msm_gpu.c 19.1 KB
Newer Older
R
Rob Clark 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19
/*
 * Copyright (C) 2013 Red Hat
 * Author: Rob Clark <robdclark@gmail.com>
 *
 * This program is free software; you can redistribute it and/or modify it
 * under the terms of the GNU General Public License version 2 as published by
 * the Free Software Foundation.
 *
 * This program is distributed in the hope that it will be useful, but WITHOUT
 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
 * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
 * more details.
 *
 * You should have received a copy of the GNU General Public License along with
 * this program.  If not, see <http://www.gnu.org/licenses/>.
 */

#include "msm_gpu.h"
#include "msm_gem.h"
20
#include "msm_mmu.h"
21
#include "msm_fence.h"
R
Rob Clark 已提交
22

23 24
#include <linux/string_helpers.h>

R
Rob Clark 已提交
25 26 27 28 29

/*
 * Power Management:
 */

30
#ifdef DOWNSTREAM_CONFIG_MSM_BUS_SCALING
R
Rob Clark 已提交
31
#include <mach/board.h>
R
Rob Clark 已提交
32
static void bs_init(struct msm_gpu *gpu)
R
Rob Clark 已提交
33
{
R
Rob Clark 已提交
34 35
	if (gpu->bus_scale_table) {
		gpu->bsc = msm_bus_scale_register_client(gpu->bus_scale_table);
R
Rob Clark 已提交
36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55
		DBG("bus scale client: %08x", gpu->bsc);
	}
}

static void bs_fini(struct msm_gpu *gpu)
{
	if (gpu->bsc) {
		msm_bus_scale_unregister_client(gpu->bsc);
		gpu->bsc = 0;
	}
}

static void bs_set(struct msm_gpu *gpu, int idx)
{
	if (gpu->bsc) {
		DBG("set bus scaling: %d", idx);
		msm_bus_scale_client_update_request(gpu->bsc, idx);
	}
}
#else
R
Rob Clark 已提交
56
static void bs_init(struct msm_gpu *gpu) {}
R
Rob Clark 已提交
57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97
static void bs_fini(struct msm_gpu *gpu) {}
static void bs_set(struct msm_gpu *gpu, int idx) {}
#endif

static int enable_pwrrail(struct msm_gpu *gpu)
{
	struct drm_device *dev = gpu->dev;
	int ret = 0;

	if (gpu->gpu_reg) {
		ret = regulator_enable(gpu->gpu_reg);
		if (ret) {
			dev_err(dev->dev, "failed to enable 'gpu_reg': %d\n", ret);
			return ret;
		}
	}

	if (gpu->gpu_cx) {
		ret = regulator_enable(gpu->gpu_cx);
		if (ret) {
			dev_err(dev->dev, "failed to enable 'gpu_cx': %d\n", ret);
			return ret;
		}
	}

	return 0;
}

static int disable_pwrrail(struct msm_gpu *gpu)
{
	if (gpu->gpu_cx)
		regulator_disable(gpu->gpu_cx);
	if (gpu->gpu_reg)
		regulator_disable(gpu->gpu_reg);
	return 0;
}

static int enable_clk(struct msm_gpu *gpu)
{
	int i;

98 99
	if (gpu->core_clk && gpu->fast_rate)
		clk_set_rate(gpu->core_clk, gpu->fast_rate);
R
Rob Clark 已提交
100

101
	/* Set the RBBM timer rate to 19.2Mhz */
102 103
	if (gpu->rbbmtimer_clk)
		clk_set_rate(gpu->rbbmtimer_clk, 19200000);
104

105
	for (i = gpu->nr_clocks - 1; i >= 0; i--)
106 107
		if (gpu->grp_clks[i])
			clk_prepare(gpu->grp_clks[i]);
R
Rob Clark 已提交
108

109
	for (i = gpu->nr_clocks - 1; i >= 0; i--)
R
Rob Clark 已提交
110 111 112 113 114 115 116 117 118 119
		if (gpu->grp_clks[i])
			clk_enable(gpu->grp_clks[i]);

	return 0;
}

static int disable_clk(struct msm_gpu *gpu)
{
	int i;

120
	for (i = gpu->nr_clocks - 1; i >= 0; i--)
121
		if (gpu->grp_clks[i])
R
Rob Clark 已提交
122 123
			clk_disable(gpu->grp_clks[i]);

124
	for (i = gpu->nr_clocks - 1; i >= 0; i--)
R
Rob Clark 已提交
125 126 127
		if (gpu->grp_clks[i])
			clk_unprepare(gpu->grp_clks[i]);

128 129 130 131 132
	/*
	 * Set the clock to a deliberately low rate. On older targets the clock
	 * speed had to be non zero to avoid problems. On newer targets this
	 * will be rounded down to zero anyway so it all works out.
	 */
133 134
	if (gpu->core_clk)
		clk_set_rate(gpu->core_clk, 27000000);
135

136 137
	if (gpu->rbbmtimer_clk)
		clk_set_rate(gpu->rbbmtimer_clk, 0);
138

R
Rob Clark 已提交
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
	return 0;
}

static int enable_axi(struct msm_gpu *gpu)
{
	if (gpu->ebi1_clk)
		clk_prepare_enable(gpu->ebi1_clk);
	if (gpu->bus_freq)
		bs_set(gpu, gpu->bus_freq);
	return 0;
}

static int disable_axi(struct msm_gpu *gpu)
{
	if (gpu->ebi1_clk)
		clk_disable_unprepare(gpu->ebi1_clk);
	if (gpu->bus_freq)
		bs_set(gpu, 0);
	return 0;
}

int msm_gpu_pm_resume(struct msm_gpu *gpu)
{
	int ret;

R
Rob Clark 已提交
164
	DBG("%s", gpu->name);
R
Rob Clark 已提交
165 166 167 168 169 170 171 172 173 174 175 176 177

	ret = enable_pwrrail(gpu);
	if (ret)
		return ret;

	ret = enable_clk(gpu);
	if (ret)
		return ret;

	ret = enable_axi(gpu);
	if (ret)
		return ret;

R
Rob Clark 已提交
178 179
	gpu->needs_hw_init = true;

R
Rob Clark 已提交
180 181 182 183 184 185 186
	return 0;
}

int msm_gpu_pm_suspend(struct msm_gpu *gpu)
{
	int ret;

R
Rob Clark 已提交
187
	DBG("%s", gpu->name);
R
Rob Clark 已提交
188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203

	ret = disable_axi(gpu);
	if (ret)
		return ret;

	ret = disable_clk(gpu);
	if (ret)
		return ret;

	ret = disable_pwrrail(gpu);
	if (ret)
		return ret;

	return 0;
}

R
Rob Clark 已提交
204
int msm_gpu_hw_init(struct msm_gpu *gpu)
205
{
R
Rob Clark 已提交
206
	int ret;
207

208 209
	WARN_ON(!mutex_is_locked(&gpu->dev->struct_mutex));

R
Rob Clark 已提交
210 211
	if (!gpu->needs_hw_init)
		return 0;
212

R
Rob Clark 已提交
213 214 215 216 217
	disable_irq(gpu->irq);
	ret = gpu->funcs->hw_init(gpu);
	if (!ret)
		gpu->needs_hw_init = false;
	enable_irq(gpu->irq);
218

R
Rob Clark 已提交
219
	return ret;
220 221
}

222 223 224 225
/*
 * Hangcheck detection for locked gpu:
 */

226 227 228 229 230 231 232 233 234 235 236 237 238 239
static void update_fences(struct msm_gpu *gpu, struct msm_ringbuffer *ring,
		uint32_t fence)
{
	struct msm_gem_submit *submit;

	list_for_each_entry(submit, &ring->submits, node) {
		if (submit->seqno > fence)
			break;

		msm_update_fence(submit->ring->fctx,
			submit->fence->seqno);
	}
}

240 241 242 243 244 245 246 247 248 249 250 251 252 253
static struct msm_gem_submit *
find_submit(struct msm_ringbuffer *ring, uint32_t fence)
{
	struct msm_gem_submit *submit;

	WARN_ON(!mutex_is_locked(&ring->gpu->dev->struct_mutex));

	list_for_each_entry(submit, &ring->submits, node)
		if (submit->seqno == fence)
			return submit;

	return NULL;
}

254
static void retire_submits(struct msm_gpu *gpu);
255

256 257 258 259
static void recover_worker(struct work_struct *work)
{
	struct msm_gpu *gpu = container_of(work, struct msm_gpu, recover_work);
	struct drm_device *dev = gpu->dev;
260
	struct msm_gem_submit *submit;
261 262 263 264 265 266 267
	struct msm_ringbuffer *cur_ring = gpu->funcs->active_ring(gpu);
	uint64_t fence;
	int i;

	/* Update all the rings with the latest and greatest fence */
	for (i = 0; i < ARRAY_SIZE(gpu->rb); i++) {
		struct msm_ringbuffer *ring = gpu->rb[i];
268

269 270 271 272 273 274 275 276 277 278 279
		fence = ring->memptrs->fence;

		/*
		 * For the current (faulting?) ring/submit advance the fence by
		 * one more to clear the faulting submit
		 */
		if (ring == cur_ring)
			fence = fence + 1;

		update_fences(gpu, ring, fence);
	}
280

281
	mutex_lock(&dev->struct_mutex);
282

283

284
	dev_err(dev->dev, "%s: hangcheck recover!\n", gpu->name);
285 286
	fence = cur_ring->memptrs->fence + 1;

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
	submit = find_submit(cur_ring, fence);
	if (submit) {
		struct task_struct *task;

		rcu_read_lock();
		task = pid_task(submit->pid, PIDTYPE_PID);
		if (task) {
			char *cmd;

			/*
			 * So slightly annoying, in other paths like
			 * mmap'ing gem buffers, mmap_sem is acquired
			 * before struct_mutex, which means we can't
			 * hold struct_mutex across the call to
			 * get_cmdline().  But submits are retired
			 * from the same in-order workqueue, so we can
			 * safely drop the lock here without worrying
			 * about the submit going away.
			 */
			mutex_unlock(&dev->struct_mutex);
			cmd = kstrdup_quotable_cmdline(task, GFP_KERNEL);
			mutex_lock(&dev->struct_mutex);

			dev_err(dev->dev, "%s: offending task: %s (%s)\n",
				gpu->name, task->comm, cmd);
312
		}
313 314
		rcu_read_unlock();

315 316 317
	}

	if (msm_gpu_active(gpu)) {
318
		/* retire completed submits, plus the one that hung: */
319
		retire_submits(gpu);
320

R
Rob Clark 已提交
321
		pm_runtime_get_sync(&gpu->pdev->dev);
322
		gpu->funcs->recover(gpu);
R
Rob Clark 已提交
323
		pm_runtime_put_sync(&gpu->pdev->dev);
324

325 326 327 328
		/*
		 * Replay all remaining submits starting with highest priority
		 * ring
		 */
329
		for (i = 0; i < gpu->nr_rings; i++) {
330 331 332 333
			struct msm_ringbuffer *ring = gpu->rb[i];

			list_for_each_entry(submit, &ring->submits, node)
				gpu->funcs->submit(gpu, submit, NULL);
334
		}
335
	}
336

337 338 339 340 341 342 343 344 345 346 347 348 349 350 351
	mutex_unlock(&dev->struct_mutex);

	msm_gpu_retire(gpu);
}

static void hangcheck_timer_reset(struct msm_gpu *gpu)
{
	DBG("%s", gpu->name);
	mod_timer(&gpu->hangcheck_timer,
			round_jiffies_up(jiffies + DRM_MSM_HANGCHECK_JIFFIES));
}

static void hangcheck_handler(unsigned long data)
{
	struct msm_gpu *gpu = (struct msm_gpu *)data;
352 353
	struct drm_device *dev = gpu->dev;
	struct msm_drm_private *priv = dev->dev_private;
354 355
	struct msm_ringbuffer *ring = gpu->funcs->active_ring(gpu);
	uint32_t fence = ring->memptrs->fence;
356

357
	if (fence != ring->hangcheck_fence) {
358
		/* some progress has been made.. ya! */
359 360
		ring->hangcheck_fence = fence;
	} else if (fence < ring->seqno) {
361
		/* no progress and not done.. hung! */
362 363 364
		ring->hangcheck_fence = fence;
		dev_err(dev->dev, "%s: hangcheck detected gpu lockup rb %d!\n",
				gpu->name, ring->id);
R
Rob Clark 已提交
365 366 367
		dev_err(dev->dev, "%s:     completed fence: %u\n",
				gpu->name, fence);
		dev_err(dev->dev, "%s:     submitted fence: %u\n",
368 369
				gpu->name, ring->seqno);

370 371 372 373
		queue_work(priv->wq, &gpu->recover_work);
	}

	/* if still more pending work, reset the hangcheck timer: */
374
	if (ring->seqno > ring->hangcheck_fence)
375
		hangcheck_timer_reset(gpu);
376 377 378

	/* workaround for missing irq: */
	queue_work(priv->wq, &gpu->retire_work);
379 380
}

381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433
/*
 * Performance Counters:
 */

/* called under perf_lock */
static int update_hw_cntrs(struct msm_gpu *gpu, uint32_t ncntrs, uint32_t *cntrs)
{
	uint32_t current_cntrs[ARRAY_SIZE(gpu->last_cntrs)];
	int i, n = min(ncntrs, gpu->num_perfcntrs);

	/* read current values: */
	for (i = 0; i < gpu->num_perfcntrs; i++)
		current_cntrs[i] = gpu_read(gpu, gpu->perfcntrs[i].sample_reg);

	/* update cntrs: */
	for (i = 0; i < n; i++)
		cntrs[i] = current_cntrs[i] - gpu->last_cntrs[i];

	/* save current values: */
	for (i = 0; i < gpu->num_perfcntrs; i++)
		gpu->last_cntrs[i] = current_cntrs[i];

	return n;
}

static void update_sw_cntrs(struct msm_gpu *gpu)
{
	ktime_t time;
	uint32_t elapsed;
	unsigned long flags;

	spin_lock_irqsave(&gpu->perf_lock, flags);
	if (!gpu->perfcntr_active)
		goto out;

	time = ktime_get();
	elapsed = ktime_to_us(ktime_sub(time, gpu->last_sample.time));

	gpu->totaltime += elapsed;
	if (gpu->last_sample.active)
		gpu->activetime += elapsed;

	gpu->last_sample.active = msm_gpu_active(gpu);
	gpu->last_sample.time = time;

out:
	spin_unlock_irqrestore(&gpu->perf_lock, flags);
}

void msm_gpu_perfcntr_start(struct msm_gpu *gpu)
{
	unsigned long flags;

R
Rob Clark 已提交
434 435
	pm_runtime_get_sync(&gpu->pdev->dev);

436 437 438 439 440 441 442 443 444 445 446 447 448
	spin_lock_irqsave(&gpu->perf_lock, flags);
	/* we could dynamically enable/disable perfcntr registers too.. */
	gpu->last_sample.active = msm_gpu_active(gpu);
	gpu->last_sample.time = ktime_get();
	gpu->activetime = gpu->totaltime = 0;
	gpu->perfcntr_active = true;
	update_hw_cntrs(gpu, 0, NULL);
	spin_unlock_irqrestore(&gpu->perf_lock, flags);
}

void msm_gpu_perfcntr_stop(struct msm_gpu *gpu)
{
	gpu->perfcntr_active = false;
R
Rob Clark 已提交
449
	pm_runtime_put_sync(&gpu->pdev->dev);
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
}

/* returns -errno or # of cntrs sampled */
int msm_gpu_perfcntr_sample(struct msm_gpu *gpu, uint32_t *activetime,
		uint32_t *totaltime, uint32_t ncntrs, uint32_t *cntrs)
{
	unsigned long flags;
	int ret;

	spin_lock_irqsave(&gpu->perf_lock, flags);

	if (!gpu->perfcntr_active) {
		ret = -EINVAL;
		goto out;
	}

	*activetime = gpu->activetime;
	*totaltime = gpu->totaltime;

	gpu->activetime = gpu->totaltime = 0;

	ret = update_hw_cntrs(gpu, ncntrs, cntrs);

out:
	spin_unlock_irqrestore(&gpu->perf_lock, flags);

	return ret;
}

R
Rob Clark 已提交
479 480 481 482
/*
 * Cmdstream submission/retirement:
 */

483 484 485 486 487 488 489 490
static void retire_submit(struct msm_gpu *gpu, struct msm_gem_submit *submit)
{
	int i;

	for (i = 0; i < submit->nr_bos; i++) {
		struct msm_gem_object *msm_obj = submit->bos[i].obj;
		/* move to inactive: */
		msm_gem_move_to_inactive(&msm_obj->base);
491
		msm_gem_put_iova(&msm_obj->base, gpu->aspace);
492 493 494
		drm_gem_object_unreference(&msm_obj->base);
	}

R
Rob Clark 已提交
495 496
	pm_runtime_mark_last_busy(&gpu->pdev->dev);
	pm_runtime_put_autosuspend(&gpu->pdev->dev);
497
	msm_gem_submit_free(submit);
498 499
}

500
static void retire_submits(struct msm_gpu *gpu)
501 502
{
	struct drm_device *dev = gpu->dev;
503 504
	struct msm_gem_submit *submit, *tmp;
	int i;
505 506 507

	WARN_ON(!mutex_is_locked(&dev->struct_mutex));

508
	/* Retire the commits starting with highest priority */
509
	for (i = 0; i < gpu->nr_rings; i++) {
510
		struct msm_ringbuffer *ring = gpu->rb[i];
511

512 513 514
		list_for_each_entry_safe(submit, tmp, &ring->submits, node) {
			if (dma_fence_is_signaled(submit->fence))
				retire_submit(gpu, submit);
515 516 517 518
		}
	}
}

R
Rob Clark 已提交
519 520 521 522
static void retire_worker(struct work_struct *work)
{
	struct msm_gpu *gpu = container_of(work, struct msm_gpu, retire_work);
	struct drm_device *dev = gpu->dev;
523
	int i;
R
Rob Clark 已提交
524

525 526
	for (i = 0; i < gpu->nr_rings; i++)
		update_fences(gpu, gpu->rb[i], gpu->rb[i]->memptrs->fence);
R
Rob Clark 已提交
527

R
Rob Clark 已提交
528
	mutex_lock(&dev->struct_mutex);
529
	retire_submits(gpu);
R
Rob Clark 已提交
530 531 532 533 534 535 536 537
	mutex_unlock(&dev->struct_mutex);
}

/* call from irq handler to schedule work to retire bo's */
void msm_gpu_retire(struct msm_gpu *gpu)
{
	struct msm_drm_private *priv = gpu->dev->dev_private;
	queue_work(priv->wq, &gpu->retire_work);
538
	update_sw_cntrs(gpu);
R
Rob Clark 已提交
539 540 541
}

/* add bo's to gpu's ring, and kick gpu: */
542
void msm_gpu_submit(struct msm_gpu *gpu, struct msm_gem_submit *submit,
R
Rob Clark 已提交
543 544 545 546
		struct msm_file_private *ctx)
{
	struct drm_device *dev = gpu->dev;
	struct msm_drm_private *priv = dev->dev_private;
547
	struct msm_ringbuffer *ring = submit->ring;
548
	int i;
R
Rob Clark 已提交
549

550 551
	WARN_ON(!mutex_is_locked(&dev->struct_mutex));

R
Rob Clark 已提交
552 553 554
	pm_runtime_get_sync(&gpu->pdev->dev);

	msm_gpu_hw_init(gpu);
555

556 557 558
	submit->seqno = ++ring->seqno;

	list_add_tail(&submit->node, &ring->submits);
559

560
	msm_rd_dump_submit(priv->rd, submit, NULL);
561

562 563
	update_sw_cntrs(gpu);

R
Rob Clark 已提交
564 565
	for (i = 0; i < submit->nr_bos; i++) {
		struct msm_gem_object *msm_obj = submit->bos[i].obj;
566
		uint64_t iova;
R
Rob Clark 已提交
567 568 569 570 571 572

		/* can't happen yet.. but when we add 2d support we'll have
		 * to deal w/ cross-ring synchronization:
		 */
		WARN_ON(is_active(msm_obj) && (msm_obj->gpu != gpu));

573 574
		/* submit takes a reference to the bo and iova until retired: */
		drm_gem_object_reference(&msm_obj->base);
575
		msm_gem_get_iova(&msm_obj->base,
576
				submit->gpu->aspace, &iova);
R
Rob Clark 已提交
577

578 579
		if (submit->bos[i].flags & MSM_SUBMIT_BO_WRITE)
			msm_gem_move_to_active(&msm_obj->base, gpu, true, submit->fence);
580 581
		else if (submit->bos[i].flags & MSM_SUBMIT_BO_READ)
			msm_gem_move_to_active(&msm_obj->base, gpu, false, submit->fence);
R
Rob Clark 已提交
582
	}
583

584
	gpu->funcs->submit(gpu, submit, ctx);
585 586
	priv->lastctx = ctx;

587
	hangcheck_timer_reset(gpu);
R
Rob Clark 已提交
588 589 590 591 592 593 594 595 596 597 598 599
}

/*
 * Init/Cleanup:
 */

static irqreturn_t irq_handler(int irq, void *data)
{
	struct msm_gpu *gpu = data;
	return gpu->funcs->irq(gpu);
}

600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628
static struct clk *get_clock(struct device *dev, const char *name)
{
	struct clk *clk = devm_clk_get(dev, name);

	return IS_ERR(clk) ? NULL : clk;
}

static int get_clocks(struct platform_device *pdev, struct msm_gpu *gpu)
{
	struct device *dev = &pdev->dev;
	struct property *prop;
	const char *name;
	int i = 0;

	gpu->nr_clocks = of_property_count_strings(dev->of_node, "clock-names");
	if (gpu->nr_clocks < 1) {
		gpu->nr_clocks = 0;
		return 0;
	}

	gpu->grp_clks = devm_kcalloc(dev, sizeof(struct clk *), gpu->nr_clocks,
		GFP_KERNEL);
	if (!gpu->grp_clks)
		return -ENOMEM;

	of_property_for_each_string(dev->of_node, "clock-names", prop, name) {
		gpu->grp_clks[i] = get_clock(dev, name);

		/* Remember the key clocks that we need to control later */
629
		if (!strcmp(name, "core") || !strcmp(name, "core_clk"))
630
			gpu->core_clk = gpu->grp_clks[i];
631
		else if (!strcmp(name, "rbbmtimer") || !strcmp(name, "rbbmtimer_clk"))
632 633 634 635 636 637 638
			gpu->rbbmtimer_clk = gpu->grp_clks[i];

		++i;
	}

	return 0;
}
R
Rob Clark 已提交
639

640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678
static struct msm_gem_address_space *
msm_gpu_create_address_space(struct msm_gpu *gpu, struct platform_device *pdev,
		uint64_t va_start, uint64_t va_end)
{
	struct iommu_domain *iommu;
	struct msm_gem_address_space *aspace;
	int ret;

	/*
	 * Setup IOMMU.. eventually we will (I think) do this once per context
	 * and have separate page tables per context.  For now, to keep things
	 * simple and to get something working, just use a single address space:
	 */
	iommu = iommu_domain_alloc(&platform_bus_type);
	if (!iommu)
		return NULL;

	iommu->geometry.aperture_start = va_start;
	iommu->geometry.aperture_end = va_end;

	dev_info(gpu->dev->dev, "%s: using IOMMU\n", gpu->name);

	aspace = msm_gem_address_space_create(&pdev->dev, iommu, "gpu");
	if (IS_ERR(aspace)) {
		dev_err(gpu->dev->dev, "failed to init iommu: %ld\n",
			PTR_ERR(aspace));
		iommu_domain_free(iommu);
		return ERR_CAST(aspace);
	}

	ret = aspace->mmu->funcs->attach(aspace->mmu, NULL, 0);
	if (ret) {
		msm_gem_address_space_put(aspace);
		return ERR_PTR(ret);
	}

	return aspace;
}

R
Rob Clark 已提交
679 680
int msm_gpu_init(struct drm_device *drm, struct platform_device *pdev,
		struct msm_gpu *gpu, const struct msm_gpu_funcs *funcs,
681
		const char *name, struct msm_gpu_config *config)
R
Rob Clark 已提交
682
{
683 684 685
	int i, ret, nr_rings = config->nr_rings;
	void *memptrs;
	uint64_t memptrs_iova;
R
Rob Clark 已提交
686

687 688 689
	if (WARN_ON(gpu->num_perfcntrs > ARRAY_SIZE(gpu->last_cntrs)))
		gpu->num_perfcntrs = ARRAY_SIZE(gpu->last_cntrs);

R
Rob Clark 已提交
690 691 692 693 694 695
	gpu->dev = drm;
	gpu->funcs = funcs;
	gpu->name = name;

	INIT_LIST_HEAD(&gpu->active_list);
	INIT_WORK(&gpu->retire_work, retire_worker);
696 697
	INIT_WORK(&gpu->recover_work, recover_worker);

698

699 700
	setup_timer(&gpu->hangcheck_timer, hangcheck_handler,
			(unsigned long)gpu);
R
Rob Clark 已提交
701

702 703
	spin_lock_init(&gpu->perf_lock);

R
Rob Clark 已提交
704 705

	/* Map registers: */
706
	gpu->mmio = msm_ioremap(pdev, config->ioname, name);
R
Rob Clark 已提交
707 708 709 710 711 712
	if (IS_ERR(gpu->mmio)) {
		ret = PTR_ERR(gpu->mmio);
		goto fail;
	}

	/* Get Interrupt: */
713
	gpu->irq = platform_get_irq_byname(pdev, config->irqname);
R
Rob Clark 已提交
714 715 716 717 718 719 720 721 722 723 724 725 726
	if (gpu->irq < 0) {
		ret = gpu->irq;
		dev_err(drm->dev, "failed to get irq: %d\n", ret);
		goto fail;
	}

	ret = devm_request_irq(&pdev->dev, gpu->irq, irq_handler,
			IRQF_TRIGGER_HIGH, gpu->name, gpu);
	if (ret) {
		dev_err(drm->dev, "failed to request IRQ%u: %d\n", gpu->irq, ret);
		goto fail;
	}

727 728 729
	ret = get_clocks(pdev, gpu);
	if (ret)
		goto fail;
R
Rob Clark 已提交
730

731
	gpu->ebi1_clk = msm_clk_get(pdev, "bus");
R
Rob Clark 已提交
732 733 734 735 736 737 738 739 740 741 742 743 744 745 746
	DBG("ebi1_clk: %p", gpu->ebi1_clk);
	if (IS_ERR(gpu->ebi1_clk))
		gpu->ebi1_clk = NULL;

	/* Acquire regulators: */
	gpu->gpu_reg = devm_regulator_get(&pdev->dev, "vdd");
	DBG("gpu_reg: %p", gpu->gpu_reg);
	if (IS_ERR(gpu->gpu_reg))
		gpu->gpu_reg = NULL;

	gpu->gpu_cx = devm_regulator_get(&pdev->dev, "vddcx");
	DBG("gpu_cx: %p", gpu->gpu_cx);
	if (IS_ERR(gpu->gpu_cx))
		gpu->gpu_cx = NULL;

747 748 749 750
	gpu->pdev = pdev;
	platform_set_drvdata(pdev, gpu);

	bs_init(gpu);
751

752 753 754 755
	gpu->aspace = msm_gpu_create_address_space(gpu, pdev,
		config->va_start, config->va_end);

	if (gpu->aspace == NULL)
756
		dev_info(drm->dev, "%s: no IOMMU, fallback to VRAM carveout!\n", name);
757 758 759
	else if (IS_ERR(gpu->aspace)) {
		ret = PTR_ERR(gpu->aspace);
		goto fail;
R
Rob Clark 已提交
760
	}
761

762
	memptrs = msm_gem_kernel_new(drm, sizeof(*gpu->memptrs_bo),
763
		MSM_BO_UNCACHED, gpu->aspace, &gpu->memptrs_bo,
764
		&memptrs_iova);
765

766 767
	if (IS_ERR(memptrs)) {
		ret = PTR_ERR(memptrs);
768 769 770 771
		dev_err(drm->dev, "could not allocate memptrs: %d\n", ret);
		goto fail;
	}

772 773 774 775
	if (nr_rings > ARRAY_SIZE(gpu->rb)) {
		DRM_DEV_INFO_ONCE(drm->dev, "Only creating %lu ringbuffers\n",
			ARRAY_SIZE(gpu->rb));
		nr_rings = ARRAY_SIZE(gpu->rb);
R
Rob Clark 已提交
776 777
	}

778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794
	/* Create ringbuffer(s): */
	for (i = 0; i < nr_rings; i++) {
		gpu->rb[i] = msm_ringbuffer_new(gpu, i, memptrs, memptrs_iova);

		if (IS_ERR(gpu->rb[i])) {
			ret = PTR_ERR(gpu->rb[i]);
			dev_err(drm->dev,
				"could not create ringbuffer %d: %d\n", i, ret);
			goto fail;
		}

		memptrs += sizeof(struct msm_rbmemptrs);
		memptrs_iova += sizeof(struct msm_rbmemptrs);
	}

	gpu->nr_rings = nr_rings;

R
Rob Clark 已提交
795 796 797
	return 0;

fail:
798 799 800 801 802
	for (i = 0; i < ARRAY_SIZE(gpu->rb); i++)  {
		msm_ringbuffer_destroy(gpu->rb[i]);
		gpu->rb[i] = NULL;
	}

803 804 805 806 807 808
	if (gpu->memptrs_bo) {
		msm_gem_put_vaddr(gpu->memptrs_bo);
		msm_gem_put_iova(gpu->memptrs_bo, gpu->aspace);
		drm_gem_object_unreference_unlocked(gpu->memptrs_bo);
	}

809
	platform_set_drvdata(pdev, NULL);
R
Rob Clark 已提交
810 811 812 813 814
	return ret;
}

void msm_gpu_cleanup(struct msm_gpu *gpu)
{
815 816
	int i;

R
Rob Clark 已提交
817 818 819 820 821 822
	DBG("%s", gpu->name);

	WARN_ON(!list_empty(&gpu->active_list));

	bs_fini(gpu);

823 824 825
	for (i = 0; i < ARRAY_SIZE(gpu->rb); i++) {
		msm_ringbuffer_destroy(gpu->rb[i]);
		gpu->rb[i] = NULL;
R
Rob Clark 已提交
826
	}
827 828 829 830 831 832 833 834

	if (gpu->memptrs_bo) {
		msm_gem_put_vaddr(gpu->memptrs_bo);
		msm_gem_put_iova(gpu->memptrs_bo, gpu->aspace);
		drm_gem_object_unreference_unlocked(gpu->memptrs_bo);
	}

	if (!IS_ERR_OR_NULL(gpu->aspace)) {
835 836 837 838
		gpu->aspace->mmu->funcs->detach(gpu->aspace->mmu,
			NULL, 0);
		msm_gem_address_space_put(gpu->aspace);
	}
R
Rob Clark 已提交
839
}
新手
引导
客服 返回
顶部