (CIK).
+ * Used for display watermark bandwidth calculations
+ * Returns the number of dram channels
+ */
+static u32 cik_get_number_of_dram_channels(struct amdgpu_device *adev)
+{
+	u32 tmp = RREG32(mmMC_SHARED_CHMAP);
+
+	switch ((tmp & MC_SHARED_CHMAP__NOOFCHAN_MASK) >> MC_SHARED_CHMAP__NOOFCHAN__SHIFT) {
+	case 0:
+	default:
+		return 1;
+	case 1:
+		return 2;
+	case 2:
+		return 4;
+	case 3:
+		return 8;
+	case 4:
+		return 3;
+	case 5:
+		return 6;
+	case 6:
+		return 10;
+	case 7:
+		return 12;
+	case 8:
+		return 16;
+	}
+}
+
+struct dce8_wm_params {
+	u32 dram_channels; /* number of dram channels */
+	u32 yclk;          /* bandwidth per dram data pin in kHz */
+	u32 sclk;          /* engine clock in kHz */
+	u32 disp_clk;      /* display clock in kHz */
+	u32 src_width;     /* viewport width */
+	u32 active_time;   /* active display time in ns */
+	u32 blank_time;    /* blank time in ns */
+	bool interlaced;    /* mode is interlaced */
+	fixed20_12 vsc;    /* vertical scale ratio */
+	u32 num_heads;     /* number of active crtcs */
+	u32 bytes_per_pixel; /* bytes per pixel display + overlay */
+	u32 lb_size;       /* line buffer allocated to pipe */
+	u32 vtaps;         /* vertical scaler taps */
+};
+
+/**
+ * dce_v8_0_dram_bandwidth - get the dram bandwidth
+ *
+ * @wm: watermark calculation data
+ *
+ * Calculate the raw dram bandwidth (CIK).
+ * Used for display watermark bandwidth calculations
+ * Returns the dram bandwidth in MBytes/s
+ */
+static u32 dce_v8_0_dram_bandwidth(struct dce8_wm_params *wm)
+{
+	/* Calculate raw DRAM Bandwidth */
+	fixed20_12 dram_efficiency; /* 0.7 */
+	fixed20_12 yclk, dram_channels, bandwidth;
+	fixed20_12 a;
+
+	a.full = dfixed_const(1000);
+	yclk.full = dfixed_const(wm->yclk);
+	yclk.full = dfixed_div(yclk, a);
+	dram_channels.full = dfixed_const(wm->dram_channels * 4);
+	a.full = dfixed_const(10);
+	dram_efficiency.full = dfixed_const(7);
+	dram_efficiency.full = dfixed_div(dram_efficiency, a);
+	bandwidth.full = dfixed_mul(dram_channels, yclk);
+	bandwidth.full = dfixed_mul(bandwidth, dram_efficiency);
+
+	return dfixed_trunc(bandwidth);
+}
+
+/**
+ * dce_v8_0_dram_bandwidth_for_display - get the dram bandwidth for display
+ *
+ * @wm: watermark calculation data
+ *
+ * Calculate the dram bandwidth used for display (CIK).
+ * Used for display watermark bandwidth calculations
+ * Returns the dram bandwidth for display in MBytes/s
+ */
+static u32 dce_v8_0_dram_bandwidth_for_display(struct dce8_wm_params *wm)
+{
+	/* Calculate DRAM Bandwidth and the part allocated to display. */
+	fixed20_12 disp_dram_allocation; /* 0.3 to 0.7 */
+	fixed20_12 yclk, dram_channels, bandwidth;
+	fixed20_12 a;
+
+	a.full = dfixed_const(1000);
+	yclk.full = dfixed_const(wm->yclk);
+	yclk.full = dfixed_div(yclk, a);
+	dram_channels.full = dfixed_const(wm->dram_channels * 4);
+	a.full = dfixed_const(10);
+	disp_dram_allocation.full = dfixed_const(3); /* XXX worse case value 0.3 */
+	disp_dram_allocation.full = dfixed_div(disp_dram_allocation, a);
+	bandwidth.full = dfixed_mul(dram_channels, yclk);
+	bandwidth.full = dfixed_mul(bandwidth, disp_dram_allocation);
+
+	return dfixed_trunc(bandwidth);
+}
+
+/**
+ * dce_v8_0_data_return_bandwidth - get the data return bandwidth
+ *
+ * @wm: watermark calculation data
+ *
+ * Calculate the data return bandwidth used for display (CIK).
+ * Used for display watermark bandwidth calculations
+ * Returns the data return bandwidth in MBytes/s
+ */
+static u32 dce_v8_0_data_return_bandwidth(struct dce8_wm_params *wm)
+{
+	/* Calculate the display Data return Bandwidth */
+	fixed20_12 return_efficiency; /* 0.8 */
+	fixed20_12 sclk, bandwidth;
+	fixed20_12 a;
+
+	a.full = dfixed_const(1000);
+	sclk.full = dfixed_const(wm->sclk);
+	sclk.full = dfixed_div(sclk, a);
+	a.full = dfixed_const(10);
+	return_efficiency.full = dfixed_const(8);
+	return_efficiency.full = dfixed_div(return_efficiency, a);
+	a.full = dfixed_const(32);
+	bandwidth.full = dfixed_mul(a, sclk);
+	bandwidth.full = dfixed_mul(bandwidth, return_efficiency);
+
+	return dfixed_trunc(bandwidth);
+}
+
+/**
+ * dce_v8_0_dmif_request_bandwidth - get the dmif bandwidth
+ *
+ * @wm: watermark calculation data
+ *
+ * Calculate the dmif bandwidth used for display (CIK).
+ * Used for display watermark bandwidth calculations
+ * Returns the dmif bandwidth in MBytes/s
+ */
+static u32 dce_v8_0_dmif_request_bandwidth(struct dce8_wm_params *wm)
+{
+	/* Calculate the DMIF Request Bandwidth */
+	fixed20_12 disp_clk_request_efficiency; /* 0.8 */
+	fixed20_12 disp_clk, bandwidth;
+	fixed20_12 a, b;
+
+	a.full = dfixed_const(1000);
+	disp_clk.full = dfixed_const(wm->disp_clk);
+	disp_clk.full = dfixed_div(disp_clk, a);
+	a.full = dfixed_const(32);
+	b.full = dfixed_mul(a, disp_clk);
+
+	a.full = dfixed_const(10);
+	disp_clk_request_efficiency.full = dfixed_const(8);
+	disp_clk_request_efficiency.full = dfixed_div(disp_clk_request_efficiency, a);
+
+	bandwidth.full = dfixed_mul(b, disp_clk_request_efficiency);
+
+	return dfixed_trunc(bandwidth);
+}
+
+/**
+ * dce_v8_0_available_bandwidth - get the min available bandwidth
+ *
+ * @wm: watermark calculation data
+ *
+ * Calculate the min available bandwidth used for display (CIK).
+ * Used for display watermark bandwidth calculations
+ * Returns the min available bandwidth in MBytes/s
+ */
+static u32 dce_v8_0_available_bandwidth(struct dce8_wm_params *wm)
+{
+	/* Calculate the Available bandwidth. Display can use this temporarily but not in average. */
+	u32 dram_bandwidth = dce_v8_0_dram_bandwidth(wm);
+	u32 data_return_bandwidth = dce_v8_0_data_return_bandwidth(wm);
+	u32 dmif_req_bandwidth = dce_v8_0_dmif_request_bandwidth(wm);
+
+	return min(dram_bandwidth, min(data_return_bandwidth, dmif_req_bandwidth));
+}
+
+/**
+ * dce_v8_0_average_bandwidth - get the average available bandwidth
+ *
+ * @wm: watermark calculation data
+ *
+ * Calculate the average available bandwidth used for display (CIK).
+ * Used for display watermark bandwidth calculations
+ * Returns the average available bandwidth in MBytes/s
+ */
+static u32 dce_v8_0_average_bandwidth(struct dce8_wm_params *wm)
+{
+	/* Calculate the display mode Average Bandwidth
+	 * DisplayMode should contain the source and destination dimensions,
+	 * timing, etc.
+	 */
+	fixed20_12 bpp;
+	fixed20_12 line_time;
+	fixed20_12 src_width;
+	fixed20_12 bandwidth;
+	fixed20_12 a;
+
+	a.full = dfixed_const(1000);
+	line_time.full = dfixed_const(wm->active_time + wm->blank_time);
+	line_time.full = dfixed_div(line_time, a);
+	bpp.full = dfixed_const(wm->bytes_per_pixel);
+	src_width.full = dfixed_const(wm->src_width);
+	bandwidth.full = dfixed_mul(src_width, bpp);
+	bandwidth.full = dfixed_mul(bandwidth, wm->vsc);
+	bandwidth.full = dfixed_div(bandwidth, line_time);
+
+	return dfixed_trunc(bandwidth);
+}
+
+/**
+ * dce_v8_0_latency_watermark - get the latency watermark
+ *
+ * @wm: watermark calculation data
+ *
+ * Calculate the latency watermark (CIK).
+ * Used for display watermark bandwidth calculations
+ * Returns the latency watermark in ns
+ */
+static u32 dce_v8_0_latency_watermark(struct dce8_wm_params *wm)
+{
+	/* First calculate the latency in ns */
+	u32 mc_latency = 2000; /* 2000 ns. */
+	u32 available_bandwidth = dce_v8_0_available_bandwidth(wm);
+	u32 worst_chunk_return_time = (512 * 8 * 1000) / available_bandwidth;
+	u32 cursor_line_pair_return_time = (128 * 4 * 1000) / available_bandwidth;
+	u32 dc_latency = 40000000 / wm->disp_clk; /* dc pipe latency */
+	u32 other_heads_data_return_time = ((wm->num_heads + 1) * worst_chunk_return_time) +
+		(wm->num_heads * cursor_line_pair_return_time);
+	u32 latency = mc_latency + other_heads_data_return_time + dc_latency;
+	u32 max_src_lines_per_dst_line, lb_fill_bw, line_fill_time;
+	u32 tmp, dmif_size = 12288;
+	fixed20_12 a, b, c;
+
+	if (wm->num_heads == 0)
+		return 0;
+
+	a.full = dfixed_const(2);
+	b.full = dfixed_const(1);
+	if ((wm->vsc.full > a.full) ||
+	    ((wm->vsc.full > b.full) && (wm->vtaps >= 3)) ||
+	    (wm->vtaps >= 5) ||
+	    ((wm->vsc.full >= a.full) && wm->interlaced))
+		max_src_lines_per_dst_line = 4;
+	else
+		max_src_lines_per_dst_line = 2;
+
+	a.full = dfixed_const(available_bandwidth);
+	b.full = dfixed_const(wm->num_heads);
+	a.full = dfixed_div(a, b);
+
+	b.full = dfixed_const(mc_latency + 512);
+	c.full = dfixed_const(wm->disp_clk);
+	b.full = dfixed_div(b, c);
+
+	c.full = dfixed_const(dmif_size);
+	b.full = dfixed_div(c, b);
+
+	tmp = min(dfixed_trunc(a), dfixed_trunc(b));
+
+	b.full = dfixed_const(1000);
+	c.full = dfixed_const(wm->disp_clk);
+	b.full = dfixed_div(c, b);
+	c.full = dfixed_const(wm->bytes_per_pixel);
+	b.full = dfixed_mul(b, c);
+
+	lb_fill_bw = min(tmp, dfixed_trunc(b));
+
+	a.full = dfixed_const(max_src_lines_per_dst_line * wm->src_width * wm->bytes_per_pixel);
+	b.full = dfixed_const(1000);
+	c.full = dfixed_const(lb_fill_bw);
+	b.full = dfixed_div(c, b);
+	a.full = dfixed_div(a, b);
+	line_fill_time = dfixed_trunc(a);
+
+	if (line_fill_time < wm->active_time)
+		return latency;
+	else
+		return latency + (line_fill_time - wm->active_time);
+
+}
+
+/**
+ * dce_v8_0_average_bandwidth_vs_dram_bandwidth_for_display - check
+ * average and available dram bandwidth
+ *
+ * @wm: watermark calculation data
+ *
+ * Check if the display average bandwidth fits in the display
+ * dram bandwidth (CIK).
+ * Used for display watermark bandwidth calculations
+ * Returns true if the display fits, false if not.
+ */
+static bool dce_v8_0_average_bandwidth_vs_dram_bandwidth_for_display(struct dce8_wm_params *wm)
+{
+	if (dce_v8_0_average_bandwidth(wm) <=
+	    (dce_v8_0_dram_bandwidth_for_display(wm) / wm->num_heads))
+		return true;
+	else
+		return false;
+}
+
+/**
+ * dce_v8_0_average_bandwidth_vs_available_bandwidth - check
+ * average and available bandwidth
+ *
+ * @wm: watermark calculation data
+ *
+ * Check if the display average bandwidth fits in the display
+ * available bandwidth (CIK).
+ * Used for display watermark bandwidth calculations
+ * Returns true if the display fits, false if not.
+ */
+static bool dce_v8_0_average_bandwidth_vs_available_bandwidth(struct dce8_wm_params *wm)
+{
+	if (dce_v8_0_average_bandwidth(wm) <=
+	    (dce_v8_0_available_bandwidth(wm) / wm->num_heads))
+		return true;
+	else
+		return false;
+}
+
+/**
+ * dce_v8_0_check_latency_hiding - check latency hiding
+ *
+ * @wm: watermark calculation data
+ *
+ * Check latency hiding (CIK).
+ * Used for display watermark bandwidth calculations
+ * Returns true if the display fits, false if not.
+ */
+static bool dce_v8_0_check_latency_hiding(struct dce8_wm_params *wm)
+{
+	u32 lb_partitions = wm->lb_size / wm->src_width;
+	u32 line_time = wm->active_time + wm->blank_time;
+	u32 latency_tolerant_lines;
+	u32 latency_hiding;
+	fixed20_12 a;
+
+	a.full = dfixed_const(1);
+	if (wm->vsc.full > a.full)
+		latency_tolerant_lines = 1;
+	else {
+		if (lb_partitions <= (wm->vtaps + 1))
+			latency_tolerant_lines = 1;
+		else
+			latency_tolerant_lines = 2;
+	}
+
+	latency_hiding = (latency_tolerant_lines * line_time + wm->blank_time);
+
+	if (dce_v8_0_latency_watermark(wm) <= latency_hiding)
+		return true;
+	else
+		return false;
+}
+
+/**
+ * dce_v8_0_program_watermarks - program display watermarks
+ *
+ * @adev: amdgpu_device pointer
+ * @amdgpu_crtc: the selected display controller
+ * @lb_size: line buffer size
+ * @num_heads: number of display controllers in use
+ *
+ * Calculate and program the display watermarks for the
+ * selected display controller (CIK).
+ */
+static void dce_v8_0_program_watermarks(struct amdgpu_device *adev,
+					struct amdgpu_crtc *amdgpu_crtc,
+					u32 lb_size, u32 num_heads)
+{
+	struct drm_display_mode *mode = &amdgpu_crtc->base.mode;
+	struct dce8_wm_params wm_low, wm_high;
+	u32 pixel_period;
+	u32 line_time = 0;
+	u32 latency_watermark_a = 0, latency_watermark_b = 0;
+	u32 tmp, wm_mask;
+
+	if (amdgpu_crtc->base.enabled && num_heads && mode) {
+		pixel_period = 1000000 / (u32)mode->clock;
+		line_time = min((u32)mode->crtc_htotal * pixel_period, (u32)65535);
+
+		/* watermark for high clocks */
+		if (adev->pm.dpm_enabled) {
+			wm_high.yclk =
+				amdgpu_dpm_get_mclk(adev, false) * 10;
+			wm_high.sclk =
+				amdgpu_dpm_get_sclk(adev, false) * 10;
+		} else {
+			wm_high.yclk = adev->pm.current_mclk * 10;
+			wm_high.sclk = adev->pm.current_sclk * 10;
+		}
+
+		wm_high.disp_clk = mode->clock;
+		wm_high.src_width = mode->crtc_hdisplay;
+		wm_high.active_time = mode->crtc_hdisplay * pixel_period;
+		wm_high.blank_time = line_time - wm_high.active_time;
+		wm_high.interlaced = false;
+		if (mode->flags & DRM_MODE_FLAG_INTERLACE)
+			wm_high.interlaced = true;
+		wm_high.vsc = amdgpu_crtc->vsc;
+		wm_high.vtaps = 1;
+		if (amdgpu_crtc->rmx_type != RMX_OFF)
+			wm_high.vtaps = 2;
+		wm_high.bytes_per_pixel = 4; /* XXX: get this from fb config */
+		wm_high.lb_size = lb_size;
+		wm_high.dram_channels = cik_get_number_of_dram_channels(adev);
+		wm_high.num_heads = num_heads;
+
+		/* set for high clocks */
+		latency_watermark_a = min(dce_v8_0_latency_watermark(&wm_high), (u32)65535);
+
+		/* possibly force display priority to high */
+		/* should really do this at mode validation time... */
+		if (!dce_v8_0_average_bandwidth_vs_dram_bandwidth_for_display(&wm_high) ||
+		    !dce_v8_0_average_bandwidth_vs_available_bandwidth(&wm_high) ||
+		    !dce_v8_0_check_latency_hiding(&wm_high) ||
+		    (adev->mode_info.disp_priority == 2)) {
+			DRM_DEBUG_KMS("force priority to high\n");
+		}
+
+		/* watermark for low clocks */
+		if (adev->pm.dpm_enabled) {
+			wm_low.yclk =
+				amdgpu_dpm_get_mclk(adev, true) * 10;
+			wm_low.sclk =
+				amdgpu_dpm_get_sclk(adev, true) * 10;
+		} else {
+			wm_low.yclk = adev->pm.current_mclk * 10;
+			wm_low.sclk = adev->pm.current_sclk * 10;
+		}
+
+		wm_low.disp_clk = mode->clock;
+		wm_low.src_width = mode->crtc_hdisplay;
+		wm_low.active_time = mode->crtc_hdisplay * pixel_period;
+		wm_low.blank_time = line_time - wm_low.active_time;
+		wm_low.interlaced = false;
+		if (mode->flags & DRM_MODE_FLAG_INTERLACE)
+			wm_low.interlaced = true;
+		wm_low.vsc = amdgpu_crtc->vsc;
+		wm_low.vtaps = 1;
+		if (amdgpu_crtc->rmx_type != RMX_OFF)
+			wm_low.vtaps = 2;
+		wm_low.bytes_per_pixel = 4; /* XXX: get this from fb config */
+		wm_low.lb_size = lb_size;
+		wm_low.dram_channels = cik_get_number_of_dram_channels(adev);
+		wm_low.num_heads = num_heads;
+
+		/* set for low clocks */
+		latency_watermark_b = min(dce_v8_0_latency_watermark(&wm_low), (u32)65535);
+
+		/* possibly force display priority to high */
+		/* should really do this at mode validation time... */
+		if (!dce_v8_0_average_bandwidth_vs_dram_bandwidth_for_display(&wm_low) ||
+		    !dce_v8_0_average_bandwidth_vs_available_bandwidth(&wm_low) ||
+		    !dce_v8_0_check_latency_hiding(&wm_low) ||
+		    (adev->mode_info.disp_priority == 2)) {
+			DRM_DEBUG_KMS("force priority to high\n");
+		}
+	}
+
+	/* select wm A */
+	wm_mask = RREG32(mmDPG_WATERMARK_MASK_CONTROL + amdgpu_crtc->crtc_offset);
+	tmp = wm_mask;
+	tmp &= ~(3 << DPG_WATERMARK_MASK_CONTROL__URGENCY_WATERMARK_MASK__SHIFT);
+	tmp |= (1 << DPG_WATERMARK_MASK_CONTROL__URGENCY_WATERMARK_MASK__SHIFT);
+	WREG32(mmDPG_WATERMARK_MASK_CONTROL + amdgpu_crtc->crtc_offset, tmp);
+	WREG32(mmDPG_PIPE_URGENCY_CONTROL + amdgpu_crtc->crtc_offset,
+	       ((latency_watermark_a << DPG_PIPE_URGENCY_CONTROL__URGENCY_LOW_WATERMARK__SHIFT) |
+		(line_time << DPG_PIPE_URGENCY_CONTROL__URGENCY_HIGH_WATERMARK__SHIFT)));
+	/* select wm B */
+	tmp = RREG32(mmDPG_WATERMARK_MASK_CONTROL + amdgpu_crtc->crtc_offset);
+	tmp &= ~(3 << DPG_WATERMARK_MASK_CONTROL__URGENCY_WATERMARK_MASK__SHIFT);
+	tmp |= (2 << DPG_WATERMARK_MASK_CONTROL__URGENCY_WATERMARK_MASK__SHIFT);
+	WREG32(mmDPG_WATERMARK_MASK_CONTROL + amdgpu_crtc->crtc_offset, tmp);
+	WREG32(mmDPG_PIPE_URGENCY_CONTROL + amdgpu_crtc->crtc_offset,
+	       ((latency_watermark_b << DPG_PIPE_URGENCY_CONTROL__URGENCY_LOW_WATERMARK__SHIFT) |
+		(line_time << DPG_PIPE_URGENCY_CONTROL__URGENCY_HIGH_WATERMARK__SHIFT)));
+	/* restore original selection */
+	WREG32(mmDPG_WATERMARK_MASK_CONTROL + amdgpu_crtc->crtc_offset, wm_mask);
+
+	/* save values for DPM */
+	amdgpu_crtc->line_time = line_time;
+	amdgpu_crtc->wm_high = latency_watermark_a;
+	amdgpu_crtc->wm_low = latency_watermark_b;
+}
+
+/**
+ * dce_v8_0_bandwidth_update - program display watermarks
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Calculate and program the display watermarks and line
+ * buffer allocation (CIK).
+ */
+static void dce_v8_0_bandwidth_update(struct amdgpu_device *adev)
+{
+	struct drm_display_mode *mode = NULL;
+	u32 num_heads = 0, lb_size;
+	int i;
+
+	amdgpu_update_display_priority(adev);
+
+	for (i = 0; i < adev->mode_info.num_crtc; i++) {
+		if (adev->mode_info.crtcs[i]->base.enabled)
+			num_heads++;
+	}
+	for (i = 0; i < adev->mode_info.num_crtc; i++) {
+		mode = &adev->mode_info.crtcs[i]->base.mode;
+		lb_size = dce_v8_0_line_buffer_adjust(adev, adev->mode_info.crtcs[i], mode);
+		dce_v8_0_program_watermarks(adev, adev->mode_info.crtcs[i],
+					    lb_size, num_heads);
+	}
+}
+
+static void dce_v8_0_audio_get_connected_pins(struct amdgpu_device *adev)
+{
+	int i;
+	u32 offset, tmp;
+
+	for (i = 0; i < adev->mode_info.audio.num_pins; i++) {
+		offset = adev->mode_info.audio.pin[i].offset;
+		tmp = RREG32_AUDIO_ENDPT(offset,
+					 ixAZALIA_F0_CODEC_PIN_CONTROL_RESPONSE_CONFIGURATION_DEFAULT);
+		if (((tmp &
+		AZALIA_F0_CODEC_PIN_CONTROL_RESPONSE_CONFIGURATION_DEFAULT__PORT_CONNECTIVITY_MASK) >>
+		AZALIA_F0_CODEC_PIN_CONTROL_RESPONSE_CONFIGURATION_DEFAULT__PORT_CONNECTIVITY__SHIFT) == 1)
+			adev->mode_info.audio.pin[i].connected = false;
+		else
+			adev->mode_info.audio.pin[i].connected = true;
+	}
+}
+
+static struct amdgpu_audio_pin *dce_v8_0_audio_get_pin(struct amdgpu_device *adev)
+{
+	int i;
+
+	dce_v8_0_audio_get_connected_pins(adev);
+
+	for (i = 0; i < adev->mode_info.audio.num_pins; i++) {
+		if (adev->mode_info.audio.pin[i].connected)
+			return &adev->mode_info.audio.pin[i];
+	}
+	DRM_ERROR("No connected audio pins found!\n");
+	return NULL;
+}
+
+static void dce_v8_0_afmt_audio_select_pin(struct drm_encoder *encoder)
+{
+	struct amdgpu_device *adev = encoder->dev->dev_private;
+	struct amdgpu_encoder *amdgpu_encoder = to_amdgpu_encoder(encoder);
+	struct amdgpu_encoder_atom_dig *dig = amdgpu_encoder->enc_priv;
+	u32 offset;
+
+	if (!dig || !dig->afmt || !dig->afmt->pin)
+		return;
+
+	offset = dig->afmt->offset;
+
+	WREG32(mmAFMT_AUDIO_SRC_CONTROL + offset,
+	       (dig->afmt->pin->id << AFMT_AUDIO_SRC_CONTROL__AFMT_AUDIO_SRC_SELECT__SHIFT));
+}
+
+static void dce_v8_0_audio_write_latency_fields(struct drm_encoder *encoder,
+						struct drm_display_mode *mode)
+{
+	struct amdgpu_device *adev = encoder->dev->dev_private;
+	struct amdgpu_encoder *amdgpu_encoder = to_amdgpu_encoder(encoder);
+	struct amdgpu_encoder_atom_dig *dig = amdgpu_encoder->enc_priv;
+	struct drm_connector *connector;
+	struct amdgpu_connector *amdgpu_connector = NULL;
+	u32 tmp = 0, offset;
+
+	if (!dig || !dig->afmt || !dig->afmt->pin)
+		return;
+
+	offset = dig->afmt->pin->offset;
+
+	list_for_each_entry(connector, &encoder->dev->mode_config.connector_list, head) {
+		if (connector->encoder == encoder) {
+			amdgpu_connector = to_amdgpu_connector(connector);
+			break;
+		}
+	}
+
+	if (!amdgpu_connector) {
+		DRM_ERROR("Couldn't find encoder's connector\n");
+		return;
+	}
+
+	if (mode->flags & DRM_MODE_FLAG_INTERLACE) {
+		if (connector->latency_present[1])
+			tmp =
+			(connector->video_latency[1] <<
+			 AZALIA_F0_CODEC_PIN_CONTROL_RESPONSE_LIPSYNC__VIDEO_LIPSYNC__SHIFT) |
+			(connector->audio_latency[1] <<
+			 AZALIA_F0_CODEC_PIN_CONTROL_RESPONSE_LIPSYNC__AUDIO_LIPSYNC__SHIFT);
+		else
+			tmp =
+			(0 <<
+			 AZALIA_F0_CODEC_PIN_CONTROL_RESPONSE_LIPSYNC__VIDEO_LIPSYNC__SHIFT) |
+			(0 <<
+			 AZALIA_F0_CODEC_PIN_CONTROL_RESPONSE_LIPSYNC__AUDIO_LIPSYNC__SHIFT);
+	} else {
+		if (connector->latency_present[0])
+			tmp =
+			(connector->video_latency[0] <<
+			 AZALIA_F0_CODEC_PIN_CONTROL_RESPONSE_LIPSYNC__VIDEO_LIPSYNC__SHIFT) |
+			(connector->audio_latency[0] <<
+			 AZALIA_F0_CODEC_PIN_CONTROL_RESPONSE_LIPSYNC__AUDIO_LIPSYNC__SHIFT);
+		else
+			tmp =
+			(0 <<
+			 AZALIA_F0_CODEC_PIN_CONTROL_RESPONSE_LIPSYNC__VIDEO_LIPSYNC__SHIFT) |
+			(0 <<
+			 AZALIA_F0_CODEC_PIN_CONTROL_RESPONSE_LIPSYNC__AUDIO_LIPSYNC__SHIFT);
+
+	}
+	WREG32_AUDIO_ENDPT(offset, ixAZALIA_F0_CODEC_PIN_CONTROL_RESPONSE_LIPSYNC, tmp);
+}
+
+static void dce_v8_0_audio_write_speaker_allocation(struct drm_encoder *encoder)
+{
+	struct amdgpu_device *adev = encoder->dev->dev_private;
+	struct amdgpu_encoder *amdgpu_encoder = to_amdgpu_encoder(encoder);
+	struct amdgpu_encoder_atom_dig *dig = amdgpu_encoder->enc_priv;
+	struct drm_connector *connector;
+	struct amdgpu_connector *amdgpu_connector = NULL;
+	u32 offset, tmp;
+	u8 *sadb = NULL;
+	int sad_count;
+
+	if (!dig || !dig->afmt || !dig->afmt->pin)
+		return;
+
+	offset = dig->afmt->pin->offset;
+
+	list_for_each_entry(connector, &encoder->dev->mode_config.connector_list, head) {
+		if (connector->encoder == encoder) {
+			amdgpu_connector = to_amdgpu_connector(connector);
+			break;
+		}
+	}
+
+	if (!amdgpu_connector) {
+		DRM_ERROR("Couldn't find encoder's connector\n");
+		return;
+	}
+
+	sad_count = drm_edid_to_speaker_allocation(amdgpu_connector_edid(connector), &sadb);
+	if (sad_count < 0) {
+		DRM_ERROR("Couldn't read Speaker Allocation Data Block: %d\n", sad_count);
+		sad_count = 0;
+	}
+
+	/* program the speaker allocation */
+	tmp = RREG32_AUDIO_ENDPT(offset, ixAZALIA_F0_CODEC_PIN_CONTROL_CHANNEL_SPEAKER);
+	tmp &= ~(AZALIA_F0_CODEC_PIN_CONTROL_CHANNEL_SPEAKER__DP_CONNECTION_MASK |
+		AZALIA_F0_CODEC_PIN_CONTROL_CHANNEL_SPEAKER__SPEAKER_ALLOCATION_MASK);
+	/* set HDMI mode */
+	tmp |= AZALIA_F0_CODEC_PIN_CONTROL_CHANNEL_SPEAKER__HDMI_CONNECTION_MASK;
+	if (sad_count)
+		tmp |= (sadb[0] << AZALIA_F0_CODEC_PIN_CONTROL_CHANNEL_SPEAKER__SPEAKER_ALLOCATION__SHIFT);
+	else
+		tmp |= (5 << AZALIA_F0_CODEC_PIN_CONTROL_CHANNEL_SPEAKER__SPEAKER_ALLOCATION__SHIFT); /* stereo */
+	WREG32_AUDIO_ENDPT(offset, ixAZALIA_F0_CODEC_PIN_CONTROL_CHANNEL_SPEAKER, tmp);
+
+	kfree(sadb);
+}
+
+static void dce_v8_0_audio_write_sad_regs(struct drm_encoder *encoder)
+{
+	struct amdgpu_device *adev = encoder->dev->dev_private;
+	struct amdgpu_encoder *amdgpu_encoder = to_amdgpu_encoder(encoder);
+	struct amdgpu_encoder_atom_dig *dig = amdgpu_encoder->enc_priv;
+	u32 offset;
+	struct drm_connector *connector;
+	struct amdgpu_connector *amdgpu_connector = NULL;
+	struct cea_sad *sads;
+	int i, sad_count;
+
+	static const u16 eld_reg_to_type[][2] = {
+		{ ixAZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR0, HDMI_AUDIO_CODING_TYPE_PCM },
+		{ ixAZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR1, HDMI_AUDIO_CODING_TYPE_AC3 },
+		{ ixAZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR2, HDMI_AUDIO_CODING_TYPE_MPEG1 },
+		{ ixAZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR3, HDMI_AUDIO_CODING_TYPE_MP3 },
+		{ ixAZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR4, HDMI_AUDIO_CODING_TYPE_MPEG2 },
+		{ ixAZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR5, HDMI_AUDIO_CODING_TYPE_AAC_LC },
+		{ ixAZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR6, HDMI_AUDIO_CODING_TYPE_DTS },
+		{ ixAZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR7, HDMI_AUDIO_CODING_TYPE_ATRAC },
+		{ ixAZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR9, HDMI_AUDIO_CODING_TYPE_EAC3 },
+		{ ixAZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR10, HDMI_AUDIO_CODING_TYPE_DTS_HD },
+		{ ixAZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR11, HDMI_AUDIO_CODING_TYPE_MLP },
+		{ ixAZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR13, HDMI_AUDIO_CODING_TYPE_WMA_PRO },
+	};
+
+	if (!dig || !dig->afmt || !dig->afmt->pin)
+		return;
+
+	offset = dig->afmt->pin->offset;
+
+	list_for_each_entry(connector, &encoder->dev->mode_config.connector_list, head) {
+		if (connector->encoder == encoder) {
+			amdgpu_connector = to_amdgpu_connector(connector);
+			break;
+		}
+	}
+
+	if (!amdgpu_connector) {
+		DRM_ERROR("Couldn't find encoder's connector\n");
+		return;
+	}
+
+	sad_count = drm_edid_to_sad(amdgpu_connector_edid(connector), &sads);
+	if (sad_count <= 0) {
+		DRM_ERROR("Couldn't read SADs: %d\n", sad_count);
+		return;
+	}
+	BUG_ON(!sads);
+
+	for (i = 0; i < ARRAY_SIZE(eld_reg_to_type); i++) {
+		u32 value = 0;
+		u8 stereo_freqs = 0;
+		int max_channels = -1;
+		int j;
+
+		for (j = 0; j < sad_count; j++) {
+			struct cea_sad *sad = &sads[j];
+
+			if (sad->format == eld_reg_to_type[i][1]) {
+				if (sad->channels > max_channels) {
+				value = (sad->channels <<
+				 AZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR0__MAX_CHANNELS__SHIFT) |
+				(sad->byte2 <<
+				 AZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR0__DESCRIPTOR_BYTE_2__SHIFT) |
+				(sad->freq <<
+				 AZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR0__SUPPORTED_FREQUENCIES__SHIFT);
+				max_channels = sad->channels;
+				}
+
+				if (sad->format == HDMI_AUDIO_CODING_TYPE_PCM)
+					stereo_freqs |= sad->freq;
+				else
+					break;
+			}
+		}
+
+		value |= (stereo_freqs <<
+			AZALIA_F0_CODEC_PIN_CONTROL_AUDIO_DESCRIPTOR0__SUPPORTED_FREQUENCIES_STEREO__SHIFT);
+
+		WREG32_AUDIO_ENDPT(offset, eld_reg_to_type[i][0], value);
+	}
+
+	kfree(sads);
+}
+
+static void dce_v8_0_audio_enable(struct amdgpu_device *adev,
+				  struct amdgpu_audio_pin *pin,
+				  bool enable)
+{
+	if (!pin)
+		return;
+
+	WREG32_AUDIO_ENDPT(pin->offset, ixAZALIA_F0_CODEC_PIN_CONTROL_HOT_PLUG_CONTROL,
+		enable ? AZALIA_F0_CODEC_PIN_CONTROL_HOT_PLUG_CONTROL__AUDIO_ENABLED_MASK : 0);
+}
+
+static const u32 pin_offsets[7] =
+{
+	(0x1780 - 0x1780),
+	(0x1786 - 0x1780),
+	(0x178c - 0x1780),
+	(0x1792 - 0x1780),
+	(0x1798 - 0x1780),
+	(0x179d - 0x1780),
+	(0x17a4 - 0x1780),
+};
+
+static int dce_v8_0_audio_init(struct amdgpu_device *adev)
+{
+	int i;
+
+	if (!amdgpu_audio)
+		return 0;
+
+	adev->mode_info.audio.enabled = true;
+
+	if (adev->asic_type == CHIP_KAVERI) /* KV: 4 streams, 7 endpoints */
+		adev->mode_info.audio.num_pins = 7;
+	else if ((adev->asic_type == CHIP_KABINI) ||
+		 (adev->asic_type == CHIP_MULLINS)) /* KB/ML: 2 streams, 3 endpoints */
+		adev->mode_info.audio.num_pins = 3;
+	else if ((adev->asic_type == CHIP_BONAIRE) ||
+		 (adev->asic_type == CHIP_HAWAII))/* BN/HW: 6 streams, 7 endpoints */
+		adev->mode_info.audio.num_pins = 7;
+	else
+		adev->mode_info.audio.num_pins = 3;
+
+	for (i = 0; i < adev->mode_info.audio.num_pins; i++) {
+		adev->mode_info.audio.pin[i].channels = -1;
+		adev->mode_info.audio.pin[i].rate = -1;
+		adev->mode_info.audio.pin[i].bits_per_sample = -1;
+		adev->mode_info.audio.pin[i].status_bits = 0;
+		adev->mode_info.audio.pin[i].category_code = 0;
+		adev->mode_info.audio.pin[i].connected = false;
+		adev->mode_info.audio.pin[i].offset = pin_offsets[i];
+		adev->mode_info.audio.pin[i].id = i;
+		/* disable audio.  it will be set up later */
+		/* XXX remove once we switch to ip funcs */
+		dce_v8_0_audio_enable(adev, &adev->mode_info.audio.pin[i], false);
+	}
+
+	return 0;
+}
+
+static void dce_v8_0_audio_fini(struct amdgpu_device *adev)
+{
+	int i;
+
+	if (!adev->mode_info.audio.enabled)
+		return;
+
+	for (i = 0; i < adev->mode_info.audio.num_pins; i++)
+		dce_v8_0_audio_enable(adev, &adev->mode_info.audio.pin[i], false);
+
+	adev->mode_info.audio.enabled = false;
+}
+
+/*
+ * update the N and CTS parameters for a given pixel clock rate
+ */
+static void dce_v8_0_afmt_update_ACR(struct drm_encoder *encoder, uint32_t clock)
+{
+	struct drm_device *dev = encoder->dev;
+	struct amdgpu_device *adev = dev->dev_private;
+	struct amdgpu_afmt_acr acr = amdgpu_afmt_acr(clock);
+	struct amdgpu_encoder *amdgpu_encoder = to_amdgpu_encoder(encoder);
+	struct amdgpu_encoder_atom_dig *dig = amdgpu_encoder->enc_priv;
+	uint32_t offset = dig->afmt->offset;
+
+	WREG32(mmHDMI_ACR_32_0 + offset, (acr.cts_32khz << HDMI_ACR_44_0__HDMI_ACR_CTS_44__SHIFT));
+	WREG32(mmHDMI_ACR_32_1 + offset, acr.n_32khz);
+
+	WREG32(mmHDMI_ACR_44_0 + offset, (acr.cts_44_1khz << HDMI_ACR_44_0__HDMI_ACR_CTS_44__SHIFT));
+	WREG32(mmHDMI_ACR_44_1 + offset, acr.n_44_1khz);
+
+	WREG32(mmHDMI_ACR_48_0 + offset, (acr.cts_48khz << HDMI_ACR_48_0__HDMI_ACR_CTS_48__SHIFT));
+	WREG32(mmHDMI_ACR_48_1 + offset, acr.n_48khz);
+}
+
+/*
+ * build a HDMI Video Info Frame
+ */
+static void dce_v8_0_afmt_update_avi_infoframe(struct drm_encoder *encoder,
+					       void *buffer, size_t size)
+{
+	struct drm_device *dev = encoder->dev;
+	struct amdgpu_device *adev = dev->dev_private;
+	struct amdgpu_encoder *amdgpu_encoder = to_amdgpu_encoder(encoder);
+	struct amdgpu_encoder_atom_dig *dig = amdgpu_encoder->enc_priv;
+	uint32_t offset = dig->afmt->offset;
+	uint8_t *frame = buffer + 3;
+	uint8_t *header = buffer;
+
+	WREG32(mmAFMT_AVI_INFO0 + offset,
+		frame[0x0] | (frame[0x1] << 8) | (frame[0x2] << 16) | (frame[0x3] << 24));
+	WREG32(mmAFMT_AVI_INFO1 + offset,
+		frame[0x4] | (frame[0x5] << 8) | (frame[0x6] << 16) | (frame[0x7] << 24));
+	WREG32(mmAFMT_AVI_INFO2 + offset,
+		frame[0x8] | (frame[0x9] << 8) | (frame[0xA] << 16) | (frame[0xB] << 24));
+	WREG32(mmAFMT_AVI_INFO3 + offset,
+		frame[0xC] | (frame[0xD] << 8) | (header[1] << 24));
+}
+
+static void dce_v8_0_audio_set_dto(struct drm_encoder *encoder, u32 clock)
+{
+	struct drm_device *dev = encoder->dev;
+	struct amdgpu_device *adev = dev->dev_private;
+	struct amdgpu_encoder *amdgpu_encoder = to_amdgpu_encoder(encoder);
+	struct amdgpu_encoder_atom_dig *dig = amdgpu_encoder->enc_priv;
+	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(encoder->crtc);
+	u32 dto_phase = 24 * 1000;
+	u32 dto_modulo = clock;
+
+	if (!dig || !dig->afmt)
+		return;
+
+	/* XXX two dtos; generally use dto0 for hdmi */
+	/* Express [24MHz / target pixel clock] as an exact rational
+	 * number (coefficient of two integer numbers.  DCCG_AUDIO_DTOx_PHASE
+	 * is the numerator, DCCG_AUDIO_DTOx_MODULE is the denominator
+	 */
+	WREG32(mmDCCG_AUDIO_DTO_SOURCE, (amdgpu_crtc->crtc_id << DCCG_AUDIO_DTO_SOURCE__DCCG_AUDIO_DTO0_SOURCE_SEL__SHIFT));
+	WREG32(mmDCCG_AUDIO_DTO0_PHASE, dto_phase);
+	WREG32(mmDCCG_AUDIO_DTO0_MODULE, dto_modulo);
+}
+
+/*
+ * update the info frames with the data from the current display mode
+ */
+static void dce_v8_0_afmt_setmode(struct drm_encoder *encoder,
+				  struct drm_display_mode *mode)
+{
+	struct drm_device *dev = encoder->dev;
+	struct amdgpu_device *adev = dev->dev_private;
+	struct amdgpu_encoder *amdgpu_encoder = to_amdgpu_encoder(encoder);
+	struct amdgpu_encoder_atom_dig *dig = amdgpu_encoder->enc_priv;
+	struct drm_connector *connector = amdgpu_get_connector_for_encoder(encoder);
+	u8 buffer[HDMI_INFOFRAME_HEADER_SIZE + HDMI_AVI_INFOFRAME_SIZE];
+	struct hdmi_avi_infoframe frame;
+	uint32_t offset, val;
+	ssize_t err;
+	int bpc = 8;
+
+	if (!dig || !dig->afmt)
+		return;
+
+	/* Silent, r600_hdmi_enable will raise WARN for us */
+	if (!dig->afmt->enabled)
+		return;
+	offset = dig->afmt->offset;
+
+	/* hdmi deep color mode general control packets setup, if bpc > 8 */
+	if (encoder->crtc) {
+		struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(encoder->crtc);
+		bpc = amdgpu_crtc->bpc;
+	}
+
+	/* disable audio prior to setting up hw */
+	dig->afmt->pin = dce_v8_0_audio_get_pin(adev);
+	dce_v8_0_audio_enable(adev, dig->afmt->pin, false);
+
+	dce_v8_0_audio_set_dto(encoder, mode->clock);
+
+	WREG32(mmHDMI_VBI_PACKET_CONTROL + offset,
+	       HDMI_VBI_PACKET_CONTROL__HDMI_NULL_SEND_MASK); /* send null packets when required */
+
+	WREG32(mmAFMT_AUDIO_CRC_CONTROL + offset, 0x1000);
+
+	val = RREG32(mmHDMI_CONTROL + offset);
+	val &= ~HDMI_CONTROL__HDMI_DEEP_COLOR_ENABLE_MASK;
+	val &= ~HDMI_CONTROL__HDMI_DEEP_COLOR_DEPTH_MASK;
+
+	switch (bpc) {
+	case 0:
+	case 6:
+	case 8:
+	case 16:
+	default:
+		DRM_DEBUG("%s: Disabling hdmi deep color for %d bpc.\n",
+			  connector->name, bpc);
+		break;
+	case 10:
+		val |= HDMI_CONTROL__HDMI_DEEP_COLOR_ENABLE_MASK;
+		val |= 1 << HDMI_CONTROL__HDMI_DEEP_COLOR_DEPTH__SHIFT;
+		DRM_DEBUG("%s: Enabling hdmi deep color 30 for 10 bpc.\n",
+			  connector->name);
+		break;
+	case 12:
+		val |= HDMI_CONTROL__HDMI_DEEP_COLOR_ENABLE_MASK;
+		val |= 2 << HDMI_CONTROL__HDMI_DEEP_COLOR_DEPTH__SHIFT;
+		DRM_DEBUG("%s: Enabling hdmi deep color 36 for 12 bpc.\n",
+			  connector->name);
+		break;
+	}
+
+	WREG32(mmHDMI_CONTROL + offset, val);
+
+	WREG32(mmHDMI_VBI_PACKET_CONTROL + offset,
+	       HDMI_VBI_PACKET_CONTROL__HDMI_NULL_SEND_MASK | /* send null packets when required */
+	       HDMI_VBI_PACKET_CONTROL__HDMI_GC_SEND_MASK | /* send general control packets */
+	       HDMI_VBI_PACKET_CONTROL__HDMI_GC_CONT_MASK); /* send general control packets every frame */
+
+	WREG32(mmHDMI_INFOFRAME_CONTROL0 + offset,
+	       HDMI_INFOFRAME_CONTROL0__HDMI_AUDIO_INFO_SEND_MASK | /* enable audio info frames (frames won't be set until audio is enabled) */
+	       HDMI_INFOFRAME_CONTROL0__HDMI_AUDIO_INFO_CONT_MASK); /* required for audio info values to be updated */
+
+	WREG32(mmAFMT_INFOFRAME_CONTROL0 + offset,
+	       AFMT_INFOFRAME_CONTROL0__AFMT_AUDIO_INFO_UPDATE_MASK); /* required for audio info values to be updated */
+
+	WREG32(mmHDMI_INFOFRAME_CONTROL1 + offset,
+	       (2 << HDMI_INFOFRAME_CONTROL1__HDMI_AUDIO_INFO_LINE__SHIFT)); /* anything other than 0 */
+
+	WREG32(mmHDMI_GC + offset, 0); /* unset HDMI_GC_AVMUTE */
+
+	WREG32(mmHDMI_AUDIO_PACKET_CONTROL + offset,
+	       (1 << HDMI_AUDIO_PACKET_CONTROL__HDMI_AUDIO_DELAY_EN__SHIFT) | /* set the default audio delay */
+	       (3 << HDMI_AUDIO_PACKET_CONTROL__HDMI_AUDIO_PACKETS_PER_LINE__SHIFT)); /* should be suffient for all audio modes and small enough for all hblanks */
+
+	WREG32(mmAFMT_AUDIO_PACKET_CONTROL + offset,
+	       AFMT_AUDIO_PACKET_CONTROL__AFMT_60958_CS_UPDATE_MASK); /* allow 60958 channel status fields to be updated */
+
+	/* fglrx clears sth in AFMT_AUDIO_PACKET_CONTROL2 here */
+
+	if (bpc > 8)
+		WREG32(mmHDMI_ACR_PACKET_CONTROL + offset,
+		       HDMI_ACR_PACKET_CONTROL__HDMI_ACR_AUTO_SEND_MASK); /* allow hw to sent ACR packets when required */
+	else
+		WREG32(mmHDMI_ACR_PACKET_CONTROL + offset,
+		       HDMI_ACR_PACKET_CONTROL__HDMI_ACR_SOURCE_MASK | /* select SW CTS value */
+		       HDMI_ACR_PACKET_CONTROL__HDMI_ACR_AUTO_SEND_MASK); /* allow hw to sent ACR packets when required */
+
+	dce_v8_0_afmt_update_ACR(encoder, mode->clock);
+
+	WREG32(mmAFMT_60958_0 + offset,
+	       (1 << AFMT_60958_0__AFMT_60958_CS_CHANNEL_NUMBER_L__SHIFT));
+
+	WREG32(mmAFMT_60958_1 + offset,
+	       (2 << AFMT_60958_1__AFMT_60958_CS_CHANNEL_NUMBER_R__SHIFT));
+
+	WREG32(mmAFMT_60958_2 + offset,
+	       (3 << AFMT_60958_2__AFMT_60958_CS_CHANNEL_NUMBER_2__SHIFT) |
+	       (4 << AFMT_60958_2__AFMT_60958_CS_CHANNEL_NUMBER_3__SHIFT) |
+	       (5 << AFMT_60958_2__AFMT_60958_CS_CHANNEL_NUMBER_4__SHIFT) |
+	       (6 << AFMT_60958_2__AFMT_60958_CS_CHANNEL_NUMBER_5__SHIFT) |
+	       (7 << AFMT_60958_2__AFMT_60958_CS_CHANNEL_NUMBER_6__SHIFT) |
+	       (8 << AFMT_60958_2__AFMT_60958_CS_CHANNEL_NUMBER_7__SHIFT));
+
+	dce_v8_0_audio_write_speaker_allocation(encoder);
+
+
+	WREG32(mmAFMT_AUDIO_PACKET_CONTROL2 + offset,
+	       (0xff << AFMT_AUDIO_PACKET_CONTROL2__AFMT_AUDIO_CHANNEL_ENABLE__SHIFT));
+
+	dce_v8_0_afmt_audio_select_pin(encoder);
+	dce_v8_0_audio_write_sad_regs(encoder);
+	dce_v8_0_audio_write_latency_fields(encoder, mode);
+
+	err = drm_hdmi_avi_infoframe_from_display_mode(&frame, mode);
+	if (err < 0) {
+		DRM_ERROR("failed to setup AVI infoframe: %zd\n", err);
+		return;
+	}
+
+	err = hdmi_avi_infoframe_pack(&frame, buffer, sizeof(buffer));
+	if (err < 0) {
+		DRM_ERROR("failed to pack AVI infoframe: %zd\n", err);
+		return;
+	}
+
+	dce_v8_0_afmt_update_avi_infoframe(encoder, buffer, sizeof(buffer));
+
+	WREG32_OR(mmHDMI_INFOFRAME_CONTROL0 + offset,
+		  HDMI_INFOFRAME_CONTROL0__HDMI_AVI_INFO_SEND_MASK | /* enable AVI info frames */
+		  HDMI_INFOFRAME_CONTROL0__HDMI_AVI_INFO_SEND_MASK); /* required for audio info values to be updated */
+
+	WREG32_P(mmHDMI_INFOFRAME_CONTROL1 + offset,
+		 (2 << HDMI_INFOFRAME_CONTROL1__HDMI_AVI_INFO_LINE__SHIFT), /* anything other than 0 */
+		 ~HDMI_INFOFRAME_CONTROL1__HDMI_AVI_INFO_LINE_MASK);
+
+	WREG32_OR(mmAFMT_AUDIO_PACKET_CONTROL + offset,
+		  AFMT_AUDIO_PACKET_CONTROL__AFMT_AUDIO_SAMPLE_SEND_MASK); /* send audio packets */
+
+	/* it's unknown what these bits do excatly, but it's indeed quite useful for debugging */
+	WREG32(mmAFMT_RAMP_CONTROL0 + offset, 0x00FFFFFF);
+	WREG32(mmAFMT_RAMP_CONTROL1 + offset, 0x007FFFFF);
+	WREG32(mmAFMT_RAMP_CONTROL2 + offset, 0x00000001);
+	WREG32(mmAFMT_RAMP_CONTROL3 + offset, 0x00000001);
+
+	/* enable audio after to setting up hw */
+	dce_v8_0_audio_enable(adev, dig->afmt->pin, true);
+}
+
+static void dce_v8_0_afmt_enable(struct drm_encoder *encoder, bool enable)
+{
+	struct drm_device *dev = encoder->dev;
+	struct amdgpu_device *adev = dev->dev_private;
+	struct amdgpu_encoder *amdgpu_encoder = to_amdgpu_encoder(encoder);
+	struct amdgpu_encoder_atom_dig *dig = amdgpu_encoder->enc_priv;
+
+	if (!dig || !dig->afmt)
+		return;
+
+	/* Silent, r600_hdmi_enable will raise WARN for us */
+	if (enable && dig->afmt->enabled)
+		return;
+	if (!enable && !dig->afmt->enabled)
+		return;
+
+	if (!enable && dig->afmt->pin) {
+		dce_v8_0_audio_enable(adev, dig->afmt->pin, false);
+		dig->afmt->pin = NULL;
+	}
+
+	dig->afmt->enabled = enable;
+
+	DRM_DEBUG("%sabling AFMT interface @ 0x%04X for encoder 0x%x\n",
+		  enable ? "En" : "Dis", dig->afmt->offset, amdgpu_encoder->encoder_id);
+}
+
+static void dce_v8_0_afmt_init(struct amdgpu_device *adev)
+{
+	int i;
+
+	for (i = 0; i < adev->mode_info.num_dig; i++)
+		adev->mode_info.afmt[i] = NULL;
+
+	/* DCE8 has audio blocks tied to DIG encoders */
+	for (i = 0; i < adev->mode_info.num_dig; i++) {
+		adev->mode_info.afmt[i] = kzalloc(sizeof(struct amdgpu_afmt), GFP_KERNEL);
+		if (adev->mode_info.afmt[i]) {
+			adev->mode_info.afmt[i]->offset = dig_offsets[i];
+			adev->mode_info.afmt[i]->id = i;
+		}
+	}
+}
+
+static void dce_v8_0_afmt_fini(struct amdgpu_device *adev)
+{
+	int i;
+
+	for (i = 0; i < adev->mode_info.num_dig; i++) {
+		kfree(adev->mode_info.afmt[i]);
+		adev->mode_info.afmt[i] = NULL;
+	}
+}
+
+static const u32 vga_control_regs[6] =
+{
+	mmD1VGA_CONTROL,
+	mmD2VGA_CONTROL,
+	mmD3VGA_CONTROL,
+	mmD4VGA_CONTROL,
+	mmD5VGA_CONTROL,
+	mmD6VGA_CONTROL,
+};
+
+static void dce_v8_0_vga_enable(struct drm_crtc *crtc, bool enable)
+{
+	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
+	struct drm_device *dev = crtc->dev;
+	struct amdgpu_device *adev = dev->dev_private;
+	u32 vga_control;
+
+	vga_control = RREG32(vga_control_regs[amdgpu_crtc->crtc_id]) & ~1;
+	if (enable)
+		WREG32(vga_control_regs[amdgpu_crtc->crtc_id], vga_control | 1);
+	else
+		WREG32(vga_control_regs[amdgpu_crtc->crtc_id], vga_control);
+}
+
+static void dce_v8_0_grph_enable(struct drm_crtc *crtc, bool enable)
+{
+	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
+	struct drm_device *dev = crtc->dev;
+	struct amdgpu_device *adev = dev->dev_private;
+
+	if (enable)
+		WREG32(mmGRPH_ENABLE + amdgpu_crtc->crtc_offset, 1);
+	else
+		WREG32(mmGRPH_ENABLE + amdgpu_crtc->crtc_offset, 0);
+}
+
+static void dce_v8_0_tiling_fields(uint64_t tiling_flags, unsigned *bankw,
+				   unsigned *bankh, unsigned *mtaspect,
+				   unsigned *tile_split)
+{
+	*bankw = (tiling_flags >> AMDGPU_TILING_EG_BANKW_SHIFT) & AMDGPU_TILING_EG_BANKW_MASK;
+	*bankh = (tiling_flags >> AMDGPU_TILING_EG_BANKH_SHIFT) & AMDGPU_TILING_EG_BANKH_MASK;
+	*mtaspect = (tiling_flags >> AMDGPU_TILING_EG_MACRO_TILE_ASPECT_SHIFT) & AMDGPU_TILING_EG_MACRO_TILE_ASPECT_MASK;
+	*tile_split = (tiling_flags >> AMDGPU_TILING_EG_TILE_SPLIT_SHIFT) & AMDGPU_TILING_EG_TILE_SPLIT_MASK;
+	switch (*bankw) {
+	default:
+	case 1:
+		*bankw = ADDR_SURF_BANK_WIDTH_1;
+		break;
+	case 2:
+		*bankw = ADDR_SURF_BANK_WIDTH_2;
+		break;
+	case 4:
+		*bankw = ADDR_SURF_BANK_WIDTH_4;
+		break;
+	case 8:
+		*bankw = ADDR_SURF_BANK_WIDTH_8;
+		break;
+	}
+	switch (*bankh) {
+	default:
+	case 1:
+		*bankh = ADDR_SURF_BANK_HEIGHT_1;
+		break;
+	case 2:
+		*bankh = ADDR_SURF_BANK_HEIGHT_2;
+		break;
+	case 4:
+		*bankh = ADDR_SURF_BANK_HEIGHT_4;
+		break;
+	case 8:
+		*bankh = ADDR_SURF_BANK_HEIGHT_8;
+		break;
+	}
+	switch (*mtaspect) {
+	default:
+	case 1:
+		*mtaspect = ADDR_SURF_MACRO_TILE_ASPECT_1;
+		break;
+	case 2:
+		*mtaspect = ADDR_SURF_MACRO_TILE_ASPECT_2;
+		break;
+	case 4:
+		*mtaspect = ADDR_SURF_MACRO_TILE_ASPECT_4;
+		break;
+	case 8:
+		*mtaspect = ADDR_SURF_MACRO_TILE_ASPECT_8;
+		break;
+	}
+}
+
+static int dce_v8_0_crtc_do_set_base(struct drm_crtc *crtc,
+				     struct drm_framebuffer *fb,
+				     int x, int y, int atomic)
+{
+	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
+	struct drm_device *dev = crtc->dev;
+	struct amdgpu_device *adev = dev->dev_private;
+	struct amdgpu_framebuffer *amdgpu_fb;
+	struct drm_framebuffer *target_fb;
+	struct drm_gem_object *obj;
+	struct amdgpu_bo *rbo;
+	uint64_t fb_location, tiling_flags;
+	uint32_t fb_format, fb_pitch_pixels;
+	unsigned bankw, bankh, mtaspect, tile_split;
+	u32 fb_swap = (GRPH_ENDIAN_NONE << GRPH_SWAP_CNTL__GRPH_ENDIAN_SWAP__SHIFT);
+	u32 pipe_config = (adev->gfx.config.tile_mode_array[10] >> 6) & 0x1f;
+	u32 tmp, viewport_w, viewport_h;
+	int r;
+	bool bypass_lut = false;
+
+	/* no fb bound */
+	if (!atomic && !crtc->primary->fb) {
+		DRM_DEBUG_KMS("No FB bound\n");
+		return 0;
+	}
+
+	if (atomic) {
+		amdgpu_fb = to_amdgpu_framebuffer(fb);
+		target_fb = fb;
+	}
+	else {
+		amdgpu_fb = to_amdgpu_framebuffer(crtc->primary->fb);
+		target_fb = crtc->primary->fb;
+	}
+
+	/* If atomic, assume fb object is pinned & idle & fenced and
+	 * just update base pointers
+	 */
+	obj = amdgpu_fb->obj;
+	rbo = gem_to_amdgpu_bo(obj);
+	r = amdgpu_bo_reserve(rbo, false);
+	if (unlikely(r != 0))
+		return r;
+
+	if (atomic)
+		fb_location = amdgpu_bo_gpu_offset(rbo);
+	else {
+		r = amdgpu_bo_pin(rbo, AMDGPU_GEM_DOMAIN_VRAM, &fb_location);
+		if (unlikely(r != 0)) {
+			amdgpu_bo_unreserve(rbo);
+			return -EINVAL;
+		}
+	}
+
+	amdgpu_bo_get_tiling_flags(rbo, &tiling_flags);
+	amdgpu_bo_unreserve(rbo);
+
+	switch (target_fb->pixel_format) {
+	case DRM_FORMAT_C8:
+		fb_format = ((GRPH_DEPTH_8BPP << GRPH_CONTROL__GRPH_DEPTH__SHIFT) |
+			     (GRPH_FORMAT_INDEXED << GRPH_CONTROL__GRPH_FORMAT__SHIFT));
+		break;
+	case DRM_FORMAT_XRGB4444:
+	case DRM_FORMAT_ARGB4444:
+		fb_format = ((GRPH_DEPTH_16BPP << GRPH_CONTROL__GRPH_DEPTH__SHIFT) |
+			     (GRPH_FORMAT_ARGB1555 << GRPH_CONTROL__GRPH_FORMAT__SHIFT));
+#ifdef __BIG_ENDIAN
+		fb_swap = (GRPH_ENDIAN_8IN16 << GRPH_SWAP_CNTL__GRPH_ENDIAN_SWAP__SHIFT);
+#endif
+		break;
+	case DRM_FORMAT_XRGB1555:
+	case DRM_FORMAT_ARGB1555:
+		fb_format = ((GRPH_DEPTH_16BPP << GRPH_CONTROL__GRPH_DEPTH__SHIFT) |
+			     (GRPH_FORMAT_ARGB1555 << GRPH_CONTROL__GRPH_FORMAT__SHIFT));
+#ifdef __BIG_ENDIAN
+		fb_swap = (GRPH_ENDIAN_8IN16 << GRPH_SWAP_CNTL__GRPH_ENDIAN_SWAP__SHIFT);
+#endif
+		break;
+	case DRM_FORMAT_BGRX5551:
+	case DRM_FORMAT_BGRA5551:
+		fb_format = ((GRPH_DEPTH_16BPP << GRPH_CONTROL__GRPH_DEPTH__SHIFT) |
+			     (GRPH_FORMAT_BGRA5551 << GRPH_CONTROL__GRPH_FORMAT__SHIFT));
+#ifdef __BIG_ENDIAN
+		fb_swap = (GRPH_ENDIAN_8IN16 << GRPH_SWAP_CNTL__GRPH_ENDIAN_SWAP__SHIFT);
+#endif
+		break;
+	case DRM_FORMAT_RGB565:
+		fb_format = ((GRPH_DEPTH_16BPP << GRPH_CONTROL__GRPH_DEPTH__SHIFT) |
+			     (GRPH_FORMAT_ARGB565 << GRPH_CONTROL__GRPH_FORMAT__SHIFT));
+#ifdef __BIG_ENDIAN
+		fb_swap = (GRPH_ENDIAN_8IN16 << GRPH_SWAP_CNTL__GRPH_ENDIAN_SWAP__SHIFT);
+#endif
+		break;
+	case DRM_FORMAT_XRGB8888:
+	case DRM_FORMAT_ARGB8888:
+		fb_format = ((GRPH_DEPTH_32BPP << GRPH_CONTROL__GRPH_DEPTH__SHIFT) |
+			     (GRPH_FORMAT_ARGB8888 << GRPH_CONTROL__GRPH_FORMAT__SHIFT));
+#ifdef __BIG_ENDIAN
+		fb_swap = (GRPH_ENDIAN_8IN32 << GRPH_SWAP_CNTL__GRPH_ENDIAN_SWAP__SHIFT);
+#endif
+		break;
+	case DRM_FORMAT_XRGB2101010:
+	case DRM_FORMAT_ARGB2101010:
+		fb_format = ((GRPH_DEPTH_32BPP << GRPH_CONTROL__GRPH_DEPTH__SHIFT) |
+			     (GRPH_FORMAT_ARGB2101010 << GRPH_CONTROL__GRPH_FORMAT__SHIFT));
+#ifdef __BIG_ENDIAN
+		fb_swap = (GRPH_ENDIAN_8IN32 << GRPH_SWAP_CNTL__GRPH_ENDIAN_SWAP__SHIFT);
+#endif
+		/* Greater 8 bpc fb needs to bypass hw-lut to retain precision */
+		bypass_lut = true;
+		break;
+	case DRM_FORMAT_BGRX1010102:
+	case DRM_FORMAT_BGRA1010102:
+		fb_format = ((GRPH_DEPTH_32BPP << GRPH_CONTROL__GRPH_DEPTH__SHIFT) |
+			     (GRPH_FORMAT_BGRA1010102 << GRPH_CONTROL__GRPH_FORMAT__SHIFT));
+#ifdef __BIG_ENDIAN
+		fb_swap = (GRPH_ENDIAN_8IN32 << GRPH_SWAP_CNTL__GRPH_ENDIAN_SWAP__SHIFT);
+#endif
+		/* Greater 8 bpc fb needs to bypass hw-lut to retain precision */
+		bypass_lut = true;
+		break;
+	default:
+		DRM_ERROR("Unsupported screen format %s\n",
+			  drm_get_format_name(target_fb->pixel_format));
+		return -EINVAL;
+	}
+
+	if (tiling_flags & AMDGPU_TILING_MACRO) {
+		unsigned tileb, index, num_banks, tile_split_bytes;
+
+		dce_v8_0_tiling_fields(tiling_flags, &bankw, &bankh, &mtaspect, &tile_split);
+		/* Set NUM_BANKS. */
+		/* Calculate the macrotile mode index. */
+		tile_split_bytes = 64 << tile_split;
+		tileb = 8 * 8 * target_fb->bits_per_pixel / 8;
+		tileb = min(tile_split_bytes, tileb);
+
+		for (index = 0; tileb > 64; index++) {
+			tileb >>= 1;
+		}
+
+		if (index >= 16) {
+			DRM_ERROR("Wrong screen bpp (%u) or tile split (%u)\n",
+				  target_fb->bits_per_pixel, tile_split);
+			return -EINVAL;
+		}
+
+		num_banks = (adev->gfx.config.macrotile_mode_array[index] >> 6) & 0x3;
+		fb_format |= (num_banks << GRPH_CONTROL__GRPH_NUM_BANKS__SHIFT);
+		fb_format |= (GRPH_ARRAY_2D_TILED_THIN1 << GRPH_CONTROL__GRPH_ARRAY_MODE__SHIFT);
+		fb_format |= (tile_split << GRPH_CONTROL__GRPH_TILE_SPLIT__SHIFT);
+		fb_format |= (bankw << GRPH_CONTROL__GRPH_BANK_WIDTH__SHIFT);
+		fb_format |= (bankh << GRPH_CONTROL__GRPH_BANK_HEIGHT__SHIFT);
+		fb_format |= (mtaspect << GRPH_CONTROL__GRPH_MACRO_TILE_ASPECT__SHIFT);
+		fb_format |= (DISPLAY_MICRO_TILING << GRPH_CONTROL__GRPH_MICRO_TILE_MODE__SHIFT);
+	} else if (tiling_flags & AMDGPU_TILING_MICRO) {
+		fb_format |= (GRPH_ARRAY_1D_TILED_THIN1 << GRPH_CONTROL__GRPH_ARRAY_MODE__SHIFT);
+	}
+
+	/* Read the pipe config from the 2D TILED SCANOUT mode.
+	 * It should be the same for the other modes too, but not all
+	 * modes set the pipe config field. */
+	fb_format |= (pipe_config << GRPH_CONTROL__GRPH_PIPE_CONFIG__SHIFT);
+
+	dce_v8_0_vga_enable(crtc, false);
+
+	WREG32(mmGRPH_PRIMARY_SURFACE_ADDRESS_HIGH + amdgpu_crtc->crtc_offset,
+	       upper_32_bits(fb_location));
+	WREG32(mmGRPH_SECONDARY_SURFACE_ADDRESS_HIGH + amdgpu_crtc->crtc_offset,
+	       upper_32_bits(fb_location));
+	WREG32(mmGRPH_PRIMARY_SURFACE_ADDRESS + amdgpu_crtc->crtc_offset,
+	       (u32)fb_location & GRPH_PRIMARY_SURFACE_ADDRESS__GRPH_PRIMARY_SURFACE_ADDRESS_MASK);
+	WREG32(mmGRPH_SECONDARY_SURFACE_ADDRESS + amdgpu_crtc->crtc_offset,
+	       (u32) fb_location & GRPH_SECONDARY_SURFACE_ADDRESS__GRPH_SECONDARY_SURFACE_ADDRESS_MASK);
+	WREG32(mmGRPH_CONTROL + amdgpu_crtc->crtc_offset, fb_format);
+	WREG32(mmGRPH_SWAP_CNTL + amdgpu_crtc->crtc_offset, fb_swap);
+
+	/*
+	 * The LUT only has 256 slots for indexing by a 8 bpc fb. Bypass the LUT
+	 * for > 8 bpc scanout to avoid truncation of fb indices to 8 msb's, to
+	 * retain the full precision throughout the pipeline.
+	 */
+	WREG32_P(mmGRPH_LUT_10BIT_BYPASS_CONTROL + amdgpu_crtc->crtc_offset,
+		 (bypass_lut ? LUT_10BIT_BYPASS_EN : 0),
+		 ~LUT_10BIT_BYPASS_EN);
+
+	if (bypass_lut)
+		DRM_DEBUG_KMS("Bypassing hardware LUT due to 10 bit fb scanout.\n");
+
+	WREG32(mmGRPH_SURFACE_OFFSET_X + amdgpu_crtc->crtc_offset, 0);
+	WREG32(mmGRPH_SURFACE_OFFSET_Y + amdgpu_crtc->crtc_offset, 0);
+	WREG32(mmGRPH_X_START + amdgpu_crtc->crtc_offset, 0);
+	WREG32(mmGRPH_Y_START + amdgpu_crtc->crtc_offset, 0);
+	WREG32(mmGRPH_X_END + amdgpu_crtc->crtc_offset, target_fb->width);
+	WREG32(mmGRPH_Y_END + amdgpu_crtc->crtc_offset, target_fb->height);
+
+	fb_pitch_pixels = target_fb->pitches[0] / (target_fb->bits_per_pixel / 8);
+	WREG32(mmGRPH_PITCH + amdgpu_crtc->crtc_offset, fb_pitch_pixels);
+
+	dce_v8_0_grph_enable(crtc, true);
+
+	WREG32(mmLB_DESKTOP_HEIGHT + amdgpu_crtc->crtc_offset,
+	       target_fb->height);
+
+	x &= ~3;
+	y &= ~1;
+	WREG32(mmVIEWPORT_START + amdgpu_crtc->crtc_offset,
+	       (x << 16) | y);
+	viewport_w = crtc->mode.hdisplay;
+	viewport_h = (crtc->mode.vdisplay + 1) & ~1;
+	WREG32(mmVIEWPORT_SIZE + amdgpu_crtc->crtc_offset,
+	       (viewport_w << 16) | viewport_h);
+
+	/* pageflip setup */
+	/* make sure flip is at vb rather than hb */
+	tmp = RREG32(mmGRPH_FLIP_CONTROL + amdgpu_crtc->crtc_offset);
+	tmp &= ~GRPH_FLIP_CONTROL__GRPH_SURFACE_UPDATE_H_RETRACE_EN_MASK;
+	WREG32(mmGRPH_FLIP_CONTROL + amdgpu_crtc->crtc_offset, tmp);
+
+	/* set pageflip to happen only at start of vblank interval (front porch) */
+	WREG32(mmMASTER_UPDATE_MODE + amdgpu_crtc->crtc_offset, 3);
+
+	if (!atomic && fb && fb != crtc->primary->fb) {
+		amdgpu_fb = to_amdgpu_framebuffer(fb);
+		rbo = gem_to_amdgpu_bo(amdgpu_fb->obj);
+		r = amdgpu_bo_reserve(rbo, false);
+		if (unlikely(r != 0))
+			return r;
+		amdgpu_bo_unpin(rbo);
+		amdgpu_bo_unreserve(rbo);
+	}
+
+	/* Bytes per pixel may have changed */
+	dce_v8_0_bandwidth_update(adev);
+
+	return 0;
+}
+
+static void dce_v8_0_set_interleave(struct drm_crtc *crtc,
+				    struct drm_display_mode *mode)
+{
+	struct drm_device *dev = crtc->dev;
+	struct amdgpu_device *adev = dev->dev_private;
+	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
+
+	if (mode->flags & DRM_MODE_FLAG_INTERLACE)
+		WREG32(mmLB_DATA_FORMAT + amdgpu_crtc->crtc_offset,
+		       LB_DATA_FORMAT__INTERLEAVE_EN__SHIFT);
+	else
+		WREG32(mmLB_DATA_FORMAT + amdgpu_crtc->crtc_offset, 0);
+}
+
+static void dce_v8_0_crtc_load_lut(struct drm_crtc *crtc)
+{
+	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
+	struct drm_device *dev = crtc->dev;
+	struct amdgpu_device *adev = dev->dev_private;
+	int i;
+
+	DRM_DEBUG_KMS("%d\n", amdgpu_crtc->crtc_id);
+
+	WREG32(mmINPUT_CSC_CONTROL + amdgpu_crtc->crtc_offset,
+	       ((INPUT_CSC_BYPASS << INPUT_CSC_CONTROL__INPUT_CSC_GRPH_MODE__SHIFT) |
+		(INPUT_CSC_BYPASS << INPUT_CSC_CONTROL__INPUT_CSC_OVL_MODE__SHIFT)));
+	WREG32(mmPRESCALE_GRPH_CONTROL + amdgpu_crtc->crtc_offset,
+	       PRESCALE_GRPH_CONTROL__GRPH_PRESCALE_BYPASS_MASK);
+	WREG32(mmPRESCALE_OVL_CONTROL + amdgpu_crtc->crtc_offset,
+	       PRESCALE_OVL_CONTROL__OVL_PRESCALE_BYPASS_MASK);
+	WREG32(mmINPUT_GAMMA_CONTROL + amdgpu_crtc->crtc_offset,
+	       ((INPUT_GAMMA_USE_LUT << INPUT_GAMMA_CONTROL__GRPH_INPUT_GAMMA_MODE__SHIFT) |
+		(INPUT_GAMMA_USE_LUT << INPUT_GAMMA_CONTROL__OVL_INPUT_GAMMA_MODE__SHIFT)));
+
+	WREG32(mmDC_LUT_CONTROL + amdgpu_crtc->crtc_offset, 0);
+
+	WREG32(mmDC_LUT_BLACK_OFFSET_BLUE + amdgpu_crtc->crtc_offset, 0);
+	WREG32(mmDC_LUT_BLACK_OFFSET_GREEN + amdgpu_crtc->crtc_offset, 0);
+	WREG32(mmDC_LUT_BLACK_OFFSET_RED + amdgpu_crtc->crtc_offset, 0);
+
+	WREG32(mmDC_LUT_WHITE_OFFSET_BLUE + amdgpu_crtc->crtc_offset, 0xffff);
+	WREG32(mmDC_LUT_WHITE_OFFSET_GREEN + amdgpu_crtc->crtc_offset, 0xffff);
+	WREG32(mmDC_LUT_WHITE_OFFSET_RED + amdgpu_crtc->crtc_offset, 0xffff);
+
+	WREG32(mmDC_LUT_RW_MODE + amdgpu_crtc->crtc_offset, 0);
+	WREG32(mmDC_LUT_WRITE_EN_MASK + amdgpu_crtc->crtc_offset, 0x00000007);
+
+	WREG32(mmDC_LUT_RW_INDEX + amdgpu_crtc->crtc_offset, 0);
+	for (i = 0; i < 256; i++) {
+		WREG32(mmDC_LUT_30_COLOR + amdgpu_crtc->crtc_offset,
+		       (amdgpu_crtc->lut_r[i] << 20) |
+		       (amdgpu_crtc->lut_g[i] << 10) |
+		       (amdgpu_crtc->lut_b[i] << 0));
+	}
+
+	WREG32(mmDEGAMMA_CONTROL + amdgpu_crtc->crtc_offset,
+	       ((DEGAMMA_BYPASS << DEGAMMA_CONTROL__GRPH_DEGAMMA_MODE__SHIFT) |
+		(DEGAMMA_BYPASS << DEGAMMA_CONTROL__OVL_DEGAMMA_MODE__SHIFT) |
+		(DEGAMMA_BYPASS << DEGAMMA_CONTROL__CURSOR_DEGAMMA_MODE__SHIFT)));
+	WREG32(mmGAMUT_REMAP_CONTROL + amdgpu_crtc->crtc_offset,
+	       ((GAMUT_REMAP_BYPASS << GAMUT_REMAP_CONTROL__GRPH_GAMUT_REMAP_MODE__SHIFT) |
+		(GAMUT_REMAP_BYPASS << GAMUT_REMAP_CONTROL__OVL_GAMUT_REMAP_MODE__SHIFT)));
+	WREG32(mmREGAMMA_CONTROL + amdgpu_crtc->crtc_offset,
+	       ((REGAMMA_BYPASS << REGAMMA_CONTROL__GRPH_REGAMMA_MODE__SHIFT) |
+		(REGAMMA_BYPASS << REGAMMA_CONTROL__OVL_REGAMMA_MODE__SHIFT)));
+	WREG32(mmOUTPUT_CSC_CONTROL + amdgpu_crtc->crtc_offset,
+	       ((OUTPUT_CSC_BYPASS << OUTPUT_CSC_CONTROL__OUTPUT_CSC_GRPH_MODE__SHIFT) |
+		(OUTPUT_CSC_BYPASS << OUTPUT_CSC_CONTROL__OUTPUT_CSC_OVL_MODE__SHIFT)));
+	/* XXX match this to the depth of the crtc fmt block, move to modeset? */
+	WREG32(0x1a50 + amdgpu_crtc->crtc_offset, 0);
+	/* XXX this only needs to be programmed once per crtc at startup,
+	 * not sure where the best place for it is
+	 */
+	WREG32(mmALPHA_CONTROL + amdgpu_crtc->crtc_offset,
+	       ALPHA_CONTROL__CURSOR_ALPHA_BLND_ENA_MASK);
+}
+
+static int dce_v8_0_pick_dig_encoder(struct drm_encoder *encoder)
+{
+	struct amdgpu_encoder *amdgpu_encoder = to_amdgpu_encoder(encoder);
+	struct amdgpu_encoder_atom_dig *dig = amdgpu_encoder->enc_priv;
+
+	switch (amdgpu_encoder->encoder_id) {
+	case ENCODER_OBJECT_ID_INTERNAL_UNIPHY:
+		if (dig->linkb)
+			return 1;
+		else
+			return 0;
+		break;
+	case ENCODER_OBJECT_ID_INTERNAL_UNIPHY1:
+		if (dig->linkb)
+			return 3;
+		else
+			return 2;
+		break;
+	case ENCODER_OBJECT_ID_INTERNAL_UNIPHY2:
+		if (dig->linkb)
+			return 5;
+		else
+			return 4;
+		break;
+	case ENCODER_OBJECT_ID_INTERNAL_UNIPHY3:
+		return 6;
+		break;
+	default:
+		DRM_ERROR("invalid encoder_id: 0x%x\n", amdgpu_encoder->encoder_id);
+		return 0;
+	}
+}
+
+/**
+ * dce_v8_0_pick_pll - Allocate a PPLL for use by the crtc.
+ *
+ * @crtc: drm crtc
+ *
+ * Returns the PPLL (Pixel PLL) to be used by the crtc.  For DP monitors
+ * a single PPLL can be used for all DP crtcs/encoders.  For non-DP
+ * monitors a dedicated PPLL must be used.  If a particular board has
+ * an external DP PLL, return ATOM_PPLL_INVALID to skip PLL programming
+ * as there is no need to program the PLL itself.  If we are not able to
+ * allocate a PLL, return ATOM_PPLL_INVALID to skip PLL programming to
+ * avoid messing up an existing monitor.
+ *
+ * Asic specific PLL information
+ *
+ * DCE 8.x
+ * KB/KV
+ * - PPLL1, PPLL2 are available for all UNIPHY (both DP and non-DP)
+ * CI
+ * - PPLL0, PPLL1, PPLL2 are available for all UNIPHY (both DP and non-DP) and DAC
+ *
+ */
+static u32 dce_v8_0_pick_pll(struct drm_crtc *crtc)
+{
+	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
+	struct drm_device *dev = crtc->dev;
+	struct amdgpu_device *adev = dev->dev_private;
+	u32 pll_in_use;
+	int pll;
+
+	if (ENCODER_MODE_IS_DP(amdgpu_atombios_encoder_get_encoder_mode(amdgpu_crtc->encoder))) {
+		if (adev->clock.dp_extclk)
+			/* skip PPLL programming if using ext clock */
+			return ATOM_PPLL_INVALID;
+		else {
+			/* use the same PPLL for all DP monitors */
+			pll = amdgpu_pll_get_shared_dp_ppll(crtc);
+			if (pll != ATOM_PPLL_INVALID)
+				return pll;
+		}
+	} else {
+		/* use the same PPLL for all monitors with the same clock */
+		pll = amdgpu_pll_get_shared_nondp_ppll(crtc);
+		if (pll != ATOM_PPLL_INVALID)
+			return pll;
+	}
+	/* otherwise, pick one of the plls */
+	if ((adev->asic_type == CHIP_KABINI) ||
+	    (adev->asic_type == CHIP_MULLINS)) {
+		/* KB/ML has PPLL1 and PPLL2 */
+		pll_in_use = amdgpu_pll_get_use_mask(crtc);
+		if (!(pll_in_use & (1 << ATOM_PPLL2)))
+			return ATOM_PPLL2;
+		if (!(pll_in_use & (1 << ATOM_PPLL1)))
+			return ATOM_PPLL1;
+		DRM_ERROR("unable to allocate a PPLL\n");
+		return ATOM_PPLL_INVALID;
+	} else {
+		/* CI/KV has PPLL0, PPLL1, and PPLL2 */
+		pll_in_use = amdgpu_pll_get_use_mask(crtc);
+		if (!(pll_in_use & (1 << ATOM_PPLL2)))
+			return ATOM_PPLL2;
+		if (!(pll_in_use & (1 << ATOM_PPLL1)))
+			return ATOM_PPLL1;
+		if (!(pll_in_use & (1 << ATOM_PPLL0)))
+			return ATOM_PPLL0;
+		DRM_ERROR("unable to allocate a PPLL\n");
+		return ATOM_PPLL_INVALID;
+	}
+	return ATOM_PPLL_INVALID;
+}
+
+static void dce_v8_0_lock_cursor(struct drm_crtc *crtc, bool lock)
+{
+	struct amdgpu_device *adev = crtc->dev->dev_private;
+	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
+	uint32_t cur_lock;
+
+	cur_lock = RREG32(mmCUR_UPDATE + amdgpu_crtc->crtc_offset);
+	if (lock)
+		cur_lock |= CUR_UPDATE__CURSOR_UPDATE_LOCK_MASK;
+	else
+		cur_lock &= ~CUR_UPDATE__CURSOR_UPDATE_LOCK_MASK;
+	WREG32(mmCUR_UPDATE + amdgpu_crtc->crtc_offset, cur_lock);
+}
+
+static void dce_v8_0_hide_cursor(struct drm_crtc *crtc)
+{
+	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
+	struct amdgpu_device *adev = crtc->dev->dev_private;
+
+	WREG32_IDX(mmCUR_CONTROL + amdgpu_crtc->crtc_offset,
+		   (CURSOR_24_8_PRE_MULT << CUR_CONTROL__CURSOR_MODE__SHIFT) |
+		   (CURSOR_URGENT_1_2 << CUR_CONTROL__CURSOR_URGENT_CONTROL__SHIFT));
+}
+
+static void dce_v8_0_show_cursor(struct drm_crtc *crtc)
+{
+	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
+	struct amdgpu_device *adev = crtc->dev->dev_private;
+
+	WREG32_IDX(mmCUR_CONTROL + amdgpu_crtc->crtc_offset,
+		   CUR_CONTROL__CURSOR_EN_MASK |
+		   (CURSOR_24_8_PRE_MULT << CUR_CONTROL__CURSOR_MODE__SHIFT) |
+		   (CURSOR_URGENT_1_2 << CUR_CONTROL__CURSOR_URGENT_CONTROL__SHIFT));
+}
+
+static void dce_v8_0_set_cursor(struct drm_crtc *crtc, struct drm_gem_object *obj,
+			      uint64_t gpu_addr)
+{
+	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
+	struct amdgpu_device *adev = crtc->dev->dev_private;
+
+	WREG32(mmCUR_SURFACE_ADDRESS_HIGH + amdgpu_crtc->crtc_offset,
+	       upper_32_bits(gpu_addr));
+	WREG32(mmCUR_SURFACE_ADDRESS + amdgpu_crtc->crtc_offset,
+	       gpu_addr & 0xffffffff);
+}
+
+static int dce_v8_0_crtc_cursor_move(struct drm_crtc *crtc,
+				     int x, int y)
+{
+	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
+	struct amdgpu_device *adev = crtc->dev->dev_private;
+	int xorigin = 0, yorigin = 0;
+
+	/* avivo cursor are offset into the total surface */
+	x += crtc->x;
+	y += crtc->y;
+	DRM_DEBUG("x %d y %d c->x %d c->y %d\n", x, y, crtc->x, crtc->y);
+
+	if (x < 0) {
+		xorigin = min(-x, amdgpu_crtc->max_cursor_width - 1);
+		x = 0;
+	}
+	if (y < 0) {
+		yorigin = min(-y, amdgpu_crtc->max_cursor_height - 1);
+		y = 0;
+	}
+
+	dce_v8_0_lock_cursor(crtc, true);
+	WREG32(mmCUR_POSITION + amdgpu_crtc->crtc_offset, (x << 16) | y);
+	WREG32(mmCUR_HOT_SPOT + amdgpu_crtc->crtc_offset, (xorigin << 16) | yorigin);
+	WREG32(mmCUR_SIZE + amdgpu_crtc->crtc_offset,
+	       ((amdgpu_crtc->cursor_width - 1) << 16) | (amdgpu_crtc->cursor_height - 1));
+	dce_v8_0_lock_cursor(crtc, false);
+
+	return 0;
+}
+
+static int dce_v8_0_crtc_cursor_set(struct drm_crtc *crtc,
+				    struct drm_file *file_priv,
+				    uint32_t handle,
+				    uint32_t width,
+				    uint32_t height)
+{
+	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
+	struct drm_gem_object *obj;
+	struct amdgpu_bo *robj;
+	uint64_t gpu_addr;
+	int ret;
+
+	if (!handle) {
+		/* turn off cursor */
+		dce_v8_0_hide_cursor(crtc);
+		obj = NULL;
+		goto unpin;
+	}
+
+	if ((width > amdgpu_crtc->max_cursor_width) ||
+	    (height > amdgpu_crtc->max_cursor_height)) {
+		DRM_ERROR("bad cursor width or height %d x %d\n", width, height);
+		return -EINVAL;
+	}
+
+	obj = drm_gem_object_lookup(crtc->dev, file_priv, handle);
+	if (!obj) {
+		DRM_ERROR("Cannot find cursor object %x for crtc %d\n", handle, amdgpu_crtc->crtc_id);
+		return -ENOENT;
+	}
+
+	robj = gem_to_amdgpu_bo(obj);
+	ret = amdgpu_bo_reserve(robj, false);
+	if (unlikely(ret != 0))
+		goto fail;
+	ret = amdgpu_bo_pin_restricted(robj, AMDGPU_GEM_DOMAIN_VRAM,
+				       0, &gpu_addr);
+	amdgpu_bo_unreserve(robj);
+	if (ret)
+		goto fail;
+
+	amdgpu_crtc->cursor_width = width;
+	amdgpu_crtc->cursor_height = height;
+
+	dce_v8_0_lock_cursor(crtc, true);
+	dce_v8_0_set_cursor(crtc, obj, gpu_addr);
+	dce_v8_0_show_cursor(crtc);
+	dce_v8_0_lock_cursor(crtc, false);
+
+unpin:
+	if (amdgpu_crtc->cursor_bo) {
+		robj = gem_to_amdgpu_bo(amdgpu_crtc->cursor_bo);
+		ret = amdgpu_bo_reserve(robj, false);
+		if (likely(ret == 0)) {
+			amdgpu_bo_unpin(robj);
+			amdgpu_bo_unreserve(robj);
+		}
+		drm_gem_object_unreference_unlocked(amdgpu_crtc->cursor_bo);
+	}
+
+	amdgpu_crtc->cursor_bo = obj;
+	return 0;
+fail:
+	drm_gem_object_unreference_unlocked(obj);
+
+	return ret;
+}
+
+static void dce_v8_0_crtc_gamma_set(struct drm_crtc *crtc, u16 *red, u16 *green,
+				    u16 *blue, uint32_t start, uint32_t size)
+{
+	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
+	int end = (start + size > 256) ? 256 : start + size, i;
+
+	/* userspace palettes are always correct as is */
+	for (i = start; i < end; i++) {
+		amdgpu_crtc->lut_r[i] = red[i] >> 6;
+		amdgpu_crtc->lut_g[i] = green[i] >> 6;
+		amdgpu_crtc->lut_b[i] = blue[i] >> 6;
+	}
+	dce_v8_0_crtc_load_lut(crtc);
+}
+
+static void dce_v8_0_crtc_destroy(struct drm_crtc *crtc)
+{
+	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
+
+	drm_crtc_cleanup(crtc);
+	destroy_workqueue(amdgpu_crtc->pflip_queue);
+	kfree(amdgpu_crtc);
+}
+
+static const struct drm_crtc_funcs dce_v8_0_crtc_funcs = {
+	.cursor_set = dce_v8_0_crtc_cursor_set,
+	.cursor_move = dce_v8_0_crtc_cursor_move,
+	.gamma_set = dce_v8_0_crtc_gamma_set,
+	.set_config = amdgpu_crtc_set_config,
+	.destroy = dce_v8_0_crtc_destroy,
+	.page_flip = amdgpu_crtc_page_flip,
+};
+
+static void dce_v8_0_crtc_dpms(struct drm_crtc *crtc, int mode)
+{
+	struct drm_device *dev = crtc->dev;
+	struct amdgpu_device *adev = dev->dev_private;
+	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
+
+	switch (mode) {
+	case DRM_MODE_DPMS_ON:
+		amdgpu_crtc->enabled = true;
+		amdgpu_atombios_crtc_enable(crtc, ATOM_ENABLE);
+		dce_v8_0_vga_enable(crtc, true);
+		amdgpu_atombios_crtc_blank(crtc, ATOM_DISABLE);
+		dce_v8_0_vga_enable(crtc, false);
+		drm_vblank_post_modeset(dev, amdgpu_crtc->crtc_id);
+		dce_v8_0_crtc_load_lut(crtc);
+		break;
+	case DRM_MODE_DPMS_STANDBY:
+	case DRM_MODE_DPMS_SUSPEND:
+	case DRM_MODE_DPMS_OFF:
+		drm_vblank_pre_modeset(dev, amdgpu_crtc->crtc_id);
+		if (amdgpu_crtc->enabled) {
+			dce_v8_0_vga_enable(crtc, true);
+			amdgpu_atombios_crtc_blank(crtc, ATOM_ENABLE);
+			dce_v8_0_vga_enable(crtc, false);
+		}
+		amdgpu_atombios_crtc_enable(crtc, ATOM_DISABLE);
+		amdgpu_crtc->enabled = false;
+		break;
+	}
+	/* adjust pm to dpms */
+	amdgpu_pm_compute_clocks(adev);
+}
+
+static void dce_v8_0_crtc_prepare(struct drm_crtc *crtc)
+{
+	/* disable crtc pair power gating before programming */
+	amdgpu_atombios_crtc_powergate(crtc, ATOM_DISABLE);
+	amdgpu_atombios_crtc_lock(crtc, ATOM_ENABLE);
+	dce_v8_0_crtc_dpms(crtc, DRM_MODE_DPMS_OFF);
+}
+
+static void dce_v8_0_crtc_commit(struct drm_crtc *crtc)
+{
+	dce_v8_0_crtc_dpms(crtc, DRM_MODE_DPMS_ON);
+	amdgpu_atombios_crtc_lock(crtc, ATOM_DISABLE);
+}
+
+static void dce_v8_0_crtc_disable(struct drm_crtc *crtc)
+{
+	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
+	struct drm_device *dev = crtc->dev;
+	struct amdgpu_device *adev = dev->dev_private;
+	struct amdgpu_atom_ss ss;
+	int i;
+
+	dce_v8_0_crtc_dpms(crtc, DRM_MODE_DPMS_OFF);
+	if (crtc->primary->fb) {
+		int r;
+		struct amdgpu_framebuffer *amdgpu_fb;
+		struct amdgpu_bo *rbo;
+
+		amdgpu_fb = to_amdgpu_framebuffer(crtc->primary->fb);
+		rbo = gem_to_amdgpu_bo(amdgpu_fb->obj);
+		r = amdgpu_bo_reserve(rbo, false);
+		if (unlikely(r))
+			DRM_ERROR("failed to reserve rbo before unpin\n");
+		else {
+			amdgpu_bo_unpin(rbo);
+			amdgpu_bo_unreserve(rbo);
+		}
+	}
+	/* disable the GRPH */
+	dce_v8_0_grph_enable(crtc, false);
+
+	amdgpu_atombios_crtc_powergate(crtc, ATOM_ENABLE);
+
+	for (i = 0; i < adev->mode_info.num_crtc; i++) {
+		if (adev->mode_info.crtcs[i] &&
+		    adev->mode_info.crtcs[i]->enabled &&
+		    i != amdgpu_crtc->crtc_id &&
+		    amdgpu_crtc->pll_id == adev->mode_info.crtcs[i]->pll_id) {
+			/* one other crtc is using this pll don't turn
+			 * off the pll
+			 */
+			goto done;
+		}
+	}
+
+	switch (amdgpu_crtc->pll_id) {
+	case ATOM_PPLL1:
+	case ATOM_PPLL2:
+		/* disable the ppll */
+		amdgpu_atombios_crtc_program_pll(crtc, amdgpu_crtc->crtc_id, amdgpu_crtc->pll_id,
+					  0, 0, ATOM_DISABLE, 0, 0, 0, 0, 0, false, &ss);
+		break;
+	case ATOM_PPLL0:
+		/* disable the ppll */
+		if ((adev->asic_type == CHIP_KAVERI) ||
+		    (adev->asic_type == CHIP_BONAIRE) ||
+		    (adev->asic_type == CHIP_HAWAII))
+			amdgpu_atombios_crtc_program_pll(crtc, amdgpu_crtc->crtc_id, amdgpu_crtc->pll_id,
+						  0, 0, ATOM_DISABLE, 0, 0, 0, 0, 0, false, &ss);
+		break;
+	default:
+		break;
+	}
+done:
+	amdgpu_crtc->pll_id = ATOM_PPLL_INVALID;
+	amdgpu_crtc->adjusted_clock = 0;
+	amdgpu_crtc->encoder = NULL;
+	amdgpu_crtc->connector = NULL;
+}
+
+static int dce_v8_0_crtc_mode_set(struct drm_crtc *crtc,
+				  struct drm_display_mode *mode,
+				  struct drm_display_mode *adjusted_mode,
+				  int x, int y, struct drm_framebuffer *old_fb)
+{
+	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
+
+	if (!amdgpu_crtc->adjusted_clock)
+		return -EINVAL;
+
+	amdgpu_atombios_crtc_set_pll(crtc, adjusted_mode);
+	amdgpu_atombios_crtc_set_dtd_timing(crtc, adjusted_mode);
+	dce_v8_0_crtc_do_set_base(crtc, old_fb, x, y, 0);
+	amdgpu_atombios_crtc_overscan_setup(crtc, mode, adjusted_mode);
+	amdgpu_atombios_crtc_scaler_setup(crtc);
+	/* update the hw version fpr dpm */
+	amdgpu_crtc->hw_mode = *adjusted_mode;
+
+	return 0;
+}
+
+static bool dce_v8_0_crtc_mode_fixup(struct drm_crtc *crtc,
+				     const struct drm_display_mode *mode,
+				     struct drm_display_mode *adjusted_mode)
+{
+	struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
+	struct drm_device *dev = crtc->dev;
+	struct drm_encoder *encoder;
+
+	/* assign the encoder to the amdgpu crtc to avoid repeated lookups later */
+	list_for_each_entry(encoder, &dev->mode_config.encoder_list, head) {
+		if (encoder->crtc == crtc) {
+			amdgpu_crtc->encoder = encoder;
+			amdgpu_crtc->connector = amdgpu_get_connector_for_encoder(encoder);
+			break;
+		}
+	}
+	if ((amdgpu_crtc->encoder == NULL) || (amdgpu_crtc->connector == NULL)) {
+		amdgpu_crtc->encoder = NULL;
+		amdgpu_crtc->connector = NULL;
+		return false;
+	}
+	if (!amdgpu_crtc_scaling_mode_fixup(crtc, mode, adjusted_mode))
+		return false;
+	if (amdgpu_atombios_crtc_prepare_pll(crtc, adjusted_mode))
+		return false;
+	/* pick pll */
+	amdgpu_crtc->pll_id = dce_v8_0_pick_pll(crtc);
+	/* if we can't get a PPLL for a non-DP encoder, fail */
+	if ((amdgpu_crtc->pll_id == ATOM_PPLL_INVALID) &&
+	    !ENCODER_MODE_IS_DP(amdgpu_atombios_encoder_get_encoder_mode(amdgpu_crtc->encoder)))
+		return false;
+
+	return true;
+}
+
+static int dce_v8_0_crtc_set_base(struct drm_crtc *crtc, int x, int y,
+				  struct drm_framebuffer *old_fb)
+{
+	return dce_v8_0_crtc_do_set_base(crtc, old_fb, x, y, 0);
+}
+
+static int dce_v8_0_crtc_set_base_atomic(struct drm_crtc *crtc,
+					 struct drm_framebuffer *fb,
+					 int x, int y, enum mode_set_atomic state)
+{
+       return dce_v8_0_crtc_do_set_base(crtc, fb, x, y, 1);
+}
+
+static const struct drm_crtc_helper_funcs dce_v8_0_crtc_helper_funcs = {
+	.dpms = dce_v8_0_crtc_dpms,
+	.mode_fixup = dce_v8_0_crtc_mode_fixup,
+	.mode_set = dce_v8_0_crtc_mode_set,
+	.mode_set_base = dce_v8_0_crtc_set_base,
+	.mode_set_base_atomic = dce_v8_0_crtc_set_base_atomic,
+	.prepare = dce_v8_0_crtc_prepare,
+	.commit = dce_v8_0_crtc_commit,
+	.load_lut = dce_v8_0_crtc_load_lut,
+	.disable = dce_v8_0_crtc_disable,
+};
+
+static int dce_v8_0_crtc_init(struct amdgpu_device *adev, int index)
+{
+	struct amdgpu_crtc *amdgpu_crtc;
+	int i;
+
+	amdgpu_crtc = kzalloc(sizeof(struct amdgpu_crtc) +
+			      (AMDGPUFB_CONN_LIMIT * sizeof(struct drm_connector *)), GFP_KERNEL);
+	if (amdgpu_crtc == NULL)
+		return -ENOMEM;
+
+	drm_crtc_init(adev->ddev, &amdgpu_crtc->base, &dce_v8_0_crtc_funcs);
+
+	drm_mode_crtc_set_gamma_size(&amdgpu_crtc->base, 256);
+	amdgpu_crtc->crtc_id = index;
+	amdgpu_crtc->pflip_queue = create_singlethread_workqueue("amdgpu-pageflip-queue");
+	adev->mode_info.crtcs[index] = amdgpu_crtc;
+
+	amdgpu_crtc->max_cursor_width = CIK_CURSOR_WIDTH;
+	amdgpu_crtc->max_cursor_height = CIK_CURSOR_HEIGHT;
+	adev->ddev->mode_config.cursor_width = amdgpu_crtc->max_cursor_width;
+	adev->ddev->mode_config.cursor_height = amdgpu_crtc->max_cursor_height;
+
+	for (i = 0; i < 256; i++) {
+		amdgpu_crtc->lut_r[i] = i << 2;
+		amdgpu_crtc->lut_g[i] = i << 2;
+		amdgpu_crtc->lut_b[i] = i << 2;
+	}
+
+	amdgpu_crtc->crtc_offset = crtc_offsets[amdgpu_crtc->crtc_id];
+
+	amdgpu_crtc->pll_id = ATOM_PPLL_INVALID;
+	amdgpu_crtc->adjusted_clock = 0;
+	amdgpu_crtc->encoder = NULL;
+	amdgpu_crtc->connector = NULL;
+	drm_crtc_helper_add(&amdgpu_crtc->base, &dce_v8_0_crtc_helper_funcs);
+
+	return 0;
+}
+
+static int dce_v8_0_early_init(struct amdgpu_device *adev)
+{
+	adev->audio_endpt_rreg = &dce_v8_0_audio_endpt_rreg;
+	adev->audio_endpt_wreg = &dce_v8_0_audio_endpt_wreg;
+
+	dce_v8_0_set_display_funcs(adev);
+	dce_v8_0_set_irq_funcs(adev);
+
+	switch (adev->asic_type) {
+	case CHIP_BONAIRE:
+	case CHIP_HAWAII:
+		adev->mode_info.num_crtc = 6;
+		adev->mode_info.num_hpd = 6;
+		adev->mode_info.num_dig = 6;
+		break;
+	case CHIP_KAVERI:
+		adev->mode_info.num_crtc = 4;
+		adev->mode_info.num_hpd = 6;
+		adev->mode_info.num_dig = 7;
+		break;
+	case CHIP_KABINI:
+	case CHIP_MULLINS:
+		adev->mode_info.num_crtc = 2;
+		adev->mode_info.num_hpd = 6;
+		adev->mode_info.num_dig = 6; /* ? */
+		break;
+	default:
+		/* FIXME: not supported yet */
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+static int dce_v8_0_sw_init(struct amdgpu_device *adev)
+{
+	int r, i;
+
+	for (i = 0; i < adev->mode_info.num_crtc; i++) {
+		r = amdgpu_irq_add_id(adev, i + 1, &adev->crtc_irq);
+		if (r)
+			return r;
+	}
+
+	for (i = 8; i < 20; i += 2) {
+		r = amdgpu_irq_add_id(adev, i, &adev->pageflip_irq);
+		if (r)
+			return r;
+	}
+
+	/* HPD hotplug */
+	r = amdgpu_irq_add_id(adev, 42, &adev->hpd_irq);
+	if (r)
+		return r;
+
+	adev->mode_info.mode_config_initialized = true;
+
+	adev->ddev->mode_config.funcs = &amdgpu_mode_funcs;
+
+	adev->ddev->mode_config.max_width = 16384;
+	adev->ddev->mode_config.max_height = 16384;
+
+	adev->ddev->mode_config.preferred_depth = 24;
+	adev->ddev->mode_config.prefer_shadow = 1;
+
+	adev->ddev->mode_config.fb_base = adev->mc.aper_base;
+
+	r = amdgpu_modeset_create_props(adev);
+	if (r)
+		return r;
+
+	adev->ddev->mode_config.max_width = 16384;
+	adev->ddev->mode_config.max_height = 16384;
+
+	/* allocate crtcs */
+	for (i = 0; i < adev->mode_info.num_crtc; i++) {
+		r = dce_v8_0_crtc_init(adev, i);
+		if (r)
+			return r;
+	}
+
+	if (amdgpu_atombios_get_connector_info_from_object_table(adev))
+		amdgpu_print_display_setup(adev->ddev);
+	else
+		return -EINVAL;
+
+	/* setup afmt */
+	dce_v8_0_afmt_init(adev);
+
+	r = dce_v8_0_audio_init(adev);
+	if (r)
+		return r;
+
+	drm_kms_helper_poll_init(adev->ddev);
+
+	return r;
+}
+
+static int dce_v8_0_sw_fini(struct amdgpu_device *adev)
+{
+	kfree(adev->mode_info.bios_hardcoded_edid);
+
+	drm_kms_helper_poll_fini(adev->ddev);
+
+	dce_v8_0_audio_fini(adev);
+
+	dce_v8_0_afmt_fini(adev);
+
+	drm_mode_config_cleanup(adev->ddev);
+	adev->mode_info.mode_config_initialized = false;
+
+	return 0;
+}
+
+static int dce_v8_0_hw_init(struct amdgpu_device *adev)
+{
+	int i;
+
+	/* init dig PHYs, disp eng pll */
+	amdgpu_atombios_encoder_init_dig(adev);
+	amdgpu_atombios_crtc_set_disp_eng_pll(adev, adev->clock.default_dispclk);
+
+	/* initialize hpd */
+	dce_v8_0_hpd_init(adev);
+
+	for (i = 0; i < adev->mode_info.audio.num_pins; i++) {
+		dce_v8_0_audio_enable(adev, &adev->mode_info.audio.pin[i], false);
+	}
+
+	return 0;
+}
+
+static int dce_v8_0_hw_fini(struct amdgpu_device *adev)
+{
+	int i;
+
+	dce_v8_0_hpd_fini(adev);
+
+	for (i = 0; i < adev->mode_info.audio.num_pins; i++) {
+		dce_v8_0_audio_enable(adev, &adev->mode_info.audio.pin[i], false);
+	}
+
+	return 0;
+}
+
+static int dce_v8_0_suspend(struct amdgpu_device *adev)
+{
+	struct drm_connector *connector;
+
+	drm_kms_helper_poll_disable(adev->ddev);
+
+	/* turn off display hw */
+	list_for_each_entry(connector, &adev->ddev->mode_config.connector_list, head) {
+		drm_helper_connector_dpms(connector, DRM_MODE_DPMS_OFF);
+	}
+
+	amdgpu_atombios_scratch_regs_save(adev);
+
+	dce_v8_0_hpd_fini(adev);
+
+	return 0;
+}
+
+static int dce_v8_0_resume(struct amdgpu_device *adev)
+{
+	struct drm_connector *connector;
+
+	amdgpu_atombios_scratch_regs_restore(adev);
+
+	/* init dig PHYs, disp eng pll */
+	amdgpu_atombios_encoder_init_dig(adev);
+	amdgpu_atombios_crtc_set_disp_eng_pll(adev, adev->clock.default_dispclk);
+	/* turn on the BL */
+	if (adev->mode_info.bl_encoder) {
+		u8 bl_level = amdgpu_display_backlight_get_level(adev,
+								  adev->mode_info.bl_encoder);
+		amdgpu_display_backlight_set_level(adev, adev->mode_info.bl_encoder,
+						    bl_level);
+	}
+
+	/* initialize hpd */
+	dce_v8_0_hpd_init(adev);
+
+	/* blat the mode back in */
+	drm_helper_resume_force_mode(adev->ddev);
+	/* turn on display hw */
+	list_for_each_entry(connector, &adev->ddev->mode_config.connector_list, head) {
+		drm_helper_connector_dpms(connector, DRM_MODE_DPMS_ON);
+	}
+
+	drm_kms_helper_poll_enable(adev->ddev);
+
+	return 0;
+}
+
+static bool dce_v8_0_is_idle(struct amdgpu_device *adev)
+{
+	/* XXX todo */
+	return true;
+}
+
+static int dce_v8_0_wait_for_idle(struct amdgpu_device *adev)
+{
+	/* XXX todo */
+	return 0;
+}
+
+static void dce_v8_0_print_status(struct amdgpu_device *adev)
+{
+	dev_info(adev->dev, "DCE 8.x registers\n");
+	/* XXX todo */
+}
+
+static int dce_v8_0_soft_reset(struct amdgpu_device *adev)
+{
+	u32 srbm_soft_reset = 0, tmp;
+
+	if (dce_v8_0_is_display_hung(adev))
+		srbm_soft_reset |= SRBM_SOFT_RESET__SOFT_RESET_DC_MASK;
+
+	if (srbm_soft_reset) {
+		dce_v8_0_print_status(adev);
+
+		tmp = RREG32(mmSRBM_SOFT_RESET);
+		tmp |= srbm_soft_reset;
+		dev_info(adev->dev, "SRBM_SOFT_RESET=0x%08X\n", tmp);
+		WREG32(mmSRBM_SOFT_RESET, tmp);
+		tmp = RREG32(mmSRBM_SOFT_RESET);
+
+		udelay(50);
+
+		tmp &= ~srbm_soft_reset;
+		WREG32(mmSRBM_SOFT_RESET, tmp);
+		tmp = RREG32(mmSRBM_SOFT_RESET);
+
+		/* Wait a little for things to settle down */
+		udelay(50);
+		dce_v8_0_print_status(adev);
+	}
+	return 0;
+}
+
+static void dce_v8_0_set_crtc_vblank_interrupt_state(struct amdgpu_device *adev,
+						     int crtc,
+						     enum amdgpu_interrupt_state state)
+{
+	u32 reg_block, lb_interrupt_mask;
+
+	if (crtc >= adev->mode_info.num_crtc) {
+		DRM_DEBUG("invalid crtc %d\n", crtc);
+		return;
+	}
+
+	switch (crtc) {
+	case 0:
+		reg_block = CRTC0_REGISTER_OFFSET;
+		break;
+	case 1:
+		reg_block = CRTC1_REGISTER_OFFSET;
+		break;
+	case 2:
+		reg_block = CRTC2_REGISTER_OFFSET;
+		break;
+	case 3:
+		reg_block = CRTC3_REGISTER_OFFSET;
+		break;
+	case 4:
+		reg_block = CRTC4_REGISTER_OFFSET;
+		break;
+	case 5:
+		reg_block = CRTC5_REGISTER_OFFSET;
+		break;
+	default:
+		DRM_DEBUG("invalid crtc %d\n", crtc);
+		return;
+	}
+
+	switch (state) {
+	case AMDGPU_IRQ_STATE_DISABLE:
+		lb_interrupt_mask = RREG32(mmLB_INTERRUPT_MASK + reg_block);
+		lb_interrupt_mask &= ~LB_INTERRUPT_MASK__VBLANK_INTERRUPT_MASK_MASK;
+		WREG32(mmLB_INTERRUPT_MASK + reg_block, lb_interrupt_mask);
+		break;
+	case AMDGPU_IRQ_STATE_ENABLE:
+		lb_interrupt_mask = RREG32(mmLB_INTERRUPT_MASK + reg_block);
+		lb_interrupt_mask |= LB_INTERRUPT_MASK__VBLANK_INTERRUPT_MASK_MASK;
+		WREG32(mmLB_INTERRUPT_MASK + reg_block, lb_interrupt_mask);
+		break;
+	default:
+		break;
+	}
+}
+
+static void dce_v8_0_set_crtc_vline_interrupt_state(struct amdgpu_device *adev,
+						    int crtc,
+						    enum amdgpu_interrupt_state state)
+{
+	u32 reg_block, lb_interrupt_mask;
+
+	if (crtc >= adev->mode_info.num_crtc) {
+		DRM_DEBUG("invalid crtc %d\n", crtc);
+		return;
+	}
+
+	switch (crtc) {
+	case 0:
+		reg_block = CRTC0_REGISTER_OFFSET;
+		break;
+	case 1:
+		reg_block = CRTC1_REGISTER_OFFSET;
+		break;
+	case 2:
+		reg_block = CRTC2_REGISTER_OFFSET;
+		break;
+	case 3:
+		reg_block = CRTC3_REGISTER_OFFSET;
+		break;
+	case 4:
+		reg_block = CRTC4_REGISTER_OFFSET;
+		break;
+	case 5:
+		reg_block = CRTC5_REGISTER_OFFSET;
+		break;
+	default:
+		DRM_DEBUG("invalid crtc %d\n", crtc);
+		return;
+	}
+
+	switch (state) {
+	case AMDGPU_IRQ_STATE_DISABLE:
+		lb_interrupt_mask = RREG32(mmLB_INTERRUPT_MASK + reg_block);
+		lb_interrupt_mask &= ~LB_INTERRUPT_MASK__VLINE_INTERRUPT_MASK_MASK;
+		WREG32(mmLB_INTERRUPT_MASK + reg_block, lb_interrupt_mask);
+		break;
+	case AMDGPU_IRQ_STATE_ENABLE:
+		lb_interrupt_mask = RREG32(mmLB_INTERRUPT_MASK + reg_block);
+		lb_interrupt_mask |= LB_INTERRUPT_MASK__VLINE_INTERRUPT_MASK_MASK;
+		WREG32(mmLB_INTERRUPT_MASK + reg_block, lb_interrupt_mask);
+		break;
+	default:
+		break;
+	}
+}
+
+static int dce_v8_0_set_hpd_interrupt_state(struct amdgpu_device *adev,
+					    struct amdgpu_irq_src *src,
+					    unsigned type,
+					    enum amdgpu_interrupt_state state)
+{
+	u32 dc_hpd_int_cntl_reg, dc_hpd_int_cntl;
+
+	switch (type) {
+	case AMDGPU_HPD_1:
+		dc_hpd_int_cntl_reg = mmDC_HPD1_INT_CONTROL;
+		break;
+	case AMDGPU_HPD_2:
+		dc_hpd_int_cntl_reg = mmDC_HPD2_INT_CONTROL;
+		break;
+	case AMDGPU_HPD_3:
+		dc_hpd_int_cntl_reg = mmDC_HPD3_INT_CONTROL;
+		break;
+	case AMDGPU_HPD_4:
+		dc_hpd_int_cntl_reg = mmDC_HPD4_INT_CONTROL;
+		break;
+	case AMDGPU_HPD_5:
+		dc_hpd_int_cntl_reg = mmDC_HPD5_INT_CONTROL;
+		break;
+	case AMDGPU_HPD_6:
+		dc_hpd_int_cntl_reg = mmDC_HPD6_INT_CONTROL;
+		break;
+	default:
+		DRM_DEBUG("invalid hdp %d\n", type);
+		return 0;
+	}
+
+	switch (state) {
+	case AMDGPU_IRQ_STATE_DISABLE:
+		dc_hpd_int_cntl = RREG32(dc_hpd_int_cntl_reg);
+		dc_hpd_int_cntl &= ~DC_HPD1_INT_CONTROL__DC_HPD1_INT_EN_MASK;
+		WREG32(dc_hpd_int_cntl_reg, dc_hpd_int_cntl);
+		break;
+	case AMDGPU_IRQ_STATE_ENABLE:
+		dc_hpd_int_cntl = RREG32(dc_hpd_int_cntl_reg);
+		dc_hpd_int_cntl |= DC_HPD1_INT_CONTROL__DC_HPD1_INT_EN_MASK;
+		WREG32(dc_hpd_int_cntl_reg, dc_hpd_int_cntl);
+		break;
+	default:
+		break;
+	}
+
+	return 0;
+}
+
+static int dce_v8_0_set_crtc_interrupt_state(struct amdgpu_device *adev,
+					     struct amdgpu_irq_src *src,
+					     unsigned type,
+					     enum amdgpu_interrupt_state state)
+{
+	switch (type) {
+	case AMDGPU_CRTC_IRQ_VBLANK1:
+		dce_v8_0_set_crtc_vblank_interrupt_state(adev, 0, state);
+		break;
+	case AMDGPU_CRTC_IRQ_VBLANK2:
+		dce_v8_0_set_crtc_vblank_interrupt_state(adev, 1, state);
+		break;
+	case AMDGPU_CRTC_IRQ_VBLANK3:
+		dce_v8_0_set_crtc_vblank_interrupt_state(adev, 2, state);
+		break;
+	case AMDGPU_CRTC_IRQ_VBLANK4:
+		dce_v8_0_set_crtc_vblank_interrupt_state(adev, 3, state);
+		break;
+	case AMDGPU_CRTC_IRQ_VBLANK5:
+		dce_v8_0_set_crtc_vblank_interrupt_state(adev, 4, state);
+		break;
+	case AMDGPU_CRTC_IRQ_VBLANK6:
+		dce_v8_0_set_crtc_vblank_interrupt_state(adev, 5, state);
+		break;
+	case AMDGPU_CRTC_IRQ_VLINE1:
+		dce_v8_0_set_crtc_vline_interrupt_state(adev, 0, state);
+		break;
+	case AMDGPU_CRTC_IRQ_VLINE2:
+		dce_v8_0_set_crtc_vline_interrupt_state(adev, 1, state);
+		break;
+	case AMDGPU_CRTC_IRQ_VLINE3:
+		dce_v8_0_set_crtc_vline_interrupt_state(adev, 2, state);
+		break;
+	case AMDGPU_CRTC_IRQ_VLINE4:
+		dce_v8_0_set_crtc_vline_interrupt_state(adev, 3, state);
+		break;
+	case AMDGPU_CRTC_IRQ_VLINE5:
+		dce_v8_0_set_crtc_vline_interrupt_state(adev, 4, state);
+		break;
+	case AMDGPU_CRTC_IRQ_VLINE6:
+		dce_v8_0_set_crtc_vline_interrupt_state(adev, 5, state);
+		break;
+	default:
+		break;
+	}
+	return 0;
+}
+
+static int dce_v8_0_crtc_irq(struct amdgpu_device *adev,
+			     struct amdgpu_irq_src *source,
+			     struct amdgpu_iv_entry *entry)
+{
+	unsigned crtc = entry->src_id - 1;
+	uint32_t disp_int = RREG32(interrupt_status_offsets[crtc].reg);
+	unsigned irq_type = amdgpu_crtc_idx_to_irq_type(adev, crtc);
+
+	switch (entry->src_data) {
+	case 0: /* vblank */
+		if (disp_int & interrupt_status_offsets[crtc].vblank) {
+			WREG32(mmLB_VBLANK_STATUS + crtc_offsets[crtc], LB_VBLANK_STATUS__VBLANK_ACK_MASK);
+			if (amdgpu_irq_enabled(adev, source, irq_type)) {
+				drm_handle_vblank(adev->ddev, crtc);
+			}
+			DRM_DEBUG("IH: D%d vblank\n", crtc + 1);
+		}
+		break;
+	case 1: /* vline */
+		if (disp_int & interrupt_status_offsets[crtc].vline) {
+			WREG32(mmLB_VLINE_STATUS + crtc_offsets[crtc], LB_VLINE_STATUS__VLINE_ACK_MASK);
+			DRM_DEBUG("IH: D%d vline\n", crtc + 1);
+		}
+		break;
+	default:
+		DRM_DEBUG("Unhandled interrupt: %d %d\n", entry->src_id, entry->src_data);
+		break;
+	}
+
+	return 0;
+}
+
+static int dce_v8_0_set_pageflip_interrupt_state(struct amdgpu_device *adev,
+						 struct amdgpu_irq_src *src,
+						 unsigned type,
+						 enum amdgpu_interrupt_state state)
+{
+	u32 reg, reg_block;
+	/* now deal with page flip IRQ */
+	switch (type) {
+		case AMDGPU_PAGEFLIP_IRQ_D1:
+			reg_block = CRTC0_REGISTER_OFFSET;
+			break;
+		case AMDGPU_PAGEFLIP_IRQ_D2:
+			reg_block = CRTC1_REGISTER_OFFSET;
+			break;
+		case AMDGPU_PAGEFLIP_IRQ_D3:
+			reg_block = CRTC2_REGISTER_OFFSET;
+			break;
+		case AMDGPU_PAGEFLIP_IRQ_D4:
+			reg_block = CRTC3_REGISTER_OFFSET;
+			break;
+		case AMDGPU_PAGEFLIP_IRQ_D5:
+			reg_block = CRTC4_REGISTER_OFFSET;
+			break;
+		case AMDGPU_PAGEFLIP_IRQ_D6:
+			reg_block = CRTC5_REGISTER_OFFSET;
+			break;
+		default:
+			DRM_ERROR("invalid pageflip crtc %d\n", type);
+			return -EINVAL;
+	}
+
+	reg = RREG32(mmGRPH_INTERRUPT_CONTROL + reg_block);
+	if (state == AMDGPU_IRQ_STATE_DISABLE)
+		WREG32(mmGRPH_INTERRUPT_CONTROL + reg_block, reg & ~GRPH_INTERRUPT_CONTROL__GRPH_PFLIP_INT_MASK_MASK);
+	else
+		WREG32(mmGRPH_INTERRUPT_CONTROL + reg_block, reg | GRPH_INTERRUPT_CONTROL__GRPH_PFLIP_INT_MASK_MASK);
+
+	return 0;
+}
+
+static int dce_v8_0_pageflip_irq(struct amdgpu_device *adev,
+				struct amdgpu_irq_src *source,
+				struct amdgpu_iv_entry *entry)
+{
+	int reg_block;
+	unsigned long flags;
+	unsigned crtc_id;
+	struct amdgpu_crtc *amdgpu_crtc;
+	struct amdgpu_flip_work *works;
+
+	crtc_id = (entry->src_id - 8) >> 1;
+	amdgpu_crtc = adev->mode_info.crtcs[crtc_id];
+
+	/* ack the interrupt */
+	switch(crtc_id){
+		case AMDGPU_PAGEFLIP_IRQ_D1:
+			reg_block = CRTC0_REGISTER_OFFSET;
+			break;
+		case AMDGPU_PAGEFLIP_IRQ_D2:
+			reg_block = CRTC1_REGISTER_OFFSET;
+			break;
+		case AMDGPU_PAGEFLIP_IRQ_D3:
+			reg_block = CRTC2_REGISTER_OFFSET;
+			break;
+		case AMDGPU_PAGEFLIP_IRQ_D4:
+			reg_block = CRTC3_REGISTER_OFFSET;
+			break;
+		case AMDGPU_PAGEFLIP_IRQ_D5:
+			reg_block = CRTC4_REGISTER_OFFSET;
+			break;
+		case AMDGPU_PAGEFLIP_IRQ_D6:
+			reg_block = CRTC5_REGISTER_OFFSET;
+			break;
+		default:
+			DRM_ERROR("invalid pageflip crtc %d\n", crtc_id);
+			return -EINVAL;
+	}
+
+	if (RREG32(mmGRPH_INTERRUPT_STATUS + reg_block) & GRPH_INTERRUPT_STATUS__GRPH_PFLIP_INT_OCCURRED_MASK)
+		WREG32(mmGRPH_INTERRUPT_STATUS + reg_block, GRPH_INTERRUPT_STATUS__GRPH_PFLIP_INT_CLEAR_MASK);
+
+	/* IRQ could occur when in initial stage */
+	if (amdgpu_crtc == NULL)
+		return 0;
+
+	spin_lock_irqsave(&adev->ddev->event_lock, flags);
+	works = amdgpu_crtc->pflip_works;
+	if (amdgpu_crtc->pflip_status != AMDGPU_FLIP_SUBMITTED){
+		DRM_DEBUG_DRIVER("amdgpu_crtc->pflip_status = %d != "
+						"AMDGPU_FLIP_SUBMITTED(%d)\n",
+						amdgpu_crtc->pflip_status,
+						AMDGPU_FLIP_SUBMITTED);
+		spin_unlock_irqrestore(&adev->ddev->event_lock, flags);
+		return 0;
+	}
+
+	/* page flip completed. clean up */
+	amdgpu_crtc->pflip_status = AMDGPU_FLIP_NONE;
+	amdgpu_crtc->pflip_works = NULL;
+
+	/* wakeup usersapce */
+	if (works->event)
+		drm_send_vblank_event(adev->ddev, crtc_id, works->event);
+
+	spin_unlock_irqrestore(&adev->ddev->event_lock, flags);
+
+	drm_vblank_put(adev->ddev, amdgpu_crtc->crtc_id);
+	amdgpu_irq_put(adev, &adev->pageflip_irq, crtc_id);
+	queue_work(amdgpu_crtc->pflip_queue, &works->unpin_work);
+
+	return 0;
+}
+
+static int dce_v8_0_hpd_irq(struct amdgpu_device *adev,
+			    struct amdgpu_irq_src *source,
+			    struct amdgpu_iv_entry *entry)
+{
+	uint32_t disp_int, mask, int_control, tmp;
+	unsigned hpd;
+
+	if (entry->src_data > 6) {
+		DRM_DEBUG("Unhandled interrupt: %d %d\n", entry->src_id, entry->src_data);
+		return 0;
+	}
+
+	hpd = entry->src_data;
+	disp_int = RREG32(interrupt_status_offsets[hpd].reg);
+	mask = interrupt_status_offsets[hpd].hpd;
+	int_control = hpd_int_control_offsets[hpd];
+
+	if (disp_int & mask) {
+		tmp = RREG32(int_control);
+		tmp |= DC_HPD1_INT_CONTROL__DC_HPD1_INT_ACK_MASK;
+		WREG32(int_control, tmp);
+		schedule_work(&adev->hotplug_work);
+		DRM_DEBUG("IH: HPD%d\n", hpd + 1);
+	}
+
+	return 0;
+
+}
+
+static int dce_v8_0_set_clockgating_state(struct amdgpu_device *adev,
+					  enum amdgpu_clockgating_state state)
+{
+	return 0;
+}
+
+static int dce_v8_0_set_powergating_state(struct amdgpu_device *adev,
+					  enum amdgpu_powergating_state state)
+{
+	return 0;
+}
+
+const struct amdgpu_ip_funcs dce_v8_0_ip_funcs = {
+	.early_init = dce_v8_0_early_init,
+	.late_init = NULL,
+	.sw_init = dce_v8_0_sw_init,
+	.sw_fini = dce_v8_0_sw_fini,
+	.hw_init = dce_v8_0_hw_init,
+	.hw_fini = dce_v8_0_hw_fini,
+	.suspend = dce_v8_0_suspend,
+	.resume = dce_v8_0_resume,
+	.is_idle = dce_v8_0_is_idle,
+	.wait_for_idle = dce_v8_0_wait_for_idle,
+	.soft_reset = dce_v8_0_soft_reset,
+	.print_status = dce_v8_0_print_status,
+	.set_clockgating_state = dce_v8_0_set_clockgating_state,
+	.set_powergating_state = dce_v8_0_set_powergating_state,
+};
+
+static void
+dce_v8_0_encoder_mode_set(struct drm_encoder *encoder,
+			  struct drm_display_mode *mode,
+			  struct drm_display_mode *adjusted_mode)
+{
+	struct amdgpu_encoder *amdgpu_encoder = to_amdgpu_encoder(encoder);
+
+	amdgpu_encoder->pixel_clock = adjusted_mode->clock;
+
+	/* need to call this here rather than in prepare() since we need some crtc info */
+	amdgpu_atombios_encoder_dpms(encoder, DRM_MODE_DPMS_OFF);
+
+	/* set scaler clears this on some chips */
+	dce_v8_0_set_interleave(encoder->crtc, mode);
+
+	if (amdgpu_atombios_encoder_get_encoder_mode(encoder) == ATOM_ENCODER_MODE_HDMI) {
+		dce_v8_0_afmt_enable(encoder, true);
+		dce_v8_0_afmt_setmode(encoder, adjusted_mode);
+	}
+}
+
+static void dce_v8_0_encoder_prepare(struct drm_encoder *encoder)
+{
+	struct amdgpu_device *adev = encoder->dev->dev_private;
+	struct amdgpu_encoder *amdgpu_encoder = to_amdgpu_encoder(encoder);
+	struct drm_connector *connector = amdgpu_get_connector_for_encoder(encoder);
+
+	if ((amdgpu_encoder->active_device &
+	     (ATOM_DEVICE_DFP_SUPPORT | ATOM_DEVICE_LCD_SUPPORT)) ||
+	    (amdgpu_encoder_get_dp_bridge_encoder_id(encoder) !=
+	     ENCODER_OBJECT_ID_NONE)) {
+		struct amdgpu_encoder_atom_dig *dig = amdgpu_encoder->enc_priv;
+		if (dig) {
+			dig->dig_encoder = dce_v8_0_pick_dig_encoder(encoder);
+			if (amdgpu_encoder->active_device & ATOM_DEVICE_DFP_SUPPORT)
+				dig->afmt = adev->mode_info.afmt[dig->dig_encoder];
+		}
+	}
+
+	amdgpu_atombios_scratch_regs_lock(adev, true);
+
+	if (connector) {
+		struct amdgpu_connector *amdgpu_connector = to_amdgpu_connector(connector);
+
+		/* select the clock/data port if it uses a router */
+		if (amdgpu_connector->router.cd_valid)
+			amdgpu_i2c_router_select_cd_port(amdgpu_connector);
+
+		/* turn eDP panel on for mode set */
+		if (connector->connector_type == DRM_MODE_CONNECTOR_eDP)
+			amdgpu_atombios_encoder_set_edp_panel_power(connector,
+							     ATOM_TRANSMITTER_ACTION_POWER_ON);
+	}
+
+	/* this is needed for the pll/ss setup to work correctly in some cases */
+	amdgpu_atombios_encoder_set_crtc_source(encoder);
+	/* set up the FMT blocks */
+	dce_v8_0_program_fmt(encoder);
+}
+
+static void dce_v8_0_encoder_commit(struct drm_encoder *encoder)
+{
+	struct drm_device *dev = encoder->dev;
+	struct amdgpu_device *adev = dev->dev_private;
+
+	/* need to call this here as we need the crtc set up */
+	amdgpu_atombios_encoder_dpms(encoder, DRM_MODE_DPMS_ON);
+	amdgpu_atombios_scratch_regs_lock(adev, false);
+}
+
+static void dce_v8_0_encoder_disable(struct drm_encoder *encoder)
+{
+	struct amdgpu_encoder *amdgpu_encoder = to_amdgpu_encoder(encoder);
+	struct amdgpu_encoder_atom_dig *dig;
+
+	amdgpu_atombios_encoder_dpms(encoder, DRM_MODE_DPMS_OFF);
+
+	if (amdgpu_atombios_encoder_is_digital(encoder)) {
+		if (amdgpu_atombios_encoder_get_encoder_mode(encoder) == ATOM_ENCODER_MODE_HDMI)
+			dce_v8_0_afmt_enable(encoder, false);
+		dig = amdgpu_encoder->enc_priv;
+		dig->dig_encoder = -1;
+	}
+	amdgpu_encoder->active_device = 0;
+}
+
+/* these are handled by the primary encoders */
+static void dce_v8_0_ext_prepare(struct drm_encoder *encoder)
+{
+
+}
+
+static void dce_v8_0_ext_commit(struct drm_encoder *encoder)
+{
+
+}
+
+static void
+dce_v8_0_ext_mode_set(struct drm_encoder *encoder,
+		      struct drm_display_mode *mode,
+		      struct drm_display_mode *adjusted_mode)
+{
+
+}
+
+static void dce_v8_0_ext_disable(struct drm_encoder *encoder)
+{
+
+}
+
+static void
+dce_v8_0_ext_dpms(struct drm_encoder *encoder, int mode)
+{
+
+}
+
+static bool dce_v8_0_ext_mode_fixup(struct drm_encoder *encoder,
+				    const struct drm_display_mode *mode,
+				    struct drm_display_mode *adjusted_mode)
+{
+	return true;
+}
+
+static const struct drm_encoder_helper_funcs dce_v8_0_ext_helper_funcs = {
+	.dpms = dce_v8_0_ext_dpms,
+	.mode_fixup = dce_v8_0_ext_mode_fixup,
+	.prepare = dce_v8_0_ext_prepare,
+	.mode_set = dce_v8_0_ext_mode_set,
+	.commit = dce_v8_0_ext_commit,
+	.disable = dce_v8_0_ext_disable,
+	/* no detect for TMDS/LVDS yet */
+};
+
+static const struct drm_encoder_helper_funcs dce_v8_0_dig_helper_funcs = {
+	.dpms = amdgpu_atombios_encoder_dpms,
+	.mode_fixup = amdgpu_atombios_encoder_mode_fixup,
+	.prepare = dce_v8_0_encoder_prepare,
+	.mode_set = dce_v8_0_encoder_mode_set,
+	.commit = dce_v8_0_encoder_commit,
+	.disable = dce_v8_0_encoder_disable,
+	.detect = amdgpu_atombios_encoder_dig_detect,
+};
+
+static const struct drm_encoder_helper_funcs dce_v8_0_dac_helper_funcs = {
+	.dpms = amdgpu_atombios_encoder_dpms,
+	.mode_fixup = amdgpu_atombios_encoder_mode_fixup,
+	.prepare = dce_v8_0_encoder_prepare,
+	.mode_set = dce_v8_0_encoder_mode_set,
+	.commit = dce_v8_0_encoder_commit,
+	.detect = amdgpu_atombios_encoder_dac_detect,
+};
+
+static void dce_v8_0_encoder_destroy(struct drm_encoder *encoder)
+{
+	struct amdgpu_encoder *amdgpu_encoder = to_amdgpu_encoder(encoder);
+	if (amdgpu_encoder->devices & (ATOM_DEVICE_LCD_SUPPORT))
+		amdgpu_atombios_encoder_fini_backlight(amdgpu_encoder);
+	kfree(amdgpu_encoder->enc_priv);
+	drm_encoder_cleanup(encoder);
+	kfree(amdgpu_encoder);
+}
+
+static const struct drm_encoder_funcs dce_v8_0_encoder_funcs = {
+	.destroy = dce_v8_0_encoder_destroy,
+};
+
+static void dce_v8_0_encoder_add(struct amdgpu_device *adev,
+				 uint32_t encoder_enum,
+				 uint32_t supported_device,
+				 u16 caps)
+{
+	struct drm_device *dev = adev->ddev;
+	struct drm_encoder *encoder;
+	struct amdgpu_encoder *amdgpu_encoder;
+
+	/* see if we already added it */
+	list_for_each_entry(encoder, &dev->mode_config.encoder_list, head) {
+		amdgpu_encoder = to_amdgpu_encoder(encoder);
+		if (amdgpu_encoder->encoder_enum == encoder_enum) {
+			amdgpu_encoder->devices |= supported_device;
+			return;
+		}
+
+	}
+
+	/* add a new one */
+	amdgpu_encoder = kzalloc(sizeof(struct amdgpu_encoder), GFP_KERNEL);
+	if (!amdgpu_encoder)
+		return;
+
+	encoder = &amdgpu_encoder->base;
+	switch (adev->mode_info.num_crtc) {
+	case 1:
+		encoder->possible_crtcs = 0x1;
+		break;
+	case 2:
+	default:
+		encoder->possible_crtcs = 0x3;
+		break;
+	case 4:
+		encoder->possible_crtcs = 0xf;
+		break;
+	case 6:
+		encoder->possible_crtcs = 0x3f;
+		break;
+	}
+
+	amdgpu_encoder->enc_priv = NULL;
+
+	amdgpu_encoder->encoder_enum = encoder_enum;
+	amdgpu_encoder->encoder_id = (encoder_enum & OBJECT_ID_MASK) >> OBJECT_ID_SHIFT;
+	amdgpu_encoder->devices = supported_device;
+	amdgpu_encoder->rmx_type = RMX_OFF;
+	amdgpu_encoder->underscan_type = UNDERSCAN_OFF;
+	amdgpu_encoder->is_ext_encoder = false;
+	amdgpu_encoder->caps = caps;
+
+	switch (amdgpu_encoder->encoder_id) {
+	case ENCODER_OBJECT_ID_INTERNAL_KLDSCP_DAC1:
+	case ENCODER_OBJECT_ID_INTERNAL_KLDSCP_DAC2:
+		drm_encoder_init(dev, encoder, &dce_v8_0_encoder_funcs,
+				 DRM_MODE_ENCODER_DAC);
+		drm_encoder_helper_add(encoder, &dce_v8_0_dac_helper_funcs);
+		break;
+	case ENCODER_OBJECT_ID_INTERNAL_KLDSCP_DVO1:
+	case ENCODER_OBJECT_ID_INTERNAL_UNIPHY:
+	case ENCODER_OBJECT_ID_INTERNAL_UNIPHY1:
+	case ENCODER_OBJECT_ID_INTERNAL_UNIPHY2:
+	case ENCODER_OBJECT_ID_INTERNAL_UNIPHY3:
+		if (amdgpu_encoder->devices & (ATOM_DEVICE_LCD_SUPPORT)) {
+			amdgpu_encoder->rmx_type = RMX_FULL;
+			drm_encoder_init(dev, encoder, &dce_v8_0_encoder_funcs,
+					 DRM_MODE_ENCODER_LVDS);
+			amdgpu_encoder->enc_priv = amdgpu_atombios_encoder_get_lcd_info(amdgpu_encoder);
+		} else if (amdgpu_encoder->devices & (ATOM_DEVICE_CRT_SUPPORT)) {
+			drm_encoder_init(dev, encoder, &dce_v8_0_encoder_funcs,
+					 DRM_MODE_ENCODER_DAC);
+			amdgpu_encoder->enc_priv = amdgpu_atombios_encoder_get_dig_info(amdgpu_encoder);
+		} else {
+			drm_encoder_init(dev, encoder, &dce_v8_0_encoder_funcs,
+					 DRM_MODE_ENCODER_TMDS);
+			amdgpu_encoder->enc_priv = amdgpu_atombios_encoder_get_dig_info(amdgpu_encoder);
+		}
+		drm_encoder_helper_add(encoder, &dce_v8_0_dig_helper_funcs);
+		break;
+	case ENCODER_OBJECT_ID_SI170B:
+	case ENCODER_OBJECT_ID_CH7303:
+	case ENCODER_OBJECT_ID_EXTERNAL_SDVOA:
+	case ENCODER_OBJECT_ID_EXTERNAL_SDVOB:
+	case ENCODER_OBJECT_ID_TITFP513:
+	case ENCODER_OBJECT_ID_VT1623:
+	case ENCODER_OBJECT_ID_HDMI_SI1930:
+	case ENCODER_OBJECT_ID_TRAVIS:
+	case ENCODER_OBJECT_ID_NUTMEG:
+		/* these are handled by the primary encoders */
+		amdgpu_encoder->is_ext_encoder = true;
+		if (amdgpu_encoder->devices & (ATOM_DEVICE_LCD_SUPPORT))
+			drm_encoder_init(dev, encoder, &dce_v8_0_encoder_funcs,
+					 DRM_MODE_ENCODER_LVDS);
+		else if (amdgpu_encoder->devices & (ATOM_DEVICE_CRT_SUPPORT))
+			drm_encoder_init(dev, encoder, &dce_v8_0_encoder_funcs,
+					 DRM_MODE_ENCODER_DAC);
+		else
+			drm_encoder_init(dev, encoder, &dce_v8_0_encoder_funcs,
+					 DRM_MODE_ENCODER_TMDS);
+		drm_encoder_helper_add(encoder, &dce_v8_0_ext_helper_funcs);
+		break;
+	}
+}
+
+static const struct amdgpu_display_funcs dce_v8_0_display_funcs = {
+	.set_vga_render_state = &dce_v8_0_set_vga_render_state,
+	.bandwidth_update = &dce_v8_0_bandwidth_update,
+	.vblank_get_counter = &dce_v8_0_vblank_get_counter,
+	.vblank_wait = &dce_v8_0_vblank_wait,
+	.is_display_hung = &dce_v8_0_is_display_hung,
+	.backlight_set_level = &amdgpu_atombios_encoder_set_backlight_level,
+	.backlight_get_level = &amdgpu_atombios_encoder_get_backlight_level,
+	.hpd_sense = &dce_v8_0_hpd_sense,
+	.hpd_set_polarity = &dce_v8_0_hpd_set_polarity,
+	.hpd_get_gpio_reg = &dce_v8_0_hpd_get_gpio_reg,
+	.page_flip = &dce_v8_0_page_flip,
+	.page_flip_get_scanoutpos = &dce_v8_0_crtc_get_scanoutpos,
+	.add_encoder = &dce_v8_0_encoder_add,
+	.add_connector = &amdgpu_connector_add,
+	.stop_mc_access = &dce_v8_0_stop_mc_access,
+	.resume_mc_access = &dce_v8_0_resume_mc_access,
+};
+
+static void dce_v8_0_set_display_funcs(struct amdgpu_device *adev)
+{
+	if (adev->mode_info.funcs == NULL)
+		adev->mode_info.funcs = &dce_v8_0_display_funcs;
+}
+
+static const struct amdgpu_irq_src_funcs dce_v8_0_crtc_irq_funcs = {
+	.set = dce_v8_0_set_crtc_interrupt_state,
+	.process = dce_v8_0_crtc_irq,
+};
+
+static const struct amdgpu_irq_src_funcs dce_v8_0_pageflip_irq_funcs = {
+	.set = dce_v8_0_set_pageflip_interrupt_state,
+	.process = dce_v8_0_pageflip_irq,
+};
+
+static const struct amdgpu_irq_src_funcs dce_v8_0_hpd_irq_funcs = {
+	.set = dce_v8_0_set_hpd_interrupt_state,
+	.process = dce_v8_0_hpd_irq,
+};
+
+static void dce_v8_0_set_irq_funcs(struct amdgpu_device *adev)
+{
+	adev->crtc_irq.num_types = AMDGPU_CRTC_IRQ_LAST;
+	adev->crtc_irq.funcs = &dce_v8_0_crtc_irq_funcs;
+
+	adev->pageflip_irq.num_types = AMDGPU_PAGEFLIP_IRQ_LAST;
+	adev->pageflip_irq.funcs = &dce_v8_0_pageflip_irq_funcs;
+
+	adev->hpd_irq.num_types = AMDGPU_HPD_LAST;
+	adev->hpd_irq.funcs = &dce_v8_0_hpd_irq_funcs;
+}

drivers/gpu/drm/amd/amdgpu/dce_v8_0.h

+/*
+ * Copyright 2014 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __DCE_V8_0_H__
+#define __DCE_V8_0_H__
+
+extern const struct amdgpu_ip_funcs dce_v8_0_ip_funcs;
+
+#endif

drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c

+/*
+ * Copyright 2014 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+#include <linux/firmware.h>
+#include "drmP.h"
+#include "amdgpu.h"
+#include "amdgpu_ih.h"
+#include "amdgpu_gfx.h"
+#include "cikd.h"
+#include "cik.h"
+#include "atom.h"
+#include "amdgpu_ucode.h"
+#include "clearstate_ci.h"
+
+#include "uvd/uvd_4_2_d.h"
+
+#include "dce/dce_8_0_d.h"
+#include "dce/dce_8_0_sh_mask.h"
+
+#include "bif/bif_4_1_d.h"
+#include "bif/bif_4_1_sh_mask.h"
+
+#include "gca/gfx_7_0_d.h"
+#include "gca/gfx_7_2_enum.h"
+#include "gca/gfx_7_2_sh_mask.h"
+
+#include "gmc/gmc_7_0_d.h"
+#include "gmc/gmc_7_0_sh_mask.h"
+
+#include "oss/oss_2_0_d.h"
+#include "oss/oss_2_0_sh_mask.h"
+
+#define GFX7_NUM_GFX_RINGS     1
+#define GFX7_NUM_COMPUTE_RINGS 8
+
+static void gfx_v7_0_set_ring_funcs(struct amdgpu_device *adev);
+static void gfx_v7_0_set_irq_funcs(struct amdgpu_device *adev);
+static void gfx_v7_0_set_gds_init(struct amdgpu_device *adev);
+int gfx_v7_0_get_cu_info(struct amdgpu_device *, struct amdgpu_cu_info *);
+
+MODULE_FIRMWARE("radeon/bonaire_pfp.bin");
+MODULE_FIRMWARE("radeon/bonaire_me.bin");
+MODULE_FIRMWARE("radeon/bonaire_ce.bin");
+MODULE_FIRMWARE("radeon/bonaire_rlc.bin");
+MODULE_FIRMWARE("radeon/bonaire_mec.bin");
+
+MODULE_FIRMWARE("radeon/hawaii_pfp.bin");
+MODULE_FIRMWARE("radeon/hawaii_me.bin");
+MODULE_FIRMWARE("radeon/hawaii_ce.bin");
+MODULE_FIRMWARE("radeon/hawaii_rlc.bin");
+MODULE_FIRMWARE("radeon/hawaii_mec.bin");
+
+MODULE_FIRMWARE("radeon/kaveri_pfp.bin");
+MODULE_FIRMWARE("radeon/kaveri_me.bin");
+MODULE_FIRMWARE("radeon/kaveri_ce.bin");
+MODULE_FIRMWARE("radeon/kaveri_rlc.bin");
+MODULE_FIRMWARE("radeon/kaveri_mec.bin");
+MODULE_FIRMWARE("radeon/kaveri_mec2.bin");
+
+MODULE_FIRMWARE("radeon/kabini_pfp.bin");
+MODULE_FIRMWARE("radeon/kabini_me.bin");
+MODULE_FIRMWARE("radeon/kabini_ce.bin");
+MODULE_FIRMWARE("radeon/kabini_rlc.bin");
+MODULE_FIRMWARE("radeon/kabini_mec.bin");
+
+MODULE_FIRMWARE("radeon/mullins_pfp.bin");
+MODULE_FIRMWARE("radeon/mullins_me.bin");
+MODULE_FIRMWARE("radeon/mullins_ce.bin");
+MODULE_FIRMWARE("radeon/mullins_rlc.bin");
+MODULE_FIRMWARE("radeon/mullins_mec.bin");
+
+static const struct amdgpu_gds_reg_offset amdgpu_gds_reg_offset[] =
+{
+	{mmGDS_VMID0_BASE, mmGDS_VMID0_SIZE, mmGDS_GWS_VMID0, mmGDS_OA_VMID0},
+	{mmGDS_VMID1_BASE, mmGDS_VMID1_SIZE, mmGDS_GWS_VMID1, mmGDS_OA_VMID1},
+	{mmGDS_VMID2_BASE, mmGDS_VMID2_SIZE, mmGDS_GWS_VMID2, mmGDS_OA_VMID2},
+	{mmGDS_VMID3_BASE, mmGDS_VMID3_SIZE, mmGDS_GWS_VMID3, mmGDS_OA_VMID3},
+	{mmGDS_VMID4_BASE, mmGDS_VMID4_SIZE, mmGDS_GWS_VMID4, mmGDS_OA_VMID4},
+	{mmGDS_VMID5_BASE, mmGDS_VMID5_SIZE, mmGDS_GWS_VMID5, mmGDS_OA_VMID5},
+	{mmGDS_VMID6_BASE, mmGDS_VMID6_SIZE, mmGDS_GWS_VMID6, mmGDS_OA_VMID6},
+	{mmGDS_VMID7_BASE, mmGDS_VMID7_SIZE, mmGDS_GWS_VMID7, mmGDS_OA_VMID7},
+	{mmGDS_VMID8_BASE, mmGDS_VMID8_SIZE, mmGDS_GWS_VMID8, mmGDS_OA_VMID8},
+	{mmGDS_VMID9_BASE, mmGDS_VMID9_SIZE, mmGDS_GWS_VMID9, mmGDS_OA_VMID9},
+	{mmGDS_VMID10_BASE, mmGDS_VMID10_SIZE, mmGDS_GWS_VMID10, mmGDS_OA_VMID10},
+	{mmGDS_VMID11_BASE, mmGDS_VMID11_SIZE, mmGDS_GWS_VMID11, mmGDS_OA_VMID11},
+	{mmGDS_VMID12_BASE, mmGDS_VMID12_SIZE, mmGDS_GWS_VMID12, mmGDS_OA_VMID12},
+	{mmGDS_VMID13_BASE, mmGDS_VMID13_SIZE, mmGDS_GWS_VMID13, mmGDS_OA_VMID13},
+	{mmGDS_VMID14_BASE, mmGDS_VMID14_SIZE, mmGDS_GWS_VMID14, mmGDS_OA_VMID14},
+	{mmGDS_VMID15_BASE, mmGDS_VMID15_SIZE, mmGDS_GWS_VMID15, mmGDS_OA_VMID15}
+};
+
+static const u32 spectre_rlc_save_restore_register_list[] =
+{
+	(0x0e00 << 16) | (0xc12c >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc140 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc150 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc15c >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc168 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc170 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc178 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc204 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc2b4 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc2b8 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc2bc >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc2c0 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x8228 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x829c >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x869c >> 2),
+	0x00000000,
+	(0x0600 << 16) | (0x98f4 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x98f8 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x9900 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc260 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x90e8 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x3c000 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x3c00c >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x8c1c >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x9700 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xcd20 >> 2),
+	0x00000000,
+	(0x4e00 << 16) | (0xcd20 >> 2),
+	0x00000000,
+	(0x5e00 << 16) | (0xcd20 >> 2),
+	0x00000000,
+	(0x6e00 << 16) | (0xcd20 >> 2),
+	0x00000000,
+	(0x7e00 << 16) | (0xcd20 >> 2),
+	0x00000000,
+	(0x8e00 << 16) | (0xcd20 >> 2),
+	0x00000000,
+	(0x9e00 << 16) | (0xcd20 >> 2),
+	0x00000000,
+	(0xae00 << 16) | (0xcd20 >> 2),
+	0x00000000,
+	(0xbe00 << 16) | (0xcd20 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x89bc >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x8900 >> 2),
+	0x00000000,
+	0x3,
+	(0x0e00 << 16) | (0xc130 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc134 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc1fc >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc208 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc264 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc268 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc26c >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc270 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc274 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc278 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc27c >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc280 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc284 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc288 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc28c >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc290 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc294 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc298 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc29c >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc2a0 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc2a4 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc2a8 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc2ac  >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc2b0 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x301d0 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x30238 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x30250 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x30254 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x30258 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x3025c >> 2),
+	0x00000000,
+	(0x4e00 << 16) | (0xc900 >> 2),
+	0x00000000,
+	(0x5e00 << 16) | (0xc900 >> 2),
+	0x00000000,
+	(0x6e00 << 16) | (0xc900 >> 2),
+	0x00000000,
+	(0x7e00 << 16) | (0xc900 >> 2),
+	0x00000000,
+	(0x8e00 << 16) | (0xc900 >> 2),
+	0x00000000,
+	(0x9e00 << 16) | (0xc900 >> 2),
+	0x00000000,
+	(0xae00 << 16) | (0xc900 >> 2),
+	0x00000000,
+	(0xbe00 << 16) | (0xc900 >> 2),
+	0x00000000,
+	(0x4e00 << 16) | (0xc904 >> 2),
+	0x00000000,
+	(0x5e00 << 16) | (0xc904 >> 2),
+	0x00000000,
+	(0x6e00 << 16) | (0xc904 >> 2),
+	0x00000000,
+	(0x7e00 << 16) | (0xc904 >> 2),
+	0x00000000,
+	(0x8e00 << 16) | (0xc904 >> 2),
+	0x00000000,
+	(0x9e00 << 16) | (0xc904 >> 2),
+	0x00000000,
+	(0xae00 << 16) | (0xc904 >> 2),
+	0x00000000,
+	(0xbe00 << 16) | (0xc904 >> 2),
+	0x00000000,
+	(0x4e00 << 16) | (0xc908 >> 2),
+	0x00000000,
+	(0x5e00 << 16) | (0xc908 >> 2),
+	0x00000000,
+	(0x6e00 << 16) | (0xc908 >> 2),
+	0x00000000,
+	(0x7e00 << 16) | (0xc908 >> 2),
+	0x00000000,
+	(0x8e00 << 16) | (0xc908 >> 2),
+	0x00000000,
+	(0x9e00 << 16) | (0xc908 >> 2),
+	0x00000000,
+	(0xae00 << 16) | (0xc908 >> 2),
+	0x00000000,
+	(0xbe00 << 16) | (0xc908 >> 2),
+	0x00000000,
+	(0x4e00 << 16) | (0xc90c >> 2),
+	0x00000000,
+	(0x5e00 << 16) | (0xc90c >> 2),
+	0x00000000,
+	(0x6e00 << 16) | (0xc90c >> 2),
+	0x00000000,
+	(0x7e00 << 16) | (0xc90c >> 2),
+	0x00000000,
+	(0x8e00 << 16) | (0xc90c >> 2),
+	0x00000000,
+	(0x9e00 << 16) | (0xc90c >> 2),
+	0x00000000,
+	(0xae00 << 16) | (0xc90c >> 2),
+	0x00000000,
+	(0xbe00 << 16) | (0xc90c >> 2),
+	0x00000000,
+	(0x4e00 << 16) | (0xc910 >> 2),
+	0x00000000,
+	(0x5e00 << 16) | (0xc910 >> 2),
+	0x00000000,
+	(0x6e00 << 16) | (0xc910 >> 2),
+	0x00000000,
+	(0x7e00 << 16) | (0xc910 >> 2),
+	0x00000000,
+	(0x8e00 << 16) | (0xc910 >> 2),
+	0x00000000,
+	(0x9e00 << 16) | (0xc910 >> 2),
+	0x00000000,
+	(0xae00 << 16) | (0xc910 >> 2),
+	0x00000000,
+	(0xbe00 << 16) | (0xc910 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc99c >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x9834 >> 2),
+	0x00000000,
+	(0x0000 << 16) | (0x30f00 >> 2),
+	0x00000000,
+	(0x0001 << 16) | (0x30f00 >> 2),
+	0x00000000,
+	(0x0000 << 16) | (0x30f04 >> 2),
+	0x00000000,
+	(0x0001 << 16) | (0x30f04 >> 2),
+	0x00000000,
+	(0x0000 << 16) | (0x30f08 >> 2),
+	0x00000000,
+	(0x0001 << 16) | (0x30f08 >> 2),
+	0x00000000,
+	(0x0000 << 16) | (0x30f0c >> 2),
+	0x00000000,
+	(0x0001 << 16) | (0x30f0c >> 2),
+	0x00000000,
+	(0x0600 << 16) | (0x9b7c >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x8a14 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x8a18 >> 2),
+	0x00000000,
+	(0x0600 << 16) | (0x30a00 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x8bf0 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x8bcc >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x8b24 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x30a04 >> 2),
+	0x00000000,
+	(0x0600 << 16) | (0x30a10 >> 2),
+	0x00000000,
+	(0x0600 << 16) | (0x30a14 >> 2),
+	0x00000000,
+	(0x0600 << 16) | (0x30a18 >> 2),
+	0x00000000,
+	(0x0600 << 16) | (0x30a2c >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc700 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc704 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc708 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc768 >> 2),
+	0x00000000,
+	(0x0400 << 16) | (0xc770 >> 2),
+	0x00000000,
+	(0x0400 << 16) | (0xc774 >> 2),
+	0x00000000,
+	(0x0400 << 16) | (0xc778 >> 2),
+	0x00000000,
+	(0x0400 << 16) | (0xc77c >> 2),
+	0x00000000,
+	(0x0400 << 16) | (0xc780 >> 2),
+	0x00000000,
+	(0x0400 << 16) | (0xc784 >> 2),
+	0x00000000,
+	(0x0400 << 16) | (0xc788 >> 2),
+	0x00000000,
+	(0x0400 << 16) | (0xc78c >> 2),
+	0x00000000,
+	(0x0400 << 16) | (0xc798 >> 2),
+	0x00000000,
+	(0x0400 << 16) | (0xc79c >> 2),
+	0x00000000,
+	(0x0400 << 16) | (0xc7a0 >> 2),
+	0x00000000,
+	(0x0400 << 16) | (0xc7a4 >> 2),
+	0x00000000,
+	(0x0400 << 16) | (0xc7a8 >> 2),
+	0x00000000,
+	(0x0400 << 16) | (0xc7ac >> 2),
+	0x00000000,
+	(0x0400 << 16) | (0xc7b0 >> 2),
+	0x00000000,
+	(0x0400 << 16) | (0xc7b4 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x9100 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x3c010 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x92a8 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x92ac >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x92b4 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x92b8 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x92bc >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x92c0 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x92c4 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x92c8 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x92cc >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x92d0 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x8c00 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x8c04 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x8c20 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x8c38 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x8c3c >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xae00 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x9604 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xac08 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xac0c >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xac10 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xac14 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xac58 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xac68 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xac6c >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xac70 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xac74 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xac78 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xac7c >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xac80 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xac84 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xac88 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xac8c >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x970c >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x9714 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x9718 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x971c >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x31068 >> 2),
+	0x00000000,
+	(0x4e00 << 16) | (0x31068 >> 2),
+	0x00000000,
+	(0x5e00 << 16) | (0x31068 >> 2),
+	0x00000000,
+	(0x6e00 << 16) | (0x31068 >> 2),
+	0x00000000,
+	(0x7e00 << 16) | (0x31068 >> 2),
+	0x00000000,
+	(0x8e00 << 16) | (0x31068 >> 2),
+	0x00000000,
+	(0x9e00 << 16) | (0x31068 >> 2),
+	0x00000000,
+	(0xae00 << 16) | (0x31068 >> 2),
+	0x00000000,
+	(0xbe00 << 16) | (0x31068 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xcd10 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xcd14 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x88b0 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x88b4 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x88b8 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x88bc >> 2),
+	0x00000000,
+	(0x0400 << 16) | (0x89c0 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x88c4 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x88c8 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x88d0 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x88d4 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x88d8 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x8980 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x30938 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x3093c >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x30940 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x89a0 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x30900 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x30904 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x89b4 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x3c210 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x3c214 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x3c218 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x8904 >> 2),
+	0x00000000,
+	0x5,
+	(0x0e00 << 16) | (0x8c28 >> 2),
+	(0x0e00 << 16) | (0x8c2c >> 2),
+	(0x0e00 << 16) | (0x8c30 >> 2),
+	(0x0e00 << 16) | (0x8c34 >> 2),
+	(0x0e00 << 16) | (0x9600 >> 2),
+};
+
+static const u32 kalindi_rlc_save_restore_register_list[] =
+{
+	(0x0e00 << 16) | (0xc12c >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc140 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc150 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc15c >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc168 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc170 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc204 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc2b4 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc2b8 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc2bc >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc2c0 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x8228 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x829c >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x869c >> 2),
+	0x00000000,
+	(0x0600 << 16) | (0x98f4 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x98f8 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x9900 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc260 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x90e8 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x3c000 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x3c00c >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x8c1c >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x9700 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xcd20 >> 2),
+	0x00000000,
+	(0x4e00 << 16) | (0xcd20 >> 2),
+	0x00000000,
+	(0x5e00 << 16) | (0xcd20 >> 2),
+	0x00000000,
+	(0x6e00 << 16) | (0xcd20 >> 2),
+	0x00000000,
+	(0x7e00 << 16) | (0xcd20 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x89bc >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x8900 >> 2),
+	0x00000000,
+	0x3,
+	(0x0e00 << 16) | (0xc130 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc134 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc1fc >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc208 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc264 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc268 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc26c >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc270 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc274 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc28c >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc290 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc294 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc298 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc2a0 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc2a4 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc2a8 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc2ac >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x301d0 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x30238 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x30250 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x30254 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x30258 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x3025c >> 2),
+	0x00000000,
+	(0x4e00 << 16) | (0xc900 >> 2),
+	0x00000000,
+	(0x5e00 << 16) | (0xc900 >> 2),
+	0x00000000,
+	(0x6e00 << 16) | (0xc900 >> 2),
+	0x00000000,
+	(0x7e00 << 16) | (0xc900 >> 2),
+	0x00000000,
+	(0x4e00 << 16) | (0xc904 >> 2),
+	0x00000000,
+	(0x5e00 << 16) | (0xc904 >> 2),
+	0x00000000,
+	(0x6e00 << 16) | (0xc904 >> 2),
+	0x00000000,
+	(0x7e00 << 16) | (0xc904 >> 2),
+	0x00000000,
+	(0x4e00 << 16) | (0xc908 >> 2),
+	0x00000000,
+	(0x5e00 << 16) | (0xc908 >> 2),
+	0x00000000,
+	(0x6e00 << 16) | (0xc908 >> 2),
+	0x00000000,
+	(0x7e00 << 16) | (0xc908 >> 2),
+	0x00000000,
+	(0x4e00 << 16) | (0xc90c >> 2),
+	0x00000000,
+	(0x5e00 << 16) | (0xc90c >> 2),
+	0x00000000,
+	(0x6e00 << 16) | (0xc90c >> 2),
+	0x00000000,
+	(0x7e00 << 16) | (0xc90c >> 2),
+	0x00000000,
+	(0x4e00 << 16) | (0xc910 >> 2),
+	0x00000000,
+	(0x5e00 << 16) | (0xc910 >> 2),
+	0x00000000,
+	(0x6e00 << 16) | (0xc910 >> 2),
+	0x00000000,
+	(0x7e00 << 16) | (0xc910 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc99c >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x9834 >> 2),
+	0x00000000,
+	(0x0000 << 16) | (0x30f00 >> 2),
+	0x00000000,
+	(0x0000 << 16) | (0x30f04 >> 2),
+	0x00000000,
+	(0x0000 << 16) | (0x30f08 >> 2),
+	0x00000000,
+	(0x0000 << 16) | (0x30f0c >> 2),
+	0x00000000,
+	(0x0600 << 16) | (0x9b7c >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x8a14 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x8a18 >> 2),
+	0x00000000,
+	(0x0600 << 16) | (0x30a00 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x8bf0 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x8bcc >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x8b24 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x30a04 >> 2),
+	0x00000000,
+	(0x0600 << 16) | (0x30a10 >> 2),
+	0x00000000,
+	(0x0600 << 16) | (0x30a14 >> 2),
+	0x00000000,
+	(0x0600 << 16) | (0x30a18 >> 2),
+	0x00000000,
+	(0x0600 << 16) | (0x30a2c >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc700 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc704 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc708 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xc768 >> 2),
+	0x00000000,
+	(0x0400 << 16) | (0xc770 >> 2),
+	0x00000000,
+	(0x0400 << 16) | (0xc774 >> 2),
+	0x00000000,
+	(0x0400 << 16) | (0xc798 >> 2),
+	0x00000000,
+	(0x0400 << 16) | (0xc79c >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x9100 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x3c010 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x8c00 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x8c04 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x8c20 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x8c38 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x8c3c >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xae00 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x9604 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xac08 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xac0c >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xac10 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xac14 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xac58 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xac68 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xac6c >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xac70 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xac74 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xac78 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xac7c >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xac80 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xac84 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xac88 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xac8c >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x970c >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x9714 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x9718 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x971c >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x31068 >> 2),
+	0x00000000,
+	(0x4e00 << 16) | (0x31068 >> 2),
+	0x00000000,
+	(0x5e00 << 16) | (0x31068 >> 2),
+	0x00000000,
+	(0x6e00 << 16) | (0x31068 >> 2),
+	0x00000000,
+	(0x7e00 << 16) | (0x31068 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xcd10 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0xcd14 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x88b0 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x88b4 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x88b8 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x88bc >> 2),
+	0x00000000,
+	(0x0400 << 16) | (0x89c0 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x88c4 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x88c8 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x88d0 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x88d4 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x88d8 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x8980 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x30938 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x3093c >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x30940 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x89a0 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x30900 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x30904 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x89b4 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x3e1fc >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x3c210 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x3c214 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x3c218 >> 2),
+	0x00000000,
+	(0x0e00 << 16) | (0x8904 >> 2),
+	0x00000000,
+	0x5,
+	(0x0e00 << 16) | (0x8c28 >> 2),
+	(0x0e00 << 16) | (0x8c2c >> 2),
+	(0x0e00 << 16) | (0x8c30 >> 2),
+	(0x0e00 << 16) | (0x8c34 >> 2),
+	(0x0e00 << 16) | (0x9600 >> 2),
+};
+
+static u32 gfx_v7_0_get_csb_size(struct amdgpu_device *adev);
+static void gfx_v7_0_get_csb_buffer(struct amdgpu_device *adev, volatile u32 *buffer);
+static void gfx_v7_0_init_cp_pg_table(struct amdgpu_device *adev);
+static void gfx_v7_0_init_pg(struct amdgpu_device *adev);
+
+/*
+ * Core functions
+ */
+/**
+ * gfx_v7_0_init_microcode - load ucode images from disk
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Use the firmware interface to load the ucode images into
+ * the driver (not loaded into hw).
+ * Returns 0 on success, error on failure.
+ */
+static int gfx_v7_0_init_microcode(struct amdgpu_device *adev)
+{
+	const char *chip_name;
+	char fw_name[30];
+	int err;
+
+	DRM_DEBUG("\n");
+
+	switch (adev->asic_type) {
+	case CHIP_BONAIRE:
+		chip_name = "bonaire";
+		break;
+	case CHIP_HAWAII:
+		chip_name = "hawaii";
+		break;
+	case CHIP_KAVERI:
+		chip_name = "kaveri";
+		break;
+	case CHIP_KABINI:
+		chip_name = "kabini";
+		break;
+	case CHIP_MULLINS:
+		chip_name = "mullins";
+		break;
+	default: BUG();
+	}
+
+	snprintf(fw_name, sizeof(fw_name), "radeon/%s_pfp.bin", chip_name);
+	err = request_firmware(&adev->gfx.pfp_fw, fw_name, adev->dev);
+	if (err)
+		goto out;
+	err = amdgpu_ucode_validate(adev->gfx.pfp_fw);
+	if (err)
+		goto out;
+
+	snprintf(fw_name, sizeof(fw_name), "radeon/%s_me.bin", chip_name);
+	err = request_firmware(&adev->gfx.me_fw, fw_name, adev->dev);
+	if (err)
+		goto out;
+	err = amdgpu_ucode_validate(adev->gfx.me_fw);
+	if (err)
+		goto out;
+
+	snprintf(fw_name, sizeof(fw_name), "radeon/%s_ce.bin", chip_name);
+	err = request_firmware(&adev->gfx.ce_fw, fw_name, adev->dev);
+	if (err)
+		goto out;
+	err = amdgpu_ucode_validate(adev->gfx.ce_fw);
+	if (err)
+		goto out;
+
+	snprintf(fw_name, sizeof(fw_name), "radeon/%s_mec.bin", chip_name);
+	err = request_firmware(&adev->gfx.mec_fw, fw_name, adev->dev);
+	if (err)
+		goto out;
+	err = amdgpu_ucode_validate(adev->gfx.mec_fw);
+	if (err)
+		goto out;
+
+	if (adev->asic_type == CHIP_KAVERI) {
+		snprintf(fw_name, sizeof(fw_name), "radeon/%s_mec2.bin", chip_name);
+		err = request_firmware(&adev->gfx.mec2_fw, fw_name, adev->dev);
+		if (err)
+			goto out;
+		err = amdgpu_ucode_validate(adev->gfx.mec2_fw);
+		if (err)
+			goto out;
+	}
+
+	snprintf(fw_name, sizeof(fw_name), "radeon/%s_rlc.bin", chip_name);
+	err = request_firmware(&adev->gfx.rlc_fw, fw_name, adev->dev);
+	if (err)
+		goto out;
+	err = amdgpu_ucode_validate(adev->gfx.rlc_fw);
+
+out:
+	if (err) {
+		printk(KERN_ERR
+		       "gfx7: Failed to load firmware \"%s\"\n",
+		       fw_name);
+		release_firmware(adev->gfx.pfp_fw);
+		adev->gfx.pfp_fw = NULL;
+		release_firmware(adev->gfx.me_fw);
+		adev->gfx.me_fw = NULL;
+		release_firmware(adev->gfx.ce_fw);
+		adev->gfx.ce_fw = NULL;
+		release_firmware(adev->gfx.mec_fw);
+		adev->gfx.mec_fw = NULL;
+		release_firmware(adev->gfx.mec2_fw);
+		adev->gfx.mec2_fw = NULL;
+		release_firmware(adev->gfx.rlc_fw);
+		adev->gfx.rlc_fw = NULL;
+	}
+	return err;
+}
+
+/**
+ * gfx_v7_0_tiling_mode_table_init - init the hw tiling table
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Starting with SI, the tiling setup is done globally in a
+ * set of 32 tiling modes.  Rather than selecting each set of
+ * parameters per surface as on older asics, we just select
+ * which index in the tiling table we want to use, and the
+ * surface uses those parameters (CIK).
+ */
+static void gfx_v7_0_tiling_mode_table_init(struct amdgpu_device *adev)
+{
+	const u32 num_tile_mode_states = 32;
+	const u32 num_secondary_tile_mode_states = 16;
+	u32 reg_offset, gb_tile_moden, split_equal_to_row_size;
+
+	switch (adev->gfx.config.mem_row_size_in_kb) {
+	case 1:
+		split_equal_to_row_size = ADDR_SURF_TILE_SPLIT_1KB;
+		break;
+	case 2:
+	default:
+		split_equal_to_row_size = ADDR_SURF_TILE_SPLIT_2KB;
+		break;
+	case 4:
+		split_equal_to_row_size = ADDR_SURF_TILE_SPLIT_4KB;
+		break;
+	}
+
+	switch (adev->asic_type) {
+	case CHIP_BONAIRE:
+		for (reg_offset = 0; reg_offset < num_tile_mode_states; reg_offset++) {
+			switch (reg_offset) {
+			case 0:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) |
+						PIPE_CONFIG(ADDR_SURF_P4_16x16) |
+						TILE_SPLIT(ADDR_SURF_TILE_SPLIT_64B) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_DEPTH_MICRO_TILING));
+				break;
+			case 1:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) |
+						PIPE_CONFIG(ADDR_SURF_P4_16x16) |
+						TILE_SPLIT(ADDR_SURF_TILE_SPLIT_128B) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_DEPTH_MICRO_TILING));
+				break;
+			case 2:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) |
+						PIPE_CONFIG(ADDR_SURF_P4_16x16) |
+						TILE_SPLIT(ADDR_SURF_TILE_SPLIT_256B) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_DEPTH_MICRO_TILING));
+				break;
+			case 3:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) |
+						PIPE_CONFIG(ADDR_SURF_P4_16x16) |
+						TILE_SPLIT(ADDR_SURF_TILE_SPLIT_512B) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_DEPTH_MICRO_TILING));
+				break;
+			case 4:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) |
+						PIPE_CONFIG(ADDR_SURF_P4_16x16) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_DEPTH_MICRO_TILING) |
+						TILE_SPLIT(split_equal_to_row_size));
+				break;
+			case 5:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_1D_TILED_THIN1) |
+						PIPE_CONFIG(ADDR_SURF_P4_16x16) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_DEPTH_MICRO_TILING));
+				break;
+			case 6:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_PRT_TILED_THIN1) |
+						PIPE_CONFIG(ADDR_SURF_P4_16x16) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_DEPTH_MICRO_TILING) |
+						TILE_SPLIT(split_equal_to_row_size));
+				break;
+			case 7:
+				gb_tile_moden = (TILE_SPLIT(split_equal_to_row_size));
+				break;
+
+			case 8:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_LINEAR_ALIGNED) |
+						PIPE_CONFIG(ADDR_SURF_P4_16x16));
+				break;
+			case 9:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_1D_TILED_THIN1) |
+						PIPE_CONFIG(ADDR_SURF_P4_16x16) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_DISPLAY_MICRO_TILING));
+				break;
+			case 10:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) |
+						PIPE_CONFIG(ADDR_SURF_P4_16x16) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_DISPLAY_MICRO_TILING) |
+						SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_2));
+				break;
+			case 11:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_PRT_TILED_THIN1) |
+						PIPE_CONFIG(ADDR_SURF_P4_16x16) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_DISPLAY_MICRO_TILING) |
+						SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_8));
+				break;
+			case 12:
+				gb_tile_moden = (TILE_SPLIT(split_equal_to_row_size));
+				break;
+			case 13:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_1D_TILED_THIN1) |
+						PIPE_CONFIG(ADDR_SURF_P4_16x16) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_THIN_MICRO_TILING));
+				break;
+			case 14:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) |
+						PIPE_CONFIG(ADDR_SURF_P4_16x16) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_THIN_MICRO_TILING) |
+						SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_2));
+				break;
+			case 15:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_3D_TILED_THIN1) |
+						PIPE_CONFIG(ADDR_SURF_P4_16x16) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_THIN_MICRO_TILING) |
+						SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_2));
+				break;
+			case 16:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_PRT_TILED_THIN1) |
+						PIPE_CONFIG(ADDR_SURF_P4_16x16) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_THIN_MICRO_TILING) |
+						SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_8));
+				break;
+			case 17:
+				gb_tile_moden = (TILE_SPLIT(split_equal_to_row_size));
+				break;
+			case 18:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_1D_TILED_THICK) |
+						PIPE_CONFIG(ADDR_SURF_P4_16x16) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_THIN_MICRO_TILING) |
+						SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_1));
+				break;
+			case 19:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_1D_TILED_THICK) |
+						PIPE_CONFIG(ADDR_SURF_P4_16x16) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_THIN_MICRO_TILING));
+				break;
+			case 20:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_2D_TILED_THICK) |
+						PIPE_CONFIG(ADDR_SURF_P4_16x16) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_THIN_MICRO_TILING) |
+						SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_1));
+				break;
+			case 21:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_3D_TILED_THICK) |
+						PIPE_CONFIG(ADDR_SURF_P4_16x16) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_THIN_MICRO_TILING) |
+						SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_1));
+				break;
+			case 22:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_PRT_TILED_THICK) |
+						PIPE_CONFIG(ADDR_SURF_P4_16x16) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_THIN_MICRO_TILING) |
+						SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_1));
+				break;
+			case 23:
+				gb_tile_moden = (TILE_SPLIT(split_equal_to_row_size));
+				break;
+			case 24:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_2D_TILED_THICK) |
+						PIPE_CONFIG(ADDR_SURF_P4_16x16) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_THIN_MICRO_TILING) |
+						SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_1));
+				break;
+			case 25:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_2D_TILED_XTHICK) |
+						PIPE_CONFIG(ADDR_SURF_P4_16x16) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_THIN_MICRO_TILING) |
+						SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_1));
+				break;
+			case 26:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_3D_TILED_XTHICK) |
+						PIPE_CONFIG(ADDR_SURF_P4_16x16) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_THIN_MICRO_TILING) |
+						SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_1));
+				break;
+			case 27:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_1D_TILED_THIN1) |
+						PIPE_CONFIG(ADDR_SURF_P4_16x16) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_ROTATED_MICRO_TILING));
+				break;
+			case 28:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) |
+						PIPE_CONFIG(ADDR_SURF_P4_16x16) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_ROTATED_MICRO_TILING) |
+						SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_2));
+				break;
+			case 29:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_PRT_TILED_THIN1) |
+						PIPE_CONFIG(ADDR_SURF_P4_16x16) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_ROTATED_MICRO_TILING) |
+						SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_8));
+				break;
+			case 30:
+				gb_tile_moden = (TILE_SPLIT(split_equal_to_row_size));
+				break;
+			default:
+				gb_tile_moden = 0;
+				break;
+			}
+			adev->gfx.config.tile_mode_array[reg_offset] = gb_tile_moden;
+			WREG32(mmGB_TILE_MODE0 + reg_offset, gb_tile_moden);
+		}
+		for (reg_offset = 0; reg_offset < num_secondary_tile_mode_states; reg_offset++) {
+			switch (reg_offset) {
+			case 0:
+				gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
+						BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_4) |
+						MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_4) |
+						NUM_BANKS(ADDR_SURF_16_BANK));
+				break;
+			case 1:
+				gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
+						BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_2) |
+						MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_4) |
+						NUM_BANKS(ADDR_SURF_16_BANK));
+				break;
+			case 2:
+				gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
+						BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_1) |
+						MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2) |
+						NUM_BANKS(ADDR_SURF_16_BANK));
+				break;
+			case 3:
+				gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
+						BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_1) |
+						MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2) |
+						NUM_BANKS(ADDR_SURF_16_BANK));
+				break;
+			case 4:
+				gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
+						BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_1) |
+						MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2) |
+						NUM_BANKS(ADDR_SURF_16_BANK));
+				break;
+			case 5:
+				gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
+						BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_1) |
+						MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2) |
+						NUM_BANKS(ADDR_SURF_8_BANK));
+				break;
+			case 6:
+				gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
+						BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_1) |
+						MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_1) |
+						NUM_BANKS(ADDR_SURF_4_BANK));
+				break;
+			case 8:
+				gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_2) |
+						BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_8) |
+						MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_4) |
+						NUM_BANKS(ADDR_SURF_16_BANK));
+				break;
+			case 9:
+				gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_2) |
+						BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_4) |
+						MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_4) |
+						NUM_BANKS(ADDR_SURF_16_BANK));
+				break;
+			case 10:
+				gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
+						BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_4) |
+						MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_4) |
+						NUM_BANKS(ADDR_SURF_16_BANK));
+				break;
+			case 11:
+				gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
+						BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_2) |
+						MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_4) |
+						NUM_BANKS(ADDR_SURF_16_BANK));
+				break;
+			case 12:
+				gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
+						BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_1) |
+						MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2) |
+						NUM_BANKS(ADDR_SURF_16_BANK));
+				break;
+			case 13:
+				gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
+						BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_1) |
+						MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2) |
+						NUM_BANKS(ADDR_SURF_8_BANK));
+				break;
+			case 14:
+				gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
+						BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_1) |
+						MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_1) |
+						NUM_BANKS(ADDR_SURF_4_BANK));
+				break;
+			default:
+				gb_tile_moden = 0;
+				break;
+			}
+			adev->gfx.config.macrotile_mode_array[reg_offset] = gb_tile_moden;
+			WREG32(mmGB_MACROTILE_MODE0 + reg_offset, gb_tile_moden);
+		}
+		break;
+	case CHIP_HAWAII:
+		for (reg_offset = 0; reg_offset < num_tile_mode_states; reg_offset++) {
+			switch (reg_offset) {
+			case 0:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) |
+						PIPE_CONFIG(ADDR_SURF_P16_32x32_16x16) |
+						TILE_SPLIT(ADDR_SURF_TILE_SPLIT_64B) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_DEPTH_MICRO_TILING));
+				break;
+			case 1:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) |
+						PIPE_CONFIG(ADDR_SURF_P16_32x32_16x16) |
+						TILE_SPLIT(ADDR_SURF_TILE_SPLIT_128B) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_DEPTH_MICRO_TILING));
+				break;
+			case 2:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) |
+						PIPE_CONFIG(ADDR_SURF_P16_32x32_16x16) |
+						TILE_SPLIT(ADDR_SURF_TILE_SPLIT_256B) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_DEPTH_MICRO_TILING));
+				break;
+			case 3:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) |
+						PIPE_CONFIG(ADDR_SURF_P16_32x32_16x16) |
+						TILE_SPLIT(ADDR_SURF_TILE_SPLIT_512B) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_DEPTH_MICRO_TILING));
+				break;
+			case 4:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) |
+						PIPE_CONFIG(ADDR_SURF_P16_32x32_16x16) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_DEPTH_MICRO_TILING) |
+						TILE_SPLIT(split_equal_to_row_size));
+				break;
+			case 5:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_1D_TILED_THIN1) |
+						PIPE_CONFIG(ADDR_SURF_P16_32x32_16x16) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_DEPTH_MICRO_TILING) |
+						TILE_SPLIT(split_equal_to_row_size));
+				break;
+			case 6:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_PRT_TILED_THIN1) |
+						PIPE_CONFIG(ADDR_SURF_P16_32x32_16x16) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_DEPTH_MICRO_TILING) |
+						TILE_SPLIT(split_equal_to_row_size));
+				break;
+			case 7:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_PRT_TILED_THIN1) |
+						PIPE_CONFIG(ADDR_SURF_P4_16x16) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_DEPTH_MICRO_TILING) |
+						TILE_SPLIT(split_equal_to_row_size));
+				break;
+
+			case 8:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_LINEAR_ALIGNED) |
+						PIPE_CONFIG(ADDR_SURF_P16_32x32_16x16));
+				break;
+			case 9:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_1D_TILED_THIN1) |
+						PIPE_CONFIG(ADDR_SURF_P16_32x32_16x16) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_DISPLAY_MICRO_TILING));
+				break;
+			case 10:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) |
+						PIPE_CONFIG(ADDR_SURF_P16_32x32_16x16) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_DISPLAY_MICRO_TILING) |
+						SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_2));
+				break;
+			case 11:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_PRT_TILED_THIN1) |
+						PIPE_CONFIG(ADDR_SURF_P16_32x32_16x16) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_DISPLAY_MICRO_TILING) |
+						SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_8));
+				break;
+			case 12:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_PRT_2D_TILED_THIN1) |
+						PIPE_CONFIG(ADDR_SURF_P4_16x16) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_DISPLAY_MICRO_TILING) |
+						SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_8));
+				break;
+			case 13:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_1D_TILED_THIN1) |
+						PIPE_CONFIG(ADDR_SURF_P16_32x32_16x16) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_THIN_MICRO_TILING));
+				break;
+			case 14:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) |
+						PIPE_CONFIG(ADDR_SURF_P16_32x32_16x16) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_THIN_MICRO_TILING) |
+						SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_2));
+				break;
+			case 15:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_3D_TILED_THIN1) |
+						PIPE_CONFIG(ADDR_SURF_P16_32x32_16x16) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_THIN_MICRO_TILING) |
+						SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_2));
+				break;
+			case 16:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_PRT_TILED_THIN1) |
+						PIPE_CONFIG(ADDR_SURF_P16_32x32_16x16) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_THIN_MICRO_TILING) |
+						SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_8));
+				break;
+			case 17:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_PRT_TILED_THIN1) |
+						PIPE_CONFIG(ADDR_SURF_P4_16x16) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_THIN_MICRO_TILING) |
+						SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_8));
+				break;
+			case 18:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_1D_TILED_THICK) |
+						PIPE_CONFIG(ADDR_SURF_P16_32x32_16x16) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_THIN_MICRO_TILING) |
+						SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_1));
+				break;
+			case 19:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_1D_TILED_THICK) |
+						PIPE_CONFIG(ADDR_SURF_P16_32x32_16x16) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_THICK_MICRO_TILING));
+				break;
+			case 20:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_2D_TILED_THICK) |
+						PIPE_CONFIG(ADDR_SURF_P16_32x32_16x16) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_THICK_MICRO_TILING) |
+						SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_1));
+				break;
+			case 21:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_3D_TILED_THICK) |
+						PIPE_CONFIG(ADDR_SURF_P16_32x32_16x16) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_THICK_MICRO_TILING) |
+						SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_1));
+				break;
+			case 22:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_PRT_TILED_THICK) |
+						PIPE_CONFIG(ADDR_SURF_P16_32x32_16x16) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_THICK_MICRO_TILING) |
+						SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_1));
+				break;
+			case 23:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_PRT_TILED_THICK) |
+						PIPE_CONFIG(ADDR_SURF_P4_16x16) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_THICK_MICRO_TILING) |
+						SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_1));
+				break;
+			case 24:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_2D_TILED_THICK) |
+						PIPE_CONFIG(ADDR_SURF_P16_32x32_16x16) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_THIN_MICRO_TILING) |
+						SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_1));
+				break;
+			case 25:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_2D_TILED_XTHICK) |
+						PIPE_CONFIG(ADDR_SURF_P16_32x32_16x16) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_THICK_MICRO_TILING) |
+						SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_1));
+				break;
+			case 26:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_3D_TILED_XTHICK) |
+						PIPE_CONFIG(ADDR_SURF_P16_32x32_16x16) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_THICK_MICRO_TILING) |
+						SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_1));
+				break;
+			case 27:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_1D_TILED_THIN1) |
+						PIPE_CONFIG(ADDR_SURF_P16_32x32_16x16) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_ROTATED_MICRO_TILING));
+				break;
+			case 28:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) |
+						PIPE_CONFIG(ADDR_SURF_P16_32x32_16x16) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_ROTATED_MICRO_TILING) |
+						SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_2));
+				break;
+			case 29:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_PRT_TILED_THIN1) |
+						PIPE_CONFIG(ADDR_SURF_P16_32x32_16x16) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_ROTATED_MICRO_TILING) |
+						SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_8));
+				break;
+			case 30:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_PRT_TILED_THIN1) |
+						PIPE_CONFIG(ADDR_SURF_P4_16x16) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_ROTATED_MICRO_TILING) |
+						SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_8));
+				break;
+			default:
+				gb_tile_moden = 0;
+				break;
+			}
+			adev->gfx.config.tile_mode_array[reg_offset] = gb_tile_moden;
+			WREG32(mmGB_TILE_MODE0 + reg_offset, gb_tile_moden);
+		}
+		for (reg_offset = 0; reg_offset < num_secondary_tile_mode_states; reg_offset++) {
+			switch (reg_offset) {
+			case 0:
+				gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
+						BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_4) |
+						MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2) |
+						NUM_BANKS(ADDR_SURF_16_BANK));
+				break;
+			case 1:
+				gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
+						BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_2) |
+						MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2) |
+						NUM_BANKS(ADDR_SURF_16_BANK));
+				break;
+			case 2:
+				gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
+						BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_1) |
+						MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_1) |
+						NUM_BANKS(ADDR_SURF_16_BANK));
+				break;
+			case 3:
+				gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
+						BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_1) |
+						MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_1) |
+						NUM_BANKS(ADDR_SURF_16_BANK));
+				break;
+			case 4:
+				gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
+						BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_1) |
+						MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_1) |
+						NUM_BANKS(ADDR_SURF_8_BANK));
+				break;
+			case 5:
+				gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
+						BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_1) |
+						MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_1) |
+						NUM_BANKS(ADDR_SURF_4_BANK));
+				break;
+			case 6:
+				gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
+						BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_1) |
+						MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_1) |
+						NUM_BANKS(ADDR_SURF_4_BANK));
+				break;
+			case 8:
+				gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
+						BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_4) |
+						MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2) |
+						NUM_BANKS(ADDR_SURF_16_BANK));
+				break;
+			case 9:
+				gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
+						BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_2) |
+						MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2) |
+						NUM_BANKS(ADDR_SURF_16_BANK));
+				break;
+			case 10:
+				gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
+						BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_1) |
+						MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_1) |
+						NUM_BANKS(ADDR_SURF_16_BANK));
+				break;
+			case 11:
+				gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
+						BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_1) |
+						MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_1) |
+						NUM_BANKS(ADDR_SURF_8_BANK));
+				break;
+			case 12:
+				gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
+						BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_1) |
+						MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2) |
+						NUM_BANKS(ADDR_SURF_16_BANK));
+				break;
+			case 13:
+				gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
+						BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_1) |
+						MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2) |
+						NUM_BANKS(ADDR_SURF_8_BANK));
+				break;
+			case 14:
+				gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
+						BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_1) |
+						MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_1) |
+						NUM_BANKS(ADDR_SURF_4_BANK));
+				break;
+			default:
+				gb_tile_moden = 0;
+				break;
+			}
+			adev->gfx.config.macrotile_mode_array[reg_offset] = gb_tile_moden;
+			WREG32(mmGB_MACROTILE_MODE0 + reg_offset, gb_tile_moden);
+		}
+		break;
+	case CHIP_KABINI:
+	case CHIP_KAVERI:
+	case CHIP_MULLINS:
+	default:
+		for (reg_offset = 0; reg_offset < num_tile_mode_states; reg_offset++) {
+			switch (reg_offset) {
+			case 0:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) |
+						PIPE_CONFIG(ADDR_SURF_P2) |
+						TILE_SPLIT(ADDR_SURF_TILE_SPLIT_64B) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_DEPTH_MICRO_TILING));
+				break;
+			case 1:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) |
+						PIPE_CONFIG(ADDR_SURF_P2) |
+						TILE_SPLIT(ADDR_SURF_TILE_SPLIT_128B) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_DEPTH_MICRO_TILING));
+				break;
+			case 2:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) |
+						PIPE_CONFIG(ADDR_SURF_P2) |
+						TILE_SPLIT(ADDR_SURF_TILE_SPLIT_256B) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_DEPTH_MICRO_TILING));
+				break;
+			case 3:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) |
+						PIPE_CONFIG(ADDR_SURF_P2) |
+						TILE_SPLIT(ADDR_SURF_TILE_SPLIT_512B) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_DEPTH_MICRO_TILING));
+				break;
+			case 4:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) |
+						PIPE_CONFIG(ADDR_SURF_P2) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_DEPTH_MICRO_TILING) |
+						TILE_SPLIT(split_equal_to_row_size));
+				break;
+			case 5:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_1D_TILED_THIN1) |
+						PIPE_CONFIG(ADDR_SURF_P2) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_DEPTH_MICRO_TILING));
+				break;
+			case 6:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_PRT_TILED_THIN1) |
+						PIPE_CONFIG(ADDR_SURF_P2) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_DEPTH_MICRO_TILING) |
+						TILE_SPLIT(split_equal_to_row_size));
+				break;
+			case 7:
+				gb_tile_moden = (TILE_SPLIT(split_equal_to_row_size));
+				break;
+
+			case 8:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_LINEAR_ALIGNED) |
+						PIPE_CONFIG(ADDR_SURF_P2));
+				break;
+			case 9:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_1D_TILED_THIN1) |
+						PIPE_CONFIG(ADDR_SURF_P2) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_DISPLAY_MICRO_TILING));
+				break;
+			case 10:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) |
+						PIPE_CONFIG(ADDR_SURF_P2) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_DISPLAY_MICRO_TILING) |
+						SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_2));
+				break;
+			case 11:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_PRT_TILED_THIN1) |
+						PIPE_CONFIG(ADDR_SURF_P2) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_DISPLAY_MICRO_TILING) |
+						SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_8));
+				break;
+			case 12:
+				gb_tile_moden = (TILE_SPLIT(split_equal_to_row_size));
+				break;
+			case 13:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_1D_TILED_THIN1) |
+						PIPE_CONFIG(ADDR_SURF_P2) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_THIN_MICRO_TILING));
+				break;
+			case 14:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) |
+						PIPE_CONFIG(ADDR_SURF_P2) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_THIN_MICRO_TILING) |
+						SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_2));
+				break;
+			case 15:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_3D_TILED_THIN1) |
+						PIPE_CONFIG(ADDR_SURF_P2) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_THIN_MICRO_TILING) |
+						SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_2));
+				break;
+			case 16:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_PRT_TILED_THIN1) |
+						PIPE_CONFIG(ADDR_SURF_P2) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_THIN_MICRO_TILING) |
+						SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_8));
+				break;
+			case 17:
+				gb_tile_moden = (TILE_SPLIT(split_equal_to_row_size));
+				break;
+			case 18:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_1D_TILED_THICK) |
+						PIPE_CONFIG(ADDR_SURF_P2) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_THIN_MICRO_TILING) |
+						SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_1));
+				break;
+			case 19:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_1D_TILED_THICK) |
+						PIPE_CONFIG(ADDR_SURF_P2) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_THICK_MICRO_TILING));
+				break;
+			case 20:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_2D_TILED_THICK) |
+						PIPE_CONFIG(ADDR_SURF_P2) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_THICK_MICRO_TILING) |
+						SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_1));
+				break;
+			case 21:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_3D_TILED_THICK) |
+						PIPE_CONFIG(ADDR_SURF_P2) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_THICK_MICRO_TILING) |
+						SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_1));
+				break;
+			case 22:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_PRT_TILED_THICK) |
+						PIPE_CONFIG(ADDR_SURF_P2) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_THICK_MICRO_TILING) |
+						SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_1));
+				break;
+			case 23:
+				gb_tile_moden = (TILE_SPLIT(split_equal_to_row_size));
+				break;
+			case 24:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_2D_TILED_THICK) |
+						PIPE_CONFIG(ADDR_SURF_P2) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_THIN_MICRO_TILING) |
+						SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_1));
+				break;
+			case 25:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_2D_TILED_XTHICK) |
+						PIPE_CONFIG(ADDR_SURF_P2) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_THICK_MICRO_TILING) |
+						SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_1));
+				break;
+			case 26:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_3D_TILED_XTHICK) |
+						PIPE_CONFIG(ADDR_SURF_P2) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_THICK_MICRO_TILING) |
+						SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_1));
+				break;
+			case 27:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_1D_TILED_THIN1) |
+						PIPE_CONFIG(ADDR_SURF_P2) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_ROTATED_MICRO_TILING));
+				break;
+			case 28:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) |
+						PIPE_CONFIG(ADDR_SURF_P2) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_ROTATED_MICRO_TILING) |
+						SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_2));
+				break;
+			case 29:
+				gb_tile_moden = (ARRAY_MODE(ARRAY_PRT_TILED_THIN1) |
+						PIPE_CONFIG(ADDR_SURF_P2) |
+						MICRO_TILE_MODE_NEW(ADDR_SURF_ROTATED_MICRO_TILING) |
+						SAMPLE_SPLIT(ADDR_SURF_SAMPLE_SPLIT_8));
+				break;
+			case 30:
+				gb_tile_moden = (TILE_SPLIT(split_equal_to_row_size));
+				break;
+			default:
+				gb_tile_moden = 0;
+				break;
+			}
+			adev->gfx.config.tile_mode_array[reg_offset] = gb_tile_moden;
+			WREG32(mmGB_TILE_MODE0 + reg_offset, gb_tile_moden);
+		}
+		for (reg_offset = 0; reg_offset < num_secondary_tile_mode_states; reg_offset++) {
+			switch (reg_offset) {
+			case 0:
+				gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
+						BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_4) |
+						MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_4) |
+						NUM_BANKS(ADDR_SURF_8_BANK));
+				break;
+			case 1:
+				gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
+						BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_2) |
+						MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_4) |
+						NUM_BANKS(ADDR_SURF_8_BANK));
+				break;
+			case 2:
+				gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
+						BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_2) |
+						MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2) |
+						NUM_BANKS(ADDR_SURF_8_BANK));
+				break;
+			case 3:
+				gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
+						BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_1) |
+						MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2) |
+						NUM_BANKS(ADDR_SURF_8_BANK));
+				break;
+			case 4:
+				gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
+						BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_1) |
+						MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2) |
+						NUM_BANKS(ADDR_SURF_8_BANK));
+				break;
+			case 5:
+				gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
+						BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_1) |
+						MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2) |
+						NUM_BANKS(ADDR_SURF_8_BANK));
+				break;
+			case 6:
+				gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
+						BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_1) |
+						MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2) |
+						NUM_BANKS(ADDR_SURF_8_BANK));
+				break;
+			case 8:
+				gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_4) |
+						BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_8) |
+						MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_4) |
+						NUM_BANKS(ADDR_SURF_16_BANK));
+				break;
+			case 9:
+				gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_4) |
+						BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_4) |
+						MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_4) |
+						NUM_BANKS(ADDR_SURF_16_BANK));
+				break;
+			case 10:
+				gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_2) |
+						BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_4) |
+						MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_4) |
+						NUM_BANKS(ADDR_SURF_16_BANK));
+				break;
+			case 11:
+				gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_2) |
+						BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_2) |
+						MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_4) |
+						NUM_BANKS(ADDR_SURF_16_BANK));
+				break;
+			case 12:
+				gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
+						BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_2) |
+						MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_4) |
+						NUM_BANKS(ADDR_SURF_16_BANK));
+				break;
+			case 13:
+				gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
+						BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_1) |
+						MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_4) |
+						NUM_BANKS(ADDR_SURF_16_BANK));
+				break;
+			case 14:
+				gb_tile_moden = (BANK_WIDTH(ADDR_SURF_BANK_WIDTH_1) |
+						BANK_HEIGHT(ADDR_SURF_BANK_HEIGHT_1) |
+						MACRO_TILE_ASPECT(ADDR_SURF_MACRO_ASPECT_2) |
+						NUM_BANKS(ADDR_SURF_8_BANK));
+				break;
+			default:
+				gb_tile_moden = 0;
+				break;
+			}
+			adev->gfx.config.macrotile_mode_array[reg_offset] = gb_tile_moden;
+			WREG32(mmGB_MACROTILE_MODE0 + reg_offset, gb_tile_moden);
+		}
+		break;
+	}
+}
+
+/**
+ * gfx_v7_0_select_se_sh - select which SE, SH to address
+ *
+ * @adev: amdgpu_device pointer
+ * @se_num: shader engine to address
+ * @sh_num: sh block to address
+ *
+ * Select which SE, SH combinations to address. Certain
+ * registers are instanced per SE or SH.  0xffffffff means
+ * broadcast to all SEs or SHs (CIK).
+ */
+void gfx_v7_0_select_se_sh(struct amdgpu_device *adev, u32 se_num, u32 sh_num)
+{
+	u32 data = GRBM_GFX_INDEX__INSTANCE_BROADCAST_WRITES_MASK;
+
+	if ((se_num == 0xffffffff) && (sh_num == 0xffffffff))
+		data |= GRBM_GFX_INDEX__SH_BROADCAST_WRITES_MASK |
+			GRBM_GFX_INDEX__SE_BROADCAST_WRITES_MASK;
+	else if (se_num == 0xffffffff)
+		data |= GRBM_GFX_INDEX__SE_BROADCAST_WRITES_MASK |
+			(sh_num << GRBM_GFX_INDEX__SH_INDEX__SHIFT);
+	else if (sh_num == 0xffffffff)
+		data |= GRBM_GFX_INDEX__SH_BROADCAST_WRITES_MASK |
+			(se_num << GRBM_GFX_INDEX__SE_INDEX__SHIFT);
+	else
+		data |= (sh_num << GRBM_GFX_INDEX__SH_INDEX__SHIFT) |
+			(se_num << GRBM_GFX_INDEX__SE_INDEX__SHIFT);
+	WREG32(mmGRBM_GFX_INDEX, data);
+}
+
+/**
+ * gfx_v7_0_create_bitmask - create a bitmask
+ *
+ * @bit_width: length of the mask
+ *
+ * create a variable length bit mask (CIK).
+ * Returns the bitmask.
+ */
+static u32 gfx_v7_0_create_bitmask(u32 bit_width)
+{
+	u32 i, mask = 0;
+
+	for (i = 0; i < bit_width; i++) {
+		mask <<= 1;
+		mask |= 1;
+	}
+	return mask;
+}
+
+/**
+ * gfx_v7_0_get_rb_disabled - computes the mask of disabled RBs
+ *
+ * @adev: amdgpu_device pointer
+ * @max_rb_num: max RBs (render backends) for the asic
+ * @se_num: number of SEs (shader engines) for the asic
+ * @sh_per_se: number of SH blocks per SE for the asic
+ *
+ * Calculates the bitmask of disabled RBs (CIK).
+ * Returns the disabled RB bitmask.
+ */
+static u32 gfx_v7_0_get_rb_disabled(struct amdgpu_device *adev,
+				    u32 max_rb_num_per_se,
+				    u32 sh_per_se)
+{
+	u32 data, mask;
+
+	data = RREG32(mmCC_RB_BACKEND_DISABLE);
+	if (data & 1)
+		data &= CC_RB_BACKEND_DISABLE__BACKEND_DISABLE_MASK;
+	else
+		data = 0;
+
+	data |= RREG32(mmGC_USER_RB_BACKEND_DISABLE);
+
+	data >>= GC_USER_RB_BACKEND_DISABLE__BACKEND_DISABLE__SHIFT;
+
+	mask = gfx_v7_0_create_bitmask(max_rb_num_per_se / sh_per_se);
+
+	return data & mask;
+}
+
+/**
+ * gfx_v7_0_setup_rb - setup the RBs on the asic
+ *
+ * @adev: amdgpu_device pointer
+ * @se_num: number of SEs (shader engines) for the asic
+ * @sh_per_se: number of SH blocks per SE for the asic
+ * @max_rb_num: max RBs (render backends) for the asic
+ *
+ * Configures per-SE/SH RB registers (CIK).
+ */
+static void gfx_v7_0_setup_rb(struct amdgpu_device *adev,
+			      u32 se_num, u32 sh_per_se,
+			      u32 max_rb_num_per_se)
+{
+	int i, j;
+	u32 data, mask;
+	u32 disabled_rbs = 0;
+	u32 enabled_rbs = 0;
+
+	mutex_lock(&adev->grbm_idx_mutex);
+	for (i = 0; i < se_num; i++) {
+		for (j = 0; j < sh_per_se; j++) {
+			gfx_v7_0_select_se_sh(adev, i, j);
+			data = gfx_v7_0_get_rb_disabled(adev, max_rb_num_per_se, sh_per_se);
+			if (adev->asic_type == CHIP_HAWAII)
+				disabled_rbs |= data << ((i * sh_per_se + j) * HAWAII_RB_BITMAP_WIDTH_PER_SH);
+			else
+				disabled_rbs |= data << ((i * sh_per_se + j) * CIK_RB_BITMAP_WIDTH_PER_SH);
+		}
+	}
+	gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff);
+	mutex_unlock(&adev->grbm_idx_mutex);
+
+	mask = 1;
+	for (i = 0; i < max_rb_num_per_se * se_num; i++) {
+		if (!(disabled_rbs & mask))
+			enabled_rbs |= mask;
+		mask <<= 1;
+	}
+
+	adev->gfx.config.backend_enable_mask = enabled_rbs;
+
+	mutex_lock(&adev->grbm_idx_mutex);
+	for (i = 0; i < se_num; i++) {
+		gfx_v7_0_select_se_sh(adev, i, 0xffffffff);
+		data = 0;
+		for (j = 0; j < sh_per_se; j++) {
+			switch (enabled_rbs & 3) {
+			case 0:
+				if (j == 0)
+					data |= (RASTER_CONFIG_RB_MAP_3 <<
+						PA_SC_RASTER_CONFIG__PKR_MAP__SHIFT);
+				else
+					data |= (RASTER_CONFIG_RB_MAP_0 <<
+						PA_SC_RASTER_CONFIG__PKR_MAP__SHIFT);
+				break;
+			case 1:
+				data |= (RASTER_CONFIG_RB_MAP_0 << (i * sh_per_se + j) * 2);
+				break;
+			case 2:
+				data |= (RASTER_CONFIG_RB_MAP_3 << (i * sh_per_se + j) * 2);
+				break;
+			case 3:
+			default:
+				data |= (RASTER_CONFIG_RB_MAP_2 << (i * sh_per_se + j) * 2);
+				break;
+			}
+			enabled_rbs >>= 2;
+		}
+		WREG32(mmPA_SC_RASTER_CONFIG, data);
+	}
+	gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff);
+	mutex_unlock(&adev->grbm_idx_mutex);
+}
+
+/**
+ * gfx_v7_0_gpu_init - setup the 3D engine
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Configures the 3D engine and tiling configuration
+ * registers so that the 3D engine is usable.
+ */
+static void gfx_v7_0_gpu_init(struct amdgpu_device *adev)
+{
+	u32 gb_addr_config;
+	u32 mc_shared_chmap, mc_arb_ramcfg;
+	u32 dimm00_addr_map, dimm01_addr_map, dimm10_addr_map, dimm11_addr_map;
+	u32 tmp;
+	int i;
+
+	switch (adev->asic_type) {
+	case CHIP_BONAIRE:
+		adev->gfx.config.max_shader_engines = 2;
+		adev->gfx.config.max_tile_pipes = 4;
+		adev->gfx.config.max_cu_per_sh = 7;
+		adev->gfx.config.max_sh_per_se = 1;
+		adev->gfx.config.max_backends_per_se = 2;
+		adev->gfx.config.max_texture_channel_caches = 4;
+		adev->gfx.config.max_gprs = 256;
+		adev->gfx.config.max_gs_threads = 32;
+		adev->gfx.config.max_hw_contexts = 8;
+
+		adev->gfx.config.sc_prim_fifo_size_frontend = 0x20;
+		adev->gfx.config.sc_prim_fifo_size_backend = 0x100;
+		adev->gfx.config.sc_hiz_tile_fifo_size = 0x30;
+		adev->gfx.config.sc_earlyz_tile_fifo_size = 0x130;
+		gb_addr_config = BONAIRE_GB_ADDR_CONFIG_GOLDEN;
+		break;
+	case CHIP_HAWAII:
+		adev->gfx.config.max_shader_engines = 4;
+		adev->gfx.config.max_tile_pipes = 16;
+		adev->gfx.config.max_cu_per_sh = 11;
+		adev->gfx.config.max_sh_per_se = 1;
+		adev->gfx.config.max_backends_per_se = 4;
+		adev->gfx.config.max_texture_channel_caches = 16;
+		adev->gfx.config.max_gprs = 256;
+		adev->gfx.config.max_gs_threads = 32;
+		adev->gfx.config.max_hw_contexts = 8;
+
+		adev->gfx.config.sc_prim_fifo_size_frontend = 0x20;
+		adev->gfx.config.sc_prim_fifo_size_backend = 0x100;
+		adev->gfx.config.sc_hiz_tile_fifo_size = 0x30;
+		adev->gfx.config.sc_earlyz_tile_fifo_size = 0x130;
+		gb_addr_config = HAWAII_GB_ADDR_CONFIG_GOLDEN;
+		break;
+	case CHIP_KAVERI:
+		adev->gfx.config.max_shader_engines = 1;
+		adev->gfx.config.max_tile_pipes = 4;
+		if ((adev->pdev->device == 0x1304) ||
+		    (adev->pdev->device == 0x1305) ||
+		    (adev->pdev->device == 0x130C) ||
+		    (adev->pdev->device == 0x130F) ||
+		    (adev->pdev->device == 0x1310) ||
+		    (adev->pdev->device == 0x1311) ||
+		    (adev->pdev->device == 0x131C)) {
+			adev->gfx.config.max_cu_per_sh = 8;
+			adev->gfx.config.max_backends_per_se = 2;
+		} else if ((adev->pdev->device == 0x1309) ||
+			   (adev->pdev->device == 0x130A) ||
+			   (adev->pdev->device == 0x130D) ||
+			   (adev->pdev->device == 0x1313) ||
+			   (adev->pdev->device == 0x131D)) {
+			adev->gfx.config.max_cu_per_sh = 6;
+			adev->gfx.config.max_backends_per_se = 2;
+		} else if ((adev->pdev->device == 0x1306) ||
+			   (adev->pdev->device == 0x1307) ||
+			   (adev->pdev->device == 0x130B) ||
+			   (adev->pdev->device == 0x130E) ||
+			   (adev->pdev->device == 0x1315) ||
+			   (adev->pdev->device == 0x131B)) {
+			adev->gfx.config.max_cu_per_sh = 4;
+			adev->gfx.config.max_backends_per_se = 1;
+		} else {
+			adev->gfx.config.max_cu_per_sh = 3;
+			adev->gfx.config.max_backends_per_se = 1;
+		}
+		adev->gfx.config.max_sh_per_se = 1;
+		adev->gfx.config.max_texture_channel_caches = 4;
+		adev->gfx.config.max_gprs = 256;
+		adev->gfx.config.max_gs_threads = 16;
+		adev->gfx.config.max_hw_contexts = 8;
+
+		adev->gfx.config.sc_prim_fifo_size_frontend = 0x20;
+		adev->gfx.config.sc_prim_fifo_size_backend = 0x100;
+		adev->gfx.config.sc_hiz_tile_fifo_size = 0x30;
+		adev->gfx.config.sc_earlyz_tile_fifo_size = 0x130;
+		gb_addr_config = BONAIRE_GB_ADDR_CONFIG_GOLDEN;
+		break;
+	case CHIP_KABINI:
+	case CHIP_MULLINS:
+	default:
+		adev->gfx.config.max_shader_engines = 1;
+		adev->gfx.config.max_tile_pipes = 2;
+		adev->gfx.config.max_cu_per_sh = 2;
+		adev->gfx.config.max_sh_per_se = 1;
+		adev->gfx.config.max_backends_per_se = 1;
+		adev->gfx.config.max_texture_channel_caches = 2;
+		adev->gfx.config.max_gprs = 256;
+		adev->gfx.config.max_gs_threads = 16;
+		adev->gfx.config.max_hw_contexts = 8;
+
+		adev->gfx.config.sc_prim_fifo_size_frontend = 0x20;
+		adev->gfx.config.sc_prim_fifo_size_backend = 0x100;
+		adev->gfx.config.sc_hiz_tile_fifo_size = 0x30;
+		adev->gfx.config.sc_earlyz_tile_fifo_size = 0x130;
+		gb_addr_config = BONAIRE_GB_ADDR_CONFIG_GOLDEN;
+		break;
+	}
+
+	WREG32(mmGRBM_CNTL, (0xff << GRBM_CNTL__READ_TIMEOUT__SHIFT));
+
+	mc_shared_chmap = RREG32(mmMC_SHARED_CHMAP);
+	adev->gfx.config.mc_arb_ramcfg = RREG32(mmMC_ARB_RAMCFG);
+	mc_arb_ramcfg = adev->gfx.config.mc_arb_ramcfg;
+
+	adev->gfx.config.num_tile_pipes = adev->gfx.config.max_tile_pipes;
+	adev->gfx.config.mem_max_burst_length_bytes = 256;
+	if (adev->flags & AMDGPU_IS_APU) {
+		/* Get memory bank mapping mode. */
+		tmp = RREG32(mmMC_FUS_DRAM0_BANK_ADDR_MAPPING);
+		dimm00_addr_map = REG_GET_FIELD(tmp, MC_FUS_DRAM0_BANK_ADDR_MAPPING, DIMM0ADDRMAP);
+		dimm01_addr_map = REG_GET_FIELD(tmp, MC_FUS_DRAM0_BANK_ADDR_MAPPING, DIMM1ADDRMAP);
+
+		tmp = RREG32(mmMC_FUS_DRAM1_BANK_ADDR_MAPPING);
+		dimm10_addr_map = REG_GET_FIELD(tmp, MC_FUS_DRAM1_BANK_ADDR_MAPPING, DIMM0ADDRMAP);
+		dimm11_addr_map = REG_GET_FIELD(tmp, MC_FUS_DRAM1_BANK_ADDR_MAPPING, DIMM1ADDRMAP);
+
+		/* Validate settings in case only one DIMM installed. */
+		if ((dimm00_addr_map == 0) || (dimm00_addr_map == 3) || (dimm00_addr_map == 4) || (dimm00_addr_map > 12))
+			dimm00_addr_map = 0;
+		if ((dimm01_addr_map == 0) || (dimm01_addr_map == 3) || (dimm01_addr_map == 4) || (dimm01_addr_map > 12))
+			dimm01_addr_map = 0;
+		if ((dimm10_addr_map == 0) || (dimm10_addr_map == 3) || (dimm10_addr_map == 4) || (dimm10_addr_map > 12))
+			dimm10_addr_map = 0;
+		if ((dimm11_addr_map == 0) || (dimm11_addr_map == 3) || (dimm11_addr_map == 4) || (dimm11_addr_map > 12))
+			dimm11_addr_map = 0;
+
+		/* If DIMM Addr map is 8GB, ROW size should be 2KB. Otherwise 1KB. */
+		/* If ROW size(DIMM1) != ROW size(DMIMM0), ROW size should be larger one. */
+		if ((dimm00_addr_map == 11) || (dimm01_addr_map == 11) || (dimm10_addr_map == 11) || (dimm11_addr_map == 11))
+			adev->gfx.config.mem_row_size_in_kb = 2;
+		else
+			adev->gfx.config.mem_row_size_in_kb = 1;
+	} else {
+		tmp = (mc_arb_ramcfg & MC_ARB_RAMCFG__NOOFCOLS_MASK) >> MC_ARB_RAMCFG__NOOFCOLS__SHIFT;
+		adev->gfx.config.mem_row_size_in_kb = (4 * (1 << (8 + tmp))) / 1024;
+		if (adev->gfx.config.mem_row_size_in_kb > 4)
+			adev->gfx.config.mem_row_size_in_kb = 4;
+	}
+	/* XXX use MC settings? */
+	adev->gfx.config.shader_engine_tile_size = 32;
+	adev->gfx.config.num_gpus = 1;
+	adev->gfx.config.multi_gpu_tile_size = 64;
+
+	/* fix up row size */
+	gb_addr_config &= ~GB_ADDR_CONFIG__ROW_SIZE_MASK;
+	switch (adev->gfx.config.mem_row_size_in_kb) {
+	case 1:
+	default:
+		gb_addr_config |= (0 << GB_ADDR_CONFIG__ROW_SIZE__SHIFT);
+		break;
+	case 2:
+		gb_addr_config |= (1 << GB_ADDR_CONFIG__ROW_SIZE__SHIFT);
+		break;
+	case 4:
+		gb_addr_config |= (2 << GB_ADDR_CONFIG__ROW_SIZE__SHIFT);
+		break;
+	}
+	adev->gfx.config.gb_addr_config = gb_addr_config;
+
+	WREG32(mmGB_ADDR_CONFIG, gb_addr_config);
+	WREG32(mmHDP_ADDR_CONFIG, gb_addr_config);
+	WREG32(mmDMIF_ADDR_CALC, gb_addr_config);
+	WREG32(mmSDMA0_TILING_CONFIG + SDMA0_REGISTER_OFFSET, gb_addr_config & 0x70);
+	WREG32(mmSDMA0_TILING_CONFIG + SDMA1_REGISTER_OFFSET, gb_addr_config & 0x70);
+	WREG32(mmUVD_UDEC_ADDR_CONFIG, gb_addr_config);
+	WREG32(mmUVD_UDEC_DB_ADDR_CONFIG, gb_addr_config);
+	WREG32(mmUVD_UDEC_DBW_ADDR_CONFIG, gb_addr_config);
+
+	gfx_v7_0_tiling_mode_table_init(adev);
+
+	gfx_v7_0_setup_rb(adev, adev->gfx.config.max_shader_engines,
+			  adev->gfx.config.max_sh_per_se,
+			  adev->gfx.config.max_backends_per_se);
+
+	/* set HW defaults for 3D engine */
+	WREG32(mmCP_MEQ_THRESHOLDS,
+			(0x30 << CP_MEQ_THRESHOLDS__MEQ1_START__SHIFT) |
+			(0x60 << CP_MEQ_THRESHOLDS__MEQ2_START__SHIFT));
+
+	mutex_lock(&adev->grbm_idx_mutex);
+	/*
+	 * making sure that the following register writes will be broadcasted
+	 * to all the shaders
+	 */
+	gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff);
+
+	/* XXX SH_MEM regs */
+	/* where to put LDS, scratch, GPUVM in FSA64 space */
+	mutex_lock(&adev->srbm_mutex);
+	for (i = 0; i < 16; i++) {
+		cik_srbm_select(adev, 0, 0, 0, i);
+		/* CP and shaders */
+		WREG32(mmSH_MEM_CONFIG, 0);
+		WREG32(mmSH_MEM_APE1_BASE, 1);
+		WREG32(mmSH_MEM_APE1_LIMIT, 0);
+		WREG32(mmSH_MEM_BASES, 0);
+	}
+	cik_srbm_select(adev, 0, 0, 0, 0);
+	mutex_unlock(&adev->srbm_mutex);
+
+	WREG32(mmSX_DEBUG_1, 0x20);
+
+	WREG32(mmTA_CNTL_AUX, 0x00010000);
+
+	tmp = RREG32(mmSPI_CONFIG_CNTL);
+	tmp |= 0x03000000;
+	WREG32(mmSPI_CONFIG_CNTL, tmp);
+
+	WREG32(mmSQ_CONFIG, 1);
+
+	WREG32(mmDB_DEBUG, 0);
+
+	tmp = RREG32(mmDB_DEBUG2) & ~0xf00fffff;
+	tmp |= 0x00000400;
+	WREG32(mmDB_DEBUG2, tmp);
+
+	tmp = RREG32(mmDB_DEBUG3) & ~0x0002021c;
+	tmp |= 0x00020200;
+	WREG32(mmDB_DEBUG3, tmp);
+
+	tmp = RREG32(mmCB_HW_CONTROL) & ~0x00010000;
+	tmp |= 0x00018208;
+	WREG32(mmCB_HW_CONTROL, tmp);
+
+	WREG32(mmSPI_CONFIG_CNTL_1, (4 << SPI_CONFIG_CNTL_1__VTX_DONE_DELAY__SHIFT));
+
+	WREG32(mmPA_SC_FIFO_SIZE,
+		((adev->gfx.config.sc_prim_fifo_size_frontend << PA_SC_FIFO_SIZE__SC_FRONTEND_PRIM_FIFO_SIZE__SHIFT) |
+		(adev->gfx.config.sc_prim_fifo_size_backend << PA_SC_FIFO_SIZE__SC_BACKEND_PRIM_FIFO_SIZE__SHIFT) |
+		(adev->gfx.config.sc_hiz_tile_fifo_size << PA_SC_FIFO_SIZE__SC_HIZ_TILE_FIFO_SIZE__SHIFT) |
+		(adev->gfx.config.sc_earlyz_tile_fifo_size << PA_SC_FIFO_SIZE__SC_EARLYZ_TILE_FIFO_SIZE__SHIFT)));
+
+	WREG32(mmVGT_NUM_INSTANCES, 1);
+
+	WREG32(mmCP_PERFMON_CNTL, 0);
+
+	WREG32(mmSQ_CONFIG, 0);
+
+	WREG32(mmPA_SC_FORCE_EOV_MAX_CNTS,
+		((4095 << PA_SC_FORCE_EOV_MAX_CNTS__FORCE_EOV_MAX_CLK_CNT__SHIFT) |
+		(255 << PA_SC_FORCE_EOV_MAX_CNTS__FORCE_EOV_MAX_REZ_CNT__SHIFT)));
+
+	WREG32(mmVGT_CACHE_INVALIDATION,
+		(VC_AND_TC << VGT_CACHE_INVALIDATION__CACHE_INVALIDATION__SHIFT) |
+		(ES_AND_GS_AUTO << VGT_CACHE_INVALIDATION__AUTO_INVLD_EN__SHIFT));
+
+	WREG32(mmVGT_GS_VERTEX_REUSE, 16);
+	WREG32(mmPA_SC_LINE_STIPPLE_STATE, 0);
+
+	WREG32(mmPA_CL_ENHANCE, PA_CL_ENHANCE__CLIP_VTX_REORDER_ENA_MASK |
+			(3 << PA_CL_ENHANCE__NUM_CLIP_SEQ__SHIFT));
+	WREG32(mmPA_SC_ENHANCE, PA_SC_ENHANCE__ENABLE_PA_SC_OUT_OF_ORDER_MASK);
+	mutex_unlock(&adev->grbm_idx_mutex);
+
+	udelay(50);
+}
+
+/*
+ * GPU scratch registers helpers function.
+ */
+/**
+ * gfx_v7_0_scratch_init - setup driver info for CP scratch regs
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Set up the number and offset of the CP scratch registers.
+ * NOTE: use of CP scratch registers is a legacy inferface and
+ * is not used by default on newer asics (r6xx+).  On newer asics,
+ * memory buffers are used for fences rather than scratch regs.
+ */
+static void gfx_v7_0_scratch_init(struct amdgpu_device *adev)
+{
+	int i;
+
+	adev->gfx.scratch.num_reg = 7;
+	adev->gfx.scratch.reg_base = mmSCRATCH_REG0;
+	for (i = 0; i < adev->gfx.scratch.num_reg; i++) {
+		adev->gfx.scratch.free[i] = true;
+		adev->gfx.scratch.reg[i] = adev->gfx.scratch.reg_base + i;
+	}
+}
+
+/**
+ * gfx_v7_0_ring_test_ring - basic gfx ring test
+ *
+ * @adev: amdgpu_device pointer
+ * @ring: amdgpu_ring structure holding ring information
+ *
+ * Allocate a scratch register and write to it using the gfx ring (CIK).
+ * Provides a basic gfx ring test to verify that the ring is working.
+ * Used by gfx_v7_0_cp_gfx_resume();
+ * Returns 0 on success, error on failure.
+ */
+static int gfx_v7_0_ring_test_ring(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	uint32_t scratch;
+	uint32_t tmp = 0;
+	unsigned i;
+	int r;
+
+	r = amdgpu_gfx_scratch_get(adev, &scratch);
+	if (r) {
+		DRM_ERROR("amdgpu: cp failed to get scratch reg (%d).\n", r);
+		return r;
+	}
+	WREG32(scratch, 0xCAFEDEAD);
+	r = amdgpu_ring_lock(ring, 3);
+	if (r) {
+		DRM_ERROR("amdgpu: cp failed to lock ring %d (%d).\n", ring->idx, r);
+		amdgpu_gfx_scratch_free(adev, scratch);
+		return r;
+	}
+	amdgpu_ring_write(ring, PACKET3(PACKET3_SET_UCONFIG_REG, 1));
+	amdgpu_ring_write(ring, (scratch - PACKET3_SET_UCONFIG_REG_START));
+	amdgpu_ring_write(ring, 0xDEADBEEF);
+	amdgpu_ring_unlock_commit(ring);
+
+	for (i = 0; i < adev->usec_timeout; i++) {
+		tmp = RREG32(scratch);
+		if (tmp == 0xDEADBEEF)
+			break;
+		DRM_UDELAY(1);
+	}
+	if (i < adev->usec_timeout) {
+		DRM_INFO("ring test on %d succeeded in %d usecs\n", ring->idx, i);
+	} else {
+		DRM_ERROR("amdgpu: ring %d test failed (scratch(0x%04X)=0x%08X)\n",
+			  ring->idx, scratch, tmp);
+		r = -EINVAL;
+	}
+	amdgpu_gfx_scratch_free(adev, scratch);
+	return r;
+}
+
+/**
+ * gfx_v7_0_hdp_flush_cp_ring_emit - emit an hdp flush on the cp
+ *
+ * @adev: amdgpu_device pointer
+ * @ridx: amdgpu ring index
+ *
+ * Emits an hdp flush on the cp.
+ */
+static void gfx_v7_0_hdp_flush_cp_ring_emit(struct amdgpu_ring *ring)
+{
+	u32 ref_and_mask;
+
+	if (ring->type == AMDGPU_RING_TYPE_COMPUTE) {
+		switch (ring->me) {
+		case 1:
+			ref_and_mask = GPU_HDP_FLUSH_DONE__CP2_MASK << ring->pipe;
+			break;
+		case 2:
+			ref_and_mask = GPU_HDP_FLUSH_DONE__CP6_MASK << ring->pipe;
+			break;
+		default:
+			return;
+		}
+	} else {
+		ref_and_mask = GPU_HDP_FLUSH_DONE__CP0_MASK;
+	}
+
+	amdgpu_ring_write(ring, PACKET3(PACKET3_WAIT_REG_MEM, 5));
+	amdgpu_ring_write(ring, (WAIT_REG_MEM_OPERATION(1) | /* write, wait, write */
+				 WAIT_REG_MEM_FUNCTION(3) |  /* == */
+				 WAIT_REG_MEM_ENGINE(1)));   /* pfp */
+	amdgpu_ring_write(ring, mmGPU_HDP_FLUSH_REQ);
+	amdgpu_ring_write(ring, mmGPU_HDP_FLUSH_DONE);
+	amdgpu_ring_write(ring, ref_and_mask);
+	amdgpu_ring_write(ring, ref_and_mask);
+	amdgpu_ring_write(ring, 0x20); /* poll interval */
+}
+
+/**
+ * gfx_v7_0_ring_emit_fence_gfx - emit a fence on the gfx ring
+ *
+ * @adev: amdgpu_device pointer
+ * @fence: amdgpu fence object
+ *
+ * Emits a fence sequnce number on the gfx ring and flushes
+ * GPU caches.
+ */
+static void gfx_v7_0_ring_emit_fence_gfx(struct amdgpu_ring *ring, u64 addr,
+					 u64 seq, bool write64bit)
+{
+	/* Workaround for cache flush problems. First send a dummy EOP
+	 * event down the pipe with seq one below.
+	 */
+	amdgpu_ring_write(ring, PACKET3(PACKET3_EVENT_WRITE_EOP, 4));
+	amdgpu_ring_write(ring, (EOP_TCL1_ACTION_EN |
+				 EOP_TC_ACTION_EN |
+				 EVENT_TYPE(CACHE_FLUSH_AND_INV_TS_EVENT) |
+				 EVENT_INDEX(5)));
+	amdgpu_ring_write(ring, addr & 0xfffffffc);
+	amdgpu_ring_write(ring, (upper_32_bits(addr) & 0xffff) |
+				DATA_SEL(1) | INT_SEL(0));
+	amdgpu_ring_write(ring, lower_32_bits(seq - 1));
+	amdgpu_ring_write(ring, upper_32_bits(seq - 1));
+
+	/* Then send the real EOP event down the pipe. */
+	amdgpu_ring_write(ring, PACKET3(PACKET3_EVENT_WRITE_EOP, 4));
+	amdgpu_ring_write(ring, (EOP_TCL1_ACTION_EN |
+				 EOP_TC_ACTION_EN |
+				 EVENT_TYPE(CACHE_FLUSH_AND_INV_TS_EVENT) |
+				 EVENT_INDEX(5)));
+	amdgpu_ring_write(ring, addr & 0xfffffffc);
+	amdgpu_ring_write(ring, (upper_32_bits(addr) & 0xffff) |
+				DATA_SEL(write64bit ? 2 : 1) | INT_SEL(2));
+	amdgpu_ring_write(ring, lower_32_bits(seq));
+	amdgpu_ring_write(ring, upper_32_bits(seq));
+}
+
+/**
+ * gfx_v7_0_ring_emit_fence_compute - emit a fence on the compute ring
+ *
+ * @adev: amdgpu_device pointer
+ * @fence: amdgpu fence object
+ *
+ * Emits a fence sequnce number on the compute ring and flushes
+ * GPU caches.
+ */
+static void gfx_v7_0_ring_emit_fence_compute(struct amdgpu_ring *ring,
+					     u64 addr, u64 seq,
+					     bool write64bits)
+{
+	/* RELEASE_MEM - flush caches, send int */
+	amdgpu_ring_write(ring, PACKET3(PACKET3_RELEASE_MEM, 5));
+	amdgpu_ring_write(ring, (EOP_TCL1_ACTION_EN |
+				 EOP_TC_ACTION_EN |
+				 EVENT_TYPE(CACHE_FLUSH_AND_INV_TS_EVENT) |
+				 EVENT_INDEX(5)));
+	amdgpu_ring_write(ring, DATA_SEL(write64bits ? 2 : 1) | INT_SEL(2));
+	amdgpu_ring_write(ring, addr & 0xfffffffc);
+	amdgpu_ring_write(ring, upper_32_bits(addr));
+	amdgpu_ring_write(ring, lower_32_bits(seq));
+	amdgpu_ring_write(ring, upper_32_bits(seq));
+}
+
+/**
+ * gfx_v7_0_ring_emit_semaphore - emit a semaphore on the CP ring
+ *
+ * @ring: amdgpu ring buffer object
+ * @semaphore: amdgpu semaphore object
+ * @emit_wait: Is this a sempahore wait?
+ *
+ * Emits a semaphore signal/wait packet to the CP ring and prevents the PFP
+ * from running ahead of semaphore waits.
+ */
+static bool gfx_v7_0_ring_emit_semaphore(struct amdgpu_ring *ring,
+					 struct amdgpu_semaphore *semaphore,
+					 bool emit_wait)
+{
+	uint64_t addr = semaphore->gpu_addr;
+	unsigned sel = emit_wait ? PACKET3_SEM_SEL_WAIT : PACKET3_SEM_SEL_SIGNAL;
+
+	amdgpu_ring_write(ring, PACKET3(PACKET3_MEM_SEMAPHORE, 1));
+	amdgpu_ring_write(ring, addr & 0xffffffff);
+	amdgpu_ring_write(ring, (upper_32_bits(addr) & 0xffff) | sel);
+
+	if (emit_wait && (ring->type == AMDGPU_RING_TYPE_GFX)) {
+		/* Prevent the PFP from running ahead of the semaphore wait */
+		amdgpu_ring_write(ring, PACKET3(PACKET3_PFP_SYNC_ME, 0));
+		amdgpu_ring_write(ring, 0x0);
+	}
+
+	return true;
+}
+
+/*
+ * IB stuff
+ */
+/**
+ * gfx_v7_0_ring_emit_ib - emit an IB (Indirect Buffer) on the ring
+ *
+ * @ring: amdgpu_ring structure holding ring information
+ * @ib: amdgpu indirect buffer object
+ *
+ * Emits an DE (drawing engine) or CE (constant engine) IB
+ * on the gfx ring.  IBs are usually generated by userspace
+ * acceleration drivers and submitted to the kernel for
+ * sheduling on the ring.  This function schedules the IB
+ * on the gfx ring for execution by the GPU.
+ */
+static void gfx_v7_0_ring_emit_ib(struct amdgpu_ring *ring,
+				  struct amdgpu_ib *ib)
+{
+	u32 header, control = 0;
+	u32 next_rptr = ring->wptr + 5;
+	if (ring->type == AMDGPU_RING_TYPE_COMPUTE)
+		control |= INDIRECT_BUFFER_VALID;
+
+	if (ib->flush_hdp_writefifo)
+		next_rptr += 7;
+
+	if (ring->need_ctx_switch && ring->type == AMDGPU_RING_TYPE_GFX)
+		next_rptr += 2;
+
+	next_rptr += 4;
+	amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3));
+	amdgpu_ring_write(ring, WRITE_DATA_DST_SEL(5) | WR_CONFIRM);
+	amdgpu_ring_write(ring, ring->next_rptr_gpu_addr & 0xfffffffc);
+	amdgpu_ring_write(ring, upper_32_bits(ring->next_rptr_gpu_addr) & 0xffffffff);
+	amdgpu_ring_write(ring, next_rptr);
+
+	if (ib->flush_hdp_writefifo)
+		gfx_v7_0_hdp_flush_cp_ring_emit(ring);
+
+	/* insert SWITCH_BUFFER packet before first IB in the ring frame */
+	if (ring->need_ctx_switch && ring->type == AMDGPU_RING_TYPE_GFX) {
+		amdgpu_ring_write(ring, PACKET3(PACKET3_SWITCH_BUFFER, 0));
+		amdgpu_ring_write(ring, 0);
+		ring->need_ctx_switch = false;
+	}
+
+	if (ib->is_const_ib)
+		header = PACKET3(PACKET3_INDIRECT_BUFFER_CONST, 2);
+	else
+		header = PACKET3(PACKET3_INDIRECT_BUFFER, 2);
+
+	control |= ib->length_dw |
+		(ib->vm ? (ib->vm->ids[ring->idx].id << 24) : 0);
+
+	amdgpu_ring_write(ring, header);
+	amdgpu_ring_write(ring,
+#ifdef __BIG_ENDIAN
+			  (2 << 0) |
+#endif
+			  (ib->gpu_addr & 0xFFFFFFFC));
+	amdgpu_ring_write(ring, upper_32_bits(ib->gpu_addr) & 0xFFFF);
+	amdgpu_ring_write(ring, control);
+}
+
+/**
+ * gfx_v7_0_ring_test_ib - basic ring IB test
+ *
+ * @ring: amdgpu_ring structure holding ring information
+ *
+ * Allocate an IB and execute it on the gfx ring (CIK).
+ * Provides a basic gfx ring test to verify that IBs are working.
+ * Returns 0 on success, error on failure.
+ */
+static int gfx_v7_0_ring_test_ib(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	struct amdgpu_ib ib;
+	uint32_t scratch;
+	uint32_t tmp = 0;
+	unsigned i;
+	int r;
+
+	r = amdgpu_gfx_scratch_get(adev, &scratch);
+	if (r) {
+		DRM_ERROR("amdgpu: failed to get scratch reg (%d).\n", r);
+		return r;
+	}
+	WREG32(scratch, 0xCAFEDEAD);
+	r = amdgpu_ib_get(ring, NULL, 256, &ib);
+	if (r) {
+		DRM_ERROR("amdgpu: failed to get ib (%d).\n", r);
+		amdgpu_gfx_scratch_free(adev, scratch);
+		return r;
+	}
+	ib.ptr[0] = PACKET3(PACKET3_SET_UCONFIG_REG, 1);
+	ib.ptr[1] = ((scratch - PACKET3_SET_UCONFIG_REG_START));
+	ib.ptr[2] = 0xDEADBEEF;
+	ib.length_dw = 3;
+	r = amdgpu_ib_schedule(adev, 1, &ib, AMDGPU_FENCE_OWNER_UNDEFINED);
+	if (r) {
+		amdgpu_gfx_scratch_free(adev, scratch);
+		amdgpu_ib_free(adev, &ib);
+		DRM_ERROR("amdgpu: failed to schedule ib (%d).\n", r);
+		return r;
+	}
+	r = amdgpu_fence_wait(ib.fence, false);
+	if (r) {
+		DRM_ERROR("amdgpu: fence wait failed (%d).\n", r);
+		amdgpu_gfx_scratch_free(adev, scratch);
+		amdgpu_ib_free(adev, &ib);
+		return r;
+	}
+	for (i = 0; i < adev->usec_timeout; i++) {
+		tmp = RREG32(scratch);
+		if (tmp == 0xDEADBEEF)
+			break;
+		DRM_UDELAY(1);
+	}
+	if (i < adev->usec_timeout) {
+		DRM_INFO("ib test on ring %d succeeded in %u usecs\n",
+			 ib.fence->ring->idx, i);
+	} else {
+		DRM_ERROR("amdgpu: ib test failed (scratch(0x%04X)=0x%08X)\n",
+			  scratch, tmp);
+		r = -EINVAL;
+	}
+	amdgpu_gfx_scratch_free(adev, scratch);
+	amdgpu_ib_free(adev, &ib);
+	return r;
+}
+
+/*
+ * CP.
+ * On CIK, gfx and compute now have independant command processors.
+ *
+ * GFX
+ * Gfx consists of a single ring and can process both gfx jobs and
+ * compute jobs.  The gfx CP consists of three microengines (ME):
+ * PFP - Pre-Fetch Parser
+ * ME - Micro Engine
+ * CE - Constant Engine
+ * The PFP and ME make up what is considered the Drawing Engine (DE).
+ * The CE is an asynchronous engine used for updating buffer desciptors
+ * used by the DE so that they can be loaded into cache in parallel
+ * while the DE is processing state update packets.
+ *
+ * Compute
+ * The compute CP consists of two microengines (ME):
+ * MEC1 - Compute MicroEngine 1
+ * MEC2 - Compute MicroEngine 2
+ * Each MEC supports 4 compute pipes and each pipe supports 8 queues.
+ * The queues are exposed to userspace and are programmed directly
+ * by the compute runtime.
+ */
+/**
+ * gfx_v7_0_cp_gfx_enable - enable/disable the gfx CP MEs
+ *
+ * @adev: amdgpu_device pointer
+ * @enable: enable or disable the MEs
+ *
+ * Halts or unhalts the gfx MEs.
+ */
+static void gfx_v7_0_cp_gfx_enable(struct amdgpu_device *adev, bool enable)
+{
+	int i;
+
+	if (enable) {
+		WREG32(mmCP_ME_CNTL, 0);
+	} else {
+		WREG32(mmCP_ME_CNTL, (CP_ME_CNTL__ME_HALT_MASK | CP_ME_CNTL__PFP_HALT_MASK | CP_ME_CNTL__CE_HALT_MASK));
+		for (i = 0; i < adev->gfx.num_gfx_rings; i++)
+			adev->gfx.gfx_ring[i].ready = false;
+	}
+	udelay(50);
+}
+
+/**
+ * gfx_v7_0_cp_gfx_load_microcode - load the gfx CP ME ucode
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Loads the gfx PFP, ME, and CE ucode.
+ * Returns 0 for success, -EINVAL if the ucode is not available.
+ */
+static int gfx_v7_0_cp_gfx_load_microcode(struct amdgpu_device *adev)
+{
+	const struct gfx_firmware_header_v1_0 *pfp_hdr;
+	const struct gfx_firmware_header_v1_0 *ce_hdr;
+	const struct gfx_firmware_header_v1_0 *me_hdr;
+	const __le32 *fw_data;
+	unsigned i, fw_size;
+
+	if (!adev->gfx.me_fw || !adev->gfx.pfp_fw || !adev->gfx.ce_fw)
+		return -EINVAL;
+
+	pfp_hdr = (const struct gfx_firmware_header_v1_0 *)adev->gfx.pfp_fw->data;
+	ce_hdr = (const struct gfx_firmware_header_v1_0 *)adev->gfx.ce_fw->data;
+	me_hdr = (const struct gfx_firmware_header_v1_0 *)adev->gfx.me_fw->data;
+
+	amdgpu_ucode_print_gfx_hdr(&pfp_hdr->header);
+	amdgpu_ucode_print_gfx_hdr(&ce_hdr->header);
+	amdgpu_ucode_print_gfx_hdr(&me_hdr->header);
+	adev->gfx.pfp_fw_version = le32_to_cpu(pfp_hdr->header.ucode_version);
+	adev->gfx.ce_fw_version = le32_to_cpu(ce_hdr->header.ucode_version);
+	adev->gfx.me_fw_version = le32_to_cpu(me_hdr->header.ucode_version);
+
+	gfx_v7_0_cp_gfx_enable(adev, false);
+
+	/* PFP */
+	fw_data = (const __le32 *)
+		(adev->gfx.pfp_fw->data +
+		 le32_to_cpu(pfp_hdr->header.ucode_array_offset_bytes));
+	fw_size = le32_to_cpu(pfp_hdr->header.ucode_size_bytes) / 4;
+	WREG32(mmCP_PFP_UCODE_ADDR, 0);
+	for (i = 0; i < fw_size; i++)
+		WREG32(mmCP_PFP_UCODE_DATA, le32_to_cpup(fw_data++));
+	WREG32(mmCP_PFP_UCODE_ADDR, adev->gfx.pfp_fw_version);
+
+	/* CE */
+	fw_data = (const __le32 *)
+		(adev->gfx.ce_fw->data +
+		 le32_to_cpu(ce_hdr->header.ucode_array_offset_bytes));
+	fw_size = le32_to_cpu(ce_hdr->header.ucode_size_bytes) / 4;
+	WREG32(mmCP_CE_UCODE_ADDR, 0);
+	for (i = 0; i < fw_size; i++)
+		WREG32(mmCP_CE_UCODE_DATA, le32_to_cpup(fw_data++));
+	WREG32(mmCP_CE_UCODE_ADDR, adev->gfx.ce_fw_version);
+
+	/* ME */
+	fw_data = (const __le32 *)
+		(adev->gfx.me_fw->data +
+		 le32_to_cpu(me_hdr->header.ucode_array_offset_bytes));
+	fw_size = le32_to_cpu(me_hdr->header.ucode_size_bytes) / 4;
+	WREG32(mmCP_ME_RAM_WADDR, 0);
+	for (i = 0; i < fw_size; i++)
+		WREG32(mmCP_ME_RAM_DATA, le32_to_cpup(fw_data++));
+	WREG32(mmCP_ME_RAM_WADDR, adev->gfx.me_fw_version);
+
+	return 0;
+}
+
+/**
+ * gfx_v7_0_cp_gfx_start - start the gfx ring
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Enables the ring and loads the clear state context and other
+ * packets required to init the ring.
+ * Returns 0 for success, error for failure.
+ */
+static int gfx_v7_0_cp_gfx_start(struct amdgpu_device *adev)
+{
+	struct amdgpu_ring *ring = &adev->gfx.gfx_ring[0];
+	const struct cs_section_def *sect = NULL;
+	const struct cs_extent_def *ext = NULL;
+	int r, i;
+
+	/* init the CP */
+	WREG32(mmCP_MAX_CONTEXT, adev->gfx.config.max_hw_contexts - 1);
+	WREG32(mmCP_ENDIAN_SWAP, 0);
+	WREG32(mmCP_DEVICE_ID, 1);
+
+	gfx_v7_0_cp_gfx_enable(adev, true);
+
+	r = amdgpu_ring_lock(ring, gfx_v7_0_get_csb_size(adev) + 8);
+	if (r) {
+		DRM_ERROR("amdgpu: cp failed to lock ring (%d).\n", r);
+		return r;
+	}
+
+	/* init the CE partitions.  CE only used for gfx on CIK */
+	amdgpu_ring_write(ring, PACKET3(PACKET3_SET_BASE, 2));
+	amdgpu_ring_write(ring, PACKET3_BASE_INDEX(CE_PARTITION_BASE));
+	amdgpu_ring_write(ring, 0x8000);
+	amdgpu_ring_write(ring, 0x8000);
+
+	/* clear state buffer */
+	amdgpu_ring_write(ring, PACKET3(PACKET3_PREAMBLE_CNTL, 0));
+	amdgpu_ring_write(ring, PACKET3_PREAMBLE_BEGIN_CLEAR_STATE);
+
+	amdgpu_ring_write(ring, PACKET3(PACKET3_CONTEXT_CONTROL, 1));
+	amdgpu_ring_write(ring, 0x80000000);
+	amdgpu_ring_write(ring, 0x80000000);
+
+	for (sect = adev->gfx.rlc.cs_data; sect->section != NULL; ++sect) {
+		for (ext = sect->section; ext->extent != NULL; ++ext) {
+			if (sect->id == SECT_CONTEXT) {
+				amdgpu_ring_write(ring,
+						  PACKET3(PACKET3_SET_CONTEXT_REG, ext->reg_count));
+				amdgpu_ring_write(ring, ext->reg_index - PACKET3_SET_CONTEXT_REG_START);
+				for (i = 0; i < ext->reg_count; i++)
+					amdgpu_ring_write(ring, ext->extent[i]);
+			}
+		}
+	}
+
+	amdgpu_ring_write(ring, PACKET3(PACKET3_SET_CONTEXT_REG, 2));
+	amdgpu_ring_write(ring, mmPA_SC_RASTER_CONFIG - PACKET3_SET_CONTEXT_REG_START);
+	switch (adev->asic_type) {
+	case CHIP_BONAIRE:
+		amdgpu_ring_write(ring, 0x16000012);
+		amdgpu_ring_write(ring, 0x00000000);
+		break;
+	case CHIP_KAVERI:
+		amdgpu_ring_write(ring, 0x00000000); /* XXX */
+		amdgpu_ring_write(ring, 0x00000000);
+		break;
+	case CHIP_KABINI:
+	case CHIP_MULLINS:
+		amdgpu_ring_write(ring, 0x00000000); /* XXX */
+		amdgpu_ring_write(ring, 0x00000000);
+		break;
+	case CHIP_HAWAII:
+		amdgpu_ring_write(ring, 0x3a00161a);
+		amdgpu_ring_write(ring, 0x0000002e);
+		break;
+	default:
+		amdgpu_ring_write(ring, 0x00000000);
+		amdgpu_ring_write(ring, 0x00000000);
+		break;
+	}
+
+	amdgpu_ring_write(ring, PACKET3(PACKET3_PREAMBLE_CNTL, 0));
+	amdgpu_ring_write(ring, PACKET3_PREAMBLE_END_CLEAR_STATE);
+
+	amdgpu_ring_write(ring, PACKET3(PACKET3_CLEAR_STATE, 0));
+	amdgpu_ring_write(ring, 0);
+
+	amdgpu_ring_write(ring, PACKET3(PACKET3_SET_CONTEXT_REG, 2));
+	amdgpu_ring_write(ring, 0x00000316);
+	amdgpu_ring_write(ring, 0x0000000e); /* VGT_VERTEX_REUSE_BLOCK_CNTL */
+	amdgpu_ring_write(ring, 0x00000010); /* VGT_OUT_DEALLOC_CNTL */
+
+	amdgpu_ring_unlock_commit(ring);
+
+	return 0;
+}
+
+/**
+ * gfx_v7_0_cp_gfx_resume - setup the gfx ring buffer registers
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Program the location and size of the gfx ring buffer
+ * and test it to make sure it's working.
+ * Returns 0 for success, error for failure.
+ */
+static int gfx_v7_0_cp_gfx_resume(struct amdgpu_device *adev)
+{
+	struct amdgpu_ring *ring;
+	u32 tmp;
+	u32 rb_bufsz;
+	u64 rb_addr, rptr_addr;
+	int r;
+
+	WREG32(mmCP_SEM_WAIT_TIMER, 0x0);
+	if (adev->asic_type != CHIP_HAWAII)
+		WREG32(mmCP_SEM_INCOMPLETE_TIMER_CNTL, 0x0);
+
+	/* Set the write pointer delay */
+	WREG32(mmCP_RB_WPTR_DELAY, 0);
+
+	/* set the RB to use vmid 0 */
+	WREG32(mmCP_RB_VMID, 0);
+
+	WREG32(mmSCRATCH_ADDR, 0);
+
+	/* ring 0 - compute and gfx */
+	/* Set ring buffer size */
+	ring = &adev->gfx.gfx_ring[0];
+	rb_bufsz = order_base_2(ring->ring_size / 8);
+	tmp = (order_base_2(AMDGPU_GPU_PAGE_SIZE/8) << 8) | rb_bufsz;
+#ifdef __BIG_ENDIAN
+	tmp |= BUF_SWAP_32BIT;
+#endif
+	WREG32(mmCP_RB0_CNTL, tmp);
+
+	/* Initialize the ring buffer's read and write pointers */
+	WREG32(mmCP_RB0_CNTL, tmp | CP_RB0_CNTL__RB_RPTR_WR_ENA_MASK);
+	ring->wptr = 0;
+	WREG32(mmCP_RB0_WPTR, ring->wptr);
+
+	/* set the wb address wether it's enabled or not */
+	rptr_addr = adev->wb.gpu_addr + (ring->rptr_offs * 4);
+	WREG32(mmCP_RB0_RPTR_ADDR, lower_32_bits(rptr_addr));
+	WREG32(mmCP_RB0_RPTR_ADDR_HI, upper_32_bits(rptr_addr) & 0xFF);
+
+	/* scratch register shadowing is no longer supported */
+	WREG32(mmSCRATCH_UMSK, 0);
+
+	mdelay(1);
+	WREG32(mmCP_RB0_CNTL, tmp);
+
+	rb_addr = ring->gpu_addr >> 8;
+	WREG32(mmCP_RB0_BASE, rb_addr);
+	WREG32(mmCP_RB0_BASE_HI, upper_32_bits(rb_addr));
+
+	/* start the ring */
+	gfx_v7_0_cp_gfx_start(adev);
+	ring->ready = true;
+	r = amdgpu_ring_test_ring(ring);
+	if (r) {
+		ring->ready = false;
+		return r;
+	}
+
+	return 0;
+}
+
+static u32 gfx_v7_0_ring_get_rptr_gfx(struct amdgpu_ring *ring)
+{
+	u32 rptr;
+
+	rptr = ring->adev->wb.wb[ring->rptr_offs];
+
+	return rptr;
+}
+
+static u32 gfx_v7_0_ring_get_wptr_gfx(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	u32 wptr;
+
+	wptr = RREG32(mmCP_RB0_WPTR);
+
+	return wptr;
+}
+
+static void gfx_v7_0_ring_set_wptr_gfx(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+
+	WREG32(mmCP_RB0_WPTR, ring->wptr);
+	(void)RREG32(mmCP_RB0_WPTR);
+}
+
+static u32 gfx_v7_0_ring_get_rptr_compute(struct amdgpu_ring *ring)
+{
+	u32 rptr;
+
+	rptr = ring->adev->wb.wb[ring->rptr_offs];
+
+	return rptr;
+}
+
+static u32 gfx_v7_0_ring_get_wptr_compute(struct amdgpu_ring *ring)
+{
+	u32 wptr;
+
+	/* XXX check if swapping is necessary on BE */
+	wptr = ring->adev->wb.wb[ring->wptr_offs];
+
+	return wptr;
+}
+
+static void gfx_v7_0_ring_set_wptr_compute(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+
+	/* XXX check if swapping is necessary on BE */
+	adev->wb.wb[ring->wptr_offs] = ring->wptr;
+	WDOORBELL32(ring->doorbell_index, ring->wptr);
+}
+
+/**
+ * gfx_v7_0_cp_compute_enable - enable/disable the compute CP MEs
+ *
+ * @adev: amdgpu_device pointer
+ * @enable: enable or disable the MEs
+ *
+ * Halts or unhalts the compute MEs.
+ */
+static void gfx_v7_0_cp_compute_enable(struct amdgpu_device *adev, bool enable)
+{
+	int i;
+
+	if (enable) {
+		WREG32(mmCP_MEC_CNTL, 0);
+	} else {
+		WREG32(mmCP_MEC_CNTL, (CP_MEC_CNTL__MEC_ME1_HALT_MASK | CP_MEC_CNTL__MEC_ME2_HALT_MASK));
+		for (i = 0; i < adev->gfx.num_compute_rings; i++)
+			adev->gfx.compute_ring[i].ready = false;
+	}
+	udelay(50);
+}
+
+/**
+ * gfx_v7_0_cp_compute_load_microcode - load the compute CP ME ucode
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Loads the compute MEC1&2 ucode.
+ * Returns 0 for success, -EINVAL if the ucode is not available.
+ */
+static int gfx_v7_0_cp_compute_load_microcode(struct amdgpu_device *adev)
+{
+	const struct gfx_firmware_header_v1_0 *mec_hdr;
+	const __le32 *fw_data;
+	unsigned i, fw_size;
+
+	if (!adev->gfx.mec_fw)
+		return -EINVAL;
+
+	mec_hdr = (const struct gfx_firmware_header_v1_0 *)adev->gfx.mec_fw->data;
+	amdgpu_ucode_print_gfx_hdr(&mec_hdr->header);
+	adev->gfx.mec_fw_version = le32_to_cpu(mec_hdr->header.ucode_version);
+
+	gfx_v7_0_cp_compute_enable(adev, false);
+
+	/* MEC1 */
+	fw_data = (const __le32 *)
+		(adev->gfx.mec_fw->data +
+		 le32_to_cpu(mec_hdr->header.ucode_array_offset_bytes));
+	fw_size = le32_to_cpu(mec_hdr->header.ucode_size_bytes) / 4;
+	WREG32(mmCP_MEC_ME1_UCODE_ADDR, 0);
+	for (i = 0; i < fw_size; i++)
+		WREG32(mmCP_MEC_ME1_UCODE_DATA, le32_to_cpup(fw_data++));
+	WREG32(mmCP_MEC_ME1_UCODE_ADDR, 0);
+
+	if (adev->asic_type == CHIP_KAVERI) {
+		const struct gfx_firmware_header_v1_0 *mec2_hdr;
+
+		if (!adev->gfx.mec2_fw)
+			return -EINVAL;
+
+		mec2_hdr = (const struct gfx_firmware_header_v1_0 *)adev->gfx.mec2_fw->data;
+		amdgpu_ucode_print_gfx_hdr(&mec2_hdr->header);
+		adev->gfx.mec2_fw_version = le32_to_cpu(mec2_hdr->header.ucode_version);
+
+		/* MEC2 */
+		fw_data = (const __le32 *)
+			(adev->gfx.mec2_fw->data +
+			 le32_to_cpu(mec2_hdr->header.ucode_array_offset_bytes));
+		fw_size = le32_to_cpu(mec2_hdr->header.ucode_size_bytes) / 4;
+		WREG32(mmCP_MEC_ME2_UCODE_ADDR, 0);
+		for (i = 0; i < fw_size; i++)
+			WREG32(mmCP_MEC_ME2_UCODE_DATA, le32_to_cpup(fw_data++));
+		WREG32(mmCP_MEC_ME2_UCODE_ADDR, 0);
+	}
+
+	return 0;
+}
+
+/**
+ * gfx_v7_0_cp_compute_start - start the compute queues
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Enable the compute queues.
+ * Returns 0 for success, error for failure.
+ */
+static int gfx_v7_0_cp_compute_start(struct amdgpu_device *adev)
+{
+	gfx_v7_0_cp_compute_enable(adev, true);
+
+	return 0;
+}
+
+/**
+ * gfx_v7_0_cp_compute_fini - stop the compute queues
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Stop the compute queues and tear down the driver queue
+ * info.
+ */
+static void gfx_v7_0_cp_compute_fini(struct amdgpu_device *adev)
+{
+	int i, r;
+
+	for (i = 0; i < adev->gfx.num_compute_rings; i++) {
+		struct amdgpu_ring *ring = &adev->gfx.compute_ring[i];
+
+		if (ring->mqd_obj) {
+			r = amdgpu_bo_reserve(ring->mqd_obj, false);
+			if (unlikely(r != 0))
+				dev_warn(adev->dev, "(%d) reserve MQD bo failed\n", r);
+
+			amdgpu_bo_unpin(ring->mqd_obj);
+			amdgpu_bo_unreserve(ring->mqd_obj);
+
+			amdgpu_bo_unref(&ring->mqd_obj);
+			ring->mqd_obj = NULL;
+		}
+	}
+}
+
+static void gfx_v7_0_mec_fini(struct amdgpu_device *adev)
+{
+	int r;
+
+	if (adev->gfx.mec.hpd_eop_obj) {
+		r = amdgpu_bo_reserve(adev->gfx.mec.hpd_eop_obj, false);
+		if (unlikely(r != 0))
+			dev_warn(adev->dev, "(%d) reserve HPD EOP bo failed\n", r);
+		amdgpu_bo_unpin(adev->gfx.mec.hpd_eop_obj);
+		amdgpu_bo_unreserve(adev->gfx.mec.hpd_eop_obj);
+
+		amdgpu_bo_unref(&adev->gfx.mec.hpd_eop_obj);
+		adev->gfx.mec.hpd_eop_obj = NULL;
+	}
+}
+
+#define MEC_HPD_SIZE 2048
+
+static int gfx_v7_0_mec_init(struct amdgpu_device *adev)
+{
+	int r;
+	u32 *hpd;
+
+	/*
+	 * KV:    2 MEC, 4 Pipes/MEC, 8 Queues/Pipe - 64 Queues total
+	 * CI/KB: 1 MEC, 4 Pipes/MEC, 8 Queues/Pipe - 32 Queues total
+	 * Nonetheless, we assign only 1 pipe because all other pipes will
+	 * be handled by KFD
+	 */
+	adev->gfx.mec.num_mec = 1;
+	adev->gfx.mec.num_pipe = 1;
+	adev->gfx.mec.num_queue = adev->gfx.mec.num_mec * adev->gfx.mec.num_pipe * 8;
+
+	if (adev->gfx.mec.hpd_eop_obj == NULL) {
+		r = amdgpu_bo_create(adev,
+				     adev->gfx.mec.num_mec *adev->gfx.mec.num_pipe * MEC_HPD_SIZE * 2,
+				     PAGE_SIZE, true,
+				     AMDGPU_GEM_DOMAIN_GTT, 0, NULL,
+				     &adev->gfx.mec.hpd_eop_obj);
+		if (r) {
+			dev_warn(adev->dev, "(%d) create HDP EOP bo failed\n", r);
+			return r;
+		}
+	}
+
+	r = amdgpu_bo_reserve(adev->gfx.mec.hpd_eop_obj, false);
+	if (unlikely(r != 0)) {
+		gfx_v7_0_mec_fini(adev);
+		return r;
+	}
+	r = amdgpu_bo_pin(adev->gfx.mec.hpd_eop_obj, AMDGPU_GEM_DOMAIN_GTT,
+			  &adev->gfx.mec.hpd_eop_gpu_addr);
+	if (r) {
+		dev_warn(adev->dev, "(%d) pin HDP EOP bo failed\n", r);
+		gfx_v7_0_mec_fini(adev);
+		return r;
+	}
+	r = amdgpu_bo_kmap(adev->gfx.mec.hpd_eop_obj, (void **)&hpd);
+	if (r) {
+		dev_warn(adev->dev, "(%d) map HDP EOP bo failed\n", r);
+		gfx_v7_0_mec_fini(adev);
+		return r;
+	}
+
+	/* clear memory.  Not sure if this is required or not */
+	memset(hpd, 0, adev->gfx.mec.num_mec *adev->gfx.mec.num_pipe * MEC_HPD_SIZE * 2);
+
+	amdgpu_bo_kunmap(adev->gfx.mec.hpd_eop_obj);
+	amdgpu_bo_unreserve(adev->gfx.mec.hpd_eop_obj);
+
+	return 0;
+}
+
+struct hqd_registers
+{
+	u32 cp_mqd_base_addr;
+	u32 cp_mqd_base_addr_hi;
+	u32 cp_hqd_active;
+	u32 cp_hqd_vmid;
+	u32 cp_hqd_persistent_state;
+	u32 cp_hqd_pipe_priority;
+	u32 cp_hqd_queue_priority;
+	u32 cp_hqd_quantum;
+	u32 cp_hqd_pq_base;
+	u32 cp_hqd_pq_base_hi;
+	u32 cp_hqd_pq_rptr;
+	u32 cp_hqd_pq_rptr_report_addr;
+	u32 cp_hqd_pq_rptr_report_addr_hi;
+	u32 cp_hqd_pq_wptr_poll_addr;
+	u32 cp_hqd_pq_wptr_poll_addr_hi;
+	u32 cp_hqd_pq_doorbell_control;
+	u32 cp_hqd_pq_wptr;
+	u32 cp_hqd_pq_control;
+	u32 cp_hqd_ib_base_addr;
+	u32 cp_hqd_ib_base_addr_hi;
+	u32 cp_hqd_ib_rptr;
+	u32 cp_hqd_ib_control;
+	u32 cp_hqd_iq_timer;
+	u32 cp_hqd_iq_rptr;
+	u32 cp_hqd_dequeue_request;
+	u32 cp_hqd_dma_offload;
+	u32 cp_hqd_sema_cmd;
+	u32 cp_hqd_msg_type;
+	u32 cp_hqd_atomic0_preop_lo;
+	u32 cp_hqd_atomic0_preop_hi;
+	u32 cp_hqd_atomic1_preop_lo;
+	u32 cp_hqd_atomic1_preop_hi;
+	u32 cp_hqd_hq_scheduler0;
+	u32 cp_hqd_hq_scheduler1;
+	u32 cp_mqd_control;
+};
+
+struct bonaire_mqd
+{
+	u32 header;
+	u32 dispatch_initiator;
+	u32 dimensions[3];
+	u32 start_idx[3];
+	u32 num_threads[3];
+	u32 pipeline_stat_enable;
+	u32 perf_counter_enable;
+	u32 pgm[2];
+	u32 tba[2];
+	u32 tma[2];
+	u32 pgm_rsrc[2];
+	u32 vmid;
+	u32 resource_limits;
+	u32 static_thread_mgmt01[2];
+	u32 tmp_ring_size;
+	u32 static_thread_mgmt23[2];
+	u32 restart[3];
+	u32 thread_trace_enable;
+	u32 reserved1;
+	u32 user_data[16];
+	u32 vgtcs_invoke_count[2];
+	struct hqd_registers queue_state;
+	u32 dequeue_cntr;
+	u32 interrupt_queue[64];
+};
+
+/**
+ * gfx_v7_0_cp_compute_resume - setup the compute queue registers
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Program the compute queues and test them to make sure they
+ * are working.
+ * Returns 0 for success, error for failure.
+ */
+static int gfx_v7_0_cp_compute_resume(struct amdgpu_device *adev)
+{
+	int r, i, j;
+	u32 tmp;
+	bool use_doorbell = true;
+	u64 hqd_gpu_addr;
+	u64 mqd_gpu_addr;
+	u64 eop_gpu_addr;
+	u64 wb_gpu_addr;
+	u32 *buf;
+	struct bonaire_mqd *mqd;
+
+	r = gfx_v7_0_cp_compute_start(adev);
+	if (r)
+		return r;
+
+	/* fix up chicken bits */
+	tmp = RREG32(mmCP_CPF_DEBUG);
+	tmp |= (1 << 23);
+	WREG32(mmCP_CPF_DEBUG, tmp);
+
+	/* init the pipes */
+	mutex_lock(&adev->srbm_mutex);
+	for (i = 0; i < (adev->gfx.mec.num_pipe * adev->gfx.mec.num_mec); i++) {
+		int me = (i < 4) ? 1 : 2;
+		int pipe = (i < 4) ? i : (i - 4);
+
+		eop_gpu_addr = adev->gfx.mec.hpd_eop_gpu_addr + (i * MEC_HPD_SIZE * 2);
+
+		cik_srbm_select(adev, me, pipe, 0, 0);
+
+		/* write the EOP addr */
+		WREG32(mmCP_HPD_EOP_BASE_ADDR, eop_gpu_addr >> 8);
+		WREG32(mmCP_HPD_EOP_BASE_ADDR_HI, upper_32_bits(eop_gpu_addr) >> 8);
+
+		/* set the VMID assigned */
+		WREG32(mmCP_HPD_EOP_VMID, 0);
+
+		/* set the EOP size, register value is 2^(EOP_SIZE+1) dwords */
+		tmp = RREG32(mmCP_HPD_EOP_CONTROL);
+		tmp &= ~CP_HPD_EOP_CONTROL__EOP_SIZE_MASK;
+		tmp |= order_base_2(MEC_HPD_SIZE / 8);
+		WREG32(mmCP_HPD_EOP_CONTROL, tmp);
+	}
+	cik_srbm_select(adev, 0, 0, 0, 0);
+	mutex_unlock(&adev->srbm_mutex);
+
+	/* init the queues.  Just two for now. */
+	for (i = 0; i < adev->gfx.num_compute_rings; i++) {
+		struct amdgpu_ring *ring = &adev->gfx.compute_ring[i];
+
+		if (ring->mqd_obj == NULL) {
+			r = amdgpu_bo_create(adev,
+					     sizeof(struct bonaire_mqd),
+					     PAGE_SIZE, true,
+					     AMDGPU_GEM_DOMAIN_GTT, 0, NULL,
+					     &ring->mqd_obj);
+			if (r) {
+				dev_warn(adev->dev, "(%d) create MQD bo failed\n", r);
+				return r;
+			}
+		}
+
+		r = amdgpu_bo_reserve(ring->mqd_obj, false);
+		if (unlikely(r != 0)) {
+			gfx_v7_0_cp_compute_fini(adev);
+			return r;
+		}
+		r = amdgpu_bo_pin(ring->mqd_obj, AMDGPU_GEM_DOMAIN_GTT,
+				  &mqd_gpu_addr);
+		if (r) {
+			dev_warn(adev->dev, "(%d) pin MQD bo failed\n", r);
+			gfx_v7_0_cp_compute_fini(adev);
+			return r;
+		}
+		r = amdgpu_bo_kmap(ring->mqd_obj, (void **)&buf);
+		if (r) {
+			dev_warn(adev->dev, "(%d) map MQD bo failed\n", r);
+			gfx_v7_0_cp_compute_fini(adev);
+			return r;
+		}
+
+		/* init the mqd struct */
+		memset(buf, 0, sizeof(struct bonaire_mqd));
+
+		mqd = (struct bonaire_mqd *)buf;
+		mqd->header = 0xC0310800;
+		mqd->static_thread_mgmt01[0] = 0xffffffff;
+		mqd->static_thread_mgmt01[1] = 0xffffffff;
+		mqd->static_thread_mgmt23[0] = 0xffffffff;
+		mqd->static_thread_mgmt23[1] = 0xffffffff;
+
+		mutex_lock(&adev->srbm_mutex);
+		cik_srbm_select(adev, ring->me,
+				ring->pipe,
+				ring->queue, 0);
+
+		/* disable wptr polling */
+		tmp = RREG32(mmCP_PQ_WPTR_POLL_CNTL);
+		tmp &= ~CP_PQ_WPTR_POLL_CNTL__EN_MASK;
+		WREG32(mmCP_PQ_WPTR_POLL_CNTL, tmp);
+
+		/* enable doorbell? */
+		mqd->queue_state.cp_hqd_pq_doorbell_control =
+			RREG32(mmCP_HQD_PQ_DOORBELL_CONTROL);
+		if (use_doorbell)
+			mqd->queue_state.cp_hqd_pq_doorbell_control |= CP_HQD_PQ_DOORBELL_CONTROL__DOORBELL_EN_MASK;
+		else
+			mqd->queue_state.cp_hqd_pq_doorbell_control &= ~CP_HQD_PQ_DOORBELL_CONTROL__DOORBELL_EN_MASK;
+		WREG32(mmCP_HQD_PQ_DOORBELL_CONTROL,
+		       mqd->queue_state.cp_hqd_pq_doorbell_control);
+
+		/* disable the queue if it's active */
+		mqd->queue_state.cp_hqd_dequeue_request = 0;
+		mqd->queue_state.cp_hqd_pq_rptr = 0;
+		mqd->queue_state.cp_hqd_pq_wptr= 0;
+		if (RREG32(mmCP_HQD_ACTIVE) & 1) {
+			WREG32(mmCP_HQD_DEQUEUE_REQUEST, 1);
+			for (j = 0; j < adev->usec_timeout; j++) {
+				if (!(RREG32(mmCP_HQD_ACTIVE) & 1))
+					break;
+				udelay(1);
+			}
+			WREG32(mmCP_HQD_DEQUEUE_REQUEST, mqd->queue_state.cp_hqd_dequeue_request);
+			WREG32(mmCP_HQD_PQ_RPTR, mqd->queue_state.cp_hqd_pq_rptr);
+			WREG32(mmCP_HQD_PQ_WPTR, mqd->queue_state.cp_hqd_pq_wptr);
+		}
+
+		/* set the pointer to the MQD */
+		mqd->queue_state.cp_mqd_base_addr = mqd_gpu_addr & 0xfffffffc;
+		mqd->queue_state.cp_mqd_base_addr_hi = upper_32_bits(mqd_gpu_addr);
+		WREG32(mmCP_MQD_BASE_ADDR, mqd->queue_state.cp_mqd_base_addr);
+		WREG32(mmCP_MQD_BASE_ADDR_HI, mqd->queue_state.cp_mqd_base_addr_hi);
+		/* set MQD vmid to 0 */
+		mqd->queue_state.cp_mqd_control = RREG32(mmCP_MQD_CONTROL);
+		mqd->queue_state.cp_mqd_control &= ~CP_MQD_CONTROL__VMID_MASK;
+		WREG32(mmCP_MQD_CONTROL, mqd->queue_state.cp_mqd_control);
+
+		/* set the pointer to the HQD, this is similar CP_RB0_BASE/_HI */
+		hqd_gpu_addr = ring->gpu_addr >> 8;
+		mqd->queue_state.cp_hqd_pq_base = hqd_gpu_addr;
+		mqd->queue_state.cp_hqd_pq_base_hi = upper_32_bits(hqd_gpu_addr);
+		WREG32(mmCP_HQD_PQ_BASE, mqd->queue_state.cp_hqd_pq_base);
+		WREG32(mmCP_HQD_PQ_BASE_HI, mqd->queue_state.cp_hqd_pq_base_hi);
+
+		/* set up the HQD, this is similar to CP_RB0_CNTL */
+		mqd->queue_state.cp_hqd_pq_control = RREG32(mmCP_HQD_PQ_CONTROL);
+		mqd->queue_state.cp_hqd_pq_control &=
+			~(CP_HQD_PQ_CONTROL__QUEUE_SIZE_MASK |
+					CP_HQD_PQ_CONTROL__RPTR_BLOCK_SIZE_MASK);
+
+		mqd->queue_state.cp_hqd_pq_control |=
+			order_base_2(ring->ring_size / 8);
+		mqd->queue_state.cp_hqd_pq_control |=
+			(order_base_2(AMDGPU_GPU_PAGE_SIZE/8) << 8);
+#ifdef __BIG_ENDIAN
+		mqd->queue_state.cp_hqd_pq_control |= BUF_SWAP_32BIT;
+#endif
+		mqd->queue_state.cp_hqd_pq_control &=
+			~(CP_HQD_PQ_CONTROL__UNORD_DISPATCH_MASK |
+				CP_HQD_PQ_CONTROL__ROQ_PQ_IB_FLIP_MASK |
+				CP_HQD_PQ_CONTROL__PQ_VOLATILE_MASK);
+		mqd->queue_state.cp_hqd_pq_control |=
+			CP_HQD_PQ_CONTROL__PRIV_STATE_MASK |
+			CP_HQD_PQ_CONTROL__KMD_QUEUE_MASK; /* assuming kernel queue control */
+		WREG32(mmCP_HQD_PQ_CONTROL, mqd->queue_state.cp_hqd_pq_control);
+
+		/* only used if CP_PQ_WPTR_POLL_CNTL.CP_PQ_WPTR_POLL_CNTL__EN_MASK=1 */
+		wb_gpu_addr = adev->wb.gpu_addr + (ring->wptr_offs * 4);
+		mqd->queue_state.cp_hqd_pq_wptr_poll_addr = wb_gpu_addr & 0xfffffffc;
+		mqd->queue_state.cp_hqd_pq_wptr_poll_addr_hi = upper_32_bits(wb_gpu_addr) & 0xffff;
+		WREG32(mmCP_HQD_PQ_WPTR_POLL_ADDR, mqd->queue_state.cp_hqd_pq_wptr_poll_addr);
+		WREG32(mmCP_HQD_PQ_WPTR_POLL_ADDR_HI,
+		       mqd->queue_state.cp_hqd_pq_wptr_poll_addr_hi);
+
+		/* set the wb address wether it's enabled or not */
+		wb_gpu_addr = adev->wb.gpu_addr + (ring->rptr_offs * 4);
+		mqd->queue_state.cp_hqd_pq_rptr_report_addr = wb_gpu_addr & 0xfffffffc;
+		mqd->queue_state.cp_hqd_pq_rptr_report_addr_hi =
+			upper_32_bits(wb_gpu_addr) & 0xffff;
+		WREG32(mmCP_HQD_PQ_RPTR_REPORT_ADDR,
+		       mqd->queue_state.cp_hqd_pq_rptr_report_addr);
+		WREG32(mmCP_HQD_PQ_RPTR_REPORT_ADDR_HI,
+		       mqd->queue_state.cp_hqd_pq_rptr_report_addr_hi);
+
+		/* enable the doorbell if requested */
+		if (use_doorbell) {
+			mqd->queue_state.cp_hqd_pq_doorbell_control =
+				RREG32(mmCP_HQD_PQ_DOORBELL_CONTROL);
+			mqd->queue_state.cp_hqd_pq_doorbell_control &=
+				~CP_HQD_PQ_DOORBELL_CONTROL__DOORBELL_OFFSET_MASK;
+			mqd->queue_state.cp_hqd_pq_doorbell_control |=
+				(ring->doorbell_index <<
+				 CP_HQD_PQ_DOORBELL_CONTROL__DOORBELL_OFFSET__SHIFT);
+			mqd->queue_state.cp_hqd_pq_doorbell_control |=
+				CP_HQD_PQ_DOORBELL_CONTROL__DOORBELL_EN_MASK;
+			mqd->queue_state.cp_hqd_pq_doorbell_control &=
+				~(CP_HQD_PQ_DOORBELL_CONTROL__DOORBELL_SOURCE_MASK |
+				CP_HQD_PQ_DOORBELL_CONTROL__DOORBELL_HIT_MASK);
+
+		} else {
+			mqd->queue_state.cp_hqd_pq_doorbell_control = 0;
+		}
+		WREG32(mmCP_HQD_PQ_DOORBELL_CONTROL,
+		       mqd->queue_state.cp_hqd_pq_doorbell_control);
+
+		/* read and write pointers, similar to CP_RB0_WPTR/_RPTR */
+		ring->wptr = 0;
+		mqd->queue_state.cp_hqd_pq_wptr = ring->wptr;
+		WREG32(mmCP_HQD_PQ_WPTR, mqd->queue_state.cp_hqd_pq_wptr);
+		mqd->queue_state.cp_hqd_pq_rptr = RREG32(mmCP_HQD_PQ_RPTR);
+
+		/* set the vmid for the queue */
+		mqd->queue_state.cp_hqd_vmid = 0;
+		WREG32(mmCP_HQD_VMID, mqd->queue_state.cp_hqd_vmid);
+
+		/* activate the queue */
+		mqd->queue_state.cp_hqd_active = 1;
+		WREG32(mmCP_HQD_ACTIVE, mqd->queue_state.cp_hqd_active);
+
+		cik_srbm_select(adev, 0, 0, 0, 0);
+		mutex_unlock(&adev->srbm_mutex);
+
+		amdgpu_bo_kunmap(ring->mqd_obj);
+		amdgpu_bo_unreserve(ring->mqd_obj);
+
+		ring->ready = true;
+		r = amdgpu_ring_test_ring(ring);
+		if (r)
+			ring->ready = false;
+	}
+
+	return 0;
+}
+
+static void gfx_v7_0_cp_enable(struct amdgpu_device *adev, bool enable)
+{
+	gfx_v7_0_cp_gfx_enable(adev, enable);
+	gfx_v7_0_cp_compute_enable(adev, enable);
+}
+
+static int gfx_v7_0_cp_load_microcode(struct amdgpu_device *adev)
+{
+	int r;
+
+	r = gfx_v7_0_cp_gfx_load_microcode(adev);
+	if (r)
+		return r;
+	r = gfx_v7_0_cp_compute_load_microcode(adev);
+	if (r)
+		return r;
+
+	return 0;
+}
+
+static void gfx_v7_0_enable_gui_idle_interrupt(struct amdgpu_device *adev,
+					       bool enable)
+{
+	u32 tmp = RREG32(mmCP_INT_CNTL_RING0);
+
+	if (enable)
+		tmp |= (CP_INT_CNTL_RING0__CNTX_BUSY_INT_ENABLE_MASK |
+				CP_INT_CNTL_RING0__CNTX_EMPTY_INT_ENABLE_MASK);
+	else
+		tmp &= ~(CP_INT_CNTL_RING0__CNTX_BUSY_INT_ENABLE_MASK |
+				CP_INT_CNTL_RING0__CNTX_EMPTY_INT_ENABLE_MASK);
+	WREG32(mmCP_INT_CNTL_RING0, tmp);
+}
+
+static int gfx_v7_0_cp_resume(struct amdgpu_device *adev)
+{
+	int r;
+
+	gfx_v7_0_enable_gui_idle_interrupt(adev, false);
+
+	r = gfx_v7_0_cp_load_microcode(adev);
+	if (r)
+		return r;
+
+	r = gfx_v7_0_cp_gfx_resume(adev);
+	if (r)
+		return r;
+	r = gfx_v7_0_cp_compute_resume(adev);
+	if (r)
+		return r;
+
+	gfx_v7_0_enable_gui_idle_interrupt(adev, true);
+
+	return 0;
+}
+
+static void gfx_v7_0_ce_sync_me(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	u64 gpu_addr = adev->wb.gpu_addr + adev->gfx.ce_sync_offs * 4;
+
+	/* instruct DE to set a magic number */
+	amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3));
+	amdgpu_ring_write(ring, (WRITE_DATA_ENGINE_SEL(0) |
+							 WRITE_DATA_DST_SEL(5)));
+	amdgpu_ring_write(ring, gpu_addr & 0xfffffffc);
+	amdgpu_ring_write(ring, upper_32_bits(gpu_addr) & 0xffffffff);
+	amdgpu_ring_write(ring, 1);
+
+	/* let CE wait till condition satisfied */
+	amdgpu_ring_write(ring, PACKET3(PACKET3_WAIT_REG_MEM, 5));
+	amdgpu_ring_write(ring, (WAIT_REG_MEM_OPERATION(0) | /* wait */
+							 WAIT_REG_MEM_MEM_SPACE(1) | /* memory */
+							 WAIT_REG_MEM_FUNCTION(3) |  /* == */
+							 WAIT_REG_MEM_ENGINE(2)));   /* ce */
+	amdgpu_ring_write(ring, gpu_addr & 0xfffffffc);
+	amdgpu_ring_write(ring, upper_32_bits(gpu_addr) & 0xffffffff);
+	amdgpu_ring_write(ring, 1);
+	amdgpu_ring_write(ring, 0xffffffff);
+	amdgpu_ring_write(ring, 4); /* poll interval */
+
+	/* instruct CE to reset wb of ce_sync to zero */
+	amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3));
+	amdgpu_ring_write(ring, (WRITE_DATA_ENGINE_SEL(2) |
+							 WRITE_DATA_DST_SEL(5) |
+							 WR_CONFIRM));
+	amdgpu_ring_write(ring, gpu_addr & 0xfffffffc);
+	amdgpu_ring_write(ring, upper_32_bits(gpu_addr) & 0xffffffff);
+	amdgpu_ring_write(ring, 0);
+}
+
+/*
+ * vm
+ * VMID 0 is the physical GPU addresses as used by the kernel.
+ * VMIDs 1-15 are used for userspace clients and are handled
+ * by the amdgpu vm/hsa code.
+ */
+/**
+ * gfx_v7_0_ring_emit_vm_flush - cik vm flush using the CP
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Update the page table base and flush the VM TLB
+ * using the CP (CIK).
+ */
+static void gfx_v7_0_ring_emit_vm_flush(struct amdgpu_ring *ring,
+					unsigned vm_id, uint64_t pd_addr)
+{
+	int usepfp = (ring->type == AMDGPU_RING_TYPE_GFX);
+
+	amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3));
+	amdgpu_ring_write(ring, (WRITE_DATA_ENGINE_SEL(usepfp) |
+				 WRITE_DATA_DST_SEL(0)));
+	if (vm_id < 8) {
+		amdgpu_ring_write(ring,
+				  (mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR + vm_id));
+	} else {
+		amdgpu_ring_write(ring,
+				  (mmVM_CONTEXT8_PAGE_TABLE_BASE_ADDR + vm_id - 8));
+	}
+	amdgpu_ring_write(ring, 0);
+	amdgpu_ring_write(ring, pd_addr >> 12);
+
+	/* update SH_MEM_* regs */
+	amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3));
+	amdgpu_ring_write(ring, (WRITE_DATA_ENGINE_SEL(0) |
+				 WRITE_DATA_DST_SEL(0)));
+	amdgpu_ring_write(ring, mmSRBM_GFX_CNTL);
+	amdgpu_ring_write(ring, 0);
+	amdgpu_ring_write(ring, VMID(vm_id));
+
+	amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 6));
+	amdgpu_ring_write(ring, (WRITE_DATA_ENGINE_SEL(0) |
+				 WRITE_DATA_DST_SEL(0)));
+	amdgpu_ring_write(ring, mmSH_MEM_BASES);
+	amdgpu_ring_write(ring, 0);
+
+	amdgpu_ring_write(ring, 0); /* SH_MEM_BASES */
+	amdgpu_ring_write(ring, 0); /* SH_MEM_CONFIG */
+	amdgpu_ring_write(ring, 1); /* SH_MEM_APE1_BASE */
+	amdgpu_ring_write(ring, 0); /* SH_MEM_APE1_LIMIT */
+
+	amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3));
+	amdgpu_ring_write(ring, (WRITE_DATA_ENGINE_SEL(0) |
+				 WRITE_DATA_DST_SEL(0)));
+	amdgpu_ring_write(ring, mmSRBM_GFX_CNTL);
+	amdgpu_ring_write(ring, 0);
+	amdgpu_ring_write(ring, VMID(0));
+
+
+	/* bits 0-15 are the VM contexts0-15 */
+	amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3));
+	amdgpu_ring_write(ring, (WRITE_DATA_ENGINE_SEL(0) |
+				 WRITE_DATA_DST_SEL(0)));
+	amdgpu_ring_write(ring, mmVM_INVALIDATE_REQUEST);
+	amdgpu_ring_write(ring, 0);
+	amdgpu_ring_write(ring, 1 << vm_id);
+
+	/* wait for the invalidate to complete */
+	amdgpu_ring_write(ring, PACKET3(PACKET3_WAIT_REG_MEM, 5));
+	amdgpu_ring_write(ring, (WAIT_REG_MEM_OPERATION(0) | /* wait */
+				 WAIT_REG_MEM_FUNCTION(0) |  /* always */
+				 WAIT_REG_MEM_ENGINE(0))); /* me */
+	amdgpu_ring_write(ring, mmVM_INVALIDATE_REQUEST);
+	amdgpu_ring_write(ring, 0);
+	amdgpu_ring_write(ring, 0); /* ref */
+	amdgpu_ring_write(ring, 0); /* mask */
+	amdgpu_ring_write(ring, 0x20); /* poll interval */
+
+	/* compute doesn't have PFP */
+	if (usepfp) {
+		/* sync PFP to ME, otherwise we might get invalid PFP reads */
+		amdgpu_ring_write(ring, PACKET3(PACKET3_PFP_SYNC_ME, 0));
+		amdgpu_ring_write(ring, 0x0);
+
+		/* synce CE with ME to prevent CE fetch CEIB before context switch done */
+		gfx_v7_0_ce_sync_me(ring);
+	}
+}
+
+/*
+ * RLC
+ * The RLC is a multi-purpose microengine that handles a
+ * variety of functions.
+ */
+static void gfx_v7_0_rlc_fini(struct amdgpu_device *adev)
+{
+	int r;
+
+	/* save restore block */
+	if (adev->gfx.rlc.save_restore_obj) {
+		r = amdgpu_bo_reserve(adev->gfx.rlc.save_restore_obj, false);
+		if (unlikely(r != 0))
+			dev_warn(adev->dev, "(%d) reserve RLC sr bo failed\n", r);
+		amdgpu_bo_unpin(adev->gfx.rlc.save_restore_obj);
+		amdgpu_bo_unreserve(adev->gfx.rlc.save_restore_obj);
+
+		amdgpu_bo_unref(&adev->gfx.rlc.save_restore_obj);
+		adev->gfx.rlc.save_restore_obj = NULL;
+	}
+
+	/* clear state block */
+	if (adev->gfx.rlc.clear_state_obj) {
+		r = amdgpu_bo_reserve(adev->gfx.rlc.clear_state_obj, false);
+		if (unlikely(r != 0))
+			dev_warn(adev->dev, "(%d) reserve RLC c bo failed\n", r);
+		amdgpu_bo_unpin(adev->gfx.rlc.clear_state_obj);
+		amdgpu_bo_unreserve(adev->gfx.rlc.clear_state_obj);
+
+		amdgpu_bo_unref(&adev->gfx.rlc.clear_state_obj);
+		adev->gfx.rlc.clear_state_obj = NULL;
+	}
+
+	/* clear state block */
+	if (adev->gfx.rlc.cp_table_obj) {
+		r = amdgpu_bo_reserve(adev->gfx.rlc.cp_table_obj, false);
+		if (unlikely(r != 0))
+			dev_warn(adev->dev, "(%d) reserve RLC cp table bo failed\n", r);
+		amdgpu_bo_unpin(adev->gfx.rlc.cp_table_obj);
+		amdgpu_bo_unreserve(adev->gfx.rlc.cp_table_obj);
+
+		amdgpu_bo_unref(&adev->gfx.rlc.cp_table_obj);
+		adev->gfx.rlc.cp_table_obj = NULL;
+	}
+}
+
+static int gfx_v7_0_rlc_init(struct amdgpu_device *adev)
+{
+	const u32 *src_ptr;
+	volatile u32 *dst_ptr;
+	u32 dws, i;
+	const struct cs_section_def *cs_data;
+	int r;
+
+	/* allocate rlc buffers */
+	if (adev->flags & AMDGPU_IS_APU) {
+		if (adev->asic_type == CHIP_KAVERI) {
+			adev->gfx.rlc.reg_list = spectre_rlc_save_restore_register_list;
+			adev->gfx.rlc.reg_list_size =
+				(u32)ARRAY_SIZE(spectre_rlc_save_restore_register_list);
+		} else {
+			adev->gfx.rlc.reg_list = kalindi_rlc_save_restore_register_list;
+			adev->gfx.rlc.reg_list_size =
+				(u32)ARRAY_SIZE(kalindi_rlc_save_restore_register_list);
+		}
+	}
+	adev->gfx.rlc.cs_data = ci_cs_data;
+	adev->gfx.rlc.cp_table_size = CP_ME_TABLE_SIZE * 5 * 4;
+
+	src_ptr = adev->gfx.rlc.reg_list;
+	dws = adev->gfx.rlc.reg_list_size;
+	dws += (5 * 16) + 48 + 48 + 64;
+
+	cs_data = adev->gfx.rlc.cs_data;
+
+	if (src_ptr) {
+		/* save restore block */
+		if (adev->gfx.rlc.save_restore_obj == NULL) {
+			r = amdgpu_bo_create(adev, dws * 4, PAGE_SIZE, true,
+					     AMDGPU_GEM_DOMAIN_VRAM, 0, NULL, &adev->gfx.rlc.save_restore_obj);
+			if (r) {
+				dev_warn(adev->dev, "(%d) create RLC sr bo failed\n", r);
+				return r;
+			}
+		}
+
+		r = amdgpu_bo_reserve(adev->gfx.rlc.save_restore_obj, false);
+		if (unlikely(r != 0)) {
+			gfx_v7_0_rlc_fini(adev);
+			return r;
+		}
+		r = amdgpu_bo_pin(adev->gfx.rlc.save_restore_obj, AMDGPU_GEM_DOMAIN_VRAM,
+				  &adev->gfx.rlc.save_restore_gpu_addr);
+		if (r) {
+			amdgpu_bo_unreserve(adev->gfx.rlc.save_restore_obj);
+			dev_warn(adev->dev, "(%d) pin RLC sr bo failed\n", r);
+			gfx_v7_0_rlc_fini(adev);
+			return r;
+		}
+
+		r = amdgpu_bo_kmap(adev->gfx.rlc.save_restore_obj, (void **)&adev->gfx.rlc.sr_ptr);
+		if (r) {
+			dev_warn(adev->dev, "(%d) map RLC sr bo failed\n", r);
+			gfx_v7_0_rlc_fini(adev);
+			return r;
+		}
+		/* write the sr buffer */
+		dst_ptr = adev->gfx.rlc.sr_ptr;
+		for (i = 0; i < adev->gfx.rlc.reg_list_size; i++)
+			dst_ptr[i] = cpu_to_le32(src_ptr[i]);
+		amdgpu_bo_kunmap(adev->gfx.rlc.save_restore_obj);
+		amdgpu_bo_unreserve(adev->gfx.rlc.save_restore_obj);
+	}
+
+	if (cs_data) {
+		/* clear state block */
+		adev->gfx.rlc.clear_state_size = dws = gfx_v7_0_get_csb_size(adev);
+
+		if (adev->gfx.rlc.clear_state_obj == NULL) {
+			r = amdgpu_bo_create(adev, dws * 4, PAGE_SIZE, true,
+					     AMDGPU_GEM_DOMAIN_VRAM, 0, NULL, &adev->gfx.rlc.clear_state_obj);
+			if (r) {
+				dev_warn(adev->dev, "(%d) create RLC c bo failed\n", r);
+				gfx_v7_0_rlc_fini(adev);
+				return r;
+			}
+		}
+		r = amdgpu_bo_reserve(adev->gfx.rlc.clear_state_obj, false);
+		if (unlikely(r != 0)) {
+			gfx_v7_0_rlc_fini(adev);
+			return r;
+		}
+		r = amdgpu_bo_pin(adev->gfx.rlc.clear_state_obj, AMDGPU_GEM_DOMAIN_VRAM,
+				  &adev->gfx.rlc.clear_state_gpu_addr);
+		if (r) {
+			amdgpu_bo_unreserve(adev->gfx.rlc.clear_state_obj);
+			dev_warn(adev->dev, "(%d) pin RLC c bo failed\n", r);
+			gfx_v7_0_rlc_fini(adev);
+			return r;
+		}
+
+		r = amdgpu_bo_kmap(adev->gfx.rlc.clear_state_obj, (void **)&adev->gfx.rlc.cs_ptr);
+		if (r) {
+			dev_warn(adev->dev, "(%d) map RLC c bo failed\n", r);
+			gfx_v7_0_rlc_fini(adev);
+			return r;
+		}
+		/* set up the cs buffer */
+		dst_ptr = adev->gfx.rlc.cs_ptr;
+		gfx_v7_0_get_csb_buffer(adev, dst_ptr);
+		amdgpu_bo_kunmap(adev->gfx.rlc.clear_state_obj);
+		amdgpu_bo_unreserve(adev->gfx.rlc.clear_state_obj);
+	}
+
+	if (adev->gfx.rlc.cp_table_size) {
+		if (adev->gfx.rlc.cp_table_obj == NULL) {
+			r = amdgpu_bo_create(adev, adev->gfx.rlc.cp_table_size, PAGE_SIZE, true,
+					     AMDGPU_GEM_DOMAIN_VRAM, 0, NULL, &adev->gfx.rlc.cp_table_obj);
+			if (r) {
+				dev_warn(adev->dev, "(%d) create RLC cp table bo failed\n", r);
+				gfx_v7_0_rlc_fini(adev);
+				return r;
+			}
+		}
+
+		r = amdgpu_bo_reserve(adev->gfx.rlc.cp_table_obj, false);
+		if (unlikely(r != 0)) {
+			dev_warn(adev->dev, "(%d) reserve RLC cp table bo failed\n", r);
+			gfx_v7_0_rlc_fini(adev);
+			return r;
+		}
+		r = amdgpu_bo_pin(adev->gfx.rlc.cp_table_obj, AMDGPU_GEM_DOMAIN_VRAM,
+				  &adev->gfx.rlc.cp_table_gpu_addr);
+		if (r) {
+			amdgpu_bo_unreserve(adev->gfx.rlc.cp_table_obj);
+			dev_warn(adev->dev, "(%d) pin RLC cp_table bo failed\n", r);
+			gfx_v7_0_rlc_fini(adev);
+			return r;
+		}
+		r = amdgpu_bo_kmap(adev->gfx.rlc.cp_table_obj, (void **)&adev->gfx.rlc.cp_table_ptr);
+		if (r) {
+			dev_warn(adev->dev, "(%d) map RLC cp table bo failed\n", r);
+			gfx_v7_0_rlc_fini(adev);
+			return r;
+		}
+
+		gfx_v7_0_init_cp_pg_table(adev);
+
+		amdgpu_bo_kunmap(adev->gfx.rlc.cp_table_obj);
+		amdgpu_bo_unreserve(adev->gfx.rlc.cp_table_obj);
+
+	}
+
+	return 0;
+}
+
+static void gfx_v7_0_enable_lbpw(struct amdgpu_device *adev, bool enable)
+{
+	u32 tmp;
+
+	tmp = RREG32(mmRLC_LB_CNTL);
+	if (enable)
+		tmp |= RLC_LB_CNTL__LOAD_BALANCE_ENABLE_MASK;
+	else
+		tmp &= ~RLC_LB_CNTL__LOAD_BALANCE_ENABLE_MASK;
+	WREG32(mmRLC_LB_CNTL, tmp);
+}
+
+static void gfx_v7_0_wait_for_rlc_serdes(struct amdgpu_device *adev)
+{
+	u32 i, j, k;
+	u32 mask;
+
+	mutex_lock(&adev->grbm_idx_mutex);
+	for (i = 0; i < adev->gfx.config.max_shader_engines; i++) {
+		for (j = 0; j < adev->gfx.config.max_sh_per_se; j++) {
+			gfx_v7_0_select_se_sh(adev, i, j);
+			for (k = 0; k < adev->usec_timeout; k++) {
+				if (RREG32(mmRLC_SERDES_CU_MASTER_BUSY) == 0)
+					break;
+				udelay(1);
+			}
+		}
+	}
+	gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff);
+	mutex_unlock(&adev->grbm_idx_mutex);
+
+	mask = RLC_SERDES_NONCU_MASTER_BUSY__SE_MASTER_BUSY_MASK |
+		RLC_SERDES_NONCU_MASTER_BUSY__GC_MASTER_BUSY_MASK |
+		RLC_SERDES_NONCU_MASTER_BUSY__TC0_MASTER_BUSY_MASK |
+		RLC_SERDES_NONCU_MASTER_BUSY__TC1_MASTER_BUSY_MASK;
+	for (k = 0; k < adev->usec_timeout; k++) {
+		if ((RREG32(mmRLC_SERDES_NONCU_MASTER_BUSY) & mask) == 0)
+			break;
+		udelay(1);
+	}
+}
+
+static void gfx_v7_0_update_rlc(struct amdgpu_device *adev, u32 rlc)
+{
+	u32 tmp;
+
+	tmp = RREG32(mmRLC_CNTL);
+	if (tmp != rlc)
+		WREG32(mmRLC_CNTL, rlc);
+}
+
+static u32 gfx_v7_0_halt_rlc(struct amdgpu_device *adev)
+{
+	u32 data, orig;
+
+	orig = data = RREG32(mmRLC_CNTL);
+
+	if (data & RLC_CNTL__RLC_ENABLE_F32_MASK) {
+		u32 i;
+
+		data &= ~RLC_CNTL__RLC_ENABLE_F32_MASK;
+		WREG32(mmRLC_CNTL, data);
+
+		for (i = 0; i < adev->usec_timeout; i++) {
+			if ((RREG32(mmRLC_GPM_STAT) & RLC_GPM_STAT__RLC_BUSY_MASK) == 0)
+				break;
+			udelay(1);
+		}
+
+		gfx_v7_0_wait_for_rlc_serdes(adev);
+	}
+
+	return orig;
+}
+
+void gfx_v7_0_enter_rlc_safe_mode(struct amdgpu_device *adev)
+{
+	u32 tmp, i, mask;
+
+	tmp = 0x1 | (1 << 1);
+	WREG32(mmRLC_GPR_REG2, tmp);
+
+	mask = RLC_GPM_STAT__GFX_POWER_STATUS_MASK |
+		RLC_GPM_STAT__GFX_CLOCK_STATUS_MASK;
+	for (i = 0; i < adev->usec_timeout; i++) {
+		if ((RREG32(mmRLC_GPM_STAT) & mask) == mask)
+			break;
+		udelay(1);
+	}
+
+	for (i = 0; i < adev->usec_timeout; i++) {
+		if ((RREG32(mmRLC_GPR_REG2) & 0x1) == 0)
+			break;
+		udelay(1);
+	}
+}
+
+void gfx_v7_0_exit_rlc_safe_mode(struct amdgpu_device *adev)
+{
+	u32 tmp;
+
+	tmp = 0x1 | (0 << 1);
+	WREG32(mmRLC_GPR_REG2, tmp);
+}
+
+/**
+ * gfx_v7_0_rlc_stop - stop the RLC ME
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Halt the RLC ME (MicroEngine) (CIK).
+ */
+void gfx_v7_0_rlc_stop(struct amdgpu_device *adev)
+{
+	WREG32(mmRLC_CNTL, 0);
+
+	gfx_v7_0_enable_gui_idle_interrupt(adev, false);
+
+	gfx_v7_0_wait_for_rlc_serdes(adev);
+}
+
+/**
+ * gfx_v7_0_rlc_start - start the RLC ME
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Unhalt the RLC ME (MicroEngine) (CIK).
+ */
+static void gfx_v7_0_rlc_start(struct amdgpu_device *adev)
+{
+	WREG32(mmRLC_CNTL, RLC_CNTL__RLC_ENABLE_F32_MASK);
+
+	gfx_v7_0_enable_gui_idle_interrupt(adev, true);
+
+	udelay(50);
+}
+
+static void gfx_v7_0_rlc_reset(struct amdgpu_device *adev)
+{
+	u32 tmp = RREG32(mmGRBM_SOFT_RESET);
+
+	tmp |= GRBM_SOFT_RESET__SOFT_RESET_RLC_MASK;
+	WREG32(mmGRBM_SOFT_RESET, tmp);
+	udelay(50);
+	tmp &= ~GRBM_SOFT_RESET__SOFT_RESET_RLC_MASK;
+	WREG32(mmGRBM_SOFT_RESET, tmp);
+	udelay(50);
+}
+
+/**
+ * gfx_v7_0_rlc_resume - setup the RLC hw
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Initialize the RLC registers, load the ucode,
+ * and start the RLC (CIK).
+ * Returns 0 for success, -EINVAL if the ucode is not available.
+ */
+static int gfx_v7_0_rlc_resume(struct amdgpu_device *adev)
+{
+	const struct rlc_firmware_header_v1_0 *hdr;
+	const __le32 *fw_data;
+	unsigned i, fw_size;
+	u32 tmp;
+
+	if (!adev->gfx.rlc_fw)
+		return -EINVAL;
+
+	hdr = (const struct rlc_firmware_header_v1_0 *)adev->gfx.rlc_fw->data;
+	amdgpu_ucode_print_rlc_hdr(&hdr->header);
+	adev->gfx.rlc_fw_version = le32_to_cpu(hdr->header.ucode_version);
+
+	gfx_v7_0_rlc_stop(adev);
+
+	/* disable CG */
+	tmp = RREG32(mmRLC_CGCG_CGLS_CTRL) & 0xfffffffc;
+	WREG32(mmRLC_CGCG_CGLS_CTRL, tmp);
+
+	gfx_v7_0_rlc_reset(adev);
+
+	gfx_v7_0_init_pg(adev);
+
+	WREG32(mmRLC_LB_CNTR_INIT, 0);
+	WREG32(mmRLC_LB_CNTR_MAX, 0x00008000);
+
+	mutex_lock(&adev->grbm_idx_mutex);
+	gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff);
+	WREG32(mmRLC_LB_INIT_CU_MASK, 0xffffffff);
+	WREG32(mmRLC_LB_PARAMS, 0x00600408);
+	WREG32(mmRLC_LB_CNTL, 0x80000004);
+	mutex_unlock(&adev->grbm_idx_mutex);
+
+	WREG32(mmRLC_MC_CNTL, 0);
+	WREG32(mmRLC_UCODE_CNTL, 0);
+
+	fw_data = (const __le32 *)
+		(adev->gfx.rlc_fw->data + le32_to_cpu(hdr->header.ucode_array_offset_bytes));
+	fw_size = le32_to_cpu(hdr->header.ucode_size_bytes) / 4;
+	WREG32(mmRLC_GPM_UCODE_ADDR, 0);
+	for (i = 0; i < fw_size; i++)
+		WREG32(mmRLC_GPM_UCODE_DATA, le32_to_cpup(fw_data++));
+	WREG32(mmRLC_GPM_UCODE_ADDR, adev->gfx.rlc_fw_version);
+
+	/* XXX - find out what chips support lbpw */
+	gfx_v7_0_enable_lbpw(adev, false);
+
+	if (adev->asic_type == CHIP_BONAIRE)
+		WREG32(mmRLC_DRIVER_CPDMA_STATUS, 0);
+
+	gfx_v7_0_rlc_start(adev);
+
+	return 0;
+}
+
+static void gfx_v7_0_enable_cgcg(struct amdgpu_device *adev, bool enable)
+{
+	u32 data, orig, tmp, tmp2;
+
+	orig = data = RREG32(mmRLC_CGCG_CGLS_CTRL);
+
+	if (enable && (adev->cg_flags & AMDGPU_CG_SUPPORT_GFX_CGCG)) {
+		gfx_v7_0_enable_gui_idle_interrupt(adev, true);
+
+		tmp = gfx_v7_0_halt_rlc(adev);
+
+		mutex_lock(&adev->grbm_idx_mutex);
+		gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff);
+		WREG32(mmRLC_SERDES_WR_CU_MASTER_MASK, 0xffffffff);
+		WREG32(mmRLC_SERDES_WR_NONCU_MASTER_MASK, 0xffffffff);
+		tmp2 = RLC_SERDES_WR_CTRL__BPM_ADDR_MASK |
+			RLC_SERDES_WR_CTRL__CGCG_OVERRIDE_0_MASK |
+			RLC_SERDES_WR_CTRL__CGLS_ENABLE_MASK;
+		WREG32(mmRLC_SERDES_WR_CTRL, tmp2);
+		mutex_unlock(&adev->grbm_idx_mutex);
+
+		gfx_v7_0_update_rlc(adev, tmp);
+
+		data |= RLC_CGCG_CGLS_CTRL__CGCG_EN_MASK | RLC_CGCG_CGLS_CTRL__CGLS_EN_MASK;
+	} else {
+		gfx_v7_0_enable_gui_idle_interrupt(adev, false);
+
+		RREG32(mmCB_CGTT_SCLK_CTRL);
+		RREG32(mmCB_CGTT_SCLK_CTRL);
+		RREG32(mmCB_CGTT_SCLK_CTRL);
+		RREG32(mmCB_CGTT_SCLK_CTRL);
+
+		data &= ~(RLC_CGCG_CGLS_CTRL__CGCG_EN_MASK | RLC_CGCG_CGLS_CTRL__CGLS_EN_MASK);
+	}
+
+	if (orig != data)
+		WREG32(mmRLC_CGCG_CGLS_CTRL, data);
+
+}
+
+static void gfx_v7_0_enable_mgcg(struct amdgpu_device *adev, bool enable)
+{
+	u32 data, orig, tmp = 0;
+
+	if (enable && (adev->cg_flags & AMDGPU_CG_SUPPORT_GFX_MGCG)) {
+		if (adev->cg_flags & AMDGPU_CG_SUPPORT_GFX_MGLS) {
+			if (adev->cg_flags & AMDGPU_CG_SUPPORT_GFX_CP_LS) {
+				orig = data = RREG32(mmCP_MEM_SLP_CNTL);
+				data |= CP_MEM_SLP_CNTL__CP_MEM_LS_EN_MASK;
+				if (orig != data)
+					WREG32(mmCP_MEM_SLP_CNTL, data);
+			}
+		}
+
+		orig = data = RREG32(mmRLC_CGTT_MGCG_OVERRIDE);
+		data |= 0x00000001;
+		data &= 0xfffffffd;
+		if (orig != data)
+			WREG32(mmRLC_CGTT_MGCG_OVERRIDE, data);
+
+		tmp = gfx_v7_0_halt_rlc(adev);
+
+		mutex_lock(&adev->grbm_idx_mutex);
+		gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff);
+		WREG32(mmRLC_SERDES_WR_CU_MASTER_MASK, 0xffffffff);
+		WREG32(mmRLC_SERDES_WR_NONCU_MASTER_MASK, 0xffffffff);
+		data = RLC_SERDES_WR_CTRL__BPM_ADDR_MASK |
+			RLC_SERDES_WR_CTRL__MGCG_OVERRIDE_0_MASK;
+		WREG32(mmRLC_SERDES_WR_CTRL, data);
+		mutex_unlock(&adev->grbm_idx_mutex);
+
+		gfx_v7_0_update_rlc(adev, tmp);
+
+		if (adev->cg_flags & AMDGPU_CG_SUPPORT_GFX_CGTS) {
+			orig = data = RREG32(mmCGTS_SM_CTRL_REG);
+			data &= ~CGTS_SM_CTRL_REG__SM_MODE_MASK;
+			data |= (0x2 << CGTS_SM_CTRL_REG__SM_MODE__SHIFT);
+			data |= CGTS_SM_CTRL_REG__SM_MODE_ENABLE_MASK;
+			data &= ~CGTS_SM_CTRL_REG__OVERRIDE_MASK;
+			if ((adev->cg_flags & AMDGPU_CG_SUPPORT_GFX_MGLS) &&
+			    (adev->cg_flags & AMDGPU_CG_SUPPORT_GFX_CGTS_LS))
+				data &= ~CGTS_SM_CTRL_REG__LS_OVERRIDE_MASK;
+			data &= ~CGTS_SM_CTRL_REG__ON_MONITOR_ADD_MASK;
+			data |= CGTS_SM_CTRL_REG__ON_MONITOR_ADD_EN_MASK;
+			data |= (0x96 << CGTS_SM_CTRL_REG__ON_MONITOR_ADD__SHIFT);
+			if (orig != data)
+				WREG32(mmCGTS_SM_CTRL_REG, data);
+		}
+	} else {
+		orig = data = RREG32(mmRLC_CGTT_MGCG_OVERRIDE);
+		data |= 0x00000003;
+		if (orig != data)
+			WREG32(mmRLC_CGTT_MGCG_OVERRIDE, data);
+
+		data = RREG32(mmRLC_MEM_SLP_CNTL);
+		if (data & RLC_MEM_SLP_CNTL__RLC_MEM_LS_EN_MASK) {
+			data &= ~RLC_MEM_SLP_CNTL__RLC_MEM_LS_EN_MASK;
+			WREG32(mmRLC_MEM_SLP_CNTL, data);
+		}
+
+		data = RREG32(mmCP_MEM_SLP_CNTL);
+		if (data & CP_MEM_SLP_CNTL__CP_MEM_LS_EN_MASK) {
+			data &= ~CP_MEM_SLP_CNTL__CP_MEM_LS_EN_MASK;
+			WREG32(mmCP_MEM_SLP_CNTL, data);
+		}
+
+		orig = data = RREG32(mmCGTS_SM_CTRL_REG);
+		data |= CGTS_SM_CTRL_REG__OVERRIDE_MASK | CGTS_SM_CTRL_REG__LS_OVERRIDE_MASK;
+		if (orig != data)
+			WREG32(mmCGTS_SM_CTRL_REG, data);
+
+		tmp = gfx_v7_0_halt_rlc(adev);
+
+		mutex_lock(&adev->grbm_idx_mutex);
+		gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff);
+		WREG32(mmRLC_SERDES_WR_CU_MASTER_MASK, 0xffffffff);
+		WREG32(mmRLC_SERDES_WR_NONCU_MASTER_MASK, 0xffffffff);
+		data = RLC_SERDES_WR_CTRL__BPM_ADDR_MASK | RLC_SERDES_WR_CTRL__MGCG_OVERRIDE_1_MASK;
+		WREG32(mmRLC_SERDES_WR_CTRL, data);
+		mutex_unlock(&adev->grbm_idx_mutex);
+
+		gfx_v7_0_update_rlc(adev, tmp);
+	}
+}
+
+static void gfx_v7_0_update_cg(struct amdgpu_device *adev,
+			       bool enable)
+{
+	gfx_v7_0_enable_gui_idle_interrupt(adev, false);
+	/* order matters! */
+	if (enable) {
+		gfx_v7_0_enable_mgcg(adev, true);
+		gfx_v7_0_enable_cgcg(adev, true);
+	} else {
+		gfx_v7_0_enable_cgcg(adev, false);
+		gfx_v7_0_enable_mgcg(adev, false);
+	}
+	gfx_v7_0_enable_gui_idle_interrupt(adev, true);
+}
+
+static void gfx_v7_0_enable_sclk_slowdown_on_pu(struct amdgpu_device *adev,
+						bool enable)
+{
+	u32 data, orig;
+
+	orig = data = RREG32(mmRLC_PG_CNTL);
+	if (enable && (adev->pg_flags & AMDGPU_PG_SUPPORT_RLC_SMU_HS))
+		data |= RLC_PG_CNTL__SMU_CLK_SLOWDOWN_ON_PU_ENABLE_MASK;
+	else
+		data &= ~RLC_PG_CNTL__SMU_CLK_SLOWDOWN_ON_PU_ENABLE_MASK;
+	if (orig != data)
+		WREG32(mmRLC_PG_CNTL, data);
+}
+
+static void gfx_v7_0_enable_sclk_slowdown_on_pd(struct amdgpu_device *adev,
+						bool enable)
+{
+	u32 data, orig;
+
+	orig = data = RREG32(mmRLC_PG_CNTL);
+	if (enable && (adev->pg_flags & AMDGPU_PG_SUPPORT_RLC_SMU_HS))
+		data |= RLC_PG_CNTL__SMU_CLK_SLOWDOWN_ON_PD_ENABLE_MASK;
+	else
+		data &= ~RLC_PG_CNTL__SMU_CLK_SLOWDOWN_ON_PD_ENABLE_MASK;
+	if (orig != data)
+		WREG32(mmRLC_PG_CNTL, data);
+}
+
+static void gfx_v7_0_enable_cp_pg(struct amdgpu_device *adev, bool enable)
+{
+	u32 data, orig;
+
+	orig = data = RREG32(mmRLC_PG_CNTL);
+	if (enable && (adev->pg_flags & AMDGPU_PG_SUPPORT_CP))
+		data &= ~0x8000;
+	else
+		data |= 0x8000;
+	if (orig != data)
+		WREG32(mmRLC_PG_CNTL, data);
+}
+
+static void gfx_v7_0_enable_gds_pg(struct amdgpu_device *adev, bool enable)
+{
+	u32 data, orig;
+
+	orig = data = RREG32(mmRLC_PG_CNTL);
+	if (enable && (adev->pg_flags & AMDGPU_PG_SUPPORT_GDS))
+		data &= ~0x2000;
+	else
+		data |= 0x2000;
+	if (orig != data)
+		WREG32(mmRLC_PG_CNTL, data);
+}
+
+static void gfx_v7_0_init_cp_pg_table(struct amdgpu_device *adev)
+{
+	const __le32 *fw_data;
+	volatile u32 *dst_ptr;
+	int me, i, max_me = 4;
+	u32 bo_offset = 0;
+	u32 table_offset, table_size;
+
+	if (adev->asic_type == CHIP_KAVERI)
+		max_me = 5;
+
+	if (adev->gfx.rlc.cp_table_ptr == NULL)
+		return;
+
+	/* write the cp table buffer */
+	dst_ptr = adev->gfx.rlc.cp_table_ptr;
+	for (me = 0; me < max_me; me++) {
+		if (me == 0) {
+			const struct gfx_firmware_header_v1_0 *hdr =
+				(const struct gfx_firmware_header_v1_0 *)adev->gfx.ce_fw->data;
+			fw_data = (const __le32 *)
+				(adev->gfx.ce_fw->data +
+				 le32_to_cpu(hdr->header.ucode_array_offset_bytes));
+			table_offset = le32_to_cpu(hdr->jt_offset);
+			table_size = le32_to_cpu(hdr->jt_size);
+		} else if (me == 1) {
+			const struct gfx_firmware_header_v1_0 *hdr =
+				(const struct gfx_firmware_header_v1_0 *)adev->gfx.pfp_fw->data;
+			fw_data = (const __le32 *)
+				(adev->gfx.pfp_fw->data +
+				 le32_to_cpu(hdr->header.ucode_array_offset_bytes));
+			table_offset = le32_to_cpu(hdr->jt_offset);
+			table_size = le32_to_cpu(hdr->jt_size);
+		} else if (me == 2) {
+			const struct gfx_firmware_header_v1_0 *hdr =
+				(const struct gfx_firmware_header_v1_0 *)adev->gfx.me_fw->data;
+			fw_data = (const __le32 *)
+				(adev->gfx.me_fw->data +
+				 le32_to_cpu(hdr->header.ucode_array_offset_bytes));
+			table_offset = le32_to_cpu(hdr->jt_offset);
+			table_size = le32_to_cpu(hdr->jt_size);
+		} else if (me == 3) {
+			const struct gfx_firmware_header_v1_0 *hdr =
+				(const struct gfx_firmware_header_v1_0 *)adev->gfx.mec_fw->data;
+			fw_data = (const __le32 *)
+				(adev->gfx.mec_fw->data +
+				 le32_to_cpu(hdr->header.ucode_array_offset_bytes));
+			table_offset = le32_to_cpu(hdr->jt_offset);
+			table_size = le32_to_cpu(hdr->jt_size);
+		} else {
+			const struct gfx_firmware_header_v1_0 *hdr =
+				(const struct gfx_firmware_header_v1_0 *)adev->gfx.mec2_fw->data;
+			fw_data = (const __le32 *)
+				(adev->gfx.mec2_fw->data +
+				 le32_to_cpu(hdr->header.ucode_array_offset_bytes));
+			table_offset = le32_to_cpu(hdr->jt_offset);
+			table_size = le32_to_cpu(hdr->jt_size);
+		}
+
+		for (i = 0; i < table_size; i ++) {
+			dst_ptr[bo_offset + i] =
+				cpu_to_le32(le32_to_cpu(fw_data[table_offset + i]));
+		}
+
+		bo_offset += table_size;
+	}
+}
+
+static void gfx_v7_0_enable_gfx_cgpg(struct amdgpu_device *adev,
+				     bool enable)
+{
+	u32 data, orig;
+
+	if (enable && (adev->pg_flags & AMDGPU_PG_SUPPORT_GFX_PG)) {
+		orig = data = RREG32(mmRLC_PG_CNTL);
+		data |= RLC_PG_CNTL__GFX_POWER_GATING_ENABLE_MASK;
+		if (orig != data)
+			WREG32(mmRLC_PG_CNTL, data);
+
+		orig = data = RREG32(mmRLC_AUTO_PG_CTRL);
+		data |= RLC_AUTO_PG_CTRL__AUTO_PG_EN_MASK;
+		if (orig != data)
+			WREG32(mmRLC_AUTO_PG_CTRL, data);
+	} else {
+		orig = data = RREG32(mmRLC_PG_CNTL);
+		data &= ~RLC_PG_CNTL__GFX_POWER_GATING_ENABLE_MASK;
+		if (orig != data)
+			WREG32(mmRLC_PG_CNTL, data);
+
+		orig = data = RREG32(mmRLC_AUTO_PG_CTRL);
+		data &= ~RLC_AUTO_PG_CTRL__AUTO_PG_EN_MASK;
+		if (orig != data)
+			WREG32(mmRLC_AUTO_PG_CTRL, data);
+
+		data = RREG32(mmDB_RENDER_CONTROL);
+	}
+}
+
+static u32 gfx_v7_0_get_cu_active_bitmap(struct amdgpu_device *adev,
+					 u32 se, u32 sh)
+{
+	u32 mask = 0, tmp, tmp1;
+	int i;
+
+	gfx_v7_0_select_se_sh(adev, se, sh);
+	tmp = RREG32(mmCC_GC_SHADER_ARRAY_CONFIG);
+	tmp1 = RREG32(mmGC_USER_SHADER_ARRAY_CONFIG);
+	gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff);
+
+	tmp &= 0xffff0000;
+
+	tmp |= tmp1;
+	tmp >>= 16;
+
+	for (i = 0; i < adev->gfx.config.max_cu_per_sh; i ++) {
+		mask <<= 1;
+		mask |= 1;
+	}
+
+	return (~tmp) & mask;
+}
+
+static void gfx_v7_0_init_ao_cu_mask(struct amdgpu_device *adev)
+{
+	uint32_t tmp, active_cu_number;
+	struct amdgpu_cu_info cu_info;
+
+	gfx_v7_0_get_cu_info(adev, &cu_info);
+	tmp = cu_info.ao_cu_mask;
+	active_cu_number = cu_info.number;
+
+	WREG32(mmRLC_PG_ALWAYS_ON_CU_MASK, tmp);
+
+	tmp = RREG32(mmRLC_MAX_PG_CU);
+	tmp &= ~RLC_MAX_PG_CU__MAX_POWERED_UP_CU_MASK;
+	tmp |= (active_cu_number << RLC_MAX_PG_CU__MAX_POWERED_UP_CU__SHIFT);
+	WREG32(mmRLC_MAX_PG_CU, tmp);
+}
+
+static void gfx_v7_0_enable_gfx_static_mgpg(struct amdgpu_device *adev,
+					    bool enable)
+{
+	u32 data, orig;
+
+	orig = data = RREG32(mmRLC_PG_CNTL);
+	if (enable && (adev->pg_flags & AMDGPU_PG_SUPPORT_GFX_SMG))
+		data |= RLC_PG_CNTL__STATIC_PER_CU_PG_ENABLE_MASK;
+	else
+		data &= ~RLC_PG_CNTL__STATIC_PER_CU_PG_ENABLE_MASK;
+	if (orig != data)
+		WREG32(mmRLC_PG_CNTL, data);
+}
+
+static void gfx_v7_0_enable_gfx_dynamic_mgpg(struct amdgpu_device *adev,
+					     bool enable)
+{
+	u32 data, orig;
+
+	orig = data = RREG32(mmRLC_PG_CNTL);
+	if (enable && (adev->pg_flags & AMDGPU_PG_SUPPORT_GFX_DMG))
+		data |= RLC_PG_CNTL__DYN_PER_CU_PG_ENABLE_MASK;
+	else
+		data &= ~RLC_PG_CNTL__DYN_PER_CU_PG_ENABLE_MASK;
+	if (orig != data)
+		WREG32(mmRLC_PG_CNTL, data);
+}
+
+#define RLC_SAVE_AND_RESTORE_STARTING_OFFSET 0x90
+#define RLC_CLEAR_STATE_DESCRIPTOR_OFFSET    0x3D
+
+static void gfx_v7_0_init_gfx_cgpg(struct amdgpu_device *adev)
+{
+	u32 data, orig;
+	u32 i;
+
+	if (adev->gfx.rlc.cs_data) {
+		WREG32(mmRLC_GPM_SCRATCH_ADDR, RLC_CLEAR_STATE_DESCRIPTOR_OFFSET);
+		WREG32(mmRLC_GPM_SCRATCH_DATA, upper_32_bits(adev->gfx.rlc.clear_state_gpu_addr));
+		WREG32(mmRLC_GPM_SCRATCH_DATA, lower_32_bits(adev->gfx.rlc.clear_state_gpu_addr));
+		WREG32(mmRLC_GPM_SCRATCH_DATA, adev->gfx.rlc.clear_state_size);
+	} else {
+		WREG32(mmRLC_GPM_SCRATCH_ADDR, RLC_CLEAR_STATE_DESCRIPTOR_OFFSET);
+		for (i = 0; i < 3; i++)
+			WREG32(mmRLC_GPM_SCRATCH_DATA, 0);
+	}
+	if (adev->gfx.rlc.reg_list) {
+		WREG32(mmRLC_GPM_SCRATCH_ADDR, RLC_SAVE_AND_RESTORE_STARTING_OFFSET);
+		for (i = 0; i < adev->gfx.rlc.reg_list_size; i++)
+			WREG32(mmRLC_GPM_SCRATCH_DATA, adev->gfx.rlc.reg_list[i]);
+	}
+
+	orig = data = RREG32(mmRLC_PG_CNTL);
+	data |= RLC_PG_CNTL__GFX_POWER_GATING_SRC_MASK;
+	if (orig != data)
+		WREG32(mmRLC_PG_CNTL, data);
+
+	WREG32(mmRLC_SAVE_AND_RESTORE_BASE, adev->gfx.rlc.save_restore_gpu_addr >> 8);
+	WREG32(mmRLC_JUMP_TABLE_RESTORE, adev->gfx.rlc.cp_table_gpu_addr >> 8);
+
+	data = RREG32(mmCP_RB_WPTR_POLL_CNTL);
+	data &= ~CP_RB_WPTR_POLL_CNTL__IDLE_POLL_COUNT_MASK;
+	data |= (0x60 << CP_RB_WPTR_POLL_CNTL__IDLE_POLL_COUNT__SHIFT);
+	WREG32(mmCP_RB_WPTR_POLL_CNTL, data);
+
+	data = 0x10101010;
+	WREG32(mmRLC_PG_DELAY, data);
+
+	data = RREG32(mmRLC_PG_DELAY_2);
+	data &= ~0xff;
+	data |= 0x3;
+	WREG32(mmRLC_PG_DELAY_2, data);
+
+	data = RREG32(mmRLC_AUTO_PG_CTRL);
+	data &= ~RLC_AUTO_PG_CTRL__GRBM_REG_SAVE_GFX_IDLE_THRESHOLD_MASK;
+	data |= (0x700 << RLC_AUTO_PG_CTRL__GRBM_REG_SAVE_GFX_IDLE_THRESHOLD__SHIFT);
+	WREG32(mmRLC_AUTO_PG_CTRL, data);
+
+}
+
+static void gfx_v7_0_update_gfx_pg(struct amdgpu_device *adev, bool enable)
+{
+	gfx_v7_0_enable_gfx_cgpg(adev, enable);
+	gfx_v7_0_enable_gfx_static_mgpg(adev, enable);
+	gfx_v7_0_enable_gfx_dynamic_mgpg(adev, enable);
+}
+
+static u32 gfx_v7_0_get_csb_size(struct amdgpu_device *adev)
+{
+	u32 count = 0;
+	const struct cs_section_def *sect = NULL;
+	const struct cs_extent_def *ext = NULL;
+
+	if (adev->gfx.rlc.cs_data == NULL)
+		return 0;
+
+	/* begin clear state */
+	count += 2;
+	/* context control state */
+	count += 3;
+
+	for (sect = adev->gfx.rlc.cs_data; sect->section != NULL; ++sect) {
+		for (ext = sect->section; ext->extent != NULL; ++ext) {
+			if (sect->id == SECT_CONTEXT)
+				count += 2 + ext->reg_count;
+			else
+				return 0;
+		}
+	}
+	/* pa_sc_raster_config/pa_sc_raster_config1 */
+	count += 4;
+	/* end clear state */
+	count += 2;
+	/* clear state */
+	count += 2;
+
+	return count;
+}
+
+static void gfx_v7_0_get_csb_buffer(struct amdgpu_device *adev,
+				    volatile u32 *buffer)
+{
+	u32 count = 0, i;
+	const struct cs_section_def *sect = NULL;
+	const struct cs_extent_def *ext = NULL;
+
+	if (adev->gfx.rlc.cs_data == NULL)
+		return;
+	if (buffer == NULL)
+		return;
+
+	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_PREAMBLE_CNTL, 0));
+	buffer[count++] = cpu_to_le32(PACKET3_PREAMBLE_BEGIN_CLEAR_STATE);
+
+	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_CONTEXT_CONTROL, 1));
+	buffer[count++] = cpu_to_le32(0x80000000);
+	buffer[count++] = cpu_to_le32(0x80000000);
+
+	for (sect = adev->gfx.rlc.cs_data; sect->section != NULL; ++sect) {
+		for (ext = sect->section; ext->extent != NULL; ++ext) {
+			if (sect->id == SECT_CONTEXT) {
+				buffer[count++] =
+					cpu_to_le32(PACKET3(PACKET3_SET_CONTEXT_REG, ext->reg_count));
+				buffer[count++] = cpu_to_le32(ext->reg_index - PACKET3_SET_CONTEXT_REG_START);
+				for (i = 0; i < ext->reg_count; i++)
+					buffer[count++] = cpu_to_le32(ext->extent[i]);
+			} else {
+				return;
+			}
+		}
+	}
+
+	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_SET_CONTEXT_REG, 2));
+	buffer[count++] = cpu_to_le32(mmPA_SC_RASTER_CONFIG - PACKET3_SET_CONTEXT_REG_START);
+	switch (adev->asic_type) {
+	case CHIP_BONAIRE:
+		buffer[count++] = cpu_to_le32(0x16000012);
+		buffer[count++] = cpu_to_le32(0x00000000);
+		break;
+	case CHIP_KAVERI:
+		buffer[count++] = cpu_to_le32(0x00000000); /* XXX */
+		buffer[count++] = cpu_to_le32(0x00000000);
+		break;
+	case CHIP_KABINI:
+	case CHIP_MULLINS:
+		buffer[count++] = cpu_to_le32(0x00000000); /* XXX */
+		buffer[count++] = cpu_to_le32(0x00000000);
+		break;
+	case CHIP_HAWAII:
+		buffer[count++] = cpu_to_le32(0x3a00161a);
+		buffer[count++] = cpu_to_le32(0x0000002e);
+		break;
+	default:
+		buffer[count++] = cpu_to_le32(0x00000000);
+		buffer[count++] = cpu_to_le32(0x00000000);
+		break;
+	}
+
+	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_PREAMBLE_CNTL, 0));
+	buffer[count++] = cpu_to_le32(PACKET3_PREAMBLE_END_CLEAR_STATE);
+
+	buffer[count++] = cpu_to_le32(PACKET3(PACKET3_CLEAR_STATE, 0));
+	buffer[count++] = cpu_to_le32(0);
+}
+
+static void gfx_v7_0_init_pg(struct amdgpu_device *adev)
+{
+	if (adev->pg_flags & (AMDGPU_PG_SUPPORT_GFX_PG |
+			      AMDGPU_PG_SUPPORT_GFX_SMG |
+			      AMDGPU_PG_SUPPORT_GFX_DMG |
+			      AMDGPU_PG_SUPPORT_CP |
+			      AMDGPU_PG_SUPPORT_GDS |
+			      AMDGPU_PG_SUPPORT_RLC_SMU_HS)) {
+		gfx_v7_0_enable_sclk_slowdown_on_pu(adev, true);
+		gfx_v7_0_enable_sclk_slowdown_on_pd(adev, true);
+		if (adev->pg_flags & AMDGPU_PG_SUPPORT_GFX_PG) {
+			gfx_v7_0_init_gfx_cgpg(adev);
+			gfx_v7_0_enable_cp_pg(adev, true);
+			gfx_v7_0_enable_gds_pg(adev, true);
+		}
+		gfx_v7_0_init_ao_cu_mask(adev);
+		gfx_v7_0_update_gfx_pg(adev, true);
+	}
+}
+
+static void gfx_v7_0_fini_pg(struct amdgpu_device *adev)
+{
+	if (adev->pg_flags & (AMDGPU_PG_SUPPORT_GFX_PG |
+			      AMDGPU_PG_SUPPORT_GFX_SMG |
+			      AMDGPU_PG_SUPPORT_GFX_DMG |
+			      AMDGPU_PG_SUPPORT_CP |
+			      AMDGPU_PG_SUPPORT_GDS |
+			      AMDGPU_PG_SUPPORT_RLC_SMU_HS)) {
+		gfx_v7_0_update_gfx_pg(adev, false);
+		if (adev->pg_flags & AMDGPU_PG_SUPPORT_GFX_PG) {
+			gfx_v7_0_enable_cp_pg(adev, false);
+			gfx_v7_0_enable_gds_pg(adev, false);
+		}
+	}
+}
+
+/**
+ * gfx_v7_0_get_gpu_clock_counter - return GPU clock counter snapshot
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Fetches a GPU clock counter snapshot (SI).
+ * Returns the 64 bit clock counter snapshot.
+ */
+uint64_t gfx_v7_0_get_gpu_clock_counter(struct amdgpu_device *adev)
+{
+	uint64_t clock;
+
+	mutex_lock(&adev->gfx.gpu_clock_mutex);
+	WREG32(mmRLC_CAPTURE_GPU_CLOCK_COUNT, 1);
+	clock = (uint64_t)RREG32(mmRLC_GPU_CLOCK_COUNT_LSB) |
+		((uint64_t)RREG32(mmRLC_GPU_CLOCK_COUNT_MSB) << 32ULL);
+	mutex_unlock(&adev->gfx.gpu_clock_mutex);
+	return clock;
+}
+
+static void gfx_v7_0_ring_emit_gds_switch(struct amdgpu_ring *ring,
+					  uint32_t vmid,
+					  uint32_t gds_base, uint32_t gds_size,
+					  uint32_t gws_base, uint32_t gws_size,
+					  uint32_t oa_base, uint32_t oa_size)
+{
+	gds_base = gds_base >> AMDGPU_GDS_SHIFT;
+	gds_size = gds_size >> AMDGPU_GDS_SHIFT;
+
+	gws_base = gws_base >> AMDGPU_GWS_SHIFT;
+	gws_size = gws_size >> AMDGPU_GWS_SHIFT;
+
+	oa_base = oa_base >> AMDGPU_OA_SHIFT;
+	oa_size = oa_size >> AMDGPU_OA_SHIFT;
+
+	/* GDS Base */
+	amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3));
+	amdgpu_ring_write(ring, (WRITE_DATA_ENGINE_SEL(0) |
+				WRITE_DATA_DST_SEL(0)));
+	amdgpu_ring_write(ring, amdgpu_gds_reg_offset[vmid].mem_base);
+	amdgpu_ring_write(ring, 0);
+	amdgpu_ring_write(ring, gds_base);
+
+	/* GDS Size */
+	amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3));
+	amdgpu_ring_write(ring, (WRITE_DATA_ENGINE_SEL(0) |
+				WRITE_DATA_DST_SEL(0)));
+	amdgpu_ring_write(ring, amdgpu_gds_reg_offset[vmid].mem_size);
+	amdgpu_ring_write(ring, 0);
+	amdgpu_ring_write(ring, gds_size);
+
+	/* GWS */
+	amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3));
+	amdgpu_ring_write(ring, (WRITE_DATA_ENGINE_SEL(0) |
+				WRITE_DATA_DST_SEL(0)));
+	amdgpu_ring_write(ring, amdgpu_gds_reg_offset[vmid].gws);
+	amdgpu_ring_write(ring, 0);
+	amdgpu_ring_write(ring, gws_size << GDS_GWS_VMID0__SIZE__SHIFT | gws_base);
+
+	/* OA */
+	amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3));
+	amdgpu_ring_write(ring, (WRITE_DATA_ENGINE_SEL(0) |
+				WRITE_DATA_DST_SEL(0)));
+	amdgpu_ring_write(ring, amdgpu_gds_reg_offset[vmid].oa);
+	amdgpu_ring_write(ring, 0);
+	amdgpu_ring_write(ring, (1 << (oa_size + oa_base)) - (1 << oa_base));
+}
+
+static int gfx_v7_0_early_init(struct amdgpu_device *adev)
+{
+
+	adev->gfx.num_gfx_rings = GFX7_NUM_GFX_RINGS;
+	adev->gfx.num_compute_rings = GFX7_NUM_COMPUTE_RINGS;
+	gfx_v7_0_set_ring_funcs(adev);
+	gfx_v7_0_set_irq_funcs(adev);
+	gfx_v7_0_set_gds_init(adev);
+
+	return 0;
+}
+
+static int gfx_v7_0_sw_init(struct amdgpu_device *adev)
+{
+	struct amdgpu_ring *ring;
+	int i, r;
+
+	/* EOP Event */
+	r = amdgpu_irq_add_id(adev, 181, &adev->gfx.eop_irq);
+	if (r)
+		return r;
+
+	/* Privileged reg */
+	r = amdgpu_irq_add_id(adev, 184, &adev->gfx.priv_reg_irq);
+	if (r)
+		return r;
+
+	/* Privileged inst */
+	r = amdgpu_irq_add_id(adev, 185, &adev->gfx.priv_inst_irq);
+	if (r)
+		return r;
+
+	gfx_v7_0_scratch_init(adev);
+
+	r = gfx_v7_0_init_microcode(adev);
+	if (r) {
+		DRM_ERROR("Failed to load gfx firmware!\n");
+		return r;
+	}
+
+	r = gfx_v7_0_rlc_init(adev);
+	if (r) {
+		DRM_ERROR("Failed to init rlc BOs!\n");
+		return r;
+	}
+
+	/* allocate mec buffers */
+	r = gfx_v7_0_mec_init(adev);
+	if (r) {
+		DRM_ERROR("Failed to init MEC BOs!\n");
+		return r;
+	}
+
+	r = amdgpu_wb_get(adev, &adev->gfx.ce_sync_offs);
+	if (r) {
+		DRM_ERROR("(%d) gfx.ce_sync_offs wb alloc failed\n", r);
+		return r;
+	}
+
+	for (i = 0; i < adev->gfx.num_gfx_rings; i++) {
+		ring = &adev->gfx.gfx_ring[i];
+		ring->ring_obj = NULL;
+		sprintf(ring->name, "gfx");
+		r = amdgpu_ring_init(adev, ring, 1024 * 1024,
+				     PACKET3(PACKET3_NOP, 0x3FFF), 0xf,
+				     &adev->gfx.eop_irq, AMDGPU_CP_IRQ_GFX_EOP,
+				     AMDGPU_RING_TYPE_GFX);
+		if (r)
+			return r;
+	}
+
+	/* set up the compute queues */
+	for (i = 0; i < adev->gfx.num_compute_rings; i++) {
+		unsigned irq_type;
+
+		/* max 32 queues per MEC */
+		if ((i >= 32) || (i >= AMDGPU_MAX_COMPUTE_RINGS)) {
+			DRM_ERROR("Too many (%d) compute rings!\n", i);
+			break;
+		}
+		ring = &adev->gfx.compute_ring[i];
+		ring->ring_obj = NULL;
+		ring->use_doorbell = true;
+		ring->doorbell_index = AMDGPU_DOORBELL_MEC_RING0 + i;
+		ring->me = 1; /* first MEC */
+		ring->pipe = i / 8;
+		ring->queue = i % 8;
+		sprintf(ring->name, "comp %d.%d.%d", ring->me, ring->pipe, ring->queue);
+		irq_type = AMDGPU_CP_IRQ_COMPUTE_MEC1_PIPE0_EOP + ring->pipe;
+		/* type-2 packets are deprecated on MEC, use type-3 instead */
+		r = amdgpu_ring_init(adev, ring, 1024 * 1024,
+				     PACKET3(PACKET3_NOP, 0x3FFF), 0xf,
+				     &adev->gfx.eop_irq, irq_type,
+				     AMDGPU_RING_TYPE_COMPUTE);
+		if (r)
+			return r;
+	}
+
+	/* reserve GDS, GWS and OA resource for gfx */
+	r = amdgpu_bo_create(adev, adev->gds.mem.gfx_partition_size,
+			PAGE_SIZE, true,
+			AMDGPU_GEM_DOMAIN_GDS, 0,
+			NULL, &adev->gds.gds_gfx_bo);
+	if (r)
+		return r;
+
+	r = amdgpu_bo_create(adev, adev->gds.gws.gfx_partition_size,
+		PAGE_SIZE, true,
+		AMDGPU_GEM_DOMAIN_GWS, 0,
+		NULL, &adev->gds.gws_gfx_bo);
+	if (r)
+		return r;
+
+	r = amdgpu_bo_create(adev, adev->gds.oa.gfx_partition_size,
+			PAGE_SIZE, true,
+			AMDGPU_GEM_DOMAIN_OA, 0,
+			NULL, &adev->gds.oa_gfx_bo);
+	if (r)
+		return r;
+
+	return r;
+}
+
+static int gfx_v7_0_sw_fini(struct amdgpu_device *adev)
+{
+	int i;
+
+	amdgpu_bo_unref(&adev->gds.oa_gfx_bo);
+	amdgpu_bo_unref(&adev->gds.gws_gfx_bo);
+	amdgpu_bo_unref(&adev->gds.gds_gfx_bo);
+
+	for (i = 0; i < adev->gfx.num_gfx_rings; i++)
+		amdgpu_ring_fini(&adev->gfx.gfx_ring[i]);
+	for (i = 0; i < adev->gfx.num_compute_rings; i++)
+		amdgpu_ring_fini(&adev->gfx.compute_ring[i]);
+
+	amdgpu_wb_free(adev, adev->gfx.ce_sync_offs);
+
+	gfx_v7_0_cp_compute_fini(adev);
+	gfx_v7_0_rlc_fini(adev);
+	gfx_v7_0_mec_fini(adev);
+
+	return 0;
+}
+
+static int gfx_v7_0_hw_init(struct amdgpu_device *adev)
+{
+	int r;
+
+	gfx_v7_0_gpu_init(adev);
+
+	/* init rlc */
+	r = gfx_v7_0_rlc_resume(adev);
+	if (r)
+		return r;
+
+	r = gfx_v7_0_cp_resume(adev);
+	if (r)
+		return r;
+
+	return r;
+}
+
+static int gfx_v7_0_hw_fini(struct amdgpu_device *adev)
+{
+	gfx_v7_0_cp_enable(adev, false);
+	gfx_v7_0_rlc_stop(adev);
+	gfx_v7_0_fini_pg(adev);
+
+	return 0;
+}
+
+static int gfx_v7_0_suspend(struct amdgpu_device *adev)
+{
+	return gfx_v7_0_hw_fini(adev);
+}
+
+static int gfx_v7_0_resume(struct amdgpu_device *adev)
+{
+	return gfx_v7_0_hw_init(adev);
+}
+
+static bool gfx_v7_0_is_idle(struct amdgpu_device *adev)
+{
+	if (RREG32(mmGRBM_STATUS) & GRBM_STATUS__GUI_ACTIVE_MASK)
+		return false;
+	else
+		return true;
+}
+
+static int gfx_v7_0_wait_for_idle(struct amdgpu_device *adev)
+{
+	unsigned i;
+	u32 tmp;
+
+	for (i = 0; i < adev->usec_timeout; i++) {
+		/* read MC_STATUS */
+		tmp = RREG32(mmGRBM_STATUS) & GRBM_STATUS__GUI_ACTIVE_MASK;
+
+		if (!tmp)
+			return 0;
+		udelay(1);
+	}
+	return -ETIMEDOUT;
+}
+
+static void gfx_v7_0_print_status(struct amdgpu_device *adev)
+{
+	int i;
+
+	dev_info(adev->dev, "GFX 7.x registers\n");
+	dev_info(adev->dev, "  GRBM_STATUS=0x%08X\n",
+		RREG32(mmGRBM_STATUS));
+	dev_info(adev->dev, "  GRBM_STATUS2=0x%08X\n",
+		RREG32(mmGRBM_STATUS2));
+	dev_info(adev->dev, "  GRBM_STATUS_SE0=0x%08X\n",
+		RREG32(mmGRBM_STATUS_SE0));
+	dev_info(adev->dev, "  GRBM_STATUS_SE1=0x%08X\n",
+		RREG32(mmGRBM_STATUS_SE1));
+	dev_info(adev->dev, "  GRBM_STATUS_SE2=0x%08X\n",
+		RREG32(mmGRBM_STATUS_SE2));
+	dev_info(adev->dev, "  GRBM_STATUS_SE3=0x%08X\n",
+		RREG32(mmGRBM_STATUS_SE3));
+	dev_info(adev->dev, "  CP_STAT = 0x%08x\n", RREG32(mmCP_STAT));
+	dev_info(adev->dev, "  CP_STALLED_STAT1 = 0x%08x\n",
+		 RREG32(mmCP_STALLED_STAT1));
+	dev_info(adev->dev, "  CP_STALLED_STAT2 = 0x%08x\n",
+		 RREG32(mmCP_STALLED_STAT2));
+	dev_info(adev->dev, "  CP_STALLED_STAT3 = 0x%08x\n",
+		 RREG32(mmCP_STALLED_STAT3));
+	dev_info(adev->dev, "  CP_CPF_BUSY_STAT = 0x%08x\n",
+		 RREG32(mmCP_CPF_BUSY_STAT));
+	dev_info(adev->dev, "  CP_CPF_STALLED_STAT1 = 0x%08x\n",
+		 RREG32(mmCP_CPF_STALLED_STAT1));
+	dev_info(adev->dev, "  CP_CPF_STATUS = 0x%08x\n", RREG32(mmCP_CPF_STATUS));
+	dev_info(adev->dev, "  CP_CPC_BUSY_STAT = 0x%08x\n", RREG32(mmCP_CPC_BUSY_STAT));
+	dev_info(adev->dev, "  CP_CPC_STALLED_STAT1 = 0x%08x\n",
+		 RREG32(mmCP_CPC_STALLED_STAT1));
+	dev_info(adev->dev, "  CP_CPC_STATUS = 0x%08x\n", RREG32(mmCP_CPC_STATUS));
+
+	for (i = 0; i < 32; i++) {
+		dev_info(adev->dev, "  GB_TILE_MODE%d=0x%08X\n",
+			 i, RREG32(mmGB_TILE_MODE0 + (i * 4)));
+	}
+	for (i = 0; i < 16; i++) {
+		dev_info(adev->dev, "  GB_MACROTILE_MODE%d=0x%08X\n",
+			 i, RREG32(mmGB_MACROTILE_MODE0 + (i * 4)));
+	}
+	for (i = 0; i < adev->gfx.config.max_shader_engines; i++) {
+		dev_info(adev->dev, "  se: %d\n", i);
+		gfx_v7_0_select_se_sh(adev, i, 0xffffffff);
+		dev_info(adev->dev, "  PA_SC_RASTER_CONFIG=0x%08X\n",
+			 RREG32(mmPA_SC_RASTER_CONFIG));
+		dev_info(adev->dev, "  PA_SC_RASTER_CONFIG_1=0x%08X\n",
+			 RREG32(mmPA_SC_RASTER_CONFIG_1));
+	}
+	gfx_v7_0_select_se_sh(adev, 0xffffffff, 0xffffffff);
+
+	dev_info(adev->dev, "  GB_ADDR_CONFIG=0x%08X\n",
+		 RREG32(mmGB_ADDR_CONFIG));
+	dev_info(adev->dev, "  HDP_ADDR_CONFIG=0x%08X\n",
+		 RREG32(mmHDP_ADDR_CONFIG));
+	dev_info(adev->dev, "  DMIF_ADDR_CALC=0x%08X\n",
+		 RREG32(mmDMIF_ADDR_CALC));
+	dev_info(adev->dev, "  SDMA0_TILING_CONFIG=0x%08X\n",
+		 RREG32(mmSDMA0_TILING_CONFIG + SDMA0_REGISTER_OFFSET));
+	dev_info(adev->dev, "  SDMA1_TILING_CONFIG=0x%08X\n",
+		 RREG32(mmSDMA0_TILING_CONFIG + SDMA1_REGISTER_OFFSET));
+	dev_info(adev->dev, "  UVD_UDEC_ADDR_CONFIG=0x%08X\n",
+		 RREG32(mmUVD_UDEC_ADDR_CONFIG));
+	dev_info(adev->dev, "  UVD_UDEC_DB_ADDR_CONFIG=0x%08X\n",
+		 RREG32(mmUVD_UDEC_DB_ADDR_CONFIG));
+	dev_info(adev->dev, "  UVD_UDEC_DBW_ADDR_CONFIG=0x%08X\n",
+		 RREG32(mmUVD_UDEC_DBW_ADDR_CONFIG));
+
+	dev_info(adev->dev, "  CP_MEQ_THRESHOLDS=0x%08X\n",
+		 RREG32(mmCP_MEQ_THRESHOLDS));
+	dev_info(adev->dev, "  SX_DEBUG_1=0x%08X\n",
+		 RREG32(mmSX_DEBUG_1));
+	dev_info(adev->dev, "  TA_CNTL_AUX=0x%08X\n",
+		 RREG32(mmTA_CNTL_AUX));
+	dev_info(adev->dev, "  SPI_CONFIG_CNTL=0x%08X\n",
+		 RREG32(mmSPI_CONFIG_CNTL));
+	dev_info(adev->dev, "  SQ_CONFIG=0x%08X\n",
+		 RREG32(mmSQ_CONFIG));
+	dev_info(adev->dev, "  DB_DEBUG=0x%08X\n",
+		 RREG32(mmDB_DEBUG));
+	dev_info(adev->dev, "  DB_DEBUG2=0x%08X\n",
+		 RREG32(mmDB_DEBUG2));
+	dev_info(adev->dev, "  DB_DEBUG3=0x%08X\n",
+		 RREG32(mmDB_DEBUG3));
+	dev_info(adev->dev, "  CB_HW_CONTROL=0x%08X\n",
+		 RREG32(mmCB_HW_CONTROL));
+	dev_info(adev->dev, "  SPI_CONFIG_CNTL_1=0x%08X\n",
+		 RREG32(mmSPI_CONFIG_CNTL_1));
+	dev_info(adev->dev, "  PA_SC_FIFO_SIZE=0x%08X\n",
+		 RREG32(mmPA_SC_FIFO_SIZE));
+	dev_info(adev->dev, "  VGT_NUM_INSTANCES=0x%08X\n",
+		 RREG32(mmVGT_NUM_INSTANCES));
+	dev_info(adev->dev, "  CP_PERFMON_CNTL=0x%08X\n",
+		 RREG32(mmCP_PERFMON_CNTL));
+	dev_info(adev->dev, "  PA_SC_FORCE_EOV_MAX_CNTS=0x%08X\n",
+		 RREG32(mmPA_SC_FORCE_EOV_MAX_CNTS));
+	dev_info(adev->dev, "  VGT_CACHE_INVALIDATION=0x%08X\n",
+		 RREG32(mmVGT_CACHE_INVALIDATION));
+	dev_info(adev->dev, "  VGT_GS_VERTEX_REUSE=0x%08X\n",
+		 RREG32(mmVGT_GS_VERTEX_REUSE));
+	dev_info(adev->dev, "  PA_SC_LINE_STIPPLE_STATE=0x%08X\n",
+		 RREG32(mmPA_SC_LINE_STIPPLE_STATE));
+	dev_info(adev->dev, "  PA_CL_ENHANCE=0x%08X\n",
+		 RREG32(mmPA_CL_ENHANCE));
+	dev_info(adev->dev, "  PA_SC_ENHANCE=0x%08X\n",
+		 RREG32(mmPA_SC_ENHANCE));
+
+	dev_info(adev->dev, "  CP_ME_CNTL=0x%08X\n",
+		 RREG32(mmCP_ME_CNTL));
+	dev_info(adev->dev, "  CP_MAX_CONTEXT=0x%08X\n",
+		 RREG32(mmCP_MAX_CONTEXT));
+	dev_info(adev->dev, "  CP_ENDIAN_SWAP=0x%08X\n",
+		 RREG32(mmCP_ENDIAN_SWAP));
+	dev_info(adev->dev, "  CP_DEVICE_ID=0x%08X\n",
+		 RREG32(mmCP_DEVICE_ID));
+
+	dev_info(adev->dev, "  CP_SEM_WAIT_TIMER=0x%08X\n",
+		 RREG32(mmCP_SEM_WAIT_TIMER));
+	if (adev->asic_type != CHIP_HAWAII)
+		dev_info(adev->dev, "  CP_SEM_INCOMPLETE_TIMER_CNTL=0x%08X\n",
+			 RREG32(mmCP_SEM_INCOMPLETE_TIMER_CNTL));
+
+	dev_info(adev->dev, "  CP_RB_WPTR_DELAY=0x%08X\n",
+		 RREG32(mmCP_RB_WPTR_DELAY));
+	dev_info(adev->dev, "  CP_RB_VMID=0x%08X\n",
+		 RREG32(mmCP_RB_VMID));
+	dev_info(adev->dev, "  CP_RB0_CNTL=0x%08X\n",
+		 RREG32(mmCP_RB0_CNTL));
+	dev_info(adev->dev, "  CP_RB0_WPTR=0x%08X\n",
+		 RREG32(mmCP_RB0_WPTR));
+	dev_info(adev->dev, "  CP_RB0_RPTR_ADDR=0x%08X\n",
+		 RREG32(mmCP_RB0_RPTR_ADDR));
+	dev_info(adev->dev, "  CP_RB0_RPTR_ADDR_HI=0x%08X\n",
+		 RREG32(mmCP_RB0_RPTR_ADDR_HI));
+	dev_info(adev->dev, "  CP_RB0_CNTL=0x%08X\n",
+		 RREG32(mmCP_RB0_CNTL));
+	dev_info(adev->dev, "  CP_RB0_BASE=0x%08X\n",
+		 RREG32(mmCP_RB0_BASE));
+	dev_info(adev->dev, "  CP_RB0_BASE_HI=0x%08X\n",
+		 RREG32(mmCP_RB0_BASE_HI));
+	dev_info(adev->dev, "  CP_MEC_CNTL=0x%08X\n",
+		 RREG32(mmCP_MEC_CNTL));
+	dev_info(adev->dev, "  CP_CPF_DEBUG=0x%08X\n",
+		 RREG32(mmCP_CPF_DEBUG));
+
+	dev_info(adev->dev, "  SCRATCH_ADDR=0x%08X\n",
+		 RREG32(mmSCRATCH_ADDR));
+	dev_info(adev->dev, "  SCRATCH_UMSK=0x%08X\n",
+		 RREG32(mmSCRATCH_UMSK));
+
+	/* init the pipes */
+	mutex_lock(&adev->srbm_mutex);
+	for (i = 0; i < (adev->gfx.mec.num_pipe * adev->gfx.mec.num_mec); i++) {
+		int me = (i < 4) ? 1 : 2;
+		int pipe = (i < 4) ? i : (i - 4);
+		int queue;
+
+		dev_info(adev->dev, "  me: %d, pipe: %d\n", me, pipe);
+		cik_srbm_select(adev, me, pipe, 0, 0);
+		dev_info(adev->dev, "  CP_HPD_EOP_BASE_ADDR=0x%08X\n",
+			 RREG32(mmCP_HPD_EOP_BASE_ADDR));
+		dev_info(adev->dev, "  CP_HPD_EOP_BASE_ADDR_HI=0x%08X\n",
+			 RREG32(mmCP_HPD_EOP_BASE_ADDR_HI));
+		dev_info(adev->dev, "  CP_HPD_EOP_VMID=0x%08X\n",
+			 RREG32(mmCP_HPD_EOP_VMID));
+		dev_info(adev->dev, "  CP_HPD_EOP_CONTROL=0x%08X\n",
+			 RREG32(mmCP_HPD_EOP_CONTROL));
+
+		for (queue = 0; queue < 8; i++) {
+			cik_srbm_select(adev, me, pipe, queue, 0);
+			dev_info(adev->dev, "  queue: %d\n", queue);
+			dev_info(adev->dev, "  CP_PQ_WPTR_POLL_CNTL=0x%08X\n",
+				 RREG32(mmCP_PQ_WPTR_POLL_CNTL));
+			dev_info(adev->dev, "  CP_HQD_PQ_DOORBELL_CONTROL=0x%08X\n",
+				 RREG32(mmCP_HQD_PQ_DOORBELL_CONTROL));
+			dev_info(adev->dev, "  CP_HQD_ACTIVE=0x%08X\n",
+				 RREG32(mmCP_HQD_ACTIVE));
+			dev_info(adev->dev, "  CP_HQD_DEQUEUE_REQUEST=0x%08X\n",
+				 RREG32(mmCP_HQD_DEQUEUE_REQUEST));
+			dev_info(adev->dev, "  CP_HQD_PQ_RPTR=0x%08X\n",
+				 RREG32(mmCP_HQD_PQ_RPTR));
+			dev_info(adev->dev, "  CP_HQD_PQ_WPTR=0x%08X\n",
+				 RREG32(mmCP_HQD_PQ_WPTR));
+			dev_info(adev->dev, "  CP_HQD_PQ_BASE=0x%08X\n",
+				 RREG32(mmCP_HQD_PQ_BASE));
+			dev_info(adev->dev, "  CP_HQD_PQ_BASE_HI=0x%08X\n",
+				 RREG32(mmCP_HQD_PQ_BASE_HI));
+			dev_info(adev->dev, "  CP_HQD_PQ_CONTROL=0x%08X\n",
+				 RREG32(mmCP_HQD_PQ_CONTROL));
+			dev_info(adev->dev, "  CP_HQD_PQ_WPTR_POLL_ADDR=0x%08X\n",
+				 RREG32(mmCP_HQD_PQ_WPTR_POLL_ADDR));
+			dev_info(adev->dev, "  CP_HQD_PQ_WPTR_POLL_ADDR_HI=0x%08X\n",
+				 RREG32(mmCP_HQD_PQ_WPTR_POLL_ADDR_HI));
+			dev_info(adev->dev, "  CP_HQD_PQ_RPTR_REPORT_ADDR=0x%08X\n",
+				 RREG32(mmCP_HQD_PQ_RPTR_REPORT_ADDR));
+			dev_info(adev->dev, "  CP_HQD_PQ_RPTR_REPORT_ADDR_HI=0x%08X\n",
+				 RREG32(mmCP_HQD_PQ_RPTR_REPORT_ADDR_HI));
+			dev_info(adev->dev, "  CP_HQD_PQ_DOORBELL_CONTROL=0x%08X\n",
+				 RREG32(mmCP_HQD_PQ_DOORBELL_CONTROL));
+			dev_info(adev->dev, "  CP_HQD_PQ_WPTR=0x%08X\n",
+				 RREG32(mmCP_HQD_PQ_WPTR));
+			dev_info(adev->dev, "  CP_HQD_VMID=0x%08X\n",
+				 RREG32(mmCP_HQD_VMID));
+			dev_info(adev->dev, "  CP_MQD_BASE_ADDR=0x%08X\n",
+				 RREG32(mmCP_MQD_BASE_ADDR));
+			dev_info(adev->dev, "  CP_MQD_BASE_ADDR_HI=0x%08X\n",
+				 RREG32(mmCP_MQD_BASE_ADDR_HI));
+			dev_info(adev->dev, "  CP_MQD_CONTROL=0x%08X\n",
+				 RREG32(mmCP_MQD_CONTROL));
+		}
+	}
+	cik_srbm_select(adev, 0, 0, 0, 0);
+	mutex_unlock(&adev->srbm_mutex);
+
+	dev_info(adev->dev, "  CP_INT_CNTL_RING0=0x%08X\n",
+		 RREG32(mmCP_INT_CNTL_RING0));
+	dev_info(adev->dev, "  RLC_LB_CNTL=0x%08X\n",
+		 RREG32(mmRLC_LB_CNTL));
+	dev_info(adev->dev, "  RLC_CNTL=0x%08X\n",
+		 RREG32(mmRLC_CNTL));
+	dev_info(adev->dev, "  RLC_CGCG_CGLS_CTRL=0x%08X\n",
+		 RREG32(mmRLC_CGCG_CGLS_CTRL));
+	dev_info(adev->dev, "  RLC_LB_CNTR_INIT=0x%08X\n",
+		 RREG32(mmRLC_LB_CNTR_INIT));
+	dev_info(adev->dev, "  RLC_LB_CNTR_MAX=0x%08X\n",
+		 RREG32(mmRLC_LB_CNTR_MAX));
+	dev_info(adev->dev, "  RLC_LB_INIT_CU_MASK=0x%08X\n",
+		 RREG32(mmRLC_LB_INIT_CU_MASK));
+	dev_info(adev->dev, "  RLC_LB_PARAMS=0x%08X\n",
+		 RREG32(mmRLC_LB_PARAMS));
+	dev_info(adev->dev, "  RLC_LB_CNTL=0x%08X\n",
+		 RREG32(mmRLC_LB_CNTL));
+	dev_info(adev->dev, "  RLC_MC_CNTL=0x%08X\n",
+		 RREG32(mmRLC_MC_CNTL));
+	dev_info(adev->dev, "  RLC_UCODE_CNTL=0x%08X\n",
+		 RREG32(mmRLC_UCODE_CNTL));
+
+	if (adev->asic_type == CHIP_BONAIRE)
+		dev_info(adev->dev, "  RLC_DRIVER_CPDMA_STATUS=0x%08X\n",
+			 RREG32(mmRLC_DRIVER_CPDMA_STATUS));
+
+	mutex_lock(&adev->srbm_mutex);
+	for (i = 0; i < 16; i++) {
+		cik_srbm_select(adev, 0, 0, 0, i);
+		dev_info(adev->dev, "  VM %d:\n", i);
+		dev_info(adev->dev, "  SH_MEM_CONFIG=0x%08X\n",
+			 RREG32(mmSH_MEM_CONFIG));
+		dev_info(adev->dev, "  SH_MEM_APE1_BASE=0x%08X\n",
+			 RREG32(mmSH_MEM_APE1_BASE));
+		dev_info(adev->dev, "  SH_MEM_APE1_LIMIT=0x%08X\n",
+			 RREG32(mmSH_MEM_APE1_LIMIT));
+		dev_info(adev->dev, "  SH_MEM_BASES=0x%08X\n",
+			 RREG32(mmSH_MEM_BASES));
+	}
+	cik_srbm_select(adev, 0, 0, 0, 0);
+	mutex_unlock(&adev->srbm_mutex);
+}
+
+static int gfx_v7_0_soft_reset(struct amdgpu_device *adev)
+{
+	u32 grbm_soft_reset = 0, srbm_soft_reset = 0;
+	u32 tmp;
+
+	/* GRBM_STATUS */
+	tmp = RREG32(mmGRBM_STATUS);
+	if (tmp & (GRBM_STATUS__PA_BUSY_MASK | GRBM_STATUS__SC_BUSY_MASK |
+		   GRBM_STATUS__BCI_BUSY_MASK | GRBM_STATUS__SX_BUSY_MASK |
+		   GRBM_STATUS__TA_BUSY_MASK | GRBM_STATUS__VGT_BUSY_MASK |
+		   GRBM_STATUS__DB_BUSY_MASK | GRBM_STATUS__CB_BUSY_MASK |
+		   GRBM_STATUS__GDS_BUSY_MASK | GRBM_STATUS__SPI_BUSY_MASK |
+		   GRBM_STATUS__IA_BUSY_MASK | GRBM_STATUS__IA_BUSY_NO_DMA_MASK))
+		grbm_soft_reset |= GRBM_SOFT_RESET__SOFT_RESET_CP_MASK |
+			GRBM_SOFT_RESET__SOFT_RESET_GFX_MASK;
+
+	if (tmp & (GRBM_STATUS__CP_BUSY_MASK | GRBM_STATUS__CP_COHERENCY_BUSY_MASK)) {
+		grbm_soft_reset |= GRBM_SOFT_RESET__SOFT_RESET_CP_MASK;
+		srbm_soft_reset |= SRBM_SOFT_RESET__SOFT_RESET_GRBM_MASK;
+	}
+
+	/* GRBM_STATUS2 */
+	tmp = RREG32(mmGRBM_STATUS2);
+	if (tmp & GRBM_STATUS2__RLC_BUSY_MASK)
+		grbm_soft_reset |= GRBM_SOFT_RESET__SOFT_RESET_RLC_MASK;
+
+	/* SRBM_STATUS */
+	tmp = RREG32(mmSRBM_STATUS);
+	if (tmp & SRBM_STATUS__GRBM_RQ_PENDING_MASK)
+		srbm_soft_reset |= SRBM_SOFT_RESET__SOFT_RESET_GRBM_MASK;
+
+	if (grbm_soft_reset || srbm_soft_reset) {
+		gfx_v7_0_print_status(adev);
+		/* disable CG/PG */
+		gfx_v7_0_fini_pg(adev);
+		gfx_v7_0_update_cg(adev, false);
+
+		/* stop the rlc */
+		gfx_v7_0_rlc_stop(adev);
+
+		/* Disable GFX parsing/prefetching */
+		WREG32(mmCP_ME_CNTL, CP_ME_CNTL__ME_HALT_MASK | CP_ME_CNTL__PFP_HALT_MASK | CP_ME_CNTL__CE_HALT_MASK);
+
+		/* Disable MEC parsing/prefetching */
+		WREG32(mmCP_MEC_CNTL, CP_MEC_CNTL__MEC_ME1_HALT_MASK | CP_MEC_CNTL__MEC_ME2_HALT_MASK);
+
+		if (grbm_soft_reset) {
+			tmp = RREG32(mmGRBM_SOFT_RESET);
+			tmp |= grbm_soft_reset;
+			dev_info(adev->dev, "GRBM_SOFT_RESET=0x%08X\n", tmp);
+			WREG32(mmGRBM_SOFT_RESET, tmp);
+			tmp = RREG32(mmGRBM_SOFT_RESET);
+
+			udelay(50);
+
+			tmp &= ~grbm_soft_reset;
+			WREG32(mmGRBM_SOFT_RESET, tmp);
+			tmp = RREG32(mmGRBM_SOFT_RESET);
+		}
+
+		if (srbm_soft_reset) {
+			tmp = RREG32(mmSRBM_SOFT_RESET);
+			tmp |= srbm_soft_reset;
+			dev_info(adev->dev, "SRBM_SOFT_RESET=0x%08X\n", tmp);
+			WREG32(mmSRBM_SOFT_RESET, tmp);
+			tmp = RREG32(mmSRBM_SOFT_RESET);
+
+			udelay(50);
+
+			tmp &= ~srbm_soft_reset;
+			WREG32(mmSRBM_SOFT_RESET, tmp);
+			tmp = RREG32(mmSRBM_SOFT_RESET);
+		}
+		/* Wait a little for things to settle down */
+		udelay(50);
+		gfx_v7_0_print_status(adev);
+	}
+	return 0;
+}
+
+static void gfx_v7_0_set_gfx_eop_interrupt_state(struct amdgpu_device *adev,
+						 enum amdgpu_interrupt_state state)
+{
+	u32 cp_int_cntl;
+
+	switch (state) {
+	case AMDGPU_IRQ_STATE_DISABLE:
+		cp_int_cntl = RREG32(mmCP_INT_CNTL_RING0);
+		cp_int_cntl &= ~CP_INT_CNTL_RING0__TIME_STAMP_INT_ENABLE_MASK;
+		WREG32(mmCP_INT_CNTL_RING0, cp_int_cntl);
+		break;
+	case AMDGPU_IRQ_STATE_ENABLE:
+		cp_int_cntl = RREG32(mmCP_INT_CNTL_RING0);
+		cp_int_cntl |= CP_INT_CNTL_RING0__TIME_STAMP_INT_ENABLE_MASK;
+		WREG32(mmCP_INT_CNTL_RING0, cp_int_cntl);
+		break;
+	default:
+		break;
+	}
+}
+
+static void gfx_v7_0_set_compute_eop_interrupt_state(struct amdgpu_device *adev,
+						     int me, int pipe,
+						     enum amdgpu_interrupt_state state)
+{
+	u32 mec_int_cntl, mec_int_cntl_reg;
+
+	/*
+	 * amdgpu controls only pipe 0 of MEC1. That's why this function only
+	 * handles the setting of interrupts for this specific pipe. All other
+	 * pipes' interrupts are set by amdkfd.
+	 */
+
+	if (me == 1) {
+		switch (pipe) {
+		case 0:
+			mec_int_cntl_reg = mmCP_ME1_PIPE0_INT_CNTL;
+			break;
+		default:
+			DRM_DEBUG("invalid pipe %d\n", pipe);
+			return;
+		}
+	} else {
+		DRM_DEBUG("invalid me %d\n", me);
+		return;
+	}
+
+	switch (state) {
+	case AMDGPU_IRQ_STATE_DISABLE:
+		mec_int_cntl = RREG32(mec_int_cntl_reg);
+		mec_int_cntl &= ~CP_INT_CNTL_RING0__TIME_STAMP_INT_ENABLE_MASK;
+		WREG32(mec_int_cntl_reg, mec_int_cntl);
+		break;
+	case AMDGPU_IRQ_STATE_ENABLE:
+		mec_int_cntl = RREG32(mec_int_cntl_reg);
+		mec_int_cntl |= CP_INT_CNTL_RING0__TIME_STAMP_INT_ENABLE_MASK;
+		WREG32(mec_int_cntl_reg, mec_int_cntl);
+		break;
+	default:
+		break;
+	}
+}
+
+static int gfx_v7_0_set_priv_reg_fault_state(struct amdgpu_device *adev,
+					     struct amdgpu_irq_src *src,
+					     unsigned type,
+					     enum amdgpu_interrupt_state state)
+{
+	u32 cp_int_cntl;
+
+	switch (state) {
+	case AMDGPU_IRQ_STATE_DISABLE:
+		cp_int_cntl = RREG32(mmCP_INT_CNTL_RING0);
+		cp_int_cntl &= ~CP_INT_CNTL_RING0__PRIV_REG_INT_ENABLE_MASK;
+		WREG32(mmCP_INT_CNTL_RING0, cp_int_cntl);
+		break;
+	case AMDGPU_IRQ_STATE_ENABLE:
+		cp_int_cntl = RREG32(mmCP_INT_CNTL_RING0);
+		cp_int_cntl |= CP_INT_CNTL_RING0__PRIV_REG_INT_ENABLE_MASK;
+		WREG32(mmCP_INT_CNTL_RING0, cp_int_cntl);
+		break;
+	default:
+		break;
+	}
+
+	return 0;
+}
+
+static int gfx_v7_0_set_priv_inst_fault_state(struct amdgpu_device *adev,
+					      struct amdgpu_irq_src *src,
+					      unsigned type,
+					      enum amdgpu_interrupt_state state)
+{
+	u32 cp_int_cntl;
+
+	switch (state) {
+	case AMDGPU_IRQ_STATE_DISABLE:
+		cp_int_cntl = RREG32(mmCP_INT_CNTL_RING0);
+		cp_int_cntl &= ~CP_INT_CNTL_RING0__PRIV_INSTR_INT_ENABLE_MASK;
+		WREG32(mmCP_INT_CNTL_RING0, cp_int_cntl);
+		break;
+	case AMDGPU_IRQ_STATE_ENABLE:
+		cp_int_cntl = RREG32(mmCP_INT_CNTL_RING0);
+		cp_int_cntl |= CP_INT_CNTL_RING0__PRIV_INSTR_INT_ENABLE_MASK;
+		WREG32(mmCP_INT_CNTL_RING0, cp_int_cntl);
+		break;
+	default:
+		break;
+	}
+
+	return 0;
+}
+
+static int gfx_v7_0_set_eop_interrupt_state(struct amdgpu_device *adev,
+					    struct amdgpu_irq_src *src,
+					    unsigned type,
+					    enum amdgpu_interrupt_state state)
+{
+	switch (type) {
+	case AMDGPU_CP_IRQ_GFX_EOP:
+		gfx_v7_0_set_gfx_eop_interrupt_state(adev, state);
+		break;
+	case AMDGPU_CP_IRQ_COMPUTE_MEC1_PIPE0_EOP:
+		gfx_v7_0_set_compute_eop_interrupt_state(adev, 1, 0, state);
+		break;
+	case AMDGPU_CP_IRQ_COMPUTE_MEC1_PIPE1_EOP:
+		gfx_v7_0_set_compute_eop_interrupt_state(adev, 1, 1, state);
+		break;
+	case AMDGPU_CP_IRQ_COMPUTE_MEC1_PIPE2_EOP:
+		gfx_v7_0_set_compute_eop_interrupt_state(adev, 1, 2, state);
+		break;
+	case AMDGPU_CP_IRQ_COMPUTE_MEC1_PIPE3_EOP:
+		gfx_v7_0_set_compute_eop_interrupt_state(adev, 1, 3, state);
+		break;
+	case AMDGPU_CP_IRQ_COMPUTE_MEC2_PIPE0_EOP:
+		gfx_v7_0_set_compute_eop_interrupt_state(adev, 2, 0, state);
+		break;
+	case AMDGPU_CP_IRQ_COMPUTE_MEC2_PIPE1_EOP:
+		gfx_v7_0_set_compute_eop_interrupt_state(adev, 2, 1, state);
+		break;
+	case AMDGPU_CP_IRQ_COMPUTE_MEC2_PIPE2_EOP:
+		gfx_v7_0_set_compute_eop_interrupt_state(adev, 2, 2, state);
+		break;
+	case AMDGPU_CP_IRQ_COMPUTE_MEC2_PIPE3_EOP:
+		gfx_v7_0_set_compute_eop_interrupt_state(adev, 2, 3, state);
+		break;
+	default:
+		break;
+	}
+	return 0;
+}
+
+static int gfx_v7_0_eop_irq(struct amdgpu_device *adev,
+			    struct amdgpu_irq_src *source,
+			    struct amdgpu_iv_entry *entry)
+{
+	u8 me_id, pipe_id;
+	struct amdgpu_ring *ring;
+	int i;
+
+	DRM_DEBUG("IH: CP EOP\n");
+	me_id = (entry->ring_id & 0x0c) >> 2;
+	pipe_id = (entry->ring_id & 0x03) >> 0;
+	switch (me_id) {
+	case 0:
+		amdgpu_fence_process(&adev->gfx.gfx_ring[0]);
+		break;
+	case 1:
+	case 2:
+		for (i = 0; i < adev->gfx.num_compute_rings; i++) {
+			ring = &adev->gfx.compute_ring[i];
+			if ((ring->me == me_id) & (ring->pipe == pipe_id))
+				amdgpu_fence_process(ring);
+		}
+		break;
+	}
+	return 0;
+}
+
+static int gfx_v7_0_priv_reg_irq(struct amdgpu_device *adev,
+				 struct amdgpu_irq_src *source,
+				 struct amdgpu_iv_entry *entry)
+{
+	DRM_ERROR("Illegal register access in command stream\n");
+	schedule_work(&adev->reset_work);
+	return 0;
+}
+
+static int gfx_v7_0_priv_inst_irq(struct amdgpu_device *adev,
+				  struct amdgpu_irq_src *source,
+				  struct amdgpu_iv_entry *entry)
+{
+	DRM_ERROR("Illegal instruction in command stream\n");
+	// XXX soft reset the gfx block only
+	schedule_work(&adev->reset_work);
+	return 0;
+}
+
+static int gfx_v7_0_set_clockgating_state(struct amdgpu_device *adev,
+					  enum amdgpu_clockgating_state state)
+{
+	bool gate = false;
+
+	if (state == AMDGPU_CG_STATE_GATE)
+		gate = true;
+
+	gfx_v7_0_enable_gui_idle_interrupt(adev, false);
+	/* order matters! */
+	if (gate) {
+		gfx_v7_0_enable_mgcg(adev, true);
+		gfx_v7_0_enable_cgcg(adev, true);
+	} else {
+		gfx_v7_0_enable_cgcg(adev, false);
+		gfx_v7_0_enable_mgcg(adev, false);
+	}
+	gfx_v7_0_enable_gui_idle_interrupt(adev, true);
+
+	return 0;
+}
+
+static int gfx_v7_0_set_powergating_state(struct amdgpu_device *adev,
+					  enum amdgpu_powergating_state state)
+{
+	bool gate = false;
+
+	if (state == AMDGPU_PG_STATE_GATE)
+		gate = true;
+
+	if (adev->pg_flags & (AMDGPU_PG_SUPPORT_GFX_PG |
+			      AMDGPU_PG_SUPPORT_GFX_SMG |
+			      AMDGPU_PG_SUPPORT_GFX_DMG |
+			      AMDGPU_PG_SUPPORT_CP |
+			      AMDGPU_PG_SUPPORT_GDS |
+			      AMDGPU_PG_SUPPORT_RLC_SMU_HS)) {
+		gfx_v7_0_update_gfx_pg(adev, gate);
+		if (adev->pg_flags & AMDGPU_PG_SUPPORT_GFX_PG) {
+			gfx_v7_0_enable_cp_pg(adev, gate);
+			gfx_v7_0_enable_gds_pg(adev, gate);
+		}
+	}
+
+	return 0;
+}
+
+const struct amdgpu_ip_funcs gfx_v7_0_ip_funcs = {
+	.early_init = gfx_v7_0_early_init,
+	.late_init = NULL,
+	.sw_init = gfx_v7_0_sw_init,
+	.sw_fini = gfx_v7_0_sw_fini,
+	.hw_init = gfx_v7_0_hw_init,
+	.hw_fini = gfx_v7_0_hw_fini,
+	.suspend = gfx_v7_0_suspend,
+	.resume = gfx_v7_0_resume,
+	.is_idle = gfx_v7_0_is_idle,
+	.wait_for_idle = gfx_v7_0_wait_for_idle,
+	.soft_reset = gfx_v7_0_soft_reset,
+	.print_status = gfx_v7_0_print_status,
+	.set_clockgating_state = gfx_v7_0_set_clockgating_state,
+	.set_powergating_state = gfx_v7_0_set_powergating_state,
+};
+
+/**
+ * gfx_v7_0_ring_is_lockup - check if the 3D engine is locked up
+ *
+ * @adev: amdgpu_device pointer
+ * @ring: amdgpu_ring structure holding ring information
+ *
+ * Check if the 3D engine is locked up (CIK).
+ * Returns true if the engine is locked, false if not.
+ */
+static bool gfx_v7_0_ring_is_lockup(struct amdgpu_ring *ring)
+{
+	if (gfx_v7_0_is_idle(ring->adev)) {
+		amdgpu_ring_lockup_update(ring);
+		return false;
+	}
+	return amdgpu_ring_test_lockup(ring);
+}
+
+static const struct amdgpu_ring_funcs gfx_v7_0_ring_funcs_gfx = {
+	.get_rptr = gfx_v7_0_ring_get_rptr_gfx,
+	.get_wptr = gfx_v7_0_ring_get_wptr_gfx,
+	.set_wptr = gfx_v7_0_ring_set_wptr_gfx,
+	.parse_cs = NULL,
+	.emit_ib = gfx_v7_0_ring_emit_ib,
+	.emit_fence = gfx_v7_0_ring_emit_fence_gfx,
+	.emit_semaphore = gfx_v7_0_ring_emit_semaphore,
+	.emit_vm_flush = gfx_v7_0_ring_emit_vm_flush,
+	.emit_gds_switch = gfx_v7_0_ring_emit_gds_switch,
+	.test_ring = gfx_v7_0_ring_test_ring,
+	.test_ib = gfx_v7_0_ring_test_ib,
+	.is_lockup = gfx_v7_0_ring_is_lockup,
+};
+
+static const struct amdgpu_ring_funcs gfx_v7_0_ring_funcs_compute = {
+	.get_rptr = gfx_v7_0_ring_get_rptr_compute,
+	.get_wptr = gfx_v7_0_ring_get_wptr_compute,
+	.set_wptr = gfx_v7_0_ring_set_wptr_compute,
+	.parse_cs = NULL,
+	.emit_ib = gfx_v7_0_ring_emit_ib,
+	.emit_fence = gfx_v7_0_ring_emit_fence_compute,
+	.emit_semaphore = gfx_v7_0_ring_emit_semaphore,
+	.emit_vm_flush = gfx_v7_0_ring_emit_vm_flush,
+	.emit_gds_switch = gfx_v7_0_ring_emit_gds_switch,
+	.test_ring = gfx_v7_0_ring_test_ring,
+	.test_ib = gfx_v7_0_ring_test_ib,
+	.is_lockup = gfx_v7_0_ring_is_lockup,
+};
+
+static void gfx_v7_0_set_ring_funcs(struct amdgpu_device *adev)
+{
+	int i;
+
+	for (i = 0; i < adev->gfx.num_gfx_rings; i++)
+		adev->gfx.gfx_ring[i].funcs = &gfx_v7_0_ring_funcs_gfx;
+	for (i = 0; i < adev->gfx.num_compute_rings; i++)
+		adev->gfx.compute_ring[i].funcs = &gfx_v7_0_ring_funcs_compute;
+}
+
+static const struct amdgpu_irq_src_funcs gfx_v7_0_eop_irq_funcs = {
+	.set = gfx_v7_0_set_eop_interrupt_state,
+	.process = gfx_v7_0_eop_irq,
+};
+
+static const struct amdgpu_irq_src_funcs gfx_v7_0_priv_reg_irq_funcs = {
+	.set = gfx_v7_0_set_priv_reg_fault_state,
+	.process = gfx_v7_0_priv_reg_irq,
+};
+
+static const struct amdgpu_irq_src_funcs gfx_v7_0_priv_inst_irq_funcs = {
+	.set = gfx_v7_0_set_priv_inst_fault_state,
+	.process = gfx_v7_0_priv_inst_irq,
+};
+
+static void gfx_v7_0_set_irq_funcs(struct amdgpu_device *adev)
+{
+	adev->gfx.eop_irq.num_types = AMDGPU_CP_IRQ_LAST;
+	adev->gfx.eop_irq.funcs = &gfx_v7_0_eop_irq_funcs;
+
+	adev->gfx.priv_reg_irq.num_types = 1;
+	adev->gfx.priv_reg_irq.funcs = &gfx_v7_0_priv_reg_irq_funcs;
+
+	adev->gfx.priv_inst_irq.num_types = 1;
+	adev->gfx.priv_inst_irq.funcs = &gfx_v7_0_priv_inst_irq_funcs;
+}
+
+static void gfx_v7_0_set_gds_init(struct amdgpu_device *adev)
+{
+	/* init asci gds info */
+	adev->gds.mem.total_size = RREG32(mmGDS_VMID0_SIZE);
+	adev->gds.gws.total_size = 64;
+	adev->gds.oa.total_size = 16;
+
+	if (adev->gds.mem.total_size == 64 * 1024) {
+		adev->gds.mem.gfx_partition_size = 4096;
+		adev->gds.mem.cs_partition_size = 4096;
+
+		adev->gds.gws.gfx_partition_size = 4;
+		adev->gds.gws.cs_partition_size = 4;
+
+		adev->gds.oa.gfx_partition_size = 4;
+		adev->gds.oa.cs_partition_size = 1;
+	} else {
+		adev->gds.mem.gfx_partition_size = 1024;
+		adev->gds.mem.cs_partition_size = 1024;
+
+		adev->gds.gws.gfx_partition_size = 16;
+		adev->gds.gws.cs_partition_size = 16;
+
+		adev->gds.oa.gfx_partition_size = 4;
+		adev->gds.oa.cs_partition_size = 4;
+	}
+}
+
+
+int gfx_v7_0_get_cu_info(struct amdgpu_device *adev,
+								   struct amdgpu_cu_info *cu_info)
+{
+	int i, j, k, counter, active_cu_number = 0;
+	u32 mask, bitmap, ao_bitmap, ao_cu_mask = 0;
+
+	if (!adev || !cu_info)
+		return -EINVAL;
+
+	mutex_lock(&adev->grbm_idx_mutex);
+	for (i = 0; i < adev->gfx.config.max_shader_engines; i++) {
+		for (j = 0; j < adev->gfx.config.max_sh_per_se; j++) {
+			mask = 1;
+			ao_bitmap = 0;
+			counter = 0;
+			bitmap = gfx_v7_0_get_cu_active_bitmap(adev, i, j);
+			cu_info->bitmap[i][j] = bitmap;
+
+			for (k = 0; k < adev->gfx.config.max_cu_per_sh; k ++) {
+				if (bitmap & mask) {
+					if (counter < 2)
+						ao_bitmap |= mask;
+					counter ++;
+				}
+				mask <<= 1;
+			}
+			active_cu_number += counter;
+			ao_cu_mask |= (ao_bitmap << (i * 16 + j * 8));
+		}
+	}
+
+	cu_info->number = active_cu_number;
+	cu_info->ao_cu_mask = ao_cu_mask;
+	mutex_unlock(&adev->grbm_idx_mutex);
+	return 0;
+}

drivers/gpu/drm/amd/amdgpu/gfx_v7_0.h

+/*
+ * Copyright 2014 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __GFX_V7_0_H__
+#define __GFX_V7_0_H__
+
+extern const struct amdgpu_ip_funcs gfx_v7_0_ip_funcs;
+
+/* XXX these shouldn't be exported */
+void gfx_v7_0_enter_rlc_safe_mode(struct amdgpu_device *adev);
+void gfx_v7_0_exit_rlc_safe_mode(struct amdgpu_device *adev);
+void gfx_v7_0_rlc_stop(struct amdgpu_device *adev);
+uint64_t gfx_v7_0_get_gpu_clock_counter(struct amdgpu_device *adev);
+void gfx_v7_0_select_se_sh(struct amdgpu_device *adev, u32 se_num, u32 sh_num);
+int gfx_v7_0_get_cu_info(struct amdgpu_device *adev, struct amdgpu_cu_info *cu_info);
+
+#endif

drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c

+/*
+ * Copyright 2014 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+#include <linux/firmware.h>
+#include "drmP.h"
+#include "amdgpu.h"
+#include "cikd.h"
+#include "cik.h"
+#include "gmc_v7_0.h"
+#include "amdgpu_ucode.h"
+
+#include "bif/bif_4_1_d.h"
+#include "bif/bif_4_1_sh_mask.h"
+
+#include "gmc/gmc_7_1_d.h"
+#include "gmc/gmc_7_1_sh_mask.h"
+
+#include "oss/oss_2_0_d.h"
+#include "oss/oss_2_0_sh_mask.h"
+
+static void gmc_v7_0_set_gart_funcs(struct amdgpu_device *adev);
+static void gmc_v7_0_set_irq_funcs(struct amdgpu_device *adev);
+
+MODULE_FIRMWARE("radeon/boniare_mc.bin");
+MODULE_FIRMWARE("radeon/hawaii_mc.bin");
+
+/**
+ * gmc8_mc_wait_for_idle - wait for MC idle callback.
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Wait for the MC (memory controller) to be idle.
+ * (evergreen+).
+ * Returns 0 if the MC is idle, -1 if not.
+ */
+int gmc_v7_0_mc_wait_for_idle(struct amdgpu_device *adev)
+{
+	unsigned i;
+	u32 tmp;
+
+	for (i = 0; i < adev->usec_timeout; i++) {
+		/* read MC_STATUS */
+		tmp = RREG32(mmSRBM_STATUS) & 0x1F00;
+		if (!tmp)
+			return 0;
+		udelay(1);
+	}
+	return -1;
+}
+
+void gmc_v7_0_mc_stop(struct amdgpu_device *adev,
+		      struct amdgpu_mode_mc_save *save)
+{
+	u32 blackout;
+
+	if (adev->mode_info.num_crtc)
+		amdgpu_display_stop_mc_access(adev, save);
+
+	amdgpu_asic_wait_for_mc_idle(adev);
+
+	blackout = RREG32(mmMC_SHARED_BLACKOUT_CNTL);
+	if (REG_GET_FIELD(blackout, MC_SHARED_BLACKOUT_CNTL, BLACKOUT_MODE) != 1) {
+		/* Block CPU access */
+		WREG32(mmBIF_FB_EN, 0);
+		/* blackout the MC */
+		blackout = REG_SET_FIELD(blackout,
+					 MC_SHARED_BLACKOUT_CNTL, BLACKOUT_MODE, 0);
+		WREG32(mmMC_SHARED_BLACKOUT_CNTL, blackout | 1);
+	}
+	/* wait for the MC to settle */
+	udelay(100);
+}
+
+void gmc_v7_0_mc_resume(struct amdgpu_device *adev,
+			struct amdgpu_mode_mc_save *save)
+{
+	u32 tmp;
+
+	/* unblackout the MC */
+	tmp = RREG32(mmMC_SHARED_BLACKOUT_CNTL);
+	tmp = REG_SET_FIELD(tmp, MC_SHARED_BLACKOUT_CNTL, BLACKOUT_MODE, 0);
+	WREG32(mmMC_SHARED_BLACKOUT_CNTL, tmp);
+	/* allow CPU access */
+	tmp = REG_SET_FIELD(0, BIF_FB_EN, FB_READ_EN, 1);
+	tmp = REG_SET_FIELD(tmp, BIF_FB_EN, FB_WRITE_EN, 1);
+	WREG32(mmBIF_FB_EN, tmp);
+
+	if (adev->mode_info.num_crtc)
+		amdgpu_display_resume_mc_access(adev, save);
+}
+
+/**
+ * gmc_v7_0_init_microcode - load ucode images from disk
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Use the firmware interface to load the ucode images into
+ * the driver (not loaded into hw).
+ * Returns 0 on success, error on failure.
+ */
+static int gmc_v7_0_init_microcode(struct amdgpu_device *adev)
+{
+	const char *chip_name;
+	char fw_name[30];
+	int err;
+
+	DRM_DEBUG("\n");
+
+	switch (adev->asic_type) {
+	case CHIP_BONAIRE:
+		chip_name = "bonaire";
+		break;
+	case CHIP_HAWAII:
+		chip_name = "hawaii";
+		break;
+	case CHIP_KAVERI:
+	case CHIP_KABINI:
+		return 0;
+	default: BUG();
+	}
+
+	snprintf(fw_name, sizeof(fw_name), "radeon/%s_mc.bin", chip_name);
+	err = request_firmware(&adev->mc.fw, fw_name, adev->dev);
+	if (err)
+		goto out;
+	err = amdgpu_ucode_validate(adev->mc.fw);
+
+out:
+	if (err) {
+		printk(KERN_ERR
+		       "cik_mc: Failed to load firmware \"%s\"\n",
+		       fw_name);
+		release_firmware(adev->mc.fw);
+		adev->mc.fw = NULL;
+	}
+	return err;
+}
+
+/**
+ * gmc_v7_0_mc_load_microcode - load MC ucode into the hw
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Load the GDDR MC ucode into the hw (CIK).
+ * Returns 0 on success, error on failure.
+ */
+static int gmc_v7_0_mc_load_microcode(struct amdgpu_device *adev)
+{
+	const struct mc_firmware_header_v1_0 *hdr;
+	const __le32 *fw_data = NULL;
+	const __le32 *io_mc_regs = NULL;
+	u32 running, blackout = 0;
+	int i, ucode_size, regs_size;
+
+	if (!adev->mc.fw)
+		return -EINVAL;
+
+	hdr = (const struct mc_firmware_header_v1_0 *)adev->mc.fw->data;
+	amdgpu_ucode_print_mc_hdr(&hdr->header);
+
+	adev->mc.fw_version = le32_to_cpu(hdr->header.ucode_version);
+	regs_size = le32_to_cpu(hdr->io_debug_size_bytes) / (4 * 2);
+	io_mc_regs = (const __le32 *)
+		(adev->mc.fw->data + le32_to_cpu(hdr->io_debug_array_offset_bytes));
+	ucode_size = le32_to_cpu(hdr->header.ucode_size_bytes) / 4;
+	fw_data = (const __le32 *)
+		(adev->mc.fw->data + le32_to_cpu(hdr->header.ucode_array_offset_bytes));
+
+	running = REG_GET_FIELD(RREG32(mmMC_SEQ_SUP_CNTL), MC_SEQ_SUP_CNTL, RUN);
+
+	if (running == 0) {
+		if (running) {
+			blackout = RREG32(mmMC_SHARED_BLACKOUT_CNTL);
+			WREG32(mmMC_SHARED_BLACKOUT_CNTL, blackout | 1);
+		}
+
+		/* reset the engine and set to writable */
+		WREG32(mmMC_SEQ_SUP_CNTL, 0x00000008);
+		WREG32(mmMC_SEQ_SUP_CNTL, 0x00000010);
+
+		/* load mc io regs */
+		for (i = 0; i < regs_size; i++) {
+			WREG32(mmMC_SEQ_IO_DEBUG_INDEX, le32_to_cpup(io_mc_regs++));
+			WREG32(mmMC_SEQ_IO_DEBUG_DATA, le32_to_cpup(io_mc_regs++));
+		}
+		/* load the MC ucode */
+		for (i = 0; i < ucode_size; i++)
+			WREG32(mmMC_SEQ_SUP_PGM, le32_to_cpup(fw_data++));
+
+		/* put the engine back into the active state */
+		WREG32(mmMC_SEQ_SUP_CNTL, 0x00000008);
+		WREG32(mmMC_SEQ_SUP_CNTL, 0x00000004);
+		WREG32(mmMC_SEQ_SUP_CNTL, 0x00000001);
+
+		/* wait for training to complete */
+		for (i = 0; i < adev->usec_timeout; i++) {
+			if (REG_GET_FIELD(RREG32(mmMC_SEQ_TRAIN_WAKEUP_CNTL),
+					  MC_SEQ_TRAIN_WAKEUP_CNTL, TRAIN_DONE_D0))
+				break;
+			udelay(1);
+		}
+		for (i = 0; i < adev->usec_timeout; i++) {
+			if (REG_GET_FIELD(RREG32(mmMC_SEQ_TRAIN_WAKEUP_CNTL),
+					  MC_SEQ_TRAIN_WAKEUP_CNTL, TRAIN_DONE_D1))
+				break;
+			udelay(1);
+		}
+
+		if (running)
+			WREG32(mmMC_SHARED_BLACKOUT_CNTL, blackout);
+	}
+
+	return 0;
+}
+
+static void gmc_v7_0_vram_gtt_location(struct amdgpu_device *adev,
+				       struct amdgpu_mc *mc)
+{
+	if (mc->mc_vram_size > 0xFFC0000000ULL) {
+		/* leave room for at least 1024M GTT */
+		dev_warn(adev->dev, "limiting VRAM\n");
+		mc->real_vram_size = 0xFFC0000000ULL;
+		mc->mc_vram_size = 0xFFC0000000ULL;
+	}
+	amdgpu_vram_location(adev, &adev->mc, 0);
+	adev->mc.gtt_base_align = 0;
+	amdgpu_gtt_location(adev, mc);
+}
+
+/**
+ * gmc_v7_0_mc_program - program the GPU memory controller
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Set the location of vram, gart, and AGP in the GPU's
+ * physical address space (CIK).
+ */
+static void gmc_v7_0_mc_program(struct amdgpu_device *adev)
+{
+	struct amdgpu_mode_mc_save save;
+	u32 tmp;
+	int i, j;
+
+	/* Initialize HDP */
+	for (i = 0, j = 0; i < 32; i++, j += 0x6) {
+		WREG32((0xb05 + j), 0x00000000);
+		WREG32((0xb06 + j), 0x00000000);
+		WREG32((0xb07 + j), 0x00000000);
+		WREG32((0xb08 + j), 0x00000000);
+		WREG32((0xb09 + j), 0x00000000);
+	}
+	WREG32(mmHDP_REG_COHERENCY_FLUSH_CNTL, 0);
+
+	if (adev->mode_info.num_crtc)
+		amdgpu_display_set_vga_render_state(adev, false);
+
+	gmc_v7_0_mc_stop(adev, &save);
+	if (amdgpu_asic_wait_for_mc_idle(adev)) {
+		dev_warn(adev->dev, "Wait for MC idle timedout !\n");
+	}
+	/* Update configuration */
+	WREG32(mmMC_VM_SYSTEM_APERTURE_LOW_ADDR,
+	       adev->mc.vram_start >> 12);
+	WREG32(mmMC_VM_SYSTEM_APERTURE_HIGH_ADDR,
+	       adev->mc.vram_end >> 12);
+	WREG32(mmMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR,
+	       adev->vram_scratch.gpu_addr >> 12);
+	tmp = ((adev->mc.vram_end >> 24) & 0xFFFF) << 16;
+	tmp |= ((adev->mc.vram_start >> 24) & 0xFFFF);
+	WREG32(mmMC_VM_FB_LOCATION, tmp);
+	/* XXX double check these! */
+	WREG32(mmHDP_NONSURFACE_BASE, (adev->mc.vram_start >> 8));
+	WREG32(mmHDP_NONSURFACE_INFO, (2 << 7) | (1 << 30));
+	WREG32(mmHDP_NONSURFACE_SIZE, 0x3FFFFFFF);
+	WREG32(mmMC_VM_AGP_BASE, 0);
+	WREG32(mmMC_VM_AGP_TOP, 0x0FFFFFFF);
+	WREG32(mmMC_VM_AGP_BOT, 0x0FFFFFFF);
+	if (amdgpu_asic_wait_for_mc_idle(adev)) {
+		dev_warn(adev->dev, "Wait for MC idle timedout !\n");
+	}
+	gmc_v7_0_mc_resume(adev, &save);
+
+	WREG32(mmBIF_FB_EN, BIF_FB_EN__FB_READ_EN_MASK | BIF_FB_EN__FB_WRITE_EN_MASK);
+
+	tmp = RREG32(mmHDP_MISC_CNTL);
+	tmp = REG_SET_FIELD(tmp, HDP_MISC_CNTL, FLUSH_INVALIDATE_CACHE, 1);
+	WREG32(mmHDP_MISC_CNTL, tmp);
+
+	tmp = RREG32(mmHDP_HOST_PATH_CNTL);
+	WREG32(mmHDP_HOST_PATH_CNTL, tmp);
+}
+
+/**
+ * gmc_v7_0_mc_init - initialize the memory controller driver params
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Look up the amount of vram, vram width, and decide how to place
+ * vram and gart within the GPU's physical address space (CIK).
+ * Returns 0 for success.
+ */
+static int gmc_v7_0_mc_init(struct amdgpu_device *adev)
+{
+	u32 tmp;
+	int chansize, numchan;
+
+	/* Get VRAM informations */
+	tmp = RREG32(mmMC_ARB_RAMCFG);
+	if (REG_GET_FIELD(tmp, MC_ARB_RAMCFG, CHANSIZE)) {
+		chansize = 64;
+	} else {
+		chansize = 32;
+	}
+	tmp = RREG32(mmMC_SHARED_CHMAP);
+	switch (REG_GET_FIELD(tmp, MC_SHARED_CHMAP, NOOFCHAN)) {
+	case 0:
+	default:
+		numchan = 1;
+		break;
+	case 1:
+		numchan = 2;
+		break;
+	case 2:
+		numchan = 4;
+		break;
+	case 3:
+		numchan = 8;
+		break;
+	case 4:
+		numchan = 3;
+		break;
+	case 5:
+		numchan = 6;
+		break;
+	case 6:
+		numchan = 10;
+		break;
+	case 7:
+		numchan = 12;
+		break;
+	case 8:
+		numchan = 16;
+		break;
+	}
+	adev->mc.vram_width = numchan * chansize;
+	/* Could aper size report 0 ? */
+	adev->mc.aper_base = pci_resource_start(adev->pdev, 0);
+	adev->mc.aper_size = pci_resource_len(adev->pdev, 0);
+	/* size in MB on si */
+	adev->mc.mc_vram_size = RREG32(mmCONFIG_MEMSIZE) * 1024ULL * 1024ULL;
+	adev->mc.real_vram_size = RREG32(mmCONFIG_MEMSIZE) * 1024ULL * 1024ULL;
+	adev->mc.visible_vram_size = adev->mc.aper_size;
+
+	/* unless the user had overridden it, set the gart
+	 * size equal to the 1024 or vram, whichever is larger.
+	 */
+	if (amdgpu_gart_size == -1)
+		adev->mc.gtt_size = max((1024ULL << 20), adev->mc.mc_vram_size);
+	else
+		adev->mc.gtt_size = (uint64_t)amdgpu_gart_size << 20;
+
+	gmc_v7_0_vram_gtt_location(adev, &adev->mc);
+
+	return 0;
+}
+
+/*
+ * GART
+ * VMID 0 is the physical GPU addresses as used by the kernel.
+ * VMIDs 1-15 are used for userspace clients and are handled
+ * by the amdgpu vm/hsa code.
+ */
+
+/**
+ * gmc_v7_0_gart_flush_gpu_tlb - gart tlb flush callback
+ *
+ * @adev: amdgpu_device pointer
+ * @vmid: vm instance to flush
+ *
+ * Flush the TLB for the requested page table (CIK).
+ */
+static void gmc_v7_0_gart_flush_gpu_tlb(struct amdgpu_device *adev,
+					uint32_t vmid)
+{
+	/* flush hdp cache */
+	WREG32(mmHDP_MEM_COHERENCY_FLUSH_CNTL, 0);
+
+	/* bits 0-15 are the VM contexts0-15 */
+	WREG32(mmVM_INVALIDATE_REQUEST, 1 << vmid);
+}
+
+/**
+ * gmc_v7_0_gart_set_pte_pde - update the page tables using MMIO
+ *
+ * @adev: amdgpu_device pointer
+ * @cpu_pt_addr: cpu address of the page table
+ * @gpu_page_idx: entry in the page table to update
+ * @addr: dst addr to write into pte/pde
+ * @flags: access flags
+ *
+ * Update the page tables using the CPU.
+ */
+static int gmc_v7_0_gart_set_pte_pde(struct amdgpu_device *adev,
+				     void *cpu_pt_addr,
+				     uint32_t gpu_page_idx,
+				     uint64_t addr,
+				     uint32_t flags)
+{
+	void __iomem *ptr = (void *)cpu_pt_addr;
+	uint64_t value;
+
+	value = addr & 0xFFFFFFFFFFFFF000ULL;
+	value |= flags;
+	writeq(value, ptr + (gpu_page_idx * 8));
+
+	return 0;
+}
+
+/**
+ * gmc_v7_0_gart_enable - gart enable
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * This sets up the TLBs, programs the page tables for VMID0,
+ * sets up the hw for VMIDs 1-15 which are allocated on
+ * demand, and sets up the global locations for the LDS, GDS,
+ * and GPUVM for FSA64 clients (CIK).
+ * Returns 0 for success, errors for failure.
+ */
+static int gmc_v7_0_gart_enable(struct amdgpu_device *adev)
+{
+	int r, i;
+	u32 tmp;
+
+	if (adev->gart.robj == NULL) {
+		dev_err(adev->dev, "No VRAM object for PCIE GART.\n");
+		return -EINVAL;
+	}
+	r = amdgpu_gart_table_vram_pin(adev);
+	if (r)
+		return r;
+	/* Setup TLB control */
+	tmp = RREG32(mmMC_VM_MX_L1_TLB_CNTL);
+	tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL, ENABLE_L1_TLB, 1);
+	tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL, ENABLE_L1_FRAGMENT_PROCESSING, 1);
+	tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL, SYSTEM_ACCESS_MODE, 3);
+	tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL, ENABLE_ADVANCED_DRIVER_MODEL, 1);
+	tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL, SYSTEM_APERTURE_UNMAPPED_ACCESS, 0);
+	WREG32(mmMC_VM_MX_L1_TLB_CNTL, tmp);
+	/* Setup L2 cache */
+	tmp = RREG32(mmVM_L2_CNTL);
+	tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, ENABLE_L2_CACHE, 1);
+	tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, ENABLE_L2_FRAGMENT_PROCESSING, 1);
+	tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, ENABLE_L2_PTE_CACHE_LRU_UPDATE_BY_WRITE, 1);
+	tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, ENABLE_L2_PDE0_CACHE_LRU_UPDATE_BY_WRITE, 1);
+	tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, EFFECTIVE_L2_QUEUE_SIZE, 7);
+	tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, CONTEXT1_IDENTITY_ACCESS_MODE, 1);
+	WREG32(mmVM_L2_CNTL, tmp);
+	tmp = REG_SET_FIELD(0, VM_L2_CNTL2, INVALIDATE_ALL_L1_TLBS, 1);
+	tmp = REG_SET_FIELD(tmp, VM_L2_CNTL2, INVALIDATE_L2_CACHE, 1);
+	WREG32(mmVM_L2_CNTL2, tmp);
+	tmp = RREG32(mmVM_L2_CNTL3);
+	tmp = REG_SET_FIELD(tmp, VM_L2_CNTL3, L2_CACHE_BIGK_ASSOCIATIVITY, 1);
+	tmp = REG_SET_FIELD(tmp, VM_L2_CNTL3, BANK_SELECT, 4);
+	tmp = REG_SET_FIELD(tmp, VM_L2_CNTL3, L2_CACHE_BIGK_FRAGMENT_SIZE, 4);
+	WREG32(mmVM_L2_CNTL3, tmp);
+	/* setup context0 */
+	WREG32(mmVM_CONTEXT0_PAGE_TABLE_START_ADDR, adev->mc.gtt_start >> 12);
+	WREG32(mmVM_CONTEXT0_PAGE_TABLE_END_ADDR, adev->mc.gtt_end >> 12);
+	WREG32(mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR, adev->gart.table_addr >> 12);
+	WREG32(mmVM_CONTEXT0_PROTECTION_FAULT_DEFAULT_ADDR,
+			(u32)(adev->dummy_page.addr >> 12));
+	WREG32(mmVM_CONTEXT0_CNTL2, 0);
+	tmp = RREG32(mmVM_CONTEXT0_CNTL);
+	tmp = REG_SET_FIELD(tmp, VM_CONTEXT0_CNTL, ENABLE_CONTEXT, 1);
+	tmp = REG_SET_FIELD(tmp, VM_CONTEXT0_CNTL, PAGE_TABLE_DEPTH, 0);
+	tmp = REG_SET_FIELD(tmp, VM_CONTEXT0_CNTL, RANGE_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+	WREG32(mmVM_CONTEXT0_CNTL, tmp);
+
+	WREG32(0x575, 0);
+	WREG32(0x576, 0);
+	WREG32(0x577, 0);
+
+	/* empty context1-15 */
+	/* FIXME start with 4G, once using 2 level pt switch to full
+	 * vm size space
+	 */
+	/* set vm size, must be a multiple of 4 */
+	WREG32(mmVM_CONTEXT1_PAGE_TABLE_START_ADDR, 0);
+	WREG32(mmVM_CONTEXT1_PAGE_TABLE_END_ADDR, adev->vm_manager.max_pfn);
+	for (i = 1; i < 16; i++) {
+		if (i < 8)
+			WREG32(mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR + i,
+			       adev->gart.table_addr >> 12);
+		else
+			WREG32(mmVM_CONTEXT8_PAGE_TABLE_BASE_ADDR + i - 8,
+			       adev->gart.table_addr >> 12);
+	}
+
+	/* enable context1-15 */
+	WREG32(mmVM_CONTEXT1_PROTECTION_FAULT_DEFAULT_ADDR,
+	       (u32)(adev->dummy_page.addr >> 12));
+	WREG32(mmVM_CONTEXT1_CNTL2, 4);
+	tmp = RREG32(mmVM_CONTEXT1_CNTL);
+	tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, ENABLE_CONTEXT, 1);
+	tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, PAGE_TABLE_DEPTH, 1);
+	tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, RANGE_PROTECTION_FAULT_ENABLE_INTERRUPT, 1);
+	tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, RANGE_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+	tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, DUMMY_PAGE_PROTECTION_FAULT_ENABLE_INTERRUPT, 1);
+	tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, DUMMY_PAGE_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+	tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, PDE0_PROTECTION_FAULT_ENABLE_INTERRUPT, 1);
+	tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, PDE0_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+	tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, VALID_PROTECTION_FAULT_ENABLE_INTERRUPT, 1);
+	tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, VALID_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+	tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, READ_PROTECTION_FAULT_ENABLE_INTERRUPT, 1);
+	tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, READ_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+	tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, WRITE_PROTECTION_FAULT_ENABLE_INTERRUPT, 1);
+	tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, WRITE_PROTECTION_FAULT_ENABLE_DEFAULT, 1);
+	tmp = REG_SET_FIELD(tmp, VM_CONTEXT1_CNTL, PAGE_TABLE_BLOCK_SIZE,
+			    amdgpu_vm_block_size - 9);
+	WREG32(mmVM_CONTEXT1_CNTL, tmp);
+
+	if (adev->asic_type == CHIP_KAVERI) {
+		tmp = RREG32(mmCHUB_CONTROL);
+		tmp &= ~BYPASS_VM;
+		WREG32(mmCHUB_CONTROL, tmp);
+	}
+
+	gmc_v7_0_gart_flush_gpu_tlb(adev, 0);
+	DRM_INFO("PCIE GART of %uM enabled (table at 0x%016llX).\n",
+		 (unsigned)(adev->mc.gtt_size >> 20),
+		 (unsigned long long)adev->gart.table_addr);
+	adev->gart.ready = true;
+	return 0;
+}
+
+static int gmc_v7_0_gart_init(struct amdgpu_device *adev)
+{
+	int r;
+
+	if (adev->gart.robj) {
+		WARN(1, "R600 PCIE GART already initialized\n");
+		return 0;
+	}
+	/* Initialize common gart structure */
+	r = amdgpu_gart_init(adev);
+	if (r)
+		return r;
+	adev->gart.table_size = adev->gart.num_gpu_pages * 8;
+	return amdgpu_gart_table_vram_alloc(adev);
+}
+
+/**
+ * gmc_v7_0_gart_disable - gart disable
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * This disables all VM page table (CIK).
+ */
+static void gmc_v7_0_gart_disable(struct amdgpu_device *adev)
+{
+	u32 tmp;
+
+	/* Disable all tables */
+	WREG32(mmVM_CONTEXT0_CNTL, 0);
+	WREG32(mmVM_CONTEXT1_CNTL, 0);
+	/* Setup TLB control */
+	tmp = RREG32(mmMC_VM_MX_L1_TLB_CNTL);
+	tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL, ENABLE_L1_TLB, 0);
+	tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL, ENABLE_L1_FRAGMENT_PROCESSING, 0);
+	tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL, ENABLE_ADVANCED_DRIVER_MODEL, 0);
+	WREG32(mmMC_VM_MX_L1_TLB_CNTL, tmp);
+	/* Setup L2 cache */
+	tmp = RREG32(mmVM_L2_CNTL);
+	tmp = REG_SET_FIELD(tmp, VM_L2_CNTL, ENABLE_L2_CACHE, 0);
+	WREG32(mmVM_L2_CNTL, tmp);
+	WREG32(mmVM_L2_CNTL2, 0);
+	amdgpu_gart_table_vram_unpin(adev);
+}
+
+/**
+ * gmc_v7_0_gart_fini - vm fini callback
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Tears down the driver GART/VM setup (CIK).
+ */
+static void gmc_v7_0_gart_fini(struct amdgpu_device *adev)
+{
+	amdgpu_gart_table_vram_free(adev);
+	amdgpu_gart_fini(adev);
+}
+
+/*
+ * vm
+ * VMID 0 is the physical GPU addresses as used by the kernel.
+ * VMIDs 1-15 are used for userspace clients and are handled
+ * by the amdgpu vm/hsa code.
+ */
+/**
+ * gmc_v7_0_vm_init - cik vm init callback
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Inits cik specific vm parameters (number of VMs, base of vram for
+ * VMIDs 1-15) (CIK).
+ * Returns 0 for success.
+ */
+static int gmc_v7_0_vm_init(struct amdgpu_device *adev)
+{
+	/*
+	 * number of VMs
+	 * VMID 0 is reserved for System
+	 * amdgpu graphics/compute will use VMIDs 1-7
+	 * amdkfd will use VMIDs 8-15
+	 */
+	adev->vm_manager.nvm = AMDGPU_NUM_OF_VMIDS;
+
+	/* base offset of vram pages */
+	if (adev->flags & AMDGPU_IS_APU) {
+		u64 tmp = RREG32(mmMC_VM_FB_OFFSET);
+		tmp <<= 22;
+		adev->vm_manager.vram_base_offset = tmp;
+	} else
+		adev->vm_manager.vram_base_offset = 0;
+
+	return 0;
+}
+
+/**
+ * gmc_v7_0_vm_fini - cik vm fini callback
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Tear down any asic specific VM setup (CIK).
+ */
+static void gmc_v7_0_vm_fini(struct amdgpu_device *adev)
+{
+}
+
+/**
+ * gmc_v7_0_vm_decode_fault - print human readable fault info
+ *
+ * @adev: amdgpu_device pointer
+ * @status: VM_CONTEXT1_PROTECTION_FAULT_STATUS register value
+ * @addr: VM_CONTEXT1_PROTECTION_FAULT_ADDR register value
+ *
+ * Print human readable fault information (CIK).
+ */
+static void gmc_v7_0_vm_decode_fault(struct amdgpu_device *adev,
+				     u32 status, u32 addr, u32 mc_client)
+{
+	u32 mc_id;
+	u32 vmid = REG_GET_FIELD(status, VM_CONTEXT1_PROTECTION_FAULT_STATUS, VMID);
+	u32 protections = REG_GET_FIELD(status, VM_CONTEXT1_PROTECTION_FAULT_STATUS,
+					PROTECTIONS);
+	char block[5] = { mc_client >> 24, (mc_client >> 16) & 0xff,
+		(mc_client >> 8) & 0xff, mc_client & 0xff, 0 };
+
+	mc_id = REG_GET_FIELD(status, VM_CONTEXT1_PROTECTION_FAULT_STATUS,
+			      MEMORY_CLIENT_ID);
+
+	printk("VM fault (0x%02x, vmid %d) at page %u, %s from '%s' (0x%08x) (%d)\n",
+	       protections, vmid, addr,
+	       REG_GET_FIELD(status, VM_CONTEXT1_PROTECTION_FAULT_STATUS,
+			     MEMORY_CLIENT_RW) ?
+	       "write" : "read", block, mc_client, mc_id);
+}
+
+
+static const u32 mc_cg_registers[] = {
+	mmMC_HUB_MISC_HUB_CG,
+	mmMC_HUB_MISC_SIP_CG,
+	mmMC_HUB_MISC_VM_CG,
+	mmMC_XPB_CLK_GAT,
+	mmATC_MISC_CG,
+	mmMC_CITF_MISC_WR_CG,
+	mmMC_CITF_MISC_RD_CG,
+	mmMC_CITF_MISC_VM_CG,
+	mmVM_L2_CG,
+};
+
+static const u32 mc_cg_ls_en[] = {
+	MC_HUB_MISC_HUB_CG__MEM_LS_ENABLE_MASK,
+	MC_HUB_MISC_SIP_CG__MEM_LS_ENABLE_MASK,
+	MC_HUB_MISC_VM_CG__MEM_LS_ENABLE_MASK,
+	MC_XPB_CLK_GAT__MEM_LS_ENABLE_MASK,
+	ATC_MISC_CG__MEM_LS_ENABLE_MASK,
+	MC_CITF_MISC_WR_CG__MEM_LS_ENABLE_MASK,
+	MC_CITF_MISC_RD_CG__MEM_LS_ENABLE_MASK,
+	MC_CITF_MISC_VM_CG__MEM_LS_ENABLE_MASK,
+	VM_L2_CG__MEM_LS_ENABLE_MASK,
+};
+
+static const u32 mc_cg_en[] = {
+	MC_HUB_MISC_HUB_CG__ENABLE_MASK,
+	MC_HUB_MISC_SIP_CG__ENABLE_MASK,
+	MC_HUB_MISC_VM_CG__ENABLE_MASK,
+	MC_XPB_CLK_GAT__ENABLE_MASK,
+	ATC_MISC_CG__ENABLE_MASK,
+	MC_CITF_MISC_WR_CG__ENABLE_MASK,
+	MC_CITF_MISC_RD_CG__ENABLE_MASK,
+	MC_CITF_MISC_VM_CG__ENABLE_MASK,
+	VM_L2_CG__ENABLE_MASK,
+};
+
+static void gmc_v7_0_enable_mc_ls(struct amdgpu_device *adev,
+				  bool enable)
+{
+	int i;
+	u32 orig, data;
+
+	for (i = 0; i < ARRAY_SIZE(mc_cg_registers); i++) {
+		orig = data = RREG32(mc_cg_registers[i]);
+		if (enable && (adev->cg_flags & AMDGPU_CG_SUPPORT_MC_LS))
+			data |= mc_cg_ls_en[i];
+		else
+			data &= ~mc_cg_ls_en[i];
+		if (data != orig)
+			WREG32(mc_cg_registers[i], data);
+	}
+}
+
+static void gmc_v7_0_enable_mc_mgcg(struct amdgpu_device *adev,
+				    bool enable)
+{
+	int i;
+	u32 orig, data;
+
+	for (i = 0; i < ARRAY_SIZE(mc_cg_registers); i++) {
+		orig = data = RREG32(mc_cg_registers[i]);
+		if (enable && (adev->cg_flags & AMDGPU_CG_SUPPORT_MC_MGCG))
+			data |= mc_cg_en[i];
+		else
+			data &= ~mc_cg_en[i];
+		if (data != orig)
+			WREG32(mc_cg_registers[i], data);
+	}
+}
+
+static void gmc_v7_0_enable_bif_mgls(struct amdgpu_device *adev,
+				     bool enable)
+{
+	u32 orig, data;
+
+	orig = data = RREG32_PCIE(ixPCIE_CNTL2);
+
+	if (enable && (adev->cg_flags & AMDGPU_CG_SUPPORT_BIF_LS)) {
+		data = REG_SET_FIELD(data, PCIE_CNTL2, SLV_MEM_LS_EN, 1);
+		data = REG_SET_FIELD(data, PCIE_CNTL2, MST_MEM_LS_EN, 1);
+		data = REG_SET_FIELD(data, PCIE_CNTL2, REPLAY_MEM_LS_EN, 1);
+		data = REG_SET_FIELD(data, PCIE_CNTL2, SLV_MEM_AGGRESSIVE_LS_EN, 1);
+	} else {
+		data = REG_SET_FIELD(data, PCIE_CNTL2, SLV_MEM_LS_EN, 0);
+		data = REG_SET_FIELD(data, PCIE_CNTL2, MST_MEM_LS_EN, 0);
+		data = REG_SET_FIELD(data, PCIE_CNTL2, REPLAY_MEM_LS_EN, 0);
+		data = REG_SET_FIELD(data, PCIE_CNTL2, SLV_MEM_AGGRESSIVE_LS_EN, 0);
+	}
+
+	if (orig != data)
+		WREG32_PCIE(ixPCIE_CNTL2, data);
+}
+
+static void gmc_v7_0_enable_hdp_mgcg(struct amdgpu_device *adev,
+				     bool enable)
+{
+	u32 orig, data;
+
+	orig = data = RREG32(mmHDP_HOST_PATH_CNTL);
+
+	if (enable && (adev->cg_flags & AMDGPU_CG_SUPPORT_HDP_MGCG))
+		data = REG_SET_FIELD(data, HDP_HOST_PATH_CNTL, CLOCK_GATING_DIS, 0);
+	else
+		data = REG_SET_FIELD(data, HDP_HOST_PATH_CNTL, CLOCK_GATING_DIS, 1);
+
+	if (orig != data)
+		WREG32(mmHDP_HOST_PATH_CNTL, data);
+}
+
+static void gmc_v7_0_enable_hdp_ls(struct amdgpu_device *adev,
+				   bool enable)
+{
+	u32 orig, data;
+
+	orig = data = RREG32(mmHDP_MEM_POWER_LS);
+
+	if (enable && (adev->cg_flags & AMDGPU_CG_SUPPORT_HDP_LS))
+		data = REG_SET_FIELD(data, HDP_MEM_POWER_LS, LS_ENABLE, 1);
+	else
+		data = REG_SET_FIELD(data, HDP_MEM_POWER_LS, LS_ENABLE, 0);
+
+	if (orig != data)
+		WREG32(mmHDP_MEM_POWER_LS, data);
+}
+
+static int gmc_v7_0_early_init(struct amdgpu_device *adev)
+{
+	gmc_v7_0_set_gart_funcs(adev);
+	gmc_v7_0_set_irq_funcs(adev);
+
+	if (adev->flags & AMDGPU_IS_APU) {
+		adev->mc.is_gddr5 = false;
+	} else {
+		u32 tmp = RREG32(mmMC_SEQ_MISC0);
+
+		if (((tmp & MC_SEQ_MISC0__GDDR5_MASK) >>
+		     MC_SEQ_MISC0__GDDR5__SHIFT) == MC_SEQ_MISC0__GDDR5_VALUE)
+			adev->mc.is_gddr5 = true;
+		else
+			adev->mc.is_gddr5 = false;
+	}
+
+	return 0;
+}
+
+static int gmc_v7_0_sw_init(struct amdgpu_device *adev)
+{
+	int r;
+	int dma_bits;
+
+	r = amdgpu_gem_init(adev);
+	if (r)
+		return r;
+
+	r = amdgpu_irq_add_id(adev, 146, &adev->mc.vm_fault);
+	if (r)
+		return r;
+
+	r = amdgpu_irq_add_id(adev, 147, &adev->mc.vm_fault);
+	if (r)
+		return r;
+
+	/* Adjust VM size here.
+	 * Currently set to 4GB ((1 << 20) 4k pages).
+	 * Max GPUVM size for cayman and SI is 40 bits.
+	 */
+	adev->vm_manager.max_pfn = amdgpu_vm_size << 18;
+
+	/* Set the internal MC address mask
+	 * This is the max address of the GPU's
+	 * internal address space.
+	 */
+	adev->mc.mc_mask = 0xffffffffffULL; /* 40 bit MC */
+
+	/* set DMA mask + need_dma32 flags.
+	 * PCIE - can handle 40-bits.
+	 * IGP - can handle 40-bits
+	 * PCI - dma32 for legacy pci gart, 40 bits on newer asics
+	 */
+	adev->need_dma32 = false;
+	dma_bits = adev->need_dma32 ? 32 : 40;
+	r = pci_set_dma_mask(adev->pdev, DMA_BIT_MASK(dma_bits));
+	if (r) {
+		adev->need_dma32 = true;
+		dma_bits = 32;
+		printk(KERN_WARNING "amdgpu: No suitable DMA available.\n");
+	}
+	r = pci_set_consistent_dma_mask(adev->pdev, DMA_BIT_MASK(dma_bits));
+	if (r) {
+		pci_set_consistent_dma_mask(adev->pdev, DMA_BIT_MASK(32));
+		printk(KERN_WARNING "amdgpu: No coherent DMA available.\n");
+	}
+
+	r = gmc_v7_0_init_microcode(adev);
+	if (r) {
+		DRM_ERROR("Failed to load mc firmware!\n");
+		return r;
+	}
+
+	r = gmc_v7_0_mc_init(adev);
+	if (r)
+		return r;
+
+	/* Memory manager */
+	r = amdgpu_bo_init(adev);
+	if (r)
+		return r;
+
+	r = gmc_v7_0_gart_init(adev);
+	if (r)
+		return r;
+
+	if (!adev->vm_manager.enabled) {
+		r = gmc_v7_0_vm_init(adev);
+		if (r) {
+			dev_err(adev->dev, "vm manager initialization failed (%d).\n", r);
+			return r;
+		}
+		adev->vm_manager.enabled = true;
+	}
+
+	return r;
+}
+
+static int gmc_v7_0_sw_fini(struct amdgpu_device *adev)
+{
+	int i;
+
+	if (adev->vm_manager.enabled) {
+		for (i = 0; i < AMDGPU_NUM_VM; ++i)
+			amdgpu_fence_unref(&adev->vm_manager.active[i]);
+		gmc_v7_0_vm_fini(adev);
+		adev->vm_manager.enabled = false;
+	}
+	gmc_v7_0_gart_fini(adev);
+	amdgpu_gem_fini(adev);
+	amdgpu_bo_fini(adev);
+
+	return 0;
+}
+
+static int gmc_v7_0_hw_init(struct amdgpu_device *adev)
+{
+	int r;
+
+	gmc_v7_0_mc_program(adev);
+
+	if (!(adev->flags & AMDGPU_IS_APU)) {
+		r = gmc_v7_0_mc_load_microcode(adev);
+		if (r) {
+			DRM_ERROR("Failed to load MC firmware!\n");
+			return r;
+		}
+	}
+
+	r = gmc_v7_0_gart_enable(adev);
+	if (r)
+		return r;
+
+	return r;
+}
+
+static int gmc_v7_0_hw_fini(struct amdgpu_device *adev)
+{
+	gmc_v7_0_gart_disable(adev);
+
+	return 0;
+}
+
+static int gmc_v7_0_suspend(struct amdgpu_device *adev)
+{
+	int i;
+
+	if (adev->vm_manager.enabled) {
+		for (i = 0; i < AMDGPU_NUM_VM; ++i)
+			amdgpu_fence_unref(&adev->vm_manager.active[i]);
+		gmc_v7_0_vm_fini(adev);
+		adev->vm_manager.enabled = false;
+	}
+	gmc_v7_0_hw_fini(adev);
+
+	return 0;
+}
+
+static int gmc_v7_0_resume(struct amdgpu_device *adev)
+{
+	int r;
+
+	r = gmc_v7_0_hw_init(adev);
+	if (r)
+		return r;
+
+	if (!adev->vm_manager.enabled) {
+		r = gmc_v7_0_vm_init(adev);
+		if (r) {
+			dev_err(adev->dev, "vm manager initialization failed (%d).\n", r);
+			return r;
+		}
+		adev->vm_manager.enabled = true;
+	}
+
+	return r;
+}
+
+static bool gmc_v7_0_is_idle(struct amdgpu_device *adev)
+{
+	u32 tmp = RREG32(mmSRBM_STATUS);
+
+	if (tmp & (SRBM_STATUS__MCB_BUSY_MASK | SRBM_STATUS__MCB_NON_DISPLAY_BUSY_MASK |
+		   SRBM_STATUS__MCC_BUSY_MASK | SRBM_STATUS__MCD_BUSY_MASK | SRBM_STATUS__VMC_BUSY_MASK))
+		return false;
+
+	return true;
+}
+
+static int gmc_v7_0_wait_for_idle(struct amdgpu_device *adev)
+{
+	unsigned i;
+	u32 tmp;
+
+	for (i = 0; i < adev->usec_timeout; i++) {
+		/* read MC_STATUS */
+		tmp = RREG32(mmSRBM_STATUS) & (SRBM_STATUS__MCB_BUSY_MASK |
+					       SRBM_STATUS__MCB_NON_DISPLAY_BUSY_MASK |
+					       SRBM_STATUS__MCC_BUSY_MASK |
+					       SRBM_STATUS__MCD_BUSY_MASK |
+					       SRBM_STATUS__VMC_BUSY_MASK);
+		if (!tmp)
+			return 0;
+		udelay(1);
+	}
+	return -ETIMEDOUT;
+
+}
+
+static void gmc_v7_0_print_status(struct amdgpu_device *adev)
+{
+	int i, j;
+
+	dev_info(adev->dev, "GMC 8.x registers\n");
+	dev_info(adev->dev, "  SRBM_STATUS=0x%08X\n",
+		RREG32(mmSRBM_STATUS));
+	dev_info(adev->dev, "  SRBM_STATUS2=0x%08X\n",
+		RREG32(mmSRBM_STATUS2));
+
+	dev_info(adev->dev, "  VM_CONTEXT1_PROTECTION_FAULT_ADDR   0x%08X\n",
+		 RREG32(mmVM_CONTEXT1_PROTECTION_FAULT_ADDR));
+	dev_info(adev->dev, "  VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x%08X\n",
+		 RREG32(mmVM_CONTEXT1_PROTECTION_FAULT_STATUS));
+	dev_info(adev->dev, "  MC_VM_MX_L1_TLB_CNTL=0x%08X\n",
+		 RREG32(mmMC_VM_MX_L1_TLB_CNTL));
+	dev_info(adev->dev, "  VM_L2_CNTL=0x%08X\n",
+		 RREG32(mmVM_L2_CNTL));
+	dev_info(adev->dev, "  VM_L2_CNTL2=0x%08X\n",
+		 RREG32(mmVM_L2_CNTL2));
+	dev_info(adev->dev, "  VM_L2_CNTL3=0x%08X\n",
+		 RREG32(mmVM_L2_CNTL3));
+	dev_info(adev->dev, "  VM_CONTEXT0_PAGE_TABLE_START_ADDR=0x%08X\n",
+		 RREG32(mmVM_CONTEXT0_PAGE_TABLE_START_ADDR));
+	dev_info(adev->dev, "  VM_CONTEXT0_PAGE_TABLE_END_ADDR=0x%08X\n",
+		 RREG32(mmVM_CONTEXT0_PAGE_TABLE_END_ADDR));
+	dev_info(adev->dev, "  VM_CONTEXT0_PROTECTION_FAULT_DEFAULT_ADDR=0x%08X\n",
+		 RREG32(mmVM_CONTEXT0_PROTECTION_FAULT_DEFAULT_ADDR));
+	dev_info(adev->dev, "  VM_CONTEXT0_CNTL2=0x%08X\n",
+		 RREG32(mmVM_CONTEXT0_CNTL2));
+	dev_info(adev->dev, "  VM_CONTEXT0_CNTL=0x%08X\n",
+		 RREG32(mmVM_CONTEXT0_CNTL));
+	dev_info(adev->dev, "  0x15D4=0x%08X\n",
+		 RREG32(0x575));
+	dev_info(adev->dev, "  0x15D8=0x%08X\n",
+		 RREG32(0x576));
+	dev_info(adev->dev, "  0x15DC=0x%08X\n",
+		 RREG32(0x577));
+	dev_info(adev->dev, "  VM_CONTEXT1_PAGE_TABLE_START_ADDR=0x%08X\n",
+		 RREG32(mmVM_CONTEXT1_PAGE_TABLE_START_ADDR));
+	dev_info(adev->dev, "  VM_CONTEXT1_PAGE_TABLE_END_ADDR=0x%08X\n",
+		 RREG32(mmVM_CONTEXT1_PAGE_TABLE_END_ADDR));
+	dev_info(adev->dev, "  VM_CONTEXT1_PROTECTION_FAULT_DEFAULT_ADDR=0x%08X\n",
+		 RREG32(mmVM_CONTEXT1_PROTECTION_FAULT_DEFAULT_ADDR));
+	dev_info(adev->dev, "  VM_CONTEXT1_CNTL2=0x%08X\n",
+		 RREG32(mmVM_CONTEXT1_CNTL2));
+	dev_info(adev->dev, "  VM_CONTEXT1_CNTL=0x%08X\n",
+		 RREG32(mmVM_CONTEXT1_CNTL));
+	for (i = 0; i < 16; i++) {
+		if (i < 8)
+			dev_info(adev->dev, "  VM_CONTEXT%d_PAGE_TABLE_BASE_ADDR=0x%08X\n",
+				 i, RREG32(mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR + i));
+		else
+			dev_info(adev->dev, "  VM_CONTEXT%d_PAGE_TABLE_BASE_ADDR=0x%08X\n",
+				 i, RREG32(mmVM_CONTEXT8_PAGE_TABLE_BASE_ADDR + i - 8));
+	}
+	dev_info(adev->dev, "  MC_VM_SYSTEM_APERTURE_LOW_ADDR=0x%08X\n",
+		 RREG32(mmMC_VM_SYSTEM_APERTURE_LOW_ADDR));
+	dev_info(adev->dev, "  MC_VM_SYSTEM_APERTURE_HIGH_ADDR=0x%08X\n",
+		 RREG32(mmMC_VM_SYSTEM_APERTURE_HIGH_ADDR));
+	dev_info(adev->dev, "  MC_VM_SYSTEM_APERTURE_DEFAULT_ADDR=0x%08X\n",
+		 RREG32(mmMC_VM_SYSTEM_APERTURE_DEFAULT_ADDR));
+	dev_info(adev->dev, "  MC_VM_FB_LOCATION=0x%08X\n",
+		 RREG32(mmMC_VM_FB_LOCATION));
+	dev_info(adev->dev, "  MC_VM_AGP_BASE=0x%08X\n",
+		 RREG32(mmMC_VM_AGP_BASE));
+	dev_info(adev->dev, "  MC_VM_AGP_TOP=0x%08X\n",
+		 RREG32(mmMC_VM_AGP_TOP));
+	dev_info(adev->dev, "  MC_VM_AGP_BOT=0x%08X\n",
+		 RREG32(mmMC_VM_AGP_BOT));
+
+	if (adev->asic_type == CHIP_KAVERI) {
+		dev_info(adev->dev, "  CHUB_CONTROL=0x%08X\n",
+			 RREG32(mmCHUB_CONTROL));
+	}
+
+	dev_info(adev->dev, "  HDP_REG_COHERENCY_FLUSH_CNTL=0x%08X\n",
+		 RREG32(mmHDP_REG_COHERENCY_FLUSH_CNTL));
+	dev_info(adev->dev, "  HDP_NONSURFACE_BASE=0x%08X\n",
+		 RREG32(mmHDP_NONSURFACE_BASE));
+	dev_info(adev->dev, "  HDP_NONSURFACE_INFO=0x%08X\n",
+		 RREG32(mmHDP_NONSURFACE_INFO));
+	dev_info(adev->dev, "  HDP_NONSURFACE_SIZE=0x%08X\n",
+		 RREG32(mmHDP_NONSURFACE_SIZE));
+	dev_info(adev->dev, "  HDP_MISC_CNTL=0x%08X\n",
+		 RREG32(mmHDP_MISC_CNTL));
+	dev_info(adev->dev, "  HDP_HOST_PATH_CNTL=0x%08X\n",
+		 RREG32(mmHDP_HOST_PATH_CNTL));
+
+	for (i = 0, j = 0; i < 32; i++, j += 0x6) {
+		dev_info(adev->dev, "  %d:\n", i);
+		dev_info(adev->dev, "  0x%04X=0x%08X\n",
+			 0xb05 + j, RREG32(0xb05 + j));
+		dev_info(adev->dev, "  0x%04X=0x%08X\n",
+			 0xb06 + j, RREG32(0xb06 + j));
+		dev_info(adev->dev, "  0x%04X=0x%08X\n",
+			 0xb07 + j, RREG32(0xb07 + j));
+		dev_info(adev->dev, "  0x%04X=0x%08X\n",
+			 0xb08 + j, RREG32(0xb08 + j));
+		dev_info(adev->dev, "  0x%04X=0x%08X\n",
+			 0xb09 + j, RREG32(0xb09 + j));
+	}
+
+	dev_info(adev->dev, "  BIF_FB_EN=0x%08X\n",
+		 RREG32(mmBIF_FB_EN));
+}
+
+static int gmc_v7_0_soft_reset(struct amdgpu_device *adev)
+{
+	struct amdgpu_mode_mc_save save;
+	u32 srbm_soft_reset = 0;
+	u32 tmp = RREG32(mmSRBM_STATUS);
+
+	if (tmp & SRBM_STATUS__VMC_BUSY_MASK)
+		srbm_soft_reset = REG_SET_FIELD(srbm_soft_reset,
+						SRBM_SOFT_RESET, SOFT_RESET_VMC, 1);
+
+	if (tmp & (SRBM_STATUS__MCB_BUSY_MASK | SRBM_STATUS__MCB_NON_DISPLAY_BUSY_MASK |
+		   SRBM_STATUS__MCC_BUSY_MASK | SRBM_STATUS__MCD_BUSY_MASK)) {
+		if (!(adev->flags & AMDGPU_IS_APU))
+			srbm_soft_reset = REG_SET_FIELD(srbm_soft_reset,
+							SRBM_SOFT_RESET, SOFT_RESET_MC, 1);
+	}
+
+	if (srbm_soft_reset) {
+		gmc_v7_0_print_status(adev);
+
+		gmc_v7_0_mc_stop(adev, &save);
+		if (gmc_v7_0_wait_for_idle(adev)) {
+			dev_warn(adev->dev, "Wait for GMC idle timed out !\n");
+		}
+
+
+		tmp = RREG32(mmSRBM_SOFT_RESET);
+		tmp |= srbm_soft_reset;
+		dev_info(adev->dev, "SRBM_SOFT_RESET=0x%08X\n", tmp);
+		WREG32(mmSRBM_SOFT_RESET, tmp);
+		tmp = RREG32(mmSRBM_SOFT_RESET);
+
+		udelay(50);
+
+		tmp &= ~srbm_soft_reset;
+		WREG32(mmSRBM_SOFT_RESET, tmp);
+		tmp = RREG32(mmSRBM_SOFT_RESET);
+
+		/* Wait a little for things to settle down */
+		udelay(50);
+
+		gmc_v7_0_mc_resume(adev, &save);
+		udelay(50);
+
+		gmc_v7_0_print_status(adev);
+	}
+
+	return 0;
+}
+
+static int gmc_v7_0_vm_fault_interrupt_state(struct amdgpu_device *adev,
+					     struct amdgpu_irq_src *src,
+					     unsigned type,
+					     enum amdgpu_interrupt_state state)
+{
+	u32 tmp;
+	u32 bits = (VM_CONTEXT1_CNTL__RANGE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK |
+		    VM_CONTEXT1_CNTL__DUMMY_PAGE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK |
+		    VM_CONTEXT1_CNTL__PDE0_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK |
+		    VM_CONTEXT1_CNTL__VALID_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK |
+		    VM_CONTEXT1_CNTL__READ_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK |
+		    VM_CONTEXT1_CNTL__WRITE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK);
+
+	switch (state) {
+	case AMDGPU_IRQ_STATE_DISABLE:
+		/* system context */
+		tmp = RREG32(mmVM_CONTEXT0_CNTL);
+		tmp &= ~bits;
+		WREG32(mmVM_CONTEXT0_CNTL, tmp);
+		/* VMs */
+		tmp = RREG32(mmVM_CONTEXT1_CNTL);
+		tmp &= ~bits;
+		WREG32(mmVM_CONTEXT1_CNTL, tmp);
+		break;
+	case AMDGPU_IRQ_STATE_ENABLE:
+		/* system context */
+		tmp = RREG32(mmVM_CONTEXT0_CNTL);
+		tmp |= bits;
+		WREG32(mmVM_CONTEXT0_CNTL, tmp);
+		/* VMs */
+		tmp = RREG32(mmVM_CONTEXT1_CNTL);
+		tmp |= bits;
+		WREG32(mmVM_CONTEXT1_CNTL, tmp);
+		break;
+	default:
+		break;
+	}
+
+	return 0;
+}
+
+static int gmc_v7_0_process_interrupt(struct amdgpu_device *adev,
+				      struct amdgpu_irq_src *source,
+				      struct amdgpu_iv_entry *entry)
+{
+	u32 addr, status, mc_client;
+
+	addr = RREG32(mmVM_CONTEXT1_PROTECTION_FAULT_ADDR);
+	status = RREG32(mmVM_CONTEXT1_PROTECTION_FAULT_STATUS);
+	mc_client = RREG32(mmVM_CONTEXT1_PROTECTION_FAULT_MCCLIENT);
+	dev_err(adev->dev, "GPU fault detected: %d 0x%08x\n",
+		entry->src_id, entry->src_data);
+	dev_err(adev->dev, "  VM_CONTEXT1_PROTECTION_FAULT_ADDR   0x%08X\n",
+		addr);
+	dev_err(adev->dev, "  VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x%08X\n",
+		status);
+	gmc_v7_0_vm_decode_fault(adev, status, addr, mc_client);
+	/* reset addr and status */
+	WREG32_P(mmVM_CONTEXT1_CNTL2, 1, ~1);
+
+	return 0;
+}
+
+static int gmc_v7_0_set_clockgating_state(struct amdgpu_device *adev,
+					  enum amdgpu_clockgating_state state)
+{
+	bool gate = false;
+
+	if (state == AMDGPU_CG_STATE_GATE)
+		gate = true;
+
+	if (!(adev->flags & AMDGPU_IS_APU)) {
+		gmc_v7_0_enable_mc_mgcg(adev, gate);
+		gmc_v7_0_enable_mc_ls(adev, gate);
+	}
+	gmc_v7_0_enable_bif_mgls(adev, gate);
+	gmc_v7_0_enable_hdp_mgcg(adev, gate);
+	gmc_v7_0_enable_hdp_ls(adev, gate);
+
+	return 0;
+}
+
+static int gmc_v7_0_set_powergating_state(struct amdgpu_device *adev,
+					  enum amdgpu_powergating_state state)
+{
+	return 0;
+}
+
+const struct amdgpu_ip_funcs gmc_v7_0_ip_funcs = {
+	.early_init = gmc_v7_0_early_init,
+	.late_init = NULL,
+	.sw_init = gmc_v7_0_sw_init,
+	.sw_fini = gmc_v7_0_sw_fini,
+	.hw_init = gmc_v7_0_hw_init,
+	.hw_fini = gmc_v7_0_hw_fini,
+	.suspend = gmc_v7_0_suspend,
+	.resume = gmc_v7_0_resume,
+	.is_idle = gmc_v7_0_is_idle,
+	.wait_for_idle = gmc_v7_0_wait_for_idle,
+	.soft_reset = gmc_v7_0_soft_reset,
+	.print_status = gmc_v7_0_print_status,
+	.set_clockgating_state = gmc_v7_0_set_clockgating_state,
+	.set_powergating_state = gmc_v7_0_set_powergating_state,
+};
+
+static const struct amdgpu_gart_funcs gmc_v7_0_gart_funcs = {
+	.flush_gpu_tlb = gmc_v7_0_gart_flush_gpu_tlb,
+	.set_pte_pde = gmc_v7_0_gart_set_pte_pde,
+};
+
+static const struct amdgpu_irq_src_funcs gmc_v7_0_irq_funcs = {
+	.set = gmc_v7_0_vm_fault_interrupt_state,
+	.process = gmc_v7_0_process_interrupt,
+};
+
+static void gmc_v7_0_set_gart_funcs(struct amdgpu_device *adev)
+{
+	if (adev->gart.gart_funcs == NULL)
+		adev->gart.gart_funcs = &gmc_v7_0_gart_funcs;
+}
+
+static void gmc_v7_0_set_irq_funcs(struct amdgpu_device *adev)
+{
+	adev->mc.vm_fault.num_types = 1;
+	adev->mc.vm_fault.funcs = &gmc_v7_0_irq_funcs;
+}

drivers/gpu/drm/amd/amdgpu/gmc_v7_0.h

+/*
+ * Copyright 2014 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __GMC_V7_0_H__
+#define __GMC_V7_0_H__
+
+extern const struct amdgpu_ip_funcs gmc_v7_0_ip_funcs;
+
+/* XXX these shouldn't be exported */
+void gmc_v7_0_mc_stop(struct amdgpu_device *adev,
+		      struct amdgpu_mode_mc_save *save);
+void gmc_v7_0_mc_resume(struct amdgpu_device *adev,
+			struct amdgpu_mode_mc_save *save);
+int gmc_v7_0_mc_wait_for_idle(struct amdgpu_device *adev);
+
+#endif

drivers/gpu/drm/amd/amdgpu/kv_dpm.c

+/*
+ * Copyright 2013 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include "drmP.h"
+#include "amdgpu.h"
+#include "amdgpu_pm.h"
+#include "cikd.h"
+#include "atom.h"
+#include "amdgpu_atombios.h"
+#include "amdgpu_dpm.h"
+#include "kv_dpm.h"
+#include "gfx_v7_0.h"
+#include <linux/seq_file.h>
+
+#include "smu/smu_7_0_0_d.h"
+#include "smu/smu_7_0_0_sh_mask.h"
+
+#include "gca/gfx_7_2_d.h"
+#include "gca/gfx_7_2_sh_mask.h"
+
+#define KV_MAX_DEEPSLEEP_DIVIDER_ID     5
+#define KV_MINIMUM_ENGINE_CLOCK         800
+#define SMC_RAM_END                     0x40000
+
+static void kv_dpm_set_dpm_funcs(struct amdgpu_device *adev);
+static void kv_dpm_set_irq_funcs(struct amdgpu_device *adev);
+static int kv_enable_nb_dpm(struct amdgpu_device *adev,
+			    bool enable);
+static void kv_init_graphics_levels(struct amdgpu_device *adev);
+static int kv_calculate_ds_divider(struct amdgpu_device *adev);
+static int kv_calculate_nbps_level_settings(struct amdgpu_device *adev);
+static int kv_calculate_dpm_settings(struct amdgpu_device *adev);
+static void kv_enable_new_levels(struct amdgpu_device *adev);
+static void kv_program_nbps_index_settings(struct amdgpu_device *adev,
+					   struct amdgpu_ps *new_rps);
+static int kv_set_enabled_level(struct amdgpu_device *adev, u32 level);
+static int kv_set_enabled_levels(struct amdgpu_device *adev);
+static int kv_force_dpm_highest(struct amdgpu_device *adev);
+static int kv_force_dpm_lowest(struct amdgpu_device *adev);
+static void kv_apply_state_adjust_rules(struct amdgpu_device *adev,
+					struct amdgpu_ps *new_rps,
+					struct amdgpu_ps *old_rps);
+static int kv_set_thermal_temperature_range(struct amdgpu_device *adev,
+					    int min_temp, int max_temp);
+static int kv_init_fps_limits(struct amdgpu_device *adev);
+
+static void kv_dpm_powergate_uvd(struct amdgpu_device *adev, bool gate);
+static void kv_dpm_powergate_vce(struct amdgpu_device *adev, bool gate);
+static void kv_dpm_powergate_samu(struct amdgpu_device *adev, bool gate);
+static void kv_dpm_powergate_acp(struct amdgpu_device *adev, bool gate);
+
+
+static u32 kv_convert_vid2_to_vid7(struct amdgpu_device *adev,
+				   struct sumo_vid_mapping_table *vid_mapping_table,
+				   u32 vid_2bit)
+{
+	struct amdgpu_clock_voltage_dependency_table *vddc_sclk_table =
+		&adev->pm.dpm.dyn_state.vddc_dependency_on_sclk;
+	u32 i;
+
+	if (vddc_sclk_table && vddc_sclk_table->count) {
+		if (vid_2bit < vddc_sclk_table->count)
+			return vddc_sclk_table->entries[vid_2bit].v;
+		else
+			return vddc_sclk_table->entries[vddc_sclk_table->count - 1].v;
+	} else {
+		for (i = 0; i < vid_mapping_table->num_entries; i++) {
+			if (vid_mapping_table->entries[i].vid_2bit == vid_2bit)
+				return vid_mapping_table->entries[i].vid_7bit;
+		}
+		return vid_mapping_table->entries[vid_mapping_table->num_entries - 1].vid_7bit;
+	}
+}
+
+static u32 kv_convert_vid7_to_vid2(struct amdgpu_device *adev,
+				   struct sumo_vid_mapping_table *vid_mapping_table,
+				   u32 vid_7bit)
+{
+	struct amdgpu_clock_voltage_dependency_table *vddc_sclk_table =
+		&adev->pm.dpm.dyn_state.vddc_dependency_on_sclk;
+	u32 i;
+
+	if (vddc_sclk_table && vddc_sclk_table->count) {
+		for (i = 0; i < vddc_sclk_table->count; i++) {
+			if (vddc_sclk_table->entries[i].v == vid_7bit)
+				return i;
+		}
+		return vddc_sclk_table->count - 1;
+	} else {
+		for (i = 0; i < vid_mapping_table->num_entries; i++) {
+			if (vid_mapping_table->entries[i].vid_7bit == vid_7bit)
+				return vid_mapping_table->entries[i].vid_2bit;
+		}
+
+		return vid_mapping_table->entries[vid_mapping_table->num_entries - 1].vid_2bit;
+	}
+}
+
+static void sumo_take_smu_control(struct amdgpu_device *adev, bool enable)
+{
+/* This bit selects who handles display phy powergating.
+ * Clear the bit to let atom handle it.
+ * Set it to let the driver handle it.
+ * For now we just let atom handle it.
+ */
+#if 0
+	u32 v = RREG32(mmDOUT_SCRATCH3);
+
+	if (enable)
+		v |= 0x4;
+	else
+		v &= 0xFFFFFFFB;
+
+	WREG32(mmDOUT_SCRATCH3, v);
+#endif
+}
+
+static u32 sumo_get_sleep_divider_from_id(u32 id)
+{
+	return 1 << id;
+}
+
+static void sumo_construct_sclk_voltage_mapping_table(struct amdgpu_device *adev,
+						      struct sumo_sclk_voltage_mapping_table *sclk_voltage_mapping_table,
+						      ATOM_AVAILABLE_SCLK_LIST *table)
+{
+	u32 i;
+	u32 n = 0;
+	u32 prev_sclk = 0;
+
+	for (i = 0; i < SUMO_MAX_HARDWARE_POWERLEVELS; i++) {
+		if (table[i].ulSupportedSCLK > prev_sclk) {
+			sclk_voltage_mapping_table->entries[n].sclk_frequency =
+				table[i].ulSupportedSCLK;
+			sclk_voltage_mapping_table->entries[n].vid_2bit =
+				table[i].usVoltageIndex;
+			prev_sclk = table[i].ulSupportedSCLK;
+			n++;
+		}
+	}
+
+	sclk_voltage_mapping_table->num_max_dpm_entries = n;
+}
+
+static void sumo_construct_vid_mapping_table(struct amdgpu_device *adev,
+					     struct sumo_vid_mapping_table *vid_mapping_table,
+					     ATOM_AVAILABLE_SCLK_LIST *table)
+{
+	u32 i, j;
+
+	for (i = 0; i < SUMO_MAX_HARDWARE_POWERLEVELS; i++) {
+		if (table[i].ulSupportedSCLK != 0) {
+			vid_mapping_table->entries[table[i].usVoltageIndex].vid_7bit =
+				table[i].usVoltageID;
+			vid_mapping_table->entries[table[i].usVoltageIndex].vid_2bit =
+				table[i].usVoltageIndex;
+		}
+	}
+
+	for (i = 0; i < SUMO_MAX_NUMBER_VOLTAGES; i++) {
+		if (vid_mapping_table->entries[i].vid_7bit == 0) {
+			for (j = i + 1; j < SUMO_MAX_NUMBER_VOLTAGES; j++) {
+				if (vid_mapping_table->entries[j].vid_7bit != 0) {
+					vid_mapping_table->entries[i] =
+						vid_mapping_table->entries[j];
+					vid_mapping_table->entries[j].vid_7bit = 0;
+					break;
+				}
+			}
+
+			if (j == SUMO_MAX_NUMBER_VOLTAGES)
+				break;
+		}
+	}
+
+	vid_mapping_table->num_entries = i;
+}
+
+static const struct kv_lcac_config_values sx_local_cac_cfg_kv[] =
+{
+	{  0,       4,        1    },
+	{  1,       4,        1    },
+	{  2,       5,        1    },
+	{  3,       4,        2    },
+	{  4,       1,        1    },
+	{  5,       5,        2    },
+	{  6,       6,        1    },
+	{  7,       9,        2    },
+	{ 0xffffffff }
+};
+
+static const struct kv_lcac_config_values mc0_local_cac_cfg_kv[] =
+{
+	{  0,       4,        1    },
+	{ 0xffffffff }
+};
+
+static const struct kv_lcac_config_values mc1_local_cac_cfg_kv[] =
+{
+	{  0,       4,        1    },
+	{ 0xffffffff }
+};
+
+static const struct kv_lcac_config_values mc2_local_cac_cfg_kv[] =
+{
+	{  0,       4,        1    },
+	{ 0xffffffff }
+};
+
+static const struct kv_lcac_config_values mc3_local_cac_cfg_kv[] =
+{
+	{  0,       4,        1    },
+	{ 0xffffffff }
+};
+
+static const struct kv_lcac_config_values cpl_local_cac_cfg_kv[] =
+{
+	{  0,       4,        1    },
+	{  1,       4,        1    },
+	{  2,       5,        1    },
+	{  3,       4,        1    },
+	{  4,       1,        1    },
+	{  5,       5,        1    },
+	{  6,       6,        1    },
+	{  7,       9,        1    },
+	{  8,       4,        1    },
+	{  9,       2,        1    },
+	{  10,      3,        1    },
+	{  11,      6,        1    },
+	{  12,      8,        2    },
+	{  13,      1,        1    },
+	{  14,      2,        1    },
+	{  15,      3,        1    },
+	{  16,      1,        1    },
+	{  17,      4,        1    },
+	{  18,      3,        1    },
+	{  19,      1,        1    },
+	{  20,      8,        1    },
+	{  21,      5,        1    },
+	{  22,      1,        1    },
+	{  23,      1,        1    },
+	{  24,      4,        1    },
+	{  27,      6,        1    },
+	{  28,      1,        1    },
+	{ 0xffffffff }
+};
+
+static const struct kv_lcac_config_reg sx0_cac_config_reg[] =
+{
+	{ 0xc0400d00, 0x003e0000, 17, 0x3fc00000, 22, 0x0001fffe, 1, 0x00000001, 0 }
+};
+
+static const struct kv_lcac_config_reg mc0_cac_config_reg[] =
+{
+	{ 0xc0400d30, 0x003e0000, 17, 0x3fc00000, 22, 0x0001fffe, 1, 0x00000001, 0 }
+};
+
+static const struct kv_lcac_config_reg mc1_cac_config_reg[] =
+{
+	{ 0xc0400d3c, 0x003e0000, 17, 0x3fc00000, 22, 0x0001fffe, 1, 0x00000001, 0 }
+};
+
+static const struct kv_lcac_config_reg mc2_cac_config_reg[] =
+{
+	{ 0xc0400d48, 0x003e0000, 17, 0x3fc00000, 22, 0x0001fffe, 1, 0x00000001, 0 }
+};
+
+static const struct kv_lcac_config_reg mc3_cac_config_reg[] =
+{
+	{ 0xc0400d54, 0x003e0000, 17, 0x3fc00000, 22, 0x0001fffe, 1, 0x00000001, 0 }
+};
+
+static const struct kv_lcac_config_reg cpl_cac_config_reg[] =
+{
+	{ 0xc0400d80, 0x003e0000, 17, 0x3fc00000, 22, 0x0001fffe, 1, 0x00000001, 0 }
+};
+
+static const struct kv_pt_config_reg didt_config_kv[] =
+{
+	{ 0x10, 0x000000ff, 0, 0x0, KV_CONFIGREG_DIDT_IND },
+	{ 0x10, 0x0000ff00, 8, 0x0, KV_CONFIGREG_DIDT_IND },
+	{ 0x10, 0x00ff0000, 16, 0x0, KV_CONFIGREG_DIDT_IND },
+	{ 0x10, 0xff000000, 24, 0x0, KV_CONFIGREG_DIDT_IND },
+	{ 0x11, 0x000000ff, 0, 0x0, KV_CONFIGREG_DIDT_IND },
+	{ 0x11, 0x0000ff00, 8, 0x0, KV_CONFIGREG_DIDT_IND },
+	{ 0x11, 0x00ff0000, 16, 0x0, KV_CONFIGREG_DIDT_IND },
+	{ 0x11, 0xff000000, 24, 0x0, KV_CONFIGREG_DIDT_IND },
+	{ 0x12, 0x000000ff, 0, 0x0, KV_CONFIGREG_DIDT_IND },
+	{ 0x12, 0x0000ff00, 8, 0x0, KV_CONFIGREG_DIDT_IND },
+	{ 0x12, 0x00ff0000, 16, 0x0, KV_CONFIGREG_DIDT_IND },
+	{ 0x12, 0xff000000, 24, 0x0, KV_CONFIGREG_DIDT_IND },
+	{ 0x2, 0x00003fff, 0, 0x4, KV_CONFIGREG_DIDT_IND },
+	{ 0x2, 0x03ff0000, 16, 0x80, KV_CONFIGREG_DIDT_IND },
+	{ 0x2, 0x78000000, 27, 0x3, KV_CONFIGREG_DIDT_IND },
+	{ 0x1, 0x0000ffff, 0, 0x3FFF, KV_CONFIGREG_DIDT_IND },
+	{ 0x1, 0xffff0000, 16, 0x3FFF, KV_CONFIGREG_DIDT_IND },
+	{ 0x0, 0x00000001, 0, 0x0, KV_CONFIGREG_DIDT_IND },
+	{ 0x30, 0x000000ff, 0, 0x0, KV_CONFIGREG_DIDT_IND },
+	{ 0x30, 0x0000ff00, 8, 0x0, KV_CONFIGREG_DIDT_IND },
+	{ 0x30, 0x00ff0000, 16, 0x0, KV_CONFIGREG_DIDT_IND },
+	{ 0x30, 0xff000000, 24, 0x0, KV_CONFIGREG_DIDT_IND },
+	{ 0x31, 0x000000ff, 0, 0x0, KV_CONFIGREG_DIDT_IND },
+	{ 0x31, 0x0000ff00, 8, 0x0, KV_CONFIGREG_DIDT_IND },
+	{ 0x31, 0x00ff0000, 16, 0x0, KV_CONFIGREG_DIDT_IND },
+	{ 0x31, 0xff000000, 24, 0x0, KV_CONFIGREG_DIDT_IND },
+	{ 0x32, 0x000000ff, 0, 0x0, KV_CONFIGREG_DIDT_IND },
+	{ 0x32, 0x0000ff00, 8, 0x0, KV_CONFIGREG_DIDT_IND },
+	{ 0x32, 0x00ff0000, 16, 0x0, KV_CONFIGREG_DIDT_IND },
+	{ 0x32, 0xff000000, 24, 0x0, KV_CONFIGREG_DIDT_IND },
+	{ 0x22, 0x00003fff, 0, 0x4, KV_CONFIGREG_DIDT_IND },
+	{ 0x22, 0x03ff0000, 16, 0x80, KV_CONFIGREG_DIDT_IND },
+	{ 0x22, 0x78000000, 27, 0x3, KV_CONFIGREG_DIDT_IND },
+	{ 0x21, 0x0000ffff, 0, 0x3FFF, KV_CONFIGREG_DIDT_IND },
+	{ 0x21, 0xffff0000, 16, 0x3FFF, KV_CONFIGREG_DIDT_IND },
+	{ 0x20, 0x00000001, 0, 0x0, KV_CONFIGREG_DIDT_IND },
+	{ 0x50, 0x000000ff, 0, 0x0, KV_CONFIGREG_DIDT_IND },
+	{ 0x50, 0x0000ff00, 8, 0x0, KV_CONFIGREG_DIDT_IND },
+	{ 0x50, 0x00ff0000, 16, 0x0, KV_CONFIGREG_DIDT_IND },
+	{ 0x50, 0xff000000, 24, 0x0, KV_CONFIGREG_DIDT_IND },
+	{ 0x51, 0x000000ff, 0, 0x0, KV_CONFIGREG_DIDT_IND },
+	{ 0x51, 0x0000ff00, 8, 0x0, KV_CONFIGREG_DIDT_IND },
+	{ 0x51, 0x00ff0000, 16, 0x0, KV_CONFIGREG_DIDT_IND },
+	{ 0x51, 0xff000000, 24, 0x0, KV_CONFIGREG_DIDT_IND },
+	{ 0x52, 0x000000ff, 0, 0x0, KV_CONFIGREG_DIDT_IND },
+	{ 0x52, 0x0000ff00, 8, 0x0, KV_CONFIGREG_DIDT_IND },
+	{ 0x52, 0x00ff0000, 16, 0x0, KV_CONFIGREG_DIDT_IND },
+	{ 0x52, 0xff000000, 24, 0x0, KV_CONFIGREG_DIDT_IND },
+	{ 0x42, 0x00003fff, 0, 0x4, KV_CONFIGREG_DIDT_IND },
+	{ 0x42, 0x03ff0000, 16, 0x80, KV_CONFIGREG_DIDT_IND },
+	{ 0x42, 0x78000000, 27, 0x3, KV_CONFIGREG_DIDT_IND },
+	{ 0x41, 0x0000ffff, 0, 0x3FFF, KV_CONFIGREG_DIDT_IND },
+	{ 0x41, 0xffff0000, 16, 0x3FFF, KV_CONFIGREG_DIDT_IND },
+	{ 0x40, 0x00000001, 0, 0x0, KV_CONFIGREG_DIDT_IND },
+	{ 0x70, 0x000000ff, 0, 0x0, KV_CONFIGREG_DIDT_IND },
+	{ 0x70, 0x0000ff00, 8, 0x0, KV_CONFIGREG_DIDT_IND },
+	{ 0x70, 0x00ff0000, 16, 0x0, KV_CONFIGREG_DIDT_IND },
+	{ 0x70, 0xff000000, 24, 0x0, KV_CONFIGREG_DIDT_IND },
+	{ 0x71, 0x000000ff, 0, 0x0, KV_CONFIGREG_DIDT_IND },
+	{ 0x71, 0x0000ff00, 8, 0x0, KV_CONFIGREG_DIDT_IND },
+	{ 0x71, 0x00ff0000, 16, 0x0, KV_CONFIGREG_DIDT_IND },
+	{ 0x71, 0xff000000, 24, 0x0, KV_CONFIGREG_DIDT_IND },
+	{ 0x72, 0x000000ff, 0, 0x0, KV_CONFIGREG_DIDT_IND },
+	{ 0x72, 0x0000ff00, 8, 0x0, KV_CONFIGREG_DIDT_IND },
+	{ 0x72, 0x00ff0000, 16, 0x0, KV_CONFIGREG_DIDT_IND },
+	{ 0x72, 0xff000000, 24, 0x0, KV_CONFIGREG_DIDT_IND },
+	{ 0x62, 0x00003fff, 0, 0x4, KV_CONFIGREG_DIDT_IND },
+	{ 0x62, 0x03ff0000, 16, 0x80, KV_CONFIGREG_DIDT_IND },
+	{ 0x62, 0x78000000, 27, 0x3, KV_CONFIGREG_DIDT_IND },
+	{ 0x61, 0x0000ffff, 0, 0x3FFF, KV_CONFIGREG_DIDT_IND },
+	{ 0x61, 0xffff0000, 16, 0x3FFF, KV_CONFIGREG_DIDT_IND },
+	{ 0x60, 0x00000001, 0, 0x0, KV_CONFIGREG_DIDT_IND },
+	{ 0xFFFFFFFF }
+};
+
+static struct kv_ps *kv_get_ps(struct amdgpu_ps *rps)
+{
+	struct kv_ps *ps = rps->ps_priv;
+
+	return ps;
+}
+
+static struct kv_power_info *kv_get_pi(struct amdgpu_device *adev)
+{
+	struct kv_power_info *pi = adev->pm.dpm.priv;
+
+	return pi;
+}
+
+#if 0
+static void kv_program_local_cac_table(struct amdgpu_device *adev,
+				       const struct kv_lcac_config_values *local_cac_table,
+				       const struct kv_lcac_config_reg *local_cac_reg)
+{
+	u32 i, count, data;
+	const struct kv_lcac_config_values *values = local_cac_table;
+
+	while (values->block_id != 0xffffffff) {
+		count = values->signal_id;
+		for (i = 0; i < count; i++) {
+			data = ((values->block_id << local_cac_reg->block_shift) &
+				local_cac_reg->block_mask);
+			data |= ((i << local_cac_reg->signal_shift) &
+				 local_cac_reg->signal_mask);
+			data |= ((values->t << local_cac_reg->t_shift) &
+				 local_cac_reg->t_mask);
+			data |= ((1 << local_cac_reg->enable_shift) &
+				 local_cac_reg->enable_mask);
+			WREG32_SMC(local_cac_reg->cntl, data);
+		}
+		values++;
+	}
+}
+#endif
+
+static int kv_program_pt_config_registers(struct amdgpu_device *adev,
+					  const struct kv_pt_config_reg *cac_config_regs)
+{
+	const struct kv_pt_config_reg *config_regs = cac_config_regs;
+	u32 data;
+	u32 cache = 0;
+
+	if (config_regs == NULL)
+		return -EINVAL;
+
+	while (config_regs->offset != 0xFFFFFFFF) {
+		if (config_regs->type == KV_CONFIGREG_CACHE) {
+			cache |= ((config_regs->value << config_regs->shift) & config_regs->mask);
+		} else {
+			switch (config_regs->type) {
+			case KV_CONFIGREG_SMC_IND:
+				data = RREG32_SMC(config_regs->offset);
+				break;
+			case KV_CONFIGREG_DIDT_IND:
+				data = RREG32_DIDT(config_regs->offset);
+				break;
+			default:
+				data = RREG32(config_regs->offset);
+				break;
+			}
+
+			data &= ~config_regs->mask;
+			data |= ((config_regs->value << config_regs->shift) & config_regs->mask);
+			data |= cache;
+			cache = 0;
+
+			switch (config_regs->type) {
+			case KV_CONFIGREG_SMC_IND:
+				WREG32_SMC(config_regs->offset, data);
+				break;
+			case KV_CONFIGREG_DIDT_IND:
+				WREG32_DIDT(config_regs->offset, data);
+				break;
+			default:
+				WREG32(config_regs->offset, data);
+				break;
+			}
+		}
+		config_regs++;
+	}
+
+	return 0;
+}
+
+static void kv_do_enable_didt(struct amdgpu_device *adev, bool enable)
+{
+	struct kv_power_info *pi = kv_get_pi(adev);
+	u32 data;
+
+	if (pi->caps_sq_ramping) {
+		data = RREG32_DIDT(ixDIDT_SQ_CTRL0);
+		if (enable)
+			data |= DIDT_SQ_CTRL0__DIDT_CTRL_EN_MASK;
+		else
+			data &= ~DIDT_SQ_CTRL0__DIDT_CTRL_EN_MASK;
+		WREG32_DIDT(ixDIDT_SQ_CTRL0, data);
+	}
+
+	if (pi->caps_db_ramping) {
+		data = RREG32_DIDT(ixDIDT_DB_CTRL0);
+		if (enable)
+			data |= DIDT_DB_CTRL0__DIDT_CTRL_EN_MASK;
+		else
+			data &= ~DIDT_DB_CTRL0__DIDT_CTRL_EN_MASK;
+		WREG32_DIDT(ixDIDT_DB_CTRL0, data);
+	}
+
+	if (pi->caps_td_ramping) {
+		data = RREG32_DIDT(ixDIDT_TD_CTRL0);
+		if (enable)
+			data |= DIDT_TD_CTRL0__DIDT_CTRL_EN_MASK;
+		else
+			data &= ~DIDT_TD_CTRL0__DIDT_CTRL_EN_MASK;
+		WREG32_DIDT(ixDIDT_TD_CTRL0, data);
+	}
+
+	if (pi->caps_tcp_ramping) {
+		data = RREG32_DIDT(ixDIDT_TCP_CTRL0);
+		if (enable)
+			data |= DIDT_TCP_CTRL0__DIDT_CTRL_EN_MASK;
+		else
+			data &= ~DIDT_TCP_CTRL0__DIDT_CTRL_EN_MASK;
+		WREG32_DIDT(ixDIDT_TCP_CTRL0, data);
+	}
+}
+
+static int kv_enable_didt(struct amdgpu_device *adev, bool enable)
+{
+	struct kv_power_info *pi = kv_get_pi(adev);
+	int ret;
+
+	if (pi->caps_sq_ramping ||
+	    pi->caps_db_ramping ||
+	    pi->caps_td_ramping ||
+	    pi->caps_tcp_ramping) {
+		gfx_v7_0_enter_rlc_safe_mode(adev);
+
+		if (enable) {
+			ret = kv_program_pt_config_registers(adev, didt_config_kv);
+			if (ret) {
+				gfx_v7_0_exit_rlc_safe_mode(adev);
+				return ret;
+			}
+		}
+
+		kv_do_enable_didt(adev, enable);
+
+		gfx_v7_0_exit_rlc_safe_mode(adev);
+	}
+
+	return 0;
+}
+
+#if 0
+static void kv_initialize_hardware_cac_manager(struct amdgpu_device *adev)
+{
+	struct kv_power_info *pi = kv_get_pi(adev);
+
+	if (pi->caps_cac) {
+		WREG32_SMC(ixLCAC_SX0_OVR_SEL, 0);
+		WREG32_SMC(ixLCAC_SX0_OVR_VAL, 0);
+		kv_program_local_cac_table(adev, sx_local_cac_cfg_kv, sx0_cac_config_reg);
+
+		WREG32_SMC(ixLCAC_MC0_OVR_SEL, 0);
+		WREG32_SMC(ixLCAC_MC0_OVR_VAL, 0);
+		kv_program_local_cac_table(adev, mc0_local_cac_cfg_kv, mc0_cac_config_reg);
+
+		WREG32_SMC(ixLCAC_MC1_OVR_SEL, 0);
+		WREG32_SMC(ixLCAC_MC1_OVR_VAL, 0);
+		kv_program_local_cac_table(adev, mc1_local_cac_cfg_kv, mc1_cac_config_reg);
+
+		WREG32_SMC(ixLCAC_MC2_OVR_SEL, 0);
+		WREG32_SMC(ixLCAC_MC2_OVR_VAL, 0);
+		kv_program_local_cac_table(adev, mc2_local_cac_cfg_kv, mc2_cac_config_reg);
+
+		WREG32_SMC(ixLCAC_MC3_OVR_SEL, 0);
+		WREG32_SMC(ixLCAC_MC3_OVR_VAL, 0);
+		kv_program_local_cac_table(adev, mc3_local_cac_cfg_kv, mc3_cac_config_reg);
+
+		WREG32_SMC(ixLCAC_CPL_OVR_SEL, 0);
+		WREG32_SMC(ixLCAC_CPL_OVR_VAL, 0);
+		kv_program_local_cac_table(adev, cpl_local_cac_cfg_kv, cpl_cac_config_reg);
+	}
+}
+#endif
+
+static int kv_enable_smc_cac(struct amdgpu_device *adev, bool enable)
+{
+	struct kv_power_info *pi = kv_get_pi(adev);
+	int ret = 0;
+
+	if (pi->caps_cac) {
+		if (enable) {
+			ret = amdgpu_kv_notify_message_to_smu(adev, PPSMC_MSG_EnableCac);
+			if (ret)
+				pi->cac_enabled = false;
+			else
+				pi->cac_enabled = true;
+		} else if (pi->cac_enabled) {
+			amdgpu_kv_notify_message_to_smu(adev, PPSMC_MSG_DisableCac);
+			pi->cac_enabled = false;
+		}
+	}
+
+	return ret;
+}
+
+static int kv_process_firmware_header(struct amdgpu_device *adev)
+{
+	struct kv_power_info *pi = kv_get_pi(adev);
+	u32 tmp;
+	int ret;
+
+	ret = amdgpu_kv_read_smc_sram_dword(adev, SMU7_FIRMWARE_HEADER_LOCATION +
+				     offsetof(SMU7_Firmware_Header, DpmTable),
+				     &tmp, pi->sram_end);
+
+	if (ret == 0)
+		pi->dpm_table_start = tmp;
+
+	ret = amdgpu_kv_read_smc_sram_dword(adev, SMU7_FIRMWARE_HEADER_LOCATION +
+				     offsetof(SMU7_Firmware_Header, SoftRegisters),
+				     &tmp, pi->sram_end);
+
+	if (ret == 0)
+		pi->soft_regs_start = tmp;
+
+	return ret;
+}
+
+static int kv_enable_dpm_voltage_scaling(struct amdgpu_device *adev)
+{
+	struct kv_power_info *pi = kv_get_pi(adev);
+	int ret;
+
+	pi->graphics_voltage_change_enable = 1;
+
+	ret = amdgpu_kv_copy_bytes_to_smc(adev,
+				   pi->dpm_table_start +
+				   offsetof(SMU7_Fusion_DpmTable, GraphicsVoltageChangeEnable),
+				   &pi->graphics_voltage_change_enable,
+				   sizeof(u8), pi->sram_end);
+
+	return ret;
+}
+
+static int kv_set_dpm_interval(struct amdgpu_device *adev)
+{
+	struct kv_power_info *pi = kv_get_pi(adev);
+	int ret;
+
+	pi->graphics_interval = 1;
+
+	ret = amdgpu_kv_copy_bytes_to_smc(adev,
+				   pi->dpm_table_start +
+				   offsetof(SMU7_Fusion_DpmTable, GraphicsInterval),
+				   &pi->graphics_interval,
+				   sizeof(u8), pi->sram_end);
+
+	return ret;
+}
+
+static int kv_set_dpm_boot_state(struct amdgpu_device *adev)
+{
+	struct kv_power_info *pi = kv_get_pi(adev);
+	int ret;
+
+	ret = amdgpu_kv_copy_bytes_to_smc(adev,
+				   pi->dpm_table_start +
+				   offsetof(SMU7_Fusion_DpmTable, GraphicsBootLevel),
+				   &pi->graphics_boot_level,
+				   sizeof(u8), pi->sram_end);
+
+	return ret;
+}
+
+static void kv_program_vc(struct amdgpu_device *adev)
+{
+	WREG32_SMC(ixCG_FREQ_TRAN_VOTING_0, 0x3FFFC100);
+}
+
+static void kv_clear_vc(struct amdgpu_device *adev)
+{
+	WREG32_SMC(ixCG_FREQ_TRAN_VOTING_0, 0);
+}
+
+static int kv_set_divider_value(struct amdgpu_device *adev,
+				u32 index, u32 sclk)
+{
+	struct kv_power_info *pi = kv_get_pi(adev);
+	struct atom_clock_dividers dividers;
+	int ret;
+
+	ret = amdgpu_atombios_get_clock_dividers(adev, COMPUTE_ENGINE_PLL_PARAM,
+						 sclk, false, &dividers);
+	if (ret)
+		return ret;
+
+	pi->graphics_level[index].SclkDid = (u8)dividers.post_div;
+	pi->graphics_level[index].SclkFrequency = cpu_to_be32(sclk);
+
+	return 0;
+}
+
+static u16 kv_convert_8bit_index_to_voltage(struct amdgpu_device *adev,
+					    u16 voltage)
+{
+	return 6200 - (voltage * 25);
+}
+
+static u16 kv_convert_2bit_index_to_voltage(struct amdgpu_device *adev,
+					    u32 vid_2bit)
+{
+	struct kv_power_info *pi = kv_get_pi(adev);
+	u32 vid_8bit = kv_convert_vid2_to_vid7(adev,
+					       &pi->sys_info.vid_mapping_table,
+					       vid_2bit);
+
+	return kv_convert_8bit_index_to_voltage(adev, (u16)vid_8bit);
+}
+
+
+static int kv_set_vid(struct amdgpu_device *adev, u32 index, u32 vid)
+{
+	struct kv_power_info *pi = kv_get_pi(adev);
+
+	pi->graphics_level[index].VoltageDownH = (u8)pi->voltage_drop_t;
+	pi->graphics_level[index].MinVddNb =
+		cpu_to_be32(kv_convert_2bit_index_to_voltage(adev, vid));
+
+	return 0;
+}
+
+static int kv_set_at(struct amdgpu_device *adev, u32 index, u32 at)
+{
+	struct kv_power_info *pi = kv_get_pi(adev);
+
+	pi->graphics_level[index].AT = cpu_to_be16((u16)at);
+
+	return 0;
+}
+
+static void kv_dpm_power_level_enable(struct amdgpu_device *adev,
+				      u32 index, bool enable)
+{
+	struct kv_power_info *pi = kv_get_pi(adev);
+
+	pi->graphics_level[index].EnabledForActivity = enable ? 1 : 0;
+}
+
+static void kv_start_dpm(struct amdgpu_device *adev)
+{
+	u32 tmp = RREG32_SMC(ixGENERAL_PWRMGT);
+
+	tmp |= GENERAL_PWRMGT__GLOBAL_PWRMGT_EN_MASK;
+	WREG32_SMC(ixGENERAL_PWRMGT, tmp);
+
+	amdgpu_kv_smc_dpm_enable(adev, true);
+}
+
+static void kv_stop_dpm(struct amdgpu_device *adev)
+{
+	amdgpu_kv_smc_dpm_enable(adev, false);
+}
+
+static void kv_start_am(struct amdgpu_device *adev)
+{
+	u32 sclk_pwrmgt_cntl = RREG32_SMC(ixSCLK_PWRMGT_CNTL);
+
+	sclk_pwrmgt_cntl &= ~(SCLK_PWRMGT_CNTL__RESET_SCLK_CNT_MASK |
+			SCLK_PWRMGT_CNTL__RESET_BUSY_CNT_MASK);
+	sclk_pwrmgt_cntl |= SCLK_PWRMGT_CNTL__DYNAMIC_PM_EN_MASK;
+
+	WREG32_SMC(ixSCLK_PWRMGT_CNTL, sclk_pwrmgt_cntl);
+}
+
+static void kv_reset_am(struct amdgpu_device *adev)
+{
+	u32 sclk_pwrmgt_cntl = RREG32_SMC(ixSCLK_PWRMGT_CNTL);
+
+	sclk_pwrmgt_cntl |= (SCLK_PWRMGT_CNTL__RESET_SCLK_CNT_MASK |
+			SCLK_PWRMGT_CNTL__RESET_BUSY_CNT_MASK);
+
+	WREG32_SMC(ixSCLK_PWRMGT_CNTL, sclk_pwrmgt_cntl);
+}
+
+static int kv_freeze_sclk_dpm(struct amdgpu_device *adev, bool freeze)
+{
+	return amdgpu_kv_notify_message_to_smu(adev, freeze ?
+					PPSMC_MSG_SCLKDPM_FreezeLevel : PPSMC_MSG_SCLKDPM_UnfreezeLevel);
+}
+
+static int kv_force_lowest_valid(struct amdgpu_device *adev)
+{
+	return kv_force_dpm_lowest(adev);
+}
+
+static int kv_unforce_levels(struct amdgpu_device *adev)
+{
+	if (adev->asic_type == CHIP_KABINI || adev->asic_type == CHIP_MULLINS)
+		return amdgpu_kv_notify_message_to_smu(adev, PPSMC_MSG_NoForcedLevel);
+	else
+		return kv_set_enabled_levels(adev);
+}
+
+static int kv_update_sclk_t(struct amdgpu_device *adev)
+{
+	struct kv_power_info *pi = kv_get_pi(adev);
+	u32 low_sclk_interrupt_t = 0;
+	int ret = 0;
+
+	if (pi->caps_sclk_throttle_low_notification) {
+		low_sclk_interrupt_t = cpu_to_be32(pi->low_sclk_interrupt_t);
+
+		ret = amdgpu_kv_copy_bytes_to_smc(adev,
+					   pi->dpm_table_start +
+					   offsetof(SMU7_Fusion_DpmTable, LowSclkInterruptT),
+					   (u8 *)&low_sclk_interrupt_t,
+					   sizeof(u32), pi->sram_end);
+	}
+	return ret;
+}
+
+static int kv_program_bootup_state(struct amdgpu_device *adev)
+{
+	struct kv_power_info *pi = kv_get_pi(adev);
+	u32 i;
+	struct amdgpu_clock_voltage_dependency_table *table =
+		&adev->pm.dpm.dyn_state.vddc_dependency_on_sclk;
+
+	if (table && table->count) {
+		for (i = pi->graphics_dpm_level_count - 1; i > 0; i--) {
+			if (table->entries[i].clk == pi->boot_pl.sclk)
+				break;
+		}
+
+		pi->graphics_boot_level = (u8)i;
+		kv_dpm_power_level_enable(adev, i, true);
+	} else {
+		struct sumo_sclk_voltage_mapping_table *table =
+			&pi->sys_info.sclk_voltage_mapping_table;
+
+		if (table->num_max_dpm_entries == 0)
+			return -EINVAL;
+
+		for (i = pi->graphics_dpm_level_count - 1; i > 0; i--) {
+			if (table->entries[i].sclk_frequency == pi->boot_pl.sclk)
+				break;
+		}
+
+		pi->graphics_boot_level = (u8)i;
+		kv_dpm_power_level_enable(adev, i, true);
+	}
+	return 0;
+}
+
+static int kv_enable_auto_thermal_throttling(struct amdgpu_device *adev)
+{
+	struct kv_power_info *pi = kv_get_pi(adev);
+	int ret;
+
+	pi->graphics_therm_throttle_enable = 1;
+
+	ret = amdgpu_kv_copy_bytes_to_smc(adev,
+				   pi->dpm_table_start +
+				   offsetof(SMU7_Fusion_DpmTable, GraphicsThermThrottleEnable),
+				   &pi->graphics_therm_throttle_enable,
+				   sizeof(u8), pi->sram_end);
+
+	return ret;
+}
+
+static int kv_upload_dpm_settings(struct amdgpu_device *adev)
+{
+	struct kv_power_info *pi = kv_get_pi(adev);
+	int ret;
+
+	ret = amdgpu_kv_copy_bytes_to_smc(adev,
+				   pi->dpm_table_start +
+				   offsetof(SMU7_Fusion_DpmTable, GraphicsLevel),
+				   (u8 *)&pi->graphics_level,
+				   sizeof(SMU7_Fusion_GraphicsLevel) * SMU7_MAX_LEVELS_GRAPHICS,
+				   pi->sram_end);
+
+	if (ret)
+		return ret;
+
+	ret = amdgpu_kv_copy_bytes_to_smc(adev,
+				   pi->dpm_table_start +
+				   offsetof(SMU7_Fusion_DpmTable, GraphicsDpmLevelCount),
+				   &pi->graphics_dpm_level_count,
+				   sizeof(u8), pi->sram_end);
+
+	return ret;
+}
+
+static u32 kv_get_clock_difference(u32 a, u32 b)
+{
+	return (a >= b) ? a - b : b - a;
+}
+
+static u32 kv_get_clk_bypass(struct amdgpu_device *adev, u32 clk)
+{
+	struct kv_power_info *pi = kv_get_pi(adev);
+	u32 value;
+
+	if (pi->caps_enable_dfs_bypass) {
+		if (kv_get_clock_difference(clk, 40000) < 200)
+			value = 3;
+		else if (kv_get_clock_difference(clk, 30000) < 200)
+			value = 2;
+		else if (kv_get_clock_difference(clk, 20000) < 200)
+			value = 7;
+		else if (kv_get_clock_difference(clk, 15000) < 200)
+			value = 6;
+		else if (kv_get_clock_difference(clk, 10000) < 200)
+			value = 8;
+		else
+			value = 0;
+	} else {
+		value = 0;
+	}
+
+	return value;
+}
+
+static int kv_populate_uvd_table(struct amdgpu_device *adev)
+{
+	struct kv_power_info *pi = kv_get_pi(adev);
+	struct amdgpu_uvd_clock_voltage_dependency_table *table =
+		&adev->pm.dpm.dyn_state.uvd_clock_voltage_dependency_table;
+	struct atom_clock_dividers dividers;
+	int ret;
+	u32 i;
+
+	if (table == NULL || table->count == 0)
+		return 0;
+
+	pi->uvd_level_count = 0;
+	for (i = 0; i < table->count; i++) {
+		if (pi->high_voltage_t &&
+		    (pi->high_voltage_t < table->entries[i].v))
+			break;
+
+		pi->uvd_level[i].VclkFrequency = cpu_to_be32(table->entries[i].vclk);
+		pi->uvd_level[i].DclkFrequency = cpu_to_be32(table->entries[i].dclk);
+		pi->uvd_level[i].MinVddNb = cpu_to_be16(table->entries[i].v);
+
+		pi->uvd_level[i].VClkBypassCntl =
+			(u8)kv_get_clk_bypass(adev, table->entries[i].vclk);
+		pi->uvd_level[i].DClkBypassCntl =
+			(u8)kv_get_clk_bypass(adev, table->entries[i].dclk);
+
+		ret = amdgpu_atombios_get_clock_dividers(adev, COMPUTE_ENGINE_PLL_PARAM,
+							 table->entries[i].vclk, false, &dividers);
+		if (ret)
+			return ret;
+		pi->uvd_level[i].VclkDivider = (u8)dividers.post_div;
+
+		ret = amdgpu_atombios_get_clock_dividers(adev, COMPUTE_ENGINE_PLL_PARAM,
+							 table->entries[i].dclk, false, &dividers);
+		if (ret)
+			return ret;
+		pi->uvd_level[i].DclkDivider = (u8)dividers.post_div;
+
+		pi->uvd_level_count++;
+	}
+
+	ret = amdgpu_kv_copy_bytes_to_smc(adev,
+				   pi->dpm_table_start +
+				   offsetof(SMU7_Fusion_DpmTable, UvdLevelCount),
+				   (u8 *)&pi->uvd_level_count,
+				   sizeof(u8), pi->sram_end);
+	if (ret)
+		return ret;
+
+	pi->uvd_interval = 1;
+
+	ret = amdgpu_kv_copy_bytes_to_smc(adev,
+				   pi->dpm_table_start +
+				   offsetof(SMU7_Fusion_DpmTable, UVDInterval),
+				   &pi->uvd_interval,
+				   sizeof(u8), pi->sram_end);
+	if (ret)
+		return ret;
+
+	ret = amdgpu_kv_copy_bytes_to_smc(adev,
+				   pi->dpm_table_start +
+				   offsetof(SMU7_Fusion_DpmTable, UvdLevel),
+				   (u8 *)&pi->uvd_level,
+				   sizeof(SMU7_Fusion_UvdLevel) * SMU7_MAX_LEVELS_UVD,
+				   pi->sram_end);
+
+	return ret;
+
+}
+
+static int kv_populate_vce_table(struct amdgpu_device *adev)
+{
+	struct kv_power_info *pi = kv_get_pi(adev);
+	int ret;
+	u32 i;
+	struct amdgpu_vce_clock_voltage_dependency_table *table =
+		&adev->pm.dpm.dyn_state.vce_clock_voltage_dependency_table;
+	struct atom_clock_dividers dividers;
+
+	if (table == NULL || table->count == 0)
+		return 0;
+
+	pi->vce_level_count = 0;
+	for (i = 0; i < table->count; i++) {
+		if (pi->high_voltage_t &&
+		    pi->high_voltage_t < table->entries[i].v)
+			break;
+
+		pi->vce_level[i].Frequency = cpu_to_be32(table->entries[i].evclk);
+		pi->vce_level[i].MinVoltage = cpu_to_be16(table->entries[i].v);
+
+		pi->vce_level[i].ClkBypassCntl =
+			(u8)kv_get_clk_bypass(adev, table->entries[i].evclk);
+
+		ret = amdgpu_atombios_get_clock_dividers(adev, COMPUTE_ENGINE_PLL_PARAM,
+							 table->entries[i].evclk, false, &dividers);
+		if (ret)
+			return ret;
+		pi->vce_level[i].Divider = (u8)dividers.post_div;
+
+		pi->vce_level_count++;
+	}
+
+	ret = amdgpu_kv_copy_bytes_to_smc(adev,
+				   pi->dpm_table_start +
+				   offsetof(SMU7_Fusion_DpmTable, VceLevelCount),
+				   (u8 *)&pi->vce_level_count,
+				   sizeof(u8),
+				   pi->sram_end);
+	if (ret)
+		return ret;
+
+	pi->vce_interval = 1;
+
+	ret = amdgpu_kv_copy_bytes_to_smc(adev,
+				   pi->dpm_table_start +
+				   offsetof(SMU7_Fusion_DpmTable, VCEInterval),
+				   (u8 *)&pi->vce_interval,
+				   sizeof(u8),
+				   pi->sram_end);
+	if (ret)
+		return ret;
+
+	ret = amdgpu_kv_copy_bytes_to_smc(adev,
+				   pi->dpm_table_start +
+				   offsetof(SMU7_Fusion_DpmTable, VceLevel),
+				   (u8 *)&pi->vce_level,
+				   sizeof(SMU7_Fusion_ExtClkLevel) * SMU7_MAX_LEVELS_VCE,
+				   pi->sram_end);
+
+	return ret;
+}
+
+static int kv_populate_samu_table(struct amdgpu_device *adev)
+{
+	struct kv_power_info *pi = kv_get_pi(adev);
+	struct amdgpu_clock_voltage_dependency_table *table =
+		&adev->pm.dpm.dyn_state.samu_clock_voltage_dependency_table;
+	struct atom_clock_dividers dividers;
+	int ret;
+	u32 i;
+
+	if (table == NULL || table->count == 0)
+		return 0;
+
+	pi->samu_level_count = 0;
+	for (i = 0; i < table->count; i++) {
+		if (pi->high_voltage_t &&
+		    pi->high_voltage_t < table->entries[i].v)
+			break;
+
+		pi->samu_level[i].Frequency = cpu_to_be32(table->entries[i].clk);
+		pi->samu_level[i].MinVoltage = cpu_to_be16(table->entries[i].v);
+
+		pi->samu_level[i].ClkBypassCntl =
+			(u8)kv_get_clk_bypass(adev, table->entries[i].clk);
+
+		ret = amdgpu_atombios_get_clock_dividers(adev, COMPUTE_ENGINE_PLL_PARAM,
+							 table->entries[i].clk, false, &dividers);
+		if (ret)
+			return ret;
+		pi->samu_level[i].Divider = (u8)dividers.post_div;
+
+		pi->samu_level_count++;
+	}
+
+	ret = amdgpu_kv_copy_bytes_to_smc(adev,
+				   pi->dpm_table_start +
+				   offsetof(SMU7_Fusion_DpmTable, SamuLevelCount),
+				   (u8 *)&pi->samu_level_count,
+				   sizeof(u8),
+				   pi->sram_end);
+	if (ret)
+		return ret;
+
+	pi->samu_interval = 1;
+
+	ret = amdgpu_kv_copy_bytes_to_smc(adev,
+				   pi->dpm_table_start +
+				   offsetof(SMU7_Fusion_DpmTable, SAMUInterval),
+				   (u8 *)&pi->samu_interval,
+				   sizeof(u8),
+				   pi->sram_end);
+	if (ret)
+		return ret;
+
+	ret = amdgpu_kv_copy_bytes_to_smc(adev,
+				   pi->dpm_table_start +
+				   offsetof(SMU7_Fusion_DpmTable, SamuLevel),
+				   (u8 *)&pi->samu_level,
+				   sizeof(SMU7_Fusion_ExtClkLevel) * SMU7_MAX_LEVELS_SAMU,
+				   pi->sram_end);
+	if (ret)
+		return ret;
+
+	return ret;
+}
+
+
+static int kv_populate_acp_table(struct amdgpu_device *adev)
+{
+	struct kv_power_info *pi = kv_get_pi(adev);
+	struct amdgpu_clock_voltage_dependency_table *table =
+		&adev->pm.dpm.dyn_state.acp_clock_voltage_dependency_table;
+	struct atom_clock_dividers dividers;
+	int ret;
+	u32 i;
+
+	if (table == NULL || table->count == 0)
+		return 0;
+
+	pi->acp_level_count = 0;
+	for (i = 0; i < table->count; i++) {
+		pi->acp_level[i].Frequency = cpu_to_be32(table->entries[i].clk);
+		pi->acp_level[i].MinVoltage = cpu_to_be16(table->entries[i].v);
+
+		ret = amdgpu_atombios_get_clock_dividers(adev, COMPUTE_ENGINE_PLL_PARAM,
+							 table->entries[i].clk, false, &dividers);
+		if (ret)
+			return ret;
+		pi->acp_level[i].Divider = (u8)dividers.post_div;
+
+		pi->acp_level_count++;
+	}
+
+	ret = amdgpu_kv_copy_bytes_to_smc(adev,
+				   pi->dpm_table_start +
+				   offsetof(SMU7_Fusion_DpmTable, AcpLevelCount),
+				   (u8 *)&pi->acp_level_count,
+				   sizeof(u8),
+				   pi->sram_end);
+	if (ret)
+		return ret;
+
+	pi->acp_interval = 1;
+
+	ret = amdgpu_kv_copy_bytes_to_smc(adev,
+				   pi->dpm_table_start +
+				   offsetof(SMU7_Fusion_DpmTable, ACPInterval),
+				   (u8 *)&pi->acp_interval,
+				   sizeof(u8),
+				   pi->sram_end);
+	if (ret)
+		return ret;
+
+	ret = amdgpu_kv_copy_bytes_to_smc(adev,
+				   pi->dpm_table_start +
+				   offsetof(SMU7_Fusion_DpmTable, AcpLevel),
+				   (u8 *)&pi->acp_level,
+				   sizeof(SMU7_Fusion_ExtClkLevel) * SMU7_MAX_LEVELS_ACP,
+				   pi->sram_end);
+	if (ret)
+		return ret;
+
+	return ret;
+}
+
+static void kv_calculate_dfs_bypass_settings(struct amdgpu_device *adev)
+{
+	struct kv_power_info *pi = kv_get_pi(adev);
+	u32 i;
+	struct amdgpu_clock_voltage_dependency_table *table =
+		&adev->pm.dpm.dyn_state.vddc_dependency_on_sclk;
+
+	if (table && table->count) {
+		for (i = 0; i < pi->graphics_dpm_level_count; i++) {
+			if (pi->caps_enable_dfs_bypass) {
+				if (kv_get_clock_difference(table->entries[i].clk, 40000) < 200)
+					pi->graphics_level[i].ClkBypassCntl = 3;
+				else if (kv_get_clock_difference(table->entries[i].clk, 30000) < 200)
+					pi->graphics_level[i].ClkBypassCntl = 2;
+				else if (kv_get_clock_difference(table->entries[i].clk, 26600) < 200)
+					pi->graphics_level[i].ClkBypassCntl = 7;
+				else if (kv_get_clock_difference(table->entries[i].clk , 20000) < 200)
+					pi->graphics_level[i].ClkBypassCntl = 6;
+				else if (kv_get_clock_difference(table->entries[i].clk , 10000) < 200)
+					pi->graphics_level[i].ClkBypassCntl = 8;
+				else
+					pi->graphics_level[i].ClkBypassCntl = 0;
+			} else {
+				pi->graphics_level[i].ClkBypassCntl = 0;
+			}
+		}
+	} else {
+		struct sumo_sclk_voltage_mapping_table *table =
+			&pi->sys_info.sclk_voltage_mapping_table;
+		for (i = 0; i < pi->graphics_dpm_level_count; i++) {
+			if (pi->caps_enable_dfs_bypass) {
+				if (kv_get_clock_difference(table->entries[i].sclk_frequency, 40000) < 200)
+					pi->graphics_level[i].ClkBypassCntl = 3;
+				else if (kv_get_clock_difference(table->entries[i].sclk_frequency, 30000) < 200)
+					pi->graphics_level[i].ClkBypassCntl = 2;
+				else if (kv_get_clock_difference(table->entries[i].sclk_frequency, 26600) < 200)
+					pi->graphics_level[i].ClkBypassCntl = 7;
+				else if (kv_get_clock_difference(table->entries[i].sclk_frequency, 20000) < 200)
+					pi->graphics_level[i].ClkBypassCntl = 6;
+				else if (kv_get_clock_difference(table->entries[i].sclk_frequency, 10000) < 200)
+					pi->graphics_level[i].ClkBypassCntl = 8;
+				else
+					pi->graphics_level[i].ClkBypassCntl = 0;
+			} else {
+				pi->graphics_level[i].ClkBypassCntl = 0;
+			}
+		}
+	}
+}
+
+static int kv_enable_ulv(struct amdgpu_device *adev, bool enable)
+{
+	return amdgpu_kv_notify_message_to_smu(adev, enable ?
+					PPSMC_MSG_EnableULV : PPSMC_MSG_DisableULV);
+}
+
+static void kv_reset_acp_boot_level(struct amdgpu_device *adev)
+{
+	struct kv_power_info *pi = kv_get_pi(adev);
+
+	pi->acp_boot_level = 0xff;
+}
+
+static void kv_update_current_ps(struct amdgpu_device *adev,
+				 struct amdgpu_ps *rps)
+{
+	struct kv_ps *new_ps = kv_get_ps(rps);
+	struct kv_power_info *pi = kv_get_pi(adev);
+
+	pi->current_rps = *rps;
+	pi->current_ps = *new_ps;
+	pi->current_rps.ps_priv = &pi->current_ps;
+}
+
+static void kv_update_requested_ps(struct amdgpu_device *adev,
+				   struct amdgpu_ps *rps)
+{
+	struct kv_ps *new_ps = kv_get_ps(rps);
+	struct kv_power_info *pi = kv_get_pi(adev);
+
+	pi->requested_rps = *rps;
+	pi->requested_ps = *new_ps;
+	pi->requested_rps.ps_priv = &pi->requested_ps;
+}
+
+static void kv_dpm_enable_bapm(struct amdgpu_device *adev, bool enable)
+{
+	struct kv_power_info *pi = kv_get_pi(adev);
+	int ret;
+
+	if (pi->bapm_enable) {
+		ret = amdgpu_kv_smc_bapm_enable(adev, enable);
+		if (ret)
+			DRM_ERROR("amdgpu_kv_smc_bapm_enable failed\n");
+	}
+}
+
+static int kv_dpm_enable(struct amdgpu_device *adev)
+{
+	struct kv_power_info *pi = kv_get_pi(adev);
+	int ret;
+
+	ret = kv_process_firmware_header(adev);
+	if (ret) {
+		DRM_ERROR("kv_process_firmware_header failed\n");
+		return ret;
+	}
+	kv_init_fps_limits(adev);
+	kv_init_graphics_levels(adev);
+	ret = kv_program_bootup_state(adev);
+	if (ret) {
+		DRM_ERROR("kv_program_bootup_state failed\n");
+		return ret;
+	}
+	kv_calculate_dfs_bypass_settings(adev);
+	ret = kv_upload_dpm_settings(adev);
+	if (ret) {
+		DRM_ERROR("kv_upload_dpm_settings failed\n");
+		return ret;
+	}
+	ret = kv_populate_uvd_table(adev);
+	if (ret) {
+		DRM_ERROR("kv_populate_uvd_table failed\n");
+		return ret;
+	}
+	ret = kv_populate_vce_table(adev);
+	if (ret) {
+		DRM_ERROR("kv_populate_vce_table failed\n");
+		return ret;
+	}
+	ret = kv_populate_samu_table(adev);
+	if (ret) {
+		DRM_ERROR("kv_populate_samu_table failed\n");
+		return ret;
+	}
+	ret = kv_populate_acp_table(adev);
+	if (ret) {
+		DRM_ERROR("kv_populate_acp_table failed\n");
+		return ret;
+	}
+	kv_program_vc(adev);
+#if 0
+	kv_initialize_hardware_cac_manager(adev);
+#endif
+	kv_start_am(adev);
+	if (pi->enable_auto_thermal_throttling) {
+		ret = kv_enable_auto_thermal_throttling(adev);
+		if (ret) {
+			DRM_ERROR("kv_enable_auto_thermal_throttling failed\n");
+			return ret;
+		}
+	}
+	ret = kv_enable_dpm_voltage_scaling(adev);
+	if (ret) {
+		DRM_ERROR("kv_enable_dpm_voltage_scaling failed\n");
+		return ret;
+	}
+	ret = kv_set_dpm_interval(adev);
+	if (ret) {
+		DRM_ERROR("kv_set_dpm_interval failed\n");
+		return ret;
+	}
+	ret = kv_set_dpm_boot_state(adev);
+	if (ret) {
+		DRM_ERROR("kv_set_dpm_boot_state failed\n");
+		return ret;
+	}
+	ret = kv_enable_ulv(adev, true);
+	if (ret) {
+		DRM_ERROR("kv_enable_ulv failed\n");
+		return ret;
+	}
+	kv_start_dpm(adev);
+	ret = kv_enable_didt(adev, true);
+	if (ret) {
+		DRM_ERROR("kv_enable_didt failed\n");
+		return ret;
+	}
+	ret = kv_enable_smc_cac(adev, true);
+	if (ret) {
+		DRM_ERROR("kv_enable_smc_cac failed\n");
+		return ret;
+	}
+
+	kv_reset_acp_boot_level(adev);
+
+	ret = amdgpu_kv_smc_bapm_enable(adev, false);
+	if (ret) {
+		DRM_ERROR("amdgpu_kv_smc_bapm_enable failed\n");
+		return ret;
+	}
+
+	kv_update_current_ps(adev, adev->pm.dpm.boot_ps);
+
+	if (adev->irq.installed &&
+	    amdgpu_is_internal_thermal_sensor(adev->pm.int_thermal_type)) {
+		ret = kv_set_thermal_temperature_range(adev, KV_TEMP_RANGE_MIN, KV_TEMP_RANGE_MAX);
+		if (ret) {
+			DRM_ERROR("kv_set_thermal_temperature_range failed\n");
+			return ret;
+		}
+		amdgpu_irq_get(adev, &adev->pm.dpm.thermal.irq,
+			       AMDGPU_THERMAL_IRQ_LOW_TO_HIGH);
+		amdgpu_irq_get(adev, &adev->pm.dpm.thermal.irq,
+			       AMDGPU_THERMAL_IRQ_HIGH_TO_LOW);
+	}
+
+	return ret;
+}
+
+static void kv_dpm_disable(struct amdgpu_device *adev)
+{
+	amdgpu_irq_put(adev, &adev->pm.dpm.thermal.irq,
+		       AMDGPU_THERMAL_IRQ_LOW_TO_HIGH);
+	amdgpu_irq_put(adev, &adev->pm.dpm.thermal.irq,
+		       AMDGPU_THERMAL_IRQ_HIGH_TO_LOW);
+
+	amdgpu_kv_smc_bapm_enable(adev, false);
+
+	if (adev->asic_type == CHIP_MULLINS)
+		kv_enable_nb_dpm(adev, false);
+
+	/* powerup blocks */
+	kv_dpm_powergate_acp(adev, false);
+	kv_dpm_powergate_samu(adev, false);
+	kv_dpm_powergate_vce(adev, false);
+	kv_dpm_powergate_uvd(adev, false);
+
+	kv_enable_smc_cac(adev, false);
+	kv_enable_didt(adev, false);
+	kv_clear_vc(adev);
+	kv_stop_dpm(adev);
+	kv_enable_ulv(adev, false);
+	kv_reset_am(adev);
+
+	kv_update_current_ps(adev, adev->pm.dpm.boot_ps);
+}
+
+#if 0
+static int kv_write_smc_soft_register(struct amdgpu_device *adev,
+				      u16 reg_offset, u32 value)
+{
+	struct kv_power_info *pi = kv_get_pi(adev);
+
+	return amdgpu_kv_copy_bytes_to_smc(adev, pi->soft_regs_start + reg_offset,
+				    (u8 *)&value, sizeof(u16), pi->sram_end);
+}
+
+static int kv_read_smc_soft_register(struct amdgpu_device *adev,
+				     u16 reg_offset, u32 *value)
+{
+	struct kv_power_info *pi = kv_get_pi(adev);
+
+	return amdgpu_kv_read_smc_sram_dword(adev, pi->soft_regs_start + reg_offset,
+				      value, pi->sram_end);
+}
+#endif
+
+static void kv_init_sclk_t(struct amdgpu_device *adev)
+{
+	struct kv_power_info *pi = kv_get_pi(adev);
+
+	pi->low_sclk_interrupt_t = 0;
+}
+
+static int kv_init_fps_limits(struct amdgpu_device *adev)
+{
+	struct kv_power_info *pi = kv_get_pi(adev);
+	int ret = 0;
+
+	if (pi->caps_fps) {
+		u16 tmp;
+
+		tmp = 45;
+		pi->fps_high_t = cpu_to_be16(tmp);
+		ret = amdgpu_kv_copy_bytes_to_smc(adev,
+					   pi->dpm_table_start +
+					   offsetof(SMU7_Fusion_DpmTable, FpsHighT),
+					   (u8 *)&pi->fps_high_t,
+					   sizeof(u16), pi->sram_end);
+
+		tmp = 30;
+		pi->fps_low_t = cpu_to_be16(tmp);
+
+		ret = amdgpu_kv_copy_bytes_to_smc(adev,
+					   pi->dpm_table_start +
+					   offsetof(SMU7_Fusion_DpmTable, FpsLowT),
+					   (u8 *)&pi->fps_low_t,
+					   sizeof(u16), pi->sram_end);
+
+	}
+	return ret;
+}
+
+static void kv_init_powergate_state(struct amdgpu_device *adev)
+{
+	struct kv_power_info *pi = kv_get_pi(adev);
+
+	pi->uvd_power_gated = false;
+	pi->vce_power_gated = false;
+	pi->samu_power_gated = false;
+	pi->acp_power_gated = false;
+
+}
+
+static int kv_enable_uvd_dpm(struct amdgpu_device *adev, bool enable)
+{
+	return amdgpu_kv_notify_message_to_smu(adev, enable ?
+					PPSMC_MSG_UVDDPM_Enable : PPSMC_MSG_UVDDPM_Disable);
+}
+
+static int kv_enable_vce_dpm(struct amdgpu_device *adev, bool enable)
+{
+	return amdgpu_kv_notify_message_to_smu(adev, enable ?
+					PPSMC_MSG_VCEDPM_Enable : PPSMC_MSG_VCEDPM_Disable);
+}
+
+static int kv_enable_samu_dpm(struct amdgpu_device *adev, bool enable)
+{
+	return amdgpu_kv_notify_message_to_smu(adev, enable ?
+					PPSMC_MSG_SAMUDPM_Enable : PPSMC_MSG_SAMUDPM_Disable);
+}
+
+static int kv_enable_acp_dpm(struct amdgpu_device *adev, bool enable)
+{
+	return amdgpu_kv_notify_message_to_smu(adev, enable ?
+					PPSMC_MSG_ACPDPM_Enable : PPSMC_MSG_ACPDPM_Disable);
+}
+
+static int kv_update_uvd_dpm(struct amdgpu_device *adev, bool gate)
+{
+	struct kv_power_info *pi = kv_get_pi(adev);
+	struct amdgpu_uvd_clock_voltage_dependency_table *table =
+		&adev->pm.dpm.dyn_state.uvd_clock_voltage_dependency_table;
+	int ret;
+	u32 mask;
+
+	if (!gate) {
+		if (table->count)
+			pi->uvd_boot_level = table->count - 1;
+		else
+			pi->uvd_boot_level = 0;
+
+		if (!pi->caps_uvd_dpm || pi->caps_stable_p_state) {
+			mask = 1 << pi->uvd_boot_level;
+		} else {
+			mask = 0x1f;
+		}
+
+		ret = amdgpu_kv_copy_bytes_to_smc(adev,
+					   pi->dpm_table_start +
+					   offsetof(SMU7_Fusion_DpmTable, UvdBootLevel),
+					   (uint8_t *)&pi->uvd_boot_level,
+					   sizeof(u8), pi->sram_end);
+		if (ret)
+			return ret;
+
+		amdgpu_kv_send_msg_to_smc_with_parameter(adev,
+						  PPSMC_MSG_UVDDPM_SetEnabledMask,
+						  mask);
+	}
+
+	return kv_enable_uvd_dpm(adev, !gate);
+}
+
+static u8 kv_get_vce_boot_level(struct amdgpu_device *adev, u32 evclk)
+{
+	u8 i;
+	struct amdgpu_vce_clock_voltage_dependency_table *table =
+		&adev->pm.dpm.dyn_state.vce_clock_voltage_dependency_table;
+
+	for (i = 0; i < table->count; i++) {
+		if (table->entries[i].evclk >= evclk)
+			break;
+	}
+
+	return i;
+}
+
+static int kv_update_vce_dpm(struct amdgpu_device *adev,
+			     struct amdgpu_ps *amdgpu_new_state,
+			     struct amdgpu_ps *amdgpu_current_state)
+{
+	struct kv_power_info *pi = kv_get_pi(adev);
+	struct amdgpu_vce_clock_voltage_dependency_table *table =
+		&adev->pm.dpm.dyn_state.vce_clock_voltage_dependency_table;
+	int ret;
+
+	if (amdgpu_new_state->evclk > 0 && amdgpu_current_state->evclk == 0) {
+		kv_dpm_powergate_vce(adev, false);
+		/* turn the clocks on when encoding */
+		ret = amdgpu_set_clockgating_state(adev, AMDGPU_IP_BLOCK_TYPE_VCE,
+						    AMDGPU_CG_STATE_UNGATE);
+		if (ret)
+			return ret;
+		if (pi->caps_stable_p_state)
+			pi->vce_boot_level = table->count - 1;
+		else
+			pi->vce_boot_level = kv_get_vce_boot_level(adev, amdgpu_new_state->evclk);
+
+		ret = amdgpu_kv_copy_bytes_to_smc(adev,
+					   pi->dpm_table_start +
+					   offsetof(SMU7_Fusion_DpmTable, VceBootLevel),
+					   (u8 *)&pi->vce_boot_level,
+					   sizeof(u8),
+					   pi->sram_end);
+		if (ret)
+			return ret;
+
+		if (pi->caps_stable_p_state)
+			amdgpu_kv_send_msg_to_smc_with_parameter(adev,
+							  PPSMC_MSG_VCEDPM_SetEnabledMask,
+							  (1 << pi->vce_boot_level));
+
+		kv_enable_vce_dpm(adev, true);
+	} else if (amdgpu_new_state->evclk == 0 && amdgpu_current_state->evclk > 0) {
+		kv_enable_vce_dpm(adev, false);
+		/* turn the clocks off when not encoding */
+		ret = amdgpu_set_clockgating_state(adev, AMDGPU_IP_BLOCK_TYPE_VCE,
+						    AMDGPU_CG_STATE_GATE);
+		if (ret)
+			return ret;
+		kv_dpm_powergate_vce(adev, true);
+	}
+
+	return 0;
+}
+
+static int kv_update_samu_dpm(struct amdgpu_device *adev, bool gate)
+{
+	struct kv_power_info *pi = kv_get_pi(adev);
+	struct amdgpu_clock_voltage_dependency_table *table =
+		&adev->pm.dpm.dyn_state.samu_clock_voltage_dependency_table;
+	int ret;
+
+	if (!gate) {
+		if (pi->caps_stable_p_state)
+			pi->samu_boot_level = table->count - 1;
+		else
+			pi->samu_boot_level = 0;
+
+		ret = amdgpu_kv_copy_bytes_to_smc(adev,
+					   pi->dpm_table_start +
+					   offsetof(SMU7_Fusion_DpmTable, SamuBootLevel),
+					   (u8 *)&pi->samu_boot_level,
+					   sizeof(u8),
+					   pi->sram_end);
+		if (ret)
+			return ret;
+
+		if (pi->caps_stable_p_state)
+			amdgpu_kv_send_msg_to_smc_with_parameter(adev,
+							  PPSMC_MSG_SAMUDPM_SetEnabledMask,
+							  (1 << pi->samu_boot_level));
+	}
+
+	return kv_enable_samu_dpm(adev, !gate);
+}
+
+static u8 kv_get_acp_boot_level(struct amdgpu_device *adev)
+{
+	u8 i;
+	struct amdgpu_clock_voltage_dependency_table *table =
+		&adev->pm.dpm.dyn_state.acp_clock_voltage_dependency_table;
+
+	for (i = 0; i < table->count; i++) {
+		if (table->entries[i].clk >= 0) /* XXX */
+			break;
+	}
+
+	if (i >= table->count)
+		i = table->count - 1;
+
+	return i;
+}
+
+static void kv_update_acp_boot_level(struct amdgpu_device *adev)
+{
+	struct kv_power_info *pi = kv_get_pi(adev);
+	u8 acp_boot_level;
+
+	if (!pi->caps_stable_p_state) {
+		acp_boot_level = kv_get_acp_boot_level(adev);
+		if (acp_boot_level != pi->acp_boot_level) {
+			pi->acp_boot_level = acp_boot_level;
+			amdgpu_kv_send_msg_to_smc_with_parameter(adev,
+							  PPSMC_MSG_ACPDPM_SetEnabledMask,
+							  (1 << pi->acp_boot_level));
+		}
+	}
+}
+
+static int kv_update_acp_dpm(struct amdgpu_device *adev, bool gate)
+{
+	struct kv_power_info *pi = kv_get_pi(adev);
+	struct amdgpu_clock_voltage_dependency_table *table =
+		&adev->pm.dpm.dyn_state.acp_clock_voltage_dependency_table;
+	int ret;
+
+	if (!gate) {
+		if (pi->caps_stable_p_state)
+			pi->acp_boot_level = table->count - 1;
+		else
+			pi->acp_boot_level = kv_get_acp_boot_level(adev);
+
+		ret = amdgpu_kv_copy_bytes_to_smc(adev,
+					   pi->dpm_table_start +
+					   offsetof(SMU7_Fusion_DpmTable, AcpBootLevel),
+					   (u8 *)&pi->acp_boot_level,
+					   sizeof(u8),
+					   pi->sram_end);
+		if (ret)
+			return ret;
+
+		if (pi->caps_stable_p_state)
+			amdgpu_kv_send_msg_to_smc_with_parameter(adev,
+							  PPSMC_MSG_ACPDPM_SetEnabledMask,
+							  (1 << pi->acp_boot_level));
+	}
+
+	return kv_enable_acp_dpm(adev, !gate);
+}
+
+static void kv_dpm_powergate_uvd(struct amdgpu_device *adev, bool gate)
+{
+	struct kv_power_info *pi = kv_get_pi(adev);
+	int ret;
+
+	if (pi->uvd_power_gated == gate)
+		return;
+
+	pi->uvd_power_gated = gate;
+
+	if (gate) {
+		if (pi->caps_uvd_pg) {
+			/* disable clockgating so we can properly shut down the block */
+			ret = amdgpu_set_clockgating_state(adev, AMDGPU_IP_BLOCK_TYPE_UVD,
+							    AMDGPU_CG_STATE_UNGATE);
+			/* shutdown the UVD block */
+			ret = amdgpu_set_powergating_state(adev, AMDGPU_IP_BLOCK_TYPE_UVD,
+							    AMDGPU_PG_STATE_GATE);
+			/* XXX: check for errors */
+		}
+		kv_update_uvd_dpm(adev, gate);
+		if (pi->caps_uvd_pg)
+			/* power off the UVD block */
+			amdgpu_kv_notify_message_to_smu(adev, PPSMC_MSG_UVDPowerOFF);
+	} else {
+		if (pi->caps_uvd_pg) {
+			/* power on the UVD block */
+			amdgpu_kv_notify_message_to_smu(adev, PPSMC_MSG_UVDPowerON);
+			/* re-init the UVD block */
+			ret = amdgpu_set_powergating_state(adev, AMDGPU_IP_BLOCK_TYPE_UVD,
+							    AMDGPU_PG_STATE_UNGATE);
+			/* enable clockgating. hw will dynamically gate/ungate clocks on the fly */
+			ret = amdgpu_set_clockgating_state(adev, AMDGPU_IP_BLOCK_TYPE_UVD,
+							    AMDGPU_CG_STATE_GATE);
+			/* XXX: check for errors */
+		}
+		kv_update_uvd_dpm(adev, gate);
+	}
+}
+
+static void kv_dpm_powergate_vce(struct amdgpu_device *adev, bool gate)
+{
+	struct kv_power_info *pi = kv_get_pi(adev);
+	int ret;
+
+	if (pi->vce_power_gated == gate)
+		return;
+
+	pi->vce_power_gated = gate;
+
+	if (gate) {
+		if (pi->caps_vce_pg) {
+			/* shutdown the VCE block */
+			ret = amdgpu_set_powergating_state(adev, AMDGPU_IP_BLOCK_TYPE_VCE,
+							    AMDGPU_PG_STATE_GATE);
+			/* XXX: check for errors */
+			/* power off the VCE block */
+			amdgpu_kv_notify_message_to_smu(adev, PPSMC_MSG_VCEPowerOFF);
+		}
+	} else {
+		if (pi->caps_vce_pg) {
+			/* power on the VCE block */
+			amdgpu_kv_notify_message_to_smu(adev, PPSMC_MSG_VCEPowerON);
+			/* re-init the VCE block */
+			ret = amdgpu_set_powergating_state(adev, AMDGPU_IP_BLOCK_TYPE_VCE,
+							    AMDGPU_PG_STATE_UNGATE);
+			/* XXX: check for errors */
+		}
+	}
+}
+
+static void kv_dpm_powergate_samu(struct amdgpu_device *adev, bool gate)
+{
+	struct kv_power_info *pi = kv_get_pi(adev);
+
+	if (pi->samu_power_gated == gate)
+		return;
+
+	pi->samu_power_gated = gate;
+
+	if (gate) {
+		kv_update_samu_dpm(adev, true);
+		if (pi->caps_samu_pg)
+			amdgpu_kv_notify_message_to_smu(adev, PPSMC_MSG_SAMPowerOFF);
+	} else {
+		if (pi->caps_samu_pg)
+			amdgpu_kv_notify_message_to_smu(adev, PPSMC_MSG_SAMPowerON);
+		kv_update_samu_dpm(adev, false);
+	}
+}
+
+static void kv_dpm_powergate_acp(struct amdgpu_device *adev, bool gate)
+{
+	struct kv_power_info *pi = kv_get_pi(adev);
+
+	if (pi->acp_power_gated == gate)
+		return;
+
+	if (adev->asic_type == CHIP_KABINI || adev->asic_type == CHIP_MULLINS)
+		return;
+
+	pi->acp_power_gated = gate;
+
+	if (gate) {
+		kv_update_acp_dpm(adev, true);
+		if (pi->caps_acp_pg)
+			amdgpu_kv_notify_message_to_smu(adev, PPSMC_MSG_ACPPowerOFF);
+	} else {
+		if (pi->caps_acp_pg)
+			amdgpu_kv_notify_message_to_smu(adev, PPSMC_MSG_ACPPowerON);
+		kv_update_acp_dpm(adev, false);
+	}
+}
+
+static void kv_set_valid_clock_range(struct amdgpu_device *adev,
+				     struct amdgpu_ps *new_rps)
+{
+	struct kv_ps *new_ps = kv_get_ps(new_rps);
+	struct kv_power_info *pi = kv_get_pi(adev);
+	u32 i;
+	struct amdgpu_clock_voltage_dependency_table *table =
+		&adev->pm.dpm.dyn_state.vddc_dependency_on_sclk;
+
+	if (table && table->count) {
+		for (i = 0; i < pi->graphics_dpm_level_count; i++) {
+			if ((table->entries[i].clk >= new_ps->levels[0].sclk) ||
+			    (i == (pi->graphics_dpm_level_count - 1))) {
+				pi->lowest_valid = i;
+				break;
+			}
+		}
+
+		for (i = pi->graphics_dpm_level_count - 1; i > 0; i--) {
+			if (table->entries[i].clk <= new_ps->levels[new_ps->num_levels - 1].sclk)
+				break;
+		}
+		pi->highest_valid = i;
+
+		if (pi->lowest_valid > pi->highest_valid) {
+			if ((new_ps->levels[0].sclk - table->entries[pi->highest_valid].clk) >
+			    (table->entries[pi->lowest_valid].clk - new_ps->levels[new_ps->num_levels - 1].sclk))
+				pi->highest_valid = pi->lowest_valid;
+			else
+				pi->lowest_valid =  pi->highest_valid;
+		}
+	} else {
+		struct sumo_sclk_voltage_mapping_table *table =
+			&pi->sys_info.sclk_voltage_mapping_table;
+
+		for (i = 0; i < (int)pi->graphics_dpm_level_count; i++) {
+			if (table->entries[i].sclk_frequency >= new_ps->levels[0].sclk ||
+			    i == (int)(pi->graphics_dpm_level_count - 1)) {
+				pi->lowest_valid = i;
+				break;
+			}
+		}
+
+		for (i = pi->graphics_dpm_level_count - 1; i > 0; i--) {
+			if (table->entries[i].sclk_frequency <=
+			    new_ps->levels[new_ps->num_levels - 1].sclk)
+				break;
+		}
+		pi->highest_valid = i;
+
+		if (pi->lowest_valid > pi->highest_valid) {
+			if ((new_ps->levels[0].sclk -
+			     table->entries[pi->highest_valid].sclk_frequency) >
+			    (table->entries[pi->lowest_valid].sclk_frequency -
+			     new_ps->levels[new_ps->num_levels -1].sclk))
+				pi->highest_valid = pi->lowest_valid;
+			else
+				pi->lowest_valid =  pi->highest_valid;
+		}
+	}
+}
+
+static int kv_update_dfs_bypass_settings(struct amdgpu_device *adev,
+					 struct amdgpu_ps *new_rps)
+{
+	struct kv_ps *new_ps = kv_get_ps(new_rps);
+	struct kv_power_info *pi = kv_get_pi(adev);
+	int ret = 0;
+	u8 clk_bypass_cntl;
+
+	if (pi->caps_enable_dfs_bypass) {
+		clk_bypass_cntl = new_ps->need_dfs_bypass ?
+			pi->graphics_level[pi->graphics_boot_level].ClkBypassCntl : 0;
+		ret = amdgpu_kv_copy_bytes_to_smc(adev,
+					   (pi->dpm_table_start +
+					    offsetof(SMU7_Fusion_DpmTable, GraphicsLevel) +
+					    (pi->graphics_boot_level * sizeof(SMU7_Fusion_GraphicsLevel)) +
+					    offsetof(SMU7_Fusion_GraphicsLevel, ClkBypassCntl)),
+					   &clk_bypass_cntl,
+					   sizeof(u8), pi->sram_end);
+	}
+
+	return ret;
+}
+
+static int kv_enable_nb_dpm(struct amdgpu_device *adev,
+			    bool enable)
+{
+	struct kv_power_info *pi = kv_get_pi(adev);
+	int ret = 0;
+
+	if (enable) {
+		if (pi->enable_nb_dpm && !pi->nb_dpm_enabled) {
+			ret = amdgpu_kv_notify_message_to_smu(adev, PPSMC_MSG_NBDPM_Enable);
+			if (ret == 0)
+				pi->nb_dpm_enabled = true;
+		}
+	} else {
+		if (pi->enable_nb_dpm && pi->nb_dpm_enabled) {
+			ret = amdgpu_kv_notify_message_to_smu(adev, PPSMC_MSG_NBDPM_Disable);
+			if (ret == 0)
+				pi->nb_dpm_enabled = false;
+		}
+	}
+
+	return ret;
+}
+
+static int kv_dpm_force_performance_level(struct amdgpu_device *adev,
+					  enum amdgpu_dpm_forced_level level)
+{
+	int ret;
+
+	if (level == AMDGPU_DPM_FORCED_LEVEL_HIGH) {
+		ret = kv_force_dpm_highest(adev);
+		if (ret)
+			return ret;
+	} else if (level == AMDGPU_DPM_FORCED_LEVEL_LOW) {
+		ret = kv_force_dpm_lowest(adev);
+		if (ret)
+			return ret;
+	} else if (level == AMDGPU_DPM_FORCED_LEVEL_AUTO) {
+		ret = kv_unforce_levels(adev);
+		if (ret)
+			return ret;
+	}
+
+	adev->pm.dpm.forced_level = level;
+
+	return 0;
+}
+
+static int kv_dpm_pre_set_power_state(struct amdgpu_device *adev)
+{
+	struct kv_power_info *pi = kv_get_pi(adev);
+	struct amdgpu_ps requested_ps = *adev->pm.dpm.requested_ps;
+	struct amdgpu_ps *new_ps = &requested_ps;
+
+	kv_update_requested_ps(adev, new_ps);
+
+	kv_apply_state_adjust_rules(adev,
+				    &pi->requested_rps,
+				    &pi->current_rps);
+
+	return 0;
+}
+
+static int kv_dpm_set_power_state(struct amdgpu_device *adev)
+{
+	struct kv_power_info *pi = kv_get_pi(adev);
+	struct amdgpu_ps *new_ps = &pi->requested_rps;
+	struct amdgpu_ps *old_ps = &pi->current_rps;
+	int ret;
+
+	if (pi->bapm_enable) {
+		ret = amdgpu_kv_smc_bapm_enable(adev, adev->pm.dpm.ac_power);
+		if (ret) {
+			DRM_ERROR("amdgpu_kv_smc_bapm_enable failed\n");
+			return ret;
+		}
+	}
+
+	if (adev->asic_type == CHIP_KABINI || adev->asic_type == CHIP_MULLINS) {
+		if (pi->enable_dpm) {
+			kv_set_valid_clock_range(adev, new_ps);
+			kv_update_dfs_bypass_settings(adev, new_ps);
+			ret = kv_calculate_ds_divider(adev);
+			if (ret) {
+				DRM_ERROR("kv_calculate_ds_divider failed\n");
+				return ret;
+			}
+			kv_calculate_nbps_level_settings(adev);
+			kv_calculate_dpm_settings(adev);
+			kv_force_lowest_valid(adev);
+			kv_enable_new_levels(adev);
+			kv_upload_dpm_settings(adev);
+			kv_program_nbps_index_settings(adev, new_ps);
+			kv_unforce_levels(adev);
+			kv_set_enabled_levels(adev);
+			kv_force_lowest_valid(adev);
+			kv_unforce_levels(adev);
+
+			ret = kv_update_vce_dpm(adev, new_ps, old_ps);
+			if (ret) {
+				DRM_ERROR("kv_update_vce_dpm failed\n");
+				return ret;
+			}
+			kv_update_sclk_t(adev);
+			if (adev->asic_type == CHIP_MULLINS)
+				kv_enable_nb_dpm(adev, true);
+		}
+	} else {
+		if (pi->enable_dpm) {
+			kv_set_valid_clock_range(adev, new_ps);
+			kv_update_dfs_bypass_settings(adev, new_ps);
+			ret = kv_calculate_ds_divider(adev);
+			if (ret) {
+				DRM_ERROR("kv_calculate_ds_divider failed\n");
+				return ret;
+			}
+			kv_calculate_nbps_level_settings(adev);
+			kv_calculate_dpm_settings(adev);
+			kv_freeze_sclk_dpm(adev, true);
+			kv_upload_dpm_settings(adev);
+			kv_program_nbps_index_settings(adev, new_ps);
+			kv_freeze_sclk_dpm(adev, false);
+			kv_set_enabled_levels(adev);
+			ret = kv_update_vce_dpm(adev, new_ps, old_ps);
+			if (ret) {
+				DRM_ERROR("kv_update_vce_dpm failed\n");
+				return ret;
+			}
+			kv_update_acp_boot_level(adev);
+			kv_update_sclk_t(adev);
+			kv_enable_nb_dpm(adev, true);
+		}
+	}
+
+	return 0;
+}
+
+static void kv_dpm_post_set_power_state(struct amdgpu_device *adev)
+{
+	struct kv_power_info *pi = kv_get_pi(adev);
+	struct amdgpu_ps *new_ps = &pi->requested_rps;
+
+	kv_update_current_ps(adev, new_ps);
+}
+
+static void kv_dpm_setup_asic(struct amdgpu_device *adev)
+{
+	sumo_take_smu_control(adev, true);
+	kv_init_powergate_state(adev);
+	kv_init_sclk_t(adev);
+}
+
+#if 0
+static void kv_dpm_reset_asic(struct amdgpu_device *adev)
+{
+	struct kv_power_info *pi = kv_get_pi(adev);
+
+	if (adev->asic_type == CHIP_KABINI || adev->asic_type == CHIP_MULLINS) {
+		kv_force_lowest_valid(adev);
+		kv_init_graphics_levels(adev);
+		kv_program_bootup_state(adev);
+		kv_upload_dpm_settings(adev);
+		kv_force_lowest_valid(adev);
+		kv_unforce_levels(adev);
+	} else {
+		kv_init_graphics_levels(adev);
+		kv_program_bootup_state(adev);
+		kv_freeze_sclk_dpm(adev, true);
+		kv_upload_dpm_settings(adev);
+		kv_freeze_sclk_dpm(adev, false);
+		kv_set_enabled_level(adev, pi->graphics_boot_level);
+	}
+}
+#endif
+
+static void kv_construct_max_power_limits_table(struct amdgpu_device *adev,
+						struct amdgpu_clock_and_voltage_limits *table)
+{
+	struct kv_power_info *pi = kv_get_pi(adev);
+
+	if (pi->sys_info.sclk_voltage_mapping_table.num_max_dpm_entries > 0) {
+		int idx = pi->sys_info.sclk_voltage_mapping_table.num_max_dpm_entries - 1;
+		table->sclk =
+			pi->sys_info.sclk_voltage_mapping_table.entries[idx].sclk_frequency;
+		table->vddc =
+			kv_convert_2bit_index_to_voltage(adev,
+							 pi->sys_info.sclk_voltage_mapping_table.entries[idx].vid_2bit);
+	}
+
+	table->mclk = pi->sys_info.nbp_memory_clock[0];
+}
+
+static void kv_patch_voltage_values(struct amdgpu_device *adev)
+{
+	int i;
+	struct amdgpu_uvd_clock_voltage_dependency_table *uvd_table =
+		&adev->pm.dpm.dyn_state.uvd_clock_voltage_dependency_table;
+	struct amdgpu_vce_clock_voltage_dependency_table *vce_table =
+		&adev->pm.dpm.dyn_state.vce_clock_voltage_dependency_table;
+	struct amdgpu_clock_voltage_dependency_table *samu_table =
+		&adev->pm.dpm.dyn_state.samu_clock_voltage_dependency_table;
+	struct amdgpu_clock_voltage_dependency_table *acp_table =
+		&adev->pm.dpm.dyn_state.acp_clock_voltage_dependency_table;
+
+	if (uvd_table->count) {
+		for (i = 0; i < uvd_table->count; i++)
+			uvd_table->entries[i].v =
+				kv_convert_8bit_index_to_voltage(adev,
+								 uvd_table->entries[i].v);
+	}
+
+	if (vce_table->count) {
+		for (i = 0; i < vce_table->count; i++)
+			vce_table->entries[i].v =
+				kv_convert_8bit_index_to_voltage(adev,
+								 vce_table->entries[i].v);
+	}
+
+	if (samu_table->count) {
+		for (i = 0; i < samu_table->count; i++)
+			samu_table->entries[i].v =
+				kv_convert_8bit_index_to_voltage(adev,
+								 samu_table->entries[i].v);
+	}
+
+	if (acp_table->count) {
+		for (i = 0; i < acp_table->count; i++)
+			acp_table->entries[i].v =
+				kv_convert_8bit_index_to_voltage(adev,
+								 acp_table->entries[i].v);
+	}
+
+}
+
+static void kv_construct_boot_state(struct amdgpu_device *adev)
+{
+	struct kv_power_info *pi = kv_get_pi(adev);
+
+	pi->boot_pl.sclk = pi->sys_info.bootup_sclk;
+	pi->boot_pl.vddc_index = pi->sys_info.bootup_nb_voltage_index;
+	pi->boot_pl.ds_divider_index = 0;
+	pi->boot_pl.ss_divider_index = 0;
+	pi->boot_pl.allow_gnb_slow = 1;
+	pi->boot_pl.force_nbp_state = 0;
+	pi->boot_pl.display_wm = 0;
+	pi->boot_pl.vce_wm = 0;
+}
+
+static int kv_force_dpm_highest(struct amdgpu_device *adev)
+{
+	int ret;
+	u32 enable_mask, i;
+
+	ret = amdgpu_kv_dpm_get_enable_mask(adev, &enable_mask);
+	if (ret)
+		return ret;
+
+	for (i = SMU7_MAX_LEVELS_GRAPHICS - 1; i > 0; i--) {
+		if (enable_mask & (1 << i))
+			break;
+	}
+
+	if (adev->asic_type == CHIP_KABINI || adev->asic_type == CHIP_MULLINS)
+		return amdgpu_kv_send_msg_to_smc_with_parameter(adev, PPSMC_MSG_DPM_ForceState, i);
+	else
+		return kv_set_enabled_level(adev, i);
+}
+
+static int kv_force_dpm_lowest(struct amdgpu_device *adev)
+{
+	int ret;
+	u32 enable_mask, i;
+
+	ret = amdgpu_kv_dpm_get_enable_mask(adev, &enable_mask);
+	if (ret)
+		return ret;
+
+	for (i = 0; i < SMU7_MAX_LEVELS_GRAPHICS; i++) {
+		if (enable_mask & (1 << i))
+			break;
+	}
+
+	if (adev->asic_type == CHIP_KABINI || adev->asic_type == CHIP_MULLINS)
+		return amdgpu_kv_send_msg_to_smc_with_parameter(adev, PPSMC_MSG_DPM_ForceState, i);
+	else
+		return kv_set_enabled_level(adev, i);
+}
+
+static u8 kv_get_sleep_divider_id_from_clock(struct amdgpu_device *adev,
+					     u32 sclk, u32 min_sclk_in_sr)
+{
+	struct kv_power_info *pi = kv_get_pi(adev);
+	u32 i;
+	u32 temp;
+	u32 min = (min_sclk_in_sr > KV_MINIMUM_ENGINE_CLOCK) ?
+		min_sclk_in_sr : KV_MINIMUM_ENGINE_CLOCK;
+
+	if (sclk < min)
+		return 0;
+
+	if (!pi->caps_sclk_ds)
+		return 0;
+
+	for (i = KV_MAX_DEEPSLEEP_DIVIDER_ID; i > 0; i--) {
+		temp = sclk / sumo_get_sleep_divider_from_id(i);
+		if (temp >= min)
+			break;
+	}
+
+	return (u8)i;
+}
+
+static int kv_get_high_voltage_limit(struct amdgpu_device *adev, int *limit)
+{
+	struct kv_power_info *pi = kv_get_pi(adev);
+	struct amdgpu_clock_voltage_dependency_table *table =
+		&adev->pm.dpm.dyn_state.vddc_dependency_on_sclk;
+	int i;
+
+	if (table && table->count) {
+		for (i = table->count - 1; i >= 0; i--) {
+			if (pi->high_voltage_t &&
+			    (kv_convert_8bit_index_to_voltage(adev, table->entries[i].v) <=
+			     pi->high_voltage_t)) {
+				*limit = i;
+				return 0;
+			}
+		}
+	} else {
+		struct sumo_sclk_voltage_mapping_table *table =
+			&pi->sys_info.sclk_voltage_mapping_table;
+
+		for (i = table->num_max_dpm_entries - 1; i >= 0; i--) {
+			if (pi->high_voltage_t &&
+			    (kv_convert_2bit_index_to_voltage(adev, table->entries[i].vid_2bit) <=
+			     pi->high_voltage_t)) {
+				*limit = i;
+				return 0;
+			}
+		}
+	}
+
+	*limit = 0;
+	return 0;
+}
+
+static void kv_apply_state_adjust_rules(struct amdgpu_device *adev,
+					struct amdgpu_ps *new_rps,
+					struct amdgpu_ps *old_rps)
+{
+	struct kv_ps *ps = kv_get_ps(new_rps);
+	struct kv_power_info *pi = kv_get_pi(adev);
+	u32 min_sclk = 10000; /* ??? */
+	u32 sclk, mclk = 0;
+	int i, limit;
+	bool force_high;
+	struct amdgpu_clock_voltage_dependency_table *table =
+		&adev->pm.dpm.dyn_state.vddc_dependency_on_sclk;
+	u32 stable_p_state_sclk = 0;
+	struct amdgpu_clock_and_voltage_limits *max_limits =
+		&adev->pm.dpm.dyn_state.max_clock_voltage_on_ac;
+
+	if (new_rps->vce_active) {
+		new_rps->evclk = adev->pm.dpm.vce_states[adev->pm.dpm.vce_level].evclk;
+		new_rps->ecclk = adev->pm.dpm.vce_states[adev->pm.dpm.vce_level].ecclk;
+	} else {
+		new_rps->evclk = 0;
+		new_rps->ecclk = 0;
+	}
+
+	mclk = max_limits->mclk;
+	sclk = min_sclk;
+
+	if (pi->caps_stable_p_state) {
+		stable_p_state_sclk = (max_limits->sclk * 75) / 100;
+
+		for (i = table->count - 1; i >= 0; i++) {
+			if (stable_p_state_sclk >= table->entries[i].clk) {
+				stable_p_state_sclk = table->entries[i].clk;
+				break;
+			}
+		}
+
+		if (i > 0)
+			stable_p_state_sclk = table->entries[0].clk;
+
+		sclk = stable_p_state_sclk;
+	}
+
+	if (new_rps->vce_active) {
+		if (sclk < adev->pm.dpm.vce_states[adev->pm.dpm.vce_level].sclk)
+			sclk = adev->pm.dpm.vce_states[adev->pm.dpm.vce_level].sclk;
+	}
+
+	ps->need_dfs_bypass = true;
+
+	for (i = 0; i < ps->num_levels; i++) {
+		if (ps->levels[i].sclk < sclk)
+			ps->levels[i].sclk = sclk;
+	}
+
+	if (table && table->count) {
+		for (i = 0; i < ps->num_levels; i++) {
+			if (pi->high_voltage_t &&
+			    (pi->high_voltage_t <
+			     kv_convert_8bit_index_to_voltage(adev, ps->levels[i].vddc_index))) {
+				kv_get_high_voltage_limit(adev, &limit);
+				ps->levels[i].sclk = table->entries[limit].clk;
+			}
+		}
+	} else {
+		struct sumo_sclk_voltage_mapping_table *table =
+			&pi->sys_info.sclk_voltage_mapping_table;
+
+		for (i = 0; i < ps->num_levels; i++) {
+			if (pi->high_voltage_t &&
+			    (pi->high_voltage_t <
+			     kv_convert_8bit_index_to_voltage(adev, ps->levels[i].vddc_index))) {
+				kv_get_high_voltage_limit(adev, &limit);
+				ps->levels[i].sclk = table->entries[limit].sclk_frequency;
+			}
+		}
+	}
+
+	if (pi->caps_stable_p_state) {
+		for (i = 0; i < ps->num_levels; i++) {
+			ps->levels[i].sclk = stable_p_state_sclk;
+		}
+	}
+
+	pi->video_start = new_rps->dclk || new_rps->vclk ||
+		new_rps->evclk || new_rps->ecclk;
+
+	if ((new_rps->class & ATOM_PPLIB_CLASSIFICATION_UI_MASK) ==
+	    ATOM_PPLIB_CLASSIFICATION_UI_BATTERY)
+		pi->battery_state = true;
+	else
+		pi->battery_state = false;
+
+	if (adev->asic_type == CHIP_KABINI || adev->asic_type == CHIP_MULLINS) {
+		ps->dpm0_pg_nb_ps_lo = 0x1;
+		ps->dpm0_pg_nb_ps_hi = 0x0;
+		ps->dpmx_nb_ps_lo = 0x1;
+		ps->dpmx_nb_ps_hi = 0x0;
+	} else {
+		ps->dpm0_pg_nb_ps_lo = 0x3;
+		ps->dpm0_pg_nb_ps_hi = 0x0;
+		ps->dpmx_nb_ps_lo = 0x3;
+		ps->dpmx_nb_ps_hi = 0x0;
+
+		if (pi->sys_info.nb_dpm_enable) {
+			force_high = (mclk >= pi->sys_info.nbp_memory_clock[3]) ||
+				pi->video_start || (adev->pm.dpm.new_active_crtc_count >= 3) ||
+				pi->disable_nb_ps3_in_battery;
+			ps->dpm0_pg_nb_ps_lo = force_high ? 0x2 : 0x3;
+			ps->dpm0_pg_nb_ps_hi = 0x2;
+			ps->dpmx_nb_ps_lo = force_high ? 0x2 : 0x3;
+			ps->dpmx_nb_ps_hi = 0x2;
+		}
+	}
+}
+
+static void kv_dpm_power_level_enabled_for_throttle(struct amdgpu_device *adev,
+						    u32 index, bool enable)
+{
+	struct kv_power_info *pi = kv_get_pi(adev);
+
+	pi->graphics_level[index].EnabledForThrottle = enable ? 1 : 0;
+}
+
+static int kv_calculate_ds_divider(struct amdgpu_device *adev)
+{
+	struct kv_power_info *pi = kv_get_pi(adev);
+	u32 sclk_in_sr = 10000; /* ??? */
+	u32 i;
+
+	if (pi->lowest_valid > pi->highest_valid)
+		return -EINVAL;
+
+	for (i = pi->lowest_valid; i <= pi->highest_valid; i++) {
+		pi->graphics_level[i].DeepSleepDivId =
+			kv_get_sleep_divider_id_from_clock(adev,
+							   be32_to_cpu(pi->graphics_level[i].SclkFrequency),
+							   sclk_in_sr);
+	}
+	return 0;
+}
+
+static int kv_calculate_nbps_level_settings(struct amdgpu_device *adev)
+{
+	struct kv_power_info *pi = kv_get_pi(adev);
+	u32 i;
+	bool force_high;
+	struct amdgpu_clock_and_voltage_limits *max_limits =
+		&adev->pm.dpm.dyn_state.max_clock_voltage_on_ac;
+	u32 mclk = max_limits->mclk;
+
+	if (pi->lowest_valid > pi->highest_valid)
+		return -EINVAL;
+
+	if (adev->asic_type == CHIP_KABINI || adev->asic_type == CHIP_MULLINS) {
+		for (i = pi->lowest_valid; i <= pi->highest_valid; i++) {
+			pi->graphics_level[i].GnbSlow = 1;
+			pi->graphics_level[i].ForceNbPs1 = 0;
+			pi->graphics_level[i].UpH = 0;
+		}
+
+		if (!pi->sys_info.nb_dpm_enable)
+			return 0;
+
+		force_high = ((mclk >= pi->sys_info.nbp_memory_clock[3]) ||
+			      (adev->pm.dpm.new_active_crtc_count >= 3) || pi->video_start);
+
+		if (force_high) {
+			for (i = pi->lowest_valid; i <= pi->highest_valid; i++)
+				pi->graphics_level[i].GnbSlow = 0;
+		} else {
+			if (pi->battery_state)
+				pi->graphics_level[0].ForceNbPs1 = 1;
+
+			pi->graphics_level[1].GnbSlow = 0;
+			pi->graphics_level[2].GnbSlow = 0;
+			pi->graphics_level[3].GnbSlow = 0;
+			pi->graphics_level[4].GnbSlow = 0;
+		}
+	} else {
+		for (i = pi->lowest_valid; i <= pi->highest_valid; i++) {
+			pi->graphics_level[i].GnbSlow = 1;
+			pi->graphics_level[i].ForceNbPs1 = 0;
+			pi->graphics_level[i].UpH = 0;
+		}
+
+		if (pi->sys_info.nb_dpm_enable && pi->battery_state) {
+			pi->graphics_level[pi->lowest_valid].UpH = 0x28;
+			pi->graphics_level[pi->lowest_valid].GnbSlow = 0;
+			if (pi->lowest_valid != pi->highest_valid)
+				pi->graphics_level[pi->lowest_valid].ForceNbPs1 = 1;
+		}
+	}
+	return 0;
+}
+
+static int kv_calculate_dpm_settings(struct amdgpu_device *adev)
+{
+	struct kv_power_info *pi = kv_get_pi(adev);
+	u32 i;
+
+	if (pi->lowest_valid > pi->highest_valid)
+		return -EINVAL;
+
+	for (i = pi->lowest_valid; i <= pi->highest_valid; i++)
+		pi->graphics_level[i].DisplayWatermark = (i == pi->highest_valid) ? 1 : 0;
+
+	return 0;
+}
+
+static void kv_init_graphics_levels(struct amdgpu_device *adev)
+{
+	struct kv_power_info *pi = kv_get_pi(adev);
+	u32 i;
+	struct amdgpu_clock_voltage_dependency_table *table =
+		&adev->pm.dpm.dyn_state.vddc_dependency_on_sclk;
+
+	if (table && table->count) {
+		u32 vid_2bit;
+
+		pi->graphics_dpm_level_count = 0;
+		for (i = 0; i < table->count; i++) {
+			if (pi->high_voltage_t &&
+			    (pi->high_voltage_t <
+			     kv_convert_8bit_index_to_voltage(adev, table->entries[i].v)))
+				break;
+
+			kv_set_divider_value(adev, i, table->entries[i].clk);
+			vid_2bit = kv_convert_vid7_to_vid2(adev,
+							   &pi->sys_info.vid_mapping_table,
+							   table->entries[i].v);
+			kv_set_vid(adev, i, vid_2bit);
+			kv_set_at(adev, i, pi->at[i]);
+			kv_dpm_power_level_enabled_for_throttle(adev, i, true);
+			pi->graphics_dpm_level_count++;
+		}
+	} else {
+		struct sumo_sclk_voltage_mapping_table *table =
+			&pi->sys_info.sclk_voltage_mapping_table;
+
+		pi->graphics_dpm_level_count = 0;
+		for (i = 0; i < table->num_max_dpm_entries; i++) {
+			if (pi->high_voltage_t &&
+			    pi->high_voltage_t <
+			    kv_convert_2bit_index_to_voltage(adev, table->entries[i].vid_2bit))
+				break;
+
+			kv_set_divider_value(adev, i, table->entries[i].sclk_frequency);
+			kv_set_vid(adev, i, table->entries[i].vid_2bit);
+			kv_set_at(adev, i, pi->at[i]);
+			kv_dpm_power_level_enabled_for_throttle(adev, i, true);
+			pi->graphics_dpm_level_count++;
+		}
+	}
+
+	for (i = 0; i < SMU7_MAX_LEVELS_GRAPHICS; i++)
+		kv_dpm_power_level_enable(adev, i, false);
+}
+
+static void kv_enable_new_levels(struct amdgpu_device *adev)
+{
+	struct kv_power_info *pi = kv_get_pi(adev);
+	u32 i;
+
+	for (i = 0; i < SMU7_MAX_LEVELS_GRAPHICS; i++) {
+		if (i >= pi->lowest_valid && i <= pi->highest_valid)
+			kv_dpm_power_level_enable(adev, i, true);
+	}
+}
+
+static int kv_set_enabled_level(struct amdgpu_device *adev, u32 level)
+{
+	u32 new_mask = (1 << level);
+
+	return amdgpu_kv_send_msg_to_smc_with_parameter(adev,
+						 PPSMC_MSG_SCLKDPM_SetEnabledMask,
+						 new_mask);
+}
+
+static int kv_set_enabled_levels(struct amdgpu_device *adev)
+{
+	struct kv_power_info *pi = kv_get_pi(adev);
+	u32 i, new_mask = 0;
+
+	for (i = pi->lowest_valid; i <= pi->highest_valid; i++)
+		new_mask |= (1 << i);
+
+	return amdgpu_kv_send_msg_to_smc_with_parameter(adev,
+						 PPSMC_MSG_SCLKDPM_SetEnabledMask,
+						 new_mask);
+}
+
+static void kv_program_nbps_index_settings(struct amdgpu_device *adev,
+					   struct amdgpu_ps *new_rps)
+{
+	struct kv_ps *new_ps = kv_get_ps(new_rps);
+	struct kv_power_info *pi = kv_get_pi(adev);
+	u32 nbdpmconfig1;
+
+	if (adev->asic_type == CHIP_KABINI || adev->asic_type == CHIP_MULLINS)
+		return;
+
+	if (pi->sys_info.nb_dpm_enable) {
+		nbdpmconfig1 = RREG32_SMC(ixNB_DPM_CONFIG_1);
+		nbdpmconfig1 &= ~(NB_DPM_CONFIG_1__Dpm0PgNbPsLo_MASK |
+				NB_DPM_CONFIG_1__Dpm0PgNbPsHi_MASK |
+				NB_DPM_CONFIG_1__DpmXNbPsLo_MASK |
+				NB_DPM_CONFIG_1__DpmXNbPsHi_MASK);
+		nbdpmconfig1 |= (new_ps->dpm0_pg_nb_ps_lo << NB_DPM_CONFIG_1__Dpm0PgNbPsLo__SHIFT) |
+				(new_ps->dpm0_pg_nb_ps_hi << NB_DPM_CONFIG_1__Dpm0PgNbPsHi__SHIFT) |
+				(new_ps->dpmx_nb_ps_lo << NB_DPM_CONFIG_1__DpmXNbPsLo__SHIFT) |
+				(new_ps->dpmx_nb_ps_hi << NB_DPM_CONFIG_1__DpmXNbPsHi__SHIFT);
+		WREG32_SMC(ixNB_DPM_CONFIG_1, nbdpmconfig1);
+	}
+}
+
+static int kv_set_thermal_temperature_range(struct amdgpu_device *adev,
+					    int min_temp, int max_temp)
+{
+	int low_temp = 0 * 1000;
+	int high_temp = 255 * 1000;
+	u32 tmp;
+
+	if (low_temp < min_temp)
+		low_temp = min_temp;
+	if (high_temp > max_temp)
+		high_temp = max_temp;
+	if (high_temp < low_temp) {
+		DRM_ERROR("invalid thermal range: %d - %d\n", low_temp, high_temp);
+		return -EINVAL;
+	}
+
+	tmp = RREG32_SMC(ixCG_THERMAL_INT_CTRL);
+	tmp &= ~(CG_THERMAL_INT_CTRL__DIG_THERM_INTH_MASK |
+		CG_THERMAL_INT_CTRL__DIG_THERM_INTL_MASK);
+	tmp |= ((49 + (high_temp / 1000)) << CG_THERMAL_INT_CTRL__DIG_THERM_INTH__SHIFT) |
+		((49 + (low_temp / 1000)) << CG_THERMAL_INT_CTRL__DIG_THERM_INTL__SHIFT);
+	WREG32_SMC(ixCG_THERMAL_INT_CTRL, tmp);
+
+	adev->pm.dpm.thermal.min_temp = low_temp;
+	adev->pm.dpm.thermal.max_temp = high_temp;
+
+	return 0;
+}
+
+union igp_info {
+	struct _ATOM_INTEGRATED_SYSTEM_INFO info;
+	struct _ATOM_INTEGRATED_SYSTEM_INFO_V2 info_2;
+	struct _ATOM_INTEGRATED_SYSTEM_INFO_V5 info_5;
+	struct _ATOM_INTEGRATED_SYSTEM_INFO_V6 info_6;
+	struct _ATOM_INTEGRATED_SYSTEM_INFO_V1_7 info_7;
+	struct _ATOM_INTEGRATED_SYSTEM_INFO_V1_8 info_8;
+};
+
+static int kv_parse_sys_info_table(struct amdgpu_device *adev)
+{
+	struct kv_power_info *pi = kv_get_pi(adev);
+	struct amdgpu_mode_info *mode_info = &adev->mode_info;
+	int index = GetIndexIntoMasterTable(DATA, IntegratedSystemInfo);
+	union igp_info *igp_info;
+	u8 frev, crev;
+	u16 data_offset;
+	int i;
+
+	if (amdgpu_atom_parse_data_header(mode_info->atom_context, index, NULL,
+				   &frev, &crev, &data_offset)) {
+		igp_info = (union igp_info *)(mode_info->atom_context->bios +
+					      data_offset);
+
+		if (crev != 8) {
+			DRM_ERROR("Unsupported IGP table: %d %d\n", frev, crev);
+			return -EINVAL;
+		}
+		pi->sys_info.bootup_sclk = le32_to_cpu(igp_info->info_8.ulBootUpEngineClock);
+		pi->sys_info.bootup_uma_clk = le32_to_cpu(igp_info->info_8.ulBootUpUMAClock);
+		pi->sys_info.bootup_nb_voltage_index =
+			le16_to_cpu(igp_info->info_8.usBootUpNBVoltage);
+		if (igp_info->info_8.ucHtcTmpLmt == 0)
+			pi->sys_info.htc_tmp_lmt = 203;
+		else
+			pi->sys_info.htc_tmp_lmt = igp_info->info_8.ucHtcTmpLmt;
+		if (igp_info->info_8.ucHtcHystLmt == 0)
+			pi->sys_info.htc_hyst_lmt = 5;
+		else
+			pi->sys_info.htc_hyst_lmt = igp_info->info_8.ucHtcHystLmt;
+		if (pi->sys_info.htc_tmp_lmt <= pi->sys_info.htc_hyst_lmt) {
+			DRM_ERROR("The htcTmpLmt should be larger than htcHystLmt.\n");
+		}
+
+		if (le32_to_cpu(igp_info->info_8.ulSystemConfig) & (1 << 3))
+			pi->sys_info.nb_dpm_enable = true;
+		else
+			pi->sys_info.nb_dpm_enable = false;
+
+		for (i = 0; i < KV_NUM_NBPSTATES; i++) {
+			pi->sys_info.nbp_memory_clock[i] =
+				le32_to_cpu(igp_info->info_8.ulNbpStateMemclkFreq[i]);
+			pi->sys_info.nbp_n_clock[i] =
+				le32_to_cpu(igp_info->info_8.ulNbpStateNClkFreq[i]);
+		}
+		if (le32_to_cpu(igp_info->info_8.ulGPUCapInfo) &
+		    SYS_INFO_GPUCAPS__ENABEL_DFS_BYPASS)
+			pi->caps_enable_dfs_bypass = true;
+
+		sumo_construct_sclk_voltage_mapping_table(adev,
+							  &pi->sys_info.sclk_voltage_mapping_table,
+							  igp_info->info_8.sAvail_SCLK);
+
+		sumo_construct_vid_mapping_table(adev,
+						 &pi->sys_info.vid_mapping_table,
+						 igp_info->info_8.sAvail_SCLK);
+
+		kv_construct_max_power_limits_table(adev,
+						    &adev->pm.dpm.dyn_state.max_clock_voltage_on_ac);
+	}
+	return 0;
+}
+
+union power_info {
+	struct _ATOM_POWERPLAY_INFO info;
+	struct _ATOM_POWERPLAY_INFO_V2 info_2;
+	struct _ATOM_POWERPLAY_INFO_V3 info_3;
+	struct _ATOM_PPLIB_POWERPLAYTABLE pplib;
+	struct _ATOM_PPLIB_POWERPLAYTABLE2 pplib2;
+	struct _ATOM_PPLIB_POWERPLAYTABLE3 pplib3;
+};
+
+union pplib_clock_info {
+	struct _ATOM_PPLIB_R600_CLOCK_INFO r600;
+	struct _ATOM_PPLIB_RS780_CLOCK_INFO rs780;
+	struct _ATOM_PPLIB_EVERGREEN_CLOCK_INFO evergreen;
+	struct _ATOM_PPLIB_SUMO_CLOCK_INFO sumo;
+};
+
+union pplib_power_state {
+	struct _ATOM_PPLIB_STATE v1;
+	struct _ATOM_PPLIB_STATE_V2 v2;
+};
+
+static void kv_patch_boot_state(struct amdgpu_device *adev,
+				struct kv_ps *ps)
+{
+	struct kv_power_info *pi = kv_get_pi(adev);
+
+	ps->num_levels = 1;
+	ps->levels[0] = pi->boot_pl;
+}
+
+static void kv_parse_pplib_non_clock_info(struct amdgpu_device *adev,
+					  struct amdgpu_ps *rps,
+					  struct _ATOM_PPLIB_NONCLOCK_INFO *non_clock_info,
+					  u8 table_rev)
+{
+	struct kv_ps *ps = kv_get_ps(rps);
+
+	rps->caps = le32_to_cpu(non_clock_info->ulCapsAndSettings);
+	rps->class = le16_to_cpu(non_clock_info->usClassification);
+	rps->class2 = le16_to_cpu(non_clock_info->usClassification2);
+
+	if (ATOM_PPLIB_NONCLOCKINFO_VER1 < table_rev) {
+		rps->vclk = le32_to_cpu(non_clock_info->ulVCLK);
+		rps->dclk = le32_to_cpu(non_clock_info->ulDCLK);
+	} else {
+		rps->vclk = 0;
+		rps->dclk = 0;
+	}
+
+	if (rps->class & ATOM_PPLIB_CLASSIFICATION_BOOT) {
+		adev->pm.dpm.boot_ps = rps;
+		kv_patch_boot_state(adev, ps);
+	}
+	if (rps->class & ATOM_PPLIB_CLASSIFICATION_UVDSTATE)
+		adev->pm.dpm.uvd_ps = rps;
+}
+
+static void kv_parse_pplib_clock_info(struct amdgpu_device *adev,
+				      struct amdgpu_ps *rps, int index,
+					union pplib_clock_info *clock_info)
+{
+	struct kv_power_info *pi = kv_get_pi(adev);
+	struct kv_ps *ps = kv_get_ps(rps);
+	struct kv_pl *pl = &ps->levels[index];
+	u32 sclk;
+
+	sclk = le16_to_cpu(clock_info->sumo.usEngineClockLow);
+	sclk |= clock_info->sumo.ucEngineClockHigh << 16;
+	pl->sclk = sclk;
+	pl->vddc_index = clock_info->sumo.vddcIndex;
+
+	ps->num_levels = index + 1;
+
+	if (pi->caps_sclk_ds) {
+		pl->ds_divider_index = 5;
+		pl->ss_divider_index = 5;
+	}
+}
+
+static int kv_parse_power_table(struct amdgpu_device *adev)
+{
+	struct amdgpu_mode_info *mode_info = &adev->mode_info;
+	struct _ATOM_PPLIB_NONCLOCK_INFO *non_clock_info;
+	union pplib_power_state *power_state;
+	int i, j, k, non_clock_array_index, clock_array_index;
+	union pplib_clock_info *clock_info;
+	struct _StateArray *state_array;
+	struct _ClockInfoArray *clock_info_array;
+	struct _NonClockInfoArray *non_clock_info_array;
+	union power_info *power_info;
+	int index = GetIndexIntoMasterTable(DATA, PowerPlayInfo);
+	u16 data_offset;
+	u8 frev, crev;
+	u8 *power_state_offset;
+	struct kv_ps *ps;
+
+	if (!amdgpu_atom_parse_data_header(mode_info->atom_context, index, NULL,
+				   &frev, &crev, &data_offset))
+		return -EINVAL;
+	power_info = (union power_info *)(mode_info->atom_context->bios + data_offset);
+
+	amdgpu_add_thermal_controller(adev);
+
+	state_array = (struct _StateArray *)
+		(mode_info->atom_context->bios + data_offset +
+		 le16_to_cpu(power_info->pplib.usStateArrayOffset));
+	clock_info_array = (struct _ClockInfoArray *)
+		(mode_info->atom_context->bios + data_offset +
+		 le16_to_cpu(power_info->pplib.usClockInfoArrayOffset));
+	non_clock_info_array = (struct _NonClockInfoArray *)
+		(mode_info->atom_context->bios + data_offset +
+		 le16_to_cpu(power_info->pplib.usNonClockInfoArrayOffset));
+
+	adev->pm.dpm.ps = kzalloc(sizeof(struct amdgpu_ps) *
+				  state_array->ucNumEntries, GFP_KERNEL);
+	if (!adev->pm.dpm.ps)
+		return -ENOMEM;
+	power_state_offset = (u8 *)state_array->states;
+	for (i = 0; i < state_array->ucNumEntries; i++) {
+		u8 *idx;
+		power_state = (union pplib_power_state *)power_state_offset;
+		non_clock_array_index = power_state->v2.nonClockInfoIndex;
+		non_clock_info = (struct _ATOM_PPLIB_NONCLOCK_INFO *)
+			&non_clock_info_array->nonClockInfo[non_clock_array_index];
+		ps = kzalloc(sizeof(struct kv_ps), GFP_KERNEL);
+		if (ps == NULL) {
+			kfree(adev->pm.dpm.ps);
+			return -ENOMEM;
+		}
+		adev->pm.dpm.ps[i].ps_priv = ps;
+		k = 0;
+		idx = (u8 *)&power_state->v2.clockInfoIndex[0];
+		for (j = 0; j < power_state->v2.ucNumDPMLevels; j++) {
+			clock_array_index = idx[j];
+			if (clock_array_index >= clock_info_array->ucNumEntries)
+				continue;
+			if (k >= SUMO_MAX_HARDWARE_POWERLEVELS)
+				break;
+			clock_info = (union pplib_clock_info *)
+				((u8 *)&clock_info_array->clockInfo[0] +
+				 (clock_array_index * clock_info_array->ucEntrySize));
+			kv_parse_pplib_clock_info(adev,
+						  &adev->pm.dpm.ps[i], k,
+						  clock_info);
+			k++;
+		}
+		kv_parse_pplib_non_clock_info(adev, &adev->pm.dpm.ps[i],
+					      non_clock_info,
+					      non_clock_info_array->ucEntrySize);
+		power_state_offset += 2 + power_state->v2.ucNumDPMLevels;
+	}
+	adev->pm.dpm.num_ps = state_array->ucNumEntries;
+
+	/* fill in the vce power states */
+	for (i = 0; i < AMDGPU_MAX_VCE_LEVELS; i++) {
+		u32 sclk;
+		clock_array_index = adev->pm.dpm.vce_states[i].clk_idx;
+		clock_info = (union pplib_clock_info *)
+			&clock_info_array->clockInfo[clock_array_index * clock_info_array->ucEntrySize];
+		sclk = le16_to_cpu(clock_info->sumo.usEngineClockLow);
+		sclk |= clock_info->sumo.ucEngineClockHigh << 16;
+		adev->pm.dpm.vce_states[i].sclk = sclk;
+		adev->pm.dpm.vce_states[i].mclk = 0;
+	}
+
+	return 0;
+}
+
+static int kv_dpm_init(struct amdgpu_device *adev)
+{
+	struct kv_power_info *pi;
+	int ret, i;
+
+	pi = kzalloc(sizeof(struct kv_power_info), GFP_KERNEL);
+	if (pi == NULL)
+		return -ENOMEM;
+	adev->pm.dpm.priv = pi;
+
+	ret = amdgpu_get_platform_caps(adev);
+	if (ret)
+		return ret;
+
+	ret = amdgpu_parse_extended_power_table(adev);
+	if (ret)
+		return ret;
+
+	for (i = 0; i < SUMO_MAX_HARDWARE_POWERLEVELS; i++)
+		pi->at[i] = TRINITY_AT_DFLT;
+
+	pi->sram_end = SMC_RAM_END;
+
+	pi->enable_nb_dpm = true;
+
+	pi->caps_power_containment = true;
+	pi->caps_cac = true;
+	pi->enable_didt = false;
+	if (pi->enable_didt) {
+		pi->caps_sq_ramping = true;
+		pi->caps_db_ramping = true;
+		pi->caps_td_ramping = true;
+		pi->caps_tcp_ramping = true;
+	}
+
+	pi->caps_sclk_ds = true;
+	pi->enable_auto_thermal_throttling = true;
+	pi->disable_nb_ps3_in_battery = false;
+	if (amdgpu_bapm == 0)
+		pi->bapm_enable = false;
+	else
+		pi->bapm_enable = true;
+	pi->voltage_drop_t = 0;
+	pi->caps_sclk_throttle_low_notification = false;
+	pi->caps_fps = false; /* true? */
+	pi->caps_uvd_pg = (adev->pg_flags & AMDGPU_PG_SUPPORT_UVD) ? true : false;
+	pi->caps_uvd_dpm = true;
+	pi->caps_vce_pg = (adev->pg_flags & AMDGPU_PG_SUPPORT_VCE) ? true : false;
+	pi->caps_samu_pg = (adev->pg_flags & AMDGPU_PG_SUPPORT_SAMU) ? true : false;
+	pi->caps_acp_pg = (adev->pg_flags & AMDGPU_PG_SUPPORT_ACP) ? true : false;
+	pi->caps_stable_p_state = false;
+
+	ret = kv_parse_sys_info_table(adev);
+	if (ret)
+		return ret;
+
+	kv_patch_voltage_values(adev);
+	kv_construct_boot_state(adev);
+
+	ret = kv_parse_power_table(adev);
+	if (ret)
+		return ret;
+
+	pi->enable_dpm = true;
+
+	return 0;
+}
+
+static void
+kv_dpm_debugfs_print_current_performance_level(struct amdgpu_device *adev,
+					       struct seq_file *m)
+{
+	struct kv_power_info *pi = kv_get_pi(adev);
+	u32 current_index =
+		(RREG32_SMC(ixTARGET_AND_CURRENT_PROFILE_INDEX) &
+		TARGET_AND_CURRENT_PROFILE_INDEX__CURR_SCLK_INDEX_MASK) >>
+		TARGET_AND_CURRENT_PROFILE_INDEX__CURR_SCLK_INDEX__SHIFT;
+	u32 sclk, tmp;
+	u16 vddc;
+
+	if (current_index >= SMU__NUM_SCLK_DPM_STATE) {
+		seq_printf(m, "invalid dpm profile %d\n", current_index);
+	} else {
+		sclk = be32_to_cpu(pi->graphics_level[current_index].SclkFrequency);
+		tmp = (RREG32_SMC(ixSMU_VOLTAGE_STATUS) &
+			SMU_VOLTAGE_STATUS__SMU_VOLTAGE_CURRENT_LEVEL_MASK) >>
+			SMU_VOLTAGE_STATUS__SMU_VOLTAGE_CURRENT_LEVEL__SHIFT;
+		vddc = kv_convert_8bit_index_to_voltage(adev, (u16)tmp);
+		seq_printf(m, "uvd    %sabled\n", pi->uvd_power_gated ? "dis" : "en");
+		seq_printf(m, "vce    %sabled\n", pi->vce_power_gated ? "dis" : "en");
+		seq_printf(m, "power level %d    sclk: %u vddc: %u\n",
+			   current_index, sclk, vddc);
+	}
+}
+
+static void
+kv_dpm_print_power_state(struct amdgpu_device *adev,
+			 struct amdgpu_ps *rps)
+{
+	int i;
+	struct kv_ps *ps = kv_get_ps(rps);
+
+	amdgpu_dpm_print_class_info(rps->class, rps->class2);
+	amdgpu_dpm_print_cap_info(rps->caps);
+	printk("\tuvd    vclk: %d dclk: %d\n", rps->vclk, rps->dclk);
+	for (i = 0; i < ps->num_levels; i++) {
+		struct kv_pl *pl = &ps->levels[i];
+		printk("\t\tpower level %d    sclk: %u vddc: %u\n",
+		       i, pl->sclk,
+		       kv_convert_8bit_index_to_voltage(adev, pl->vddc_index));
+	}
+	amdgpu_dpm_print_ps_status(adev, rps);
+}
+
+static void kv_dpm_fini(struct amdgpu_device *adev)
+{
+	int i;
+
+	for (i = 0; i < adev->pm.dpm.num_ps; i++) {
+		kfree(adev->pm.dpm.ps[i].ps_priv);
+	}
+	kfree(adev->pm.dpm.ps);
+	kfree(adev->pm.dpm.priv);
+	amdgpu_free_extended_power_table(adev);
+}
+
+static void kv_dpm_display_configuration_changed(struct amdgpu_device *adev)
+{
+
+}
+
+static u32 kv_dpm_get_sclk(struct amdgpu_device *adev, bool low)
+{
+	struct kv_power_info *pi = kv_get_pi(adev);
+	struct kv_ps *requested_state = kv_get_ps(&pi->requested_rps);
+
+	if (low)
+		return requested_state->levels[0].sclk;
+	else
+		return requested_state->levels[requested_state->num_levels - 1].sclk;
+}
+
+static u32 kv_dpm_get_mclk(struct amdgpu_device *adev, bool low)
+{
+	struct kv_power_info *pi = kv_get_pi(adev);
+
+	return pi->sys_info.bootup_uma_clk;
+}
+
+/* get temperature in millidegrees */
+static int kv_dpm_get_temp(struct amdgpu_device *adev)
+{
+	u32 temp;
+	int actual_temp = 0;
+
+	temp = RREG32_SMC(0xC0300E0C);
+
+	if (temp)
+		actual_temp = (temp / 8) - 49;
+	else
+		actual_temp = 0;
+
+	actual_temp = actual_temp * 1000;
+
+	return actual_temp;
+}
+
+static int kv_dpm_early_init(struct amdgpu_device *adev)
+{
+	kv_dpm_set_dpm_funcs(adev);
+	kv_dpm_set_irq_funcs(adev);
+
+	return 0;
+}
+
+static int kv_dpm_late_init(struct amdgpu_device *adev)
+{
+	if (!amdgpu_dpm)
+		return 0;
+
+	/* powerdown unused blocks for now */
+	kv_dpm_powergate_acp(adev, true);
+	kv_dpm_powergate_samu(adev, true);
+	kv_dpm_powergate_vce(adev, true);
+	kv_dpm_powergate_uvd(adev, true);
+
+	return 0;
+}
+
+static int kv_dpm_sw_init(struct amdgpu_device *adev)
+{
+	int ret;
+
+	ret = amdgpu_irq_add_id(adev, 230, &adev->pm.dpm.thermal.irq);
+	if (ret)
+		return ret;
+
+	ret = amdgpu_irq_add_id(adev, 231, &adev->pm.dpm.thermal.irq);
+	if (ret)
+		return ret;
+
+	/* default to balanced state */
+	adev->pm.dpm.state = POWER_STATE_TYPE_BALANCED;
+	adev->pm.dpm.user_state = POWER_STATE_TYPE_BALANCED;
+	adev->pm.dpm.forced_level = AMDGPU_DPM_FORCED_LEVEL_AUTO;
+	adev->pm.default_sclk = adev->clock.default_sclk;
+	adev->pm.default_mclk = adev->clock.default_mclk;
+	adev->pm.current_sclk = adev->clock.default_sclk;
+	adev->pm.current_mclk = adev->clock.default_mclk;
+	adev->pm.int_thermal_type = THERMAL_TYPE_NONE;
+
+	if (amdgpu_dpm == 0)
+		return 0;
+
+	INIT_WORK(&adev->pm.dpm.thermal.work, amdgpu_dpm_thermal_work_handler);
+	mutex_lock(&adev->pm.mutex);
+	ret = kv_dpm_init(adev);
+	if (ret)
+		goto dpm_failed;
+	adev->pm.dpm.current_ps = adev->pm.dpm.requested_ps = adev->pm.dpm.boot_ps;
+	if (amdgpu_dpm == 1)
+		amdgpu_pm_print_power_states(adev);
+	ret = amdgpu_pm_sysfs_init(adev);
+	if (ret)
+		goto dpm_failed;
+	mutex_unlock(&adev->pm.mutex);
+	DRM_INFO("amdgpu: dpm initialized\n");
+
+	return 0;
+
+dpm_failed:
+	kv_dpm_fini(adev);
+	mutex_unlock(&adev->pm.mutex);
+	DRM_ERROR("amdgpu: dpm initialization failed\n");
+	return ret;
+}
+
+static int kv_dpm_sw_fini(struct amdgpu_device *adev)
+{
+	mutex_lock(&adev->pm.mutex);
+	amdgpu_pm_sysfs_fini(adev);
+	kv_dpm_fini(adev);
+	mutex_unlock(&adev->pm.mutex);
+
+	return 0;
+}
+
+static int kv_dpm_hw_init(struct amdgpu_device *adev)
+{
+	int ret;
+
+	if (!amdgpu_dpm)
+		return 0;
+
+	mutex_lock(&adev->pm.mutex);
+	kv_dpm_setup_asic(adev);
+	ret = kv_dpm_enable(adev);
+	if (ret)
+		adev->pm.dpm_enabled = false;
+	else
+		adev->pm.dpm_enabled = true;
+	mutex_unlock(&adev->pm.mutex);
+
+	return ret;
+}
+
+static int kv_dpm_hw_fini(struct amdgpu_device *adev)
+{
+	if (adev->pm.dpm_enabled) {
+		mutex_lock(&adev->pm.mutex);
+		kv_dpm_disable(adev);
+		mutex_unlock(&adev->pm.mutex);
+	}
+
+	return 0;
+}
+
+static int kv_dpm_suspend(struct amdgpu_device *adev)
+{
+	if (adev->pm.dpm_enabled) {
+		mutex_lock(&adev->pm.mutex);
+		/* disable dpm */
+		kv_dpm_disable(adev);
+		/* reset the power state */
+		adev->pm.dpm.current_ps = adev->pm.dpm.requested_ps = adev->pm.dpm.boot_ps;
+		mutex_unlock(&adev->pm.mutex);
+	}
+	return 0;
+}
+
+static int kv_dpm_resume(struct amdgpu_device *adev)
+{
+	int ret;
+
+	if (adev->pm.dpm_enabled) {
+		/* asic init will reset to the boot state */
+		mutex_lock(&adev->pm.mutex);
+		kv_dpm_setup_asic(adev);
+		ret = kv_dpm_enable(adev);
+		if (ret)
+			adev->pm.dpm_enabled = false;
+		else
+			adev->pm.dpm_enabled = true;
+		mutex_unlock(&adev->pm.mutex);
+		if (adev->pm.dpm_enabled)
+			amdgpu_pm_compute_clocks(adev);
+	}
+	return 0;
+}
+
+static bool kv_dpm_is_idle(struct amdgpu_device *adev)
+{
+	/* XXX */
+	return true;
+}
+
+static int kv_dpm_wait_for_idle(struct amdgpu_device *adev)
+{
+	/* XXX */
+	return 0;
+}
+
+static void kv_dpm_print_status(struct amdgpu_device *adev)
+{
+	dev_info(adev->dev, "KV/KB DPM registers\n");
+	dev_info(adev->dev, "  DIDT_SQ_CTRL0=0x%08X\n",
+		 RREG32_DIDT(ixDIDT_SQ_CTRL0));
+	dev_info(adev->dev, "  DIDT_DB_CTRL0=0x%08X\n",
+		 RREG32_DIDT(ixDIDT_DB_CTRL0));
+	dev_info(adev->dev, "  DIDT_TD_CTRL0=0x%08X\n",
+		 RREG32_DIDT(ixDIDT_TD_CTRL0));
+	dev_info(adev->dev, "  DIDT_TCP_CTRL0=0x%08X\n",
+		 RREG32_DIDT(ixDIDT_TCP_CTRL0));
+	dev_info(adev->dev, "  LCAC_SX0_OVR_SEL=0x%08X\n",
+		 RREG32_SMC(ixLCAC_SX0_OVR_SEL));
+	dev_info(adev->dev, "  LCAC_SX0_OVR_VAL=0x%08X\n",
+		 RREG32_SMC(ixLCAC_SX0_OVR_VAL));
+	dev_info(adev->dev, "  LCAC_MC0_OVR_SEL=0x%08X\n",
+		 RREG32_SMC(ixLCAC_MC0_OVR_SEL));
+	dev_info(adev->dev, "  LCAC_MC0_OVR_VAL=0x%08X\n",
+		 RREG32_SMC(ixLCAC_MC0_OVR_VAL));
+	dev_info(adev->dev, "  LCAC_MC1_OVR_SEL=0x%08X\n",
+		 RREG32_SMC(ixLCAC_MC1_OVR_SEL));
+	dev_info(adev->dev, "  LCAC_MC1_OVR_VAL=0x%08X\n",
+		 RREG32_SMC(ixLCAC_MC1_OVR_VAL));
+	dev_info(adev->dev, "  LCAC_MC2_OVR_SEL=0x%08X\n",
+		 RREG32_SMC(ixLCAC_MC2_OVR_SEL));
+	dev_info(adev->dev, "  LCAC_MC2_OVR_VAL=0x%08X\n",
+		 RREG32_SMC(ixLCAC_MC2_OVR_VAL));
+	dev_info(adev->dev, "  LCAC_MC3_OVR_SEL=0x%08X\n",
+		 RREG32_SMC(ixLCAC_MC3_OVR_SEL));
+	dev_info(adev->dev, "  LCAC_MC3_OVR_VAL=0x%08X\n",
+		 RREG32_SMC(ixLCAC_MC3_OVR_VAL));
+	dev_info(adev->dev, "  LCAC_CPL_OVR_SEL=0x%08X\n",
+		 RREG32_SMC(ixLCAC_CPL_OVR_SEL));
+	dev_info(adev->dev, "  LCAC_CPL_OVR_VAL=0x%08X\n",
+		 RREG32_SMC(ixLCAC_CPL_OVR_VAL));
+	dev_info(adev->dev, "  CG_FREQ_TRAN_VOTING_0=0x%08X\n",
+		 RREG32_SMC(ixCG_FREQ_TRAN_VOTING_0));
+	dev_info(adev->dev, "  GENERAL_PWRMGT=0x%08X\n",
+		 RREG32_SMC(ixGENERAL_PWRMGT));
+	dev_info(adev->dev, "  SCLK_PWRMGT_CNTL=0x%08X\n",
+		 RREG32_SMC(ixSCLK_PWRMGT_CNTL));
+	dev_info(adev->dev, "  SMC_MESSAGE_0=0x%08X\n",
+		 RREG32(mmSMC_MESSAGE_0));
+	dev_info(adev->dev, "  SMC_RESP_0=0x%08X\n",
+		 RREG32(mmSMC_RESP_0));
+	dev_info(adev->dev, "  SMC_MSG_ARG_0=0x%08X\n",
+		 RREG32(mmSMC_MSG_ARG_0));
+	dev_info(adev->dev, "  SMC_IND_INDEX_0=0x%08X\n",
+		 RREG32(mmSMC_IND_INDEX_0));
+	dev_info(adev->dev, "  SMC_IND_DATA_0=0x%08X\n",
+		 RREG32(mmSMC_IND_DATA_0));
+	dev_info(adev->dev, "  SMC_IND_ACCESS_CNTL=0x%08X\n",
+		 RREG32(mmSMC_IND_ACCESS_CNTL));
+}
+
+static int kv_dpm_soft_reset(struct amdgpu_device *adev)
+{
+	return 0;
+}
+
+static int kv_dpm_set_interrupt_state(struct amdgpu_device *adev,
+				      struct amdgpu_irq_src *src,
+				      unsigned type,
+				      enum amdgpu_interrupt_state state)
+{
+	u32 cg_thermal_int;
+
+	switch (type) {
+	case AMDGPU_THERMAL_IRQ_LOW_TO_HIGH:
+		switch (state) {
+		case AMDGPU_IRQ_STATE_DISABLE:
+			cg_thermal_int = RREG32_SMC(ixCG_THERMAL_INT_CTRL);
+			cg_thermal_int &= ~CG_THERMAL_INT_CTRL__THERM_INTH_MASK_MASK;
+			WREG32_SMC(ixCG_THERMAL_INT_CTRL, cg_thermal_int);
+			break;
+		case AMDGPU_IRQ_STATE_ENABLE:
+			cg_thermal_int = RREG32_SMC(ixCG_THERMAL_INT_CTRL);
+			cg_thermal_int |= CG_THERMAL_INT_CTRL__THERM_INTH_MASK_MASK;
+			WREG32_SMC(ixCG_THERMAL_INT_CTRL, cg_thermal_int);
+			break;
+		default:
+			break;
+		}
+		break;
+
+	case AMDGPU_THERMAL_IRQ_HIGH_TO_LOW:
+		switch (state) {
+		case AMDGPU_IRQ_STATE_DISABLE:
+			cg_thermal_int = RREG32_SMC(ixCG_THERMAL_INT_CTRL);
+			cg_thermal_int &= ~CG_THERMAL_INT_CTRL__THERM_INTL_MASK_MASK;
+			WREG32_SMC(ixCG_THERMAL_INT_CTRL, cg_thermal_int);
+			break;
+		case AMDGPU_IRQ_STATE_ENABLE:
+			cg_thermal_int = RREG32_SMC(ixCG_THERMAL_INT_CTRL);
+			cg_thermal_int |= CG_THERMAL_INT_CTRL__THERM_INTL_MASK_MASK;
+			WREG32_SMC(ixCG_THERMAL_INT_CTRL, cg_thermal_int);
+			break;
+		default:
+			break;
+		}
+		break;
+
+	default:
+		break;
+	}
+	return 0;
+}
+
+static int kv_dpm_process_interrupt(struct amdgpu_device *adev,
+				    struct amdgpu_irq_src *source,
+				    struct amdgpu_iv_entry *entry)
+{
+	bool queue_thermal = false;
+
+	if (entry == NULL)
+		return -EINVAL;
+
+	switch (entry->src_id) {
+	case 230: /* thermal low to high */
+		DRM_DEBUG("IH: thermal low to high\n");
+		adev->pm.dpm.thermal.high_to_low = false;
+		queue_thermal = true;
+		break;
+	case 231: /* thermal high to low */
+		DRM_DEBUG("IH: thermal high to low\n");
+		adev->pm.dpm.thermal.high_to_low = true;
+		queue_thermal = true;
+		break;
+	default:
+		break;
+	}
+
+	if (queue_thermal)
+		schedule_work(&adev->pm.dpm.thermal.work);
+
+	return 0;
+}
+
+static int kv_dpm_set_clockgating_state(struct amdgpu_device *adev,
+					  enum amdgpu_clockgating_state state)
+{
+	return 0;
+}
+
+static int kv_dpm_set_powergating_state(struct amdgpu_device *adev,
+					  enum amdgpu_powergating_state state)
+{
+	return 0;
+}
+
+const struct amdgpu_ip_funcs kv_dpm_ip_funcs = {
+	.early_init = kv_dpm_early_init,
+	.late_init = kv_dpm_late_init,
+	.sw_init = kv_dpm_sw_init,
+	.sw_fini = kv_dpm_sw_fini,
+	.hw_init = kv_dpm_hw_init,
+	.hw_fini = kv_dpm_hw_fini,
+	.suspend = kv_dpm_suspend,
+	.resume = kv_dpm_resume,
+	.is_idle = kv_dpm_is_idle,
+	.wait_for_idle = kv_dpm_wait_for_idle,
+	.soft_reset = kv_dpm_soft_reset,
+	.print_status = kv_dpm_print_status,
+	.set_clockgating_state = kv_dpm_set_clockgating_state,
+	.set_powergating_state = kv_dpm_set_powergating_state,
+};
+
+static const struct amdgpu_dpm_funcs kv_dpm_funcs = {
+	.get_temperature = &kv_dpm_get_temp,
+	.pre_set_power_state = &kv_dpm_pre_set_power_state,
+	.set_power_state = &kv_dpm_set_power_state,
+	.post_set_power_state = &kv_dpm_post_set_power_state,
+	.display_configuration_changed = &kv_dpm_display_configuration_changed,
+	.get_sclk = &kv_dpm_get_sclk,
+	.get_mclk = &kv_dpm_get_mclk,
+	.print_power_state = &kv_dpm_print_power_state,
+	.debugfs_print_current_performance_level = &kv_dpm_debugfs_print_current_performance_level,
+	.force_performance_level = &kv_dpm_force_performance_level,
+	.powergate_uvd = &kv_dpm_powergate_uvd,
+	.enable_bapm = &kv_dpm_enable_bapm,
+};
+
+static void kv_dpm_set_dpm_funcs(struct amdgpu_device *adev)
+{
+	if (adev->pm.funcs == NULL)
+		adev->pm.funcs = &kv_dpm_funcs;
+}
+
+static const struct amdgpu_irq_src_funcs kv_dpm_irq_funcs = {
+	.set = kv_dpm_set_interrupt_state,
+	.process = kv_dpm_process_interrupt,
+};
+
+static void kv_dpm_set_irq_funcs(struct amdgpu_device *adev)
+{
+	adev->pm.dpm.thermal.irq.num_types = AMDGPU_THERMAL_IRQ_LAST;
+	adev->pm.dpm.thermal.irq.funcs = &kv_dpm_irq_funcs;
+}

drivers/gpu/drm/amd/amdgpu/kv_dpm.h

+/*
+ * Copyright 2013 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+#ifndef __KV_DPM_H__
+#define __KV_DPM_H__
+
+#define SMU__NUM_SCLK_DPM_STATE  8
+#define SMU__NUM_MCLK_DPM_LEVELS 4
+#define SMU__NUM_LCLK_DPM_LEVELS 8
+#define SMU__NUM_PCIE_DPM_LEVELS 0 /* ??? */
+#include "smu7_fusion.h"
+#include "ppsmc.h"
+
+#define SUMO_MAX_HARDWARE_POWERLEVELS 5
+
+#define SUMO_MAX_NUMBER_VOLTAGES    4
+
+struct sumo_vid_mapping_entry {
+	u16 vid_2bit;
+	u16 vid_7bit;
+};
+
+struct sumo_vid_mapping_table {
+	u32 num_entries;
+	struct sumo_vid_mapping_entry entries[SUMO_MAX_NUMBER_VOLTAGES];
+};
+
+struct sumo_sclk_voltage_mapping_entry {
+	u32 sclk_frequency;
+	u16 vid_2bit;
+	u16 rsv;
+};
+
+struct sumo_sclk_voltage_mapping_table {
+	u32 num_max_dpm_entries;
+	struct sumo_sclk_voltage_mapping_entry entries[SUMO_MAX_HARDWARE_POWERLEVELS];
+};
+
+#define TRINITY_AT_DFLT            30
+
+#define KV_NUM_NBPSTATES   4
+
+enum kv_pt_config_reg_type {
+	KV_CONFIGREG_MMR = 0,
+	KV_CONFIGREG_SMC_IND,
+	KV_CONFIGREG_DIDT_IND,
+	KV_CONFIGREG_CACHE,
+	KV_CONFIGREG_MAX
+};
+
+struct kv_pt_config_reg {
+	u32 offset;
+	u32 mask;
+	u32 shift;
+	u32 value;
+	enum kv_pt_config_reg_type type;
+};
+
+struct kv_lcac_config_values {
+	u32 block_id;
+	u32 signal_id;
+	u32 t;
+};
+
+struct kv_lcac_config_reg {
+	u32 cntl;
+	u32 block_mask;
+	u32 block_shift;
+	u32 signal_mask;
+	u32 signal_shift;
+	u32 t_mask;
+	u32 t_shift;
+	u32 enable_mask;
+	u32 enable_shift;
+};
+
+struct kv_pl {
+	u32 sclk;
+	u8 vddc_index;
+	u8 ds_divider_index;
+	u8 ss_divider_index;
+	u8 allow_gnb_slow;
+	u8 force_nbp_state;
+	u8 display_wm;
+	u8 vce_wm;
+};
+
+struct kv_ps {
+	struct kv_pl levels[SUMO_MAX_HARDWARE_POWERLEVELS];
+	u32 num_levels;
+	bool need_dfs_bypass;
+	u8 dpm0_pg_nb_ps_lo;
+	u8 dpm0_pg_nb_ps_hi;
+	u8 dpmx_nb_ps_lo;
+	u8 dpmx_nb_ps_hi;
+};
+
+struct kv_sys_info {
+	u32 bootup_uma_clk;
+	u32 bootup_sclk;
+	u32 dentist_vco_freq;
+	u32 nb_dpm_enable;
+	u32 nbp_memory_clock[KV_NUM_NBPSTATES];
+	u32 nbp_n_clock[KV_NUM_NBPSTATES];
+	u16 bootup_nb_voltage_index;
+	u8 htc_tmp_lmt;
+	u8 htc_hyst_lmt;
+	struct sumo_sclk_voltage_mapping_table sclk_voltage_mapping_table;
+	struct sumo_vid_mapping_table vid_mapping_table;
+	u32 uma_channel_number;
+};
+
+struct kv_power_info {
+	u32 at[SUMO_MAX_HARDWARE_POWERLEVELS];
+	u32 voltage_drop_t;
+	struct kv_sys_info sys_info;
+	struct kv_pl boot_pl;
+	bool enable_nb_ps_policy;
+	bool disable_nb_ps3_in_battery;
+	bool video_start;
+	bool battery_state;
+	u32 lowest_valid;
+	u32 highest_valid;
+	u16 high_voltage_t;
+	bool cac_enabled;
+	bool bapm_enable;
+	/* smc offsets */
+	u32 sram_end;
+	u32 dpm_table_start;
+	u32 soft_regs_start;
+	/* dpm SMU tables */
+	u8 graphics_dpm_level_count;
+	u8 uvd_level_count;
+	u8 vce_level_count;
+	u8 acp_level_count;
+	u8 samu_level_count;
+	u16 fps_high_t;
+	SMU7_Fusion_GraphicsLevel graphics_level[SMU__NUM_SCLK_DPM_STATE];
+	SMU7_Fusion_ACPILevel acpi_level;
+	SMU7_Fusion_UvdLevel uvd_level[SMU7_MAX_LEVELS_UVD];
+	SMU7_Fusion_ExtClkLevel vce_level[SMU7_MAX_LEVELS_VCE];
+	SMU7_Fusion_ExtClkLevel acp_level[SMU7_MAX_LEVELS_ACP];
+	SMU7_Fusion_ExtClkLevel samu_level[SMU7_MAX_LEVELS_SAMU];
+	u8 uvd_boot_level;
+	u8 vce_boot_level;
+	u8 acp_boot_level;
+	u8 samu_boot_level;
+	u8 uvd_interval;
+	u8 vce_interval;
+	u8 acp_interval;
+	u8 samu_interval;
+	u8 graphics_boot_level;
+	u8 graphics_interval;
+	u8 graphics_therm_throttle_enable;
+	u8 graphics_voltage_change_enable;
+	u8 graphics_clk_slow_enable;
+	u8 graphics_clk_slow_divider;
+	u8 fps_low_t;
+	u32 low_sclk_interrupt_t;
+	bool uvd_power_gated;
+	bool vce_power_gated;
+	bool acp_power_gated;
+	bool samu_power_gated;
+	bool nb_dpm_enabled;
+	/* flags */
+	bool enable_didt;
+	bool enable_dpm;
+	bool enable_auto_thermal_throttling;
+	bool enable_nb_dpm;
+	/* caps */
+	bool caps_cac;
+	bool caps_power_containment;
+	bool caps_sq_ramping;
+	bool caps_db_ramping;
+	bool caps_td_ramping;
+	bool caps_tcp_ramping;
+	bool caps_sclk_throttle_low_notification;
+	bool caps_fps;
+	bool caps_uvd_dpm;
+	bool caps_uvd_pg;
+	bool caps_vce_pg;
+	bool caps_samu_pg;
+	bool caps_acp_pg;
+	bool caps_stable_p_state;
+	bool caps_enable_dfs_bypass;
+	bool caps_sclk_ds;
+	struct amdgpu_ps current_rps;
+	struct kv_ps current_ps;
+	struct amdgpu_ps requested_rps;
+	struct kv_ps requested_ps;
+};
+
+/* XXX are these ok? */
+#define KV_TEMP_RANGE_MIN (90 * 1000)
+#define KV_TEMP_RANGE_MAX (120 * 1000)
+
+/* kv_smc.c */
+int amdgpu_kv_notify_message_to_smu(struct amdgpu_device *adev, u32 id);
+int amdgpu_kv_dpm_get_enable_mask(struct amdgpu_device *adev, u32 *enable_mask);
+int amdgpu_kv_send_msg_to_smc_with_parameter(struct amdgpu_device *adev,
+				      PPSMC_Msg msg, u32 parameter);
+int amdgpu_kv_read_smc_sram_dword(struct amdgpu_device *adev, u32 smc_address,
+			   u32 *value, u32 limit);
+int amdgpu_kv_smc_dpm_enable(struct amdgpu_device *adev, bool enable);
+int amdgpu_kv_smc_bapm_enable(struct amdgpu_device *adev, bool enable);
+int amdgpu_kv_copy_bytes_to_smc(struct amdgpu_device *adev,
+			 u32 smc_start_address,
+			 const u8 *src, u32 byte_count, u32 limit);
+
+#endif

drivers/gpu/drm/amd/amdgpu/kv_smc.c

+/*
+ * Copyright 2013 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Alex Deucher
+ */
+
+#include "drmP.h"
+#include "amdgpu.h"
+#include "cikd.h"
+#include "kv_dpm.h"
+
+#include "smu/smu_7_0_0_d.h"
+#include "smu/smu_7_0_0_sh_mask.h"
+
+int amdgpu_kv_notify_message_to_smu(struct amdgpu_device *adev, u32 id)
+{
+	u32 i;
+	u32 tmp = 0;
+
+	WREG32(mmSMC_MESSAGE_0, id & SMC_MESSAGE_0__SMC_MSG_MASK);
+
+	for (i = 0; i < adev->usec_timeout; i++) {
+		if ((RREG32(mmSMC_RESP_0) & SMC_RESP_0__SMC_RESP_MASK) != 0)
+			break;
+		udelay(1);
+	}
+	tmp = RREG32(mmSMC_RESP_0) & SMC_RESP_0__SMC_RESP_MASK;
+
+	if (tmp != 1) {
+		if (tmp == 0xFF)
+			return -EINVAL;
+		else if (tmp == 0xFE)
+			return -EINVAL;
+	}
+
+	return 0;
+}
+
+int amdgpu_kv_dpm_get_enable_mask(struct amdgpu_device *adev, u32 *enable_mask)
+{
+	int ret;
+
+	ret = amdgpu_kv_notify_message_to_smu(adev, PPSMC_MSG_SCLKDPM_GetEnabledMask);
+
+	if (ret == 0)
+		*enable_mask = RREG32_SMC(ixSMC_SYSCON_MSG_ARG_0);
+
+	return ret;
+}
+
+int amdgpu_kv_send_msg_to_smc_with_parameter(struct amdgpu_device *adev,
+				      PPSMC_Msg msg, u32 parameter)
+{
+
+	WREG32(mmSMC_MSG_ARG_0, parameter);
+
+	return amdgpu_kv_notify_message_to_smu(adev, msg);
+}
+
+static int kv_set_smc_sram_address(struct amdgpu_device *adev,
+				   u32 smc_address, u32 limit)
+{
+	if (smc_address & 3)
+		return -EINVAL;
+	if ((smc_address + 3) > limit)
+		return -EINVAL;
+
+	WREG32(mmSMC_IND_INDEX_0, smc_address);
+	WREG32_P(mmSMC_IND_ACCESS_CNTL, 0,
+			~SMC_IND_ACCESS_CNTL__AUTO_INCREMENT_IND_0_MASK);
+
+	return 0;
+}
+
+int amdgpu_kv_read_smc_sram_dword(struct amdgpu_device *adev, u32 smc_address,
+			   u32 *value, u32 limit)
+{
+	int ret;
+
+	ret = kv_set_smc_sram_address(adev, smc_address, limit);
+	if (ret)
+		return ret;
+
+	*value = RREG32(mmSMC_IND_DATA_0);
+	return 0;
+}
+
+int amdgpu_kv_smc_dpm_enable(struct amdgpu_device *adev, bool enable)
+{
+	if (enable)
+		return amdgpu_kv_notify_message_to_smu(adev, PPSMC_MSG_DPM_Enable);
+	else
+		return amdgpu_kv_notify_message_to_smu(adev, PPSMC_MSG_DPM_Disable);
+}
+
+int amdgpu_kv_smc_bapm_enable(struct amdgpu_device *adev, bool enable)
+{
+	if (enable)
+		return amdgpu_kv_notify_message_to_smu(adev, PPSMC_MSG_EnableBAPM);
+	else
+		return amdgpu_kv_notify_message_to_smu(adev, PPSMC_MSG_DisableBAPM);
+}
+
+int amdgpu_kv_copy_bytes_to_smc(struct amdgpu_device *adev,
+			 u32 smc_start_address,
+			 const u8 *src, u32 byte_count, u32 limit)
+{
+	int ret;
+	u32 data, original_data, addr, extra_shift, t_byte, count, mask;
+
+	if ((smc_start_address + byte_count) > limit)
+		return -EINVAL;
+
+	addr = smc_start_address;
+	t_byte = addr & 3;
+
+	/* RMW for the initial bytes */
+	if  (t_byte != 0) {
+		addr -= t_byte;
+
+		ret = kv_set_smc_sram_address(adev, addr, limit);
+		if (ret)
+			return ret;
+
+		original_data = RREG32(mmSMC_IND_DATA_0);
+
+		data = 0;
+		mask = 0;
+		count = 4;
+		while (count > 0) {
+			if (t_byte > 0) {
+				mask = (mask << 8) | 0xff;
+				t_byte--;
+			} else if (byte_count > 0) {
+				data = (data << 8) + *src++;
+				byte_count--;
+				mask <<= 8;
+			} else {
+				data <<= 8;
+				mask = (mask << 8) | 0xff;
+			}
+			count--;
+		}
+
+		data |= original_data & mask;
+
+		ret = kv_set_smc_sram_address(adev, addr, limit);
+		if (ret)
+			return ret;
+
+		WREG32(mmSMC_IND_DATA_0, data);
+
+		addr += 4;
+	}
+
+	while (byte_count >= 4) {
+		/* SMC address space is BE */
+		data = (src[0] << 24) + (src[1] << 16) + (src[2] << 8) + src[3];
+
+		ret = kv_set_smc_sram_address(adev, addr, limit);
+		if (ret)
+			return ret;
+
+		WREG32(mmSMC_IND_DATA_0, data);
+
+		src += 4;
+		byte_count -= 4;
+		addr += 4;
+	}
+
+	/* RMW for the final bytes */
+	if (byte_count > 0) {
+		data = 0;
+
+		ret = kv_set_smc_sram_address(adev, addr, limit);
+		if (ret)
+			return ret;
+
+		original_data = RREG32(mmSMC_IND_DATA_0);
+
+		extra_shift = 8 * (4 - byte_count);
+
+		while (byte_count > 0) {
+			/* SMC address space is BE */
+			data = (data << 8) + *src++;
+			byte_count--;
+		}
+
+		data <<= extra_shift;
+
+		data |= (original_data & ~((~0UL) << extra_shift));
+
+		ret = kv_set_smc_sram_address(adev, addr, limit);
+		if (ret)
+			return ret;
+
+		WREG32(mmSMC_IND_DATA_0, data);
+	}
+	return 0;
+}
+

drivers/gpu/drm/amd/amdgpu/smu7.h

+/*
+ * Copyright 2013 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef SMU7_H
+#define SMU7_H
+
+#pragma pack(push, 1)
+
+#define SMU7_CONTEXT_ID_SMC        1
+#define SMU7_CONTEXT_ID_VBIOS      2
+
+
+#define SMU7_CONTEXT_ID_SMC        1
+#define SMU7_CONTEXT_ID_VBIOS      2
+
+#define SMU7_MAX_LEVELS_VDDC            8
+#define SMU7_MAX_LEVELS_VDDCI           4
+#define SMU7_MAX_LEVELS_MVDD            4
+#define SMU7_MAX_LEVELS_VDDNB           8
+
+#define SMU7_MAX_LEVELS_GRAPHICS        SMU__NUM_SCLK_DPM_STATE   // SCLK + SQ DPM + ULV
+#define SMU7_MAX_LEVELS_MEMORY          SMU__NUM_MCLK_DPM_LEVELS   // MCLK Levels DPM
+#define SMU7_MAX_LEVELS_GIO             SMU__NUM_LCLK_DPM_LEVELS  // LCLK Levels
+#define SMU7_MAX_LEVELS_LINK            SMU__NUM_PCIE_DPM_LEVELS  // PCIe speed and number of lanes.
+#define SMU7_MAX_LEVELS_UVD             8   // VCLK/DCLK levels for UVD.
+#define SMU7_MAX_LEVELS_VCE             8   // ECLK levels for VCE.
+#define SMU7_MAX_LEVELS_ACP             8   // ACLK levels for ACP.
+#define SMU7_MAX_LEVELS_SAMU            8   // SAMCLK levels for SAMU.
+#define SMU7_MAX_ENTRIES_SMIO           32  // Number of entries in SMIO table.
+
+#define DPM_NO_LIMIT 0
+#define DPM_NO_UP 1
+#define DPM_GO_DOWN 2
+#define DPM_GO_UP 3
+
+#define SMU7_FIRST_DPM_GRAPHICS_LEVEL    0
+#define SMU7_FIRST_DPM_MEMORY_LEVEL      0
+
+#define GPIO_CLAMP_MODE_VRHOT      1
+#define GPIO_CLAMP_MODE_THERM      2
+#define GPIO_CLAMP_MODE_DC         4
+
+#define SCRATCH_B_TARG_PCIE_INDEX_SHIFT 0
+#define SCRATCH_B_TARG_PCIE_INDEX_MASK  (0x7<<SCRATCH_B_TARG_PCIE_INDEX_SHIFT)
+#define SCRATCH_B_CURR_PCIE_INDEX_SHIFT 3
+#define SCRATCH_B_CURR_PCIE_INDEX_MASK  (0x7<<SCRATCH_B_CURR_PCIE_INDEX_SHIFT)
+#define SCRATCH_B_TARG_UVD_INDEX_SHIFT  6
+#define SCRATCH_B_TARG_UVD_INDEX_MASK   (0x7<<SCRATCH_B_TARG_UVD_INDEX_SHIFT)
+#define SCRATCH_B_CURR_UVD_INDEX_SHIFT  9
+#define SCRATCH_B_CURR_UVD_INDEX_MASK   (0x7<<SCRATCH_B_CURR_UVD_INDEX_SHIFT)
+#define SCRATCH_B_TARG_VCE_INDEX_SHIFT  12
+#define SCRATCH_B_TARG_VCE_INDEX_MASK   (0x7<<SCRATCH_B_TARG_VCE_INDEX_SHIFT)
+#define SCRATCH_B_CURR_VCE_INDEX_SHIFT  15
+#define SCRATCH_B_CURR_VCE_INDEX_MASK   (0x7<<SCRATCH_B_CURR_VCE_INDEX_SHIFT)
+#define SCRATCH_B_TARG_ACP_INDEX_SHIFT  18
+#define SCRATCH_B_TARG_ACP_INDEX_MASK   (0x7<<SCRATCH_B_TARG_ACP_INDEX_SHIFT)
+#define SCRATCH_B_CURR_ACP_INDEX_SHIFT  21
+#define SCRATCH_B_CURR_ACP_INDEX_MASK   (0x7<<SCRATCH_B_CURR_ACP_INDEX_SHIFT)
+#define SCRATCH_B_TARG_SAMU_INDEX_SHIFT 24
+#define SCRATCH_B_TARG_SAMU_INDEX_MASK  (0x7<<SCRATCH_B_TARG_SAMU_INDEX_SHIFT)
+#define SCRATCH_B_CURR_SAMU_INDEX_SHIFT 27
+#define SCRATCH_B_CURR_SAMU_INDEX_MASK  (0x7<<SCRATCH_B_CURR_SAMU_INDEX_SHIFT)
+
+
+struct SMU7_PIDController
+{
+    uint32_t Ki;
+    int32_t LFWindupUL;
+    int32_t LFWindupLL;
+    uint32_t StatePrecision;
+    uint32_t LfPrecision;
+    uint32_t LfOffset;
+    uint32_t MaxState;
+    uint32_t MaxLfFraction;
+    uint32_t StateShift;
+};
+
+typedef struct SMU7_PIDController SMU7_PIDController;
+
+// -------------------------------------------------------------------------------------------------------------------------
+#define SMU7_MAX_PCIE_LINK_SPEEDS 3 /* 0:Gen1 1:Gen2 2:Gen3 */
+
+#define SMU7_SCLK_DPM_CONFIG_MASK                        0x01
+#define SMU7_VOLTAGE_CONTROLLER_CONFIG_MASK              0x02
+#define SMU7_THERMAL_CONTROLLER_CONFIG_MASK              0x04
+#define SMU7_MCLK_DPM_CONFIG_MASK                        0x08
+#define SMU7_UVD_DPM_CONFIG_MASK                         0x10
+#define SMU7_VCE_DPM_CONFIG_MASK                         0x20
+#define SMU7_ACP_DPM_CONFIG_MASK                         0x40
+#define SMU7_SAMU_DPM_CONFIG_MASK                        0x80
+#define SMU7_PCIEGEN_DPM_CONFIG_MASK                    0x100
+
+#define SMU7_ACP_MCLK_HANDSHAKE_DISABLE                  0x00000001
+#define SMU7_ACP_SCLK_HANDSHAKE_DISABLE                  0x00000002
+#define SMU7_UVD_MCLK_HANDSHAKE_DISABLE                  0x00000100
+#define SMU7_UVD_SCLK_HANDSHAKE_DISABLE                  0x00000200
+#define SMU7_VCE_MCLK_HANDSHAKE_DISABLE                  0x00010000
+#define SMU7_VCE_SCLK_HANDSHAKE_DISABLE                  0x00020000
+
+struct SMU7_Firmware_Header
+{
+    uint32_t Digest[5];
+    uint32_t Version;
+    uint32_t HeaderSize;
+    uint32_t Flags;
+    uint32_t EntryPoint;
+    uint32_t CodeSize;
+    uint32_t ImageSize;
+
+    uint32_t Rtos;
+    uint32_t SoftRegisters;
+    uint32_t DpmTable;
+    uint32_t FanTable;
+    uint32_t CacConfigTable;
+    uint32_t CacStatusTable;
+
+    uint32_t mcRegisterTable;
+
+    uint32_t mcArbDramTimingTable;
+
+    uint32_t PmFuseTable;
+    uint32_t Globals;
+    uint32_t Reserved[42];
+    uint32_t Signature;
+};
+
+typedef struct SMU7_Firmware_Header SMU7_Firmware_Header;
+
+#define SMU7_FIRMWARE_HEADER_LOCATION 0x20000
+
+enum  DisplayConfig {
+    PowerDown = 1,
+    DP54x4,
+    DP54x2,
+    DP54x1,
+    DP27x4,
+    DP27x2,
+    DP27x1,
+    HDMI297,
+    HDMI162,
+    LVDS,
+    DP324x4,
+    DP324x2,
+    DP324x1
+};
+
+#pragma pack(pop)
+
+#endif
+

drivers/gpu/drm/amd/amdgpu/smu7_discrete.h

+/*
+ * Copyright 2013 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef SMU7_DISCRETE_H
+#define SMU7_DISCRETE_H
+
+#include "smu7.h"
+
+#pragma pack(push, 1)
+
+#define SMU7_DTE_ITERATIONS 5
+#define SMU7_DTE_SOURCES 3
+#define SMU7_DTE_SINKS 1
+#define SMU7_NUM_CPU_TES 0
+#define SMU7_NUM_GPU_TES 1
+#define SMU7_NUM_NON_TES 2
+
+struct SMU7_SoftRegisters
+{
+    uint32_t        RefClockFrequency;
+    uint32_t        PmTimerP;
+    uint32_t        FeatureEnables;
+    uint32_t        PreVBlankGap;
+    uint32_t        VBlankTimeout;
+    uint32_t        TrainTimeGap;
+
+    uint32_t        MvddSwitchTime;
+    uint32_t        LongestAcpiTrainTime;
+    uint32_t        AcpiDelay;
+    uint32_t        G5TrainTime;
+    uint32_t        DelayMpllPwron;
+    uint32_t        VoltageChangeTimeout;
+    uint32_t        HandshakeDisables;
+
+    uint8_t         DisplayPhy1Config;
+    uint8_t         DisplayPhy2Config;
+    uint8_t         DisplayPhy3Config;
+    uint8_t         DisplayPhy4Config;
+
+    uint8_t         DisplayPhy5Config;
+    uint8_t         DisplayPhy6Config;
+    uint8_t         DisplayPhy7Config;
+    uint8_t         DisplayPhy8Config;
+
+    uint32_t        AverageGraphicsA;
+    uint32_t        AverageMemoryA;
+    uint32_t        AverageGioA;
+
+    uint8_t         SClkDpmEnabledLevels;
+    uint8_t         MClkDpmEnabledLevels;
+    uint8_t         LClkDpmEnabledLevels;
+    uint8_t         PCIeDpmEnabledLevels;
+
+    uint8_t         UVDDpmEnabledLevels;
+    uint8_t         SAMUDpmEnabledLevels;
+    uint8_t         ACPDpmEnabledLevels;
+    uint8_t         VCEDpmEnabledLevels;
+
+    uint32_t        DRAM_LOG_ADDR_H;
+    uint32_t        DRAM_LOG_ADDR_L;
+    uint32_t        DRAM_LOG_PHY_ADDR_H;
+    uint32_t        DRAM_LOG_PHY_ADDR_L;
+    uint32_t        DRAM_LOG_BUFF_SIZE;
+    uint32_t        UlvEnterC;
+    uint32_t        UlvTime;
+    uint32_t        Reserved[3];
+
+};
+
+typedef struct SMU7_SoftRegisters SMU7_SoftRegisters;
+
+struct SMU7_Discrete_VoltageLevel
+{
+    uint16_t    Voltage;
+    uint16_t    StdVoltageHiSidd;
+    uint16_t    StdVoltageLoSidd;
+    uint8_t     Smio;
+    uint8_t     padding;
+};
+
+typedef struct SMU7_Discrete_VoltageLevel SMU7_Discrete_VoltageLevel;
+
+struct SMU7_Discrete_GraphicsLevel
+{
+    uint32_t    Flags;
+    uint32_t    MinVddc;
+    uint32_t    MinVddcPhases;
+
+    uint32_t    SclkFrequency;
+
+    uint8_t     padding1[2];
+    uint16_t    ActivityLevel;
+
+    uint32_t    CgSpllFuncCntl3;
+    uint32_t    CgSpllFuncCntl4;
+    uint32_t    SpllSpreadSpectrum;
+    uint32_t    SpllSpreadSpectrum2;
+    uint32_t    CcPwrDynRm;
+    uint32_t    CcPwrDynRm1;
+    uint8_t     SclkDid;
+    uint8_t     DisplayWatermark;
+    uint8_t     EnabledForActivity;
+    uint8_t     EnabledForThrottle;
+    uint8_t     UpH;
+    uint8_t     DownH;
+    uint8_t     VoltageDownH;
+    uint8_t     PowerThrottle;
+    uint8_t     DeepSleepDivId;
+    uint8_t     padding[3];
+};
+
+typedef struct SMU7_Discrete_GraphicsLevel SMU7_Discrete_GraphicsLevel;
+
+struct SMU7_Discrete_ACPILevel
+{
+    uint32_t    Flags;
+    uint32_t    MinVddc;
+    uint32_t    MinVddcPhases;
+    uint32_t    SclkFrequency;
+    uint8_t     SclkDid;
+    uint8_t     DisplayWatermark;
+    uint8_t     DeepSleepDivId;
+    uint8_t     padding;
+    uint32_t    CgSpllFuncCntl;
+    uint32_t    CgSpllFuncCntl2;
+    uint32_t    CgSpllFuncCntl3;
+    uint32_t    CgSpllFuncCntl4;
+    uint32_t    SpllSpreadSpectrum;
+    uint32_t    SpllSpreadSpectrum2;
+    uint32_t    CcPwrDynRm;
+    uint32_t    CcPwrDynRm1;
+};
+
+typedef struct SMU7_Discrete_ACPILevel SMU7_Discrete_ACPILevel;
+
+struct SMU7_Discrete_Ulv
+{
+    uint32_t    CcPwrDynRm;
+    uint32_t    CcPwrDynRm1;
+    uint16_t    VddcOffset;
+    uint8_t     VddcOffsetVid;
+    uint8_t     VddcPhase;
+    uint32_t    Reserved;
+};
+
+typedef struct SMU7_Discrete_Ulv SMU7_Discrete_Ulv;
+
+struct SMU7_Discrete_MemoryLevel
+{
+    uint32_t    MinVddc;
+    uint32_t    MinVddcPhases;
+    uint32_t    MinVddci;
+    uint32_t    MinMvdd;
+
+    uint32_t    MclkFrequency;
+
+    uint8_t     EdcReadEnable;
+    uint8_t     EdcWriteEnable;
+    uint8_t     RttEnable;
+    uint8_t     StutterEnable;
+
+    uint8_t     StrobeEnable;
+    uint8_t     StrobeRatio;
+    uint8_t     EnabledForThrottle;
+    uint8_t     EnabledForActivity;
+
+    uint8_t     UpH;
+    uint8_t     DownH;
+    uint8_t     VoltageDownH;
+    uint8_t     padding;
+
+    uint16_t    ActivityLevel;
+    uint8_t     DisplayWatermark;
+    uint8_t     padding1;
+
+    uint32_t    MpllFuncCntl;
+    uint32_t    MpllFuncCntl_1;
+    uint32_t    MpllFuncCntl_2;
+    uint32_t    MpllAdFuncCntl;
+    uint32_t    MpllDqFuncCntl;
+    uint32_t    MclkPwrmgtCntl;
+    uint32_t    DllCntl;
+    uint32_t    MpllSs1;
+    uint32_t    MpllSs2;
+};
+
+typedef struct SMU7_Discrete_MemoryLevel SMU7_Discrete_MemoryLevel;
+
+struct SMU7_Discrete_LinkLevel
+{
+    uint8_t     PcieGenSpeed;
+    uint8_t     PcieLaneCount;
+    uint8_t     EnabledForActivity;
+    uint8_t     Padding;
+    uint32_t    DownT;
+    uint32_t    UpT;
+    uint32_t    Reserved;
+};
+
+typedef struct SMU7_Discrete_LinkLevel SMU7_Discrete_LinkLevel;
+
+
+struct SMU7_Discrete_MCArbDramTimingTableEntry
+{
+    uint32_t McArbDramTiming;
+    uint32_t McArbDramTiming2;
+    uint8_t  McArbBurstTime;
+    uint8_t  padding[3];
+};
+
+typedef struct SMU7_Discrete_MCArbDramTimingTableEntry SMU7_Discrete_MCArbDramTimingTableEntry;
+
+struct SMU7_Discrete_MCArbDramTimingTable
+{
+    SMU7_Discrete_MCArbDramTimingTableEntry entries[SMU__NUM_SCLK_DPM_STATE][SMU__NUM_MCLK_DPM_LEVELS];
+};
+
+typedef struct SMU7_Discrete_MCArbDramTimingTable SMU7_Discrete_MCArbDramTimingTable;
+
+struct SMU7_Discrete_UvdLevel
+{
+    uint32_t VclkFrequency;
+    uint32_t DclkFrequency;
+    uint16_t MinVddc;
+    uint8_t  MinVddcPhases;
+    uint8_t  VclkDivider;
+    uint8_t  DclkDivider;
+    uint8_t  padding[3];
+};
+
+typedef struct SMU7_Discrete_UvdLevel SMU7_Discrete_UvdLevel;
+
+struct SMU7_Discrete_ExtClkLevel
+{
+    uint32_t Frequency;
+    uint16_t MinVoltage;
+    uint8_t  MinPhases;
+    uint8_t  Divider;
+};
+
+typedef struct SMU7_Discrete_ExtClkLevel SMU7_Discrete_ExtClkLevel;
+
+struct SMU7_Discrete_StateInfo
+{
+    uint32_t SclkFrequency;
+    uint32_t MclkFrequency;
+    uint32_t VclkFrequency;
+    uint32_t DclkFrequency;
+    uint32_t SamclkFrequency;
+    uint32_t AclkFrequency;
+    uint32_t EclkFrequency;
+    uint16_t MvddVoltage;
+    uint16_t padding16;
+    uint8_t  DisplayWatermark;
+    uint8_t  McArbIndex;
+    uint8_t  McRegIndex;
+    uint8_t  SeqIndex;
+    uint8_t  SclkDid;
+    int8_t   SclkIndex;
+    int8_t   MclkIndex;
+    uint8_t  PCIeGen;
+
+};
+
+typedef struct SMU7_Discrete_StateInfo SMU7_Discrete_StateInfo;
+
+
+struct SMU7_Discrete_DpmTable
+{
+    SMU7_PIDController                  GraphicsPIDController;
+    SMU7_PIDController                  MemoryPIDController;
+    SMU7_PIDController                  LinkPIDController;
+
+    uint32_t                            SystemFlags;
+
+
+    uint32_t                            SmioMaskVddcVid;
+    uint32_t                            SmioMaskVddcPhase;
+    uint32_t                            SmioMaskVddciVid;
+    uint32_t                            SmioMaskMvddVid;
+
+    uint32_t                            VddcLevelCount;
+    uint32_t                            VddciLevelCount;
+    uint32_t                            MvddLevelCount;
+
+    SMU7_Discrete_VoltageLevel          VddcLevel               [SMU7_MAX_LEVELS_VDDC];
+//    SMU7_Discrete_VoltageLevel          VddcStandardReference   [SMU7_MAX_LEVELS_VDDC];
+    SMU7_Discrete_VoltageLevel          VddciLevel              [SMU7_MAX_LEVELS_VDDCI];
+    SMU7_Discrete_VoltageLevel          MvddLevel               [SMU7_MAX_LEVELS_MVDD];
+
+    uint8_t                             GraphicsDpmLevelCount;
+    uint8_t                             MemoryDpmLevelCount;
+    uint8_t                             LinkLevelCount;
+    uint8_t                             UvdLevelCount;
+    uint8_t                             VceLevelCount;
+    uint8_t                             AcpLevelCount;
+    uint8_t                             SamuLevelCount;
+    uint8_t                             MasterDeepSleepControl;
+    uint32_t                            Reserved[5];
+//    uint32_t                            SamuDefaultLevel;
+
+    SMU7_Discrete_GraphicsLevel         GraphicsLevel           [SMU7_MAX_LEVELS_GRAPHICS];
+    SMU7_Discrete_MemoryLevel           MemoryACPILevel;
+    SMU7_Discrete_MemoryLevel           MemoryLevel             [SMU7_MAX_LEVELS_MEMORY];
+    SMU7_Discrete_LinkLevel             LinkLevel               [SMU7_MAX_LEVELS_LINK];
+    SMU7_Discrete_ACPILevel             ACPILevel;
+    SMU7_Discrete_UvdLevel              UvdLevel                [SMU7_MAX_LEVELS_UVD];
+    SMU7_Discrete_ExtClkLevel           VceLevel                [SMU7_MAX_LEVELS_VCE];
+    SMU7_Discrete_ExtClkLevel           AcpLevel                [SMU7_MAX_LEVELS_ACP];
+    SMU7_Discrete_ExtClkLevel           SamuLevel               [SMU7_MAX_LEVELS_SAMU];
+    SMU7_Discrete_Ulv                   Ulv;
+
+    uint32_t                            SclkStepSize;
+    uint32_t                            Smio                    [SMU7_MAX_ENTRIES_SMIO];
+
+    uint8_t                             UvdBootLevel;
+    uint8_t                             VceBootLevel;
+    uint8_t                             AcpBootLevel;
+    uint8_t                             SamuBootLevel;
+
+    uint8_t                             UVDInterval;
+    uint8_t                             VCEInterval;
+    uint8_t                             ACPInterval;
+    uint8_t                             SAMUInterval;
+
+    uint8_t                             GraphicsBootLevel;
+    uint8_t                             GraphicsVoltageChangeEnable;
+    uint8_t                             GraphicsThermThrottleEnable;
+    uint8_t                             GraphicsInterval;
+
+    uint8_t                             VoltageInterval;
+    uint8_t                             ThermalInterval;
+    uint16_t                            TemperatureLimitHigh;
+
+    uint16_t                            TemperatureLimitLow;
+    uint8_t                             MemoryBootLevel;
+    uint8_t                             MemoryVoltageChangeEnable;
+
+    uint8_t                             MemoryInterval;
+    uint8_t                             MemoryThermThrottleEnable;
+    uint16_t                            VddcVddciDelta;
+
+    uint16_t                            VoltageResponseTime;
+    uint16_t                            PhaseResponseTime;
+
+    uint8_t                             PCIeBootLinkLevel;
+    uint8_t                             PCIeGenInterval;
+    uint8_t                             DTEInterval;
+    uint8_t                             DTEMode;
+
+    uint8_t                             SVI2Enable;
+    uint8_t                             VRHotGpio;
+    uint8_t                             AcDcGpio;
+    uint8_t                             ThermGpio;
+
+    uint16_t                            PPM_PkgPwrLimit;
+    uint16_t                            PPM_TemperatureLimit;
+
+    uint16_t                            DefaultTdp;
+    uint16_t                            TargetTdp;
+
+    uint16_t                            FpsHighT;
+    uint16_t                            FpsLowT;
+
+    uint16_t                            BAPMTI_R  [SMU7_DTE_ITERATIONS][SMU7_DTE_SOURCES][SMU7_DTE_SINKS];
+    uint16_t                            BAPMTI_RC [SMU7_DTE_ITERATIONS][SMU7_DTE_SOURCES][SMU7_DTE_SINKS];
+
+    uint8_t                             DTEAmbientTempBase;
+    uint8_t                             DTETjOffset;
+    uint8_t                             GpuTjMax;
+    uint8_t                             GpuTjHyst;
+
+    uint16_t                            BootVddc;
+    uint16_t                            BootVddci;
+
+    uint16_t                            BootMVdd;
+    uint16_t                            padding;
+
+    uint32_t                            BAPM_TEMP_GRADIENT;
+
+    uint32_t                            LowSclkInterruptT;
+};
+
+typedef struct SMU7_Discrete_DpmTable SMU7_Discrete_DpmTable;
+
+#define SMU7_DISCRETE_MC_REGISTER_ARRAY_SIZE 16
+#define SMU7_DISCRETE_MC_REGISTER_ARRAY_SET_COUNT SMU7_MAX_LEVELS_MEMORY
+
+struct SMU7_Discrete_MCRegisterAddress
+{
+    uint16_t s0;
+    uint16_t s1;
+};
+
+typedef struct SMU7_Discrete_MCRegisterAddress SMU7_Discrete_MCRegisterAddress;
+
+struct SMU7_Discrete_MCRegisterSet
+{
+    uint32_t value[SMU7_DISCRETE_MC_REGISTER_ARRAY_SIZE];
+};
+
+typedef struct SMU7_Discrete_MCRegisterSet SMU7_Discrete_MCRegisterSet;
+
+struct SMU7_Discrete_MCRegisters
+{
+    uint8_t                             last;
+    uint8_t                             reserved[3];
+    SMU7_Discrete_MCRegisterAddress     address[SMU7_DISCRETE_MC_REGISTER_ARRAY_SIZE];
+    SMU7_Discrete_MCRegisterSet         data[SMU7_DISCRETE_MC_REGISTER_ARRAY_SET_COUNT];
+};
+
+typedef struct SMU7_Discrete_MCRegisters SMU7_Discrete_MCRegisters;
+
+struct SMU7_Discrete_FanTable
+{
+	uint16_t FdoMode;
+	int16_t  TempMin;
+	int16_t  TempMed;
+	int16_t  TempMax;
+	int16_t  Slope1;
+	int16_t  Slope2;
+	int16_t  FdoMin;
+	int16_t  HystUp;
+	int16_t  HystDown;
+	int16_t  HystSlope;
+	int16_t  TempRespLim;
+	int16_t  TempCurr;
+	int16_t  SlopeCurr;
+	int16_t  PwmCurr;
+	uint32_t RefreshPeriod;
+	int16_t  FdoMax;
+	uint8_t  TempSrc;
+	int8_t   Padding;
+};
+
+typedef struct SMU7_Discrete_FanTable SMU7_Discrete_FanTable;
+
+
+struct SMU7_Discrete_PmFuses {
+  // dw0-dw1
+  uint8_t BapmVddCVidHiSidd[8];
+
+  // dw2-dw3
+  uint8_t BapmVddCVidLoSidd[8];
+
+  // dw4-dw5
+  uint8_t VddCVid[8];
+
+  // dw6
+  uint8_t SviLoadLineEn;
+  uint8_t SviLoadLineVddC;
+  uint8_t SviLoadLineTrimVddC;
+  uint8_t SviLoadLineOffsetVddC;
+
+  // dw7
+  uint16_t TDC_VDDC_PkgLimit;
+  uint8_t TDC_VDDC_ThrottleReleaseLimitPerc;
+  uint8_t TDC_MAWt;
+
+  // dw8
+  uint8_t TdcWaterfallCtl;
+  uint8_t LPMLTemperatureMin;
+  uint8_t LPMLTemperatureMax;
+  uint8_t Reserved;
+
+  // dw9-dw10
+  uint8_t BapmVddCVidHiSidd2[8];
+
+  // dw11-dw12
+  int16_t FuzzyFan_ErrorSetDelta;
+  int16_t FuzzyFan_ErrorRateSetDelta;
+  int16_t FuzzyFan_PwmSetDelta;
+  uint16_t CalcMeasPowerBlend;
+
+  // dw13-dw16
+  uint8_t GnbLPML[16];
+
+  // dw17
+  uint8_t GnbLPMLMaxVid;
+  uint8_t GnbLPMLMinVid;
+  uint8_t Reserved1[2];
+
+  // dw18
+  uint16_t BapmVddCBaseLeakageHiSidd;
+  uint16_t BapmVddCBaseLeakageLoSidd;
+};
+
+typedef struct SMU7_Discrete_PmFuses SMU7_Discrete_PmFuses;
+
+
+#pragma pack(pop)
+
+#endif
+

drivers/gpu/drm/amd/amdgpu/smu7_fusion.h

+/*
+ * Copyright 2013 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef SMU7_FUSION_H
+#define SMU7_FUSION_H
+
+#include "smu7.h"
+
+#pragma pack(push, 1)
+
+#define SMU7_DTE_ITERATIONS 5
+#define SMU7_DTE_SOURCES 5
+#define SMU7_DTE_SINKS 3
+#define SMU7_NUM_CPU_TES 2
+#define SMU7_NUM_GPU_TES 1
+#define SMU7_NUM_NON_TES 2
+
+// All 'soft registers' should be uint32_t.
+struct SMU7_SoftRegisters
+{
+    uint32_t        RefClockFrequency;
+    uint32_t        PmTimerP;
+    uint32_t        FeatureEnables;
+    uint32_t        HandshakeDisables;
+
+    uint8_t         DisplayPhy1Config;
+    uint8_t         DisplayPhy2Config;
+    uint8_t         DisplayPhy3Config;
+    uint8_t         DisplayPhy4Config;
+
+    uint8_t         DisplayPhy5Config;
+    uint8_t         DisplayPhy6Config;
+    uint8_t         DisplayPhy7Config;
+    uint8_t         DisplayPhy8Config;
+
+    uint32_t        AverageGraphicsA;
+    uint32_t        AverageMemoryA;
+    uint32_t        AverageGioA;
+
+    uint8_t         SClkDpmEnabledLevels;
+    uint8_t         MClkDpmEnabledLevels;
+    uint8_t         LClkDpmEnabledLevels;
+    uint8_t         PCIeDpmEnabledLevels;
+
+    uint8_t         UVDDpmEnabledLevels;
+    uint8_t         SAMUDpmEnabledLevels;
+    uint8_t         ACPDpmEnabledLevels;
+    uint8_t         VCEDpmEnabledLevels;
+
+    uint32_t        DRAM_LOG_ADDR_H;
+    uint32_t        DRAM_LOG_ADDR_L;
+    uint32_t        DRAM_LOG_PHY_ADDR_H;
+    uint32_t        DRAM_LOG_PHY_ADDR_L;
+    uint32_t        DRAM_LOG_BUFF_SIZE;
+    uint32_t        UlvEnterC;
+    uint32_t        UlvTime;
+    uint32_t        Reserved[3];
+
+};
+
+typedef struct SMU7_SoftRegisters SMU7_SoftRegisters;
+
+struct SMU7_Fusion_GraphicsLevel
+{
+    uint32_t    MinVddNb;
+
+    uint32_t    SclkFrequency;
+
+    uint8_t     Vid;
+    uint8_t     VidOffset;
+    uint16_t    AT;
+
+    uint8_t     PowerThrottle;
+    uint8_t     GnbSlow;
+    uint8_t     ForceNbPs1;
+    uint8_t     SclkDid;
+
+    uint8_t     DisplayWatermark;
+    uint8_t     EnabledForActivity;
+    uint8_t     EnabledForThrottle;
+    uint8_t     UpH;
+
+    uint8_t     DownH;
+    uint8_t     VoltageDownH;
+    uint8_t     DeepSleepDivId;
+
+    uint8_t     ClkBypassCntl;
+
+    uint32_t    reserved;
+};
+
+typedef struct SMU7_Fusion_GraphicsLevel SMU7_Fusion_GraphicsLevel;
+
+struct SMU7_Fusion_GIOLevel
+{
+    uint8_t     EnabledForActivity;
+    uint8_t     LclkDid;
+    uint8_t     Vid;
+    uint8_t     VoltageDownH;
+
+    uint32_t    MinVddNb;
+
+    uint16_t    ResidencyCounter;
+    uint8_t     UpH;
+    uint8_t     DownH;
+
+    uint32_t    LclkFrequency;
+
+    uint8_t     ActivityLevel;
+    uint8_t     EnabledForThrottle;
+
+    uint8_t     ClkBypassCntl;
+
+    uint8_t     padding;
+};
+
+typedef struct SMU7_Fusion_GIOLevel SMU7_Fusion_GIOLevel;
+
+// UVD VCLK/DCLK state (level) definition.
+struct SMU7_Fusion_UvdLevel
+{
+    uint32_t VclkFrequency;
+    uint32_t DclkFrequency;
+    uint16_t MinVddNb;
+    uint8_t  VclkDivider;
+    uint8_t  DclkDivider;
+
+    uint8_t     VClkBypassCntl;
+    uint8_t     DClkBypassCntl;
+
+    uint8_t     padding[2];
+
+};
+
+typedef struct SMU7_Fusion_UvdLevel SMU7_Fusion_UvdLevel;
+
+// Clocks for other external blocks (VCE, ACP, SAMU).
+struct SMU7_Fusion_ExtClkLevel
+{
+    uint32_t Frequency;
+    uint16_t MinVoltage;
+    uint8_t  Divider;
+    uint8_t  ClkBypassCntl;
+
+    uint32_t Reserved;
+};
+typedef struct SMU7_Fusion_ExtClkLevel SMU7_Fusion_ExtClkLevel;
+
+struct SMU7_Fusion_ACPILevel
+{
+    uint32_t    Flags;
+    uint32_t    MinVddNb;
+    uint32_t    SclkFrequency;
+    uint8_t     SclkDid;
+    uint8_t     GnbSlow;
+    uint8_t     ForceNbPs1;
+    uint8_t     DisplayWatermark;
+    uint8_t     DeepSleepDivId;
+    uint8_t     padding[3];
+};
+
+typedef struct SMU7_Fusion_ACPILevel SMU7_Fusion_ACPILevel;
+
+struct SMU7_Fusion_NbDpm
+{
+    uint8_t DpmXNbPsHi;
+    uint8_t DpmXNbPsLo;
+    uint8_t Dpm0PgNbPsHi;
+    uint8_t Dpm0PgNbPsLo;
+    uint8_t EnablePsi1;
+    uint8_t SkipDPM0;
+    uint8_t SkipPG;
+    uint8_t Hysteresis;
+    uint8_t EnableDpmPstatePoll;
+    uint8_t padding[3];
+};
+
+typedef struct SMU7_Fusion_NbDpm SMU7_Fusion_NbDpm;
+
+struct SMU7_Fusion_StateInfo
+{
+    uint32_t SclkFrequency;
+    uint32_t LclkFrequency;
+    uint32_t VclkFrequency;
+    uint32_t DclkFrequency;
+    uint32_t SamclkFrequency;
+    uint32_t AclkFrequency;
+    uint32_t EclkFrequency;
+    uint8_t  DisplayWatermark;
+    uint8_t  McArbIndex;
+    int8_t   SclkIndex;
+    int8_t   MclkIndex;
+};
+
+typedef struct SMU7_Fusion_StateInfo SMU7_Fusion_StateInfo;
+
+struct SMU7_Fusion_DpmTable
+{
+    uint32_t                            SystemFlags;
+
+    SMU7_PIDController                  GraphicsPIDController;
+    SMU7_PIDController                  GioPIDController;
+
+    uint8_t                            GraphicsDpmLevelCount;
+    uint8_t                            GIOLevelCount;
+    uint8_t                            UvdLevelCount;
+    uint8_t                            VceLevelCount;
+
+    uint8_t                            AcpLevelCount;
+    uint8_t                            SamuLevelCount;
+    uint16_t                           FpsHighT;
+
+    SMU7_Fusion_GraphicsLevel         GraphicsLevel           [SMU__NUM_SCLK_DPM_STATE];
+    SMU7_Fusion_ACPILevel             ACPILevel;
+    SMU7_Fusion_UvdLevel              UvdLevel                [SMU7_MAX_LEVELS_UVD];
+    SMU7_Fusion_ExtClkLevel           VceLevel                [SMU7_MAX_LEVELS_VCE];
+    SMU7_Fusion_ExtClkLevel           AcpLevel                [SMU7_MAX_LEVELS_ACP];
+    SMU7_Fusion_ExtClkLevel           SamuLevel               [SMU7_MAX_LEVELS_SAMU];
+
+    uint8_t                           UvdBootLevel;
+    uint8_t                           VceBootLevel;
+    uint8_t                           AcpBootLevel;
+    uint8_t                           SamuBootLevel;
+    uint8_t                           UVDInterval;
+    uint8_t                           VCEInterval;
+    uint8_t                           ACPInterval;
+    uint8_t                           SAMUInterval;
+
+    uint8_t                           GraphicsBootLevel;
+    uint8_t                           GraphicsInterval;
+    uint8_t                           GraphicsThermThrottleEnable;
+    uint8_t                           GraphicsVoltageChangeEnable;
+
+    uint8_t                           GraphicsClkSlowEnable;
+    uint8_t                           GraphicsClkSlowDivider;
+    uint16_t                          FpsLowT;
+
+    uint32_t                          DisplayCac;
+    uint32_t                          LowSclkInterruptT;
+
+    uint32_t                          DRAM_LOG_ADDR_H;
+    uint32_t                          DRAM_LOG_ADDR_L;
+    uint32_t                          DRAM_LOG_PHY_ADDR_H;
+    uint32_t                          DRAM_LOG_PHY_ADDR_L;
+    uint32_t                          DRAM_LOG_BUFF_SIZE;
+
+};
+
+struct SMU7_Fusion_GIODpmTable
+{
+
+    SMU7_Fusion_GIOLevel              GIOLevel                [SMU7_MAX_LEVELS_GIO];
+
+    SMU7_PIDController                GioPIDController;
+
+    uint32_t                          GIOLevelCount;
+
+    uint8_t                           Enable;
+    uint8_t                           GIOVoltageChangeEnable;
+    uint8_t                           GIOBootLevel;
+    uint8_t                           padding;
+    uint8_t                           padding1[2];
+    uint8_t                           TargetState;
+    uint8_t                           CurrenttState;
+    uint8_t                           ThrottleOnHtc;
+    uint8_t                           ThermThrottleStatus;
+    uint8_t                           ThermThrottleTempSelect;
+    uint8_t                           ThermThrottleEnable;
+    uint16_t                          TemperatureLimitHigh;
+    uint16_t                          TemperatureLimitLow;
+
+};
+
+typedef struct SMU7_Fusion_DpmTable SMU7_Fusion_DpmTable;
+typedef struct SMU7_Fusion_GIODpmTable SMU7_Fusion_GIODpmTable;
+
+#pragma pack(pop)
+
+#endif
+

drivers/gpu/drm/amd/amdgpu/uvd_v4_2.c

+/*
+ * Copyright 2013 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Christian König <christian.koenig@amd.com>
+ */
+
+#include <linux/firmware.h>
+#include <drm/drmP.h>
+#include "amdgpu.h"
+#include "amdgpu_uvd.h"
+#include "cikd.h"
+
+#include "uvd/uvd_4_2_d.h"
+#include "uvd/uvd_4_2_sh_mask.h"
+
+#include "oss/oss_2_0_d.h"
+#include "oss/oss_2_0_sh_mask.h"
+
+static void uvd_v4_2_mc_resume(struct amdgpu_device *adev);
+static void uvd_v4_2_init_cg(struct amdgpu_device *adev);
+static void uvd_v4_2_set_ring_funcs(struct amdgpu_device *adev);
+static void uvd_v4_2_set_irq_funcs(struct amdgpu_device *adev);
+static int uvd_v4_2_start(struct amdgpu_device *adev);
+static void uvd_v4_2_stop(struct amdgpu_device *adev);
+
+/**
+ * uvd_v4_2_ring_get_rptr - get read pointer
+ *
+ * @ring: amdgpu_ring pointer
+ *
+ * Returns the current hardware read pointer
+ */
+static uint32_t uvd_v4_2_ring_get_rptr(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+
+	return RREG32(mmUVD_RBC_RB_RPTR);
+}
+
+/**
+ * uvd_v4_2_ring_get_wptr - get write pointer
+ *
+ * @ring: amdgpu_ring pointer
+ *
+ * Returns the current hardware write pointer
+ */
+static uint32_t uvd_v4_2_ring_get_wptr(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+
+	return RREG32(mmUVD_RBC_RB_WPTR);
+}
+
+/**
+ * uvd_v4_2_ring_set_wptr - set write pointer
+ *
+ * @ring: amdgpu_ring pointer
+ *
+ * Commits the write pointer to the hardware
+ */
+static void uvd_v4_2_ring_set_wptr(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+
+	WREG32(mmUVD_RBC_RB_WPTR, ring->wptr);
+}
+
+static int uvd_v4_2_early_init(struct amdgpu_device *adev)
+{
+	uvd_v4_2_set_ring_funcs(adev);
+	uvd_v4_2_set_irq_funcs(adev);
+
+	return 0;
+}
+
+static int uvd_v4_2_sw_init(struct amdgpu_device *adev)
+{
+	struct amdgpu_ring *ring;
+	int r;
+
+	/* UVD TRAP */
+	r = amdgpu_irq_add_id(adev, 124, &adev->uvd.irq);
+	if (r)
+		return r;
+
+	r = amdgpu_uvd_sw_init(adev);
+	if (r)
+		return r;
+
+	r = amdgpu_uvd_resume(adev);
+	if (r)
+		return r;
+
+	ring = &adev->uvd.ring;
+	sprintf(ring->name, "uvd");
+	r = amdgpu_ring_init(adev, ring, 4096, CP_PACKET2, 0xf,
+			     &adev->uvd.irq, 0, AMDGPU_RING_TYPE_UVD);
+
+	return r;
+}
+
+static int uvd_v4_2_sw_fini(struct amdgpu_device *adev)
+{
+	int r;
+
+	r = amdgpu_uvd_suspend(adev);
+	if (r)
+		return r;
+
+	r = amdgpu_uvd_sw_fini(adev);
+	if (r)
+		return r;
+
+	return r;
+}
+
+/**
+ * uvd_v4_2_hw_init - start and test UVD block
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Initialize the hardware, boot up the VCPU and do some testing
+ */
+static int uvd_v4_2_hw_init(struct amdgpu_device *adev)
+{
+	struct amdgpu_ring *ring = &adev->uvd.ring;
+	uint32_t tmp;
+	int r;
+
+	/* raise clocks while booting up the VCPU */
+	amdgpu_asic_set_uvd_clocks(adev, 53300, 40000);
+
+	r = uvd_v4_2_start(adev);
+	if (r)
+		goto done;
+
+	ring->ready = true;
+	r = amdgpu_ring_test_ring(ring);
+	if (r) {
+		ring->ready = false;
+		goto done;
+	}
+
+	r = amdgpu_ring_lock(ring, 10);
+	if (r) {
+		DRM_ERROR("amdgpu: ring failed to lock UVD ring (%d).\n", r);
+		goto done;
+	}
+
+	tmp = PACKET0(mmUVD_SEMA_WAIT_FAULT_TIMEOUT_CNTL, 0);
+	amdgpu_ring_write(ring, tmp);
+	amdgpu_ring_write(ring, 0xFFFFF);
+
+	tmp = PACKET0(mmUVD_SEMA_WAIT_INCOMPLETE_TIMEOUT_CNTL, 0);
+	amdgpu_ring_write(ring, tmp);
+	amdgpu_ring_write(ring, 0xFFFFF);
+
+	tmp = PACKET0(mmUVD_SEMA_SIGNAL_INCOMPLETE_TIMEOUT_CNTL, 0);
+	amdgpu_ring_write(ring, tmp);
+	amdgpu_ring_write(ring, 0xFFFFF);
+
+	/* Clear timeout status bits */
+	amdgpu_ring_write(ring, PACKET0(mmUVD_SEMA_TIMEOUT_STATUS, 0));
+	amdgpu_ring_write(ring, 0x8);
+
+	amdgpu_ring_write(ring, PACKET0(mmUVD_SEMA_CNTL, 0));
+	amdgpu_ring_write(ring, 3);
+
+	amdgpu_ring_unlock_commit(ring);
+
+done:
+	/* lower clocks again */
+	amdgpu_asic_set_uvd_clocks(adev, 0, 0);
+
+	if (!r)
+		DRM_INFO("UVD initialized successfully.\n");
+
+	return r;
+}
+
+/**
+ * uvd_v4_2_hw_fini - stop the hardware block
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Stop the UVD block, mark ring as not ready any more
+ */
+static int uvd_v4_2_hw_fini(struct amdgpu_device *adev)
+{
+	struct amdgpu_ring *ring = &adev->uvd.ring;
+
+	uvd_v4_2_stop(adev);
+	ring->ready = false;
+
+	return 0;
+}
+
+static int uvd_v4_2_suspend(struct amdgpu_device *adev)
+{
+	int r;
+
+	r = uvd_v4_2_hw_fini(adev);
+	if (r)
+		return r;
+
+	r = amdgpu_uvd_suspend(adev);
+	if (r)
+		return r;
+
+	return r;
+}
+
+static int uvd_v4_2_resume(struct amdgpu_device *adev)
+{
+	int r;
+
+	r = amdgpu_uvd_resume(adev);
+	if (r)
+		return r;
+
+	r = uvd_v4_2_hw_init(adev);
+	if (r)
+		return r;
+
+	return r;
+}
+
+/**
+ * uvd_v4_2_start - start UVD block
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Setup and start the UVD block
+ */
+static int uvd_v4_2_start(struct amdgpu_device *adev)
+{
+	struct amdgpu_ring *ring = &adev->uvd.ring;
+	uint32_t rb_bufsz;
+	int i, j, r;
+
+	/* disable byte swapping */
+	u32 lmi_swap_cntl = 0;
+	u32 mp_swap_cntl = 0;
+
+	uvd_v4_2_mc_resume(adev);
+
+	/* disable clock gating */
+	WREG32(mmUVD_CGC_GATE, 0);
+
+	/* disable interupt */
+	WREG32_P(mmUVD_MASTINT_EN, 0, ~(1 << 1));
+
+	/* Stall UMC and register bus before resetting VCPU */
+	WREG32_P(mmUVD_LMI_CTRL2, 1 << 8, ~(1 << 8));
+	mdelay(1);
+
+	/* put LMI, VCPU, RBC etc... into reset */
+	WREG32(mmUVD_SOFT_RESET, UVD_SOFT_RESET__LMI_SOFT_RESET_MASK |
+		UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK | UVD_SOFT_RESET__LBSI_SOFT_RESET_MASK |
+		UVD_SOFT_RESET__RBC_SOFT_RESET_MASK | UVD_SOFT_RESET__CSM_SOFT_RESET_MASK |
+		UVD_SOFT_RESET__CXW_SOFT_RESET_MASK | UVD_SOFT_RESET__TAP_SOFT_RESET_MASK |
+		UVD_SOFT_RESET__LMI_UMC_SOFT_RESET_MASK);
+	mdelay(5);
+
+	/* take UVD block out of reset */
+	WREG32_P(mmSRBM_SOFT_RESET, 0, ~SRBM_SOFT_RESET__SOFT_RESET_UVD_MASK);
+	mdelay(5);
+
+	/* initialize UVD memory controller */
+	WREG32(mmUVD_LMI_CTRL, 0x40 | (1 << 8) | (1 << 13) |
+			     (1 << 21) | (1 << 9) | (1 << 20));
+
+#ifdef __BIG_ENDIAN
+	/* swap (8 in 32) RB and IB */
+	lmi_swap_cntl = 0xa;
+	mp_swap_cntl = 0;
+#endif
+	WREG32(mmUVD_LMI_SWAP_CNTL, lmi_swap_cntl);
+	WREG32(mmUVD_MP_SWAP_CNTL, mp_swap_cntl);
+
+	WREG32(mmUVD_MPC_SET_MUXA0, 0x40c2040);
+	WREG32(mmUVD_MPC_SET_MUXA1, 0x0);
+	WREG32(mmUVD_MPC_SET_MUXB0, 0x40c2040);
+	WREG32(mmUVD_MPC_SET_MUXB1, 0x0);
+	WREG32(mmUVD_MPC_SET_ALU, 0);
+	WREG32(mmUVD_MPC_SET_MUX, 0x88);
+
+	/* take all subblocks out of reset, except VCPU */
+	WREG32(mmUVD_SOFT_RESET, UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK);
+	mdelay(5);
+
+	/* enable VCPU clock */
+	WREG32(mmUVD_VCPU_CNTL,  1 << 9);
+
+	/* enable UMC */
+	WREG32_P(mmUVD_LMI_CTRL2, 0, ~(1 << 8));
+
+	/* boot up the VCPU */
+	WREG32(mmUVD_SOFT_RESET, 0);
+	mdelay(10);
+
+	for (i = 0; i < 10; ++i) {
+		uint32_t status;
+		for (j = 0; j < 100; ++j) {
+			status = RREG32(mmUVD_STATUS);
+			if (status & 2)
+				break;
+			mdelay(10);
+		}
+		r = 0;
+		if (status & 2)
+			break;
+
+		DRM_ERROR("UVD not responding, trying to reset the VCPU!!!\n");
+		WREG32_P(mmUVD_SOFT_RESET, UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK,
+				~UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK);
+		mdelay(10);
+		WREG32_P(mmUVD_SOFT_RESET, 0, ~UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK);
+		mdelay(10);
+		r = -1;
+	}
+
+	if (r) {
+		DRM_ERROR("UVD not responding, giving up!!!\n");
+		return r;
+	}
+
+	/* enable interupt */
+	WREG32_P(mmUVD_MASTINT_EN, 3<<1, ~(3 << 1));
+
+	/* force RBC into idle state */
+	WREG32(mmUVD_RBC_RB_CNTL, 0x11010101);
+
+	/* Set the write pointer delay */
+	WREG32(mmUVD_RBC_RB_WPTR_CNTL, 0);
+
+	/* programm the 4GB memory segment for rptr and ring buffer */
+	WREG32(mmUVD_LMI_EXT40_ADDR, upper_32_bits(ring->gpu_addr) |
+				   (0x7 << 16) | (0x1 << 31));
+
+	/* Initialize the ring buffer's read and write pointers */
+	WREG32(mmUVD_RBC_RB_RPTR, 0x0);
+
+	ring->wptr = RREG32(mmUVD_RBC_RB_RPTR);
+	WREG32(mmUVD_RBC_RB_WPTR, ring->wptr);
+
+	/* set the ring address */
+	WREG32(mmUVD_RBC_RB_BASE, ring->gpu_addr);
+
+	/* Set ring buffer size */
+	rb_bufsz = order_base_2(ring->ring_size);
+	rb_bufsz = (0x1 << 8) | rb_bufsz;
+	WREG32_P(mmUVD_RBC_RB_CNTL, rb_bufsz, ~0x11f1f);
+
+	return 0;
+}
+
+/**
+ * uvd_v4_2_stop - stop UVD block
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * stop the UVD block
+ */
+static void uvd_v4_2_stop(struct amdgpu_device *adev)
+{
+	/* force RBC into idle state */
+	WREG32(mmUVD_RBC_RB_CNTL, 0x11010101);
+
+	/* Stall UMC and register bus before resetting VCPU */
+	WREG32_P(mmUVD_LMI_CTRL2, 1 << 8, ~(1 << 8));
+	mdelay(1);
+
+	/* put VCPU into reset */
+	WREG32(mmUVD_SOFT_RESET, UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK);
+	mdelay(5);
+
+	/* disable VCPU clock */
+	WREG32(mmUVD_VCPU_CNTL, 0x0);
+
+	/* Unstall UMC and register bus */
+	WREG32_P(mmUVD_LMI_CTRL2, 0, ~(1 << 8));
+}
+
+/**
+ * uvd_v4_2_ring_emit_fence - emit an fence & trap command
+ *
+ * @ring: amdgpu_ring pointer
+ * @fence: fence to emit
+ *
+ * Write a fence and a trap command to the ring.
+ */
+static void uvd_v4_2_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, u64 seq,
+				     bool write64bit)
+{
+	WARN_ON(write64bit);
+
+	amdgpu_ring_write(ring, PACKET0(mmUVD_CONTEXT_ID, 0));
+	amdgpu_ring_write(ring, seq);
+	amdgpu_ring_write(ring, PACKET0(mmUVD_GPCOM_VCPU_DATA0, 0));
+	amdgpu_ring_write(ring, addr & 0xffffffff);
+	amdgpu_ring_write(ring, PACKET0(mmUVD_GPCOM_VCPU_DATA1, 0));
+	amdgpu_ring_write(ring, upper_32_bits(addr) & 0xff);
+	amdgpu_ring_write(ring, PACKET0(mmUVD_GPCOM_VCPU_CMD, 0));
+	amdgpu_ring_write(ring, 0);
+
+	amdgpu_ring_write(ring, PACKET0(mmUVD_GPCOM_VCPU_DATA0, 0));
+	amdgpu_ring_write(ring, 0);
+	amdgpu_ring_write(ring, PACKET0(mmUVD_GPCOM_VCPU_DATA1, 0));
+	amdgpu_ring_write(ring, 0);
+	amdgpu_ring_write(ring, PACKET0(mmUVD_GPCOM_VCPU_CMD, 0));
+	amdgpu_ring_write(ring, 2);
+}
+
+/**
+ * uvd_v4_2_ring_emit_semaphore - emit semaphore command
+ *
+ * @ring: amdgpu_ring pointer
+ * @semaphore: semaphore to emit commands for
+ * @emit_wait: true if we should emit a wait command
+ *
+ * Emit a semaphore command (either wait or signal) to the UVD ring.
+ */
+static bool uvd_v4_2_ring_emit_semaphore(struct amdgpu_ring *ring,
+					 struct amdgpu_semaphore *semaphore,
+					 bool emit_wait)
+{
+	uint64_t addr = semaphore->gpu_addr;
+
+	amdgpu_ring_write(ring, PACKET0(mmUVD_SEMA_ADDR_LOW, 0));
+	amdgpu_ring_write(ring, (addr >> 3) & 0x000FFFFF);
+
+	amdgpu_ring_write(ring, PACKET0(mmUVD_SEMA_ADDR_HIGH, 0));
+	amdgpu_ring_write(ring, (addr >> 23) & 0x000FFFFF);
+
+	amdgpu_ring_write(ring, PACKET0(mmUVD_SEMA_CMD, 0));
+	amdgpu_ring_write(ring, 0x80 | (emit_wait ? 1 : 0));
+
+	return true;
+}
+
+/**
+ * uvd_v4_2_ring_test_ring - register write test
+ *
+ * @ring: amdgpu_ring pointer
+ *
+ * Test if we can successfully write to the context register
+ */
+static int uvd_v4_2_ring_test_ring(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	uint32_t tmp = 0;
+	unsigned i;
+	int r;
+
+	WREG32(mmUVD_CONTEXT_ID, 0xCAFEDEAD);
+	r = amdgpu_ring_lock(ring, 3);
+	if (r) {
+		DRM_ERROR("amdgpu: cp failed to lock ring %d (%d).\n",
+			  ring->idx, r);
+		return r;
+	}
+	amdgpu_ring_write(ring, PACKET0(mmUVD_CONTEXT_ID, 0));
+	amdgpu_ring_write(ring, 0xDEADBEEF);
+	amdgpu_ring_unlock_commit(ring);
+	for (i = 0; i < adev->usec_timeout; i++) {
+		tmp = RREG32(mmUVD_CONTEXT_ID);
+		if (tmp == 0xDEADBEEF)
+			break;
+		DRM_UDELAY(1);
+	}
+
+	if (i < adev->usec_timeout) {
+		DRM_INFO("ring test on %d succeeded in %d usecs\n",
+			 ring->idx, i);
+	} else {
+		DRM_ERROR("amdgpu: ring %d test failed (0x%08X)\n",
+			  ring->idx, tmp);
+		r = -EINVAL;
+	}
+	return r;
+}
+
+/**
+ * uvd_v4_2_ring_emit_ib - execute indirect buffer
+ *
+ * @ring: amdgpu_ring pointer
+ * @ib: indirect buffer to execute
+ *
+ * Write ring commands to execute the indirect buffer
+ */
+static void uvd_v4_2_ring_emit_ib(struct amdgpu_ring *ring,
+				  struct amdgpu_ib *ib)
+{
+	amdgpu_ring_write(ring, PACKET0(mmUVD_RBC_IB_BASE, 0));
+	amdgpu_ring_write(ring, ib->gpu_addr);
+	amdgpu_ring_write(ring, PACKET0(mmUVD_RBC_IB_SIZE, 0));
+	amdgpu_ring_write(ring, ib->length_dw);
+}
+
+/**
+ * uvd_v4_2_ring_test_ib - test ib execution
+ *
+ * @ring: amdgpu_ring pointer
+ *
+ * Test if we can successfully execute an IB
+ */
+static int uvd_v4_2_ring_test_ib(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	struct amdgpu_fence *fence = NULL;
+	int r;
+
+	r = amdgpu_asic_set_uvd_clocks(adev, 53300, 40000);
+	if (r) {
+		DRM_ERROR("amdgpu: failed to raise UVD clocks (%d).\n", r);
+		return r;
+	}
+
+	r = amdgpu_uvd_get_create_msg(ring, 1, NULL);
+	if (r) {
+		DRM_ERROR("amdgpu: failed to get create msg (%d).\n", r);
+		goto error;
+	}
+
+	r = amdgpu_uvd_get_destroy_msg(ring, 1, &fence);
+	if (r) {
+		DRM_ERROR("amdgpu: failed to get destroy ib (%d).\n", r);
+		goto error;
+	}
+
+	r = amdgpu_fence_wait(fence, false);
+	if (r) {
+		DRM_ERROR("amdgpu: fence wait failed (%d).\n", r);
+		goto error;
+	}
+	DRM_INFO("ib test on ring %d succeeded\n",  ring->idx);
+error:
+	amdgpu_fence_unref(&fence);
+	amdgpu_asic_set_uvd_clocks(adev, 0, 0);
+	return r;
+}
+
+/**
+ * uvd_v4_2_mc_resume - memory controller programming
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Let the UVD memory controller know it's offsets
+ */
+static void uvd_v4_2_mc_resume(struct amdgpu_device *adev)
+{
+	uint64_t addr;
+	uint32_t size;
+
+	/* programm the VCPU memory controller bits 0-27 */
+	addr = (adev->uvd.gpu_addr + AMDGPU_UVD_FIRMWARE_OFFSET) >> 3;
+	size = AMDGPU_GPU_PAGE_ALIGN(adev->uvd.fw->size + 4) >> 3;
+	WREG32(mmUVD_VCPU_CACHE_OFFSET0, addr);
+	WREG32(mmUVD_VCPU_CACHE_SIZE0, size);
+
+	addr += size;
+	size = AMDGPU_UVD_STACK_SIZE >> 3;
+	WREG32(mmUVD_VCPU_CACHE_OFFSET1, addr);
+	WREG32(mmUVD_VCPU_CACHE_SIZE1, size);
+
+	addr += size;
+	size = AMDGPU_UVD_HEAP_SIZE >> 3;
+	WREG32(mmUVD_VCPU_CACHE_OFFSET2, addr);
+	WREG32(mmUVD_VCPU_CACHE_SIZE2, size);
+
+	/* bits 28-31 */
+	addr = (adev->uvd.gpu_addr >> 28) & 0xF;
+	WREG32(mmUVD_LMI_ADDR_EXT, (addr << 12) | (addr << 0));
+
+	/* bits 32-39 */
+	addr = (adev->uvd.gpu_addr >> 32) & 0xFF;
+	WREG32(mmUVD_LMI_EXT40_ADDR, addr | (0x9 << 16) | (0x1 << 31));
+
+	uvd_v4_2_init_cg(adev);
+}
+
+static void uvd_v4_2_enable_mgcg(struct amdgpu_device *adev,
+				 bool enable)
+{
+	u32 orig, data;
+
+	if (enable && (adev->cg_flags & AMDGPU_CG_SUPPORT_UVD_MGCG)) {
+		data = RREG32_UVD_CTX(ixUVD_CGC_MEM_CTRL);
+		data = 0xfff;
+		WREG32_UVD_CTX(ixUVD_CGC_MEM_CTRL, data);
+
+		orig = data = RREG32(mmUVD_CGC_CTRL);
+		data |= UVD_CGC_CTRL__DYN_CLOCK_MODE_MASK;
+		if (orig != data)
+			WREG32(mmUVD_CGC_CTRL, data);
+	} else {
+		data = RREG32_UVD_CTX(ixUVD_CGC_MEM_CTRL);
+		data &= ~0xfff;
+		WREG32_UVD_CTX(ixUVD_CGC_MEM_CTRL, data);
+
+		orig = data = RREG32(mmUVD_CGC_CTRL);
+		data &= ~UVD_CGC_CTRL__DYN_CLOCK_MODE_MASK;
+		if (orig != data)
+			WREG32(mmUVD_CGC_CTRL, data);
+	}
+}
+
+static void uvd_v4_2_set_dcm(struct amdgpu_device *adev,
+			     bool sw_mode)
+{
+	u32 tmp, tmp2;
+
+	tmp = RREG32(mmUVD_CGC_CTRL);
+	tmp &= ~(UVD_CGC_CTRL__CLK_OFF_DELAY_MASK | UVD_CGC_CTRL__CLK_GATE_DLY_TIMER_MASK);
+	tmp |= UVD_CGC_CTRL__DYN_CLOCK_MODE_MASK |
+		(1 << UVD_CGC_CTRL__CLK_GATE_DLY_TIMER__SHIFT) |
+		(4 << UVD_CGC_CTRL__CLK_OFF_DELAY__SHIFT);
+
+	if (sw_mode) {
+		tmp &= ~0x7ffff800;
+		tmp2 = UVD_CGC_CTRL2__DYN_OCLK_RAMP_EN_MASK |
+			UVD_CGC_CTRL2__DYN_RCLK_RAMP_EN_MASK |
+			(7 << UVD_CGC_CTRL2__GATER_DIV_ID__SHIFT);
+	} else {
+		tmp |= 0x7ffff800;
+		tmp2 = 0;
+	}
+
+	WREG32(mmUVD_CGC_CTRL, tmp);
+	WREG32_UVD_CTX(ixUVD_CGC_CTRL2, tmp2);
+}
+
+static void uvd_v4_2_init_cg(struct amdgpu_device *adev)
+{
+	bool hw_mode = true;
+
+	if (hw_mode) {
+		uvd_v4_2_set_dcm(adev, false);
+	} else {
+		u32 tmp = RREG32(mmUVD_CGC_CTRL);
+		tmp &= ~UVD_CGC_CTRL__DYN_CLOCK_MODE_MASK;
+		WREG32(mmUVD_CGC_CTRL, tmp);
+	}
+}
+
+static bool uvd_v4_2_is_idle(struct amdgpu_device *adev)
+{
+	return !(RREG32(mmSRBM_STATUS) & SRBM_STATUS__UVD_BUSY_MASK);
+}
+
+static int uvd_v4_2_wait_for_idle(struct amdgpu_device *adev)
+{
+	unsigned i;
+
+	for (i = 0; i < adev->usec_timeout; i++) {
+		if (!(RREG32(mmSRBM_STATUS) & SRBM_STATUS__UVD_BUSY_MASK))
+			return 0;
+	}
+	return -ETIMEDOUT;
+}
+
+static int uvd_v4_2_soft_reset(struct amdgpu_device *adev)
+{
+	uvd_v4_2_stop(adev);
+
+	WREG32_P(mmSRBM_SOFT_RESET, SRBM_SOFT_RESET__SOFT_RESET_UVD_MASK,
+			~SRBM_SOFT_RESET__SOFT_RESET_UVD_MASK);
+	mdelay(5);
+
+	return uvd_v4_2_start(adev);
+}
+
+static void uvd_v4_2_print_status(struct amdgpu_device *adev)
+{
+	dev_info(adev->dev, "UVD 4.2 registers\n");
+	dev_info(adev->dev, "  UVD_SEMA_ADDR_LOW=0x%08X\n",
+		 RREG32(mmUVD_SEMA_ADDR_LOW));
+	dev_info(adev->dev, "  UVD_SEMA_ADDR_HIGH=0x%08X\n",
+		 RREG32(mmUVD_SEMA_ADDR_HIGH));
+	dev_info(adev->dev, "  UVD_SEMA_CMD=0x%08X\n",
+		 RREG32(mmUVD_SEMA_CMD));
+	dev_info(adev->dev, "  UVD_GPCOM_VCPU_CMD=0x%08X\n",
+		 RREG32(mmUVD_GPCOM_VCPU_CMD));
+	dev_info(adev->dev, "  UVD_GPCOM_VCPU_DATA0=0x%08X\n",
+		 RREG32(mmUVD_GPCOM_VCPU_DATA0));
+	dev_info(adev->dev, "  UVD_GPCOM_VCPU_DATA1=0x%08X\n",
+		 RREG32(mmUVD_GPCOM_VCPU_DATA1));
+	dev_info(adev->dev, "  UVD_ENGINE_CNTL=0x%08X\n",
+		 RREG32(mmUVD_ENGINE_CNTL));
+	dev_info(adev->dev, "  UVD_UDEC_ADDR_CONFIG=0x%08X\n",
+		 RREG32(mmUVD_UDEC_ADDR_CONFIG));
+	dev_info(adev->dev, "  UVD_UDEC_DB_ADDR_CONFIG=0x%08X\n",
+		 RREG32(mmUVD_UDEC_DB_ADDR_CONFIG));
+	dev_info(adev->dev, "  UVD_UDEC_DBW_ADDR_CONFIG=0x%08X\n",
+		 RREG32(mmUVD_UDEC_DBW_ADDR_CONFIG));
+	dev_info(adev->dev, "  UVD_SEMA_CNTL=0x%08X\n",
+		 RREG32(mmUVD_SEMA_CNTL));
+	dev_info(adev->dev, "  UVD_LMI_EXT40_ADDR=0x%08X\n",
+		 RREG32(mmUVD_LMI_EXT40_ADDR));
+	dev_info(adev->dev, "  UVD_CTX_INDEX=0x%08X\n",
+		 RREG32(mmUVD_CTX_INDEX));
+	dev_info(adev->dev, "  UVD_CTX_DATA=0x%08X\n",
+		 RREG32(mmUVD_CTX_DATA));
+	dev_info(adev->dev, "  UVD_CGC_GATE=0x%08X\n",
+		 RREG32(mmUVD_CGC_GATE));
+	dev_info(adev->dev, "  UVD_CGC_CTRL=0x%08X\n",
+		 RREG32(mmUVD_CGC_CTRL));
+	dev_info(adev->dev, "  UVD_LMI_CTRL2=0x%08X\n",
+		 RREG32(mmUVD_LMI_CTRL2));
+	dev_info(adev->dev, "  UVD_MASTINT_EN=0x%08X\n",
+		 RREG32(mmUVD_MASTINT_EN));
+	dev_info(adev->dev, "  UVD_LMI_ADDR_EXT=0x%08X\n",
+		 RREG32(mmUVD_LMI_ADDR_EXT));
+	dev_info(adev->dev, "  UVD_LMI_CTRL=0x%08X\n",
+		 RREG32(mmUVD_LMI_CTRL));
+	dev_info(adev->dev, "  UVD_LMI_SWAP_CNTL=0x%08X\n",
+		 RREG32(mmUVD_LMI_SWAP_CNTL));
+	dev_info(adev->dev, "  UVD_MP_SWAP_CNTL=0x%08X\n",
+		 RREG32(mmUVD_MP_SWAP_CNTL));
+	dev_info(adev->dev, "  UVD_MPC_SET_MUXA0=0x%08X\n",
+		 RREG32(mmUVD_MPC_SET_MUXA0));
+	dev_info(adev->dev, "  UVD_MPC_SET_MUXA1=0x%08X\n",
+		 RREG32(mmUVD_MPC_SET_MUXA1));
+	dev_info(adev->dev, "  UVD_MPC_SET_MUXB0=0x%08X\n",
+		 RREG32(mmUVD_MPC_SET_MUXB0));
+	dev_info(adev->dev, "  UVD_MPC_SET_MUXB1=0x%08X\n",
+		 RREG32(mmUVD_MPC_SET_MUXB1));
+	dev_info(adev->dev, "  UVD_MPC_SET_MUX=0x%08X\n",
+		 RREG32(mmUVD_MPC_SET_MUX));
+	dev_info(adev->dev, "  UVD_MPC_SET_ALU=0x%08X\n",
+		 RREG32(mmUVD_MPC_SET_ALU));
+	dev_info(adev->dev, "  UVD_VCPU_CACHE_OFFSET0=0x%08X\n",
+		 RREG32(mmUVD_VCPU_CACHE_OFFSET0));
+	dev_info(adev->dev, "  UVD_VCPU_CACHE_SIZE0=0x%08X\n",
+		 RREG32(mmUVD_VCPU_CACHE_SIZE0));
+	dev_info(adev->dev, "  UVD_VCPU_CACHE_OFFSET1=0x%08X\n",
+		 RREG32(mmUVD_VCPU_CACHE_OFFSET1));
+	dev_info(adev->dev, "  UVD_VCPU_CACHE_SIZE1=0x%08X\n",
+		 RREG32(mmUVD_VCPU_CACHE_SIZE1));
+	dev_info(adev->dev, "  UVD_VCPU_CACHE_OFFSET2=0x%08X\n",
+		 RREG32(mmUVD_VCPU_CACHE_OFFSET2));
+	dev_info(adev->dev, "  UVD_VCPU_CACHE_SIZE2=0x%08X\n",
+		 RREG32(mmUVD_VCPU_CACHE_SIZE2));
+	dev_info(adev->dev, "  UVD_VCPU_CNTL=0x%08X\n",
+		 RREG32(mmUVD_VCPU_CNTL));
+	dev_info(adev->dev, "  UVD_SOFT_RESET=0x%08X\n",
+		 RREG32(mmUVD_SOFT_RESET));
+	dev_info(adev->dev, "  UVD_RBC_IB_BASE=0x%08X\n",
+		 RREG32(mmUVD_RBC_IB_BASE));
+	dev_info(adev->dev, "  UVD_RBC_IB_SIZE=0x%08X\n",
+		 RREG32(mmUVD_RBC_IB_SIZE));
+	dev_info(adev->dev, "  UVD_RBC_RB_BASE=0x%08X\n",
+		 RREG32(mmUVD_RBC_RB_BASE));
+	dev_info(adev->dev, "  UVD_RBC_RB_RPTR=0x%08X\n",
+		 RREG32(mmUVD_RBC_RB_RPTR));
+	dev_info(adev->dev, "  UVD_RBC_RB_WPTR=0x%08X\n",
+		 RREG32(mmUVD_RBC_RB_WPTR));
+	dev_info(adev->dev, "  UVD_RBC_RB_WPTR_CNTL=0x%08X\n",
+		 RREG32(mmUVD_RBC_RB_WPTR_CNTL));
+	dev_info(adev->dev, "  UVD_RBC_RB_CNTL=0x%08X\n",
+		 RREG32(mmUVD_RBC_RB_CNTL));
+	dev_info(adev->dev, "  UVD_STATUS=0x%08X\n",
+		 RREG32(mmUVD_STATUS));
+	dev_info(adev->dev, "  UVD_SEMA_TIMEOUT_STATUS=0x%08X\n",
+		 RREG32(mmUVD_SEMA_TIMEOUT_STATUS));
+	dev_info(adev->dev, "  UVD_SEMA_WAIT_INCOMPLETE_TIMEOUT_CNTL=0x%08X\n",
+		 RREG32(mmUVD_SEMA_WAIT_INCOMPLETE_TIMEOUT_CNTL));
+	dev_info(adev->dev, "  UVD_SEMA_WAIT_FAULT_TIMEOUT_CNTL=0x%08X\n",
+		 RREG32(mmUVD_SEMA_WAIT_FAULT_TIMEOUT_CNTL));
+	dev_info(adev->dev, "  UVD_SEMA_SIGNAL_INCOMPLETE_TIMEOUT_CNTL=0x%08X\n",
+		 RREG32(mmUVD_SEMA_SIGNAL_INCOMPLETE_TIMEOUT_CNTL));
+	dev_info(adev->dev, "  UVD_CONTEXT_ID=0x%08X\n",
+		 RREG32(mmUVD_CONTEXT_ID));
+}
+
+static int uvd_v4_2_set_interrupt_state(struct amdgpu_device *adev,
+					struct amdgpu_irq_src *source,
+					unsigned type,
+					enum amdgpu_interrupt_state state)
+{
+	// TODO
+	return 0;
+}
+
+static int uvd_v4_2_process_interrupt(struct amdgpu_device *adev,
+				      struct amdgpu_irq_src *source,
+				      struct amdgpu_iv_entry *entry)
+{
+	DRM_DEBUG("IH: UVD TRAP\n");
+	amdgpu_fence_process(&adev->uvd.ring);
+	return 0;
+}
+
+static int uvd_v4_2_set_clockgating_state(struct amdgpu_device *adev,
+					  enum amdgpu_clockgating_state state)
+{
+	bool gate = false;
+
+	if (state == AMDGPU_CG_STATE_GATE)
+		gate = true;
+
+	uvd_v4_2_enable_mgcg(adev, gate);
+
+	return 0;
+}
+
+static int uvd_v4_2_set_powergating_state(struct amdgpu_device *adev,
+					  enum amdgpu_powergating_state state)
+{
+	/* This doesn't actually powergate the UVD block.
+	 * That's done in the dpm code via the SMC.  This
+	 * just re-inits the block as necessary.  The actual
+	 * gating still happens in the dpm code.  We should
+	 * revisit this when there is a cleaner line between
+	 * the smc and the hw blocks
+	 */
+	if (state == AMDGPU_PG_STATE_GATE) {
+		uvd_v4_2_stop(adev);
+		return 0;
+	} else {
+		return uvd_v4_2_start(adev);
+	}
+}
+
+const struct amdgpu_ip_funcs uvd_v4_2_ip_funcs = {
+	.early_init = uvd_v4_2_early_init,
+	.late_init = NULL,
+	.sw_init = uvd_v4_2_sw_init,
+	.sw_fini = uvd_v4_2_sw_fini,
+	.hw_init = uvd_v4_2_hw_init,
+	.hw_fini = uvd_v4_2_hw_fini,
+	.suspend = uvd_v4_2_suspend,
+	.resume = uvd_v4_2_resume,
+	.is_idle = uvd_v4_2_is_idle,
+	.wait_for_idle = uvd_v4_2_wait_for_idle,
+	.soft_reset = uvd_v4_2_soft_reset,
+	.print_status = uvd_v4_2_print_status,
+	.set_clockgating_state = uvd_v4_2_set_clockgating_state,
+	.set_powergating_state = uvd_v4_2_set_powergating_state,
+};
+
+static const struct amdgpu_ring_funcs uvd_v4_2_ring_funcs = {
+	.get_rptr = uvd_v4_2_ring_get_rptr,
+	.get_wptr = uvd_v4_2_ring_get_wptr,
+	.set_wptr = uvd_v4_2_ring_set_wptr,
+	.parse_cs = amdgpu_uvd_ring_parse_cs,
+	.emit_ib = uvd_v4_2_ring_emit_ib,
+	.emit_fence = uvd_v4_2_ring_emit_fence,
+	.emit_semaphore = uvd_v4_2_ring_emit_semaphore,
+	.test_ring = uvd_v4_2_ring_test_ring,
+	.test_ib = uvd_v4_2_ring_test_ib,
+	.is_lockup = amdgpu_ring_test_lockup,
+};
+
+static void uvd_v4_2_set_ring_funcs(struct amdgpu_device *adev)
+{
+	adev->uvd.ring.funcs = &uvd_v4_2_ring_funcs;
+}
+
+static const struct amdgpu_irq_src_funcs uvd_v4_2_irq_funcs = {
+	.set = uvd_v4_2_set_interrupt_state,
+	.process = uvd_v4_2_process_interrupt,
+};
+
+static void uvd_v4_2_set_irq_funcs(struct amdgpu_device *adev)
+{
+	adev->uvd.irq.num_types = 1;
+	adev->uvd.irq.funcs = &uvd_v4_2_irq_funcs;
+}

drivers/gpu/drm/amd/amdgpu/uvd_v4_2.h

+/*
+ * Copyright 2014 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __UVD_V4_2_H__
+#define __UVD_V4_2_H__
+
+extern const struct amdgpu_ip_funcs uvd_v4_2_ip_funcs;
+
+#endif

drivers/gpu/drm/amd/amdgpu/vce_v2_0.c

+/*
+ * Copyright 2013 Advanced Micro Devices, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * Authors: Christian König <christian.koenig@amd.com>
+ */
+
+#include <linux/firmware.h>
+#include <drm/drmP.h>
+#include "amdgpu.h"
+#include "amdgpu_vce.h"
+#include "cikd.h"
+
+#include "vce/vce_2_0_d.h"
+#include "vce/vce_2_0_sh_mask.h"
+
+#include "oss/oss_2_0_d.h"
+#include "oss/oss_2_0_sh_mask.h"
+
+static void vce_v2_0_mc_resume(struct amdgpu_device *adev);
+static void vce_v2_0_set_ring_funcs(struct amdgpu_device *adev);
+static void vce_v2_0_set_irq_funcs(struct amdgpu_device *adev);
+
+/**
+ * vce_v2_0_ring_get_rptr - get read pointer
+ *
+ * @ring: amdgpu_ring pointer
+ *
+ * Returns the current hardware read pointer
+ */
+static uint32_t vce_v2_0_ring_get_rptr(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+
+	if (ring == &adev->vce.ring[0])
+		return RREG32(mmVCE_RB_RPTR);
+	else
+		return RREG32(mmVCE_RB_RPTR2);
+}
+
+/**
+ * vce_v2_0_ring_get_wptr - get write pointer
+ *
+ * @ring: amdgpu_ring pointer
+ *
+ * Returns the current hardware write pointer
+ */
+static uint32_t vce_v2_0_ring_get_wptr(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+
+	if (ring == &adev->vce.ring[0])
+		return RREG32(mmVCE_RB_WPTR);
+	else
+		return RREG32(mmVCE_RB_WPTR2);
+}
+
+/**
+ * vce_v2_0_ring_set_wptr - set write pointer
+ *
+ * @ring: amdgpu_ring pointer
+ *
+ * Commits the write pointer to the hardware
+ */
+static void vce_v2_0_ring_set_wptr(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+
+	if (ring == &adev->vce.ring[0])
+		WREG32(mmVCE_RB_WPTR, ring->wptr);
+	else
+		WREG32(mmVCE_RB_WPTR2, ring->wptr);
+}
+
+/**
+ * vce_v2_0_start - start VCE block
+ *
+ * @adev: amdgpu_device pointer
+ *
+ * Setup and start the VCE block
+ */
+static int vce_v2_0_start(struct amdgpu_device *adev)
+{
+	struct amdgpu_ring *ring;
+	int i, j, r;
+
+	vce_v2_0_mc_resume(adev);
+
+	/* set BUSY flag */
+	WREG32_P(mmVCE_STATUS, 1, ~1);
+
+	ring = &adev->vce.ring[0];
+	WREG32(mmVCE_RB_RPTR, ring->wptr);
+	WREG32(mmVCE_RB_WPTR, ring->wptr);
+	WREG32(mmVCE_RB_BASE_LO, ring->gpu_addr);
+	WREG32(mmVCE_RB_BASE_HI, upper_32_bits(ring->gpu_addr));
+	WREG32(mmVCE_RB_SIZE, ring->ring_size / 4);
+
+	ring = &adev->vce.ring[1];
+	WREG32(mmVCE_RB_RPTR2, ring->wptr);
+	WREG32(mmVCE_RB_WPTR2, ring->wptr);
+	WREG32(mmVCE_RB_BASE_LO2, ring->gpu_addr);
+	WREG32(mmVCE_RB_BASE_HI2, upper_32_bits(ring->gpu_addr));
+	WREG32(mmVCE_RB_SIZE2, ring->ring_size / 4);
+
+	WREG32_P(mmVCE_VCPU_CNTL, VCE_VCPU_CNTL__CLK_EN_MASK, ~VCE_VCPU_CNTL__CLK_EN_MASK);
+
+	WREG32_P(mmVCE_SOFT_RESET,
+		 VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK,
+		 ~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK);
+
+	mdelay(100);
+
+	WREG32_P(mmVCE_SOFT_RESET, 0, ~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK);
+
+	for (i = 0; i < 10; ++i) {
+		uint32_t status;
+		for (j = 0; j < 100; ++j) {
+			status = RREG32(mmVCE_STATUS);
+			if (status & 2)
+				break;
+			mdelay(10);
+		}
+		r = 0;
+		if (status & 2)
+			break;
+
+		DRM_ERROR("VCE not responding, trying to reset the ECPU!!!\n");
+		WREG32_P(mmVCE_SOFT_RESET, VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK,
+				~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK);
+		mdelay(10);
+		WREG32_P(mmVCE_SOFT_RESET, 0, ~VCE_SOFT_RESET__ECPU_SOFT_RESET_MASK);
+		mdelay(10);
+		r = -1;
+	}
+
+	/* clear BUSY flag */
+	WREG32_P(mmVCE_STATUS, 0, ~1);
+
+	if (r) {
+		DRM_ERROR("VCE not responding, giving up!!!\n");
+		return r;
+	}
+
+	return 0;
+}
+
+static int vce_v2_0_early_init(struct amdgpu_device *adev)
+{
+	vce_v2_0_set_ring_funcs(adev);
+	vce_v2_0_set_irq_funcs(adev);
+
+	return 0;
+}
+
+static int vce_v2_0_sw_init(struct amdgpu_device *adev)
+{
+	struct amdgpu_ring *ring;
+	int r;
+
+	/* VCE */
+	r = amdgpu_irq_add_id(adev, 167, &adev->vce.irq);
+	if (r)
+		return r;
+
+	r = amdgpu_vce_sw_init(adev);
+	if (r)
+		return r;
+
+	r = amdgpu_vce_resume(adev);
+	if (r)
+		return r;
+
+	ring = &adev->vce.ring[0];
+	sprintf(ring->name, "vce0");
+	r = amdgpu_ring_init(adev, ring, 4096, VCE_CMD_NO_OP, 0xf,
+			     &adev->vce.irq, 0, AMDGPU_RING_TYPE_VCE);
+	if (r)
+		return r;
+
+	ring = &adev->vce.ring[1];
+	sprintf(ring->name, "vce1");
+	r = amdgpu_ring_init(adev, ring, 4096, VCE_CMD_NO_OP, 0xf,
+			     &adev->vce.irq, 0, AMDGPU_RING_TYPE_VCE);
+	if (r)
+		return r;
+
+	return r;
+}
+
+static int vce_v2_0_sw_fini(struct amdgpu_device *adev)
+{
+	int r;
+
+	r = amdgpu_vce_suspend(adev);
+	if (r)
+		return r;
+
+	r = amdgpu_vce_sw_fini(adev);
+	if (r)
+		return r;
+
+	return r;
+}
+
+static int vce_v2_0_hw_init(struct amdgpu_device *adev)
+{
+	struct amdgpu_ring *ring;
+	int r;
+
+	r = vce_v2_0_start(adev);
+	if (r)
+		return r;
+
+	ring = &adev->vce.ring[0];
+	ring->ready = true;
+	r = amdgpu_ring_test_ring(ring);
+	if (r) {
+		ring->ready = false;
+		return r;
+	}
+
+	ring = &adev->vce.ring[1];
+	ring->ready = true;
+	r = amdgpu_ring_test_ring(ring);
+	if (r) {
+		ring->ready = false;
+		return r;
+	}
+
+	DRM_INFO("VCE initialized successfully.\n");
+
+	return 0;
+}
+
+static int vce_v2_0_hw_fini(struct amdgpu_device *adev)
+{
+	// TODO
+	return 0;
+}
+
+static int vce_v2_0_suspend(struct amdgpu_device *adev)
+{
+	int r;
+
+	r = vce_v2_0_hw_fini(adev);
+	if (r)
+		return r;
+
+	r = amdgpu_vce_suspend(adev);
+	if (r)
+		return r;
+
+	return r;
+}
+
+static int vce_v2_0_resume(struct amdgpu_device *adev)
+{
+	int r;
+
+	r = amdgpu_vce_resume(adev);
+	if (r)
+		return r;
+
+	r = vce_v2_0_hw_init(adev);
+	if (r)
+		return r;
+
+	return r;
+}
+
+static void vce_v2_0_set_sw_cg(struct amdgpu_device *adev, bool gated)
+{
+	u32 tmp;
+
+	if (gated) {
+		tmp = RREG32(mmVCE_CLOCK_GATING_B);
+		tmp |= 0xe70000;
+		WREG32(mmVCE_CLOCK_GATING_B, tmp);
+
+		tmp = RREG32(mmVCE_UENC_CLOCK_GATING);
+		tmp |= 0xff000000;
+		WREG32(mmVCE_UENC_CLOCK_GATING, tmp);
+
+		tmp = RREG32(mmVCE_UENC_REG_CLOCK_GATING);
+		tmp &= ~0x3fc;
+		WREG32(mmVCE_UENC_REG_CLOCK_GATING, tmp);
+
+		WREG32(mmVCE_CGTT_CLK_OVERRIDE, 0);
+    } else {
+		tmp = RREG32(mmVCE_CLOCK_GATING_B);
+		tmp |= 0xe7;
+		tmp &= ~0xe70000;
+		WREG32(mmVCE_CLOCK_GATING_B, tmp);
+
+		tmp = RREG32(mmVCE_UENC_CLOCK_GATING);
+		tmp |= 0x1fe000;
+		tmp &= ~0xff000000;
+		WREG32(mmVCE_UENC_CLOCK_GATING, tmp);
+
+		tmp = RREG32(mmVCE_UENC_REG_CLOCK_GATING);
+		tmp |= 0x3fc;
+		WREG32(mmVCE_UENC_REG_CLOCK_GATING, tmp);
+	}
+}
+
+static void vce_v2_0_set_dyn_cg(struct amdgpu_device *adev, bool gated)
+{
+	u32 orig, tmp;
+
+	tmp = RREG32(mmVCE_CLOCK_GATING_B);
+	tmp &= ~0x00060006;
+	if (gated) {
+		tmp |= 0xe10000;
+	} else {
+		tmp |= 0xe1;
+		tmp &= ~0xe10000;
+	}
+	WREG32(mmVCE_CLOCK_GATING_B, tmp);
+
+	orig = tmp = RREG32(mmVCE_UENC_CLOCK_GATING);
+	tmp &= ~0x1fe000;
+	tmp &= ~0xff000000;
+	if (tmp != orig)
+		WREG32(mmVCE_UENC_CLOCK_GATING, tmp);
+
+	orig = tmp = RREG32(mmVCE_UENC_REG_CLOCK_GATING);
+	tmp &= ~0x3fc;
+	if (tmp != orig)
+		WREG32(mmVCE_UENC_REG_CLOCK_GATING, tmp);
+
+	if (gated)
+		WREG32(mmVCE_CGTT_CLK_OVERRIDE, 0);
+}
+
+static void vce_v2_0_disable_cg(struct amdgpu_device *adev)
+{
+	WREG32(mmVCE_CGTT_CLK_OVERRIDE, 7);
+}
+
+static void vce_v2_0_enable_mgcg(struct amdgpu_device *adev, bool enable)
+{
+	bool sw_cg = false;
+
+	if (enable && (adev->cg_flags & AMDGPU_CG_SUPPORT_VCE_MGCG)) {
+		if (sw_cg)
+			vce_v2_0_set_sw_cg(adev, true);
+		else
+			vce_v2_0_set_dyn_cg(adev, true);
+	} else {
+		vce_v2_0_disable_cg(adev);
+
+		if (sw_cg)
+			vce_v2_0_set_sw_cg(adev, false);
+		else
+			vce_v2_0_set_dyn_cg(adev, false);
+	}
+}
+
+static void vce_v2_0_init_cg(struct amdgpu_device *adev)
+{
+	u32 tmp;
+
+	tmp = RREG32(mmVCE_CLOCK_GATING_A);
+	tmp &= ~0xfff;
+	tmp |= ((0 << 0) | (4 << 4));
+	tmp |= 0x40000;
+	WREG32(mmVCE_CLOCK_GATING_A, tmp);
+
+	tmp = RREG32(mmVCE_UENC_CLOCK_GATING);
+	tmp &= ~0xfff;
+	tmp |= ((0 << 0) | (4 << 4));
+	WREG32(mmVCE_UENC_CLOCK_GATING, tmp);
+
+	tmp = RREG32(mmVCE_CLOCK_GATING_B);
+	tmp |= 0x10;
+	tmp &= ~0x100000;
+	WREG32(mmVCE_CLOCK_GATING_B, tmp);
+}
+
+static void vce_v2_0_mc_resume(struct amdgpu_device *adev)
+{
+	uint64_t addr = adev->vce.gpu_addr;
+	uint32_t size;
+
+	WREG32_P(mmVCE_CLOCK_GATING_A, 0, ~(1 << 16));
+	WREG32_P(mmVCE_UENC_CLOCK_GATING, 0x1FF000, ~0xFF9FF000);
+	WREG32_P(mmVCE_UENC_REG_CLOCK_GATING, 0x3F, ~0x3F);
+	WREG32(mmVCE_CLOCK_GATING_B, 0xf7);
+
+	WREG32(mmVCE_LMI_CTRL, 0x00398000);
+	WREG32_P(mmVCE_LMI_CACHE_CTRL, 0x0, ~0x1);
+	WREG32(mmVCE_LMI_SWAP_CNTL, 0);
+	WREG32(mmVCE_LMI_SWAP_CNTL1, 0);
+	WREG32(mmVCE_LMI_VM_CTRL, 0);
+
+	addr += AMDGPU_VCE_FIRMWARE_OFFSET;
+	size = AMDGPU_GPU_PAGE_ALIGN(adev->vce.fw->size);
+	WREG32(mmVCE_VCPU_CACHE_OFFSET0, addr & 0x7fffffff);
+	WREG32(mmVCE_VCPU_CACHE_SIZE0, size);
+
+	addr += size;
+	size = AMDGPU_VCE_STACK_SIZE;
+	WREG32(mmVCE_VCPU_CACHE_OFFSET1, addr & 0x7fffffff);
+	WREG32(mmVCE_VCPU_CACHE_SIZE1, size);
+
+	addr += size;
+	size = AMDGPU_VCE_HEAP_SIZE;
+	WREG32(mmVCE_VCPU_CACHE_OFFSET2, addr & 0x7fffffff);
+	WREG32(mmVCE_VCPU_CACHE_SIZE2, size);
+
+	WREG32_P(mmVCE_LMI_CTRL2, 0x0, ~0x100);
+
+	WREG32_P(mmVCE_SYS_INT_EN, VCE_SYS_INT_EN__VCE_SYS_INT_TRAP_INTERRUPT_EN_MASK,
+		 ~VCE_SYS_INT_EN__VCE_SYS_INT_TRAP_INTERRUPT_EN_MASK);
+
+	vce_v2_0_init_cg(adev);
+}
+
+static bool vce_v2_0_is_idle(struct amdgpu_device *adev)
+{
+	return !(RREG32(mmSRBM_STATUS2) & SRBM_STATUS2__VCE_BUSY_MASK);
+}
+
+static int vce_v2_0_wait_for_idle(struct amdgpu_device *adev)
+{
+	unsigned i;
+
+	for (i = 0; i < adev->usec_timeout; i++) {
+		if (!(RREG32(mmSRBM_STATUS2) & SRBM_STATUS2__VCE_BUSY_MASK))
+			return 0;
+	}
+	return -ETIMEDOUT;
+}
+
+static int vce_v2_0_soft_reset(struct amdgpu_device *adev)
+{
+	WREG32_P(mmSRBM_SOFT_RESET, SRBM_SOFT_RESET__SOFT_RESET_VCE_MASK,
+			~SRBM_SOFT_RESET__SOFT_RESET_VCE_MASK);
+	mdelay(5);
+
+	return vce_v2_0_start(adev);
+}
+
+static void vce_v2_0_print_status(struct amdgpu_device *adev)
+{
+	dev_info(adev->dev, "VCE 2.0 registers\n");
+	dev_info(adev->dev, "  VCE_STATUS=0x%08X\n",
+		 RREG32(mmVCE_STATUS));
+	dev_info(adev->dev, "  VCE_VCPU_CNTL=0x%08X\n",
+		 RREG32(mmVCE_VCPU_CNTL));
+	dev_info(adev->dev, "  VCE_VCPU_CACHE_OFFSET0=0x%08X\n",
+		 RREG32(mmVCE_VCPU_CACHE_OFFSET0));
+	dev_info(adev->dev, "  VCE_VCPU_CACHE_SIZE0=0x%08X\n",
+		 RREG32(mmVCE_VCPU_CACHE_SIZE0));
+	dev_info(adev->dev, "  VCE_VCPU_CACHE_OFFSET1=0x%08X\n",
+		 RREG32(mmVCE_VCPU_CACHE_OFFSET1));
+	dev_info(adev->dev, "  VCE_VCPU_CACHE_SIZE1=0x%08X\n",
+		 RREG32(mmVCE_VCPU_CACHE_SIZE1));
+	dev_info(adev->dev, "  VCE_VCPU_CACHE_OFFSET2=0x%08X\n",
+		 RREG32(mmVCE_VCPU_CACHE_OFFSET2));
+	dev_info(adev->dev, "  VCE_VCPU_CACHE_SIZE2=0x%08X\n",
+		 RREG32(mmVCE_VCPU_CACHE_SIZE2));
+	dev_info(adev->dev, "  VCE_SOFT_RESET=0x%08X\n",
+		 RREG32(mmVCE_SOFT_RESET));
+	dev_info(adev->dev, "  VCE_RB_BASE_LO2=0x%08X\n",
+		 RREG32(mmVCE_RB_BASE_LO2));
+	dev_info(adev->dev, "  VCE_RB_BASE_HI2=0x%08X\n",
+		 RREG32(mmVCE_RB_BASE_HI2));
+	dev_info(adev->dev, "  VCE_RB_SIZE2=0x%08X\n",
+		 RREG32(mmVCE_RB_SIZE2));
+	dev_info(adev->dev, "  VCE_RB_RPTR2=0x%08X\n",
+		 RREG32(mmVCE_RB_RPTR2));
+	dev_info(adev->dev, "  VCE_RB_WPTR2=0x%08X\n",
+		 RREG32(mmVCE_RB_WPTR2));
+	dev_info(adev->dev, "  VCE_RB_BASE_LO=0x%08X\n",
+		 RREG32(mmVCE_RB_BASE_LO));
+	dev_info(adev->dev, "  VCE_RB_BASE_HI=0x%08X\n",
+		 RREG32(mmVCE_RB_BASE_HI));
+	dev_info(adev->dev, "  VCE_RB_SIZE=0x%08X\n",
+		 RREG32(mmVCE_RB_SIZE));
+	dev_info(adev->dev, "  VCE_RB_RPTR=0x%08X\n",
+		 RREG32(mmVCE_RB_RPTR));
+	dev_info(adev->dev, "  VCE_RB_WPTR=0x%08X\n",
+		 RREG32(mmVCE_RB_WPTR));
+	dev_info(adev->dev, "  VCE_CLOCK_GATING_A=0x%08X\n",
+		 RREG32(mmVCE_CLOCK_GATING_A));
+	dev_info(adev->dev, "  VCE_CLOCK_GATING_B=0x%08X\n",
+		 RREG32(mmVCE_CLOCK_GATING_B));
+	dev_info(adev->dev, "  VCE_CGTT_CLK_OVERRIDE=0x%08X\n",
+		 RREG32(mmVCE_CGTT_CLK_OVERRIDE));
+	dev_info(adev->dev, "  VCE_UENC_CLOCK_GATING=0x%08X\n",
+		 RREG32(mmVCE_UENC_CLOCK_GATING));
+	dev_info(adev->dev, "  VCE_UENC_REG_CLOCK_GATING=0x%08X\n",
+		 RREG32(mmVCE_UENC_REG_CLOCK_GATING));
+	dev_info(adev->dev, "  VCE_SYS_INT_EN=0x%08X\n",
+		 RREG32(mmVCE_SYS_INT_EN));
+	dev_info(adev->dev, "  VCE_LMI_CTRL2=0x%08X\n",
+		 RREG32(mmVCE_LMI_CTRL2));
+	dev_info(adev->dev, "  VCE_LMI_CTRL=0x%08X\n",
+		 RREG32(mmVCE_LMI_CTRL));
+	dev_info(adev->dev, "  VCE_LMI_VM_CTRL=0x%08X\n",
+		 RREG32(mmVCE_LMI_VM_CTRL));
+	dev_info(adev->dev, "  VCE_LMI_SWAP_CNTL=0x%08X\n",
+		 RREG32(mmVCE_LMI_SWAP_CNTL));
+	dev_info(adev->dev, "  VCE_LMI_SWAP_CNTL1=0x%08X\n",
+		 RREG32(mmVCE_LMI_SWAP_CNTL1));
+	dev_info(adev->dev, "  VCE_LMI_CACHE_CTRL=0x%08X\n",
+		 RREG32(mmVCE_LMI_CACHE_CTRL));
+}
+
+static int vce_v2_0_set_interrupt_state(struct amdgpu_device *adev,
+					struct amdgpu_irq_src *source,
+					unsigned type,
+					enum amdgpu_interrupt_state state)
+{
+	uint32_t val = 0;
+
+	if (state == AMDGPU_IRQ_STATE_ENABLE)
+		val |= VCE_SYS_INT_EN__VCE_SYS_INT_TRAP_INTERRUPT_EN_MASK;
+
+	WREG32_P(mmVCE_SYS_INT_EN, val, ~VCE_SYS_INT_EN__VCE_SYS_INT_TRAP_INTERRUPT_EN_MASK);
+	return 0;
+}
+
+static int vce_v2_0_process_interrupt(struct amdgpu_device *adev,
+				      struct amdgpu_irq_src *source,
+				      struct amdgpu_iv_entry *entry)
+{
+	DRM_DEBUG("IH: VCE\n");
+	switch (entry->src_data) {
+	case 0:
+		amdgpu_fence_process(&adev->vce.ring[0]);
+		break;
+	case 1:
+		amdgpu_fence_process(&adev->vce.ring[1]);
+		break;
+	default:
+		DRM_ERROR("Unhandled interrupt: %d %d\n",
+			  entry->src_id, entry->src_data);
+		break;
+	}
+
+	return 0;
+}
+
+static int vce_v2_0_set_clockgating_state(struct amdgpu_device *adev,
+					  enum amdgpu_clockgating_state state)
+{
+	bool gate = false;
+
+	if (state == AMDGPU_CG_STATE_GATE)
+		gate = true;
+
+	vce_v2_0_enable_mgcg(adev, gate);
+
+	return 0;
+}
+
+static int vce_v2_0_set_powergating_state(struct amdgpu_device *adev,
+					  enum amdgpu_powergating_state state)
+{
+	/* This doesn't actually powergate the VCE block.
+	 * That's done in the dpm code via the SMC.  This
+	 * just re-inits the block as necessary.  The actual
+	 * gating still happens in the dpm code.  We should
+	 * revisit this when there is a cleaner line between
+	 * the smc and the hw blocks
+	 */
+	if (state == AMDGPU_PG_STATE_GATE)
+		/* XXX do we need a vce_v2_0_stop()? */
+		return 0;
+	else
+		return vce_v2_0_start(adev);
+}
+
+const struct amdgpu_ip_funcs vce_v2_0_ip_funcs = {
+	.early_init = vce_v2_0_early_init,
+	.late_init = NULL,
+	.sw_init = vce_v2_0_sw_init,
+	.sw_fini = vce_v2_0_sw_fini,
+	.hw_init = vce_v2_0_hw_init,
+	.hw_fini = vce_v2_0_hw_fini,
+	.suspend = vce_v2_0_suspend,
+	.resume = vce_v2_0_resume,
+	.is_idle = vce_v2_0_is_idle,
+	.wait_for_idle = vce_v2_0_wait_for_idle,
+	.soft_reset = vce_v2_0_soft_reset,
+	.print_status = vce_v2_0_print_status,
+	.set_clockgating_state = vce_v2_0_set_clockgating_state,
+	.set_powergating_state = vce_v2_0_set_powergating_state,
+};
+
+static const struct amdgpu_ring_funcs vce_v2_0_ring_funcs = {
+	.get_rptr = vce_v2_0_ring_get_rptr,
+	.get_wptr = vce_v2_0_ring_get_wptr,
+	.set_wptr = vce_v2_0_ring_set_wptr,
+	.parse_cs = amdgpu_vce_ring_parse_cs,
+	.emit_ib = amdgpu_vce_ring_emit_ib,
+	.emit_fence = amdgpu_vce_ring_emit_fence,
+	.emit_semaphore = amdgpu_vce_ring_emit_semaphore,
+	.test_ring = amdgpu_vce_ring_test_ring,
+	.test_ib = amdgpu_vce_ring_test_ib,
+	.is_lockup = amdgpu_ring_test_lockup,
+};
+
+static void vce_v2_0_set_ring_funcs(struct amdgpu_device *adev)
+{
+	adev->vce.ring[0].funcs = &vce_v2_0_ring_funcs;
+	adev->vce.ring[1].funcs = &vce_v2_0_ring_funcs;
+}
+
+static const struct amdgpu_irq_src_funcs vce_v2_0_irq_funcs = {
+	.set = vce_v2_0_set_interrupt_state,
+	.process = vce_v2_0_process_interrupt,
+};
+
+static void vce_v2_0_set_irq_funcs(struct amdgpu_device *adev)
+{
+	adev->vce.irq.num_types = 1;
+	adev->vce.irq.funcs = &vce_v2_0_irq_funcs;
+};

drivers/gpu/drm/amd/amdgpu/vce_v2_0.h

+/*
+ * Copyright 2014 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#ifndef __VCE_V2_0_H__
+#define __VCE_V2_0_H__
+
+extern const struct amdgpu_ip_funcs vce_v2_0_ip_funcs;
+
+#endif