amdgpu_device.c 92.8 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 33 34 35 36 37 38
#include <linux/console.h>
#include <linux/slab.h>
#include <linux/debugfs.h>
#include <drm/drmP.h>
#include <drm/drm_crtc_helper.h>
#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>
A
Alex Deucher 已提交
56

57
MODULE_FIRMWARE("amdgpu/vega10_gpu_info.bin");
58
MODULE_FIRMWARE("amdgpu/raven_gpu_info.bin");
59

A
Alex Deucher 已提交
60 61 62 63
static int amdgpu_debugfs_regs_init(struct amdgpu_device *adev);
static void amdgpu_debugfs_regs_cleanup(struct amdgpu_device *adev);

static const char *amdgpu_asic_name[] = {
64 65 66 67 68
	"TAHITI",
	"PITCAIRN",
	"VERDE",
	"OLAND",
	"HAINAN",
A
Alex Deucher 已提交
69 70 71 72 73 74 75
	"BONAIRE",
	"KAVERI",
	"KABINI",
	"HAWAII",
	"MULLINS",
	"TOPAZ",
	"TONGA",
76
	"FIJI",
A
Alex Deucher 已提交
77
	"CARRIZO",
S
Samuel Li 已提交
78
	"STONEY",
79 80
	"POLARIS10",
	"POLARIS11",
81
	"POLARIS12",
K
Ken Wang 已提交
82
	"VEGA10",
83
	"RAVEN",
A
Alex Deucher 已提交
84 85 86 87 88 89 90
	"LAST",
};

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

91
	if (adev->flags & AMD_IS_PX)
A
Alex Deucher 已提交
92 93 94 95 96 97 98 99
		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
100
			uint32_t acc_flags)
A
Alex Deucher 已提交
101
{
102 103
	uint32_t ret;

M
Monk Liu 已提交
104
	if (!(acc_flags & AMDGPU_REGS_NO_KIQ) && amdgpu_sriov_runtime(adev)) {
105 106 107 108
		BUG_ON(in_interrupt());
		return amdgpu_virt_kiq_rreg(adev, reg);
	}

M
Monk Liu 已提交
109
	if ((reg * 4) < adev->rmmio_size && !(acc_flags & AMDGPU_REGS_IDX))
110
		ret = readl(((void __iomem *)adev->rmmio) + (reg * 4));
A
Alex Deucher 已提交
111 112 113 114 115 116 117 118
	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);
	}
119 120
	trace_amdgpu_mm_rreg(adev->pdev->device, reg, ret);
	return ret;
A
Alex Deucher 已提交
121 122 123
}

void amdgpu_mm_wreg(struct amdgpu_device *adev, uint32_t reg, uint32_t v,
M
Monk Liu 已提交
124
		    uint32_t acc_flags)
A
Alex Deucher 已提交
125
{
126
	trace_amdgpu_mm_wreg(adev->pdev->device, reg, v);
127

M
Monk Liu 已提交
128
	if (!(acc_flags & AMDGPU_REGS_NO_KIQ) && amdgpu_sriov_runtime(adev)) {
129 130 131 132
		BUG_ON(in_interrupt());
		return amdgpu_virt_kiq_wreg(adev, reg, v);
	}

M
Monk Liu 已提交
133
	if ((reg * 4) < adev->rmmio_size && !(acc_flags & AMDGPU_REGS_IDX))
A
Alex Deucher 已提交
134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203
		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);
	}
}

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)
{

	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));
	}
}

/**
 * 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);
	}
}

204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241
/**
 * 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
242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321
/**
 * 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();
}

static int amdgpu_vram_scratch_init(struct amdgpu_device *adev)
{
	int r;

	if (adev->vram_scratch.robj == NULL) {
		r = amdgpu_bo_create(adev, AMDGPU_GPU_PAGE_SIZE,
322
				     PAGE_SIZE, true, AMDGPU_GEM_DOMAIN_VRAM,
323 324
				     AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED |
				     AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS,
325
				     NULL, NULL, &adev->vram_scratch.robj);
A
Alex Deucher 已提交
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
		if (r) {
			return r;
		}
	}

	r = amdgpu_bo_reserve(adev->vram_scratch.robj, false);
	if (unlikely(r != 0))
		return r;
	r = amdgpu_bo_pin(adev->vram_scratch.robj,
			  AMDGPU_GEM_DOMAIN_VRAM, &adev->vram_scratch.gpu_addr);
	if (r) {
		amdgpu_bo_unreserve(adev->vram_scratch.robj);
		return r;
	}
	r = amdgpu_bo_kmap(adev->vram_scratch.robj,
				(void **)&adev->vram_scratch.ptr);
	if (r)
		amdgpu_bo_unpin(adev->vram_scratch.robj);
	amdgpu_bo_unreserve(adev->vram_scratch.robj);

	return r;
}

static void amdgpu_vram_scratch_fini(struct amdgpu_device *adev)
{
	int r;

	if (adev->vram_scratch.robj == NULL) {
		return;
	}
356
	r = amdgpu_bo_reserve(adev->vram_scratch.robj, true);
A
Alex Deucher 已提交
357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422
	if (likely(r == 0)) {
		amdgpu_bo_kunmap(adev->vram_scratch.robj);
		amdgpu_bo_unpin(adev->vram_scratch.robj);
		amdgpu_bo_unreserve(adev->vram_scratch.robj);
	}
	amdgpu_bo_unref(&adev->vram_scratch.robj);
}

/**
 * amdgpu_program_register_sequence - program an array of registers.
 *
 * @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.
 */
void amdgpu_program_register_sequence(struct amdgpu_device *adev,
				      const u32 *registers,
				      const u32 array_size)
{
	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);
	}
}

void amdgpu_pci_config_reset(struct amdgpu_device *adev)
{
	pci_write_config_dword(adev->pdev, 0x7c, AMDGPU_ASIC_RESET_DATA);
}

/*
 * GPU doorbell aperture helpers function.
 */
/**
 * amdgpu_doorbell_init - Init doorbell driver information.
 *
 * @adev: amdgpu_device pointer
 *
 * Init doorbell driver information (CIK)
 * Returns 0 on success, error on failure.
 */
static int amdgpu_doorbell_init(struct amdgpu_device *adev)
{
	/* doorbell bar mapping */
	adev->doorbell.base = pci_resource_start(adev->pdev, 2);
	adev->doorbell.size = pci_resource_len(adev->pdev, 2);

423
	adev->doorbell.num_doorbells = min_t(u32, adev->doorbell.size / sizeof(u32),
A
Alex Deucher 已提交
424 425 426 427
					     AMDGPU_DOORBELL_MAX_ASSIGNMENT+1);
	if (adev->doorbell.num_doorbells == 0)
		return -EINVAL;

428 429 430 431
	adev->doorbell.ptr = ioremap(adev->doorbell.base,
				     adev->doorbell.num_doorbells *
				     sizeof(u32));
	if (adev->doorbell.ptr == NULL)
A
Alex Deucher 已提交
432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484
		return -ENOMEM;

	return 0;
}

/**
 * amdgpu_doorbell_fini - Tear down doorbell driver information.
 *
 * @adev: amdgpu_device pointer
 *
 * Tear down doorbell driver information (CIK)
 */
static void amdgpu_doorbell_fini(struct amdgpu_device *adev)
{
	iounmap(adev->doorbell.ptr);
	adev->doorbell.ptr = NULL;
}

/**
 * amdgpu_doorbell_get_kfd_info - Report doorbell configuration required to
 *                                setup amdkfd
 *
 * @adev: amdgpu_device pointer
 * @aperture_base: output returning doorbell aperture base physical address
 * @aperture_size: output returning doorbell aperture size in bytes
 * @start_offset: output returning # of doorbell bytes reserved for amdgpu.
 *
 * amdgpu and amdkfd share the doorbell aperture. amdgpu sets it up,
 * takes doorbells required for its own rings and reports the setup to amdkfd.
 * amdgpu reserved doorbells are at the start of the doorbell aperture.
 */
void amdgpu_doorbell_get_kfd_info(struct amdgpu_device *adev,
				phys_addr_t *aperture_base,
				size_t *aperture_size,
				size_t *start_offset)
{
	/*
	 * The first num_doorbells are used by amdgpu.
	 * amdkfd takes whatever's left in the aperture.
	 */
	if (adev->doorbell.size > adev->doorbell.num_doorbells * sizeof(u32)) {
		*aperture_base = adev->doorbell.base;
		*aperture_size = adev->doorbell.size;
		*start_offset = adev->doorbell.num_doorbells * sizeof(u32);
	} else {
		*aperture_base = 0;
		*aperture_size = 0;
		*start_offset = 0;
	}
}

/*
 * amdgpu_wb_*()
485
 * Writeback is the method by which the GPU updates special pages in memory
A
Alex Xie 已提交
486
 * with the status of certain GPU events (fences, ring pointers,etc.).
A
Alex Deucher 已提交
487 488 489 490 491 492 493 494 495 496 497 498 499
 */

/**
 * amdgpu_wb_fini - Disable Writeback and free memory
 *
 * @adev: amdgpu_device pointer
 *
 * Disables Writeback and frees the Writeback memory (all asics).
 * Used at driver shutdown.
 */
static void amdgpu_wb_fini(struct amdgpu_device *adev)
{
	if (adev->wb.wb_obj) {
500 501 502
		amdgpu_bo_free_kernel(&adev->wb.wb_obj,
				      &adev->wb.gpu_addr,
				      (void **)&adev->wb.wb);
A
Alex Deucher 已提交
503 504 505 506 507 508 509 510 511
		adev->wb.wb_obj = NULL;
	}
}

/**
 * amdgpu_wb_init- Init Writeback driver info and allocate memory
 *
 * @adev: amdgpu_device pointer
 *
512
 * Initializes writeback and allocates writeback memory (all asics).
A
Alex Deucher 已提交
513 514 515 516 517 518 519 520
 * Used at driver startup.
 * Returns 0 on success or an -error on failure.
 */
static int amdgpu_wb_init(struct amdgpu_device *adev)
{
	int r;

	if (adev->wb.wb_obj == NULL) {
H
Huang Rui 已提交
521
		r = amdgpu_bo_create_kernel(adev, AMDGPU_MAX_WB * sizeof(uint32_t),
522 523 524
					    PAGE_SIZE, AMDGPU_GEM_DOMAIN_GTT,
					    &adev->wb.wb_obj, &adev->wb.gpu_addr,
					    (void **)&adev->wb.wb);
A
Alex Deucher 已提交
525 526 527 528 529 530 531 532 533
		if (r) {
			dev_warn(adev->dev, "(%d) create WB bo failed\n", r);
			return r;
		}

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

		/* clear wb memory */
H
Huang Rui 已提交
534
		memset((char *)adev->wb.wb, 0, AMDGPU_MAX_WB * sizeof(uint32_t));
A
Alex Deucher 已提交
535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560
	}

	return 0;
}

/**
 * amdgpu_wb_get - Allocate a wb entry
 *
 * @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.
 */
int amdgpu_wb_get(struct amdgpu_device *adev, u32 *wb)
{
	unsigned long offset = find_first_zero_bit(adev->wb.used, adev->wb.num_wb);
	if (offset < adev->wb.num_wb) {
		__set_bit(offset, adev->wb.used);
		*wb = offset;
		return 0;
	} else {
		return -EINVAL;
	}
}

K
Ken Wang 已提交
561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583
/**
 * amdgpu_wb_get_64bit - Allocate a wb entry
 *
 * @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.
 */
int amdgpu_wb_get_64bit(struct amdgpu_device *adev, u32 *wb)
{
	unsigned long offset = bitmap_find_next_zero_area_off(adev->wb.used,
				adev->wb.num_wb, 0, 2, 7, 0);
	if ((offset + 1) < adev->wb.num_wb) {
		__set_bit(offset, adev->wb.used);
		__set_bit(offset + 1, adev->wb.used);
		*wb = offset;
		return 0;
	} else {
		return -EINVAL;
	}
}

A
Alex Deucher 已提交
584 585 586 587 588 589 590 591 592 593 594 595 596 597
/**
 * amdgpu_wb_free - Free a wb entry
 *
 * @adev: amdgpu_device pointer
 * @wb: wb index
 *
 * Free a wb slot allocated for use by the driver (all asics)
 */
void amdgpu_wb_free(struct amdgpu_device *adev, u32 wb)
{
	if (wb < adev->wb.num_wb)
		__clear_bit(wb, adev->wb.used);
}

K
Ken Wang 已提交
598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613
/**
 * amdgpu_wb_free_64bit - Free a wb entry
 *
 * @adev: amdgpu_device pointer
 * @wb: wb index
 *
 * Free a wb slot allocated for use by the driver (all asics)
 */
void amdgpu_wb_free_64bit(struct amdgpu_device *adev, u32 wb)
{
	if ((wb + 1) < adev->wb.num_wb) {
		__clear_bit(wb, adev->wb.used);
		__clear_bit(wb + 1, adev->wb.used);
	}
}

A
Alex Deucher 已提交
614 615 616 617 618 619
/**
 * amdgpu_vram_location - try to find VRAM location
 * @adev: amdgpu device structure holding all necessary informations
 * @mc: memory controller structure holding memory informations
 * @base: base address at which to put VRAM
 *
620
 * Function will try to place VRAM at base address provided
A
Alex Deucher 已提交
621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641
 * as parameter (which is so far either PCI aperture address or
 * for IGP TOM base address).
 *
 * If there is not enough space to fit the unvisible VRAM in the 32bits
 * address space then we limit the VRAM size to the aperture.
 *
 * Note: We don't explicitly enforce VRAM start to be aligned on VRAM size,
 * this shouldn't be a problem as we are using the PCI aperture as a reference.
 * Otherwise this would be needed for rv280, all r3xx, and all r4xx, but
 * not IGP.
 *
 * Note: we use mc_vram_size as on some board we need to program the mc to
 * cover the whole aperture even if VRAM size is inferior to aperture size
 * Novell bug 204882 + along with lots of ubuntu ones
 *
 * Note: when limiting vram it's safe to overwritte real_vram_size because
 * we are not in case where real_vram_size is inferior to mc_vram_size (ie
 * note afected by bogus hw of Novell bug 204882 + along with lots of ubuntu
 * ones)
 *
 * Note: IGP TOM addr should be the same as the aperture addr, we don't
642
 * explicitly check for that though.
A
Alex Deucher 已提交
643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686
 *
 * FIXME: when reducing VRAM size align new size on power of 2.
 */
void amdgpu_vram_location(struct amdgpu_device *adev, struct amdgpu_mc *mc, u64 base)
{
	uint64_t limit = (uint64_t)amdgpu_vram_limit << 20;

	mc->vram_start = base;
	if (mc->mc_vram_size > (adev->mc.mc_mask - base + 1)) {
		dev_warn(adev->dev, "limiting VRAM to PCI aperture size\n");
		mc->real_vram_size = mc->aper_size;
		mc->mc_vram_size = mc->aper_size;
	}
	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);
}

/**
 * amdgpu_gtt_location - try to find GTT location
 * @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.
 */
void amdgpu_gtt_location(struct amdgpu_device *adev, struct amdgpu_mc *mc)
{
	u64 size_af, size_bf;

	size_af = ((adev->mc.mc_mask - mc->vram_end) + mc->gtt_base_align) & ~mc->gtt_base_align;
	size_bf = mc->vram_start & ~mc->gtt_base_align;
	if (size_bf > size_af) {
		if (mc->gtt_size > size_bf) {
			dev_warn(adev->dev, "limiting GTT\n");
			mc->gtt_size = size_bf;
		}
687
		mc->gtt_start = 0;
A
Alex Deucher 已提交
688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703
	} else {
		if (mc->gtt_size > size_af) {
			dev_warn(adev->dev, "limiting GTT\n");
			mc->gtt_size = size_af;
		}
		mc->gtt_start = (mc->vram_end + 1 + mc->gtt_base_align) & ~mc->gtt_base_align;
	}
	mc->gtt_end = mc->gtt_start + mc->gtt_size - 1;
	dev_info(adev->dev, "GTT: %lluM 0x%016llX - 0x%016llX\n",
			mc->gtt_size >> 20, mc->gtt_start, mc->gtt_end);
}

/*
 * GPU helpers function.
 */
/**
704
 * amdgpu_need_post - check if the hw need post or not
A
Alex Deucher 已提交
705 706 707
 *
 * @adev: amdgpu_device pointer
 *
708 709 710
 * 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
711
 */
712
bool amdgpu_need_post(struct amdgpu_device *adev)
A
Alex Deucher 已提交
713 714 715
{
	uint32_t reg;

716 717 718 719
	if (adev->has_hw_reset) {
		adev->has_hw_reset = false;
		return true;
	}
A
Alex Deucher 已提交
720
	/* then check MEM_SIZE, in case the crtcs are off */
721
	reg = amdgpu_asic_get_config_memsize(adev);
A
Alex Deucher 已提交
722

723
	if ((reg != 0) && (reg != 0xffffffff))
724
		return false;
A
Alex Deucher 已提交
725

726
	return true;
A
Alex Deucher 已提交
727 728 729

}

730 731 732 733 734 735
static bool amdgpu_vpost_needed(struct amdgpu_device *adev)
{
	if (amdgpu_sriov_vf(adev))
		return false;

	if (amdgpu_passthrough(adev)) {
M
Monk Liu 已提交
736 737 738 739
		/* 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
740 741 742 743 744 745 746 747 748 749
		 */
		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
750 751
			if (fw_ver < 0x00160e00)
				return true;
752 753
		}
	}
754
	return amdgpu_need_post(adev);
755 756
}

A
Alex Deucher 已提交
757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946
/**
 * amdgpu_dummy_page_init - init dummy page used by the driver
 *
 * @adev: amdgpu_device pointer
 *
 * Allocate the dummy page used by the driver (all asics).
 * This dummy page is used by the driver as a filler for gart entries
 * when pages are taken out of the GART
 * Returns 0 on sucess, -ENOMEM on failure.
 */
int amdgpu_dummy_page_init(struct amdgpu_device *adev)
{
	if (adev->dummy_page.page)
		return 0;
	adev->dummy_page.page = alloc_page(GFP_DMA32 | GFP_KERNEL | __GFP_ZERO);
	if (adev->dummy_page.page == NULL)
		return -ENOMEM;
	adev->dummy_page.addr = pci_map_page(adev->pdev, adev->dummy_page.page,
					0, PAGE_SIZE, PCI_DMA_BIDIRECTIONAL);
	if (pci_dma_mapping_error(adev->pdev, adev->dummy_page.addr)) {
		dev_err(&adev->pdev->dev, "Failed to DMA MAP the dummy page\n");
		__free_page(adev->dummy_page.page);
		adev->dummy_page.page = NULL;
		return -ENOMEM;
	}
	return 0;
}

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


/* ATOM accessor methods */
/*
 * ATOM is an interpreted byte code stored in tables in the vbios.  The
 * driver registers callbacks to access registers and the interpreter
 * in the driver parses the tables and executes then to program specific
 * actions (set display modes, asic init, etc.).  See amdgpu_atombios.c,
 * atombios.h, and atom.c
 */

/**
 * cail_pll_read - read PLL register
 *
 * @info: atom card_info pointer
 * @reg: PLL register offset
 *
 * Provides a PLL register accessor for the atom interpreter (r4xx+).
 * Returns the value of the PLL register.
 */
static uint32_t cail_pll_read(struct card_info *info, uint32_t reg)
{
	return 0;
}

/**
 * cail_pll_write - write PLL register
 *
 * @info: atom card_info pointer
 * @reg: PLL register offset
 * @val: value to write to the pll register
 *
 * Provides a PLL register accessor for the atom interpreter (r4xx+).
 */
static void cail_pll_write(struct card_info *info, uint32_t reg, uint32_t val)
{

}

/**
 * cail_mc_read - read MC (Memory Controller) register
 *
 * @info: atom card_info pointer
 * @reg: MC register offset
 *
 * Provides an MC register accessor for the atom interpreter (r4xx+).
 * Returns the value of the MC register.
 */
static uint32_t cail_mc_read(struct card_info *info, uint32_t reg)
{
	return 0;
}

/**
 * cail_mc_write - write MC (Memory Controller) register
 *
 * @info: atom card_info pointer
 * @reg: MC register offset
 * @val: value to write to the pll register
 *
 * Provides a MC register accessor for the atom interpreter (r4xx+).
 */
static void cail_mc_write(struct card_info *info, uint32_t reg, uint32_t val)
{

}

/**
 * cail_reg_write - write MMIO register
 *
 * @info: atom card_info pointer
 * @reg: MMIO register offset
 * @val: value to write to the pll register
 *
 * Provides a MMIO register accessor for the atom interpreter (r4xx+).
 */
static void cail_reg_write(struct card_info *info, uint32_t reg, uint32_t val)
{
	struct amdgpu_device *adev = info->dev->dev_private;

	WREG32(reg, val);
}

/**
 * cail_reg_read - read MMIO register
 *
 * @info: atom card_info pointer
 * @reg: MMIO register offset
 *
 * Provides an MMIO register accessor for the atom interpreter (r4xx+).
 * Returns the value of the MMIO register.
 */
static uint32_t cail_reg_read(struct card_info *info, uint32_t reg)
{
	struct amdgpu_device *adev = info->dev->dev_private;
	uint32_t r;

	r = RREG32(reg);
	return r;
}

/**
 * cail_ioreg_write - write IO register
 *
 * @info: atom card_info pointer
 * @reg: IO register offset
 * @val: value to write to the pll register
 *
 * Provides a IO register accessor for the atom interpreter (r4xx+).
 */
static void cail_ioreg_write(struct card_info *info, uint32_t reg, uint32_t val)
{
	struct amdgpu_device *adev = info->dev->dev_private;

	WREG32_IO(reg, val);
}

/**
 * cail_ioreg_read - read IO register
 *
 * @info: atom card_info pointer
 * @reg: IO register offset
 *
 * Provides an IO register accessor for the atom interpreter (r4xx+).
 * Returns the value of the IO register.
 */
static uint32_t cail_ioreg_read(struct card_info *info, uint32_t reg)
{
	struct amdgpu_device *adev = info->dev->dev_private;
	uint32_t r;

	r = RREG32_IO(reg);
	return r;
}

/**
 * amdgpu_atombios_fini - free the driver info and callbacks for atombios
 *
 * @adev: amdgpu_device pointer
 *
 * Frees the driver info and register access callbacks for the ATOM
 * interpreter (r4xx+).
 * Called at driver shutdown.
 */
static void amdgpu_atombios_fini(struct amdgpu_device *adev)
{
M
Monk Liu 已提交
947
	if (adev->mode_info.atom_context) {
A
Alex Deucher 已提交
948
		kfree(adev->mode_info.atom_context->scratch);
M
Monk Liu 已提交
949 950
		kfree(adev->mode_info.atom_context->iio);
	}
A
Alex Deucher 已提交
951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983
	kfree(adev->mode_info.atom_context);
	adev->mode_info.atom_context = NULL;
	kfree(adev->mode_info.atom_card_info);
	adev->mode_info.atom_card_info = NULL;
}

/**
 * amdgpu_atombios_init - init the driver info and callbacks for atombios
 *
 * @adev: amdgpu_device pointer
 *
 * Initializes the driver info and register access callbacks for the
 * ATOM interpreter (r4xx+).
 * Returns 0 on sucess, -ENOMEM on failure.
 * Called at driver startup.
 */
static int amdgpu_atombios_init(struct amdgpu_device *adev)
{
	struct card_info *atom_card_info =
	    kzalloc(sizeof(struct card_info), GFP_KERNEL);

	if (!atom_card_info)
		return -ENOMEM;

	adev->mode_info.atom_card_info = atom_card_info;
	atom_card_info->dev = adev->ddev;
	atom_card_info->reg_read = cail_reg_read;
	atom_card_info->reg_write = cail_reg_write;
	/* needed for iio ops */
	if (adev->rio_mem) {
		atom_card_info->ioreg_read = cail_ioreg_read;
		atom_card_info->ioreg_write = cail_ioreg_write;
	} else {
984
		DRM_INFO("PCI I/O BAR is not found. Using MMIO to access ATOM BIOS\n");
A
Alex Deucher 已提交
985 986 987 988 989 990 991 992 993 994 995 996 997 998 999
		atom_card_info->ioreg_read = cail_reg_read;
		atom_card_info->ioreg_write = cail_reg_write;
	}
	atom_card_info->mc_read = cail_mc_read;
	atom_card_info->mc_write = cail_mc_write;
	atom_card_info->pll_read = cail_pll_read;
	atom_card_info->pll_write = cail_pll_write;

	adev->mode_info.atom_context = amdgpu_atom_parse(atom_card_info, adev->bios);
	if (!adev->mode_info.atom_context) {
		amdgpu_atombios_fini(adev);
		return -ENOMEM;
	}

	mutex_init(&adev->mode_info.atom_context->mutex);
1000 1001 1002 1003 1004 1005 1006
	if (adev->is_atom_fw) {
		amdgpu_atomfirmware_scratch_regs_init(adev);
		amdgpu_atomfirmware_allocate_fb_scratch(adev);
	} else {
		amdgpu_atombios_scratch_regs_init(adev);
		amdgpu_atombios_allocate_fb_scratch(adev);
	}
A
Alex Deucher 已提交
1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043
	return 0;
}

/* if we get transitioned to only one device, take VGA back */
/**
 * amdgpu_vga_set_decode - enable/disable vga decode
 *
 * @cookie: amdgpu_device pointer
 * @state: enable/disable vga decode
 *
 * Enable/disable vga decode (all asics).
 * Returns VGA resource flags.
 */
static unsigned int amdgpu_vga_set_decode(void *cookie, bool state)
{
	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;
}

/**
 * amdgpu_check_pot_argument - check that argument is a power of two
 *
 * @arg: value to check
 *
 * Validates that a certain argument is a power of two (all asics).
 * Returns true if argument is valid.
 */
static bool amdgpu_check_pot_argument(int arg)
{
	return (arg & (arg - 1)) == 0;
}

1044
static void amdgpu_check_block_size(struct amdgpu_device *adev)
1045 1046 1047 1048
{
	/* 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 */
1049 1050
	if (amdgpu_vm_block_size == -1)
		return;
1051

1052
	if (amdgpu_vm_block_size < 9) {
1053 1054
		dev_warn(adev->dev, "VM page table size (%d) too small\n",
			 amdgpu_vm_block_size);
1055
		goto def_value;
1056 1057 1058 1059 1060 1061
	}

	if (amdgpu_vm_block_size > 24 ||
	    (amdgpu_vm_size * 1024) < (1ull << amdgpu_vm_block_size)) {
		dev_warn(adev->dev, "VM page table size (%d) too large\n",
			 amdgpu_vm_block_size);
1062
		goto def_value;
1063
	}
1064 1065 1066 1067 1068

	return;

def_value:
	amdgpu_vm_block_size = -1;
1069 1070
}

1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096
static void amdgpu_check_vm_size(struct amdgpu_device *adev)
{
	if (!amdgpu_check_pot_argument(amdgpu_vm_size)) {
		dev_warn(adev->dev, "VM size (%d) must be a power of 2\n",
			 amdgpu_vm_size);
		goto def_value;
	}

	if (amdgpu_vm_size < 1) {
		dev_warn(adev->dev, "VM size (%d) too small, min is 1GB\n",
			 amdgpu_vm_size);
		goto def_value;
	}

	/*
	 * Max GPUVM size for Cayman, SI, CI VI are 40 bits.
	 */
	if (amdgpu_vm_size > 1024) {
		dev_warn(adev->dev, "VM size (%d) too large, max is 1TB\n",
			 amdgpu_vm_size);
		goto def_value;
	}

	return;

def_value:
1097
	amdgpu_vm_size = -1;
1098 1099
}

A
Alex Deucher 已提交
1100 1101 1102 1103 1104 1105 1106 1107 1108 1109
/**
 * amdgpu_check_arguments - validate module params
 *
 * @adev: amdgpu_device pointer
 *
 * Validates certain module parameters and updates
 * the associated values used by the driver (all asics).
 */
static void amdgpu_check_arguments(struct amdgpu_device *adev)
{
1110 1111 1112 1113 1114 1115 1116 1117 1118
	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;
	} else if (!amdgpu_check_pot_argument(amdgpu_sched_jobs)){
		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
1119 1120

	if (amdgpu_gart_size != -1) {
1121
		/* gtt size must be greater or equal to 32M */
A
Alex Deucher 已提交
1122 1123 1124 1125 1126 1127 1128
		if (amdgpu_gart_size < 32) {
			dev_warn(adev->dev, "gart size (%d) too small\n",
				 amdgpu_gart_size);
			amdgpu_gart_size = -1;
		}
	}

1129
	amdgpu_check_vm_size(adev);
A
Alex Deucher 已提交
1130

1131
	amdgpu_check_block_size(adev);
C
Christian König 已提交
1132

1133 1134
	if (amdgpu_vram_page_split != -1 && (amdgpu_vram_page_split < 16 ||
	    !amdgpu_check_pot_argument(amdgpu_vram_page_split))) {
C
Christian König 已提交
1135 1136 1137 1138
		dev_warn(adev->dev, "invalid VRAM page split (%d)\n",
			 amdgpu_vram_page_split);
		amdgpu_vram_page_split = 1024;
	}
A
Alex Deucher 已提交
1139 1140 1141 1142 1143 1144
}

/**
 * amdgpu_switcheroo_set_state - set switcheroo state
 *
 * @pdev: pci dev pointer
1145
 * @state: vga_switcheroo state
A
Alex Deucher 已提交
1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159
 *
 * 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) {
		unsigned d3_delay = dev->pdev->d3_delay;

1160
		pr_info("amdgpu: switched on\n");
A
Alex Deucher 已提交
1161 1162 1163
		/* don't suspend or resume card normally */
		dev->switch_power_state = DRM_SWITCH_POWER_CHANGING;

1164
		amdgpu_device_resume(dev, true, true);
A
Alex Deucher 已提交
1165 1166 1167 1168 1169 1170

		dev->pdev->d3_delay = d3_delay;

		dev->switch_power_state = DRM_SWITCH_POWER_ON;
		drm_kms_helper_poll_enable(dev);
	} else {
1171
		pr_info("amdgpu: switched off\n");
A
Alex Deucher 已提交
1172 1173
		drm_kms_helper_poll_disable(dev);
		dev->switch_power_state = DRM_SWITCH_POWER_CHANGING;
1174
		amdgpu_device_suspend(dev, true, true);
A
Alex Deucher 已提交
1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206
		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,
};

int amdgpu_set_clockgating_state(struct amdgpu_device *adev,
1207 1208
				  enum amd_ip_block_type block_type,
				  enum amd_clockgating_state state)
A
Alex Deucher 已提交
1209 1210 1211 1212
{
	int i, r = 0;

	for (i = 0; i < adev->num_ip_blocks; i++) {
1213
		if (!adev->ip_blocks[i].status.valid)
1214
			continue;
1215 1216 1217 1218 1219 1220 1221 1222 1223
		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
1224 1225 1226 1227 1228
	}
	return r;
}

int amdgpu_set_powergating_state(struct amdgpu_device *adev,
1229 1230
				  enum amd_ip_block_type block_type,
				  enum amd_powergating_state state)
A
Alex Deucher 已提交
1231 1232 1233 1234
{
	int i, r = 0;

	for (i = 0; i < adev->num_ip_blocks; i++) {
1235
		if (!adev->ip_blocks[i].status.valid)
1236
			continue;
1237 1238 1239 1240 1241 1242 1243 1244 1245
		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
1246 1247 1248 1249
	}
	return r;
}

1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261
void amdgpu_get_clockgating_state(struct amdgpu_device *adev, u32 *flags)
{
	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);
	}
}

1262 1263 1264 1265 1266 1267
int amdgpu_wait_for_idle(struct amdgpu_device *adev,
			 enum amd_ip_block_type block_type)
{
	int i, r;

	for (i = 0; i < adev->num_ip_blocks; i++) {
1268
		if (!adev->ip_blocks[i].status.valid)
1269
			continue;
1270 1271
		if (adev->ip_blocks[i].version->type == block_type) {
			r = adev->ip_blocks[i].version->funcs->wait_for_idle((void *)adev);
1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286
			if (r)
				return r;
			break;
		}
	}
	return 0;

}

bool amdgpu_is_idle(struct amdgpu_device *adev,
		    enum amd_ip_block_type block_type)
{
	int i;

	for (i = 0; i < adev->num_ip_blocks; i++) {
1287
		if (!adev->ip_blocks[i].status.valid)
1288
			continue;
1289 1290
		if (adev->ip_blocks[i].version->type == block_type)
			return adev->ip_blocks[i].version->funcs->is_idle((void *)adev);
1291 1292 1293 1294 1295
	}
	return true;

}

1296 1297
struct amdgpu_ip_block * amdgpu_get_ip_block(struct amdgpu_device *adev,
					     enum amd_ip_block_type type)
A
Alex Deucher 已提交
1298 1299 1300 1301
{
	int i;

	for (i = 0; i < adev->num_ip_blocks; i++)
1302
		if (adev->ip_blocks[i].version->type == type)
A
Alex Deucher 已提交
1303 1304 1305 1306 1307 1308 1309 1310 1311
			return &adev->ip_blocks[i];

	return NULL;
}

/**
 * amdgpu_ip_block_version_cmp
 *
 * @adev: amdgpu_device pointer
1312
 * @type: enum amd_ip_block_type
A
Alex Deucher 已提交
1313 1314 1315 1316 1317 1318 1319
 * @major: major version
 * @minor: minor version
 *
 * return 0 if equal or greater
 * return 1 if smaller or the ip_block doesn't exist
 */
int amdgpu_ip_block_version_cmp(struct amdgpu_device *adev,
1320
				enum amd_ip_block_type type,
A
Alex Deucher 已提交
1321 1322
				u32 major, u32 minor)
{
1323
	struct amdgpu_ip_block *ip_block = amdgpu_get_ip_block(adev, type);
A
Alex Deucher 已提交
1324

1325 1326 1327
	if (ip_block && ((ip_block->version->major > major) ||
			((ip_block->version->major == major) &&
			(ip_block->version->minor >= minor))))
A
Alex Deucher 已提交
1328 1329 1330 1331 1332
		return 0;

	return 1;
}

1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352
/**
 * amdgpu_ip_block_add
 *
 * @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.
 */
int amdgpu_ip_block_add(struct amdgpu_device *adev,
			const struct amdgpu_ip_block_version *ip_block_version)
{
	if (!ip_block_version)
		return -EINVAL;

	adev->ip_blocks[adev->num_ip_blocks++].version = ip_block_version;

	return 0;
}

1353
static void amdgpu_device_enable_virtual_display(struct amdgpu_device *adev)
1354 1355 1356 1357 1358 1359
{
	adev->enable_virtual_display = false;

	if (amdgpu_virtual_display) {
		struct drm_device *ddev = adev->ddev;
		const char *pci_address_name = pci_name(ddev->pdev);
1360
		char *pciaddstr, *pciaddstr_tmp, *pciaddname_tmp, *pciaddname;
1361 1362 1363

		pciaddstr = kstrdup(amdgpu_virtual_display, GFP_KERNEL);
		pciaddstr_tmp = pciaddstr;
1364 1365
		while ((pciaddname_tmp = strsep(&pciaddstr_tmp, ";"))) {
			pciaddname = strsep(&pciaddname_tmp, ",");
1366 1367
			if (!strcmp("all", pciaddname)
			    || !strcmp(pci_address_name, pciaddname)) {
1368 1369 1370
				long num_crtc;
				int res = -1;

1371
				adev->enable_virtual_display = true;
1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385

				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;
				}
1386 1387 1388 1389
				break;
			}
		}

1390 1391 1392
		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);
1393 1394 1395 1396 1397

		kfree(pciaddstr);
	}
}

1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433
static int amdgpu_device_parse_gpu_info_fw(struct amdgpu_device *adev)
{
	const struct firmware *fw;
	const char *chip_name;
	char fw_name[30];
	int err;
	const struct gpu_info_firmware_header_v1_0 *hdr;

	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;
1434 1435 1436
	case CHIP_RAVEN:
		chip_name = "raven";
		break;
1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492
	}

	snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_gpu_info.bin", chip_name);
	err = request_firmware(&fw, fw_name, adev->dev);
	if (err) {
		dev_err(adev->dev,
			"Failed to load gpu_info firmware \"%s\"\n",
			fw_name);
		goto out;
	}
	err = amdgpu_ucode_validate(fw);
	if (err) {
		dev_err(adev->dev,
			"Failed to validate gpu_info firmware \"%s\"\n",
			fw_name);
		goto out;
	}

	hdr = (const struct gpu_info_firmware_header_v1_0 *)fw->data;
	amdgpu_ucode_print_gpu_info_hdr(&hdr->header);

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

		adev->gfx.config.max_shader_engines = gpu_info_fw->gc_num_se;
		adev->gfx.config.max_cu_per_sh = gpu_info_fw->gc_num_cu_per_sh;
		adev->gfx.config.max_sh_per_se = gpu_info_fw->gc_num_sh_per_se;
		adev->gfx.config.max_backends_per_se = gpu_info_fw->gc_num_rb_per_se;
		adev->gfx.config.max_texture_channel_caches =
			gpu_info_fw->gc_num_tccs;
		adev->gfx.config.max_gprs = gpu_info_fw->gc_num_gprs;
		adev->gfx.config.max_gs_threads = gpu_info_fw->gc_num_max_gs_thds;
		adev->gfx.config.gs_vgt_table_depth = gpu_info_fw->gc_gs_table_depth;
		adev->gfx.config.gs_prim_buffer_depth = gpu_info_fw->gc_gsprim_buff_depth;
		adev->gfx.config.double_offchip_lds_buf =
			gpu_info_fw->gc_double_offchip_lds_buffer;
		adev->gfx.cu_info.wave_front_size = gpu_info_fw->gc_wave_size;
		break;
	}
	default:
		dev_err(adev->dev,
			"Unsupported gpu_info table %d\n", hdr->header.ucode_version);
		err = -EINVAL;
		goto out;
	}
out:
	release_firmware(fw);
	fw = NULL;

	return err;
}

A
Alex Deucher 已提交
1493 1494
static int amdgpu_early_init(struct amdgpu_device *adev)
{
1495
	int i, r;
A
Alex Deucher 已提交
1496

1497
	amdgpu_device_enable_virtual_display(adev);
1498

A
Alex Deucher 已提交
1499
	switch (adev->asic_type) {
1500 1501
	case CHIP_TOPAZ:
	case CHIP_TONGA:
1502
	case CHIP_FIJI:
1503 1504
	case CHIP_POLARIS11:
	case CHIP_POLARIS10:
1505
	case CHIP_POLARIS12:
1506
	case CHIP_CARRIZO:
1507 1508
	case CHIP_STONEY:
		if (adev->asic_type == CHIP_CARRIZO || adev->asic_type == CHIP_STONEY)
1509 1510 1511 1512 1513 1514 1515 1516
			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
1517 1518 1519 1520 1521 1522
#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
1523
		adev->family = AMDGPU_FAMILY_SI;
K
Ken Wang 已提交
1524 1525 1526 1527 1528
		r = si_set_ip_blocks(adev);
		if (r)
			return r;
		break;
#endif
1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544
#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
1545 1546 1547 1548 1549 1550
	case  CHIP_VEGA10:
	case  CHIP_RAVEN:
		if (adev->asic_type == CHIP_RAVEN)
			adev->family = AMDGPU_FAMILY_RV;
		else
			adev->family = AMDGPU_FAMILY_AI;
1551 1552 1553 1554 1555

		r = soc15_set_ip_blocks(adev);
		if (r)
			return r;
		break;
A
Alex Deucher 已提交
1556 1557 1558 1559 1560
	default:
		/* FIXME: not supported yet */
		return -EINVAL;
	}

1561 1562 1563 1564
	r = amdgpu_device_parse_gpu_info_fw(adev);
	if (r)
		return r;

1565 1566 1567 1568 1569 1570
	if (amdgpu_sriov_vf(adev)) {
		r = amdgpu_virt_request_full_gpu(adev, true);
		if (r)
			return r;
	}

A
Alex Deucher 已提交
1571 1572 1573
	for (i = 0; i < adev->num_ip_blocks; i++) {
		if ((amdgpu_ip_block_mask & (1 << i)) == 0) {
			DRM_ERROR("disabled ip block: %d\n", i);
1574
			adev->ip_blocks[i].status.valid = false;
A
Alex Deucher 已提交
1575
		} else {
1576 1577
			if (adev->ip_blocks[i].version->funcs->early_init) {
				r = adev->ip_blocks[i].version->funcs->early_init((void *)adev);
1578
				if (r == -ENOENT) {
1579
					adev->ip_blocks[i].status.valid = false;
1580
				} else if (r) {
1581 1582
					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
1583
					return r;
1584
				} else {
1585
					adev->ip_blocks[i].status.valid = true;
1586
				}
1587
			} else {
1588
				adev->ip_blocks[i].status.valid = true;
A
Alex Deucher 已提交
1589 1590 1591 1592
			}
		}
	}

1593 1594 1595
	adev->cg_flags &= amdgpu_cg_mask;
	adev->pg_flags &= amdgpu_pg_mask;

A
Alex Deucher 已提交
1596 1597 1598 1599 1600 1601 1602 1603
	return 0;
}

static int amdgpu_init(struct amdgpu_device *adev)
{
	int i, r;

	for (i = 0; i < adev->num_ip_blocks; i++) {
1604
		if (!adev->ip_blocks[i].status.valid)
A
Alex Deucher 已提交
1605
			continue;
1606
		r = adev->ip_blocks[i].version->funcs->sw_init((void *)adev);
1607
		if (r) {
1608 1609
			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
1610
			return r;
1611
		}
1612
		adev->ip_blocks[i].status.sw = true;
A
Alex Deucher 已提交
1613
		/* need to do gmc hw init early so we can allocate gpu mem */
1614
		if (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_GMC) {
A
Alex Deucher 已提交
1615
			r = amdgpu_vram_scratch_init(adev);
1616 1617
			if (r) {
				DRM_ERROR("amdgpu_vram_scratch_init failed %d\n", r);
A
Alex Deucher 已提交
1618
				return r;
1619
			}
1620
			r = adev->ip_blocks[i].version->funcs->hw_init((void *)adev);
1621 1622
			if (r) {
				DRM_ERROR("hw_init %d failed %d\n", i, r);
A
Alex Deucher 已提交
1623
				return r;
1624
			}
A
Alex Deucher 已提交
1625
			r = amdgpu_wb_init(adev);
1626 1627
			if (r) {
				DRM_ERROR("amdgpu_wb_init failed %d\n", r);
A
Alex Deucher 已提交
1628
				return r;
1629
			}
1630
			adev->ip_blocks[i].status.hw = true;
M
Monk Liu 已提交
1631 1632 1633 1634 1635 1636 1637 1638 1639

			/* 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
1640 1641 1642 1643
		}
	}

	for (i = 0; i < adev->num_ip_blocks; i++) {
1644
		if (!adev->ip_blocks[i].status.sw)
A
Alex Deucher 已提交
1645 1646
			continue;
		/* gmc hw init is done early */
1647
		if (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_GMC)
A
Alex Deucher 已提交
1648
			continue;
1649
		r = adev->ip_blocks[i].version->funcs->hw_init((void *)adev);
1650
		if (r) {
1651 1652
			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
1653
			return r;
1654
		}
1655
		adev->ip_blocks[i].status.hw = true;
A
Alex Deucher 已提交
1656 1657 1658 1659 1660
	}

	return 0;
}

1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671
static void amdgpu_fill_reset_magic(struct amdgpu_device *adev)
{
	memcpy(adev->reset_magic, adev->gart.ptr, AMDGPU_RESET_MAGIC_NUM);
}

static bool amdgpu_check_vram_lost(struct amdgpu_device *adev)
{
	return !!memcmp(adev->gart.ptr, adev->reset_magic,
			AMDGPU_RESET_MAGIC_NUM);
}

A
Alex Deucher 已提交
1672 1673 1674 1675 1676
static int amdgpu_late_init(struct amdgpu_device *adev)
{
	int i = 0, r;

	for (i = 0; i < adev->num_ip_blocks; i++) {
1677
		if (!adev->ip_blocks[i].status.valid)
A
Alex Deucher 已提交
1678
			continue;
1679 1680
		if (adev->ip_blocks[i].version->funcs->late_init) {
			r = adev->ip_blocks[i].version->funcs->late_init((void *)adev);
1681
			if (r) {
1682 1683
				DRM_ERROR("late_init of IP block <%s> failed %d\n",
					  adev->ip_blocks[i].version->funcs->name, r);
A
Alex Deucher 已提交
1684
				return r;
1685
			}
1686
			adev->ip_blocks[i].status.late_initialized = true;
A
Alex Deucher 已提交
1687
		}
1688
		/* skip CG for VCE/UVD, it's handled specially */
1689 1690
		if (adev->ip_blocks[i].version->type != AMD_IP_BLOCK_TYPE_UVD &&
		    adev->ip_blocks[i].version->type != AMD_IP_BLOCK_TYPE_VCE) {
1691
			/* enable clockgating to save power */
1692 1693
			r = adev->ip_blocks[i].version->funcs->set_clockgating_state((void *)adev,
										     AMD_CG_STATE_GATE);
1694 1695
			if (r) {
				DRM_ERROR("set_clockgating_state(gate) of IP block <%s> failed %d\n",
1696
					  adev->ip_blocks[i].version->funcs->name, r);
1697 1698
				return r;
			}
1699
		}
A
Alex Deucher 已提交
1700 1701
	}

1702 1703
	amdgpu_fill_reset_magic(adev);

A
Alex Deucher 已提交
1704 1705 1706 1707 1708 1709 1710
	return 0;
}

static int amdgpu_fini(struct amdgpu_device *adev)
{
	int i, r;

1711 1712
	/* need to disable SMC first */
	for (i = 0; i < adev->num_ip_blocks; i++) {
1713
		if (!adev->ip_blocks[i].status.hw)
1714
			continue;
1715
		if (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_SMC) {
1716
			/* ungate blocks before hw fini so that we can shutdown the blocks safely */
1717 1718
			r = adev->ip_blocks[i].version->funcs->set_clockgating_state((void *)adev,
										     AMD_CG_STATE_UNGATE);
1719 1720
			if (r) {
				DRM_ERROR("set_clockgating_state(ungate) of IP block <%s> failed %d\n",
1721
					  adev->ip_blocks[i].version->funcs->name, r);
1722 1723
				return r;
			}
1724
			r = adev->ip_blocks[i].version->funcs->hw_fini((void *)adev);
1725 1726 1727
			/* XXX handle errors */
			if (r) {
				DRM_DEBUG("hw_fini of IP block <%s> failed %d\n",
1728
					  adev->ip_blocks[i].version->funcs->name, r);
1729
			}
1730
			adev->ip_blocks[i].status.hw = false;
1731 1732 1733 1734
			break;
		}
	}

A
Alex Deucher 已提交
1735
	for (i = adev->num_ip_blocks - 1; i >= 0; i--) {
1736
		if (!adev->ip_blocks[i].status.hw)
A
Alex Deucher 已提交
1737
			continue;
1738
		if (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_GMC) {
A
Alex Deucher 已提交
1739 1740 1741
			amdgpu_wb_fini(adev);
			amdgpu_vram_scratch_fini(adev);
		}
1742 1743 1744 1745 1746 1747 1748 1749 1750 1751 1752

		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;
			}
1753
		}
1754

1755
		r = adev->ip_blocks[i].version->funcs->hw_fini((void *)adev);
A
Alex Deucher 已提交
1756
		/* XXX handle errors */
1757
		if (r) {
1758 1759
			DRM_DEBUG("hw_fini of IP block <%s> failed %d\n",
				  adev->ip_blocks[i].version->funcs->name, r);
1760
		}
1761

1762
		adev->ip_blocks[i].status.hw = false;
A
Alex Deucher 已提交
1763 1764 1765
	}

	for (i = adev->num_ip_blocks - 1; i >= 0; i--) {
1766
		if (!adev->ip_blocks[i].status.sw)
A
Alex Deucher 已提交
1767
			continue;
1768
		r = adev->ip_blocks[i].version->funcs->sw_fini((void *)adev);
A
Alex Deucher 已提交
1769
		/* XXX handle errors */
1770
		if (r) {
1771 1772
			DRM_DEBUG("sw_fini of IP block <%s> failed %d\n",
				  adev->ip_blocks[i].version->funcs->name, r);
1773
		}
1774 1775
		adev->ip_blocks[i].status.sw = false;
		adev->ip_blocks[i].status.valid = false;
A
Alex Deucher 已提交
1776 1777
	}

M
Monk Liu 已提交
1778
	for (i = adev->num_ip_blocks - 1; i >= 0; i--) {
1779
		if (!adev->ip_blocks[i].status.late_initialized)
1780
			continue;
1781 1782 1783
		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
1784 1785
	}

1786
	if (amdgpu_sriov_vf(adev)) {
M
Monk Liu 已提交
1787
		amdgpu_bo_free_kernel(&adev->virt.csa_obj, &adev->virt.csa_vmid0_addr, NULL);
1788 1789
		amdgpu_virt_release_full_gpu(adev, false);
	}
M
Monk Liu 已提交
1790

A
Alex Deucher 已提交
1791 1792 1793
	return 0;
}

1794
int amdgpu_suspend(struct amdgpu_device *adev)
A
Alex Deucher 已提交
1795 1796 1797
{
	int i, r;

1798 1799 1800
	if (amdgpu_sriov_vf(adev))
		amdgpu_virt_request_full_gpu(adev, false);

1801 1802 1803 1804 1805 1806 1807
	/* ungate SMC block first */
	r = amdgpu_set_clockgating_state(adev, AMD_IP_BLOCK_TYPE_SMC,
					 AMD_CG_STATE_UNGATE);
	if (r) {
		DRM_ERROR("set_clockgating_state(ungate) SMC failed %d\n",r);
	}

A
Alex Deucher 已提交
1808
	for (i = adev->num_ip_blocks - 1; i >= 0; i--) {
1809
		if (!adev->ip_blocks[i].status.valid)
A
Alex Deucher 已提交
1810 1811
			continue;
		/* ungate blocks so that suspend can properly shut them down */
1812
		if (i != AMD_IP_BLOCK_TYPE_SMC) {
1813 1814
			r = adev->ip_blocks[i].version->funcs->set_clockgating_state((void *)adev,
										     AMD_CG_STATE_UNGATE);
1815
			if (r) {
1816 1817
				DRM_ERROR("set_clockgating_state(ungate) of IP block <%s> failed %d\n",
					  adev->ip_blocks[i].version->funcs->name, r);
1818
			}
1819
		}
A
Alex Deucher 已提交
1820
		/* XXX handle errors */
1821
		r = adev->ip_blocks[i].version->funcs->suspend(adev);
A
Alex Deucher 已提交
1822
		/* XXX handle errors */
1823
		if (r) {
1824 1825
			DRM_ERROR("suspend of IP block <%s> failed %d\n",
				  adev->ip_blocks[i].version->funcs->name, r);
1826
		}
A
Alex Deucher 已提交
1827 1828
	}

1829 1830 1831
	if (amdgpu_sriov_vf(adev))
		amdgpu_virt_release_full_gpu(adev, false);

A
Alex Deucher 已提交
1832 1833 1834
	return 0;
}

1835
static int amdgpu_sriov_reinit_early(struct amdgpu_device *adev)
1836 1837 1838
{
	int i, r;

1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859
	static enum amd_ip_block_type ip_order[] = {
		AMD_IP_BLOCK_TYPE_GMC,
		AMD_IP_BLOCK_TYPE_COMMON,
		AMD_IP_BLOCK_TYPE_GFXHUB,
		AMD_IP_BLOCK_TYPE_MMHUB,
		AMD_IP_BLOCK_TYPE_IH,
	};

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

		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");
1860 1861 1862 1863 1864 1865
		}
	}

	return 0;
}

1866
static int amdgpu_sriov_reinit_late(struct amdgpu_device *adev)
1867 1868 1869
{
	int i, r;

1870 1871 1872 1873 1874 1875 1876
	static enum amd_ip_block_type ip_order[] = {
		AMD_IP_BLOCK_TYPE_SMC,
		AMD_IP_BLOCK_TYPE_DCE,
		AMD_IP_BLOCK_TYPE_GFX,
		AMD_IP_BLOCK_TYPE_SDMA,
		AMD_IP_BLOCK_TYPE_VCE,
	};
1877

1878 1879 1880
	for (i = 0; i < ARRAY_SIZE(ip_order); i++) {
		int j;
		struct amdgpu_ip_block *block;
1881

1882 1883 1884 1885 1886 1887 1888 1889 1890
		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");
1891 1892 1893 1894 1895 1896
		}
	}

	return 0;
}

1897
static int amdgpu_resume_phase1(struct amdgpu_device *adev)
A
Alex Deucher 已提交
1898 1899 1900 1901
{
	int i, r;

	for (i = 0; i < adev->num_ip_blocks; i++) {
1902
		if (!adev->ip_blocks[i].status.valid)
A
Alex Deucher 已提交
1903
			continue;
1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930
		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) {
			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;
			}
		}
	}

	return 0;
}

static int amdgpu_resume_phase2(struct amdgpu_device *adev)
{
	int i, r;

	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 ||
				adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_IH )
			continue;
1931
		r = adev->ip_blocks[i].version->funcs->resume(adev);
1932
		if (r) {
1933 1934
			DRM_ERROR("resume of IP block <%s> failed %d\n",
				  adev->ip_blocks[i].version->funcs->name, r);
A
Alex Deucher 已提交
1935
			return r;
1936
		}
A
Alex Deucher 已提交
1937 1938 1939 1940 1941
	}

	return 0;
}

1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953
static int amdgpu_resume(struct amdgpu_device *adev)
{
	int r;

	r = amdgpu_resume_phase1(adev);
	if (r)
		return r;
	r = amdgpu_resume_phase2(adev);

	return r;
}

1954
static void amdgpu_device_detect_sriov_bios(struct amdgpu_device *adev)
1955
{
1956 1957 1958 1959 1960 1961 1962
	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;
	}
1963 1964
}

A
Alex Deucher 已提交
1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983
/**
 * 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;
1984
	u32 max_MBps;
A
Alex Deucher 已提交
1985 1986 1987 1988 1989 1990

	adev->shutdown = false;
	adev->dev = &pdev->dev;
	adev->ddev = ddev;
	adev->pdev = pdev;
	adev->flags = flags;
1991
	adev->asic_type = flags & AMD_ASIC_MASK;
A
Alex Deucher 已提交
1992 1993 1994 1995 1996 1997 1998
	adev->usec_timeout = AMDGPU_MAX_USEC_TIMEOUT;
	adev->mc.gtt_size = 512 * 1024 * 1024;
	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;
1999
	adev->vm_manager.vm_pte_num_rings = 0;
A
Alex Deucher 已提交
2000
	adev->gart.gart_funcs = NULL;
2001
	adev->fence_context = dma_fence_context_alloc(AMDGPU_MAX_RINGS);
A
Alex Deucher 已提交
2002 2003 2004 2005 2006

	adev->smc_rreg = &amdgpu_invalid_rreg;
	adev->smc_wreg = &amdgpu_invalid_wreg;
	adev->pcie_rreg = &amdgpu_invalid_rreg;
	adev->pcie_wreg = &amdgpu_invalid_wreg;
2007 2008
	adev->pciep_rreg = &amdgpu_invalid_rreg;
	adev->pciep_wreg = &amdgpu_invalid_wreg;
A
Alex Deucher 已提交
2009 2010 2011 2012
	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;
2013 2014
	adev->gc_cac_rreg = &amdgpu_invalid_rreg;
	adev->gc_cac_wreg = &amdgpu_invalid_wreg;
A
Alex Deucher 已提交
2015 2016 2017
	adev->audio_endpt_rreg = &amdgpu_block_invalid_rreg;
	adev->audio_endpt_wreg = &amdgpu_block_invalid_wreg;

2018

2019 2020 2021
	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
2022 2023 2024 2025

	/* mutex initialization are all done here so we
	 * can recall function without having locking issues */
	atomic_set(&adev->irq.ih.lock, 0);
2026
	mutex_init(&adev->firmware.mutex);
A
Alex Deucher 已提交
2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042
	mutex_init(&adev->pm.mutex);
	mutex_init(&adev->gfx.gpu_clock_mutex);
	mutex_init(&adev->srbm_mutex);
	mutex_init(&adev->grbm_idx_mutex);
	mutex_init(&adev->mn_lock);
	hash_init(adev->mn_hash);

	amdgpu_check_arguments(adev);

	/* Registers mapping */
	/* TODO: block userspace mapping of io register */
	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);
2043
	spin_lock_init(&adev->gc_cac_idx_lock);
A
Alex Deucher 已提交
2044
	spin_lock_init(&adev->audio_endpt_idx_lock);
2045
	spin_lock_init(&adev->mm_stats.lock);
A
Alex Deucher 已提交
2046

2047 2048 2049
	INIT_LIST_HEAD(&adev->shadow_list);
	mutex_init(&adev->shadow_list_lock);

2050 2051 2052
	INIT_LIST_HEAD(&adev->gtt_list);
	spin_lock_init(&adev->gtt_list_lock);

2053 2054 2055 2056 2057 2058 2059
	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
2060 2061 2062 2063 2064 2065 2066 2067

	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);

2068 2069 2070
	if (adev->asic_type >= CHIP_BONAIRE)
		/* doorbell bar mapping */
		amdgpu_doorbell_init(adev);
A
Alex Deucher 已提交
2071 2072 2073 2074 2075 2076 2077 2078 2079 2080

	/* 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)
2081
		DRM_INFO("PCI I/O BAR is not found.\n");
A
Alex Deucher 已提交
2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 2092 2093 2094

	/* early init functions */
	r = amdgpu_early_init(adev);
	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 */
	vga_client_register(adev->pdev, adev, NULL, amdgpu_vga_set_decode);

	if (amdgpu_runtime_pm == 1)
		runtime = true;
2095
	if (amdgpu_device_is_px(ddev))
A
Alex Deucher 已提交
2096
		runtime = true;
2097 2098 2099
	if (!pci_is_thunderbolt_attached(adev->pdev))
		vga_switcheroo_register_client(adev->pdev,
					       &amdgpu_switcheroo_ops, runtime);
A
Alex Deucher 已提交
2100 2101 2102 2103
	if (runtime)
		vga_switcheroo_init_domain_pm_ops(adev->dev, &adev->vga_pm_domain);

	/* Read BIOS */
2104 2105 2106 2107
	if (!amdgpu_get_bios(adev)) {
		r = -EINVAL;
		goto failed;
	}
2108

A
Alex Deucher 已提交
2109
	r = amdgpu_atombios_init(adev);
2110 2111
	if (r) {
		dev_err(adev->dev, "amdgpu_atombios_init failed\n");
2112
		goto failed;
2113
	}
A
Alex Deucher 已提交
2114

2115 2116
	/* detect if we are with an SRIOV vbios */
	amdgpu_device_detect_sriov_bios(adev);
2117

A
Alex Deucher 已提交
2118
	/* Post card if necessary */
2119
	if (amdgpu_vpost_needed(adev)) {
A
Alex Deucher 已提交
2120
		if (!adev->bios) {
2121
			dev_err(adev->dev, "no vBIOS found\n");
2122 2123
			r = -EINVAL;
			goto failed;
A
Alex Deucher 已提交
2124
		}
2125
		DRM_INFO("GPU posting now...\n");
2126 2127 2128 2129 2130 2131 2132
		r = amdgpu_atom_asic_init(adev->mode_info.atom_context);
		if (r) {
			dev_err(adev->dev, "gpu post error!\n");
			goto failed;
		}
	} else {
		DRM_INFO("GPU post is not needed\n");
A
Alex Deucher 已提交
2133 2134
	}

2135 2136 2137 2138 2139 2140 2141 2142 2143
	if (!adev->is_atom_fw) {
		/* Initialize clocks */
		r = amdgpu_atombios_get_clock_info(adev);
		if (r) {
			dev_err(adev->dev, "amdgpu_atombios_get_clock_info failed\n");
			return r;
		}
		/* init i2c buses */
		amdgpu_atombios_i2c_init(adev);
2144
	}
A
Alex Deucher 已提交
2145 2146 2147

	/* Fence driver */
	r = amdgpu_fence_driver_init(adev);
2148 2149
	if (r) {
		dev_err(adev->dev, "amdgpu_fence_driver_init failed\n");
2150
		goto failed;
2151
	}
A
Alex Deucher 已提交
2152 2153 2154 2155 2156 2157

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

	r = amdgpu_init(adev);
	if (r) {
2158
		dev_err(adev->dev, "amdgpu_init failed\n");
A
Alex Deucher 已提交
2159
		amdgpu_fini(adev);
2160
		goto failed;
A
Alex Deucher 已提交
2161 2162 2163 2164
	}

	adev->accel_working = true;

2165 2166 2167 2168 2169 2170 2171 2172
	/* 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
2173 2174 2175
	r = amdgpu_ib_pool_init(adev);
	if (r) {
		dev_err(adev->dev, "IB initialization failed (%d).\n", r);
2176
		goto failed;
A
Alex Deucher 已提交
2177 2178 2179 2180 2181 2182
	}

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

2183 2184
	amdgpu_fbdev_init(adev);

A
Alex Deucher 已提交
2185
	r = amdgpu_gem_debugfs_init(adev);
M
Monk Liu 已提交
2186
	if (r)
A
Alex Deucher 已提交
2187 2188 2189
		DRM_ERROR("registering gem debugfs failed (%d).\n", r);

	r = amdgpu_debugfs_regs_init(adev);
M
Monk Liu 已提交
2190
	if (r)
A
Alex Deucher 已提交
2191 2192
		DRM_ERROR("registering register debugfs failed (%d).\n", r);

2193
	r = amdgpu_debugfs_firmware_init(adev);
M
Monk Liu 已提交
2194
	if (r)
2195 2196
		DRM_ERROR("registering firmware debugfs failed (%d).\n", r);

A
Alex Deucher 已提交
2197 2198 2199 2200 2201 2202 2203 2204 2205 2206 2207 2208 2209 2210 2211 2212 2213
	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.
	 */
	r = amdgpu_late_init(adev);
2214 2215
	if (r) {
		dev_err(adev->dev, "amdgpu_late_init failed\n");
2216
		goto failed;
2217
	}
A
Alex Deucher 已提交
2218 2219

	return 0;
2220 2221 2222 2223 2224

failed:
	if (runtime)
		vga_switcheroo_fini_domain_pm_ops(adev->dev);
	return r;
A
Alex Deucher 已提交
2225 2226 2227 2228 2229 2230 2231 2232 2233 2234 2235 2236 2237 2238 2239 2240
}

/**
 * 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;
2241 2242
	if (adev->mode_info.mode_config_initialized)
		drm_crtc_force_disable_all(adev->ddev);
A
Alex Deucher 已提交
2243 2244 2245 2246 2247 2248 2249 2250 2251 2252 2253 2254
	/* evict vram memory */
	amdgpu_bo_evict_vram(adev);
	amdgpu_ib_pool_fini(adev);
	amdgpu_fence_driver_fini(adev);
	amdgpu_fbdev_fini(adev);
	r = amdgpu_fini(adev);
	adev->accel_working = false;
	/* free i2c buses */
	amdgpu_i2c_fini(adev);
	amdgpu_atombios_fini(adev);
	kfree(adev->bios);
	adev->bios = NULL;
2255 2256
	if (!pci_is_thunderbolt_attached(adev->pdev))
		vga_switcheroo_unregister_client(adev->pdev);
2257 2258
	if (adev->flags & AMD_IS_PX)
		vga_switcheroo_fini_domain_pm_ops(adev->dev);
A
Alex Deucher 已提交
2259 2260 2261 2262 2263 2264
	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;
2265 2266
	if (adev->asic_type >= CHIP_BONAIRE)
		amdgpu_doorbell_fini(adev);
A
Alex Deucher 已提交
2267 2268 2269 2270 2271 2272 2273 2274
	amdgpu_debugfs_regs_cleanup(adev);
}


/*
 * Suspend & resume.
 */
/**
2275
 * amdgpu_device_suspend - initiate device suspend
A
Alex Deucher 已提交
2276 2277 2278 2279 2280 2281 2282 2283
 *
 * @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.
 */
2284
int amdgpu_device_suspend(struct drm_device *dev, bool suspend, bool fbcon)
A
Alex Deucher 已提交
2285 2286 2287 2288
{
	struct amdgpu_device *adev;
	struct drm_crtc *crtc;
	struct drm_connector *connector;
2289
	int r;
A
Alex Deucher 已提交
2290 2291 2292 2293 2294 2295 2296 2297 2298 2299 2300 2301 2302

	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);

	/* turn off display hw */
2303
	drm_modeset_lock_all(dev);
A
Alex Deucher 已提交
2304 2305 2306
	list_for_each_entry(connector, &dev->mode_config.connector_list, head) {
		drm_helper_connector_dpms(connector, DRM_MODE_DPMS_OFF);
	}
2307
	drm_modeset_unlock_all(dev);
A
Alex Deucher 已提交
2308

2309
	/* unpin the front buffers and cursors */
A
Alex Deucher 已提交
2310
	list_for_each_entry(crtc, &dev->mode_config.crtc_list, head) {
2311
		struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc);
A
Alex Deucher 已提交
2312 2313 2314
		struct amdgpu_framebuffer *rfb = to_amdgpu_framebuffer(crtc->primary->fb);
		struct amdgpu_bo *robj;

2315 2316
		if (amdgpu_crtc->cursor_bo) {
			struct amdgpu_bo *aobj = gem_to_amdgpu_bo(amdgpu_crtc->cursor_bo);
2317
			r = amdgpu_bo_reserve(aobj, true);
2318 2319 2320 2321 2322 2323
			if (r == 0) {
				amdgpu_bo_unpin(aobj);
				amdgpu_bo_unreserve(aobj);
			}
		}

A
Alex Deucher 已提交
2324 2325 2326 2327 2328 2329
		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)) {
2330
			r = amdgpu_bo_reserve(robj, true);
A
Alex Deucher 已提交
2331 2332 2333 2334 2335 2336 2337 2338 2339
			if (r == 0) {
				amdgpu_bo_unpin(robj);
				amdgpu_bo_unreserve(robj);
			}
		}
	}
	/* evict vram memory */
	amdgpu_bo_evict_vram(adev);

2340
	amdgpu_fence_driver_suspend(adev);
A
Alex Deucher 已提交
2341 2342 2343

	r = amdgpu_suspend(adev);

2344 2345 2346 2347
	/* 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
2348 2349
	amdgpu_bo_evict_vram(adev);

2350 2351 2352 2353
	if (adev->is_atom_fw)
		amdgpu_atomfirmware_scratch_regs_save(adev);
	else
		amdgpu_atombios_scratch_regs_save(adev);
A
Alex Deucher 已提交
2354 2355 2356 2357 2358
	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
2359 2360 2361 2362
	} else {
		r = amdgpu_asic_reset(adev);
		if (r)
			DRM_ERROR("amdgpu asic reset failed\n");
A
Alex Deucher 已提交
2363 2364 2365 2366 2367 2368 2369 2370 2371 2372 2373
	}

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

/**
2374
 * amdgpu_device_resume - initiate device resume
A
Alex Deucher 已提交
2375 2376 2377 2378 2379 2380 2381
 *
 * @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.
 */
2382
int amdgpu_device_resume(struct drm_device *dev, bool resume, bool fbcon)
A
Alex Deucher 已提交
2383 2384 2385
{
	struct drm_connector *connector;
	struct amdgpu_device *adev = dev->dev_private;
2386
	struct drm_crtc *crtc;
2387
	int r = 0;
A
Alex Deucher 已提交
2388 2389 2390 2391

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

J
jimqu 已提交
2392
	if (fbcon)
A
Alex Deucher 已提交
2393
		console_lock();
J
jimqu 已提交
2394

A
Alex Deucher 已提交
2395 2396 2397
	if (resume) {
		pci_set_power_state(dev->pdev, PCI_D0);
		pci_restore_state(dev->pdev);
J
jimqu 已提交
2398
		r = pci_enable_device(dev->pdev);
2399 2400
		if (r)
			goto unlock;
A
Alex Deucher 已提交
2401
	}
2402 2403 2404 2405
	if (adev->is_atom_fw)
		amdgpu_atomfirmware_scratch_regs_restore(adev);
	else
		amdgpu_atombios_scratch_regs_restore(adev);
A
Alex Deucher 已提交
2406 2407

	/* post card */
2408
	if (amdgpu_need_post(adev)) {
J
jimqu 已提交
2409 2410 2411 2412
		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
2413 2414

	r = amdgpu_resume(adev);
2415
	if (r) {
F
Flora Cui 已提交
2416
		DRM_ERROR("amdgpu_resume failed (%d).\n", r);
2417
		goto unlock;
2418
	}
2419 2420
	amdgpu_fence_driver_resume(adev);

F
Flora Cui 已提交
2421 2422 2423 2424 2425
	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
2426 2427

	r = amdgpu_late_init(adev);
2428 2429
	if (r)
		goto unlock;
A
Alex Deucher 已提交
2430

2431 2432 2433 2434 2435 2436
	/* 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);
2437
			r = amdgpu_bo_reserve(aobj, true);
2438 2439 2440 2441 2442 2443 2444 2445 2446 2447 2448
			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);
			}
		}
	}

A
Alex Deucher 已提交
2449 2450 2451 2452
	/* blat the mode back in */
	if (fbcon) {
		drm_helper_resume_force_mode(dev);
		/* turn on display hw */
2453
		drm_modeset_lock_all(dev);
A
Alex Deucher 已提交
2454 2455 2456
		list_for_each_entry(connector, &dev->mode_config.connector_list, head) {
			drm_helper_connector_dpms(connector, DRM_MODE_DPMS_ON);
		}
2457
		drm_modeset_unlock_all(dev);
A
Alex Deucher 已提交
2458 2459 2460
	}

	drm_kms_helper_poll_enable(dev);
2461 2462 2463 2464 2465 2466 2467 2468 2469 2470 2471 2472 2473

	/*
	 * 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
2474
	drm_helper_hpd_irq_event(dev);
2475 2476 2477
#ifdef CONFIG_PM
	dev->dev->power.disable_depth--;
#endif
A
Alex Deucher 已提交
2478

2479
	if (fbcon)
A
Alex Deucher 已提交
2480
		amdgpu_fbdev_set_suspend(adev, 0);
2481 2482 2483

unlock:
	if (fbcon)
A
Alex Deucher 已提交
2484 2485
		console_unlock();

2486
	return r;
A
Alex Deucher 已提交
2487 2488
}

2489 2490 2491 2492 2493 2494
static bool amdgpu_check_soft_reset(struct amdgpu_device *adev)
{
	int i;
	bool asic_hang = false;

	for (i = 0; i < adev->num_ip_blocks; i++) {
2495
		if (!adev->ip_blocks[i].status.valid)
2496
			continue;
2497 2498 2499 2500 2501
		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);
2502 2503 2504 2505 2506 2507
			asic_hang = true;
		}
	}
	return asic_hang;
}

2508
static int amdgpu_pre_soft_reset(struct amdgpu_device *adev)
2509 2510 2511 2512
{
	int i, r = 0;

	for (i = 0; i < adev->num_ip_blocks; i++) {
2513
		if (!adev->ip_blocks[i].status.valid)
2514
			continue;
2515 2516 2517
		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);
2518 2519 2520 2521 2522 2523 2524 2525
			if (r)
				return r;
		}
	}

	return 0;
}

2526 2527
static bool amdgpu_need_full_reset(struct amdgpu_device *adev)
{
2528 2529 2530
	int i;

	for (i = 0; i < adev->num_ip_blocks; i++) {
2531
		if (!adev->ip_blocks[i].status.valid)
2532
			continue;
2533 2534 2535 2536 2537
		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) ||
		    (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_DCE)) {
			if (adev->ip_blocks[i].status.hang) {
2538 2539 2540 2541
				DRM_INFO("Some block need full reset!\n");
				return true;
			}
		}
2542 2543 2544 2545 2546 2547 2548 2549 2550
	}
	return false;
}

static int amdgpu_soft_reset(struct amdgpu_device *adev)
{
	int i, r = 0;

	for (i = 0; i < adev->num_ip_blocks; i++) {
2551
		if (!adev->ip_blocks[i].status.valid)
2552
			continue;
2553 2554 2555
		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);
2556 2557 2558 2559 2560 2561 2562 2563 2564 2565 2566 2567 2568
			if (r)
				return r;
		}
	}

	return 0;
}

static int amdgpu_post_soft_reset(struct amdgpu_device *adev)
{
	int i, r = 0;

	for (i = 0; i < adev->num_ip_blocks; i++) {
2569
		if (!adev->ip_blocks[i].status.valid)
2570
			continue;
2571 2572 2573
		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);
2574 2575 2576 2577 2578 2579 2580
		if (r)
			return r;
	}

	return 0;
}

2581 2582 2583 2584 2585 2586 2587 2588
bool amdgpu_need_backup(struct amdgpu_device *adev)
{
	if (adev->flags & AMD_IS_APU)
		return false;

	return amdgpu_lockup_timeout > 0 ? true : false;
}

2589 2590 2591
static int amdgpu_recover_vram_from_shadow(struct amdgpu_device *adev,
					   struct amdgpu_ring *ring,
					   struct amdgpu_bo *bo,
2592
					   struct dma_fence **fence)
2593 2594 2595 2596
{
	uint32_t domain;
	int r;

R
Roger.He 已提交
2597 2598 2599
	if (!bo->shadow)
		return 0;

2600
	r = amdgpu_bo_reserve(bo, true);
R
Roger.He 已提交
2601 2602 2603 2604 2605
	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) {
2606 2607 2608 2609 2610 2611 2612 2613 2614 2615 2616 2617
		r = amdgpu_bo_validate(bo->shadow);
		if (r) {
			DRM_ERROR("bo validate failed!\n");
			goto err;
		}

		r = amdgpu_ttm_bind(&bo->shadow->tbo, &bo->shadow->tbo.mem);
		if (r) {
			DRM_ERROR("%p bind failed\n", bo->shadow);
			goto err;
		}

R
Roger.He 已提交
2618
		r = amdgpu_bo_restore_from_shadow(adev, ring, bo,
2619
						 NULL, fence, true);
R
Roger.He 已提交
2620 2621 2622 2623 2624
		if (r) {
			DRM_ERROR("recover page table failed!\n");
			goto err;
		}
	}
2625
err:
R
Roger.He 已提交
2626 2627
	amdgpu_bo_unreserve(bo);
	return r;
2628 2629
}

2630 2631 2632 2633
/**
 * amdgpu_sriov_gpu_reset - reset the asic
 *
 * @adev: amdgpu device pointer
2634
 * @job: which job trigger hang
2635 2636 2637 2638 2639
 *
 * Attempt the reset the GPU if it has hung (all asics).
 * for SRIOV case.
 * Returns 0 for success or an error on failure.
 */
2640
int amdgpu_sriov_gpu_reset(struct amdgpu_device *adev, struct amdgpu_job *job)
2641
{
2642
	int i, j, r = 0;
2643 2644 2645 2646 2647
	int resched;
	struct amdgpu_bo *bo, *tmp;
	struct amdgpu_ring *ring;
	struct dma_fence *fence = NULL, *next = NULL;

M
Monk Liu 已提交
2648
	mutex_lock(&adev->virt.lock_reset);
2649
	atomic_inc(&adev->gpu_reset_counter);
2650
	adev->gfx.in_reset = true;
2651 2652 2653 2654

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

2655 2656
	/* we start from the ring trigger GPU hang */
	j = job ? job->ring->idx : 0;
2657

2658 2659 2660
	/* block scheduler */
	for (i = j; i < j + AMDGPU_MAX_RINGS; ++i) {
		ring = adev->rings[i % AMDGPU_MAX_RINGS];
2661 2662 2663 2664
		if (!ring || !ring->sched.thread)
			continue;

		kthread_park(ring->sched.thread);
2665 2666 2667 2668

		if (job && j != i)
			continue;

2669
		/* here give the last chance to check if job removed from mirror-list
2670
		 * since we already pay some time on kthread_park */
2671
		if (job && list_empty(&job->base.node)) {
2672 2673 2674 2675 2676 2677 2678 2679
			kthread_unpark(ring->sched.thread);
			goto give_up_reset;
		}

		if (amd_sched_invalidate_job(&job->base, amdgpu_job_hang_limit))
			amd_sched_job_kickout(&job->base);

		/* only do job_reset on the hang ring if @job not NULL */
2680 2681
		amd_sched_hw_job_reset(&ring->sched);

2682 2683 2684
		/* after all hw jobs are reset, hw fence is meaningless, so force_completion */
		amdgpu_fence_driver_force_completion_ring(ring);
	}
2685 2686

	/* request to take full control of GPU before re-initialization  */
2687
	if (job)
2688 2689 2690 2691 2692 2693
		amdgpu_virt_reset_gpu(adev);
	else
		amdgpu_virt_request_full_gpu(adev, true);


	/* Resume IP prior to SMC */
2694
	amdgpu_sriov_reinit_early(adev);
2695 2696 2697 2698 2699

	/* we need recover gart prior to run SMC/CP/SDMA resume */
	amdgpu_ttm_recover_gart(adev);

	/* now we are okay to resume SMC/CP/SDMA */
2700
	amdgpu_sriov_reinit_late(adev);
2701 2702 2703 2704 2705 2706 2707 2708 2709 2710 2711 2712 2713 2714

	amdgpu_irq_gpu_reset_resume_helper(adev);

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

	/* release full control of GPU after ib test */
	amdgpu_virt_release_full_gpu(adev, true);

	DRM_INFO("recover vram bo from shadow\n");

	ring = adev->mman.buffer_funcs_ring;
	mutex_lock(&adev->shadow_list_lock);
	list_for_each_entry_safe(bo, tmp, &adev->shadow_list, shadow_list) {
M
Monk Liu 已提交
2715
		next = NULL;
2716 2717 2718 2719 2720 2721 2722 2723 2724 2725 2726 2727 2728 2729 2730 2731 2732 2733 2734 2735 2736
		amdgpu_recover_vram_from_shadow(adev, ring, bo, &next);
		if (fence) {
			r = dma_fence_wait(fence, false);
			if (r) {
				WARN(r, "recovery from shadow isn't completed\n");
				break;
			}
		}

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

	if (fence) {
		r = dma_fence_wait(fence, false);
		if (r)
			WARN(r, "recovery from shadow isn't completed\n");
	}
	dma_fence_put(fence);

2737 2738
	for (i = j; i < j + AMDGPU_MAX_RINGS; ++i) {
		ring = adev->rings[i % AMDGPU_MAX_RINGS];
2739 2740 2741
		if (!ring || !ring->sched.thread)
			continue;

2742 2743 2744 2745 2746
		if (job && j != i) {
			kthread_unpark(ring->sched.thread);
			continue;
		}

2747 2748 2749 2750 2751
		amd_sched_job_recovery(&ring->sched);
		kthread_unpark(ring->sched.thread);
	}

	drm_helper_resume_force_mode(adev->ddev);
2752
give_up_reset:
2753 2754 2755 2756
	ttm_bo_unlock_delayed_workqueue(&adev->mman.bdev, resched);
	if (r) {
		/* bad news, how to tell it to userspace ? */
		dev_info(adev->dev, "GPU reset failed\n");
2757 2758
	} else {
		dev_info(adev->dev, "GPU reset successed!\n");
2759 2760
	}

2761
	adev->gfx.in_reset = false;
M
Monk Liu 已提交
2762
	mutex_unlock(&adev->virt.lock_reset);
2763 2764 2765
	return r;
}

A
Alex Deucher 已提交
2766 2767 2768 2769 2770 2771 2772 2773 2774 2775 2776 2777
/**
 * amdgpu_gpu_reset - reset the asic
 *
 * @adev: amdgpu device pointer
 *
 * Attempt the reset the GPU if it has hung (all asics).
 * Returns 0 for success or an error on failure.
 */
int amdgpu_gpu_reset(struct amdgpu_device *adev)
{
	int i, r;
	int resched;
2778
	bool need_full_reset, vram_lost = false;
A
Alex Deucher 已提交
2779

2780 2781 2782 2783
	if (!amdgpu_check_soft_reset(adev)) {
		DRM_INFO("No hardware hang detected. Did some blocks stall?\n");
		return 0;
	}
A
Alex Deucher 已提交
2784

2785
	atomic_inc(&adev->gpu_reset_counter);
A
Alex Deucher 已提交
2786

2787 2788 2789
	/* block TTM */
	resched = ttm_bo_lock_delayed_workqueue(&adev->mman.bdev);

2790 2791 2792 2793
	/* block scheduler */
	for (i = 0; i < AMDGPU_MAX_RINGS; ++i) {
		struct amdgpu_ring *ring = adev->rings[i];

C
Chunming Zhou 已提交
2794
		if (!ring || !ring->sched.thread)
2795 2796
			continue;
		kthread_park(ring->sched.thread);
2797
		amd_sched_hw_job_reset(&ring->sched);
2798
	}
2799 2800
	/* after all hw jobs are reset, hw fence is meaningless, so force_completion */
	amdgpu_fence_driver_force_completion(adev);
A
Alex Deucher 已提交
2801

2802
	need_full_reset = amdgpu_need_full_reset(adev);
A
Alex Deucher 已提交
2803

2804 2805 2806 2807 2808 2809 2810 2811
	if (!need_full_reset) {
		amdgpu_pre_soft_reset(adev);
		r = amdgpu_soft_reset(adev);
		amdgpu_post_soft_reset(adev);
		if (r || amdgpu_check_soft_reset(adev)) {
			DRM_INFO("soft reset failed, will fallback to full reset!\n");
			need_full_reset = true;
		}
2812 2813
	}

2814 2815
	if (need_full_reset) {
		r = amdgpu_suspend(adev);
2816

2817 2818 2819 2820 2821 2822 2823
retry:
		/* Disable fb access */
		if (adev->mode_info.num_crtc) {
			struct amdgpu_mode_mc_save save;
			amdgpu_display_stop_mc_access(adev, &save);
			amdgpu_wait_for_idle(adev, AMD_IP_BLOCK_TYPE_GMC);
		}
2824 2825 2826 2827
		if (adev->is_atom_fw)
			amdgpu_atomfirmware_scratch_regs_save(adev);
		else
			amdgpu_atombios_scratch_regs_save(adev);
2828
		r = amdgpu_asic_reset(adev);
2829 2830 2831 2832
		if (adev->is_atom_fw)
			amdgpu_atomfirmware_scratch_regs_restore(adev);
		else
			amdgpu_atombios_scratch_regs_restore(adev);
2833 2834 2835 2836 2837
		/* post card */
		amdgpu_atom_asic_init(adev->mode_info.atom_context);

		if (!r) {
			dev_info(adev->dev, "GPU reset succeeded, trying to resume\n");
2838 2839 2840
			r = amdgpu_resume_phase1(adev);
			if (r)
				goto out;
2841
			vram_lost = amdgpu_check_vram_lost(adev);
2842
			if (vram_lost) {
2843
				DRM_ERROR("VRAM is lost!\n");
2844 2845
				atomic_inc(&adev->vram_lost_counter);
			}
2846 2847 2848 2849 2850 2851
			r = amdgpu_ttm_recover_gart(adev);
			if (r)
				goto out;
			r = amdgpu_resume_phase2(adev);
			if (r)
				goto out;
2852 2853
			if (vram_lost)
				amdgpu_fill_reset_magic(adev);
2854
		}
A
Alex Deucher 已提交
2855
	}
2856
out:
A
Alex Deucher 已提交
2857
	if (!r) {
2858
		amdgpu_irq_gpu_reset_resume_helper(adev);
2859 2860 2861
		r = amdgpu_ib_ring_tests(adev);
		if (r) {
			dev_err(adev->dev, "ib ring test failed (%d).\n", r);
2862
			r = amdgpu_suspend(adev);
2863
			need_full_reset = true;
2864
			goto retry;
2865
		}
2866 2867 2868 2869 2870 2871 2872
		/**
		 * recovery vm page tables, since we cannot depend on VRAM is
		 * consistent after gpu full reset.
		 */
		if (need_full_reset && amdgpu_need_backup(adev)) {
			struct amdgpu_ring *ring = adev->mman.buffer_funcs_ring;
			struct amdgpu_bo *bo, *tmp;
2873
			struct dma_fence *fence = NULL, *next = NULL;
2874 2875 2876 2877

			DRM_INFO("recover vram bo from shadow\n");
			mutex_lock(&adev->shadow_list_lock);
			list_for_each_entry_safe(bo, tmp, &adev->shadow_list, shadow_list) {
M
Monk Liu 已提交
2878
				next = NULL;
2879 2880
				amdgpu_recover_vram_from_shadow(adev, ring, bo, &next);
				if (fence) {
2881
					r = dma_fence_wait(fence, false);
2882
					if (r) {
M
Monk Liu 已提交
2883
						WARN(r, "recovery from shadow isn't completed\n");
2884 2885 2886
						break;
					}
				}
2887

2888
				dma_fence_put(fence);
2889 2890 2891 2892
				fence = next;
			}
			mutex_unlock(&adev->shadow_list_lock);
			if (fence) {
2893
				r = dma_fence_wait(fence, false);
2894
				if (r)
M
Monk Liu 已提交
2895
					WARN(r, "recovery from shadow isn't completed\n");
2896
			}
2897
			dma_fence_put(fence);
2898
		}
A
Alex Deucher 已提交
2899 2900
		for (i = 0; i < AMDGPU_MAX_RINGS; ++i) {
			struct amdgpu_ring *ring = adev->rings[i];
C
Chunming Zhou 已提交
2901 2902

			if (!ring || !ring->sched.thread)
A
Alex Deucher 已提交
2903
				continue;
2904

2905
			amd_sched_job_recovery(&ring->sched);
2906
			kthread_unpark(ring->sched.thread);
A
Alex Deucher 已提交
2907 2908
		}
	} else {
2909
		dev_err(adev->dev, "asic resume failed (%d).\n", r);
A
Alex Deucher 已提交
2910
		for (i = 0; i < AMDGPU_MAX_RINGS; ++i) {
C
Chunming Zhou 已提交
2911
			if (adev->rings[i] && adev->rings[i]->sched.thread) {
2912 2913
				kthread_unpark(adev->rings[i]->sched.thread);
			}
A
Alex Deucher 已提交
2914 2915 2916 2917 2918 2919
		}
	}

	drm_helper_resume_force_mode(adev->ddev);

	ttm_bo_unlock_delayed_workqueue(&adev->mman.bdev, resched);
2920
	if (r)
A
Alex Deucher 已提交
2921 2922
		/* bad news, how to tell it to userspace ? */
		dev_info(adev->dev, "GPU reset failed\n");
2923 2924
	else
		dev_info(adev->dev, "GPU reset successed!\n");
A
Alex Deucher 已提交
2925 2926 2927 2928

	return r;
}

2929 2930 2931 2932 2933
void amdgpu_get_pcie_info(struct amdgpu_device *adev)
{
	u32 mask;
	int ret;

2934 2935
	if (amdgpu_pcie_gen_cap)
		adev->pm.pcie_gen_mask = amdgpu_pcie_gen_cap;
2936

2937 2938
	if (amdgpu_pcie_lane_cap)
		adev->pm.pcie_mlw_mask = amdgpu_pcie_lane_cap;
2939

2940 2941 2942 2943 2944 2945
	/* 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;
2946
		return;
2947
	}
2948

2949 2950 2951 2952 2953 2954 2955 2956 2957 2958 2959 2960 2961 2962 2963 2964 2965 2966 2967 2968 2969 2970 2971 2972 2973 2974 2975 2976 2977 2978 2979 2980 2981 2982 2983 2984 2985 2986 2987 2988 2989 2990 2991 2992 2993 2994 2995 2996 2997 2998 2999 3000 3001 3002 3003 3004 3005 3006 3007 3008 3009 3010 3011 3012 3013 3014 3015 3016
	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;
3017 3018 3019
		}
	}
}
A
Alex Deucher 已提交
3020 3021 3022 3023 3024

/*
 * Debugfs
 */
int amdgpu_debugfs_add_files(struct amdgpu_device *adev,
3025
			     const struct drm_info_list *files,
A
Alex Deucher 已提交
3026 3027 3028 3029 3030 3031 3032 3033 3034 3035 3036 3037 3038 3039 3040 3041 3042 3043 3044 3045 3046 3047 3048 3049 3050 3051 3052 3053 3054 3055 3056 3057 3058 3059
			     unsigned nfiles)
{
	unsigned i;

	for (i = 0; i < adev->debugfs_count; i++) {
		if (adev->debugfs[i].files == files) {
			/* Already registered */
			return 0;
		}
	}

	i = adev->debugfs_count + 1;
	if (i > AMDGPU_DEBUGFS_MAX_COMPONENTS) {
		DRM_ERROR("Reached maximum number of debugfs components.\n");
		DRM_ERROR("Report so we increase "
			  "AMDGPU_DEBUGFS_MAX_COMPONENTS.\n");
		return -EINVAL;
	}
	adev->debugfs[adev->debugfs_count].files = files;
	adev->debugfs[adev->debugfs_count].num_files = nfiles;
	adev->debugfs_count = i;
#if defined(CONFIG_DEBUG_FS)
	drm_debugfs_create_files(files, nfiles,
				 adev->ddev->primary->debugfs_root,
				 adev->ddev->primary);
#endif
	return 0;
}

#if defined(CONFIG_DEBUG_FS)

static ssize_t amdgpu_debugfs_regs_read(struct file *f, char __user *buf,
					size_t size, loff_t *pos)
{
A
Al Viro 已提交
3060
	struct amdgpu_device *adev = file_inode(f)->i_private;
A
Alex Deucher 已提交
3061 3062
	ssize_t result = 0;
	int r;
3063
	bool pm_pg_lock, use_bank;
3064
	unsigned instance_bank, sh_bank, se_bank;
A
Alex Deucher 已提交
3065 3066 3067 3068

	if (size & 0x3 || *pos & 0x3)
		return -EINVAL;

3069 3070 3071
	/* are we reading registers for which a PG lock is necessary? */
	pm_pg_lock = (*pos >> 23) & 1;

3072 3073 3074 3075
	if (*pos & (1ULL << 62)) {
		se_bank = (*pos >> 24) & 0x3FF;
		sh_bank = (*pos >> 34) & 0x3FF;
		instance_bank = (*pos >> 44) & 0x3FF;
3076 3077 3078 3079 3080 3081 3082

		if (se_bank == 0x3FF)
			se_bank = 0xFFFFFFFF;
		if (sh_bank == 0x3FF)
			sh_bank = 0xFFFFFFFF;
		if (instance_bank == 0x3FF)
			instance_bank = 0xFFFFFFFF;
3083 3084 3085 3086 3087
		use_bank = 1;
	} else {
		use_bank = 0;
	}

3088
	*pos &= (1UL << 22) - 1;
3089

3090
	if (use_bank) {
3091 3092
		if ((sh_bank != 0xFFFFFFFF && sh_bank >= adev->gfx.config.max_sh_per_se) ||
		    (se_bank != 0xFFFFFFFF && se_bank >= adev->gfx.config.max_shader_engines))
3093 3094 3095 3096 3097 3098
			return -EINVAL;
		mutex_lock(&adev->grbm_idx_mutex);
		amdgpu_gfx_select_se_sh(adev, se_bank,
					sh_bank, instance_bank);
	}

3099 3100 3101
	if (pm_pg_lock)
		mutex_lock(&adev->pm.mutex);

A
Alex Deucher 已提交
3102 3103 3104 3105
	while (size) {
		uint32_t value;

		if (*pos > adev->rmmio_size)
3106
			goto end;
A
Alex Deucher 已提交
3107 3108 3109

		value = RREG32(*pos >> 2);
		r = put_user(value, (uint32_t *)buf);
3110 3111 3112 3113
		if (r) {
			result = r;
			goto end;
		}
A
Alex Deucher 已提交
3114 3115 3116 3117 3118 3119 3120

		result += 4;
		buf += 4;
		*pos += 4;
		size -= 4;
	}

3121 3122 3123 3124 3125 3126
end:
	if (use_bank) {
		amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
		mutex_unlock(&adev->grbm_idx_mutex);
	}

3127 3128 3129
	if (pm_pg_lock)
		mutex_unlock(&adev->pm.mutex);

A
Alex Deucher 已提交
3130 3131 3132 3133 3134 3135
	return result;
}

static ssize_t amdgpu_debugfs_regs_write(struct file *f, const char __user *buf,
					 size_t size, loff_t *pos)
{
A
Al Viro 已提交
3136
	struct amdgpu_device *adev = file_inode(f)->i_private;
A
Alex Deucher 已提交
3137 3138
	ssize_t result = 0;
	int r;
3139 3140
	bool pm_pg_lock, use_bank;
	unsigned instance_bank, sh_bank, se_bank;
A
Alex Deucher 已提交
3141 3142 3143 3144

	if (size & 0x3 || *pos & 0x3)
		return -EINVAL;

3145 3146 3147 3148 3149 3150 3151 3152 3153 3154 3155 3156 3157 3158 3159 3160 3161 3162 3163
	/* are we reading registers for which a PG lock is necessary? */
	pm_pg_lock = (*pos >> 23) & 1;

	if (*pos & (1ULL << 62)) {
		se_bank = (*pos >> 24) & 0x3FF;
		sh_bank = (*pos >> 34) & 0x3FF;
		instance_bank = (*pos >> 44) & 0x3FF;

		if (se_bank == 0x3FF)
			se_bank = 0xFFFFFFFF;
		if (sh_bank == 0x3FF)
			sh_bank = 0xFFFFFFFF;
		if (instance_bank == 0x3FF)
			instance_bank = 0xFFFFFFFF;
		use_bank = 1;
	} else {
		use_bank = 0;
	}

3164
	*pos &= (1UL << 22) - 1;
3165 3166 3167 3168 3169 3170 3171 3172 3173 3174 3175 3176 3177

	if (use_bank) {
		if ((sh_bank != 0xFFFFFFFF && sh_bank >= adev->gfx.config.max_sh_per_se) ||
		    (se_bank != 0xFFFFFFFF && se_bank >= adev->gfx.config.max_shader_engines))
			return -EINVAL;
		mutex_lock(&adev->grbm_idx_mutex);
		amdgpu_gfx_select_se_sh(adev, se_bank,
					sh_bank, instance_bank);
	}

	if (pm_pg_lock)
		mutex_lock(&adev->pm.mutex);

A
Alex Deucher 已提交
3178 3179 3180 3181 3182 3183 3184 3185 3186 3187 3188 3189 3190 3191 3192 3193 3194 3195
	while (size) {
		uint32_t value;

		if (*pos > adev->rmmio_size)
			return result;

		r = get_user(value, (uint32_t *)buf);
		if (r)
			return r;

		WREG32(*pos >> 2, value);

		result += 4;
		buf += 4;
		*pos += 4;
		size -= 4;
	}

3196 3197 3198 3199 3200 3201 3202 3203
	if (use_bank) {
		amdgpu_gfx_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
		mutex_unlock(&adev->grbm_idx_mutex);
	}

	if (pm_pg_lock)
		mutex_unlock(&adev->pm.mutex);

A
Alex Deucher 已提交
3204 3205 3206
	return result;
}

3207 3208 3209
static ssize_t amdgpu_debugfs_regs_pcie_read(struct file *f, char __user *buf,
					size_t size, loff_t *pos)
{
A
Al Viro 已提交
3210
	struct amdgpu_device *adev = file_inode(f)->i_private;
3211 3212 3213 3214 3215 3216 3217 3218 3219 3220 3221 3222 3223 3224 3225 3226 3227 3228 3229 3230 3231 3232 3233 3234 3235 3236
	ssize_t result = 0;
	int r;

	if (size & 0x3 || *pos & 0x3)
		return -EINVAL;

	while (size) {
		uint32_t value;

		value = RREG32_PCIE(*pos >> 2);
		r = put_user(value, (uint32_t *)buf);
		if (r)
			return r;

		result += 4;
		buf += 4;
		*pos += 4;
		size -= 4;
	}

	return result;
}

static ssize_t amdgpu_debugfs_regs_pcie_write(struct file *f, const char __user *buf,
					 size_t size, loff_t *pos)
{
A
Al Viro 已提交
3237
	struct amdgpu_device *adev = file_inode(f)->i_private;
3238 3239 3240 3241 3242 3243 3244 3245 3246 3247 3248 3249 3250 3251 3252 3253 3254 3255 3256 3257 3258 3259 3260 3261 3262 3263 3264
	ssize_t result = 0;
	int r;

	if (size & 0x3 || *pos & 0x3)
		return -EINVAL;

	while (size) {
		uint32_t value;

		r = get_user(value, (uint32_t *)buf);
		if (r)
			return r;

		WREG32_PCIE(*pos >> 2, value);

		result += 4;
		buf += 4;
		*pos += 4;
		size -= 4;
	}

	return result;
}

static ssize_t amdgpu_debugfs_regs_didt_read(struct file *f, char __user *buf,
					size_t size, loff_t *pos)
{
A
Al Viro 已提交
3265
	struct amdgpu_device *adev = file_inode(f)->i_private;
3266 3267 3268 3269 3270 3271 3272 3273 3274 3275 3276 3277 3278 3279 3280 3281 3282 3283 3284 3285 3286 3287 3288 3289 3290 3291
	ssize_t result = 0;
	int r;

	if (size & 0x3 || *pos & 0x3)
		return -EINVAL;

	while (size) {
		uint32_t value;

		value = RREG32_DIDT(*pos >> 2);
		r = put_user(value, (uint32_t *)buf);
		if (r)
			return r;

		result += 4;
		buf += 4;
		*pos += 4;
		size -= 4;
	}

	return result;
}

static ssize_t amdgpu_debugfs_regs_didt_write(struct file *f, const char __user *buf,
					 size_t size, loff_t *pos)
{
A
Al Viro 已提交
3292
	struct amdgpu_device *adev = file_inode(f)->i_private;
3293 3294 3295 3296 3297 3298 3299 3300 3301 3302 3303 3304 3305 3306 3307 3308 3309 3310 3311 3312 3313 3314 3315 3316 3317 3318 3319
	ssize_t result = 0;
	int r;

	if (size & 0x3 || *pos & 0x3)
		return -EINVAL;

	while (size) {
		uint32_t value;

		r = get_user(value, (uint32_t *)buf);
		if (r)
			return r;

		WREG32_DIDT(*pos >> 2, value);

		result += 4;
		buf += 4;
		*pos += 4;
		size -= 4;
	}

	return result;
}

static ssize_t amdgpu_debugfs_regs_smc_read(struct file *f, char __user *buf,
					size_t size, loff_t *pos)
{
A
Al Viro 已提交
3320
	struct amdgpu_device *adev = file_inode(f)->i_private;
3321 3322 3323 3324 3325 3326 3327 3328 3329
	ssize_t result = 0;
	int r;

	if (size & 0x3 || *pos & 0x3)
		return -EINVAL;

	while (size) {
		uint32_t value;

3330
		value = RREG32_SMC(*pos);
3331 3332 3333 3334 3335 3336 3337 3338 3339 3340 3341 3342 3343 3344 3345 3346
		r = put_user(value, (uint32_t *)buf);
		if (r)
			return r;

		result += 4;
		buf += 4;
		*pos += 4;
		size -= 4;
	}

	return result;
}

static ssize_t amdgpu_debugfs_regs_smc_write(struct file *f, const char __user *buf,
					 size_t size, loff_t *pos)
{
A
Al Viro 已提交
3347
	struct amdgpu_device *adev = file_inode(f)->i_private;
3348 3349 3350 3351 3352 3353 3354 3355 3356 3357 3358 3359 3360
	ssize_t result = 0;
	int r;

	if (size & 0x3 || *pos & 0x3)
		return -EINVAL;

	while (size) {
		uint32_t value;

		r = get_user(value, (uint32_t *)buf);
		if (r)
			return r;

3361
		WREG32_SMC(*pos, value);
3362 3363 3364 3365 3366 3367 3368 3369 3370 3371

		result += 4;
		buf += 4;
		*pos += 4;
		size -= 4;
	}

	return result;
}

3372 3373 3374
static ssize_t amdgpu_debugfs_gca_config_read(struct file *f, char __user *buf,
					size_t size, loff_t *pos)
{
A
Al Viro 已提交
3375
	struct amdgpu_device *adev = file_inode(f)->i_private;
3376 3377 3378 3379 3380 3381 3382
	ssize_t result = 0;
	int r;
	uint32_t *config, no_regs = 0;

	if (size & 0x3 || *pos & 0x3)
		return -EINVAL;

3383
	config = kmalloc_array(256, sizeof(*config), GFP_KERNEL);
3384 3385 3386 3387
	if (!config)
		return -ENOMEM;

	/* version, increment each time something is added */
3388
	config[no_regs++] = 3;
3389 3390 3391 3392 3393 3394 3395 3396 3397 3398 3399 3400 3401 3402 3403 3404 3405 3406 3407 3408 3409 3410 3411 3412
	config[no_regs++] = adev->gfx.config.max_shader_engines;
	config[no_regs++] = adev->gfx.config.max_tile_pipes;
	config[no_regs++] = adev->gfx.config.max_cu_per_sh;
	config[no_regs++] = adev->gfx.config.max_sh_per_se;
	config[no_regs++] = adev->gfx.config.max_backends_per_se;
	config[no_regs++] = adev->gfx.config.max_texture_channel_caches;
	config[no_regs++] = adev->gfx.config.max_gprs;
	config[no_regs++] = adev->gfx.config.max_gs_threads;
	config[no_regs++] = adev->gfx.config.max_hw_contexts;
	config[no_regs++] = adev->gfx.config.sc_prim_fifo_size_frontend;
	config[no_regs++] = adev->gfx.config.sc_prim_fifo_size_backend;
	config[no_regs++] = adev->gfx.config.sc_hiz_tile_fifo_size;
	config[no_regs++] = adev->gfx.config.sc_earlyz_tile_fifo_size;
	config[no_regs++] = adev->gfx.config.num_tile_pipes;
	config[no_regs++] = adev->gfx.config.backend_enable_mask;
	config[no_regs++] = adev->gfx.config.mem_max_burst_length_bytes;
	config[no_regs++] = adev->gfx.config.mem_row_size_in_kb;
	config[no_regs++] = adev->gfx.config.shader_engine_tile_size;
	config[no_regs++] = adev->gfx.config.num_gpus;
	config[no_regs++] = adev->gfx.config.multi_gpu_tile_size;
	config[no_regs++] = adev->gfx.config.mc_arb_ramcfg;
	config[no_regs++] = adev->gfx.config.gb_addr_config;
	config[no_regs++] = adev->gfx.config.num_rbs;

3413 3414 3415 3416 3417
	/* rev==1 */
	config[no_regs++] = adev->rev_id;
	config[no_regs++] = adev->pg_flags;
	config[no_regs++] = adev->cg_flags;

3418 3419 3420 3421
	/* rev==2 */
	config[no_regs++] = adev->family;
	config[no_regs++] = adev->external_rev_id;

3422 3423 3424 3425 3426 3427
	/* rev==3 */
	config[no_regs++] = adev->pdev->device;
	config[no_regs++] = adev->pdev->revision;
	config[no_regs++] = adev->pdev->subsystem_device;
	config[no_regs++] = adev->pdev->subsystem_vendor;

3428 3429 3430 3431 3432 3433 3434 3435 3436 3437 3438 3439 3440 3441 3442 3443 3444 3445 3446 3447
	while (size && (*pos < no_regs * 4)) {
		uint32_t value;

		value = config[*pos >> 2];
		r = put_user(value, (uint32_t *)buf);
		if (r) {
			kfree(config);
			return r;
		}

		result += 4;
		buf += 4;
		*pos += 4;
		size -= 4;
	}

	kfree(config);
	return result;
}

3448 3449 3450
static ssize_t amdgpu_debugfs_sensor_read(struct file *f, char __user *buf,
					size_t size, loff_t *pos)
{
A
Al Viro 已提交
3451
	struct amdgpu_device *adev = file_inode(f)->i_private;
3452 3453
	int idx, x, outsize, r, valuesize;
	uint32_t values[16];
3454

3455
	if (size & 3 || *pos & 0x3)
3456 3457
		return -EINVAL;

3458 3459 3460
	if (amdgpu_dpm == 0)
		return -EINVAL;

3461 3462 3463
	/* convert offset to sensor number */
	idx = *pos >> 2;

3464
	valuesize = sizeof(values);
3465
	if (adev->powerplay.pp_funcs && adev->powerplay.pp_funcs->read_sensor)
3466
		r = adev->powerplay.pp_funcs->read_sensor(adev->powerplay.pp_handle, idx, &values[0], &valuesize);
3467 3468 3469
	else if (adev->pm.funcs && adev->pm.funcs->read_sensor)
		r = adev->pm.funcs->read_sensor(adev, idx, &values[0],
						&valuesize);
3470 3471 3472
	else
		return -EINVAL;

3473 3474 3475 3476 3477 3478 3479 3480 3481 3482 3483 3484 3485
	if (size > valuesize)
		return -EINVAL;

	outsize = 0;
	x = 0;
	if (!r) {
		while (size) {
			r = put_user(values[x++], (int32_t *)buf);
			buf += 4;
			size -= 4;
			outsize += 4;
		}
	}
3486

3487
	return !r ? outsize : r;
3488
}
3489

3490 3491 3492 3493 3494 3495
static ssize_t amdgpu_debugfs_wave_read(struct file *f, char __user *buf,
					size_t size, loff_t *pos)
{
	struct amdgpu_device *adev = f->f_inode->i_private;
	int r, x;
	ssize_t result=0;
3496
	uint32_t offset, se, sh, cu, wave, simd, data[32];
3497 3498 3499 3500 3501 3502 3503 3504 3505 3506 3507 3508 3509 3510 3511 3512 3513

	if (size & 3 || *pos & 3)
		return -EINVAL;

	/* decode offset */
	offset = (*pos & 0x7F);
	se = ((*pos >> 7) & 0xFF);
	sh = ((*pos >> 15) & 0xFF);
	cu = ((*pos >> 23) & 0xFF);
	wave = ((*pos >> 31) & 0xFF);
	simd = ((*pos >> 37) & 0xFF);

	/* switch to the specific se/sh/cu */
	mutex_lock(&adev->grbm_idx_mutex);
	amdgpu_gfx_select_se_sh(adev, se, sh, cu);

	x = 0;
3514 3515
	if (adev->gfx.funcs->read_wave_data)
		adev->gfx.funcs->read_wave_data(adev, simd, wave, data, &x);
3516 3517 3518 3519

	amdgpu_gfx_select_se_sh(adev, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF);
	mutex_unlock(&adev->grbm_idx_mutex);

3520 3521 3522
	if (!x)
		return -EINVAL;

3523
	while (size && (offset < x * 4)) {
3524 3525
		uint32_t value;

3526
		value = data[offset >> 2];
3527 3528 3529 3530 3531 3532
		r = put_user(value, (uint32_t *)buf);
		if (r)
			return r;

		result += 4;
		buf += 4;
3533
		offset += 4;
3534 3535 3536 3537 3538 3539
		size -= 4;
	}

	return result;
}

3540 3541 3542 3543 3544 3545 3546 3547 3548 3549 3550 3551 3552 3553 3554 3555 3556 3557 3558 3559 3560 3561 3562 3563 3564 3565 3566 3567 3568 3569 3570 3571 3572 3573 3574 3575 3576 3577 3578 3579 3580 3581 3582 3583 3584 3585 3586 3587 3588 3589 3590 3591 3592 3593 3594 3595 3596 3597 3598 3599
static ssize_t amdgpu_debugfs_gpr_read(struct file *f, char __user *buf,
					size_t size, loff_t *pos)
{
	struct amdgpu_device *adev = f->f_inode->i_private;
	int r;
	ssize_t result = 0;
	uint32_t offset, se, sh, cu, wave, simd, thread, bank, *data;

	if (size & 3 || *pos & 3)
		return -EINVAL;

	/* decode offset */
	offset = (*pos & 0xFFF);       /* in dwords */
	se = ((*pos >> 12) & 0xFF);
	sh = ((*pos >> 20) & 0xFF);
	cu = ((*pos >> 28) & 0xFF);
	wave = ((*pos >> 36) & 0xFF);
	simd = ((*pos >> 44) & 0xFF);
	thread = ((*pos >> 52) & 0xFF);
	bank = ((*pos >> 60) & 1);

	data = kmalloc_array(1024, sizeof(*data), GFP_KERNEL);
	if (!data)
		return -ENOMEM;

	/* switch to the specific se/sh/cu */
	mutex_lock(&adev->grbm_idx_mutex);
	amdgpu_gfx_select_se_sh(adev, se, sh, cu);

	if (bank == 0) {
		if (adev->gfx.funcs->read_wave_vgprs)
			adev->gfx.funcs->read_wave_vgprs(adev, simd, wave, thread, offset, size>>2, data);
	} else {
		if (adev->gfx.funcs->read_wave_sgprs)
			adev->gfx.funcs->read_wave_sgprs(adev, simd, wave, offset, size>>2, data);
	}

	amdgpu_gfx_select_se_sh(adev, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF);
	mutex_unlock(&adev->grbm_idx_mutex);

	while (size) {
		uint32_t value;

		value = data[offset++];
		r = put_user(value, (uint32_t *)buf);
		if (r) {
			result = r;
			goto err;
		}

		result += 4;
		buf += 4;
		size -= 4;
	}

err:
	kfree(data);
	return result;
}

A
Alex Deucher 已提交
3600 3601 3602 3603 3604 3605
static const struct file_operations amdgpu_debugfs_regs_fops = {
	.owner = THIS_MODULE,
	.read = amdgpu_debugfs_regs_read,
	.write = amdgpu_debugfs_regs_write,
	.llseek = default_llseek
};
3606 3607 3608 3609 3610 3611 3612 3613 3614 3615 3616 3617 3618 3619 3620 3621 3622 3623 3624
static const struct file_operations amdgpu_debugfs_regs_didt_fops = {
	.owner = THIS_MODULE,
	.read = amdgpu_debugfs_regs_didt_read,
	.write = amdgpu_debugfs_regs_didt_write,
	.llseek = default_llseek
};
static const struct file_operations amdgpu_debugfs_regs_pcie_fops = {
	.owner = THIS_MODULE,
	.read = amdgpu_debugfs_regs_pcie_read,
	.write = amdgpu_debugfs_regs_pcie_write,
	.llseek = default_llseek
};
static const struct file_operations amdgpu_debugfs_regs_smc_fops = {
	.owner = THIS_MODULE,
	.read = amdgpu_debugfs_regs_smc_read,
	.write = amdgpu_debugfs_regs_smc_write,
	.llseek = default_llseek
};

3625 3626 3627 3628 3629 3630
static const struct file_operations amdgpu_debugfs_gca_config_fops = {
	.owner = THIS_MODULE,
	.read = amdgpu_debugfs_gca_config_read,
	.llseek = default_llseek
};

3631 3632 3633 3634 3635 3636
static const struct file_operations amdgpu_debugfs_sensors_fops = {
	.owner = THIS_MODULE,
	.read = amdgpu_debugfs_sensor_read,
	.llseek = default_llseek
};

3637 3638 3639 3640 3641
static const struct file_operations amdgpu_debugfs_wave_fops = {
	.owner = THIS_MODULE,
	.read = amdgpu_debugfs_wave_read,
	.llseek = default_llseek
};
3642 3643 3644 3645 3646
static const struct file_operations amdgpu_debugfs_gpr_fops = {
	.owner = THIS_MODULE,
	.read = amdgpu_debugfs_gpr_read,
	.llseek = default_llseek
};
3647

3648 3649 3650 3651 3652
static const struct file_operations *debugfs_regs[] = {
	&amdgpu_debugfs_regs_fops,
	&amdgpu_debugfs_regs_didt_fops,
	&amdgpu_debugfs_regs_pcie_fops,
	&amdgpu_debugfs_regs_smc_fops,
3653
	&amdgpu_debugfs_gca_config_fops,
3654
	&amdgpu_debugfs_sensors_fops,
3655
	&amdgpu_debugfs_wave_fops,
3656
	&amdgpu_debugfs_gpr_fops,
3657 3658 3659 3660 3661 3662 3663
};

static const char *debugfs_regs_names[] = {
	"amdgpu_regs",
	"amdgpu_regs_didt",
	"amdgpu_regs_pcie",
	"amdgpu_regs_smc",
3664
	"amdgpu_gca_config",
3665
	"amdgpu_sensors",
3666
	"amdgpu_wave",
3667
	"amdgpu_gpr",
3668
};
A
Alex Deucher 已提交
3669 3670 3671 3672 3673

static int amdgpu_debugfs_regs_init(struct amdgpu_device *adev)
{
	struct drm_minor *minor = adev->ddev->primary;
	struct dentry *ent, *root = minor->debugfs_root;
3674 3675 3676 3677 3678 3679 3680 3681 3682 3683 3684 3685 3686
	unsigned i, j;

	for (i = 0; i < ARRAY_SIZE(debugfs_regs); i++) {
		ent = debugfs_create_file(debugfs_regs_names[i],
					  S_IFREG | S_IRUGO, root,
					  adev, debugfs_regs[i]);
		if (IS_ERR(ent)) {
			for (j = 0; j < i; j++) {
				debugfs_remove(adev->debugfs_regs[i]);
				adev->debugfs_regs[i] = NULL;
			}
			return PTR_ERR(ent);
		}
A
Alex Deucher 已提交
3687

3688 3689 3690 3691
		if (!i)
			i_size_write(ent->d_inode, adev->rmmio_size);
		adev->debugfs_regs[i] = ent;
	}
A
Alex Deucher 已提交
3692 3693 3694 3695 3696 3697

	return 0;
}

static void amdgpu_debugfs_regs_cleanup(struct amdgpu_device *adev)
{
3698 3699 3700 3701 3702 3703 3704 3705
	unsigned i;

	for (i = 0; i < ARRAY_SIZE(debugfs_regs); i++) {
		if (adev->debugfs_regs[i]) {
			debugfs_remove(adev->debugfs_regs[i]);
			adev->debugfs_regs[i] = NULL;
		}
	}
A
Alex Deucher 已提交
3706 3707 3708 3709 3710 3711
}

int amdgpu_debugfs_init(struct drm_minor *minor)
{
	return 0;
}
3712 3713 3714 3715 3716 3717
#else
static int amdgpu_debugfs_regs_init(struct amdgpu_device *adev)
{
	return 0;
}
static void amdgpu_debugfs_regs_cleanup(struct amdgpu_device *adev) { }
A
Alex Deucher 已提交
3718
#endif