amdgpu_device.c 75.4 KB
Newer Older
A
Alex Deucher 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27
/*
 * Copyright 2008 Advanced Micro Devices, Inc.
 * Copyright 2008 Red Hat Inc.
 * Copyright 2009 Jerome Glisse.
 *
 * 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: Dave Airlie
 *          Alex Deucher
 *          Jerome Glisse
 */
28
#include <linux/kthread.h>
A
Alex Deucher 已提交
29 30 31 32
#include <linux/console.h>
#include <linux/slab.h>
#include <drm/drmP.h>
#include <drm/drm_crtc_helper.h>
33
#include <drm/drm_atomic_helper.h>
A
Alex Deucher 已提交
34 35 36 37 38
#include <drm/amdgpu_drm.h>
#include <linux/vgaarb.h>
#include <linux/vga_switcheroo.h>
#include <linux/efi.h>
#include "amdgpu.h"
39
#include "amdgpu_trace.h"
A
Alex Deucher 已提交
40 41 42
#include "amdgpu_i2c.h"
#include "atom.h"
#include "amdgpu_atombios.h"
43
#include "amdgpu_atomfirmware.h"
44
#include "amd_pcie.h"
K
Ken Wang 已提交
45 46 47
#ifdef CONFIG_DRM_AMDGPU_SI
#include "si.h"
#endif
48 49 50
#ifdef CONFIG_DRM_AMDGPU_CIK
#include "cik.h"
#endif
51
#include "vi.h"
52
#include "soc15.h"
A
Alex Deucher 已提交
53
#include "bif/bif_4_1_d.h"
54
#include <linux/pci.h>
55
#include <linux/firmware.h>
56
#include "amdgpu_vf_error.h"
A
Alex Deucher 已提交
57

58
#include "amdgpu_amdkfd.h"
59
#include "amdgpu_pm.h"
A
Alex Deucher 已提交
60

61
MODULE_FIRMWARE("amdgpu/vega10_gpu_info.bin");
62
MODULE_FIRMWARE("amdgpu/raven_gpu_info.bin");
63

64 65
#define AMDGPU_RESUME_MS		2000

A
Alex Deucher 已提交
66
static const char *amdgpu_asic_name[] = {
67 68 69 70 71
	"TAHITI",
	"PITCAIRN",
	"VERDE",
	"OLAND",
	"HAINAN",
A
Alex Deucher 已提交
72 73 74 75 76 77 78
	"BONAIRE",
	"KAVERI",
	"KABINI",
	"HAWAII",
	"MULLINS",
	"TOPAZ",
	"TONGA",
79
	"FIJI",
A
Alex Deucher 已提交
80
	"CARRIZO",
S
Samuel Li 已提交
81
	"STONEY",
82 83
	"POLARIS10",
	"POLARIS11",
84
	"POLARIS12",
K
Ken Wang 已提交
85
	"VEGA10",
86
	"RAVEN",
A
Alex Deucher 已提交
87 88 89 90 91 92 93
	"LAST",
};

bool amdgpu_device_is_px(struct drm_device *dev)
{
	struct amdgpu_device *adev = dev->dev_private;

94
	if (adev->flags & AMD_IS_PX)
A
Alex Deucher 已提交
95 96 97 98 99 100 101 102
		return true;
	return false;
}

/*
 * MMIO register access helper functions.
 */
uint32_t amdgpu_mm_rreg(struct amdgpu_device *adev, uint32_t reg,
M
Monk Liu 已提交
103
			uint32_t acc_flags)
A
Alex Deucher 已提交
104
{
105 106
	uint32_t ret;

107
	if (!(acc_flags & AMDGPU_REGS_NO_KIQ) && amdgpu_sriov_runtime(adev))
108 109
		return amdgpu_virt_kiq_rreg(adev, reg);

M
Monk Liu 已提交
110
	if ((reg * 4) < adev->rmmio_size && !(acc_flags & AMDGPU_REGS_IDX))
111
		ret = readl(((void __iomem *)adev->rmmio) + (reg * 4));
A
Alex Deucher 已提交
112 113 114 115 116 117 118 119
	else {
		unsigned long flags;

		spin_lock_irqsave(&adev->mmio_idx_lock, flags);
		writel((reg * 4), ((void __iomem *)adev->rmmio) + (mmMM_INDEX * 4));
		ret = readl(((void __iomem *)adev->rmmio) + (mmMM_DATA * 4));
		spin_unlock_irqrestore(&adev->mmio_idx_lock, flags);
	}
120 121
	trace_amdgpu_mm_rreg(adev->pdev->device, reg, ret);
	return ret;
A
Alex Deucher 已提交
122 123 124
}

void amdgpu_mm_wreg(struct amdgpu_device *adev, uint32_t reg, uint32_t v,
M
Monk Liu 已提交
125
		    uint32_t acc_flags)
A
Alex Deucher 已提交
126
{
127
	trace_amdgpu_mm_wreg(adev->pdev->device, reg, v);
128

129 130 131 132
	if (adev->asic_type >= CHIP_VEGA10 && reg == 0) {
		adev->last_mm_index = v;
	}

133
	if (!(acc_flags & AMDGPU_REGS_NO_KIQ) && amdgpu_sriov_runtime(adev))
134 135
		return amdgpu_virt_kiq_wreg(adev, reg, v);

M
Monk Liu 已提交
136
	if ((reg * 4) < adev->rmmio_size && !(acc_flags & AMDGPU_REGS_IDX))
A
Alex Deucher 已提交
137 138 139 140 141 142 143 144 145
		writel(v, ((void __iomem *)adev->rmmio) + (reg * 4));
	else {
		unsigned long flags;

		spin_lock_irqsave(&adev->mmio_idx_lock, flags);
		writel((reg * 4), ((void __iomem *)adev->rmmio) + (mmMM_INDEX * 4));
		writel(v, ((void __iomem *)adev->rmmio) + (mmMM_DATA * 4));
		spin_unlock_irqrestore(&adev->mmio_idx_lock, flags);
	}
146 147 148 149

	if (adev->asic_type >= CHIP_VEGA10 && reg == 1 && adev->last_mm_index == 0x5702C) {
		udelay(500);
	}
A
Alex Deucher 已提交
150 151 152 153 154 155 156 157 158 159 160 161 162 163
}

u32 amdgpu_io_rreg(struct amdgpu_device *adev, u32 reg)
{
	if ((reg * 4) < adev->rio_mem_size)
		return ioread32(adev->rio_mem + (reg * 4));
	else {
		iowrite32((reg * 4), adev->rio_mem + (mmMM_INDEX * 4));
		return ioread32(adev->rio_mem + (mmMM_DATA * 4));
	}
}

void amdgpu_io_wreg(struct amdgpu_device *adev, u32 reg, u32 v)
{
164 165 166
	if (adev->asic_type >= CHIP_VEGA10 && reg == 0) {
		adev->last_mm_index = v;
	}
A
Alex Deucher 已提交
167 168 169 170 171 172 173

	if ((reg * 4) < adev->rio_mem_size)
		iowrite32(v, adev->rio_mem + (reg * 4));
	else {
		iowrite32((reg * 4), adev->rio_mem + (mmMM_INDEX * 4));
		iowrite32(v, adev->rio_mem + (mmMM_DATA * 4));
	}
174 175 176 177

	if (adev->asic_type >= CHIP_VEGA10 && reg == 1 && adev->last_mm_index == 0x5702C) {
		udelay(500);
	}
A
Alex Deucher 已提交
178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217
}

/**
 * amdgpu_mm_rdoorbell - read a doorbell dword
 *
 * @adev: amdgpu_device pointer
 * @index: doorbell index
 *
 * Returns the value in the doorbell aperture at the
 * requested doorbell index (CIK).
 */
u32 amdgpu_mm_rdoorbell(struct amdgpu_device *adev, u32 index)
{
	if (index < adev->doorbell.num_doorbells) {
		return readl(adev->doorbell.ptr + index);
	} else {
		DRM_ERROR("reading beyond doorbell aperture: 0x%08x!\n", index);
		return 0;
	}
}

/**
 * amdgpu_mm_wdoorbell - write a doorbell dword
 *
 * @adev: amdgpu_device pointer
 * @index: doorbell index
 * @v: value to write
 *
 * Writes @v to the doorbell aperture at the
 * requested doorbell index (CIK).
 */
void amdgpu_mm_wdoorbell(struct amdgpu_device *adev, u32 index, u32 v)
{
	if (index < adev->doorbell.num_doorbells) {
		writel(v, adev->doorbell.ptr + index);
	} else {
		DRM_ERROR("writing beyond doorbell aperture: 0x%08x!\n", index);
	}
}

218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255
/**
 * amdgpu_mm_rdoorbell64 - read a doorbell Qword
 *
 * @adev: amdgpu_device pointer
 * @index: doorbell index
 *
 * Returns the value in the doorbell aperture at the
 * requested doorbell index (VEGA10+).
 */
u64 amdgpu_mm_rdoorbell64(struct amdgpu_device *adev, u32 index)
{
	if (index < adev->doorbell.num_doorbells) {
		return atomic64_read((atomic64_t *)(adev->doorbell.ptr + index));
	} else {
		DRM_ERROR("reading beyond doorbell aperture: 0x%08x!\n", index);
		return 0;
	}
}

/**
 * amdgpu_mm_wdoorbell64 - write a doorbell Qword
 *
 * @adev: amdgpu_device pointer
 * @index: doorbell index
 * @v: value to write
 *
 * Writes @v to the doorbell aperture at the
 * requested doorbell index (VEGA10+).
 */
void amdgpu_mm_wdoorbell64(struct amdgpu_device *adev, u32 index, u64 v)
{
	if (index < adev->doorbell.num_doorbells) {
		atomic64_set((atomic64_t *)(adev->doorbell.ptr + index), v);
	} else {
		DRM_ERROR("writing beyond doorbell aperture: 0x%08x!\n", index);
	}
}

A
Alex Deucher 已提交
256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329
/**
 * amdgpu_invalid_rreg - dummy reg read function
 *
 * @adev: amdgpu device pointer
 * @reg: offset of register
 *
 * Dummy register read function.  Used for register blocks
 * that certain asics don't have (all asics).
 * Returns the value in the register.
 */
static uint32_t amdgpu_invalid_rreg(struct amdgpu_device *adev, uint32_t reg)
{
	DRM_ERROR("Invalid callback to read register 0x%04X\n", reg);
	BUG();
	return 0;
}

/**
 * amdgpu_invalid_wreg - dummy reg write function
 *
 * @adev: amdgpu device pointer
 * @reg: offset of register
 * @v: value to write to the register
 *
 * Dummy register read function.  Used for register blocks
 * that certain asics don't have (all asics).
 */
static void amdgpu_invalid_wreg(struct amdgpu_device *adev, uint32_t reg, uint32_t v)
{
	DRM_ERROR("Invalid callback to write register 0x%04X with 0x%08X\n",
		  reg, v);
	BUG();
}

/**
 * amdgpu_block_invalid_rreg - dummy reg read function
 *
 * @adev: amdgpu device pointer
 * @block: offset of instance
 * @reg: offset of register
 *
 * Dummy register read function.  Used for register blocks
 * that certain asics don't have (all asics).
 * Returns the value in the register.
 */
static uint32_t amdgpu_block_invalid_rreg(struct amdgpu_device *adev,
					  uint32_t block, uint32_t reg)
{
	DRM_ERROR("Invalid callback to read register 0x%04X in block 0x%04X\n",
		  reg, block);
	BUG();
	return 0;
}

/**
 * amdgpu_block_invalid_wreg - dummy reg write function
 *
 * @adev: amdgpu device pointer
 * @block: offset of instance
 * @reg: offset of register
 * @v: value to write to the register
 *
 * Dummy register read function.  Used for register blocks
 * that certain asics don't have (all asics).
 */
static void amdgpu_block_invalid_wreg(struct amdgpu_device *adev,
				      uint32_t block,
				      uint32_t reg, uint32_t v)
{
	DRM_ERROR("Invalid block callback to write register 0x%04X in block 0x%04X with 0x%08X\n",
		  reg, block, v);
	BUG();
}

330
static int amdgpu_device_vram_scratch_init(struct amdgpu_device *adev)
A
Alex Deucher 已提交
331
{
332 333 334 335 336
	return amdgpu_bo_create_kernel(adev, AMDGPU_GPU_PAGE_SIZE,
				       PAGE_SIZE, AMDGPU_GEM_DOMAIN_VRAM,
				       &adev->vram_scratch.robj,
				       &adev->vram_scratch.gpu_addr,
				       (void **)&adev->vram_scratch.ptr);
A
Alex Deucher 已提交
337 338
}

339
static void amdgpu_device_vram_scratch_fini(struct amdgpu_device *adev)
A
Alex Deucher 已提交
340
{
341
	amdgpu_bo_free_kernel(&adev->vram_scratch.robj, NULL, NULL);
A
Alex Deucher 已提交
342 343 344
}

/**
345
 * amdgpu_device_program_register_sequence - program an array of registers.
A
Alex Deucher 已提交
346 347 348 349 350 351 352 353
 *
 * @adev: amdgpu_device pointer
 * @registers: pointer to the register array
 * @array_size: size of the register array
 *
 * Programs an array or registers with and and or masks.
 * This is a helper for setting golden registers.
 */
354 355 356
void amdgpu_device_program_register_sequence(struct amdgpu_device *adev,
					     const u32 *registers,
					     const u32 array_size)
A
Alex Deucher 已提交
357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379
{
	u32 tmp, reg, and_mask, or_mask;
	int i;

	if (array_size % 3)
		return;

	for (i = 0; i < array_size; i +=3) {
		reg = registers[i + 0];
		and_mask = registers[i + 1];
		or_mask = registers[i + 2];

		if (and_mask == 0xffffffff) {
			tmp = or_mask;
		} else {
			tmp = RREG32(reg);
			tmp &= ~and_mask;
			tmp |= or_mask;
		}
		WREG32(reg, tmp);
	}
}

380
void amdgpu_device_pci_config_reset(struct amdgpu_device *adev)
A
Alex Deucher 已提交
381 382 383 384 385 386 387 388
{
	pci_write_config_dword(adev->pdev, 0x7c, AMDGPU_ASIC_RESET_DATA);
}

/*
 * GPU doorbell aperture helpers function.
 */
/**
389
 * amdgpu_device_doorbell_init - Init doorbell driver information.
A
Alex Deucher 已提交
390 391 392 393 394 395
 *
 * @adev: amdgpu_device pointer
 *
 * Init doorbell driver information (CIK)
 * Returns 0 on success, error on failure.
 */
396
static int amdgpu_device_doorbell_init(struct amdgpu_device *adev)
A
Alex Deucher 已提交
397
{
398 399 400 401 402 403 404 405 406
	/* No doorbell on SI hardware generation */
	if (adev->asic_type < CHIP_BONAIRE) {
		adev->doorbell.base = 0;
		adev->doorbell.size = 0;
		adev->doorbell.num_doorbells = 0;
		adev->doorbell.ptr = NULL;
		return 0;
	}

407 408 409
	if (pci_resource_flags(adev->pdev, 2) & IORESOURCE_UNSET)
		return -EINVAL;

A
Alex Deucher 已提交
410 411 412 413
	/* doorbell bar mapping */
	adev->doorbell.base = pci_resource_start(adev->pdev, 2);
	adev->doorbell.size = pci_resource_len(adev->pdev, 2);

414
	adev->doorbell.num_doorbells = min_t(u32, adev->doorbell.size / sizeof(u32),
A
Alex Deucher 已提交
415 416 417 418
					     AMDGPU_DOORBELL_MAX_ASSIGNMENT+1);
	if (adev->doorbell.num_doorbells == 0)
		return -EINVAL;

419 420 421 422
	adev->doorbell.ptr = ioremap(adev->doorbell.base,
				     adev->doorbell.num_doorbells *
				     sizeof(u32));
	if (adev->doorbell.ptr == NULL)
A
Alex Deucher 已提交
423 424 425 426 427 428
		return -ENOMEM;

	return 0;
}

/**
429
 * amdgpu_device_doorbell_fini - Tear down doorbell driver information.
A
Alex Deucher 已提交
430 431 432 433 434
 *
 * @adev: amdgpu_device pointer
 *
 * Tear down doorbell driver information (CIK)
 */
435
static void amdgpu_device_doorbell_fini(struct amdgpu_device *adev)
A
Alex Deucher 已提交
436 437 438 439 440
{
	iounmap(adev->doorbell.ptr);
	adev->doorbell.ptr = NULL;
}

441

A
Alex Deucher 已提交
442 443

/*
444
 * amdgpu_device_wb_*()
445
 * Writeback is the method by which the GPU updates special pages in memory
A
Alex Xie 已提交
446
 * with the status of certain GPU events (fences, ring pointers,etc.).
A
Alex Deucher 已提交
447 448 449
 */

/**
450
 * amdgpu_device_wb_fini - Disable Writeback and free memory
A
Alex Deucher 已提交
451 452 453 454 455 456
 *
 * @adev: amdgpu_device pointer
 *
 * Disables Writeback and frees the Writeback memory (all asics).
 * Used at driver shutdown.
 */
457
static void amdgpu_device_wb_fini(struct amdgpu_device *adev)
A
Alex Deucher 已提交
458 459
{
	if (adev->wb.wb_obj) {
460 461 462
		amdgpu_bo_free_kernel(&adev->wb.wb_obj,
				      &adev->wb.gpu_addr,
				      (void **)&adev->wb.wb);
A
Alex Deucher 已提交
463 464 465 466 467
		adev->wb.wb_obj = NULL;
	}
}

/**
468
 * amdgpu_device_wb_init- Init Writeback driver info and allocate memory
A
Alex Deucher 已提交
469 470 471
 *
 * @adev: amdgpu_device pointer
 *
472
 * Initializes writeback and allocates writeback memory (all asics).
A
Alex Deucher 已提交
473 474 475
 * Used at driver startup.
 * Returns 0 on success or an -error on failure.
 */
476
static int amdgpu_device_wb_init(struct amdgpu_device *adev)
A
Alex Deucher 已提交
477 478 479 480
{
	int r;

	if (adev->wb.wb_obj == NULL) {
481 482
		/* AMDGPU_MAX_WB * sizeof(uint32_t) * 8 = AMDGPU_MAX_WB 256bit slots */
		r = amdgpu_bo_create_kernel(adev, AMDGPU_MAX_WB * sizeof(uint32_t) * 8,
483 484 485
					    PAGE_SIZE, AMDGPU_GEM_DOMAIN_GTT,
					    &adev->wb.wb_obj, &adev->wb.gpu_addr,
					    (void **)&adev->wb.wb);
A
Alex Deucher 已提交
486 487 488 489 490 491 492 493 494
		if (r) {
			dev_warn(adev->dev, "(%d) create WB bo failed\n", r);
			return r;
		}

		adev->wb.num_wb = AMDGPU_MAX_WB;
		memset(&adev->wb.used, 0, sizeof(adev->wb.used));

		/* clear wb memory */
H
Huang Rui 已提交
495
		memset((char *)adev->wb.wb, 0, AMDGPU_MAX_WB * sizeof(uint32_t));
A
Alex Deucher 已提交
496 497 498 499 500 501
	}

	return 0;
}

/**
502
 * amdgpu_device_wb_get - Allocate a wb entry
A
Alex Deucher 已提交
503 504 505 506 507 508 509
 *
 * @adev: amdgpu_device pointer
 * @wb: wb index
 *
 * Allocate a wb slot for use by the driver (all asics).
 * Returns 0 on success or -EINVAL on failure.
 */
510
int amdgpu_device_wb_get(struct amdgpu_device *adev, u32 *wb)
A
Alex Deucher 已提交
511 512 513
{
	unsigned long offset = find_first_zero_bit(adev->wb.used, adev->wb.num_wb);

514
	if (offset < adev->wb.num_wb) {
K
Ken Wang 已提交
515
		__set_bit(offset, adev->wb.used);
M
Monk Liu 已提交
516
		*wb = offset << 3; /* convert to dw offset */
517 518 519 520 521 522
		return 0;
	} else {
		return -EINVAL;
	}
}

A
Alex Deucher 已提交
523
/**
524
 * amdgpu_device_wb_free - Free a wb entry
A
Alex Deucher 已提交
525 526 527 528 529 530
 *
 * @adev: amdgpu_device pointer
 * @wb: wb index
 *
 * Free a wb slot allocated for use by the driver (all asics)
 */
531
void amdgpu_device_wb_free(struct amdgpu_device *adev, u32 wb)
A
Alex Deucher 已提交
532 533
{
	if (wb < adev->wb.num_wb)
M
Monk Liu 已提交
534
		__clear_bit(wb >> 3, adev->wb.used);
A
Alex Deucher 已提交
535 536 537
}

/**
538
 * amdgpu_device_vram_location - try to find VRAM location
A
Alex Deucher 已提交
539 540 541 542
 * @adev: amdgpu device structure holding all necessary informations
 * @mc: memory controller structure holding memory informations
 * @base: base address at which to put VRAM
 *
543
 * Function will try to place VRAM at base address provided
544
 * as parameter.
A
Alex Deucher 已提交
545
 */
546 547
void amdgpu_device_vram_location(struct amdgpu_device *adev,
				 struct amdgpu_mc *mc, u64 base)
A
Alex Deucher 已提交
548 549 550 551 552 553 554 555 556 557 558 559 560
{
	uint64_t limit = (uint64_t)amdgpu_vram_limit << 20;

	mc->vram_start = base;
	mc->vram_end = mc->vram_start + mc->mc_vram_size - 1;
	if (limit && limit < mc->real_vram_size)
		mc->real_vram_size = limit;
	dev_info(adev->dev, "VRAM: %lluM 0x%016llX - 0x%016llX (%lluM used)\n",
			mc->mc_vram_size >> 20, mc->vram_start,
			mc->vram_end, mc->real_vram_size >> 20);
}

/**
561
 * amdgpu_device_gart_location - try to find GTT location
A
Alex Deucher 已提交
562 563 564 565 566 567 568 569 570 571
 * @adev: amdgpu device structure holding all necessary informations
 * @mc: memory controller structure holding memory informations
 *
 * Function will place try to place GTT before or after VRAM.
 *
 * If GTT size is bigger than space left then we ajust GTT size.
 * Thus function will never fails.
 *
 * FIXME: when reducing GTT size align new size on power of 2.
 */
572 573
void amdgpu_device_gart_location(struct amdgpu_device *adev,
				 struct amdgpu_mc *mc)
A
Alex Deucher 已提交
574 575 576
{
	u64 size_af, size_bf;

577 578
	size_af = adev->mc.mc_mask - mc->vram_end;
	size_bf = mc->vram_start;
A
Alex Deucher 已提交
579
	if (size_bf > size_af) {
580
		if (mc->gart_size > size_bf) {
A
Alex Deucher 已提交
581
			dev_warn(adev->dev, "limiting GTT\n");
582
			mc->gart_size = size_bf;
A
Alex Deucher 已提交
583
		}
584
		mc->gart_start = 0;
A
Alex Deucher 已提交
585
	} else {
586
		if (mc->gart_size > size_af) {
A
Alex Deucher 已提交
587
			dev_warn(adev->dev, "limiting GTT\n");
588
			mc->gart_size = size_af;
A
Alex Deucher 已提交
589
		}
590 591 592 593
		/* VCE doesn't like it when BOs cross a 4GB segment, so align
		 * the GART base on a 4GB boundary as well.
		 */
		mc->gart_start = ALIGN(mc->vram_end + 1, 0x100000000ULL);
A
Alex Deucher 已提交
594
	}
595
	mc->gart_end = mc->gart_start + mc->gart_size - 1;
A
Alex Deucher 已提交
596
	dev_info(adev->dev, "GTT: %lluM 0x%016llX - 0x%016llX\n",
597
			mc->gart_size >> 20, mc->gart_start, mc->gart_end);
A
Alex Deucher 已提交
598 599
}

600 601 602 603 604 605 606 607 608 609 610 611 612
/**
 * amdgpu_device_resize_fb_bar - try to resize FB BAR
 *
 * @adev: amdgpu_device pointer
 *
 * Try to resize FB BAR to make all VRAM CPU accessible. We try very hard not
 * to fail, but if any of the BARs is not accessible after the size we abort
 * driver loading by returning -ENODEV.
 */
int amdgpu_device_resize_fb_bar(struct amdgpu_device *adev)
{
	u64 space_needed = roundup_pow_of_two(adev->mc.real_vram_size);
	u32 rbar_size = order_base_2(((space_needed >> 20) | 1)) - 1;
613 614 615
	struct pci_bus *root;
	struct resource *res;
	unsigned i;
616 617 618
	u16 cmd;
	int r;

619 620 621 622
	/* Bypass for VF */
	if (amdgpu_sriov_vf(adev))
		return 0;

623 624 625 626 627 628 629 630 631 632 633 634 635 636 637
	/* Check if the root BUS has 64bit memory resources */
	root = adev->pdev->bus;
	while (root->parent)
		root = root->parent;

	pci_bus_for_each_resource(root, res, i) {
		if (res && res->flags & IORESOURCE_MEM_64 &&
		    res->start > 0x100000000ull)
			break;
	}

	/* Trying to resize is pointless without a root hub window above 4GB */
	if (!res)
		return 0;

638 639 640 641 642 643
	/* Disable memory decoding while we change the BAR addresses and size */
	pci_read_config_word(adev->pdev, PCI_COMMAND, &cmd);
	pci_write_config_word(adev->pdev, PCI_COMMAND,
			      cmd & ~PCI_COMMAND_MEMORY);

	/* Free the VRAM and doorbell BAR, we most likely need to move both. */
644
	amdgpu_device_doorbell_fini(adev);
645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660
	if (adev->asic_type >= CHIP_BONAIRE)
		pci_release_resource(adev->pdev, 2);

	pci_release_resource(adev->pdev, 0);

	r = pci_resize_resource(adev->pdev, 0, rbar_size);
	if (r == -ENOSPC)
		DRM_INFO("Not enough PCI address space for a large BAR.");
	else if (r && r != -ENOTSUPP)
		DRM_ERROR("Problem resizing BAR0 (%d).", r);

	pci_assign_unassigned_bus_resources(adev->pdev->bus);

	/* When the doorbell or fb BAR isn't available we have no chance of
	 * using the device.
	 */
661
	r = amdgpu_device_doorbell_init(adev);
662 663 664 665 666 667 668
	if (r || (pci_resource_flags(adev->pdev, 0) & IORESOURCE_UNSET))
		return -ENODEV;

	pci_write_config_word(adev->pdev, PCI_COMMAND, cmd);

	return 0;
}
669

A
Alex Deucher 已提交
670 671 672 673
/*
 * GPU helpers function.
 */
/**
674
 * amdgpu_need_post - check if the hw need post or not
A
Alex Deucher 已提交
675 676 677
 *
 * @adev: amdgpu_device pointer
 *
678 679 680
 * Check if the asic has been initialized (all asics) at driver startup
 * or post is needed if  hw reset is performed.
 * Returns true if need or false if not.
A
Alex Deucher 已提交
681
 */
682
bool amdgpu_need_post(struct amdgpu_device *adev)
A
Alex Deucher 已提交
683 684 685
{
	uint32_t reg;

686 687 688 689
	if (amdgpu_sriov_vf(adev))
		return false;

	if (amdgpu_passthrough(adev)) {
M
Monk Liu 已提交
690 691 692 693
		/* for FIJI: In whole GPU pass-through virtualization case, after VM reboot
		 * some old smc fw still need driver do vPost otherwise gpu hang, while
		 * those smc fw version above 22.15 doesn't have this flaw, so we force
		 * vpost executed for smc version below 22.15
694 695 696 697 698 699 700 701 702 703
		 */
		if (adev->asic_type == CHIP_FIJI) {
			int err;
			uint32_t fw_ver;
			err = request_firmware(&adev->pm.fw, "amdgpu/fiji_smc.bin", adev->dev);
			/* force vPost if error occured */
			if (err)
				return true;

			fw_ver = *((uint32_t *)adev->pm.fw->data + 69);
M
Monk Liu 已提交
704 705
			if (fw_ver < 0x00160e00)
				return true;
706 707
		}
	}
708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724

	if (adev->has_hw_reset) {
		adev->has_hw_reset = false;
		return true;
	}

	/* bios scratch used on CIK+ */
	if (adev->asic_type >= CHIP_BONAIRE)
		return amdgpu_atombios_scratch_need_asic_init(adev);

	/* check MEM_SIZE for older asics */
	reg = amdgpu_asic_get_config_memsize(adev);

	if ((reg != 0) && (reg != 0xffffffff))
		return false;

	return true;
725 726
}

A
Alex Deucher 已提交
727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773
/**
 * amdgpu_dummy_page_init - init dummy page used by the driver
 *
 * @adev: amdgpu_device pointer
 *
 * Allocate the dummy page used by the driver (all asics).
 * This dummy page is used by the driver as a filler for gart entries
 * when pages are taken out of the GART
 * Returns 0 on sucess, -ENOMEM on failure.
 */
int amdgpu_dummy_page_init(struct amdgpu_device *adev)
{
	if (adev->dummy_page.page)
		return 0;
	adev->dummy_page.page = alloc_page(GFP_DMA32 | GFP_KERNEL | __GFP_ZERO);
	if (adev->dummy_page.page == NULL)
		return -ENOMEM;
	adev->dummy_page.addr = pci_map_page(adev->pdev, adev->dummy_page.page,
					0, PAGE_SIZE, PCI_DMA_BIDIRECTIONAL);
	if (pci_dma_mapping_error(adev->pdev, adev->dummy_page.addr)) {
		dev_err(&adev->pdev->dev, "Failed to DMA MAP the dummy page\n");
		__free_page(adev->dummy_page.page);
		adev->dummy_page.page = NULL;
		return -ENOMEM;
	}
	return 0;
}

/**
 * amdgpu_dummy_page_fini - free dummy page used by the driver
 *
 * @adev: amdgpu_device pointer
 *
 * Frees the dummy page used by the driver (all asics).
 */
void amdgpu_dummy_page_fini(struct amdgpu_device *adev)
{
	if (adev->dummy_page.page == NULL)
		return;
	pci_unmap_page(adev->pdev, adev->dummy_page.addr,
			PAGE_SIZE, PCI_DMA_BIDIRECTIONAL);
	__free_page(adev->dummy_page.page);
	adev->dummy_page.page = NULL;
}

/* if we get transitioned to only one device, take VGA back */
/**
774
 * amdgpu_device_vga_set_decode - enable/disable vga decode
A
Alex Deucher 已提交
775 776 777 778 779 780 781
 *
 * @cookie: amdgpu_device pointer
 * @state: enable/disable vga decode
 *
 * Enable/disable vga decode (all asics).
 * Returns VGA resource flags.
 */
782
static unsigned int amdgpu_device_vga_set_decode(void *cookie, bool state)
A
Alex Deucher 已提交
783 784 785 786 787 788 789 790 791 792
{
	struct amdgpu_device *adev = cookie;
	amdgpu_asic_set_vga_state(adev, state);
	if (state)
		return VGA_RSRC_LEGACY_IO | VGA_RSRC_LEGACY_MEM |
		       VGA_RSRC_NORMAL_IO | VGA_RSRC_NORMAL_MEM;
	else
		return VGA_RSRC_NORMAL_IO | VGA_RSRC_NORMAL_MEM;
}

793
static void amdgpu_device_check_block_size(struct amdgpu_device *adev)
794 795 796 797
{
	/* defines number of bits in page table versus page directory,
	 * a page is 4KB so we have 12 bits offset, minimum 9 bits in the
	 * page table and the remaining bits are in the page directory */
798 799
	if (amdgpu_vm_block_size == -1)
		return;
800

801
	if (amdgpu_vm_block_size < 9) {
802 803
		dev_warn(adev->dev, "VM page table size (%d) too small\n",
			 amdgpu_vm_block_size);
804
		amdgpu_vm_block_size = -1;
805 806 807
	}
}

808
static void amdgpu_device_check_vm_size(struct amdgpu_device *adev)
809
{
810 811 812 813
	/* no need to check the default value */
	if (amdgpu_vm_size == -1)
		return;

814 815 816
	if (amdgpu_vm_size < 1) {
		dev_warn(adev->dev, "VM size (%d) too small, min is 1GB\n",
			 amdgpu_vm_size);
817
		amdgpu_vm_size = -1;
818 819 820
	}
}

A
Alex Deucher 已提交
821
/**
822
 * amdgpu_device_check_arguments - validate module params
A
Alex Deucher 已提交
823 824 825 826 827 828
 *
 * @adev: amdgpu_device pointer
 *
 * Validates certain module parameters and updates
 * the associated values used by the driver (all asics).
 */
829
static void amdgpu_device_check_arguments(struct amdgpu_device *adev)
A
Alex Deucher 已提交
830
{
831 832 833 834
	if (amdgpu_sched_jobs < 4) {
		dev_warn(adev->dev, "sched jobs (%d) must be at least 4\n",
			 amdgpu_sched_jobs);
		amdgpu_sched_jobs = 4;
835
	} else if (!is_power_of_2(amdgpu_sched_jobs)){
836 837 838 839
		dev_warn(adev->dev, "sched jobs (%d) must be a power of 2\n",
			 amdgpu_sched_jobs);
		amdgpu_sched_jobs = roundup_pow_of_two(amdgpu_sched_jobs);
	}
A
Alex Deucher 已提交
840

841
	if (amdgpu_gart_size != -1 && amdgpu_gart_size < 32) {
842 843 844
		/* gart size must be greater or equal to 32M */
		dev_warn(adev->dev, "gart size (%d) too small\n",
			 amdgpu_gart_size);
845
		amdgpu_gart_size = -1;
A
Alex Deucher 已提交
846 847
	}

848
	if (amdgpu_gtt_size != -1 && amdgpu_gtt_size < 32) {
849
		/* gtt size must be greater or equal to 32M */
850 851 852
		dev_warn(adev->dev, "gtt size (%d) too small\n",
				 amdgpu_gtt_size);
		amdgpu_gtt_size = -1;
A
Alex Deucher 已提交
853 854
	}

855 856 857 858 859 860 861
	/* valid range is between 4 and 9 inclusive */
	if (amdgpu_vm_fragment_size != -1 &&
	    (amdgpu_vm_fragment_size > 9 || amdgpu_vm_fragment_size < 4)) {
		dev_warn(adev->dev, "valid range is between 4 and 9\n");
		amdgpu_vm_fragment_size = -1;
	}

862
	amdgpu_device_check_vm_size(adev);
A
Alex Deucher 已提交
863

864
	amdgpu_device_check_block_size(adev);
C
Christian König 已提交
865

866
	if (amdgpu_vram_page_split != -1 && (amdgpu_vram_page_split < 16 ||
867
	    !is_power_of_2(amdgpu_vram_page_split))) {
C
Christian König 已提交
868 869 870 871
		dev_warn(adev->dev, "invalid VRAM page split (%d)\n",
			 amdgpu_vram_page_split);
		amdgpu_vram_page_split = 1024;
	}
872 873 874 875 876

	if (amdgpu_lockup_timeout == 0) {
		dev_warn(adev->dev, "lockup_timeout msut be > 0, adjusting to 10000\n");
		amdgpu_lockup_timeout = 10000;
	}
A
Alex Deucher 已提交
877 878 879 880 881 882
}

/**
 * amdgpu_switcheroo_set_state - set switcheroo state
 *
 * @pdev: pci dev pointer
883
 * @state: vga_switcheroo state
A
Alex Deucher 已提交
884 885 886 887 888 889 890 891 892 893 894 895
 *
 * Callback for the switcheroo driver.  Suspends or resumes the
 * the asics before or after it is powered up using ACPI methods.
 */
static void amdgpu_switcheroo_set_state(struct pci_dev *pdev, enum vga_switcheroo_state state)
{
	struct drm_device *dev = pci_get_drvdata(pdev);

	if (amdgpu_device_is_px(dev) && state == VGA_SWITCHEROO_OFF)
		return;

	if (state == VGA_SWITCHEROO_ON) {
896
		pr_info("amdgpu: switched on\n");
A
Alex Deucher 已提交
897 898 899
		/* don't suspend or resume card normally */
		dev->switch_power_state = DRM_SWITCH_POWER_CHANGING;

900
		amdgpu_device_resume(dev, true, true);
A
Alex Deucher 已提交
901 902 903 904

		dev->switch_power_state = DRM_SWITCH_POWER_ON;
		drm_kms_helper_poll_enable(dev);
	} else {
905
		pr_info("amdgpu: switched off\n");
A
Alex Deucher 已提交
906 907
		drm_kms_helper_poll_disable(dev);
		dev->switch_power_state = DRM_SWITCH_POWER_CHANGING;
908
		amdgpu_device_suspend(dev, true, true);
A
Alex Deucher 已提交
909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939
		dev->switch_power_state = DRM_SWITCH_POWER_OFF;
	}
}

/**
 * amdgpu_switcheroo_can_switch - see if switcheroo state can change
 *
 * @pdev: pci dev pointer
 *
 * Callback for the switcheroo driver.  Check of the switcheroo
 * state can be changed.
 * Returns true if the state can be changed, false if not.
 */
static bool amdgpu_switcheroo_can_switch(struct pci_dev *pdev)
{
	struct drm_device *dev = pci_get_drvdata(pdev);

	/*
	* FIXME: open_count is protected by drm_global_mutex but that would lead to
	* locking inversion with the driver load path. And the access here is
	* completely racy anyway. So don't bother with locking for now.
	*/
	return dev->open_count == 0;
}

static const struct vga_switcheroo_client_ops amdgpu_switcheroo_ops = {
	.set_gpu_state = amdgpu_switcheroo_set_state,
	.reprobe = NULL,
	.can_switch = amdgpu_switcheroo_can_switch,
};

940 941 942
int amdgpu_device_ip_set_clockgating_state(struct amdgpu_device *adev,
					   enum amd_ip_block_type block_type,
					   enum amd_clockgating_state state)
A
Alex Deucher 已提交
943 944 945 946
{
	int i, r = 0;

	for (i = 0; i < adev->num_ip_blocks; i++) {
947
		if (!adev->ip_blocks[i].status.valid)
948
			continue;
949 950 951 952 953 954 955 956 957
		if (adev->ip_blocks[i].version->type != block_type)
			continue;
		if (!adev->ip_blocks[i].version->funcs->set_clockgating_state)
			continue;
		r = adev->ip_blocks[i].version->funcs->set_clockgating_state(
			(void *)adev, state);
		if (r)
			DRM_ERROR("set_clockgating_state of IP block <%s> failed %d\n",
				  adev->ip_blocks[i].version->funcs->name, r);
A
Alex Deucher 已提交
958 959 960 961
	}
	return r;
}

962 963 964
int amdgpu_device_ip_set_powergating_state(struct amdgpu_device *adev,
					   enum amd_ip_block_type block_type,
					   enum amd_powergating_state state)
A
Alex Deucher 已提交
965 966 967 968
{
	int i, r = 0;

	for (i = 0; i < adev->num_ip_blocks; i++) {
969
		if (!adev->ip_blocks[i].status.valid)
970
			continue;
971 972 973 974 975 976 977 978 979
		if (adev->ip_blocks[i].version->type != block_type)
			continue;
		if (!adev->ip_blocks[i].version->funcs->set_powergating_state)
			continue;
		r = adev->ip_blocks[i].version->funcs->set_powergating_state(
			(void *)adev, state);
		if (r)
			DRM_ERROR("set_powergating_state of IP block <%s> failed %d\n",
				  adev->ip_blocks[i].version->funcs->name, r);
A
Alex Deucher 已提交
980 981 982 983
	}
	return r;
}

984 985
void amdgpu_device_ip_get_clockgating_state(struct amdgpu_device *adev,
					    u32 *flags)
986 987 988 989 990 991 992 993 994 995 996
{
	int i;

	for (i = 0; i < adev->num_ip_blocks; i++) {
		if (!adev->ip_blocks[i].status.valid)
			continue;
		if (adev->ip_blocks[i].version->funcs->get_clockgating_state)
			adev->ip_blocks[i].version->funcs->get_clockgating_state((void *)adev, flags);
	}
}

997 998
int amdgpu_device_ip_wait_for_idle(struct amdgpu_device *adev,
				   enum amd_ip_block_type block_type)
999 1000 1001 1002
{
	int i, r;

	for (i = 0; i < adev->num_ip_blocks; i++) {
1003
		if (!adev->ip_blocks[i].status.valid)
1004
			continue;
1005 1006
		if (adev->ip_blocks[i].version->type == block_type) {
			r = adev->ip_blocks[i].version->funcs->wait_for_idle((void *)adev);
1007 1008 1009 1010 1011 1012 1013 1014 1015
			if (r)
				return r;
			break;
		}
	}
	return 0;

}

1016 1017
bool amdgpu_device_ip_is_idle(struct amdgpu_device *adev,
			      enum amd_ip_block_type block_type)
1018 1019 1020 1021
{
	int i;

	for (i = 0; i < adev->num_ip_blocks; i++) {
1022
		if (!adev->ip_blocks[i].status.valid)
1023
			continue;
1024 1025
		if (adev->ip_blocks[i].version->type == block_type)
			return adev->ip_blocks[i].version->funcs->is_idle((void *)adev);
1026 1027 1028 1029 1030
	}
	return true;

}

1031 1032 1033
struct amdgpu_ip_block *
amdgpu_device_ip_get_ip_block(struct amdgpu_device *adev,
			      enum amd_ip_block_type type)
A
Alex Deucher 已提交
1034 1035 1036 1037
{
	int i;

	for (i = 0; i < adev->num_ip_blocks; i++)
1038
		if (adev->ip_blocks[i].version->type == type)
A
Alex Deucher 已提交
1039 1040 1041 1042 1043 1044
			return &adev->ip_blocks[i];

	return NULL;
}

/**
1045
 * amdgpu_device_ip_block_version_cmp
A
Alex Deucher 已提交
1046 1047
 *
 * @adev: amdgpu_device pointer
1048
 * @type: enum amd_ip_block_type
A
Alex Deucher 已提交
1049 1050 1051 1052 1053 1054
 * @major: major version
 * @minor: minor version
 *
 * return 0 if equal or greater
 * return 1 if smaller or the ip_block doesn't exist
 */
1055 1056 1057
int amdgpu_device_ip_block_version_cmp(struct amdgpu_device *adev,
				       enum amd_ip_block_type type,
				       u32 major, u32 minor)
A
Alex Deucher 已提交
1058
{
1059
	struct amdgpu_ip_block *ip_block = amdgpu_device_ip_get_ip_block(adev, type);
A
Alex Deucher 已提交
1060

1061 1062 1063
	if (ip_block && ((ip_block->version->major > major) ||
			((ip_block->version->major == major) &&
			(ip_block->version->minor >= minor))))
A
Alex Deucher 已提交
1064 1065 1066 1067 1068
		return 0;

	return 1;
}

1069
/**
1070
 * amdgpu_device_ip_block_add
1071 1072 1073 1074 1075 1076 1077
 *
 * @adev: amdgpu_device pointer
 * @ip_block_version: pointer to the IP to add
 *
 * Adds the IP block driver information to the collection of IPs
 * on the asic.
 */
1078 1079
int amdgpu_device_ip_block_add(struct amdgpu_device *adev,
			       const struct amdgpu_ip_block_version *ip_block_version)
1080 1081 1082 1083
{
	if (!ip_block_version)
		return -EINVAL;

1084 1085 1086
	DRM_DEBUG("add ip block number %d <%s>\n", adev->num_ip_blocks,
		  ip_block_version->funcs->name);

1087 1088 1089 1090 1091
	adev->ip_blocks[adev->num_ip_blocks++].version = ip_block_version;

	return 0;
}

1092
static void amdgpu_device_enable_virtual_display(struct amdgpu_device *adev)
1093 1094 1095 1096 1097 1098
{
	adev->enable_virtual_display = false;

	if (amdgpu_virtual_display) {
		struct drm_device *ddev = adev->ddev;
		const char *pci_address_name = pci_name(ddev->pdev);
1099
		char *pciaddstr, *pciaddstr_tmp, *pciaddname_tmp, *pciaddname;
1100 1101 1102

		pciaddstr = kstrdup(amdgpu_virtual_display, GFP_KERNEL);
		pciaddstr_tmp = pciaddstr;
1103 1104
		while ((pciaddname_tmp = strsep(&pciaddstr_tmp, ";"))) {
			pciaddname = strsep(&pciaddname_tmp, ",");
1105 1106
			if (!strcmp("all", pciaddname)
			    || !strcmp(pci_address_name, pciaddname)) {
1107 1108 1109
				long num_crtc;
				int res = -1;

1110
				adev->enable_virtual_display = true;
1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124

				if (pciaddname_tmp)
					res = kstrtol(pciaddname_tmp, 10,
						      &num_crtc);

				if (!res) {
					if (num_crtc < 1)
						num_crtc = 1;
					if (num_crtc > 6)
						num_crtc = 6;
					adev->mode_info.num_crtc = num_crtc;
				} else {
					adev->mode_info.num_crtc = 1;
				}
1125 1126 1127 1128
				break;
			}
		}

1129 1130 1131
		DRM_INFO("virtual display string:%s, %s:virtual_display:%d, num_crtc:%d\n",
			 amdgpu_virtual_display, pci_address_name,
			 adev->enable_virtual_display, adev->mode_info.num_crtc);
1132 1133 1134 1135 1136

		kfree(pciaddstr);
	}
}

1137 1138 1139 1140 1141 1142 1143
static int amdgpu_device_parse_gpu_info_fw(struct amdgpu_device *adev)
{
	const char *chip_name;
	char fw_name[30];
	int err;
	const struct gpu_info_firmware_header_v1_0 *hdr;

1144 1145
	adev->firmware.gpu_info_fw = NULL;

1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173
	switch (adev->asic_type) {
	case CHIP_TOPAZ:
	case CHIP_TONGA:
	case CHIP_FIJI:
	case CHIP_POLARIS11:
	case CHIP_POLARIS10:
	case CHIP_POLARIS12:
	case CHIP_CARRIZO:
	case CHIP_STONEY:
#ifdef CONFIG_DRM_AMDGPU_SI
	case CHIP_VERDE:
	case CHIP_TAHITI:
	case CHIP_PITCAIRN:
	case CHIP_OLAND:
	case CHIP_HAINAN:
#endif
#ifdef CONFIG_DRM_AMDGPU_CIK
	case CHIP_BONAIRE:
	case CHIP_HAWAII:
	case CHIP_KAVERI:
	case CHIP_KABINI:
	case CHIP_MULLINS:
#endif
	default:
		return 0;
	case CHIP_VEGA10:
		chip_name = "vega10";
		break;
1174 1175 1176
	case CHIP_RAVEN:
		chip_name = "raven";
		break;
1177 1178 1179
	}

	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_gpu_info.bin", chip_name);
1180
	err = request_firmware(&adev->firmware.gpu_info_fw, fw_name, adev->dev);
1181 1182 1183 1184 1185 1186
	if (err) {
		dev_err(adev->dev,
			"Failed to load gpu_info firmware \"%s\"\n",
			fw_name);
		goto out;
	}
1187
	err = amdgpu_ucode_validate(adev->firmware.gpu_info_fw);
1188 1189 1190 1191 1192 1193 1194
	if (err) {
		dev_err(adev->dev,
			"Failed to validate gpu_info firmware \"%s\"\n",
			fw_name);
		goto out;
	}

1195
	hdr = (const struct gpu_info_firmware_header_v1_0 *)adev->firmware.gpu_info_fw->data;
1196 1197 1198 1199 1200 1201
	amdgpu_ucode_print_gpu_info_hdr(&hdr->header);

	switch (hdr->version_major) {
	case 1:
	{
		const struct gpu_info_firmware_v1_0 *gpu_info_fw =
1202
			(const struct gpu_info_firmware_v1_0 *)(adev->firmware.gpu_info_fw->data +
1203 1204
								le32_to_cpu(hdr->header.ucode_array_offset_bytes));

1205 1206 1207 1208
		adev->gfx.config.max_shader_engines = le32_to_cpu(gpu_info_fw->gc_num_se);
		adev->gfx.config.max_cu_per_sh = le32_to_cpu(gpu_info_fw->gc_num_cu_per_sh);
		adev->gfx.config.max_sh_per_se = le32_to_cpu(gpu_info_fw->gc_num_sh_per_se);
		adev->gfx.config.max_backends_per_se = le32_to_cpu(gpu_info_fw->gc_num_rb_per_se);
1209
		adev->gfx.config.max_texture_channel_caches =
1210 1211 1212 1213 1214
			le32_to_cpu(gpu_info_fw->gc_num_tccs);
		adev->gfx.config.max_gprs = le32_to_cpu(gpu_info_fw->gc_num_gprs);
		adev->gfx.config.max_gs_threads = le32_to_cpu(gpu_info_fw->gc_num_max_gs_thds);
		adev->gfx.config.gs_vgt_table_depth = le32_to_cpu(gpu_info_fw->gc_gs_table_depth);
		adev->gfx.config.gs_prim_buffer_depth = le32_to_cpu(gpu_info_fw->gc_gsprim_buff_depth);
1215
		adev->gfx.config.double_offchip_lds_buf =
1216 1217
			le32_to_cpu(gpu_info_fw->gc_double_offchip_lds_buffer);
		adev->gfx.cu_info.wave_front_size = le32_to_cpu(gpu_info_fw->gc_wave_size);
1218 1219 1220 1221 1222
		adev->gfx.cu_info.max_waves_per_simd =
			le32_to_cpu(gpu_info_fw->gc_max_waves_per_simd);
		adev->gfx.cu_info.max_scratch_slots_per_cu =
			le32_to_cpu(gpu_info_fw->gc_max_scratch_slots_per_cu);
		adev->gfx.cu_info.lds_size = le32_to_cpu(gpu_info_fw->gc_lds_size);
1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234
		break;
	}
	default:
		dev_err(adev->dev,
			"Unsupported gpu_info table %d\n", hdr->header.ucode_version);
		err = -EINVAL;
		goto out;
	}
out:
	return err;
}

1235
static int amdgpu_device_ip_early_init(struct amdgpu_device *adev)
A
Alex Deucher 已提交
1236
{
1237
	int i, r;
A
Alex Deucher 已提交
1238

1239
	amdgpu_device_enable_virtual_display(adev);
1240

A
Alex Deucher 已提交
1241
	switch (adev->asic_type) {
1242 1243
	case CHIP_TOPAZ:
	case CHIP_TONGA:
1244
	case CHIP_FIJI:
1245 1246
	case CHIP_POLARIS11:
	case CHIP_POLARIS10:
1247
	case CHIP_POLARIS12:
1248
	case CHIP_CARRIZO:
1249 1250
	case CHIP_STONEY:
		if (adev->asic_type == CHIP_CARRIZO || adev->asic_type == CHIP_STONEY)
1251 1252 1253 1254 1255 1256 1257 1258
			adev->family = AMDGPU_FAMILY_CZ;
		else
			adev->family = AMDGPU_FAMILY_VI;

		r = vi_set_ip_blocks(adev);
		if (r)
			return r;
		break;
K
Ken Wang 已提交
1259 1260 1261 1262 1263 1264
#ifdef CONFIG_DRM_AMDGPU_SI
	case CHIP_VERDE:
	case CHIP_TAHITI:
	case CHIP_PITCAIRN:
	case CHIP_OLAND:
	case CHIP_HAINAN:
K
Ken Wang 已提交
1265
		adev->family = AMDGPU_FAMILY_SI;
K
Ken Wang 已提交
1266 1267 1268 1269 1270
		r = si_set_ip_blocks(adev);
		if (r)
			return r;
		break;
#endif
1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286
#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
1287 1288 1289 1290 1291 1292
	case  CHIP_VEGA10:
	case  CHIP_RAVEN:
		if (adev->asic_type == CHIP_RAVEN)
			adev->family = AMDGPU_FAMILY_RV;
		else
			adev->family = AMDGPU_FAMILY_AI;
1293 1294 1295 1296 1297

		r = soc15_set_ip_blocks(adev);
		if (r)
			return r;
		break;
A
Alex Deucher 已提交
1298 1299 1300 1301 1302
	default:
		/* FIXME: not supported yet */
		return -EINVAL;
	}

1303 1304 1305 1306
	r = amdgpu_device_parse_gpu_info_fw(adev);
	if (r)
		return r;

1307 1308
	amdgpu_amdkfd_device_probe(adev);

1309 1310 1311
	if (amdgpu_sriov_vf(adev)) {
		r = amdgpu_virt_request_full_gpu(adev, true);
		if (r)
1312
			return -EAGAIN;
1313 1314
	}

A
Alex Deucher 已提交
1315 1316
	for (i = 0; i < adev->num_ip_blocks; i++) {
		if ((amdgpu_ip_block_mask & (1 << i)) == 0) {
1317 1318
			DRM_ERROR("disabled ip block: %d <%s>\n",
				  i, adev->ip_blocks[i].version->funcs->name);
1319
			adev->ip_blocks[i].status.valid = false;
A
Alex Deucher 已提交
1320
		} else {
1321 1322
			if (adev->ip_blocks[i].version->funcs->early_init) {
				r = adev->ip_blocks[i].version->funcs->early_init((void *)adev);
1323
				if (r == -ENOENT) {
1324
					adev->ip_blocks[i].status.valid = false;
1325
				} else if (r) {
1326 1327
					DRM_ERROR("early_init of IP block <%s> failed %d\n",
						  adev->ip_blocks[i].version->funcs->name, r);
A
Alex Deucher 已提交
1328
					return r;
1329
				} else {
1330
					adev->ip_blocks[i].status.valid = true;
1331
				}
1332
			} else {
1333
				adev->ip_blocks[i].status.valid = true;
A
Alex Deucher 已提交
1334 1335 1336 1337
			}
		}
	}

1338 1339 1340
	adev->cg_flags &= amdgpu_cg_mask;
	adev->pg_flags &= amdgpu_pg_mask;

A
Alex Deucher 已提交
1341 1342 1343
	return 0;
}

1344
static int amdgpu_device_ip_init(struct amdgpu_device *adev)
A
Alex Deucher 已提交
1345 1346 1347 1348
{
	int i, r;

	for (i = 0; i < adev->num_ip_blocks; i++) {
1349
		if (!adev->ip_blocks[i].status.valid)
A
Alex Deucher 已提交
1350
			continue;
1351
		r = adev->ip_blocks[i].version->funcs->sw_init((void *)adev);
1352
		if (r) {
1353 1354
			DRM_ERROR("sw_init of IP block <%s> failed %d\n",
				  adev->ip_blocks[i].version->funcs->name, r);
A
Alex Deucher 已提交
1355
			return r;
1356
		}
1357
		adev->ip_blocks[i].status.sw = true;
A
Alex Deucher 已提交
1358
		/* need to do gmc hw init early so we can allocate gpu mem */
1359
		if (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_GMC) {
1360
			r = amdgpu_device_vram_scratch_init(adev);
1361 1362
			if (r) {
				DRM_ERROR("amdgpu_vram_scratch_init failed %d\n", r);
A
Alex Deucher 已提交
1363
				return r;
1364
			}
1365
			r = adev->ip_blocks[i].version->funcs->hw_init((void *)adev);
1366 1367
			if (r) {
				DRM_ERROR("hw_init %d failed %d\n", i, r);
A
Alex Deucher 已提交
1368
				return r;
1369
			}
1370
			r = amdgpu_device_wb_init(adev);
1371
			if (r) {
1372
				DRM_ERROR("amdgpu_device_wb_init failed %d\n", r);
A
Alex Deucher 已提交
1373
				return r;
1374
			}
1375
			adev->ip_blocks[i].status.hw = true;
M
Monk Liu 已提交
1376 1377 1378 1379 1380 1381 1382 1383 1384

			/* right after GMC hw init, we create CSA */
			if (amdgpu_sriov_vf(adev)) {
				r = amdgpu_allocate_static_csa(adev);
				if (r) {
					DRM_ERROR("allocate CSA failed %d\n", r);
					return r;
				}
			}
A
Alex Deucher 已提交
1385 1386 1387 1388
		}
	}

	for (i = 0; i < adev->num_ip_blocks; i++) {
1389
		if (!adev->ip_blocks[i].status.sw)
A
Alex Deucher 已提交
1390 1391
			continue;
		/* gmc hw init is done early */
1392
		if (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_GMC)
A
Alex Deucher 已提交
1393
			continue;
1394
		r = adev->ip_blocks[i].version->funcs->hw_init((void *)adev);
1395
		if (r) {
1396 1397
			DRM_ERROR("hw_init of IP block <%s> failed %d\n",
				  adev->ip_blocks[i].version->funcs->name, r);
A
Alex Deucher 已提交
1398
			return r;
1399
		}
1400
		adev->ip_blocks[i].status.hw = true;
A
Alex Deucher 已提交
1401 1402
	}

1403
	amdgpu_amdkfd_device_init(adev);
1404 1405 1406 1407

	if (amdgpu_sriov_vf(adev))
		amdgpu_virt_release_full_gpu(adev, true);

A
Alex Deucher 已提交
1408 1409 1410
	return 0;
}

1411
static void amdgpu_device_fill_reset_magic(struct amdgpu_device *adev)
1412 1413 1414 1415
{
	memcpy(adev->reset_magic, adev->gart.ptr, AMDGPU_RESET_MAGIC_NUM);
}

1416
static bool amdgpu_device_check_vram_lost(struct amdgpu_device *adev)
1417 1418 1419 1420 1421
{
	return !!memcmp(adev->gart.ptr, adev->reset_magic,
			AMDGPU_RESET_MAGIC_NUM);
}

1422
static int amdgpu_device_ip_late_set_cg_state(struct amdgpu_device *adev)
A
Alex Deucher 已提交
1423 1424 1425 1426
{
	int i = 0, r;

	for (i = 0; i < adev->num_ip_blocks; i++) {
1427
		if (!adev->ip_blocks[i].status.valid)
A
Alex Deucher 已提交
1428
			continue;
1429
		/* skip CG for VCE/UVD, it's handled specially */
1430 1431
		if (adev->ip_blocks[i].version->type != AMD_IP_BLOCK_TYPE_UVD &&
		    adev->ip_blocks[i].version->type != AMD_IP_BLOCK_TYPE_VCE) {
1432
			/* enable clockgating to save power */
1433 1434
			r = adev->ip_blocks[i].version->funcs->set_clockgating_state((void *)adev,
										     AMD_CG_STATE_GATE);
1435 1436
			if (r) {
				DRM_ERROR("set_clockgating_state(gate) of IP block <%s> failed %d\n",
1437
					  adev->ip_blocks[i].version->funcs->name, r);
1438 1439
				return r;
			}
1440
		}
A
Alex Deucher 已提交
1441
	}
1442 1443 1444
	return 0;
}

1445
static int amdgpu_device_ip_late_init(struct amdgpu_device *adev)
1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464
{
	int i = 0, r;

	for (i = 0; i < adev->num_ip_blocks; i++) {
		if (!adev->ip_blocks[i].status.valid)
			continue;
		if (adev->ip_blocks[i].version->funcs->late_init) {
			r = adev->ip_blocks[i].version->funcs->late_init((void *)adev);
			if (r) {
				DRM_ERROR("late_init of IP block <%s> failed %d\n",
					  adev->ip_blocks[i].version->funcs->name, r);
				return r;
			}
			adev->ip_blocks[i].status.late_initialized = true;
		}
	}

	mod_delayed_work(system_wq, &adev->late_init_work,
			msecs_to_jiffies(AMDGPU_RESUME_MS));
A
Alex Deucher 已提交
1465

1466
	amdgpu_device_fill_reset_magic(adev);
A
Alex Deucher 已提交
1467 1468 1469 1470

	return 0;
}

1471
static int amdgpu_device_ip_fini(struct amdgpu_device *adev)
A
Alex Deucher 已提交
1472 1473 1474
{
	int i, r;

1475
	amdgpu_amdkfd_device_fini(adev);
1476 1477
	/* need to disable SMC first */
	for (i = 0; i < adev->num_ip_blocks; i++) {
1478
		if (!adev->ip_blocks[i].status.hw)
1479
			continue;
1480
		if (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_SMC) {
1481
			/* ungate blocks before hw fini so that we can shutdown the blocks safely */
1482 1483
			r = adev->ip_blocks[i].version->funcs->set_clockgating_state((void *)adev,
										     AMD_CG_STATE_UNGATE);
1484 1485
			if (r) {
				DRM_ERROR("set_clockgating_state(ungate) of IP block <%s> failed %d\n",
1486
					  adev->ip_blocks[i].version->funcs->name, r);
1487 1488
				return r;
			}
1489
			r = adev->ip_blocks[i].version->funcs->hw_fini((void *)adev);
1490 1491 1492
			/* XXX handle errors */
			if (r) {
				DRM_DEBUG("hw_fini of IP block <%s> failed %d\n",
1493
					  adev->ip_blocks[i].version->funcs->name, r);
1494
			}
1495
			adev->ip_blocks[i].status.hw = false;
1496 1497 1498 1499
			break;
		}
	}

A
Alex Deucher 已提交
1500
	for (i = adev->num_ip_blocks - 1; i >= 0; i--) {
1501
		if (!adev->ip_blocks[i].status.hw)
A
Alex Deucher 已提交
1502
			continue;
1503
		if (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_GMC) {
M
Monk Liu 已提交
1504
			amdgpu_free_static_csa(adev);
1505 1506
			amdgpu_device_wb_fini(adev);
			amdgpu_device_vram_scratch_fini(adev);
A
Alex Deucher 已提交
1507
		}
1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518

		if (adev->ip_blocks[i].version->type != AMD_IP_BLOCK_TYPE_UVD &&
			adev->ip_blocks[i].version->type != AMD_IP_BLOCK_TYPE_VCE) {
			/* ungate blocks before hw fini so that we can shutdown the blocks safely */
			r = adev->ip_blocks[i].version->funcs->set_clockgating_state((void *)adev,
										     AMD_CG_STATE_UNGATE);
			if (r) {
				DRM_ERROR("set_clockgating_state(ungate) of IP block <%s> failed %d\n",
					  adev->ip_blocks[i].version->funcs->name, r);
				return r;
			}
1519
		}
1520

1521
		r = adev->ip_blocks[i].version->funcs->hw_fini((void *)adev);
A
Alex Deucher 已提交
1522
		/* XXX handle errors */
1523
		if (r) {
1524 1525
			DRM_DEBUG("hw_fini of IP block <%s> failed %d\n",
				  adev->ip_blocks[i].version->funcs->name, r);
1526
		}
1527

1528
		adev->ip_blocks[i].status.hw = false;
A
Alex Deucher 已提交
1529 1530 1531
	}

	for (i = adev->num_ip_blocks - 1; i >= 0; i--) {
1532
		if (!adev->ip_blocks[i].status.sw)
A
Alex Deucher 已提交
1533
			continue;
1534
		r = adev->ip_blocks[i].version->funcs->sw_fini((void *)adev);
A
Alex Deucher 已提交
1535
		/* XXX handle errors */
1536
		if (r) {
1537 1538
			DRM_DEBUG("sw_fini of IP block <%s> failed %d\n",
				  adev->ip_blocks[i].version->funcs->name, r);
1539
		}
1540 1541
		adev->ip_blocks[i].status.sw = false;
		adev->ip_blocks[i].status.valid = false;
A
Alex Deucher 已提交
1542 1543
	}

M
Monk Liu 已提交
1544
	for (i = adev->num_ip_blocks - 1; i >= 0; i--) {
1545
		if (!adev->ip_blocks[i].status.late_initialized)
1546
			continue;
1547 1548 1549
		if (adev->ip_blocks[i].version->funcs->late_fini)
			adev->ip_blocks[i].version->funcs->late_fini((void *)adev);
		adev->ip_blocks[i].status.late_initialized = false;
M
Monk Liu 已提交
1550 1551
	}

1552
	if (amdgpu_sriov_vf(adev))
1553 1554
		if (amdgpu_virt_release_full_gpu(adev, false))
			DRM_ERROR("failed to release exclusive mode on fini\n");
M
Monk Liu 已提交
1555

A
Alex Deucher 已提交
1556 1557 1558
	return 0;
}

1559
static void amdgpu_device_ip_late_init_func_handler(struct work_struct *work)
1560 1561 1562
{
	struct amdgpu_device *adev =
		container_of(work, struct amdgpu_device, late_init_work.work);
1563
	amdgpu_device_ip_late_set_cg_state(adev);
1564 1565
}

1566
int amdgpu_device_ip_suspend(struct amdgpu_device *adev)
A
Alex Deucher 已提交
1567 1568 1569
{
	int i, r;

1570 1571 1572
	if (amdgpu_sriov_vf(adev))
		amdgpu_virt_request_full_gpu(adev, false);

1573
	/* ungate SMC block first */
1574 1575
	r = amdgpu_device_ip_set_clockgating_state(adev, AMD_IP_BLOCK_TYPE_SMC,
						   AMD_CG_STATE_UNGATE);
1576
	if (r) {
1577
		DRM_ERROR("set_clockgating_state(ungate) SMC failed %d\n", r);
1578 1579
	}

A
Alex Deucher 已提交
1580
	for (i = adev->num_ip_blocks - 1; i >= 0; i--) {
1581
		if (!adev->ip_blocks[i].status.valid)
A
Alex Deucher 已提交
1582 1583
			continue;
		/* ungate blocks so that suspend can properly shut them down */
1584
		if (i != AMD_IP_BLOCK_TYPE_SMC) {
1585 1586
			r = adev->ip_blocks[i].version->funcs->set_clockgating_state((void *)adev,
										     AMD_CG_STATE_UNGATE);
1587
			if (r) {
1588 1589
				DRM_ERROR("set_clockgating_state(ungate) of IP block <%s> failed %d\n",
					  adev->ip_blocks[i].version->funcs->name, r);
1590
			}
1591
		}
A
Alex Deucher 已提交
1592
		/* XXX handle errors */
1593
		r = adev->ip_blocks[i].version->funcs->suspend(adev);
A
Alex Deucher 已提交
1594
		/* XXX handle errors */
1595
		if (r) {
1596 1597
			DRM_ERROR("suspend of IP block <%s> failed %d\n",
				  adev->ip_blocks[i].version->funcs->name, r);
1598
		}
A
Alex Deucher 已提交
1599 1600
	}

1601 1602 1603
	if (amdgpu_sriov_vf(adev))
		amdgpu_virt_release_full_gpu(adev, false);

A
Alex Deucher 已提交
1604 1605 1606
	return 0;
}

1607
static int amdgpu_device_ip_reinit_early_sriov(struct amdgpu_device *adev)
1608 1609 1610
{
	int i, r;

1611 1612 1613 1614 1615
	static enum amd_ip_block_type ip_order[] = {
		AMD_IP_BLOCK_TYPE_GMC,
		AMD_IP_BLOCK_TYPE_COMMON,
		AMD_IP_BLOCK_TYPE_IH,
	};
1616

1617 1618 1619
	for (i = 0; i < ARRAY_SIZE(ip_order); i++) {
		int j;
		struct amdgpu_ip_block *block;
1620

1621 1622 1623 1624 1625 1626 1627 1628 1629
		for (j = 0; j < adev->num_ip_blocks; j++) {
			block = &adev->ip_blocks[j];

			if (block->version->type != ip_order[i] ||
				!block->status.valid)
				continue;

			r = block->version->funcs->hw_init(adev);
			DRM_INFO("RE-INIT: %s %s\n", block->version->funcs->name, r?"failed":"successed");
1630 1631 1632 1633 1634 1635
		}
	}

	return 0;
}

1636
static int amdgpu_device_ip_reinit_late_sriov(struct amdgpu_device *adev)
1637 1638 1639
{
	int i, r;

1640 1641
	static enum amd_ip_block_type ip_order[] = {
		AMD_IP_BLOCK_TYPE_SMC,
M
Monk Liu 已提交
1642
		AMD_IP_BLOCK_TYPE_PSP,
1643 1644 1645
		AMD_IP_BLOCK_TYPE_DCE,
		AMD_IP_BLOCK_TYPE_GFX,
		AMD_IP_BLOCK_TYPE_SDMA,
1646 1647
		AMD_IP_BLOCK_TYPE_UVD,
		AMD_IP_BLOCK_TYPE_VCE
1648
	};
1649

1650 1651 1652
	for (i = 0; i < ARRAY_SIZE(ip_order); i++) {
		int j;
		struct amdgpu_ip_block *block;
1653

1654 1655 1656 1657 1658 1659 1660 1661 1662
		for (j = 0; j < adev->num_ip_blocks; j++) {
			block = &adev->ip_blocks[j];

			if (block->version->type != ip_order[i] ||
				!block->status.valid)
				continue;

			r = block->version->funcs->hw_init(adev);
			DRM_INFO("RE-INIT: %s %s\n", block->version->funcs->name, r?"failed":"successed");
1663 1664 1665 1666 1667 1668
		}
	}

	return 0;
}

1669
static int amdgpu_device_ip_resume_phase1(struct amdgpu_device *adev)
A
Alex Deucher 已提交
1670 1671 1672
{
	int i, r;

1673 1674 1675 1676 1677
	for (i = 0; i < adev->num_ip_blocks; i++) {
		if (!adev->ip_blocks[i].status.valid)
			continue;
		if (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_COMMON ||
				adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_GMC ||
1678 1679 1680 1681 1682 1683 1684 1685
				adev->ip_blocks[i].version->type ==
				AMD_IP_BLOCK_TYPE_IH) {
			r = adev->ip_blocks[i].version->funcs->resume(adev);
			if (r) {
				DRM_ERROR("resume of IP block <%s> failed %d\n",
					  adev->ip_blocks[i].version->funcs->name, r);
				return r;
			}
1686 1687 1688 1689 1690 1691
		}
	}

	return 0;
}

1692
static int amdgpu_device_ip_resume_phase2(struct amdgpu_device *adev)
A
Alex Deucher 已提交
1693 1694 1695 1696
{
	int i, r;

	for (i = 0; i < adev->num_ip_blocks; i++) {
1697
		if (!adev->ip_blocks[i].status.valid)
A
Alex Deucher 已提交
1698
			continue;
1699 1700 1701 1702
		if (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_COMMON ||
				adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_GMC ||
				adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_IH )
			continue;
1703
		r = adev->ip_blocks[i].version->funcs->resume(adev);
1704
		if (r) {
1705 1706
			DRM_ERROR("resume of IP block <%s> failed %d\n",
				  adev->ip_blocks[i].version->funcs->name, r);
A
Alex Deucher 已提交
1707
			return r;
1708
		}
A
Alex Deucher 已提交
1709 1710 1711 1712 1713
	}

	return 0;
}

1714
static int amdgpu_device_ip_resume(struct amdgpu_device *adev)
1715 1716 1717
{
	int r;

1718
	r = amdgpu_device_ip_resume_phase1(adev);
1719 1720
	if (r)
		return r;
1721
	r = amdgpu_device_ip_resume_phase2(adev);
1722 1723 1724 1725

	return r;
}

1726
static void amdgpu_device_detect_sriov_bios(struct amdgpu_device *adev)
1727
{
M
Monk Liu 已提交
1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738
	if (amdgpu_sriov_vf(adev)) {
		if (adev->is_atom_fw) {
			if (amdgpu_atomfirmware_gpu_supports_virtualization(adev))
				adev->virt.caps |= AMDGPU_SRIOV_CAPS_SRIOV_VBIOS;
		} else {
			if (amdgpu_atombios_has_gpu_virtualization_table(adev))
				adev->virt.caps |= AMDGPU_SRIOV_CAPS_SRIOV_VBIOS;
		}

		if (!(adev->virt.caps & AMDGPU_SRIOV_CAPS_SRIOV_VBIOS))
			amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_NO_VBIOS, 0, 0);
1739
	}
1740 1741
}

1742 1743 1744 1745 1746 1747
bool amdgpu_device_asic_has_dc_support(enum amd_asic_type asic_type)
{
	switch (asic_type) {
#if defined(CONFIG_DRM_AMD_DC)
	case CHIP_BONAIRE:
	case CHIP_HAWAII:
1748
	case CHIP_KAVERI:
1749 1750 1751 1752
	case CHIP_CARRIZO:
	case CHIP_STONEY:
	case CHIP_POLARIS11:
	case CHIP_POLARIS10:
1753
	case CHIP_POLARIS12:
1754 1755 1756 1757 1758
	case CHIP_TONGA:
	case CHIP_FIJI:
#if defined(CONFIG_DRM_AMD_DC_PRE_VEGA)
		return amdgpu_dc != 0;
#endif
1759 1760 1761
	case CHIP_KABINI:
	case CHIP_MULLINS:
		return amdgpu_dc > 0;
1762 1763
	case CHIP_VEGA10:
#if defined(CONFIG_DRM_AMD_DC_DCN1_0)
1764
	case CHIP_RAVEN:
1765
#endif
1766
		return amdgpu_dc != 0;
1767 1768 1769 1770 1771 1772 1773 1774 1775 1776 1777 1778 1779 1780 1781
#endif
	default:
		return false;
	}
}

/**
 * amdgpu_device_has_dc_support - check if dc is supported
 *
 * @adev: amdgpu_device_pointer
 *
 * Returns true for supported, false for not supported
 */
bool amdgpu_device_has_dc_support(struct amdgpu_device *adev)
{
X
Xiangliang Yu 已提交
1782 1783 1784
	if (amdgpu_sriov_vf(adev))
		return false;

1785 1786 1787
	return amdgpu_device_asic_has_dc_support(adev->asic_type);
}

A
Alex Deucher 已提交
1788 1789 1790 1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806
/**
 * amdgpu_device_init - initialize the driver
 *
 * @adev: amdgpu_device pointer
 * @pdev: drm dev pointer
 * @pdev: pci dev pointer
 * @flags: driver flags
 *
 * Initializes the driver info and hw (all asics).
 * Returns 0 for success or an error on failure.
 * Called at driver startup.
 */
int amdgpu_device_init(struct amdgpu_device *adev,
		       struct drm_device *ddev,
		       struct pci_dev *pdev,
		       uint32_t flags)
{
	int r, i;
	bool runtime = false;
1807
	u32 max_MBps;
A
Alex Deucher 已提交
1808 1809 1810 1811 1812 1813

	adev->shutdown = false;
	adev->dev = &pdev->dev;
	adev->ddev = ddev;
	adev->pdev = pdev;
	adev->flags = flags;
1814
	adev->asic_type = flags & AMD_ASIC_MASK;
A
Alex Deucher 已提交
1815
	adev->usec_timeout = AMDGPU_MAX_USEC_TIMEOUT;
1816
	adev->mc.gart_size = 512 * 1024 * 1024;
A
Alex Deucher 已提交
1817 1818 1819 1820 1821
	adev->accel_working = false;
	adev->num_rings = 0;
	adev->mman.buffer_funcs = NULL;
	adev->mman.buffer_funcs_ring = NULL;
	adev->vm_manager.vm_pte_funcs = NULL;
1822
	adev->vm_manager.vm_pte_num_rings = 0;
A
Alex Deucher 已提交
1823
	adev->gart.gart_funcs = NULL;
1824
	adev->fence_context = dma_fence_context_alloc(AMDGPU_MAX_RINGS);
1825
	bitmap_zero(adev->gfx.pipe_reserve_bitmap, AMDGPU_MAX_COMPUTE_QUEUES);
A
Alex Deucher 已提交
1826 1827 1828 1829 1830

	adev->smc_rreg = &amdgpu_invalid_rreg;
	adev->smc_wreg = &amdgpu_invalid_wreg;
	adev->pcie_rreg = &amdgpu_invalid_rreg;
	adev->pcie_wreg = &amdgpu_invalid_wreg;
1831 1832
	adev->pciep_rreg = &amdgpu_invalid_rreg;
	adev->pciep_wreg = &amdgpu_invalid_wreg;
A
Alex Deucher 已提交
1833 1834 1835 1836
	adev->uvd_ctx_rreg = &amdgpu_invalid_rreg;
	adev->uvd_ctx_wreg = &amdgpu_invalid_wreg;
	adev->didt_rreg = &amdgpu_invalid_rreg;
	adev->didt_wreg = &amdgpu_invalid_wreg;
1837 1838
	adev->gc_cac_rreg = &amdgpu_invalid_rreg;
	adev->gc_cac_wreg = &amdgpu_invalid_wreg;
A
Alex Deucher 已提交
1839 1840 1841
	adev->audio_endpt_rreg = &amdgpu_block_invalid_rreg;
	adev->audio_endpt_wreg = &amdgpu_block_invalid_wreg;

1842 1843 1844
	DRM_INFO("initializing kernel modesetting (%s 0x%04X:0x%04X 0x%04X:0x%04X 0x%02X).\n",
		 amdgpu_asic_name[adev->asic_type], pdev->vendor, pdev->device,
		 pdev->subsystem_vendor, pdev->subsystem_device, pdev->revision);
A
Alex Deucher 已提交
1845 1846 1847 1848

	/* mutex initialization are all done here so we
	 * can recall function without having locking issues */
	atomic_set(&adev->irq.ih.lock, 0);
1849
	mutex_init(&adev->firmware.mutex);
A
Alex Deucher 已提交
1850 1851 1852
	mutex_init(&adev->pm.mutex);
	mutex_init(&adev->gfx.gpu_clock_mutex);
	mutex_init(&adev->srbm_mutex);
1853
	mutex_init(&adev->gfx.pipe_reserve_mutex);
A
Alex Deucher 已提交
1854 1855
	mutex_init(&adev->grbm_idx_mutex);
	mutex_init(&adev->mn_lock);
A
Alex Deucher 已提交
1856
	mutex_init(&adev->virt.vf_errors.lock);
A
Alex Deucher 已提交
1857
	hash_init(adev->mn_hash);
1858
	mutex_init(&adev->lock_reset);
A
Alex Deucher 已提交
1859

1860
	amdgpu_device_check_arguments(adev);
A
Alex Deucher 已提交
1861 1862 1863 1864 1865 1866

	spin_lock_init(&adev->mmio_idx_lock);
	spin_lock_init(&adev->smc_idx_lock);
	spin_lock_init(&adev->pcie_idx_lock);
	spin_lock_init(&adev->uvd_ctx_idx_lock);
	spin_lock_init(&adev->didt_idx_lock);
1867
	spin_lock_init(&adev->gc_cac_idx_lock);
1868
	spin_lock_init(&adev->se_cac_idx_lock);
A
Alex Deucher 已提交
1869
	spin_lock_init(&adev->audio_endpt_idx_lock);
1870
	spin_lock_init(&adev->mm_stats.lock);
A
Alex Deucher 已提交
1871

1872 1873 1874
	INIT_LIST_HEAD(&adev->shadow_list);
	mutex_init(&adev->shadow_list_lock);

1875 1876 1877
	INIT_LIST_HEAD(&adev->ring_lru_list);
	spin_lock_init(&adev->ring_lru_list_lock);

1878 1879
	INIT_DELAYED_WORK(&adev->late_init_work,
			  amdgpu_device_ip_late_init_func_handler);
1880

1881 1882
	/* Registers mapping */
	/* TODO: block userspace mapping of io register */
1883 1884 1885 1886 1887 1888 1889
	if (adev->asic_type >= CHIP_BONAIRE) {
		adev->rmmio_base = pci_resource_start(adev->pdev, 5);
		adev->rmmio_size = pci_resource_len(adev->pdev, 5);
	} else {
		adev->rmmio_base = pci_resource_start(adev->pdev, 2);
		adev->rmmio_size = pci_resource_len(adev->pdev, 2);
	}
A
Alex Deucher 已提交
1890 1891 1892 1893 1894 1895 1896 1897

	adev->rmmio = ioremap(adev->rmmio_base, adev->rmmio_size);
	if (adev->rmmio == NULL) {
		return -ENOMEM;
	}
	DRM_INFO("register mmio base: 0x%08X\n", (uint32_t)adev->rmmio_base);
	DRM_INFO("register mmio size: %u\n", (unsigned)adev->rmmio_size);

1898
	/* doorbell bar mapping */
1899
	amdgpu_device_doorbell_init(adev);
A
Alex Deucher 已提交
1900 1901 1902 1903 1904 1905 1906 1907 1908 1909

	/* io port mapping */
	for (i = 0; i < DEVICE_COUNT_RESOURCE; i++) {
		if (pci_resource_flags(adev->pdev, i) & IORESOURCE_IO) {
			adev->rio_mem_size = pci_resource_len(adev->pdev, i);
			adev->rio_mem = pci_iomap(adev->pdev, i, adev->rio_mem_size);
			break;
		}
	}
	if (adev->rio_mem == NULL)
1910
		DRM_INFO("PCI I/O BAR is not found.\n");
A
Alex Deucher 已提交
1911 1912

	/* early init functions */
1913
	r = amdgpu_device_ip_early_init(adev);
A
Alex Deucher 已提交
1914 1915 1916 1917 1918 1919
	if (r)
		return r;

	/* if we have > 1 VGA cards, then disable the amdgpu VGA resources */
	/* this will fail for cards that aren't VGA class devices, just
	 * ignore it */
1920
	vga_client_register(adev->pdev, adev, NULL, amdgpu_device_vga_set_decode);
A
Alex Deucher 已提交
1921 1922 1923

	if (amdgpu_runtime_pm == 1)
		runtime = true;
1924
	if (amdgpu_device_is_px(ddev))
A
Alex Deucher 已提交
1925
		runtime = true;
1926 1927 1928
	if (!pci_is_thunderbolt_attached(adev->pdev))
		vga_switcheroo_register_client(adev->pdev,
					       &amdgpu_switcheroo_ops, runtime);
A
Alex Deucher 已提交
1929 1930 1931 1932
	if (runtime)
		vga_switcheroo_init_domain_pm_ops(adev->dev, &adev->vga_pm_domain);

	/* Read BIOS */
1933 1934 1935 1936
	if (!amdgpu_get_bios(adev)) {
		r = -EINVAL;
		goto failed;
	}
1937

A
Alex Deucher 已提交
1938
	r = amdgpu_atombios_init(adev);
1939 1940
	if (r) {
		dev_err(adev->dev, "amdgpu_atombios_init failed\n");
A
Alex Deucher 已提交
1941
		amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_ATOMBIOS_INIT_FAIL, 0, 0);
1942
		goto failed;
1943
	}
A
Alex Deucher 已提交
1944

1945 1946
	/* detect if we are with an SRIOV vbios */
	amdgpu_device_detect_sriov_bios(adev);
1947

A
Alex Deucher 已提交
1948
	/* Post card if necessary */
1949
	if (amdgpu_need_post(adev)) {
A
Alex Deucher 已提交
1950
		if (!adev->bios) {
1951
			dev_err(adev->dev, "no vBIOS found\n");
1952 1953
			r = -EINVAL;
			goto failed;
A
Alex Deucher 已提交
1954
		}
1955
		DRM_INFO("GPU posting now...\n");
1956 1957 1958 1959 1960
		r = amdgpu_atom_asic_init(adev->mode_info.atom_context);
		if (r) {
			dev_err(adev->dev, "gpu post error!\n");
			goto failed;
		}
A
Alex Deucher 已提交
1961 1962
	}

1963 1964 1965 1966 1967
	if (adev->is_atom_fw) {
		/* Initialize clocks */
		r = amdgpu_atomfirmware_get_clock_info(adev);
		if (r) {
			dev_err(adev->dev, "amdgpu_atomfirmware_get_clock_info failed\n");
A
Alex Deucher 已提交
1968
			amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_ATOMBIOS_GET_CLOCK_FAIL, 0, 0);
1969 1970 1971
			goto failed;
		}
	} else {
1972 1973 1974 1975
		/* Initialize clocks */
		r = amdgpu_atombios_get_clock_info(adev);
		if (r) {
			dev_err(adev->dev, "amdgpu_atombios_get_clock_info failed\n");
A
Alex Deucher 已提交
1976
			amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_ATOMBIOS_GET_CLOCK_FAIL, 0, 0);
1977
			goto failed;
1978 1979
		}
		/* init i2c buses */
1980 1981
		if (!amdgpu_device_has_dc_support(adev))
			amdgpu_atombios_i2c_init(adev);
1982
	}
A
Alex Deucher 已提交
1983 1984 1985

	/* Fence driver */
	r = amdgpu_fence_driver_init(adev);
1986 1987
	if (r) {
		dev_err(adev->dev, "amdgpu_fence_driver_init failed\n");
A
Alex Deucher 已提交
1988
		amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_FENCE_INIT_FAIL, 0, 0);
1989
		goto failed;
1990
	}
A
Alex Deucher 已提交
1991 1992 1993 1994

	/* init the mode config */
	drm_mode_config_init(adev->ddev);

1995
	r = amdgpu_device_ip_init(adev);
A
Alex Deucher 已提交
1996
	if (r) {
1997 1998 1999 2000 2001 2002
		/* failed in exclusive mode due to timeout */
		if (amdgpu_sriov_vf(adev) &&
		    !amdgpu_sriov_runtime(adev) &&
		    amdgpu_virt_mmio_blocked(adev) &&
		    !amdgpu_virt_wait_reset(adev)) {
			dev_err(adev->dev, "VF exclusive mode timeout\n");
2003 2004 2005
			/* Don't send request since VF is inactive. */
			adev->virt.caps &= ~AMDGPU_SRIOV_CAPS_RUNTIME;
			adev->virt.ops = NULL;
2006 2007 2008
			r = -EAGAIN;
			goto failed;
		}
2009
		dev_err(adev->dev, "amdgpu_device_ip_init failed\n");
A
Alex Deucher 已提交
2010
		amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_AMDGPU_INIT_FAIL, 0, 0);
2011
		amdgpu_device_ip_fini(adev);
2012
		goto failed;
A
Alex Deucher 已提交
2013 2014 2015 2016
	}

	adev->accel_working = true;

2017 2018
	amdgpu_vm_check_compute_bug(adev);

2019 2020 2021 2022 2023 2024 2025 2026
	/* Initialize the buffer migration limit. */
	if (amdgpu_moverate >= 0)
		max_MBps = amdgpu_moverate;
	else
		max_MBps = 8; /* Allow 8 MB/s. */
	/* Get a log2 for easy divisions. */
	adev->mm_stats.log2_max_MBps = ilog2(max(1u, max_MBps));

A
Alex Deucher 已提交
2027 2028 2029
	r = amdgpu_ib_pool_init(adev);
	if (r) {
		dev_err(adev->dev, "IB initialization failed (%d).\n", r);
A
Alex Deucher 已提交
2030
		amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_IB_INIT_FAIL, 0, r);
2031
		goto failed;
A
Alex Deucher 已提交
2032 2033 2034 2035 2036 2037
	}

	r = amdgpu_ib_ring_tests(adev);
	if (r)
		DRM_ERROR("ib ring test failed (%d).\n", r);

2038 2039 2040
	if (amdgpu_sriov_vf(adev))
		amdgpu_virt_init_data_exchange(adev);

2041 2042
	amdgpu_fbdev_init(adev);

2043 2044 2045 2046
	r = amdgpu_pm_sysfs_init(adev);
	if (r)
		DRM_ERROR("registering pm debugfs failed (%d).\n", r);

2047
	r = amdgpu_debugfs_gem_init(adev);
M
Monk Liu 已提交
2048
	if (r)
A
Alex Deucher 已提交
2049 2050 2051
		DRM_ERROR("registering gem debugfs failed (%d).\n", r);

	r = amdgpu_debugfs_regs_init(adev);
M
Monk Liu 已提交
2052
	if (r)
A
Alex Deucher 已提交
2053 2054
		DRM_ERROR("registering register debugfs failed (%d).\n", r);

2055
	r = amdgpu_debugfs_firmware_init(adev);
M
Monk Liu 已提交
2056
	if (r)
2057 2058
		DRM_ERROR("registering firmware debugfs failed (%d).\n", r);

2059
	r = amdgpu_debugfs_init(adev);
2060
	if (r)
2061
		DRM_ERROR("Creating debugfs files failed (%d).\n", r);
2062

A
Alex Deucher 已提交
2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076 2077 2078
	if ((amdgpu_testing & 1)) {
		if (adev->accel_working)
			amdgpu_test_moves(adev);
		else
			DRM_INFO("amdgpu: acceleration disabled, skipping move tests\n");
	}
	if (amdgpu_benchmarking) {
		if (adev->accel_working)
			amdgpu_benchmark(adev, amdgpu_benchmarking);
		else
			DRM_INFO("amdgpu: acceleration disabled, skipping benchmarks\n");
	}

	/* enable clockgating, etc. after ib tests, etc. since some blocks require
	 * explicit gating rather than handling it automatically.
	 */
2079
	r = amdgpu_device_ip_late_init(adev);
2080
	if (r) {
2081
		dev_err(adev->dev, "amdgpu_device_ip_late_init failed\n");
A
Alex Deucher 已提交
2082
		amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_AMDGPU_LATE_INIT_FAIL, 0, r);
2083
		goto failed;
2084
	}
A
Alex Deucher 已提交
2085 2086

	return 0;
2087 2088

failed:
2089
	amdgpu_vf_error_trans_all(adev);
2090 2091
	if (runtime)
		vga_switcheroo_fini_domain_pm_ops(adev->dev);
2092

2093
	return r;
A
Alex Deucher 已提交
2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107 2108 2109
}

/**
 * amdgpu_device_fini - tear down the driver
 *
 * @adev: amdgpu_device pointer
 *
 * Tear down the driver info (all asics).
 * Called at driver shutdown.
 */
void amdgpu_device_fini(struct amdgpu_device *adev)
{
	int r;

	DRM_INFO("amdgpu: finishing device.\n");
	adev->shutdown = true;
2110 2111
	if (adev->mode_info.mode_config_initialized)
		drm_crtc_force_disable_all(adev->ddev);
2112

A
Alex Deucher 已提交
2113 2114 2115
	amdgpu_ib_pool_fini(adev);
	amdgpu_fence_driver_fini(adev);
	amdgpu_fbdev_fini(adev);
2116
	r = amdgpu_device_ip_fini(adev);
2117 2118 2119 2120
	if (adev->firmware.gpu_info_fw) {
		release_firmware(adev->firmware.gpu_info_fw);
		adev->firmware.gpu_info_fw = NULL;
	}
A
Alex Deucher 已提交
2121
	adev->accel_working = false;
2122
	cancel_delayed_work_sync(&adev->late_init_work);
A
Alex Deucher 已提交
2123
	/* free i2c buses */
2124 2125
	if (!amdgpu_device_has_dc_support(adev))
		amdgpu_i2c_fini(adev);
A
Alex Deucher 已提交
2126 2127 2128
	amdgpu_atombios_fini(adev);
	kfree(adev->bios);
	adev->bios = NULL;
2129 2130
	if (!pci_is_thunderbolt_attached(adev->pdev))
		vga_switcheroo_unregister_client(adev->pdev);
2131 2132
	if (adev->flags & AMD_IS_PX)
		vga_switcheroo_fini_domain_pm_ops(adev->dev);
A
Alex Deucher 已提交
2133 2134 2135 2136 2137 2138
	vga_client_register(adev->pdev, NULL, NULL, NULL);
	if (adev->rio_mem)
		pci_iounmap(adev->pdev, adev->rio_mem);
	adev->rio_mem = NULL;
	iounmap(adev->rmmio);
	adev->rmmio = NULL;
2139
	amdgpu_device_doorbell_fini(adev);
2140
	amdgpu_pm_sysfs_fini(adev);
A
Alex Deucher 已提交
2141 2142 2143 2144 2145 2146 2147 2148
	amdgpu_debugfs_regs_cleanup(adev);
}


/*
 * Suspend & resume.
 */
/**
2149
 * amdgpu_device_suspend - initiate device suspend
A
Alex Deucher 已提交
2150 2151 2152 2153 2154 2155 2156 2157
 *
 * @pdev: drm dev pointer
 * @state: suspend state
 *
 * Puts the hw in the suspend state (all asics).
 * Returns 0 for success or an error on failure.
 * Called at driver suspend.
 */
2158
int amdgpu_device_suspend(struct drm_device *dev, bool suspend, bool fbcon)
A
Alex Deucher 已提交
2159 2160 2161 2162
{
	struct amdgpu_device *adev;
	struct drm_crtc *crtc;
	struct drm_connector *connector;
2163
	int r;
A
Alex Deucher 已提交
2164 2165 2166 2167 2168 2169 2170 2171 2172 2173 2174 2175

	if (dev == NULL || dev->dev_private == NULL) {
		return -ENODEV;
	}

	adev = dev->dev_private;

	if (dev->switch_power_state == DRM_SWITCH_POWER_OFF)
		return 0;

	drm_kms_helper_poll_disable(dev);

2176 2177 2178 2179 2180 2181 2182
	if (!amdgpu_device_has_dc_support(adev)) {
		/* turn off display hw */
		drm_modeset_lock_all(dev);
		list_for_each_entry(connector, &dev->mode_config.connector_list, head) {
			drm_helper_connector_dpms(connector, DRM_MODE_DPMS_OFF);
		}
		drm_modeset_unlock_all(dev);
A
Alex Deucher 已提交
2183 2184
	}

2185 2186
	amdgpu_amdkfd_suspend(adev);

2187
	/* unpin the front buffers and cursors */
A
Alex Deucher 已提交
2188
	list_for_each_entry(crtc, &dev->mode_config.crtc_list, head) {
2189
		struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
A
Alex Deucher 已提交
2190 2191 2192
		struct amdgpu_framebuffer *rfb = to_amdgpu_framebuffer(crtc->primary->fb);
		struct amdgpu_bo *robj;

2193 2194
		if (amdgpu_crtc->cursor_bo) {
			struct amdgpu_bo *aobj = gem_to_amdgpu_bo(amdgpu_crtc->cursor_bo);
2195
			r = amdgpu_bo_reserve(aobj, true);
2196 2197 2198 2199 2200 2201
			if (r == 0) {
				amdgpu_bo_unpin(aobj);
				amdgpu_bo_unreserve(aobj);
			}
		}

A
Alex Deucher 已提交
2202 2203 2204 2205 2206 2207
		if (rfb == NULL || rfb->obj == NULL) {
			continue;
		}
		robj = gem_to_amdgpu_bo(rfb->obj);
		/* don't unpin kernel fb objects */
		if (!amdgpu_fbdev_robj_is_fb(adev, robj)) {
2208
			r = amdgpu_bo_reserve(robj, true);
A
Alex Deucher 已提交
2209 2210 2211 2212 2213 2214 2215 2216 2217
			if (r == 0) {
				amdgpu_bo_unpin(robj);
				amdgpu_bo_unreserve(robj);
			}
		}
	}
	/* evict vram memory */
	amdgpu_bo_evict_vram(adev);

2218
	amdgpu_fence_driver_suspend(adev);
A
Alex Deucher 已提交
2219

2220
	r = amdgpu_device_ip_suspend(adev);
A
Alex Deucher 已提交
2221

2222 2223 2224 2225
	/* evict remaining vram memory
	 * This second call to evict vram is to evict the gart page table
	 * using the CPU.
	 */
A
Alex Deucher 已提交
2226 2227 2228 2229 2230 2231 2232
	amdgpu_bo_evict_vram(adev);

	pci_save_state(dev->pdev);
	if (suspend) {
		/* Shut down the device */
		pci_disable_device(dev->pdev);
		pci_set_power_state(dev->pdev, PCI_D3hot);
J
jimqu 已提交
2233 2234 2235 2236
	} else {
		r = amdgpu_asic_reset(adev);
		if (r)
			DRM_ERROR("amdgpu asic reset failed\n");
A
Alex Deucher 已提交
2237 2238 2239 2240 2241 2242 2243 2244 2245 2246 2247
	}

	if (fbcon) {
		console_lock();
		amdgpu_fbdev_set_suspend(adev, 1);
		console_unlock();
	}
	return 0;
}

/**
2248
 * amdgpu_device_resume - initiate device resume
A
Alex Deucher 已提交
2249 2250 2251 2252 2253 2254 2255
 *
 * @pdev: drm dev pointer
 *
 * Bring the hw back to operating state (all asics).
 * Returns 0 for success or an error on failure.
 * Called at driver resume.
 */
2256
int amdgpu_device_resume(struct drm_device *dev, bool resume, bool fbcon)
A
Alex Deucher 已提交
2257 2258 2259
{
	struct drm_connector *connector;
	struct amdgpu_device *adev = dev->dev_private;
2260
	struct drm_crtc *crtc;
2261
	int r = 0;
A
Alex Deucher 已提交
2262 2263 2264 2265

	if (dev->switch_power_state == DRM_SWITCH_POWER_OFF)
		return 0;

J
jimqu 已提交
2266
	if (fbcon)
A
Alex Deucher 已提交
2267
		console_lock();
J
jimqu 已提交
2268

A
Alex Deucher 已提交
2269 2270 2271
	if (resume) {
		pci_set_power_state(dev->pdev, PCI_D0);
		pci_restore_state(dev->pdev);
J
jimqu 已提交
2272
		r = pci_enable_device(dev->pdev);
2273 2274
		if (r)
			goto unlock;
A
Alex Deucher 已提交
2275 2276 2277
	}

	/* post card */
2278
	if (amdgpu_need_post(adev)) {
J
jimqu 已提交
2279 2280 2281 2282
		r = amdgpu_atom_asic_init(adev->mode_info.atom_context);
		if (r)
			DRM_ERROR("amdgpu asic init failed\n");
	}
A
Alex Deucher 已提交
2283

2284
	r = amdgpu_device_ip_resume(adev);
2285
	if (r) {
2286
		DRM_ERROR("amdgpu_device_ip_resume failed (%d).\n", r);
2287
		goto unlock;
2288
	}
2289 2290
	amdgpu_fence_driver_resume(adev);

F
Flora Cui 已提交
2291 2292 2293 2294 2295
	if (resume) {
		r = amdgpu_ib_ring_tests(adev);
		if (r)
			DRM_ERROR("ib ring test failed (%d).\n", r);
	}
A
Alex Deucher 已提交
2296

2297
	r = amdgpu_device_ip_late_init(adev);
2298 2299
	if (r)
		goto unlock;
A
Alex Deucher 已提交
2300

2301 2302 2303 2304 2305 2306
	/* pin cursors */
	list_for_each_entry(crtc, &dev->mode_config.crtc_list, head) {
		struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);

		if (amdgpu_crtc->cursor_bo) {
			struct amdgpu_bo *aobj = gem_to_amdgpu_bo(amdgpu_crtc->cursor_bo);
2307
			r = amdgpu_bo_reserve(aobj, true);
2308 2309 2310 2311 2312 2313 2314 2315 2316 2317
			if (r == 0) {
				r = amdgpu_bo_pin(aobj,
						  AMDGPU_GEM_DOMAIN_VRAM,
						  &amdgpu_crtc->cursor_addr);
				if (r != 0)
					DRM_ERROR("Failed to pin cursor BO (%d)\n", r);
				amdgpu_bo_unreserve(aobj);
			}
		}
	}
2318 2319 2320
	r = amdgpu_amdkfd_resume(adev);
	if (r)
		return r;
2321

A
Alex Deucher 已提交
2322 2323
	/* blat the mode back in */
	if (fbcon) {
2324 2325 2326 2327 2328 2329 2330 2331 2332 2333 2334 2335 2336 2337 2338 2339 2340 2341
		if (!amdgpu_device_has_dc_support(adev)) {
			/* pre DCE11 */
			drm_helper_resume_force_mode(dev);

			/* turn on display hw */
			drm_modeset_lock_all(dev);
			list_for_each_entry(connector, &dev->mode_config.connector_list, head) {
				drm_helper_connector_dpms(connector, DRM_MODE_DPMS_ON);
			}
			drm_modeset_unlock_all(dev);
		} else {
			/*
			 * There is no equivalent atomic helper to turn on
			 * display, so we defined our own function for this,
			 * once suspend resume is supported by the atomic
			 * framework this will be reworked
			 */
			amdgpu_dm_display_resume(adev);
A
Alex Deucher 已提交
2342 2343 2344 2345
		}
	}

	drm_kms_helper_poll_enable(dev);
2346 2347 2348 2349 2350 2351 2352 2353 2354 2355 2356 2357 2358

	/*
	 * Most of the connector probing functions try to acquire runtime pm
	 * refs to ensure that the GPU is powered on when connector polling is
	 * performed. Since we're calling this from a runtime PM callback,
	 * trying to acquire rpm refs will cause us to deadlock.
	 *
	 * Since we're guaranteed to be holding the rpm lock, it's safe to
	 * temporarily disable the rpm helpers so this doesn't deadlock us.
	 */
#ifdef CONFIG_PM
	dev->dev->power.disable_depth++;
#endif
2359 2360 2361 2362
	if (!amdgpu_device_has_dc_support(adev))
		drm_helper_hpd_irq_event(dev);
	else
		drm_kms_helper_hotplug_event(dev);
2363 2364 2365
#ifdef CONFIG_PM
	dev->dev->power.disable_depth--;
#endif
A
Alex Deucher 已提交
2366

2367
	if (fbcon)
A
Alex Deucher 已提交
2368
		amdgpu_fbdev_set_suspend(adev, 0);
2369 2370 2371

unlock:
	if (fbcon)
A
Alex Deucher 已提交
2372 2373
		console_unlock();

2374
	return r;
A
Alex Deucher 已提交
2375 2376
}

2377
static bool amdgpu_device_ip_check_soft_reset(struct amdgpu_device *adev)
2378 2379 2380 2381
{
	int i;
	bool asic_hang = false;

2382 2383 2384
	if (amdgpu_sriov_vf(adev))
		return true;

2385
	for (i = 0; i < adev->num_ip_blocks; i++) {
2386
		if (!adev->ip_blocks[i].status.valid)
2387
			continue;
2388 2389 2390 2391 2392
		if (adev->ip_blocks[i].version->funcs->check_soft_reset)
			adev->ip_blocks[i].status.hang =
				adev->ip_blocks[i].version->funcs->check_soft_reset(adev);
		if (adev->ip_blocks[i].status.hang) {
			DRM_INFO("IP block:%s is hung!\n", adev->ip_blocks[i].version->funcs->name);
2393 2394 2395 2396 2397 2398
			asic_hang = true;
		}
	}
	return asic_hang;
}

2399
static int amdgpu_device_ip_pre_soft_reset(struct amdgpu_device *adev)
2400 2401 2402 2403
{
	int i, r = 0;

	for (i = 0; i < adev->num_ip_blocks; i++) {
2404
		if (!adev->ip_blocks[i].status.valid)
2405
			continue;
2406 2407 2408
		if (adev->ip_blocks[i].status.hang &&
		    adev->ip_blocks[i].version->funcs->pre_soft_reset) {
			r = adev->ip_blocks[i].version->funcs->pre_soft_reset(adev);
2409 2410 2411 2412 2413 2414 2415 2416
			if (r)
				return r;
		}
	}

	return 0;
}

2417
static bool amdgpu_device_ip_need_full_reset(struct amdgpu_device *adev)
2418
{
2419 2420 2421
	int i;

	for (i = 0; i < adev->num_ip_blocks; i++) {
2422
		if (!adev->ip_blocks[i].status.valid)
2423
			continue;
2424 2425 2426
		if ((adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_GMC) ||
		    (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_SMC) ||
		    (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_ACP) ||
2427 2428
		    (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_DCE) ||
		     adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_PSP) {
2429
			if (adev->ip_blocks[i].status.hang) {
2430 2431 2432 2433
				DRM_INFO("Some block need full reset!\n");
				return true;
			}
		}
2434 2435 2436 2437
	}
	return false;
}

2438
static int amdgpu_device_ip_soft_reset(struct amdgpu_device *adev)
2439 2440 2441 2442
{
	int i, r = 0;

	for (i = 0; i < adev->num_ip_blocks; i++) {
2443
		if (!adev->ip_blocks[i].status.valid)
2444
			continue;
2445 2446 2447
		if (adev->ip_blocks[i].status.hang &&
		    adev->ip_blocks[i].version->funcs->soft_reset) {
			r = adev->ip_blocks[i].version->funcs->soft_reset(adev);
2448 2449 2450 2451 2452 2453 2454 2455
			if (r)
				return r;
		}
	}

	return 0;
}

2456
static int amdgpu_device_ip_post_soft_reset(struct amdgpu_device *adev)
2457 2458 2459 2460
{
	int i, r = 0;

	for (i = 0; i < adev->num_ip_blocks; i++) {
2461
		if (!adev->ip_blocks[i].status.valid)
2462
			continue;
2463 2464 2465
		if (adev->ip_blocks[i].status.hang &&
		    adev->ip_blocks[i].version->funcs->post_soft_reset)
			r = adev->ip_blocks[i].version->funcs->post_soft_reset(adev);
2466 2467 2468 2469 2470 2471 2472
		if (r)
			return r;
	}

	return 0;
}

2473 2474 2475 2476 2477
bool amdgpu_need_backup(struct amdgpu_device *adev)
{
	if (adev->flags & AMD_IS_APU)
		return false;

2478
	return amdgpu_gpu_recovery;
2479 2480
}

2481 2482 2483 2484
static int amdgpu_device_recover_vram_from_shadow(struct amdgpu_device *adev,
						  struct amdgpu_ring *ring,
						  struct amdgpu_bo *bo,
						  struct dma_fence **fence)
2485 2486 2487 2488
{
	uint32_t domain;
	int r;

R
Roger.He 已提交
2489 2490 2491
	if (!bo->shadow)
		return 0;

2492
	r = amdgpu_bo_reserve(bo, true);
R
Roger.He 已提交
2493 2494 2495 2496 2497
	if (r)
		return r;
	domain = amdgpu_mem_type_to_domain(bo->tbo.mem.mem_type);
	/* if bo has been evicted, then no need to recover */
	if (domain == AMDGPU_GEM_DOMAIN_VRAM) {
2498 2499 2500 2501 2502 2503
		r = amdgpu_bo_validate(bo->shadow);
		if (r) {
			DRM_ERROR("bo validate failed!\n");
			goto err;
		}

R
Roger.He 已提交
2504
		r = amdgpu_bo_restore_from_shadow(adev, ring, bo,
2505
						 NULL, fence, true);
R
Roger.He 已提交
2506 2507 2508 2509 2510
		if (r) {
			DRM_ERROR("recover page table failed!\n");
			goto err;
		}
	}
2511
err:
R
Roger.He 已提交
2512 2513
	amdgpu_bo_unreserve(bo);
	return r;
2514 2515
}

2516
/*
2517
 * amdgpu_device_reset - reset ASIC/GPU for bare-metal or passthrough
2518 2519
 *
 * @adev: amdgpu device pointer
2520
 * @reset_flags: output param tells caller the reset result
2521
 *
2522 2523 2524
 * attempt to do soft-reset or full-reset and reinitialize Asic
 * return 0 means successed otherwise failed
*/
2525 2526
static int amdgpu_device_reset(struct amdgpu_device *adev,
			       uint64_t* reset_flags)
2527
{
2528 2529
	bool need_full_reset, vram_lost = 0;
	int r;
2530

2531
	need_full_reset = amdgpu_device_ip_need_full_reset(adev);
2532

2533
	if (!need_full_reset) {
2534 2535 2536 2537
		amdgpu_device_ip_pre_soft_reset(adev);
		r = amdgpu_device_ip_soft_reset(adev);
		amdgpu_device_ip_post_soft_reset(adev);
		if (r || amdgpu_device_ip_check_soft_reset(adev)) {
2538 2539 2540
			DRM_INFO("soft reset failed, will fallback to full reset!\n");
			need_full_reset = true;
		}
2541

2542
	}
2543

2544
	if (need_full_reset) {
2545
		r = amdgpu_device_ip_suspend(adev);
2546

2547 2548 2549 2550
retry:
		r = amdgpu_asic_reset(adev);
		/* post card */
		amdgpu_atom_asic_init(adev->mode_info.atom_context);
2551

2552 2553
		if (!r) {
			dev_info(adev->dev, "GPU reset succeeded, trying to resume\n");
2554
			r = amdgpu_device_ip_resume_phase1(adev);
2555 2556
			if (r)
				goto out;
2557

2558
			vram_lost = amdgpu_device_check_vram_lost(adev);
2559 2560 2561 2562 2563
			if (vram_lost) {
				DRM_ERROR("VRAM is lost!\n");
				atomic_inc(&adev->vram_lost_counter);
			}

2564 2565
			r = amdgpu_gtt_mgr_recover(
				&adev->mman.bdev.man[TTM_PL_TT]);
2566 2567 2568
			if (r)
				goto out;

2569
			r = amdgpu_device_ip_resume_phase2(adev);
2570 2571 2572 2573
			if (r)
				goto out;

			if (vram_lost)
2574
				amdgpu_device_fill_reset_magic(adev);
2575
		}
2576
	}
2577

2578 2579 2580 2581 2582 2583
out:
	if (!r) {
		amdgpu_irq_gpu_reset_resume_helper(adev);
		r = amdgpu_ib_ring_tests(adev);
		if (r) {
			dev_err(adev->dev, "ib ring test failed (%d).\n", r);
2584
			r = amdgpu_device_ip_suspend(adev);
2585 2586 2587 2588
			need_full_reset = true;
			goto retry;
		}
	}
2589

2590 2591 2592
	if (reset_flags) {
		if (vram_lost)
			(*reset_flags) |= AMDGPU_RESET_INFO_VRAM_LOST;
2593

2594 2595
		if (need_full_reset)
			(*reset_flags) |= AMDGPU_RESET_INFO_FULLRESET;
2596
	}
2597

2598 2599
	return r;
}
2600

2601
/*
2602
 * amdgpu_device_reset_sriov - reset ASIC for SR-IOV vf
2603 2604 2605 2606 2607 2608 2609
 *
 * @adev: amdgpu device pointer
 * @reset_flags: output param tells caller the reset result
 *
 * do VF FLR and reinitialize Asic
 * return 0 means successed otherwise failed
*/
2610 2611 2612
static int amdgpu_device_reset_sriov(struct amdgpu_device *adev,
				     uint64_t *reset_flags,
				     bool from_hypervisor)
2613 2614 2615 2616 2617 2618 2619 2620 2621
{
	int r;

	if (from_hypervisor)
		r = amdgpu_virt_request_full_gpu(adev, true);
	else
		r = amdgpu_virt_reset_gpu(adev);
	if (r)
		return r;
2622 2623

	/* Resume IP prior to SMC */
2624
	r = amdgpu_device_ip_reinit_early_sriov(adev);
2625 2626
	if (r)
		goto error;
2627 2628

	/* we need recover gart prior to run SMC/CP/SDMA resume */
2629
	amdgpu_gtt_mgr_recover(&adev->mman.bdev.man[TTM_PL_TT]);
2630 2631

	/* now we are okay to resume SMC/CP/SDMA */
2632
	r = amdgpu_device_ip_reinit_late_sriov(adev);
2633 2634
	if (r)
		goto error;
2635 2636

	amdgpu_irq_gpu_reset_resume_helper(adev);
2637 2638
	r = amdgpu_ib_ring_tests(adev);
	if (r)
2639 2640
		dev_err(adev->dev, "[GPU_RESET] ib ring test failed (%d).\n", r);

2641
error:
2642 2643 2644
	/* release full control of GPU after ib test */
	amdgpu_virt_release_full_gpu(adev, true);

2645
	if (reset_flags) {
M
Monk Liu 已提交
2646 2647 2648 2649
		if (adev->virt.gim_feature & AMDGIM_FEATURE_GIM_FLR_VRAMLOST) {
			(*reset_flags) |= AMDGPU_RESET_INFO_VRAM_LOST;
			atomic_inc(&adev->vram_lost_counter);
		}
2650

2651 2652
		/* VF FLR or hotlink reset is always full-reset */
		(*reset_flags) |= AMDGPU_RESET_INFO_FULLRESET;
2653 2654 2655 2656 2657
	}

	return r;
}

A
Alex Deucher 已提交
2658
/**
2659
 * amdgpu_gpu_recover - reset the asic and recover scheduler
A
Alex Deucher 已提交
2660 2661
 *
 * @adev: amdgpu device pointer
2662
 * @job: which job trigger hang
2663
 * @force forces reset regardless of amdgpu_gpu_recovery
A
Alex Deucher 已提交
2664
 *
2665
 * Attempt to reset the GPU if it has hung (all asics).
A
Alex Deucher 已提交
2666 2667
 * Returns 0 for success or an error on failure.
 */
2668
int amdgpu_gpu_recover(struct amdgpu_device *adev, struct amdgpu_job *job, bool force)
A
Alex Deucher 已提交
2669
{
2670
	struct drm_atomic_state *state = NULL;
2671 2672
	uint64_t reset_flags = 0;
	int i, r, resched;
2673

2674
	if (!amdgpu_device_ip_check_soft_reset(adev)) {
2675 2676 2677
		DRM_INFO("No hardware hang detected. Did some blocks stall?\n");
		return 0;
	}
A
Alex Deucher 已提交
2678

2679 2680 2681 2682 2683 2684
	if (!force && (amdgpu_gpu_recovery == 0 ||
			(amdgpu_gpu_recovery == -1  && !amdgpu_sriov_vf(adev)))) {
		DRM_INFO("GPU recovery disabled.\n");
		return 0;
	}

2685 2686
	dev_info(adev->dev, "GPU reset begin!\n");

2687
	mutex_lock(&adev->lock_reset);
2688
	atomic_inc(&adev->gpu_reset_counter);
2689
	adev->in_gpu_reset = 1;
A
Alex Deucher 已提交
2690

2691 2692
	/* block TTM */
	resched = ttm_bo_lock_delayed_workqueue(&adev->mman.bdev);
2693 2694 2695
	/* store modesetting */
	if (amdgpu_device_has_dc_support(adev))
		state = drm_atomic_helper_suspend(adev->ddev);
2696

2697 2698 2699 2700
	/* block scheduler */
	for (i = 0; i < AMDGPU_MAX_RINGS; ++i) {
		struct amdgpu_ring *ring = adev->rings[i];

C
Chunming Zhou 已提交
2701
		if (!ring || !ring->sched.thread)
2702
			continue;
2703 2704 2705 2706 2707

		/* only focus on the ring hit timeout if &job not NULL */
		if (job && job->ring->idx != i)
			continue;

2708
		kthread_park(ring->sched.thread);
2709
		drm_sched_hw_job_reset(&ring->sched, &job->base);
2710

M
Monk Liu 已提交
2711 2712
		/* after all hw jobs are reset, hw fence is meaningless, so force_completion */
		amdgpu_fence_driver_force_completion(ring);
2713
	}
A
Alex Deucher 已提交
2714

2715
	if (amdgpu_sriov_vf(adev))
2716
		r = amdgpu_device_reset_sriov(adev, &reset_flags, job ? false : true);
2717
	else
2718
		r = amdgpu_device_reset(adev, &reset_flags);
2719

A
Alex Deucher 已提交
2720
	if (!r) {
2721 2722
		if (((reset_flags & AMDGPU_RESET_INFO_FULLRESET) && !(adev->flags & AMD_IS_APU)) ||
			(reset_flags & AMDGPU_RESET_INFO_VRAM_LOST)) {
2723 2724
			struct amdgpu_ring *ring = adev->mman.buffer_funcs_ring;
			struct amdgpu_bo *bo, *tmp;
2725
			struct dma_fence *fence = NULL, *next = NULL;
2726 2727 2728 2729

			DRM_INFO("recover vram bo from shadow\n");
			mutex_lock(&adev->shadow_list_lock);
			list_for_each_entry_safe(bo, tmp, &adev->shadow_list, shadow_list) {
M
Monk Liu 已提交
2730
				next = NULL;
2731
				amdgpu_device_recover_vram_from_shadow(adev, ring, bo, &next);
2732
				if (fence) {
2733
					r = dma_fence_wait(fence, false);
2734
					if (r) {
M
Monk Liu 已提交
2735
						WARN(r, "recovery from shadow isn't completed\n");
2736 2737 2738
						break;
					}
				}
2739

2740
				dma_fence_put(fence);
2741 2742 2743 2744
				fence = next;
			}
			mutex_unlock(&adev->shadow_list_lock);
			if (fence) {
2745
				r = dma_fence_wait(fence, false);
2746
				if (r)
M
Monk Liu 已提交
2747
					WARN(r, "recovery from shadow isn't completed\n");
2748
			}
2749
			dma_fence_put(fence);
2750
		}
2751

A
Alex Deucher 已提交
2752 2753
		for (i = 0; i < AMDGPU_MAX_RINGS; ++i) {
			struct amdgpu_ring *ring = adev->rings[i];
C
Chunming Zhou 已提交
2754 2755

			if (!ring || !ring->sched.thread)
A
Alex Deucher 已提交
2756
				continue;
2757

2758 2759 2760 2761
			/* only focus on the ring hit timeout if &job not NULL */
			if (job && job->ring->idx != i)
				continue;

2762
			drm_sched_job_recovery(&ring->sched);
2763
			kthread_unpark(ring->sched.thread);
A
Alex Deucher 已提交
2764 2765 2766
		}
	} else {
		for (i = 0; i < AMDGPU_MAX_RINGS; ++i) {
2767 2768 2769 2770 2771 2772 2773 2774 2775 2776
			struct amdgpu_ring *ring = adev->rings[i];

			if (!ring || !ring->sched.thread)
				continue;

			/* only focus on the ring hit timeout if &job not NULL */
			if (job && job->ring->idx != i)
				continue;

			kthread_unpark(adev->rings[i]->sched.thread);
A
Alex Deucher 已提交
2777 2778 2779
		}
	}

2780
	if (amdgpu_device_has_dc_support(adev)) {
2781 2782
		if (drm_atomic_helper_resume(adev->ddev, state))
			dev_info(adev->dev, "drm resume failed:%d\n", r);
2783
		amdgpu_dm_display_resume(adev);
2784
	} else {
2785
		drm_helper_resume_force_mode(adev->ddev);
2786
	}
A
Alex Deucher 已提交
2787 2788

	ttm_bo_unlock_delayed_workqueue(&adev->mman.bdev, resched);
2789

2790
	if (r) {
A
Alex Deucher 已提交
2791
		/* bad news, how to tell it to userspace ? */
2792 2793 2794 2795
		dev_info(adev->dev, "GPU reset(%d) failed\n", atomic_read(&adev->gpu_reset_counter));
		amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_GPU_RESET_FAIL, 0, r);
	} else {
		dev_info(adev->dev, "GPU reset(%d) successed!\n",atomic_read(&adev->gpu_reset_counter));
2796
	}
A
Alex Deucher 已提交
2797

2798
	amdgpu_vf_error_trans_all(adev);
2799 2800
	adev->in_gpu_reset = 0;
	mutex_unlock(&adev->lock_reset);
A
Alex Deucher 已提交
2801 2802 2803
	return r;
}

2804 2805 2806 2807 2808
void amdgpu_get_pcie_info(struct amdgpu_device *adev)
{
	u32 mask;
	int ret;

2809 2810
	if (amdgpu_pcie_gen_cap)
		adev->pm.pcie_gen_mask = amdgpu_pcie_gen_cap;
2811

2812 2813
	if (amdgpu_pcie_lane_cap)
		adev->pm.pcie_mlw_mask = amdgpu_pcie_lane_cap;
2814

2815 2816 2817 2818 2819 2820
	/* covers APUs as well */
	if (pci_is_root_bus(adev->pdev->bus)) {
		if (adev->pm.pcie_gen_mask == 0)
			adev->pm.pcie_gen_mask = AMDGPU_DEFAULT_PCIE_GEN_MASK;
		if (adev->pm.pcie_mlw_mask == 0)
			adev->pm.pcie_mlw_mask = AMDGPU_DEFAULT_PCIE_MLW_MASK;
2821
		return;
2822
	}
2823

2824 2825 2826 2827 2828 2829 2830 2831 2832 2833 2834 2835 2836 2837 2838 2839 2840 2841 2842 2843 2844 2845 2846 2847 2848 2849 2850 2851 2852 2853 2854 2855 2856 2857 2858 2859 2860 2861 2862 2863 2864 2865 2866 2867 2868 2869 2870 2871 2872 2873 2874 2875 2876 2877 2878 2879 2880 2881 2882 2883 2884 2885 2886 2887 2888 2889 2890 2891
	if (adev->pm.pcie_gen_mask == 0) {
		ret = drm_pcie_get_speed_cap_mask(adev->ddev, &mask);
		if (!ret) {
			adev->pm.pcie_gen_mask = (CAIL_ASIC_PCIE_LINK_SPEED_SUPPORT_GEN1 |
						  CAIL_ASIC_PCIE_LINK_SPEED_SUPPORT_GEN2 |
						  CAIL_ASIC_PCIE_LINK_SPEED_SUPPORT_GEN3);

			if (mask & DRM_PCIE_SPEED_25)
				adev->pm.pcie_gen_mask |= CAIL_PCIE_LINK_SPEED_SUPPORT_GEN1;
			if (mask & DRM_PCIE_SPEED_50)
				adev->pm.pcie_gen_mask |= CAIL_PCIE_LINK_SPEED_SUPPORT_GEN2;
			if (mask & DRM_PCIE_SPEED_80)
				adev->pm.pcie_gen_mask |= CAIL_PCIE_LINK_SPEED_SUPPORT_GEN3;
		} else {
			adev->pm.pcie_gen_mask = AMDGPU_DEFAULT_PCIE_GEN_MASK;
		}
	}
	if (adev->pm.pcie_mlw_mask == 0) {
		ret = drm_pcie_get_max_link_width(adev->ddev, &mask);
		if (!ret) {
			switch (mask) {
			case 32:
				adev->pm.pcie_mlw_mask = (CAIL_PCIE_LINK_WIDTH_SUPPORT_X32 |
							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X16 |
							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X12 |
							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X8 |
							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X4 |
							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X2 |
							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X1);
				break;
			case 16:
				adev->pm.pcie_mlw_mask = (CAIL_PCIE_LINK_WIDTH_SUPPORT_X16 |
							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X12 |
							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X8 |
							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X4 |
							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X2 |
							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X1);
				break;
			case 12:
				adev->pm.pcie_mlw_mask = (CAIL_PCIE_LINK_WIDTH_SUPPORT_X12 |
							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X8 |
							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X4 |
							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X2 |
							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X1);
				break;
			case 8:
				adev->pm.pcie_mlw_mask = (CAIL_PCIE_LINK_WIDTH_SUPPORT_X8 |
							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X4 |
							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X2 |
							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X1);
				break;
			case 4:
				adev->pm.pcie_mlw_mask = (CAIL_PCIE_LINK_WIDTH_SUPPORT_X4 |
							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X2 |
							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X1);
				break;
			case 2:
				adev->pm.pcie_mlw_mask = (CAIL_PCIE_LINK_WIDTH_SUPPORT_X2 |
							  CAIL_PCIE_LINK_WIDTH_SUPPORT_X1);
				break;
			case 1:
				adev->pm.pcie_mlw_mask = CAIL_PCIE_LINK_WIDTH_SUPPORT_X1;
				break;
			default:
				break;
			}
		} else {
			adev->pm.pcie_mlw_mask = AMDGPU_DEFAULT_PCIE_MLW_MASK;
2892 2893 2894
		}
	}
}
A
Alex Deucher 已提交
2895