amdgpu_device.c 74.7 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
	"LAST",
};

90 91
static void amdgpu_device_get_pcie_info(struct amdgpu_device *adev);

A
Alex Deucher 已提交
92 93 94 95
bool amdgpu_device_is_px(struct drm_device *dev)
{
	struct amdgpu_device *adev = dev->dev_private;

96
	if (adev->flags & AMD_IS_PX)
A
Alex Deucher 已提交
97 98 99 100 101 102 103 104
		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
105
			uint32_t acc_flags)
A
Alex Deucher 已提交
106
{
107 108
	uint32_t ret;

109
	if (!(acc_flags & AMDGPU_REGS_NO_KIQ) && amdgpu_sriov_runtime(adev))
110 111
		return amdgpu_virt_kiq_rreg(adev, reg);

M
Monk Liu 已提交
112
	if ((reg * 4) < adev->rmmio_size && !(acc_flags & AMDGPU_REGS_IDX))
113
		ret = readl(((void __iomem *)adev->rmmio) + (reg * 4));
A
Alex Deucher 已提交
114 115 116 117 118 119 120 121
	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);
	}
122 123
	trace_amdgpu_mm_rreg(adev->pdev->device, reg, ret);
	return ret;
A
Alex Deucher 已提交
124 125
}

126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151
/*
 * MMIO register read with bytes helper functions
 * @offset:bytes offset from MMIO start
 *
*/

uint8_t amdgpu_mm_rreg8(struct amdgpu_device *adev, uint32_t offset) {
	if (offset < adev->rmmio_size)
		return (readb(adev->rmmio + offset));
	BUG();
}

/*
 * MMIO register write with bytes helper functions
 * @offset:bytes offset from MMIO start
 * @value: the value want to be written to the register
 *
*/
void amdgpu_mm_wreg8(struct amdgpu_device *adev, uint32_t offset, uint8_t value) {
	if (offset < adev->rmmio_size)
		writeb(value, adev->rmmio + offset);
	else
		BUG();
}


A
Alex Deucher 已提交
152
void amdgpu_mm_wreg(struct amdgpu_device *adev, uint32_t reg, uint32_t v,
M
Monk Liu 已提交
153
		    uint32_t acc_flags)
A
Alex Deucher 已提交
154
{
155
	trace_amdgpu_mm_wreg(adev->pdev->device, reg, v);
156

157 158 159 160
	if (adev->asic_type >= CHIP_VEGA10 && reg == 0) {
		adev->last_mm_index = v;
	}

161
	if (!(acc_flags & AMDGPU_REGS_NO_KIQ) && amdgpu_sriov_runtime(adev))
162 163
		return amdgpu_virt_kiq_wreg(adev, reg, v);

M
Monk Liu 已提交
164
	if ((reg * 4) < adev->rmmio_size && !(acc_flags & AMDGPU_REGS_IDX))
A
Alex Deucher 已提交
165 166 167 168 169 170 171 172 173
		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);
	}
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
}

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)
{
192 193 194
	if (adev->asic_type >= CHIP_VEGA10 && reg == 0) {
		adev->last_mm_index = v;
	}
A
Alex Deucher 已提交
195 196 197 198 199 200 201

	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));
	}
202 203 204 205

	if (adev->asic_type >= CHIP_VEGA10 && reg == 1 && adev->last_mm_index == 0x5702C) {
		udelay(500);
	}
A
Alex Deucher 已提交
206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245
}

/**
 * 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);
	}
}

246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283
/**
 * 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
284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357
/**
 * 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();
}

358
static int amdgpu_device_vram_scratch_init(struct amdgpu_device *adev)
A
Alex Deucher 已提交
359
{
360 361 362 363 364
	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
365 366
}

367
static void amdgpu_device_vram_scratch_fini(struct amdgpu_device *adev)
A
Alex Deucher 已提交
368
{
369
	amdgpu_bo_free_kernel(&adev->vram_scratch.robj, NULL, NULL);
A
Alex Deucher 已提交
370 371 372
}

/**
373
 * amdgpu_device_program_register_sequence - program an array of registers.
A
Alex Deucher 已提交
374 375 376 377 378 379 380 381
 *
 * @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.
 */
382 383 384
void amdgpu_device_program_register_sequence(struct amdgpu_device *adev,
					     const u32 *registers,
					     const u32 array_size)
A
Alex Deucher 已提交
385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407
{
	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);
	}
}

408
void amdgpu_device_pci_config_reset(struct amdgpu_device *adev)
A
Alex Deucher 已提交
409 410 411 412 413 414 415 416
{
	pci_write_config_dword(adev->pdev, 0x7c, AMDGPU_ASIC_RESET_DATA);
}

/*
 * GPU doorbell aperture helpers function.
 */
/**
417
 * amdgpu_device_doorbell_init - Init doorbell driver information.
A
Alex Deucher 已提交
418 419 420 421 422 423
 *
 * @adev: amdgpu_device pointer
 *
 * Init doorbell driver information (CIK)
 * Returns 0 on success, error on failure.
 */
424
static int amdgpu_device_doorbell_init(struct amdgpu_device *adev)
A
Alex Deucher 已提交
425
{
426 427 428 429 430 431 432 433 434
	/* 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;
	}

435 436 437
	if (pci_resource_flags(adev->pdev, 2) & IORESOURCE_UNSET)
		return -EINVAL;

A
Alex Deucher 已提交
438 439 440 441
	/* doorbell bar mapping */
	adev->doorbell.base = pci_resource_start(adev->pdev, 2);
	adev->doorbell.size = pci_resource_len(adev->pdev, 2);

442
	adev->doorbell.num_doorbells = min_t(u32, adev->doorbell.size / sizeof(u32),
A
Alex Deucher 已提交
443 444 445 446
					     AMDGPU_DOORBELL_MAX_ASSIGNMENT+1);
	if (adev->doorbell.num_doorbells == 0)
		return -EINVAL;

447 448 449 450
	adev->doorbell.ptr = ioremap(adev->doorbell.base,
				     adev->doorbell.num_doorbells *
				     sizeof(u32));
	if (adev->doorbell.ptr == NULL)
A
Alex Deucher 已提交
451 452 453 454 455 456
		return -ENOMEM;

	return 0;
}

/**
457
 * amdgpu_device_doorbell_fini - Tear down doorbell driver information.
A
Alex Deucher 已提交
458 459 460 461 462
 *
 * @adev: amdgpu_device pointer
 *
 * Tear down doorbell driver information (CIK)
 */
463
static void amdgpu_device_doorbell_fini(struct amdgpu_device *adev)
A
Alex Deucher 已提交
464 465 466 467 468
{
	iounmap(adev->doorbell.ptr);
	adev->doorbell.ptr = NULL;
}

469

A
Alex Deucher 已提交
470 471

/*
472
 * amdgpu_device_wb_*()
473
 * Writeback is the method by which the GPU updates special pages in memory
A
Alex Xie 已提交
474
 * with the status of certain GPU events (fences, ring pointers,etc.).
A
Alex Deucher 已提交
475 476 477
 */

/**
478
 * amdgpu_device_wb_fini - Disable Writeback and free memory
A
Alex Deucher 已提交
479 480 481 482 483 484
 *
 * @adev: amdgpu_device pointer
 *
 * Disables Writeback and frees the Writeback memory (all asics).
 * Used at driver shutdown.
 */
485
static void amdgpu_device_wb_fini(struct amdgpu_device *adev)
A
Alex Deucher 已提交
486 487
{
	if (adev->wb.wb_obj) {
488 489 490
		amdgpu_bo_free_kernel(&adev->wb.wb_obj,
				      &adev->wb.gpu_addr,
				      (void **)&adev->wb.wb);
A
Alex Deucher 已提交
491 492 493 494 495
		adev->wb.wb_obj = NULL;
	}
}

/**
496
 * amdgpu_device_wb_init- Init Writeback driver info and allocate memory
A
Alex Deucher 已提交
497 498 499
 *
 * @adev: amdgpu_device pointer
 *
500
 * Initializes writeback and allocates writeback memory (all asics).
A
Alex Deucher 已提交
501 502 503
 * Used at driver startup.
 * Returns 0 on success or an -error on failure.
 */
504
static int amdgpu_device_wb_init(struct amdgpu_device *adev)
A
Alex Deucher 已提交
505 506 507 508
{
	int r;

	if (adev->wb.wb_obj == NULL) {
509 510
		/* 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,
511 512 513
					    PAGE_SIZE, AMDGPU_GEM_DOMAIN_GTT,
					    &adev->wb.wb_obj, &adev->wb.gpu_addr,
					    (void **)&adev->wb.wb);
A
Alex Deucher 已提交
514 515 516 517 518 519 520 521 522
		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 */
M
Monk Liu 已提交
523
		memset((char *)adev->wb.wb, 0, AMDGPU_MAX_WB * sizeof(uint32_t) * 8);
A
Alex Deucher 已提交
524 525 526 527 528 529
	}

	return 0;
}

/**
530
 * amdgpu_device_wb_get - Allocate a wb entry
A
Alex Deucher 已提交
531 532 533 534 535 536 537
 *
 * @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.
 */
538
int amdgpu_device_wb_get(struct amdgpu_device *adev, u32 *wb)
A
Alex Deucher 已提交
539 540 541
{
	unsigned long offset = find_first_zero_bit(adev->wb.used, adev->wb.num_wb);

542
	if (offset < adev->wb.num_wb) {
K
Ken Wang 已提交
543
		__set_bit(offset, adev->wb.used);
M
Monk Liu 已提交
544
		*wb = offset << 3; /* convert to dw offset */
545 546 547 548 549 550
		return 0;
	} else {
		return -EINVAL;
	}
}

A
Alex Deucher 已提交
551
/**
552
 * amdgpu_device_wb_free - Free a wb entry
A
Alex Deucher 已提交
553 554 555 556 557 558
 *
 * @adev: amdgpu_device pointer
 * @wb: wb index
 *
 * Free a wb slot allocated for use by the driver (all asics)
 */
559
void amdgpu_device_wb_free(struct amdgpu_device *adev, u32 wb)
A
Alex Deucher 已提交
560
{
M
Monk Liu 已提交
561
	wb >>= 3;
A
Alex Deucher 已提交
562
	if (wb < adev->wb.num_wb)
M
Monk Liu 已提交
563
		__clear_bit(wb, adev->wb.used);
A
Alex Deucher 已提交
564 565 566
}

/**
567
 * amdgpu_device_vram_location - try to find VRAM location
A
Alex Deucher 已提交
568 569 570 571
 * @adev: amdgpu device structure holding all necessary informations
 * @mc: memory controller structure holding memory informations
 * @base: base address at which to put VRAM
 *
572
 * Function will try to place VRAM at base address provided
573
 * as parameter.
A
Alex Deucher 已提交
574
 */
575
void amdgpu_device_vram_location(struct amdgpu_device *adev,
576
				 struct amdgpu_gmc *mc, u64 base)
A
Alex Deucher 已提交
577 578 579 580 581 582 583 584 585 586 587 588 589
{
	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);
}

/**
590
 * amdgpu_device_gart_location - try to find GTT location
A
Alex Deucher 已提交
591 592 593 594 595 596 597 598 599 600
 * @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.
 */
601
void amdgpu_device_gart_location(struct amdgpu_device *adev,
602
				 struct amdgpu_gmc *mc)
A
Alex Deucher 已提交
603 604 605
{
	u64 size_af, size_bf;

606
	size_af = adev->gmc.mc_mask - mc->vram_end;
607
	size_bf = mc->vram_start;
A
Alex Deucher 已提交
608
	if (size_bf > size_af) {
609
		if (mc->gart_size > size_bf) {
A
Alex Deucher 已提交
610
			dev_warn(adev->dev, "limiting GTT\n");
611
			mc->gart_size = size_bf;
A
Alex Deucher 已提交
612
		}
613
		mc->gart_start = 0;
A
Alex Deucher 已提交
614
	} else {
615
		if (mc->gart_size > size_af) {
A
Alex Deucher 已提交
616
			dev_warn(adev->dev, "limiting GTT\n");
617
			mc->gart_size = size_af;
A
Alex Deucher 已提交
618
		}
619 620 621 622
		/* 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
623
	}
624
	mc->gart_end = mc->gart_start + mc->gart_size - 1;
A
Alex Deucher 已提交
625
	dev_info(adev->dev, "GTT: %lluM 0x%016llX - 0x%016llX\n",
626
			mc->gart_size >> 20, mc->gart_start, mc->gart_end);
A
Alex Deucher 已提交
627 628
}

629 630 631 632 633 634 635 636 637 638 639
/**
 * 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)
{
640
	u64 space_needed = roundup_pow_of_two(adev->gmc.real_vram_size);
641
	u32 rbar_size = order_base_2(((space_needed >> 20) | 1)) - 1;
642 643 644
	struct pci_bus *root;
	struct resource *res;
	unsigned i;
645 646 647
	u16 cmd;
	int r;

648 649 650 651
	/* Bypass for VF */
	if (amdgpu_sriov_vf(adev))
		return 0;

652 653 654 655 656 657
	/* 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) {
658
		if (res && res->flags & (IORESOURCE_MEM | IORESOURCE_MEM_64) &&
659 660 661 662 663 664 665 666
		    res->start > 0x100000000ull)
			break;
	}

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

667 668 669 670 671 672
	/* 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. */
673
	amdgpu_device_doorbell_fini(adev);
674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689
	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.
	 */
690
	r = amdgpu_device_doorbell_init(adev);
691 692 693 694 695 696 697
	if (r || (pci_resource_flags(adev->pdev, 0) & IORESOURCE_UNSET))
		return -ENODEV;

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

	return 0;
}
698

A
Alex Deucher 已提交
699 700 701 702
/*
 * GPU helpers function.
 */
/**
A
Alex Deucher 已提交
703
 * amdgpu_device_need_post - check if the hw need post or not
A
Alex Deucher 已提交
704 705 706
 *
 * @adev: amdgpu_device pointer
 *
707 708 709
 * 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
710
 */
A
Alex Deucher 已提交
711
bool amdgpu_device_need_post(struct amdgpu_device *adev)
A
Alex Deucher 已提交
712 713 714
{
	uint32_t reg;

715 716 717 718
	if (amdgpu_sriov_vf(adev))
		return false;

	if (amdgpu_passthrough(adev)) {
M
Monk Liu 已提交
719 720 721 722
		/* 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
723 724 725 726 727 728 729 730 731 732
		 */
		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
733 734
			if (fw_ver < 0x00160e00)
				return true;
735 736
		}
	}
737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753

	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;
754 755
}

A
Alex Deucher 已提交
756 757
/* if we get transitioned to only one device, take VGA back */
/**
758
 * amdgpu_device_vga_set_decode - enable/disable vga decode
A
Alex Deucher 已提交
759 760 761 762 763 764 765
 *
 * @cookie: amdgpu_device pointer
 * @state: enable/disable vga decode
 *
 * Enable/disable vga decode (all asics).
 * Returns VGA resource flags.
 */
766
static unsigned int amdgpu_device_vga_set_decode(void *cookie, bool state)
A
Alex Deucher 已提交
767 768 769 770 771 772 773 774 775 776
{
	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;
}

777
static void amdgpu_device_check_block_size(struct amdgpu_device *adev)
778 779 780 781
{
	/* 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 */
782 783
	if (amdgpu_vm_block_size == -1)
		return;
784

785
	if (amdgpu_vm_block_size < 9) {
786 787
		dev_warn(adev->dev, "VM page table size (%d) too small\n",
			 amdgpu_vm_block_size);
788
		amdgpu_vm_block_size = -1;
789 790 791
	}
}

792
static void amdgpu_device_check_vm_size(struct amdgpu_device *adev)
793
{
794 795 796 797
	/* no need to check the default value */
	if (amdgpu_vm_size == -1)
		return;

798 799 800
	if (amdgpu_vm_size < 1) {
		dev_warn(adev->dev, "VM size (%d) too small, min is 1GB\n",
			 amdgpu_vm_size);
801
		amdgpu_vm_size = -1;
802 803 804
	}
}

A
Alex Deucher 已提交
805
/**
806
 * amdgpu_device_check_arguments - validate module params
A
Alex Deucher 已提交
807 808 809 810 811 812
 *
 * @adev: amdgpu_device pointer
 *
 * Validates certain module parameters and updates
 * the associated values used by the driver (all asics).
 */
813
static void amdgpu_device_check_arguments(struct amdgpu_device *adev)
A
Alex Deucher 已提交
814
{
815 816 817 818
	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;
819
	} else if (!is_power_of_2(amdgpu_sched_jobs)){
820 821 822 823
		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
824

825
	if (amdgpu_gart_size != -1 && amdgpu_gart_size < 32) {
826 827 828
		/* gart size must be greater or equal to 32M */
		dev_warn(adev->dev, "gart size (%d) too small\n",
			 amdgpu_gart_size);
829
		amdgpu_gart_size = -1;
A
Alex Deucher 已提交
830 831
	}

832
	if (amdgpu_gtt_size != -1 && amdgpu_gtt_size < 32) {
833
		/* gtt size must be greater or equal to 32M */
834 835 836
		dev_warn(adev->dev, "gtt size (%d) too small\n",
				 amdgpu_gtt_size);
		amdgpu_gtt_size = -1;
A
Alex Deucher 已提交
837 838
	}

839 840 841 842 843 844 845
	/* 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;
	}

846
	amdgpu_device_check_vm_size(adev);
A
Alex Deucher 已提交
847

848
	amdgpu_device_check_block_size(adev);
C
Christian König 已提交
849

850
	if (amdgpu_vram_page_split != -1 && (amdgpu_vram_page_split < 16 ||
851
	    !is_power_of_2(amdgpu_vram_page_split))) {
C
Christian König 已提交
852 853 854 855
		dev_warn(adev->dev, "invalid VRAM page split (%d)\n",
			 amdgpu_vram_page_split);
		amdgpu_vram_page_split = 1024;
	}
856 857 858 859 860

	if (amdgpu_lockup_timeout == 0) {
		dev_warn(adev->dev, "lockup_timeout msut be > 0, adjusting to 10000\n");
		amdgpu_lockup_timeout = 10000;
	}
861 862

	adev->firmware.load_type = amdgpu_ucode_get_load_type(adev, amdgpu_fw_load_type);
A
Alex Deucher 已提交
863 864 865 866 867 868
}

/**
 * amdgpu_switcheroo_set_state - set switcheroo state
 *
 * @pdev: pci dev pointer
869
 * @state: vga_switcheroo state
A
Alex Deucher 已提交
870 871 872 873 874 875 876 877 878 879 880 881
 *
 * 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) {
882
		pr_info("amdgpu: switched on\n");
A
Alex Deucher 已提交
883 884 885
		/* don't suspend or resume card normally */
		dev->switch_power_state = DRM_SWITCH_POWER_CHANGING;

886
		amdgpu_device_resume(dev, true, true);
A
Alex Deucher 已提交
887 888 889 890

		dev->switch_power_state = DRM_SWITCH_POWER_ON;
		drm_kms_helper_poll_enable(dev);
	} else {
891
		pr_info("amdgpu: switched off\n");
A
Alex Deucher 已提交
892 893
		drm_kms_helper_poll_disable(dev);
		dev->switch_power_state = DRM_SWITCH_POWER_CHANGING;
894
		amdgpu_device_suspend(dev, true, true);
A
Alex Deucher 已提交
895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925
		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,
};

926 927 928
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
929 930 931 932
{
	int i, r = 0;

	for (i = 0; i < adev->num_ip_blocks; i++) {
933
		if (!adev->ip_blocks[i].status.valid)
934
			continue;
935 936 937 938 939 940 941 942 943
		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
944 945 946 947
	}
	return r;
}

948 949 950
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
951 952 953 954
{
	int i, r = 0;

	for (i = 0; i < adev->num_ip_blocks; i++) {
955
		if (!adev->ip_blocks[i].status.valid)
956
			continue;
957 958 959 960 961 962 963 964 965
		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
966 967 968 969
	}
	return r;
}

970 971
void amdgpu_device_ip_get_clockgating_state(struct amdgpu_device *adev,
					    u32 *flags)
972 973 974 975 976 977 978 979 980 981 982
{
	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);
	}
}

983 984
int amdgpu_device_ip_wait_for_idle(struct amdgpu_device *adev,
				   enum amd_ip_block_type block_type)
985 986 987 988
{
	int i, r;

	for (i = 0; i < adev->num_ip_blocks; i++) {
989
		if (!adev->ip_blocks[i].status.valid)
990
			continue;
991 992
		if (adev->ip_blocks[i].version->type == block_type) {
			r = adev->ip_blocks[i].version->funcs->wait_for_idle((void *)adev);
993 994 995 996 997 998 999 1000 1001
			if (r)
				return r;
			break;
		}
	}
	return 0;

}

1002 1003
bool amdgpu_device_ip_is_idle(struct amdgpu_device *adev,
			      enum amd_ip_block_type block_type)
1004 1005 1006 1007
{
	int i;

	for (i = 0; i < adev->num_ip_blocks; i++) {
1008
		if (!adev->ip_blocks[i].status.valid)
1009
			continue;
1010 1011
		if (adev->ip_blocks[i].version->type == block_type)
			return adev->ip_blocks[i].version->funcs->is_idle((void *)adev);
1012 1013 1014 1015 1016
	}
	return true;

}

1017 1018 1019
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
1020 1021 1022 1023
{
	int i;

	for (i = 0; i < adev->num_ip_blocks; i++)
1024
		if (adev->ip_blocks[i].version->type == type)
A
Alex Deucher 已提交
1025 1026 1027 1028 1029 1030
			return &adev->ip_blocks[i];

	return NULL;
}

/**
1031
 * amdgpu_device_ip_block_version_cmp
A
Alex Deucher 已提交
1032 1033
 *
 * @adev: amdgpu_device pointer
1034
 * @type: enum amd_ip_block_type
A
Alex Deucher 已提交
1035 1036 1037 1038 1039 1040
 * @major: major version
 * @minor: minor version
 *
 * return 0 if equal or greater
 * return 1 if smaller or the ip_block doesn't exist
 */
1041 1042 1043
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
1044
{
1045
	struct amdgpu_ip_block *ip_block = amdgpu_device_ip_get_ip_block(adev, type);
A
Alex Deucher 已提交
1046

1047 1048 1049
	if (ip_block && ((ip_block->version->major > major) ||
			((ip_block->version->major == major) &&
			(ip_block->version->minor >= minor))))
A
Alex Deucher 已提交
1050 1051 1052 1053 1054
		return 0;

	return 1;
}

1055
/**
1056
 * amdgpu_device_ip_block_add
1057 1058 1059 1060 1061 1062 1063
 *
 * @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.
 */
1064 1065
int amdgpu_device_ip_block_add(struct amdgpu_device *adev,
			       const struct amdgpu_ip_block_version *ip_block_version)
1066 1067 1068 1069
{
	if (!ip_block_version)
		return -EINVAL;

1070
	DRM_INFO("add ip block number %d <%s>\n", adev->num_ip_blocks,
1071 1072
		  ip_block_version->funcs->name);

1073 1074 1075 1076 1077
	adev->ip_blocks[adev->num_ip_blocks++].version = ip_block_version;

	return 0;
}

1078
static void amdgpu_device_enable_virtual_display(struct amdgpu_device *adev)
1079 1080 1081 1082 1083 1084
{
	adev->enable_virtual_display = false;

	if (amdgpu_virtual_display) {
		struct drm_device *ddev = adev->ddev;
		const char *pci_address_name = pci_name(ddev->pdev);
1085
		char *pciaddstr, *pciaddstr_tmp, *pciaddname_tmp, *pciaddname;
1086 1087 1088

		pciaddstr = kstrdup(amdgpu_virtual_display, GFP_KERNEL);
		pciaddstr_tmp = pciaddstr;
1089 1090
		while ((pciaddname_tmp = strsep(&pciaddstr_tmp, ";"))) {
			pciaddname = strsep(&pciaddname_tmp, ",");
1091 1092
			if (!strcmp("all", pciaddname)
			    || !strcmp(pci_address_name, pciaddname)) {
1093 1094 1095
				long num_crtc;
				int res = -1;

1096
				adev->enable_virtual_display = true;
1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110

				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;
				}
1111 1112 1113 1114
				break;
			}
		}

1115 1116 1117
		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);
1118 1119 1120 1121 1122

		kfree(pciaddstr);
	}
}

1123 1124 1125 1126 1127 1128 1129
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;

1130 1131
	adev->firmware.gpu_info_fw = NULL;

1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159
	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;
1160 1161 1162
	case CHIP_RAVEN:
		chip_name = "raven";
		break;
1163 1164 1165
	}

	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_gpu_info.bin", chip_name);
1166
	err = request_firmware(&adev->firmware.gpu_info_fw, fw_name, adev->dev);
1167 1168 1169 1170 1171 1172
	if (err) {
		dev_err(adev->dev,
			"Failed to load gpu_info firmware \"%s\"\n",
			fw_name);
		goto out;
	}
1173
	err = amdgpu_ucode_validate(adev->firmware.gpu_info_fw);
1174 1175 1176 1177 1178 1179 1180
	if (err) {
		dev_err(adev->dev,
			"Failed to validate gpu_info firmware \"%s\"\n",
			fw_name);
		goto out;
	}

1181
	hdr = (const struct gpu_info_firmware_header_v1_0 *)adev->firmware.gpu_info_fw->data;
1182 1183 1184 1185 1186 1187
	amdgpu_ucode_print_gpu_info_hdr(&hdr->header);

	switch (hdr->version_major) {
	case 1:
	{
		const struct gpu_info_firmware_v1_0 *gpu_info_fw =
1188
			(const struct gpu_info_firmware_v1_0 *)(adev->firmware.gpu_info_fw->data +
1189 1190
								le32_to_cpu(hdr->header.ucode_array_offset_bytes));

1191 1192 1193 1194
		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);
1195
		adev->gfx.config.max_texture_channel_caches =
1196 1197 1198 1199 1200
			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);
1201
		adev->gfx.config.double_offchip_lds_buf =
1202 1203
			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);
1204 1205 1206 1207 1208
		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);
1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220
		break;
	}
	default:
		dev_err(adev->dev,
			"Unsupported gpu_info table %d\n", hdr->header.ucode_version);
		err = -EINVAL;
		goto out;
	}
out:
	return err;
}

1221
static int amdgpu_device_ip_early_init(struct amdgpu_device *adev)
A
Alex Deucher 已提交
1222
{
1223
	int i, r;
A
Alex Deucher 已提交
1224

1225
	amdgpu_device_enable_virtual_display(adev);
1226

A
Alex Deucher 已提交
1227
	switch (adev->asic_type) {
1228 1229
	case CHIP_TOPAZ:
	case CHIP_TONGA:
1230
	case CHIP_FIJI:
1231 1232
	case CHIP_POLARIS11:
	case CHIP_POLARIS10:
1233
	case CHIP_POLARIS12:
1234
	case CHIP_CARRIZO:
1235 1236
	case CHIP_STONEY:
		if (adev->asic_type == CHIP_CARRIZO || adev->asic_type == CHIP_STONEY)
1237 1238 1239 1240 1241 1242 1243 1244
			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
1245 1246 1247 1248 1249 1250
#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
1251
		adev->family = AMDGPU_FAMILY_SI;
K
Ken Wang 已提交
1252 1253 1254 1255 1256
		r = si_set_ip_blocks(adev);
		if (r)
			return r;
		break;
#endif
1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272
#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
1273 1274 1275 1276 1277 1278
	case  CHIP_VEGA10:
	case  CHIP_RAVEN:
		if (adev->asic_type == CHIP_RAVEN)
			adev->family = AMDGPU_FAMILY_RV;
		else
			adev->family = AMDGPU_FAMILY_AI;
1279 1280 1281 1282 1283

		r = soc15_set_ip_blocks(adev);
		if (r)
			return r;
		break;
A
Alex Deucher 已提交
1284 1285 1286 1287 1288
	default:
		/* FIXME: not supported yet */
		return -EINVAL;
	}

1289 1290 1291 1292
	r = amdgpu_device_parse_gpu_info_fw(adev);
	if (r)
		return r;

1293 1294
	amdgpu_amdkfd_device_probe(adev);

1295 1296 1297
	if (amdgpu_sriov_vf(adev)) {
		r = amdgpu_virt_request_full_gpu(adev, true);
		if (r)
1298
			return -EAGAIN;
1299 1300
	}

A
Alex Deucher 已提交
1301 1302
	for (i = 0; i < adev->num_ip_blocks; i++) {
		if ((amdgpu_ip_block_mask & (1 << i)) == 0) {
1303 1304
			DRM_ERROR("disabled ip block: %d <%s>\n",
				  i, adev->ip_blocks[i].version->funcs->name);
1305
			adev->ip_blocks[i].status.valid = false;
A
Alex Deucher 已提交
1306
		} else {
1307 1308
			if (adev->ip_blocks[i].version->funcs->early_init) {
				r = adev->ip_blocks[i].version->funcs->early_init((void *)adev);
1309
				if (r == -ENOENT) {
1310
					adev->ip_blocks[i].status.valid = false;
1311
				} else if (r) {
1312 1313
					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
1314
					return r;
1315
				} else {
1316
					adev->ip_blocks[i].status.valid = true;
1317
				}
1318
			} else {
1319
				adev->ip_blocks[i].status.valid = true;
A
Alex Deucher 已提交
1320 1321 1322 1323
			}
		}
	}

1324 1325 1326
	adev->cg_flags &= amdgpu_cg_mask;
	adev->pg_flags &= amdgpu_pg_mask;

A
Alex Deucher 已提交
1327 1328 1329
	return 0;
}

1330
static int amdgpu_device_ip_init(struct amdgpu_device *adev)
A
Alex Deucher 已提交
1331 1332 1333 1334
{
	int i, r;

	for (i = 0; i < adev->num_ip_blocks; i++) {
1335
		if (!adev->ip_blocks[i].status.valid)
A
Alex Deucher 已提交
1336
			continue;
1337
		r = adev->ip_blocks[i].version->funcs->sw_init((void *)adev);
1338
		if (r) {
1339 1340
			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
1341
			return r;
1342
		}
1343
		adev->ip_blocks[i].status.sw = true;
1344

A
Alex Deucher 已提交
1345
		/* need to do gmc hw init early so we can allocate gpu mem */
1346
		if (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_GMC) {
1347
			r = amdgpu_device_vram_scratch_init(adev);
1348 1349
			if (r) {
				DRM_ERROR("amdgpu_vram_scratch_init failed %d\n", r);
A
Alex Deucher 已提交
1350
				return r;
1351
			}
1352
			r = adev->ip_blocks[i].version->funcs->hw_init((void *)adev);
1353 1354
			if (r) {
				DRM_ERROR("hw_init %d failed %d\n", i, r);
A
Alex Deucher 已提交
1355
				return r;
1356
			}
1357
			r = amdgpu_device_wb_init(adev);
1358
			if (r) {
1359
				DRM_ERROR("amdgpu_device_wb_init failed %d\n", r);
A
Alex Deucher 已提交
1360
				return r;
1361
			}
1362
			adev->ip_blocks[i].status.hw = true;
M
Monk Liu 已提交
1363 1364 1365 1366 1367 1368 1369 1370 1371

			/* 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
1372 1373 1374 1375
		}
	}

	for (i = 0; i < adev->num_ip_blocks; i++) {
1376
		if (!adev->ip_blocks[i].status.sw)
A
Alex Deucher 已提交
1377
			continue;
1378
		if (adev->ip_blocks[i].status.hw)
A
Alex Deucher 已提交
1379
			continue;
1380
		r = adev->ip_blocks[i].version->funcs->hw_init((void *)adev);
1381
		if (r) {
1382 1383
			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
1384
			return r;
1385
		}
1386
		adev->ip_blocks[i].status.hw = true;
A
Alex Deucher 已提交
1387 1388
	}

1389
	amdgpu_amdkfd_device_init(adev);
1390 1391 1392 1393

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

A
Alex Deucher 已提交
1394 1395 1396
	return 0;
}

1397
static void amdgpu_device_fill_reset_magic(struct amdgpu_device *adev)
1398 1399 1400 1401
{
	memcpy(adev->reset_magic, adev->gart.ptr, AMDGPU_RESET_MAGIC_NUM);
}

1402
static bool amdgpu_device_check_vram_lost(struct amdgpu_device *adev)
1403 1404 1405 1406 1407
{
	return !!memcmp(adev->gart.ptr, adev->reset_magic,
			AMDGPU_RESET_MAGIC_NUM);
}

1408
static int amdgpu_device_ip_late_set_cg_state(struct amdgpu_device *adev)
A
Alex Deucher 已提交
1409 1410 1411
{
	int i = 0, r;

1412 1413 1414
	if (amdgpu_emu_mode == 1)
		return 0;

A
Alex Deucher 已提交
1415
	for (i = 0; i < adev->num_ip_blocks; i++) {
1416
		if (!adev->ip_blocks[i].status.valid)
A
Alex Deucher 已提交
1417
			continue;
1418
		/* skip CG for VCE/UVD, it's handled specially */
1419
		if (adev->ip_blocks[i].version->type != AMD_IP_BLOCK_TYPE_UVD &&
1420 1421
		    adev->ip_blocks[i].version->type != AMD_IP_BLOCK_TYPE_VCE &&
		    adev->ip_blocks[i].version->funcs->set_clockgating_state) {
1422
			/* enable clockgating to save power */
1423 1424
			r = adev->ip_blocks[i].version->funcs->set_clockgating_state((void *)adev,
										     AMD_CG_STATE_GATE);
1425 1426
			if (r) {
				DRM_ERROR("set_clockgating_state(gate) of IP block <%s> failed %d\n",
1427
					  adev->ip_blocks[i].version->funcs->name, r);
1428 1429
				return r;
			}
1430
		}
A
Alex Deucher 已提交
1431
	}
1432 1433 1434
	return 0;
}

1435
static int amdgpu_device_ip_late_init(struct amdgpu_device *adev)
1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454
{
	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
1455

1456
	amdgpu_device_fill_reset_magic(adev);
A
Alex Deucher 已提交
1457 1458 1459 1460

	return 0;
}

1461
static int amdgpu_device_ip_fini(struct amdgpu_device *adev)
A
Alex Deucher 已提交
1462 1463 1464
{
	int i, r;

1465
	amdgpu_amdkfd_device_fini(adev);
1466 1467
	/* need to disable SMC first */
	for (i = 0; i < adev->num_ip_blocks; i++) {
1468
		if (!adev->ip_blocks[i].status.hw)
1469
			continue;
1470 1471
		if (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_SMC &&
			adev->ip_blocks[i].version->funcs->set_clockgating_state) {
1472
			/* ungate blocks before hw fini so that we can shutdown the blocks safely */
1473 1474
			r = adev->ip_blocks[i].version->funcs->set_clockgating_state((void *)adev,
										     AMD_CG_STATE_UNGATE);
1475 1476
			if (r) {
				DRM_ERROR("set_clockgating_state(ungate) of IP block <%s> failed %d\n",
1477
					  adev->ip_blocks[i].version->funcs->name, r);
1478 1479
				return r;
			}
1480
			r = adev->ip_blocks[i].version->funcs->hw_fini((void *)adev);
1481 1482 1483
			/* XXX handle errors */
			if (r) {
				DRM_DEBUG("hw_fini of IP block <%s> failed %d\n",
1484
					  adev->ip_blocks[i].version->funcs->name, r);
1485
			}
1486
			adev->ip_blocks[i].status.hw = false;
1487 1488 1489 1490
			break;
		}
	}

A
Alex Deucher 已提交
1491
	for (i = adev->num_ip_blocks - 1; i >= 0; i--) {
1492 1493 1494
		if (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_SMC &&
			adev->firmware.load_type == AMDGPU_FW_LOAD_SMU)
			amdgpu_ucode_fini_bo(adev);
1495
		if (!adev->ip_blocks[i].status.hw)
A
Alex Deucher 已提交
1496
			continue;
1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507

		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;
			}
1508
		}
1509

1510
		r = adev->ip_blocks[i].version->funcs->hw_fini((void *)adev);
A
Alex Deucher 已提交
1511
		/* XXX handle errors */
1512
		if (r) {
1513 1514
			DRM_DEBUG("hw_fini of IP block <%s> failed %d\n",
				  adev->ip_blocks[i].version->funcs->name, r);
1515
		}
1516

1517
		adev->ip_blocks[i].status.hw = false;
A
Alex Deucher 已提交
1518 1519
	}

1520 1521 1522
	/* disable all interrupts */
	amdgpu_irq_disable_all(adev);

A
Alex Deucher 已提交
1523
	for (i = adev->num_ip_blocks - 1; i >= 0; i--) {
1524
		if (!adev->ip_blocks[i].status.sw)
A
Alex Deucher 已提交
1525
			continue;
1526 1527 1528 1529 1530 1531 1532

		if (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_GMC) {
			amdgpu_free_static_csa(adev);
			amdgpu_device_wb_fini(adev);
			amdgpu_device_vram_scratch_fini(adev);
		}

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

M
Monk Liu 已提交
1543
	for (i = adev->num_ip_blocks - 1; i >= 0; i--) {
1544
		if (!adev->ip_blocks[i].status.late_initialized)
1545
			continue;
1546 1547 1548
		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
1549 1550
	}

1551
	if (amdgpu_sriov_vf(adev))
1552 1553
		if (amdgpu_virt_release_full_gpu(adev, false))
			DRM_ERROR("failed to release exclusive mode on fini\n");
M
Monk Liu 已提交
1554

A
Alex Deucher 已提交
1555 1556 1557
	return 0;
}

1558
static void amdgpu_device_ip_late_init_func_handler(struct work_struct *work)
1559 1560 1561
{
	struct amdgpu_device *adev =
		container_of(work, struct amdgpu_device, late_init_work.work);
1562
	amdgpu_device_ip_late_set_cg_state(adev);
1563 1564
}

1565
int amdgpu_device_ip_suspend(struct amdgpu_device *adev)
A
Alex Deucher 已提交
1566 1567 1568
{
	int i, r;

1569 1570 1571
	if (amdgpu_sriov_vf(adev))
		amdgpu_virt_request_full_gpu(adev, false);

1572
	/* ungate SMC block first */
1573 1574
	r = amdgpu_device_ip_set_clockgating_state(adev, AMD_IP_BLOCK_TYPE_SMC,
						   AMD_CG_STATE_UNGATE);
1575
	if (r) {
1576
		DRM_ERROR("set_clockgating_state(ungate) SMC failed %d\n", r);
1577 1578
	}

A
Alex Deucher 已提交
1579
	for (i = adev->num_ip_blocks - 1; i >= 0; i--) {
1580
		if (!adev->ip_blocks[i].status.valid)
A
Alex Deucher 已提交
1581 1582
			continue;
		/* ungate blocks so that suspend can properly shut them down */
1583 1584
		if (i != AMD_IP_BLOCK_TYPE_SMC &&
			adev->ip_blocks[i].version->funcs->set_clockgating_state) {
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
			if (r)
				return r;
1632 1633 1634 1635 1636 1637
		}
	}

	return 0;
}

1638
static int amdgpu_device_ip_reinit_late_sriov(struct amdgpu_device *adev)
1639 1640 1641
{
	int i, r;

1642 1643
	static enum amd_ip_block_type ip_order[] = {
		AMD_IP_BLOCK_TYPE_SMC,
M
Monk Liu 已提交
1644
		AMD_IP_BLOCK_TYPE_PSP,
1645 1646 1647
		AMD_IP_BLOCK_TYPE_DCE,
		AMD_IP_BLOCK_TYPE_GFX,
		AMD_IP_BLOCK_TYPE_SDMA,
1648 1649
		AMD_IP_BLOCK_TYPE_UVD,
		AMD_IP_BLOCK_TYPE_VCE
1650
	};
1651

1652 1653 1654
	for (i = 0; i < ARRAY_SIZE(ip_order); i++) {
		int j;
		struct amdgpu_ip_block *block;
1655

1656 1657 1658 1659 1660 1661 1662 1663 1664
		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");
1665 1666
			if (r)
				return r;
1667 1668 1669 1670 1671 1672
		}
	}

	return 0;
}

1673
static int amdgpu_device_ip_resume_phase1(struct amdgpu_device *adev)
A
Alex Deucher 已提交
1674 1675 1676
{
	int i, r;

1677 1678 1679 1680 1681
	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 ||
1682 1683 1684 1685 1686 1687 1688 1689
				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;
			}
1690 1691 1692 1693 1694 1695
		}
	}

	return 0;
}

1696
static int amdgpu_device_ip_resume_phase2(struct amdgpu_device *adev)
A
Alex Deucher 已提交
1697 1698 1699 1700
{
	int i, r;

	for (i = 0; i < adev->num_ip_blocks; i++) {
1701
		if (!adev->ip_blocks[i].status.valid)
A
Alex Deucher 已提交
1702
			continue;
1703 1704 1705 1706
		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;
1707
		r = adev->ip_blocks[i].version->funcs->resume(adev);
1708
		if (r) {
1709 1710
			DRM_ERROR("resume of IP block <%s> failed %d\n",
				  adev->ip_blocks[i].version->funcs->name, r);
A
Alex Deucher 已提交
1711
			return r;
1712
		}
A
Alex Deucher 已提交
1713 1714 1715 1716 1717
	}

	return 0;
}

1718
static int amdgpu_device_ip_resume(struct amdgpu_device *adev)
1719 1720 1721
{
	int r;

1722
	r = amdgpu_device_ip_resume_phase1(adev);
1723 1724
	if (r)
		return r;
1725
	r = amdgpu_device_ip_resume_phase2(adev);
1726 1727 1728 1729

	return r;
}

1730
static void amdgpu_device_detect_sriov_bios(struct amdgpu_device *adev)
1731
{
M
Monk Liu 已提交
1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742
	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);
1743
	}
1744 1745
}

1746 1747 1748 1749 1750 1751
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:
1752
	case CHIP_KAVERI:
1753 1754
	case CHIP_KABINI:
	case CHIP_MULLINS:
1755 1756 1757 1758
	case CHIP_CARRIZO:
	case CHIP_STONEY:
	case CHIP_POLARIS11:
	case CHIP_POLARIS10:
1759
	case CHIP_POLARIS12:
1760 1761 1762 1763 1764
	case CHIP_TONGA:
	case CHIP_FIJI:
#if defined(CONFIG_DRM_AMD_DC_PRE_VEGA)
		return amdgpu_dc != 0;
#endif
1765 1766
	case CHIP_VEGA10:
#if defined(CONFIG_DRM_AMD_DC_DCN1_0)
1767
	case CHIP_RAVEN:
1768
#endif
1769
		return amdgpu_dc != 0;
1770 1771 1772 1773 1774 1775 1776 1777 1778 1779 1780 1781 1782 1783 1784
#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
1785 1786 1787
	if (amdgpu_sriov_vf(adev))
		return false;

1788 1789 1790
	return amdgpu_device_asic_has_dc_support(adev->asic_type);
}

A
Alex Deucher 已提交
1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809
/**
 * 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;
1810
	u32 max_MBps;
A
Alex Deucher 已提交
1811 1812 1813 1814 1815 1816

	adev->shutdown = false;
	adev->dev = &pdev->dev;
	adev->ddev = ddev;
	adev->pdev = pdev;
	adev->flags = flags;
1817
	adev->asic_type = flags & AMD_ASIC_MASK;
A
Alex Deucher 已提交
1818
	adev->usec_timeout = AMDGPU_MAX_USEC_TIMEOUT;
1819 1820
	if (amdgpu_emu_mode == 1)
		adev->usec_timeout *= 2;
1821
	adev->gmc.gart_size = 512 * 1024 * 1024;
A
Alex Deucher 已提交
1822 1823 1824 1825 1826
	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;
1827
	adev->vm_manager.vm_pte_num_rings = 0;
1828
	adev->gmc.gmc_funcs = NULL;
1829
	adev->fence_context = dma_fence_context_alloc(AMDGPU_MAX_RINGS);
1830
	bitmap_zero(adev->gfx.pipe_reserve_bitmap, AMDGPU_MAX_COMPUTE_QUEUES);
A
Alex Deucher 已提交
1831 1832 1833 1834 1835

	adev->smc_rreg = &amdgpu_invalid_rreg;
	adev->smc_wreg = &amdgpu_invalid_wreg;
	adev->pcie_rreg = &amdgpu_invalid_rreg;
	adev->pcie_wreg = &amdgpu_invalid_wreg;
1836 1837
	adev->pciep_rreg = &amdgpu_invalid_rreg;
	adev->pciep_wreg = &amdgpu_invalid_wreg;
A
Alex Deucher 已提交
1838 1839 1840 1841
	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;
1842 1843
	adev->gc_cac_rreg = &amdgpu_invalid_rreg;
	adev->gc_cac_wreg = &amdgpu_invalid_wreg;
A
Alex Deucher 已提交
1844 1845 1846
	adev->audio_endpt_rreg = &amdgpu_block_invalid_rreg;
	adev->audio_endpt_wreg = &amdgpu_block_invalid_wreg;

1847 1848 1849
	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
1850 1851 1852 1853

	/* mutex initialization are all done here so we
	 * can recall function without having locking issues */
	atomic_set(&adev->irq.ih.lock, 0);
1854
	mutex_init(&adev->firmware.mutex);
A
Alex Deucher 已提交
1855 1856 1857
	mutex_init(&adev->pm.mutex);
	mutex_init(&adev->gfx.gpu_clock_mutex);
	mutex_init(&adev->srbm_mutex);
1858
	mutex_init(&adev->gfx.pipe_reserve_mutex);
A
Alex Deucher 已提交
1859 1860
	mutex_init(&adev->grbm_idx_mutex);
	mutex_init(&adev->mn_lock);
A
Alex Deucher 已提交
1861
	mutex_init(&adev->virt.vf_errors.lock);
A
Alex Deucher 已提交
1862
	hash_init(adev->mn_hash);
1863
	mutex_init(&adev->lock_reset);
A
Alex Deucher 已提交
1864

1865
	amdgpu_device_check_arguments(adev);
A
Alex Deucher 已提交
1866 1867 1868 1869 1870 1871

	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);
1872
	spin_lock_init(&adev->gc_cac_idx_lock);
1873
	spin_lock_init(&adev->se_cac_idx_lock);
A
Alex Deucher 已提交
1874
	spin_lock_init(&adev->audio_endpt_idx_lock);
1875
	spin_lock_init(&adev->mm_stats.lock);
A
Alex Deucher 已提交
1876

1877 1878 1879
	INIT_LIST_HEAD(&adev->shadow_list);
	mutex_init(&adev->shadow_list_lock);

1880 1881 1882
	INIT_LIST_HEAD(&adev->ring_lru_list);
	spin_lock_init(&adev->ring_lru_list_lock);

1883 1884
	INIT_DELAYED_WORK(&adev->late_init_work,
			  amdgpu_device_ip_late_init_func_handler);
1885

1886 1887
	/* Registers mapping */
	/* TODO: block userspace mapping of io register */
1888 1889 1890 1891 1892 1893 1894
	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
1895 1896 1897 1898 1899 1900 1901 1902

	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);

1903
	/* doorbell bar mapping */
1904
	amdgpu_device_doorbell_init(adev);
A
Alex Deucher 已提交
1905 1906 1907 1908 1909 1910 1911 1912 1913 1914

	/* 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)
1915
		DRM_INFO("PCI I/O BAR is not found.\n");
A
Alex Deucher 已提交
1916

1917 1918
	amdgpu_device_get_pcie_info(adev);

A
Alex Deucher 已提交
1919
	/* early init functions */
1920
	r = amdgpu_device_ip_early_init(adev);
A
Alex Deucher 已提交
1921 1922 1923 1924 1925 1926
	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 */
1927
	vga_client_register(adev->pdev, adev, NULL, amdgpu_device_vga_set_decode);
A
Alex Deucher 已提交
1928

1929
	if (amdgpu_device_is_px(ddev))
A
Alex Deucher 已提交
1930
		runtime = true;
1931 1932 1933
	if (!pci_is_thunderbolt_attached(adev->pdev))
		vga_switcheroo_register_client(adev->pdev,
					       &amdgpu_switcheroo_ops, runtime);
A
Alex Deucher 已提交
1934 1935 1936
	if (runtime)
		vga_switcheroo_init_domain_pm_ops(adev->dev, &adev->vga_pm_domain);

1937 1938 1939
	if (amdgpu_emu_mode == 1) {
		/* post the asic on emulation mode */
		emu_soc_asic_init(adev);
1940
		goto fence_driver_init;
1941
	}
1942

A
Alex Deucher 已提交
1943
	/* Read BIOS */
1944 1945 1946 1947
	if (!amdgpu_get_bios(adev)) {
		r = -EINVAL;
		goto failed;
	}
1948

A
Alex Deucher 已提交
1949
	r = amdgpu_atombios_init(adev);
1950 1951
	if (r) {
		dev_err(adev->dev, "amdgpu_atombios_init failed\n");
A
Alex Deucher 已提交
1952
		amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_ATOMBIOS_INIT_FAIL, 0, 0);
1953
		goto failed;
1954
	}
A
Alex Deucher 已提交
1955

1956 1957
	/* detect if we are with an SRIOV vbios */
	amdgpu_device_detect_sriov_bios(adev);
1958

A
Alex Deucher 已提交
1959
	/* Post card if necessary */
A
Alex Deucher 已提交
1960
	if (amdgpu_device_need_post(adev)) {
A
Alex Deucher 已提交
1961
		if (!adev->bios) {
1962
			dev_err(adev->dev, "no vBIOS found\n");
1963 1964
			r = -EINVAL;
			goto failed;
A
Alex Deucher 已提交
1965
		}
1966
		DRM_INFO("GPU posting now...\n");
1967 1968 1969 1970 1971
		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
1972 1973
	}

1974 1975 1976 1977 1978
	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
1979
			amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_ATOMBIOS_GET_CLOCK_FAIL, 0, 0);
1980 1981 1982
			goto failed;
		}
	} else {
1983 1984 1985 1986
		/* 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
1987
			amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_ATOMBIOS_GET_CLOCK_FAIL, 0, 0);
1988
			goto failed;
1989 1990
		}
		/* init i2c buses */
1991 1992
		if (!amdgpu_device_has_dc_support(adev))
			amdgpu_atombios_i2c_init(adev);
1993
	}
A
Alex Deucher 已提交
1994

1995
fence_driver_init:
A
Alex Deucher 已提交
1996 1997
	/* Fence driver */
	r = amdgpu_fence_driver_init(adev);
1998 1999
	if (r) {
		dev_err(adev->dev, "amdgpu_fence_driver_init failed\n");
A
Alex Deucher 已提交
2000
		amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_FENCE_INIT_FAIL, 0, 0);
2001
		goto failed;
2002
	}
A
Alex Deucher 已提交
2003 2004 2005 2006

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

2007
	r = amdgpu_device_ip_init(adev);
A
Alex Deucher 已提交
2008
	if (r) {
2009 2010 2011 2012 2013 2014
		/* 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");
2015 2016 2017
			/* Don't send request since VF is inactive. */
			adev->virt.caps &= ~AMDGPU_SRIOV_CAPS_RUNTIME;
			adev->virt.ops = NULL;
2018 2019 2020
			r = -EAGAIN;
			goto failed;
		}
2021
		dev_err(adev->dev, "amdgpu_device_ip_init failed\n");
A
Alex Deucher 已提交
2022
		amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_AMDGPU_INIT_FAIL, 0, 0);
2023
		amdgpu_device_ip_fini(adev);
2024
		goto failed;
A
Alex Deucher 已提交
2025 2026 2027 2028
	}

	adev->accel_working = true;

2029 2030
	amdgpu_vm_check_compute_bug(adev);

2031 2032 2033 2034 2035 2036 2037 2038
	/* 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
2039 2040 2041
	r = amdgpu_ib_pool_init(adev);
	if (r) {
		dev_err(adev->dev, "IB initialization failed (%d).\n", r);
A
Alex Deucher 已提交
2042
		amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_IB_INIT_FAIL, 0, r);
2043
		goto failed;
A
Alex Deucher 已提交
2044 2045 2046 2047 2048 2049
	}

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

2050 2051 2052
	if (amdgpu_sriov_vf(adev))
		amdgpu_virt_init_data_exchange(adev);

2053 2054
	amdgpu_fbdev_init(adev);

2055 2056 2057 2058
	r = amdgpu_pm_sysfs_init(adev);
	if (r)
		DRM_ERROR("registering pm debugfs failed (%d).\n", r);

2059
	r = amdgpu_debugfs_gem_init(adev);
M
Monk Liu 已提交
2060
	if (r)
A
Alex Deucher 已提交
2061 2062 2063
		DRM_ERROR("registering gem debugfs failed (%d).\n", r);

	r = amdgpu_debugfs_regs_init(adev);
M
Monk Liu 已提交
2064
	if (r)
A
Alex Deucher 已提交
2065 2066
		DRM_ERROR("registering register debugfs failed (%d).\n", r);

2067
	r = amdgpu_debugfs_firmware_init(adev);
M
Monk Liu 已提交
2068
	if (r)
2069 2070
		DRM_ERROR("registering firmware debugfs failed (%d).\n", r);

2071
	r = amdgpu_debugfs_init(adev);
2072
	if (r)
2073
		DRM_ERROR("Creating debugfs files failed (%d).\n", r);
2074

A
Alex Deucher 已提交
2075 2076 2077 2078 2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090
	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.
	 */
2091
	r = amdgpu_device_ip_late_init(adev);
2092
	if (r) {
2093
		dev_err(adev->dev, "amdgpu_device_ip_late_init failed\n");
A
Alex Deucher 已提交
2094
		amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_AMDGPU_LATE_INIT_FAIL, 0, r);
2095
		goto failed;
2096
	}
A
Alex Deucher 已提交
2097 2098

	return 0;
2099 2100

failed:
2101
	amdgpu_vf_error_trans_all(adev);
2102 2103
	if (runtime)
		vga_switcheroo_fini_domain_pm_ops(adev->dev);
2104

2105
	return r;
A
Alex Deucher 已提交
2106 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121
}

/**
 * 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;
2122 2123
	if (adev->mode_info.mode_config_initialized)
		drm_crtc_force_disable_all(adev->ddev);
2124

A
Alex Deucher 已提交
2125 2126
	amdgpu_ib_pool_fini(adev);
	amdgpu_fence_driver_fini(adev);
2127
	amdgpu_pm_sysfs_fini(adev);
A
Alex Deucher 已提交
2128
	amdgpu_fbdev_fini(adev);
2129
	r = amdgpu_device_ip_fini(adev);
2130 2131 2132 2133
	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
2134
	adev->accel_working = false;
2135
	cancel_delayed_work_sync(&adev->late_init_work);
A
Alex Deucher 已提交
2136
	/* free i2c buses */
2137 2138
	if (!amdgpu_device_has_dc_support(adev))
		amdgpu_i2c_fini(adev);
2139 2140 2141 2142

	if (amdgpu_emu_mode != 1)
		amdgpu_atombios_fini(adev);

A
Alex Deucher 已提交
2143 2144
	kfree(adev->bios);
	adev->bios = NULL;
2145 2146
	if (!pci_is_thunderbolt_attached(adev->pdev))
		vga_switcheroo_unregister_client(adev->pdev);
2147 2148
	if (adev->flags & AMD_IS_PX)
		vga_switcheroo_fini_domain_pm_ops(adev->dev);
A
Alex Deucher 已提交
2149 2150 2151 2152 2153 2154
	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;
2155
	amdgpu_device_doorbell_fini(adev);
A
Alex Deucher 已提交
2156 2157 2158 2159 2160 2161 2162 2163
	amdgpu_debugfs_regs_cleanup(adev);
}


/*
 * Suspend & resume.
 */
/**
2164
 * amdgpu_device_suspend - initiate device suspend
A
Alex Deucher 已提交
2165 2166 2167 2168 2169 2170 2171 2172
 *
 * @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.
 */
2173
int amdgpu_device_suspend(struct drm_device *dev, bool suspend, bool fbcon)
A
Alex Deucher 已提交
2174 2175 2176 2177
{
	struct amdgpu_device *adev;
	struct drm_crtc *crtc;
	struct drm_connector *connector;
2178
	int r;
A
Alex Deucher 已提交
2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190

	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);

2191 2192 2193 2194 2195 2196 2197
	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
2198 2199
	}

2200 2201
	amdgpu_amdkfd_suspend(adev);

2202
	/* unpin the front buffers and cursors */
A
Alex Deucher 已提交
2203
	list_for_each_entry(crtc, &dev->mode_config.crtc_list, head) {
2204
		struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
A
Alex Deucher 已提交
2205 2206 2207
		struct amdgpu_framebuffer *rfb = to_amdgpu_framebuffer(crtc->primary->fb);
		struct amdgpu_bo *robj;

2208 2209
		if (amdgpu_crtc->cursor_bo) {
			struct amdgpu_bo *aobj = gem_to_amdgpu_bo(amdgpu_crtc->cursor_bo);
2210
			r = amdgpu_bo_reserve(aobj, true);
2211 2212 2213 2214 2215 2216
			if (r == 0) {
				amdgpu_bo_unpin(aobj);
				amdgpu_bo_unreserve(aobj);
			}
		}

A
Alex Deucher 已提交
2217 2218 2219 2220 2221 2222
		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)) {
2223
			r = amdgpu_bo_reserve(robj, true);
A
Alex Deucher 已提交
2224 2225 2226 2227 2228 2229 2230 2231 2232
			if (r == 0) {
				amdgpu_bo_unpin(robj);
				amdgpu_bo_unreserve(robj);
			}
		}
	}
	/* evict vram memory */
	amdgpu_bo_evict_vram(adev);

2233
	amdgpu_fence_driver_suspend(adev);
A
Alex Deucher 已提交
2234

2235
	r = amdgpu_device_ip_suspend(adev);
A
Alex Deucher 已提交
2236

2237 2238 2239 2240
	/* 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
2241 2242 2243 2244 2245 2246 2247
	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
2248 2249 2250 2251
	} else {
		r = amdgpu_asic_reset(adev);
		if (r)
			DRM_ERROR("amdgpu asic reset failed\n");
A
Alex Deucher 已提交
2252 2253 2254 2255 2256 2257 2258 2259 2260 2261 2262
	}

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

/**
2263
 * amdgpu_device_resume - initiate device resume
A
Alex Deucher 已提交
2264 2265 2266 2267 2268 2269 2270
 *
 * @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.
 */
2271
int amdgpu_device_resume(struct drm_device *dev, bool resume, bool fbcon)
A
Alex Deucher 已提交
2272 2273 2274
{
	struct drm_connector *connector;
	struct amdgpu_device *adev = dev->dev_private;
2275
	struct drm_crtc *crtc;
2276
	int r = 0;
A
Alex Deucher 已提交
2277 2278 2279 2280

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

J
jimqu 已提交
2281
	if (fbcon)
A
Alex Deucher 已提交
2282
		console_lock();
J
jimqu 已提交
2283

A
Alex Deucher 已提交
2284 2285 2286
	if (resume) {
		pci_set_power_state(dev->pdev, PCI_D0);
		pci_restore_state(dev->pdev);
J
jimqu 已提交
2287
		r = pci_enable_device(dev->pdev);
2288 2289
		if (r)
			goto unlock;
A
Alex Deucher 已提交
2290 2291 2292
	}

	/* post card */
A
Alex Deucher 已提交
2293
	if (amdgpu_device_need_post(adev)) {
J
jimqu 已提交
2294 2295 2296 2297
		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
2298

2299
	r = amdgpu_device_ip_resume(adev);
2300
	if (r) {
2301
		DRM_ERROR("amdgpu_device_ip_resume failed (%d).\n", r);
2302
		goto unlock;
2303
	}
2304 2305
	amdgpu_fence_driver_resume(adev);

F
Flora Cui 已提交
2306 2307 2308 2309 2310
	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
2311

2312
	r = amdgpu_device_ip_late_init(adev);
2313 2314
	if (r)
		goto unlock;
A
Alex Deucher 已提交
2315

2316 2317 2318 2319 2320 2321
	/* 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);
2322
			r = amdgpu_bo_reserve(aobj, true);
2323 2324 2325 2326 2327 2328 2329 2330 2331 2332
			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);
			}
		}
	}
2333 2334 2335
	r = amdgpu_amdkfd_resume(adev);
	if (r)
		return r;
2336

A
Alex Deucher 已提交
2337 2338
	/* blat the mode back in */
	if (fbcon) {
2339 2340 2341 2342 2343 2344 2345 2346 2347 2348
		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);
A
Alex Deucher 已提交
2349 2350 2351 2352
		}
	}

	drm_kms_helper_poll_enable(dev);
2353 2354 2355 2356 2357 2358 2359 2360 2361 2362 2363 2364 2365

	/*
	 * 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
2366 2367 2368 2369
	if (!amdgpu_device_has_dc_support(adev))
		drm_helper_hpd_irq_event(dev);
	else
		drm_kms_helper_hotplug_event(dev);
2370 2371 2372
#ifdef CONFIG_PM
	dev->dev->power.disable_depth--;
#endif
A
Alex Deucher 已提交
2373

2374
	if (fbcon)
A
Alex Deucher 已提交
2375
		amdgpu_fbdev_set_suspend(adev, 0);
2376 2377 2378

unlock:
	if (fbcon)
A
Alex Deucher 已提交
2379 2380
		console_unlock();

2381
	return r;
A
Alex Deucher 已提交
2382 2383
}

2384
static bool amdgpu_device_ip_check_soft_reset(struct amdgpu_device *adev)
2385 2386 2387 2388
{
	int i;
	bool asic_hang = false;

2389 2390 2391
	if (amdgpu_sriov_vf(adev))
		return true;

2392
	for (i = 0; i < adev->num_ip_blocks; i++) {
2393
		if (!adev->ip_blocks[i].status.valid)
2394
			continue;
2395 2396 2397 2398 2399
		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);
2400 2401 2402 2403 2404 2405
			asic_hang = true;
		}
	}
	return asic_hang;
}

2406
static int amdgpu_device_ip_pre_soft_reset(struct amdgpu_device *adev)
2407 2408 2409 2410
{
	int i, r = 0;

	for (i = 0; i < adev->num_ip_blocks; i++) {
2411
		if (!adev->ip_blocks[i].status.valid)
2412
			continue;
2413 2414 2415
		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);
2416 2417 2418 2419 2420 2421 2422 2423
			if (r)
				return r;
		}
	}

	return 0;
}

2424
static bool amdgpu_device_ip_need_full_reset(struct amdgpu_device *adev)
2425
{
2426 2427 2428
	int i;

	for (i = 0; i < adev->num_ip_blocks; i++) {
2429
		if (!adev->ip_blocks[i].status.valid)
2430
			continue;
2431 2432 2433
		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) ||
2434 2435
		    (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_DCE) ||
		     adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_PSP) {
2436
			if (adev->ip_blocks[i].status.hang) {
2437 2438 2439 2440
				DRM_INFO("Some block need full reset!\n");
				return true;
			}
		}
2441 2442 2443 2444
	}
	return false;
}

2445
static int amdgpu_device_ip_soft_reset(struct amdgpu_device *adev)
2446 2447 2448 2449
{
	int i, r = 0;

	for (i = 0; i < adev->num_ip_blocks; i++) {
2450
		if (!adev->ip_blocks[i].status.valid)
2451
			continue;
2452 2453 2454
		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);
2455 2456 2457 2458 2459 2460 2461 2462
			if (r)
				return r;
		}
	}

	return 0;
}

2463
static int amdgpu_device_ip_post_soft_reset(struct amdgpu_device *adev)
2464 2465 2466 2467
{
	int i, r = 0;

	for (i = 0; i < adev->num_ip_blocks; i++) {
2468
		if (!adev->ip_blocks[i].status.valid)
2469
			continue;
2470 2471 2472
		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);
2473 2474 2475 2476 2477 2478 2479
		if (r)
			return r;
	}

	return 0;
}

2480 2481 2482 2483
static int amdgpu_device_recover_vram_from_shadow(struct amdgpu_device *adev,
						  struct amdgpu_ring *ring,
						  struct amdgpu_bo *bo,
						  struct dma_fence **fence)
2484 2485 2486 2487
{
	uint32_t domain;
	int r;

R
Roger.He 已提交
2488 2489 2490
	if (!bo->shadow)
		return 0;

2491
	r = amdgpu_bo_reserve(bo, true);
R
Roger.He 已提交
2492 2493 2494 2495 2496
	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) {
2497 2498 2499 2500 2501 2502
		r = amdgpu_bo_validate(bo->shadow);
		if (r) {
			DRM_ERROR("bo validate failed!\n");
			goto err;
		}

R
Roger.He 已提交
2503
		r = amdgpu_bo_restore_from_shadow(adev, ring, bo,
2504
						 NULL, fence, true);
R
Roger.He 已提交
2505 2506 2507 2508 2509
		if (r) {
			DRM_ERROR("recover page table failed!\n");
			goto err;
		}
	}
2510
err:
R
Roger.He 已提交
2511 2512
	amdgpu_bo_unreserve(bo);
	return r;
2513 2514
}

2515 2516 2517 2518 2519 2520 2521 2522 2523 2524 2525 2526 2527 2528 2529 2530 2531 2532 2533 2534 2535 2536 2537 2538 2539 2540 2541 2542 2543 2544 2545 2546 2547 2548 2549 2550 2551 2552 2553 2554 2555 2556 2557 2558 2559 2560 2561 2562 2563 2564 2565 2566 2567 2568 2569 2570
static int amdgpu_device_handle_vram_lost(struct amdgpu_device *adev)
{
	struct amdgpu_ring *ring = adev->mman.buffer_funcs_ring;
	struct amdgpu_bo *bo, *tmp;
	struct dma_fence *fence = NULL, *next = NULL;
	long r = 1;
	int i = 0;
	long tmo;

	if (amdgpu_sriov_runtime(adev))
		tmo = msecs_to_jiffies(amdgpu_lockup_timeout);
	else
		tmo = msecs_to_jiffies(100);

	DRM_INFO("recover vram bo from shadow start\n");
	mutex_lock(&adev->shadow_list_lock);
	list_for_each_entry_safe(bo, tmp, &adev->shadow_list, shadow_list) {
		next = NULL;
		amdgpu_device_recover_vram_from_shadow(adev, ring, bo, &next);
		if (fence) {
			r = dma_fence_wait_timeout(fence, false, tmo);
			if (r == 0)
				pr_err("wait fence %p[%d] timeout\n", fence, i);
			else if (r < 0)
				pr_err("wait fence %p[%d] interrupted\n", fence, i);
			if (r < 1) {
				dma_fence_put(fence);
				fence = next;
				break;
			}
			i++;
		}

		dma_fence_put(fence);
		fence = next;
	}
	mutex_unlock(&adev->shadow_list_lock);

	if (fence) {
		r = dma_fence_wait_timeout(fence, false, tmo);
		if (r == 0)
			pr_err("wait fence %p[%d] timeout\n", fence, i);
		else if (r < 0)
			pr_err("wait fence %p[%d] interrupted\n", fence, i);

	}
	dma_fence_put(fence);

	if (r > 0)
		DRM_INFO("recover vram bo from shadow done\n");
	else
		DRM_ERROR("recover vram bo from shadow failed\n");

	return (r > 0?0:1);
}

2571
/*
2572
 * amdgpu_device_reset - reset ASIC/GPU for bare-metal or passthrough
2573 2574 2575
 *
 * @adev: amdgpu device pointer
 *
2576 2577 2578
 * attempt to do soft-reset or full-reset and reinitialize Asic
 * return 0 means successed otherwise failed
*/
2579
static int amdgpu_device_reset(struct amdgpu_device *adev)
2580
{
2581 2582
	bool need_full_reset, vram_lost = 0;
	int r;
2583

2584
	need_full_reset = amdgpu_device_ip_need_full_reset(adev);
2585

2586
	if (!need_full_reset) {
2587 2588 2589 2590
		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)) {
2591 2592 2593 2594
			DRM_INFO("soft reset failed, will fallback to full reset!\n");
			need_full_reset = true;
		}
	}
2595

2596
	if (need_full_reset) {
2597
		r = amdgpu_device_ip_suspend(adev);
2598

2599 2600 2601 2602
retry:
		r = amdgpu_asic_reset(adev);
		/* post card */
		amdgpu_atom_asic_init(adev->mode_info.atom_context);
2603

2604 2605
		if (!r) {
			dev_info(adev->dev, "GPU reset succeeded, trying to resume\n");
2606
			r = amdgpu_device_ip_resume_phase1(adev);
2607 2608
			if (r)
				goto out;
2609

2610
			vram_lost = amdgpu_device_check_vram_lost(adev);
2611 2612 2613 2614 2615
			if (vram_lost) {
				DRM_ERROR("VRAM is lost!\n");
				atomic_inc(&adev->vram_lost_counter);
			}

2616 2617
			r = amdgpu_gtt_mgr_recover(
				&adev->mman.bdev.man[TTM_PL_TT]);
2618 2619 2620
			if (r)
				goto out;

2621
			r = amdgpu_device_ip_resume_phase2(adev);
2622 2623 2624 2625
			if (r)
				goto out;

			if (vram_lost)
2626
				amdgpu_device_fill_reset_magic(adev);
2627
		}
2628
	}
2629

2630 2631 2632 2633 2634 2635
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);
2636
			r = amdgpu_device_ip_suspend(adev);
2637 2638 2639 2640
			need_full_reset = true;
			goto retry;
		}
	}
2641

2642 2643
	if (!r && ((need_full_reset && !(adev->flags & AMD_IS_APU)) || vram_lost))
		r = amdgpu_device_handle_vram_lost(adev);
2644

2645 2646
	return r;
}
2647

2648
/*
2649
 * amdgpu_device_reset_sriov - reset ASIC for SR-IOV vf
2650 2651 2652 2653 2654 2655
 *
 * @adev: amdgpu device pointer
 *
 * do VF FLR and reinitialize Asic
 * return 0 means successed otherwise failed
*/
2656
static int amdgpu_device_reset_sriov(struct amdgpu_device *adev, bool from_hypervisor)
2657 2658 2659 2660 2661 2662 2663 2664 2665
{
	int r;

	if (from_hypervisor)
		r = amdgpu_virt_request_full_gpu(adev, true);
	else
		r = amdgpu_virt_reset_gpu(adev);
	if (r)
		return r;
2666 2667

	/* Resume IP prior to SMC */
2668
	r = amdgpu_device_ip_reinit_early_sriov(adev);
2669 2670
	if (r)
		goto error;
2671 2672

	/* we need recover gart prior to run SMC/CP/SDMA resume */
2673
	amdgpu_gtt_mgr_recover(&adev->mman.bdev.man[TTM_PL_TT]);
2674 2675

	/* now we are okay to resume SMC/CP/SDMA */
2676
	r = amdgpu_device_ip_reinit_late_sriov(adev);
2677
	amdgpu_virt_release_full_gpu(adev, true);
2678 2679
	if (r)
		goto error;
2680 2681

	amdgpu_irq_gpu_reset_resume_helper(adev);
2682
	r = amdgpu_ib_ring_tests(adev);
2683

2684 2685 2686
	if (!r && adev->virt.gim_feature & AMDGIM_FEATURE_GIM_FLR_VRAMLOST) {
		atomic_inc(&adev->vram_lost_counter);
		r = amdgpu_device_handle_vram_lost(adev);
2687 2688
	}

2689 2690
error:

2691 2692 2693
	return r;
}

A
Alex Deucher 已提交
2694
/**
2695
 * amdgpu_device_gpu_recover - reset the asic and recover scheduler
A
Alex Deucher 已提交
2696 2697
 *
 * @adev: amdgpu device pointer
2698
 * @job: which job trigger hang
2699
 * @force forces reset regardless of amdgpu_gpu_recovery
A
Alex Deucher 已提交
2700
 *
2701
 * Attempt to reset the GPU if it has hung (all asics).
A
Alex Deucher 已提交
2702 2703
 * Returns 0 for success or an error on failure.
 */
2704 2705
int amdgpu_device_gpu_recover(struct amdgpu_device *adev,
			      struct amdgpu_job *job, bool force)
A
Alex Deucher 已提交
2706
{
2707
	struct drm_atomic_state *state = NULL;
2708
	int i, r, resched;
2709

2710
	if (!force && !amdgpu_device_ip_check_soft_reset(adev)) {
2711 2712 2713
		DRM_INFO("No hardware hang detected. Did some blocks stall?\n");
		return 0;
	}
A
Alex Deucher 已提交
2714

2715 2716 2717 2718 2719 2720
	if (!force && (amdgpu_gpu_recovery == 0 ||
			(amdgpu_gpu_recovery == -1  && !amdgpu_sriov_vf(adev)))) {
		DRM_INFO("GPU recovery disabled.\n");
		return 0;
	}

2721 2722
	dev_info(adev->dev, "GPU reset begin!\n");

2723
	mutex_lock(&adev->lock_reset);
2724
	atomic_inc(&adev->gpu_reset_counter);
2725
	adev->in_gpu_reset = 1;
A
Alex Deucher 已提交
2726

2727 2728
	/* block TTM */
	resched = ttm_bo_lock_delayed_workqueue(&adev->mman.bdev);
2729

2730 2731 2732
	/* store modesetting */
	if (amdgpu_device_has_dc_support(adev))
		state = drm_atomic_helper_suspend(adev->ddev);
2733

2734
	/* block all schedulers and reset given job's ring */
2735 2736 2737
	for (i = 0; i < AMDGPU_MAX_RINGS; ++i) {
		struct amdgpu_ring *ring = adev->rings[i];

C
Chunming Zhou 已提交
2738
		if (!ring || !ring->sched.thread)
2739
			continue;
2740

2741 2742
		kthread_park(ring->sched.thread);

2743 2744 2745
		if (job && job->ring->idx != i)
			continue;

2746
		drm_sched_hw_job_reset(&ring->sched, &job->base);
2747

M
Monk Liu 已提交
2748 2749
		/* after all hw jobs are reset, hw fence is meaningless, so force_completion */
		amdgpu_fence_driver_force_completion(ring);
2750
	}
A
Alex Deucher 已提交
2751

2752
	if (amdgpu_sriov_vf(adev))
2753
		r = amdgpu_device_reset_sriov(adev, job ? false : true);
2754
	else
2755
		r = amdgpu_device_reset(adev);
2756

2757 2758
	for (i = 0; i < AMDGPU_MAX_RINGS; ++i) {
		struct amdgpu_ring *ring = adev->rings[i];
2759

2760 2761
		if (!ring || !ring->sched.thread)
			continue;
2762

2763 2764 2765 2766 2767
		/* only need recovery sched of the given job's ring
		 * or all rings (in the case @job is NULL)
		 * after above amdgpu_reset accomplished
		 */
		if ((!job || job->ring->idx == i) && !r)
2768
			drm_sched_job_recovery(&ring->sched);
2769

2770
		kthread_unpark(ring->sched.thread);
A
Alex Deucher 已提交
2771 2772
	}

2773
	if (amdgpu_device_has_dc_support(adev)) {
2774 2775 2776
		if (drm_atomic_helper_resume(adev->ddev, state))
			dev_info(adev->dev, "drm resume failed:%d\n", r);
	} else {
2777
		drm_helper_resume_force_mode(adev->ddev);
2778
	}
A
Alex Deucher 已提交
2779 2780

	ttm_bo_unlock_delayed_workqueue(&adev->mman.bdev, resched);
2781

2782
	if (r) {
A
Alex Deucher 已提交
2783
		/* bad news, how to tell it to userspace ? */
2784 2785 2786 2787
		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));
2788
	}
A
Alex Deucher 已提交
2789

2790
	amdgpu_vf_error_trans_all(adev);
2791 2792
	adev->in_gpu_reset = 0;
	mutex_unlock(&adev->lock_reset);
A
Alex Deucher 已提交
2793 2794 2795
	return r;
}

2796
static void amdgpu_device_get_pcie_info(struct amdgpu_device *adev)
2797 2798 2799 2800
{
	u32 mask;
	int ret;

2801 2802
	if (amdgpu_pcie_gen_cap)
		adev->pm.pcie_gen_mask = amdgpu_pcie_gen_cap;
2803

2804 2805
	if (amdgpu_pcie_lane_cap)
		adev->pm.pcie_mlw_mask = amdgpu_pcie_lane_cap;
2806

2807 2808 2809 2810 2811 2812
	/* 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;
2813
		return;
2814
	}
2815

2816 2817 2818 2819 2820 2821 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
	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;
2884 2885 2886
		}
	}
}
A
Alex Deucher 已提交
2887