ttm_bo.c 43.1 KB
Newer Older
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 28 29 30
/**************************************************************************
 *
 * Copyright (c) 2006-2009 VMware, Inc., Palo Alto, CA., USA
 * All Rights Reserved.
 *
 * Permission is hereby granted, free of charge, to any person obtaining a
 * copy of this software and associated documentation files (the
 * "Software"), to deal in the Software without restriction, including
 * without limitation the rights to use, copy, modify, merge, publish,
 * distribute, sub license, and/or sell copies of the Software, and to
 * permit persons to whom the Software is furnished to do so, subject to
 * the following conditions:
 *
 * The above copyright notice and this permission notice (including the
 * next paragraph) 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 NON-INFRINGEMENT. IN NO EVENT SHALL
 * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
 * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
 * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
 * USE OR OTHER DEALINGS IN THE SOFTWARE.
 *
 **************************************************************************/
/*
 * Authors: Thomas Hellstrom <thellstrom-at-vmware-dot-com>
 */

J
Joe Perches 已提交
31 32
#define pr_fmt(fmt) "[TTM] " fmt

33 34 35
#include <drm/ttm/ttm_module.h>
#include <drm/ttm/ttm_bo_driver.h>
#include <drm/ttm/ttm_placement.h>
36 37 38 39 40 41
#include <linux/jiffies.h>
#include <linux/slab.h>
#include <linux/sched.h>
#include <linux/mm.h>
#include <linux/file.h>
#include <linux/module.h>
A
Arun Sharma 已提交
42
#include <linux/atomic.h>
43
#include <linux/reservation.h>
44 45 46 47 48 49

#define TTM_ASSERT_LOCKED(param)
#define TTM_DEBUG(fmt, arg...)
#define TTM_BO_HASH_ORDER 13

static int ttm_bo_swapout(struct ttm_mem_shrink *shrink);
50 51 52 53 54 55 56
static void ttm_bo_global_kobj_release(struct kobject *kobj);

static struct attribute ttm_bo_count = {
	.name = "bo_count",
	.mode = S_IRUGO
};

57 58
static inline int ttm_mem_type_from_place(const struct ttm_place *place,
					  uint32_t *mem_type)
59
{
60
	int pos;
61

62 63 64 65 66 67
	pos = ffs(place->flags & TTM_PL_MASK_MEM);
	if (unlikely(!pos))
		return -EINVAL;

	*mem_type = pos - 1;
	return 0;
68 69
}

70
static void ttm_mem_type_debug(struct ttm_bo_device *bdev, int mem_type)
71
{
72 73
	struct ttm_mem_type_manager *man = &bdev->man[mem_type];

J
Joe Perches 已提交
74 75 76
	pr_err("    has_type: %d\n", man->has_type);
	pr_err("    use_type: %d\n", man->use_type);
	pr_err("    flags: 0x%08X\n", man->flags);
77
	pr_err("    gpu_offset: 0x%08llX\n", man->gpu_offset);
J
Joe Perches 已提交
78 79 80
	pr_err("    size: %llu\n", man->size);
	pr_err("    available_caching: 0x%08X\n", man->available_caching);
	pr_err("    default_caching: 0x%08X\n", man->default_caching);
81 82
	if (mem_type != TTM_PL_SYSTEM)
		(*man->func->debug)(man, TTM_PFX);
83 84 85 86 87 88 89
}

static void ttm_bo_mem_space_debug(struct ttm_buffer_object *bo,
					struct ttm_placement *placement)
{
	int i, ret, mem_type;

J
Joe Perches 已提交
90 91 92
	pr_err("No space for %p (%lu pages, %luK, %luM)\n",
	       bo, bo->mem.num_pages, bo->mem.size >> 10,
	       bo->mem.size >> 20);
93
	for (i = 0; i < placement->num_placement; i++) {
94
		ret = ttm_mem_type_from_place(&placement->placement[i],
95 96 97
						&mem_type);
		if (ret)
			return;
J
Joe Perches 已提交
98
		pr_err("  placement[%d]=0x%08X (%d)\n",
99
		       i, placement->placement[i].flags, mem_type);
100
		ttm_mem_type_debug(bo->bdev, mem_type);
101 102 103
	}
}

104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119
static ssize_t ttm_bo_global_show(struct kobject *kobj,
				  struct attribute *attr,
				  char *buffer)
{
	struct ttm_bo_global *glob =
		container_of(kobj, struct ttm_bo_global, kobj);

	return snprintf(buffer, PAGE_SIZE, "%lu\n",
			(unsigned long) atomic_read(&glob->bo_count));
}

static struct attribute *ttm_bo_global_attrs[] = {
	&ttm_bo_count,
	NULL
};

120
static const struct sysfs_ops ttm_bo_global_ops = {
121 122 123 124 125 126 127 128 129
	.show = &ttm_bo_global_show
};

static struct kobj_type ttm_bo_glob_kobj_type  = {
	.release = &ttm_bo_global_kobj_release,
	.sysfs_ops = &ttm_bo_global_ops,
	.default_attrs = ttm_bo_global_attrs
};

130 131 132 133 134 135 136 137 138 139 140

static inline uint32_t ttm_bo_type_flags(unsigned type)
{
	return 1 << (type);
}

static void ttm_bo_release_list(struct kref *list_kref)
{
	struct ttm_buffer_object *bo =
	    container_of(list_kref, struct ttm_buffer_object, list_kref);
	struct ttm_bo_device *bdev = bo->bdev;
141
	size_t acc_size = bo->acc_size;
142 143 144 145 146 147 148

	BUG_ON(atomic_read(&bo->list_kref.refcount));
	BUG_ON(atomic_read(&bo->kref.refcount));
	BUG_ON(atomic_read(&bo->cpu_writers));
	BUG_ON(bo->mem.mm_node != NULL);
	BUG_ON(!list_empty(&bo->lru));
	BUG_ON(!list_empty(&bo->ddestroy));
149
	ttm_tt_destroy(bo->ttm);
150
	atomic_dec(&bo->glob->bo_count);
151
	dma_fence_put(bo->moving);
152 153
	if (bo->resv == &bo->ttm_resv)
		reservation_object_fini(&bo->ttm_resv);
154
	mutex_destroy(&bo->wu_mutex);
155 156 157 158 159
	if (bo->destroy)
		bo->destroy(bo);
	else {
		kfree(bo);
	}
160
	ttm_mem_global_free(bdev->glob->mem_glob, acc_size);
161 162
}

163
void ttm_bo_add_to_lru(struct ttm_buffer_object *bo)
164 165 166
{
	struct ttm_bo_device *bdev = bo->bdev;

167
	lockdep_assert_held(&bo->resv->lock.base);
168 169 170 171 172

	if (!(bo->mem.placement & TTM_PL_FLAG_NO_EVICT)) {

		BUG_ON(!list_empty(&bo->lru));

173
		list_add(&bo->lru, bdev->driver->lru_tail(bo));
174 175
		kref_get(&bo->list_kref);

176
		if (bo->ttm && !(bo->ttm->page_flags & TTM_PAGE_FLAG_SG)) {
177
			list_add(&bo->swap, bdev->driver->swap_lru_tail(bo));
178 179 180 181
			kref_get(&bo->list_kref);
		}
	}
}
182
EXPORT_SYMBOL(ttm_bo_add_to_lru);
183

184
int ttm_bo_del_from_lru(struct ttm_buffer_object *bo)
185
{
186
	struct ttm_bo_device *bdev = bo->bdev;
187 188
	int put_count = 0;

189 190 191
	if (bdev->driver->lru_removal)
		bdev->driver->lru_removal(bo);

192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208
	if (!list_empty(&bo->swap)) {
		list_del_init(&bo->swap);
		++put_count;
	}
	if (!list_empty(&bo->lru)) {
		list_del_init(&bo->lru);
		++put_count;
	}

	return put_count;
}

static void ttm_bo_ref_bug(struct kref *list_kref)
{
	BUG();
}

209 210 211
void ttm_bo_list_ref_sub(struct ttm_buffer_object *bo, int count,
			 bool never_free)
{
212 213
	kref_sub(&bo->list_kref, count,
		 (never_free) ? ttm_bo_ref_bug : ttm_bo_release_list);
214 215
}

216
void ttm_bo_del_sub_from_lru(struct ttm_buffer_object *bo)
217
{
218
	int put_count;
219

220 221 222 223
	spin_lock(&bo->glob->lru_lock);
	put_count = ttm_bo_del_from_lru(bo);
	spin_unlock(&bo->glob->lru_lock);
	ttm_bo_list_ref_sub(bo, put_count, true);
224
}
225
EXPORT_SYMBOL(ttm_bo_del_sub_from_lru);
226

227 228 229
void ttm_bo_move_to_lru_tail(struct ttm_buffer_object *bo)
{
	struct ttm_bo_device *bdev = bo->bdev;
230
	int put_count = 0;
231 232 233

	lockdep_assert_held(&bo->resv->lock.base);

234 235 236
	if (bdev->driver->lru_removal)
		bdev->driver->lru_removal(bo);

237 238 239
	put_count = ttm_bo_del_from_lru(bo);
	ttm_bo_list_ref_sub(bo, put_count, true);
	ttm_bo_add_to_lru(bo);
240 241 242
}
EXPORT_SYMBOL(ttm_bo_move_to_lru_tail);

243 244
struct list_head *ttm_bo_default_lru_tail(struct ttm_buffer_object *bo)
{
245
	return bo->bdev->man[bo->mem.mem_type].lru[bo->priority].prev;
246 247 248 249 250
}
EXPORT_SYMBOL(ttm_bo_default_lru_tail);

struct list_head *ttm_bo_default_swap_lru_tail(struct ttm_buffer_object *bo)
{
251
	return bo->glob->swap_lru[bo->priority].prev;
252 253 254
}
EXPORT_SYMBOL(ttm_bo_default_swap_lru_tail);

255 256 257 258 259 260
/*
 * Call bo->mutex locked.
 */
static int ttm_bo_add_ttm(struct ttm_buffer_object *bo, bool zero_alloc)
{
	struct ttm_bo_device *bdev = bo->bdev;
261
	struct ttm_bo_global *glob = bo->glob;
262 263 264 265 266 267
	int ret = 0;
	uint32_t page_flags = 0;

	TTM_ASSERT_LOCKED(&bo->mutex);
	bo->ttm = NULL;

D
Dave Airlie 已提交
268 269 270
	if (bdev->need_dma32)
		page_flags |= TTM_PAGE_FLAG_DMA32;

271 272 273 274 275
	switch (bo->type) {
	case ttm_bo_type_device:
		if (zero_alloc)
			page_flags |= TTM_PAGE_FLAG_ZERO_ALLOC;
	case ttm_bo_type_kernel:
276 277
		bo->ttm = bdev->driver->ttm_tt_create(bdev, bo->num_pages << PAGE_SHIFT,
						      page_flags, glob->dummy_read_page);
278 279 280
		if (unlikely(bo->ttm == NULL))
			ret = -ENOMEM;
		break;
281 282 283 284 285 286 287 288 289 290
	case ttm_bo_type_sg:
		bo->ttm = bdev->driver->ttm_tt_create(bdev, bo->num_pages << PAGE_SHIFT,
						      page_flags | TTM_PAGE_FLAG_SG,
						      glob->dummy_read_page);
		if (unlikely(bo->ttm == NULL)) {
			ret = -ENOMEM;
			break;
		}
		bo->ttm->sg = bo->sg;
		break;
291
	default:
J
Joe Perches 已提交
292
		pr_err("Illegal buffer object type\n");
293 294 295 296 297 298 299 300 301
		ret = -EINVAL;
		break;
	}

	return ret;
}

static int ttm_bo_handle_move_mem(struct ttm_buffer_object *bo,
				  struct ttm_mem_reg *mem,
302
				  bool evict, bool interruptible,
303
				  bool no_wait_gpu)
304 305 306 307 308 309 310 311 312
{
	struct ttm_bo_device *bdev = bo->bdev;
	bool old_is_pci = ttm_mem_reg_is_pci(bdev, &bo->mem);
	bool new_is_pci = ttm_mem_reg_is_pci(bdev, mem);
	struct ttm_mem_type_manager *old_man = &bdev->man[bo->mem.mem_type];
	struct ttm_mem_type_manager *new_man = &bdev->man[mem->mem_type];
	int ret = 0;

	if (old_is_pci || new_is_pci ||
313 314 315 316 317 318 319
	    ((mem->placement & bo->mem.placement & TTM_PL_MASK_CACHING) == 0)) {
		ret = ttm_mem_io_lock(old_man, true);
		if (unlikely(ret != 0))
			goto out_err;
		ttm_bo_unmap_virtual_locked(bo);
		ttm_mem_io_unlock(old_man);
	}
320 321 322 323 324

	/*
	 * Create and bind a ttm if required.
	 */

325 326
	if (!(new_man->flags & TTM_MEMTYPE_FLAG_FIXED)) {
		if (bo->ttm == NULL) {
327 328
			bool zero = !(old_man->flags & TTM_MEMTYPE_FLAG_FIXED);
			ret = ttm_bo_add_ttm(bo, zero);
329 330 331
			if (ret)
				goto out_err;
		}
332 333 334

		ret = ttm_tt_set_placement_caching(bo->ttm, mem->placement);
		if (ret)
335
			goto out_err;
336 337 338 339 340 341 342 343

		if (mem->mem_type != TTM_PL_SYSTEM) {
			ret = ttm_tt_bind(bo->ttm, mem);
			if (ret)
				goto out_err;
		}

		if (bo->mem.mem_type == TTM_PL_SYSTEM) {
344
			if (bdev->driver->move_notify)
345
				bdev->driver->move_notify(bo, evict, mem);
346
			bo->mem = *mem;
347 348 349 350 351
			mem->mm_node = NULL;
			goto moved;
		}
	}

352
	if (bdev->driver->move_notify)
353
		bdev->driver->move_notify(bo, evict, mem);
354

355 356
	if (!(old_man->flags & TTM_MEMTYPE_FLAG_FIXED) &&
	    !(new_man->flags & TTM_MEMTYPE_FLAG_FIXED))
357
		ret = ttm_bo_move_ttm(bo, interruptible, no_wait_gpu, mem);
358 359
	else if (bdev->driver->move)
		ret = bdev->driver->move(bo, evict, interruptible,
360
					 no_wait_gpu, mem);
361
	else
362
		ret = ttm_bo_move_memcpy(bo, interruptible, no_wait_gpu, mem);
363

364 365 366 367 368
	if (ret) {
		if (bdev->driver->move_notify) {
			struct ttm_mem_reg tmp_mem = *mem;
			*mem = bo->mem;
			bo->mem = tmp_mem;
369
			bdev->driver->move_notify(bo, false, mem);
370
			bo->mem = *mem;
371
			*mem = tmp_mem;
372
		}
373

374 375
		goto out_err;
	}
376

377 378
moved:
	if (bo->evicted) {
379 380 381 382 383
		if (bdev->driver->invalidate_caches) {
			ret = bdev->driver->invalidate_caches(bdev, bo->mem.placement);
			if (ret)
				pr_err("Can not flush read caches\n");
		}
384 385 386 387
		bo->evicted = false;
	}

	if (bo->mem.mm_node) {
388
		bo->offset = (bo->mem.start << PAGE_SHIFT) +
389 390
		    bdev->man[bo->mem.mem_type].gpu_offset;
		bo->cur_placement = bo->mem.placement;
391 392
	} else
		bo->offset = 0;
393 394 395 396 397

	return 0;

out_err:
	new_man = &bdev->man[bo->mem.mem_type];
398
	if (new_man->flags & TTM_MEMTYPE_FLAG_FIXED) {
399 400 401 402 403 404 405
		ttm_tt_destroy(bo->ttm);
		bo->ttm = NULL;
	}

	return ret;
}

406
/**
407
 * Call bo::reserved.
408
 * Will release GPU memory type usage on destruction.
409 410 411
 * This is the place to put in driver specific hooks to release
 * driver private resources.
 * Will release the bo::reserved lock.
412 413 414 415
 */

static void ttm_bo_cleanup_memtype_use(struct ttm_buffer_object *bo)
{
416
	if (bo->bdev->driver->move_notify)
417
		bo->bdev->driver->move_notify(bo, false, NULL);
418

419 420
	ttm_tt_destroy(bo->ttm);
	bo->ttm = NULL;
421
	ttm_bo_mem_put(bo, &bo->mem);
422

423
	ww_mutex_unlock (&bo->resv->lock);
424 425
}

426 427 428
static void ttm_bo_flush_all_fences(struct ttm_buffer_object *bo)
{
	struct reservation_object_list *fobj;
429
	struct dma_fence *fence;
430 431 432 433 434
	int i;

	fobj = reservation_object_get_list(bo->resv);
	fence = reservation_object_get_excl(bo->resv);
	if (fence && !fence->ops->signaled)
435
		dma_fence_enable_sw_signaling(fence);
436 437 438 439 440 441

	for (i = 0; fobj && i < fobj->shared_count; ++i) {
		fence = rcu_dereference_protected(fobj->shared[i],
					reservation_object_held(bo->resv));

		if (!fence->ops->signaled)
442
			dma_fence_enable_sw_signaling(fence);
443 444 445
	}
}

446
static void ttm_bo_cleanup_refs_or_queue(struct ttm_buffer_object *bo)
447 448
{
	struct ttm_bo_device *bdev = bo->bdev;
449
	struct ttm_bo_global *glob = bo->glob;
450
	int put_count;
451 452
	int ret;

453
	spin_lock(&glob->lru_lock);
454
	ret = __ttm_bo_reserve(bo, false, true, NULL);
455

M
Maarten Lankhorst 已提交
456
	if (!ret) {
457
		if (!ttm_bo_wait(bo, false, true)) {
M
Maarten Lankhorst 已提交
458
			put_count = ttm_bo_del_from_lru(bo);
459

M
Maarten Lankhorst 已提交
460 461
			spin_unlock(&glob->lru_lock);
			ttm_bo_cleanup_memtype_use(bo);
462

M
Maarten Lankhorst 已提交
463
			ttm_bo_list_ref_sub(bo, put_count, true);
464

M
Maarten Lankhorst 已提交
465
			return;
466 467
		} else
			ttm_bo_flush_all_fences(bo);
468 469 470 471 472 473 474 475 476 477 478

		/*
		 * Make NO_EVICT bos immediately available to
		 * shrinkers, now that they are queued for
		 * destruction.
		 */
		if (bo->mem.placement & TTM_PL_FLAG_NO_EVICT) {
			bo->mem.placement &= ~TTM_PL_FLAG_NO_EVICT;
			ttm_bo_add_to_lru(bo);
		}

479
		__ttm_bo_unreserve(bo);
480
	}
481 482 483 484 485 486 487 488 489 490

	kref_get(&bo->list_kref);
	list_add_tail(&bo->ddestroy, &bdev->ddestroy);
	spin_unlock(&glob->lru_lock);

	schedule_delayed_work(&bdev->wq,
			      ((HZ / 100) < 1) ? 1 : HZ / 100);
}

/**
491
 * function ttm_bo_cleanup_refs_and_unlock
492 493 494
 * If bo idle, remove from delayed- and lru lists, and unref.
 * If not idle, do nothing.
 *
495 496 497
 * Must be called with lru_lock and reservation held, this function
 * will drop both before returning.
 *
498 499 500 501
 * @interruptible         Any sleeps should occur interruptibly.
 * @no_wait_gpu           Never wait for gpu. Return -EBUSY instead.
 */

502 503 504
static int ttm_bo_cleanup_refs_and_unlock(struct ttm_buffer_object *bo,
					  bool interruptible,
					  bool no_wait_gpu)
505 506 507
{
	struct ttm_bo_global *glob = bo->glob;
	int put_count;
508
	int ret;
509

510
	ret = ttm_bo_wait(bo, false, true);
511

512
	if (ret && !no_wait_gpu) {
M
Maarten Lankhorst 已提交
513 514 515 516 517 518 519 520 521 522 523 524 525
		long lret;
		ww_mutex_unlock(&bo->resv->lock);
		spin_unlock(&glob->lru_lock);

		lret = reservation_object_wait_timeout_rcu(bo->resv,
							   true,
							   interruptible,
							   30 * HZ);

		if (lret < 0)
			return lret;
		else if (lret == 0)
			return -EBUSY;
526

527
		spin_lock(&glob->lru_lock);
528
		ret = __ttm_bo_reserve(bo, false, true, NULL);
529

530 531 532 533 534 535 536 537 538 539 540 541
		/*
		 * We raced, and lost, someone else holds the reservation now,
		 * and is probably busy in ttm_bo_cleanup_memtype_use.
		 *
		 * Even if it's not the case, because we finished waiting any
		 * delayed destruction would succeed, so just return success
		 * here.
		 */
		if (ret) {
			spin_unlock(&glob->lru_lock);
			return 0;
		}
542 543 544 545 546

		/*
		 * remove sync_obj with ttm_bo_wait, the wait should be
		 * finished, and no new wait object should have been added.
		 */
547
		ret = ttm_bo_wait(bo, false, true);
548 549
		WARN_ON(ret);
	}
550

551
	if (ret || unlikely(list_empty(&bo->ddestroy))) {
552
		__ttm_bo_unreserve(bo);
553
		spin_unlock(&glob->lru_lock);
554
		return ret;
555 556
	}

557 558 559 560 561 562 563
	put_count = ttm_bo_del_from_lru(bo);
	list_del_init(&bo->ddestroy);
	++put_count;

	spin_unlock(&glob->lru_lock);
	ttm_bo_cleanup_memtype_use(bo);

564
	ttm_bo_list_ref_sub(bo, put_count, true);
565 566

	return 0;
567 568 569 570 571 572 573 574 575
}

/**
 * Traverse the delayed list, and call ttm_bo_cleanup_refs on all
 * encountered buffers.
 */

static int ttm_bo_delayed_delete(struct ttm_bo_device *bdev, bool remove_all)
{
576
	struct ttm_bo_global *glob = bdev->glob;
577 578
	struct ttm_buffer_object *entry = NULL;
	int ret = 0;
579

580
	spin_lock(&glob->lru_lock);
581 582 583 584 585 586 587 588 589 590 591 592 593
	if (list_empty(&bdev->ddestroy))
		goto out_unlock;

	entry = list_first_entry(&bdev->ddestroy,
		struct ttm_buffer_object, ddestroy);
	kref_get(&entry->list_kref);

	for (;;) {
		struct ttm_buffer_object *nentry = NULL;

		if (entry->ddestroy.next != &bdev->ddestroy) {
			nentry = list_first_entry(&entry->ddestroy,
				struct ttm_buffer_object, ddestroy);
594 595 596
			kref_get(&nentry->list_kref);
		}

597
		ret = __ttm_bo_reserve(entry, false, true, NULL);
598 599
		if (remove_all && ret) {
			spin_unlock(&glob->lru_lock);
600
			ret = __ttm_bo_reserve(entry, false, false, NULL);
601 602 603
			spin_lock(&glob->lru_lock);
		}

604 605 606 607 608 609
		if (!ret)
			ret = ttm_bo_cleanup_refs_and_unlock(entry, false,
							     !remove_all);
		else
			spin_unlock(&glob->lru_lock);

610
		kref_put(&entry->list_kref, ttm_bo_release_list);
611 612 613 614
		entry = nentry;

		if (ret || !entry)
			goto out;
615

616
		spin_lock(&glob->lru_lock);
617
		if (list_empty(&entry->ddestroy))
618 619 620
			break;
	}

621 622 623 624 625
out_unlock:
	spin_unlock(&glob->lru_lock);
out:
	if (entry)
		kref_put(&entry->list_kref, ttm_bo_release_list);
626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644
	return ret;
}

static void ttm_bo_delayed_workqueue(struct work_struct *work)
{
	struct ttm_bo_device *bdev =
	    container_of(work, struct ttm_bo_device, wq.work);

	if (ttm_bo_delayed_delete(bdev, false)) {
		schedule_delayed_work(&bdev->wq,
				      ((HZ / 100) < 1) ? 1 : HZ / 100);
	}
}

static void ttm_bo_release(struct kref *kref)
{
	struct ttm_buffer_object *bo =
	    container_of(kref, struct ttm_buffer_object, kref);
	struct ttm_bo_device *bdev = bo->bdev;
645
	struct ttm_mem_type_manager *man = &bdev->man[bo->mem.mem_type];
646

647
	drm_vma_offset_remove(&bdev->vma_manager, &bo->vma_node);
648 649 650
	ttm_mem_io_lock(man, false);
	ttm_mem_io_free_vm(bo);
	ttm_mem_io_unlock(man);
651
	ttm_bo_cleanup_refs_or_queue(bo);
652 653 654 655 656 657 658 659 660 661 662 663
	kref_put(&bo->list_kref, ttm_bo_release_list);
}

void ttm_bo_unref(struct ttm_buffer_object **p_bo)
{
	struct ttm_buffer_object *bo = *p_bo;

	*p_bo = NULL;
	kref_put(&bo->kref, ttm_bo_release);
}
EXPORT_SYMBOL(ttm_bo_unref);

664 665 666 667 668 669 670 671 672 673 674 675 676 677
int ttm_bo_lock_delayed_workqueue(struct ttm_bo_device *bdev)
{
	return cancel_delayed_work_sync(&bdev->wq);
}
EXPORT_SYMBOL(ttm_bo_lock_delayed_workqueue);

void ttm_bo_unlock_delayed_workqueue(struct ttm_bo_device *bdev, int resched)
{
	if (resched)
		schedule_delayed_work(&bdev->wq,
				      ((HZ / 100) < 1) ? 1 : HZ / 100);
}
EXPORT_SYMBOL(ttm_bo_unlock_delayed_workqueue);

678
static int ttm_bo_evict(struct ttm_buffer_object *bo, bool interruptible,
679
			bool no_wait_gpu)
680 681 682
{
	struct ttm_bo_device *bdev = bo->bdev;
	struct ttm_mem_reg evict_mem;
683 684
	struct ttm_placement placement;
	int ret = 0;
685

686
	lockdep_assert_held(&bo->resv->lock.base);
687 688 689

	evict_mem = bo->mem;
	evict_mem.mm_node = NULL;
690 691
	evict_mem.bus.io_reserved_vm = false;
	evict_mem.bus.io_reserved_count = 0;
692

693 694
	placement.num_placement = 0;
	placement.num_busy_placement = 0;
695 696
	bdev->driver->evict_flags(bo, &placement);
	ret = ttm_bo_mem_space(bo, &placement, &evict_mem, interruptible,
697
				no_wait_gpu);
698
	if (ret) {
699
		if (ret != -ERESTARTSYS) {
J
Joe Perches 已提交
700 701
			pr_err("Failed to find memory space for buffer 0x%p eviction\n",
			       bo);
702 703
			ttm_bo_mem_space_debug(bo, &placement);
		}
704 705 706 707
		goto out;
	}

	ret = ttm_bo_handle_move_mem(bo, &evict_mem, true, interruptible,
708
				     no_wait_gpu);
709
	if (unlikely(ret)) {
710
		if (ret != -ERESTARTSYS)
J
Joe Perches 已提交
711
			pr_err("Buffer eviction failed\n");
712
		ttm_bo_mem_put(bo, &evict_mem);
713 714
		goto out;
	}
715 716 717 718 719
	bo->evicted = true;
out:
	return ret;
}

720 721 722 723 724 725 726 727 728 729 730 731 732 733
bool ttm_bo_eviction_valuable(struct ttm_buffer_object *bo,
			      const struct ttm_place *place)
{
	/* Don't evict this BO if it's outside of the
	 * requested placement range
	 */
	if (place->fpfn >= (bo->mem.start + bo->mem.size) ||
	    (place->lpfn && place->lpfn <= bo->mem.start))
		return false;

	return true;
}
EXPORT_SYMBOL(ttm_bo_eviction_valuable);

734 735
static int ttm_mem_evict_first(struct ttm_bo_device *bdev,
				uint32_t mem_type,
736
				const struct ttm_place *place,
737
				bool interruptible,
738
				bool no_wait_gpu)
739 740 741 742
{
	struct ttm_bo_global *glob = bdev->glob;
	struct ttm_mem_type_manager *man = &bdev->man[mem_type];
	struct ttm_buffer_object *bo;
743
	int ret = -EBUSY, put_count;
744
	unsigned i;
745

746
	spin_lock(&glob->lru_lock);
747 748 749 750 751
	for (i = 0; i < TTM_MAX_BO_PRIORITY; ++i) {
		list_for_each_entry(bo, &man->lru[i], lru) {
			ret = __ttm_bo_reserve(bo, false, true, NULL);
			if (ret)
				continue;
752

753 754 755 756 757 758 759 760
			if (place && !bdev->driver->eviction_valuable(bo,
								      place)) {
				__ttm_bo_unreserve(bo);
				ret = -EBUSY;
				continue;
			}

			break;
761
		}
762

763 764
		if (!ret)
			break;
765 766 767
	}

	if (ret) {
768
		spin_unlock(&glob->lru_lock);
769
		return ret;
770 771
	}

772
	kref_get(&bo->list_kref);
773

774
	if (!list_empty(&bo->ddestroy)) {
775 776
		ret = ttm_bo_cleanup_refs_and_unlock(bo, interruptible,
						     no_wait_gpu);
777
		kref_put(&bo->list_kref, ttm_bo_release_list);
778
		return ret;
779 780
	}

781
	put_count = ttm_bo_del_from_lru(bo);
782
	spin_unlock(&glob->lru_lock);
783 784 785

	BUG_ON(ret != 0);

786
	ttm_bo_list_ref_sub(bo, put_count, true);
787

788
	ret = ttm_bo_evict(bo, interruptible, no_wait_gpu);
789
	ttm_bo_unreserve(bo);
790

791
	kref_put(&bo->list_kref, ttm_bo_release_list);
792 793 794
	return ret;
}

795 796
void ttm_bo_mem_put(struct ttm_buffer_object *bo, struct ttm_mem_reg *mem)
{
797
	struct ttm_mem_type_manager *man = &bo->bdev->man[mem->mem_type];
798

799 800
	if (mem->mm_node)
		(*man->func->put_node)(man, mem);
801 802 803
}
EXPORT_SYMBOL(ttm_bo_mem_put);

804 805 806 807 808 809 810
/**
 * Add the last move fence to the BO and reserve a new shared slot.
 */
static int ttm_bo_add_move_fence(struct ttm_buffer_object *bo,
				 struct ttm_mem_type_manager *man,
				 struct ttm_mem_reg *mem)
{
811
	struct dma_fence *fence;
812 813 814
	int ret;

	spin_lock(&man->move_lock);
815
	fence = dma_fence_get(man->move);
816 817 818 819 820 821 822 823 824
	spin_unlock(&man->move_lock);

	if (fence) {
		reservation_object_add_shared_fence(bo->resv, fence);

		ret = reservation_object_reserve_shared(bo->resv);
		if (unlikely(ret))
			return ret;

825
		dma_fence_put(bo->moving);
826 827 828 829 830 831
		bo->moving = fence;
	}

	return 0;
}

832 833 834 835
/**
 * Repeatedly evict memory from the LRU for @mem_type until we create enough
 * space, or we've evicted everything and there isn't enough space.
 */
836 837
static int ttm_bo_mem_force_space(struct ttm_buffer_object *bo,
					uint32_t mem_type,
838
					const struct ttm_place *place,
839
					struct ttm_mem_reg *mem,
840 841
					bool interruptible,
					bool no_wait_gpu)
842
{
843
	struct ttm_bo_device *bdev = bo->bdev;
844 845 846 847
	struct ttm_mem_type_manager *man = &bdev->man[mem_type];
	int ret;

	do {
848
		ret = (*man->func->get_node)(man, bo, place, mem);
849 850
		if (unlikely(ret != 0))
			return ret;
851
		if (mem->mm_node)
852
			break;
853
		ret = ttm_mem_evict_first(bdev, mem_type, place,
854
					  interruptible, no_wait_gpu);
855 856 857 858
		if (unlikely(ret != 0))
			return ret;
	} while (1);
	mem->mem_type = mem_type;
859
	return ttm_bo_add_move_fence(bo, man, mem);
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
static uint32_t ttm_bo_select_caching(struct ttm_mem_type_manager *man,
				      uint32_t cur_placement,
				      uint32_t proposed_placement)
{
	uint32_t caching = proposed_placement & TTM_PL_MASK_CACHING;
	uint32_t result = proposed_placement & ~TTM_PL_MASK_CACHING;

	/**
	 * Keep current caching if possible.
	 */

	if ((cur_placement & caching) != 0)
		result |= (cur_placement & caching);
	else if ((man->default_caching & caching) != 0)
		result |= man->default_caching;
	else if ((TTM_PL_FLAG_CACHED & caching) != 0)
		result |= TTM_PL_FLAG_CACHED;
	else if ((TTM_PL_FLAG_WC & caching) != 0)
		result |= TTM_PL_FLAG_WC;
	else if ((TTM_PL_FLAG_UNCACHED & caching) != 0)
		result |= TTM_PL_FLAG_UNCACHED;

	return result;
}

887 888
static bool ttm_bo_mt_compatible(struct ttm_mem_type_manager *man,
				 uint32_t mem_type,
889
				 const struct ttm_place *place,
890
				 uint32_t *masked_placement)
891 892 893
{
	uint32_t cur_flags = ttm_bo_type_flags(mem_type);

894
	if ((cur_flags & place->flags & TTM_PL_MASK_MEM) == 0)
895 896
		return false;

897
	if ((place->flags & man->available_caching) == 0)
898 899
		return false;

900
	cur_flags |= (place->flags & man->available_caching);
901 902

	*masked_placement = cur_flags;
903 904 905 906 907 908 909 910 911 912 913 914
	return true;
}

/**
 * Creates space for memory region @mem according to its type.
 *
 * This function first searches for free space in compatible memory types in
 * the priority order defined by the driver.  If free space isn't found, then
 * ttm_bo_mem_force_space is attempted in priority order to evict and find
 * space.
 */
int ttm_bo_mem_space(struct ttm_buffer_object *bo,
915 916
			struct ttm_placement *placement,
			struct ttm_mem_reg *mem,
917
			bool interruptible,
918
			bool no_wait_gpu)
919 920 921 922 923 924 925
{
	struct ttm_bo_device *bdev = bo->bdev;
	struct ttm_mem_type_manager *man;
	uint32_t mem_type = TTM_PL_SYSTEM;
	uint32_t cur_flags = 0;
	bool type_found = false;
	bool type_ok = false;
926
	bool has_erestartsys = false;
927
	int i, ret;
928

929 930 931 932
	ret = reservation_object_reserve_shared(bo->resv);
	if (unlikely(ret))
		return ret;

933
	mem->mm_node = NULL;
934
	for (i = 0; i < placement->num_placement; ++i) {
935 936 937
		const struct ttm_place *place = &placement->placement[i];

		ret = ttm_mem_type_from_place(place, &mem_type);
938 939
		if (ret)
			return ret;
940
		man = &bdev->man[mem_type];
941 942
		if (!man->has_type || !man->use_type)
			continue;
943

944
		type_ok = ttm_bo_mt_compatible(man, mem_type, place,
945
						&cur_flags);
946 947 948 949

		if (!type_ok)
			continue;

950
		type_found = true;
951 952
		cur_flags = ttm_bo_select_caching(man, bo->mem.placement,
						  cur_flags);
953 954 955 956
		/*
		 * Use the access and other non-mapping-related flag bits from
		 * the memory placement flags to the current flags
		 */
957
		ttm_flag_masked(&cur_flags, place->flags,
958
				~TTM_PL_MASK_MEMTYPE);
959

960 961 962
		if (mem_type == TTM_PL_SYSTEM)
			break;

963 964 965
		ret = (*man->func->get_node)(man, bo, place, mem);
		if (unlikely(ret))
			return ret;
966 967 968 969 970 971 972

		if (mem->mm_node) {
			ret = ttm_bo_add_move_fence(bo, man, mem);
			if (unlikely(ret)) {
				(*man->func->put_node)(man, mem);
				return ret;
			}
973
			break;
974
		}
975 976
	}

977
	if ((type_ok && (mem_type == TTM_PL_SYSTEM)) || mem->mm_node) {
978 979 980 981 982
		mem->mem_type = mem_type;
		mem->placement = cur_flags;
		return 0;
	}

983
	for (i = 0; i < placement->num_busy_placement; ++i) {
984 985 986
		const struct ttm_place *place = &placement->busy_placement[i];

		ret = ttm_mem_type_from_place(place, &mem_type);
987 988
		if (ret)
			return ret;
989
		man = &bdev->man[mem_type];
990
		if (!man->has_type || !man->use_type)
991
			continue;
992
		if (!ttm_bo_mt_compatible(man, mem_type, place, &cur_flags))
993 994
			continue;

995
		type_found = true;
996 997
		cur_flags = ttm_bo_select_caching(man, bo->mem.placement,
						  cur_flags);
998 999 1000 1001
		/*
		 * Use the access and other non-mapping-related flag bits from
		 * the memory placement flags to the current flags
		 */
1002
		ttm_flag_masked(&cur_flags, place->flags,
1003
				~TTM_PL_MASK_MEMTYPE);
1004

1005 1006 1007 1008 1009 1010 1011
		if (mem_type == TTM_PL_SYSTEM) {
			mem->mem_type = mem_type;
			mem->placement = cur_flags;
			mem->mm_node = NULL;
			return 0;
		}

1012
		ret = ttm_bo_mem_force_space(bo, mem_type, place, mem,
1013
						interruptible, no_wait_gpu);
1014 1015 1016 1017
		if (ret == 0 && mem->mm_node) {
			mem->placement = cur_flags;
			return 0;
		}
1018 1019
		if (ret == -ERESTARTSYS)
			has_erestartsys = true;
1020
	}
1021 1022 1023 1024 1025 1026 1027

	if (!type_found) {
		printk(KERN_ERR TTM_PFX "No compatible memory type found.\n");
		return -EINVAL;
	}

	return (has_erestartsys) ? -ERESTARTSYS : -ENOMEM;
1028 1029 1030
}
EXPORT_SYMBOL(ttm_bo_mem_space);

1031
static int ttm_bo_move_buffer(struct ttm_buffer_object *bo,
1032
			struct ttm_placement *placement,
1033
			bool interruptible,
1034
			bool no_wait_gpu)
1035 1036 1037 1038
{
	int ret = 0;
	struct ttm_mem_reg mem;

1039
	lockdep_assert_held(&bo->resv->lock.base);
1040 1041 1042 1043

	mem.num_pages = bo->num_pages;
	mem.size = mem.num_pages << PAGE_SHIFT;
	mem.page_alignment = bo->mem.page_alignment;
1044 1045
	mem.bus.io_reserved_vm = false;
	mem.bus.io_reserved_count = 0;
1046 1047 1048
	/*
	 * Determine where to move the buffer.
	 */
1049 1050
	ret = ttm_bo_mem_space(bo, placement, &mem,
			       interruptible, no_wait_gpu);
1051 1052
	if (ret)
		goto out_unlock;
1053 1054
	ret = ttm_bo_handle_move_mem(bo, &mem, false,
				     interruptible, no_wait_gpu);
1055
out_unlock:
1056 1057
	if (ret && mem.mm_node)
		ttm_bo_mem_put(bo, &mem);
1058 1059 1060
	return ret;
}

1061 1062 1063
bool ttm_bo_mem_compat(struct ttm_placement *placement,
		       struct ttm_mem_reg *mem,
		       uint32_t *new_flags)
1064
{
1065
	int i;
1066

1067
	for (i = 0; i < placement->num_placement; i++) {
1068
		const struct ttm_place *heap = &placement->placement[i];
1069
		if (mem->mm_node &&
1070
		    (mem->start < heap->fpfn ||
1071
		     (heap->lpfn != 0 && (mem->start + mem->num_pages) > heap->lpfn)))
1072 1073 1074
			continue;

		*new_flags = heap->flags;
1075 1076 1077 1078 1079 1080
		if ((*new_flags & mem->placement & TTM_PL_MASK_CACHING) &&
		    (*new_flags & mem->placement & TTM_PL_MASK_MEM))
			return true;
	}

	for (i = 0; i < placement->num_busy_placement; i++) {
1081
		const struct ttm_place *heap = &placement->busy_placement[i];
1082
		if (mem->mm_node &&
1083
		    (mem->start < heap->fpfn ||
1084
		     (heap->lpfn != 0 && (mem->start + mem->num_pages) > heap->lpfn)))
1085 1086 1087
			continue;

		*new_flags = heap->flags;
1088 1089 1090
		if ((*new_flags & mem->placement & TTM_PL_MASK_CACHING) &&
		    (*new_flags & mem->placement & TTM_PL_MASK_MEM))
			return true;
1091
	}
1092 1093

	return false;
1094
}
1095
EXPORT_SYMBOL(ttm_bo_mem_compat);
1096

1097 1098
int ttm_bo_validate(struct ttm_buffer_object *bo,
			struct ttm_placement *placement,
1099
			bool interruptible,
1100
			bool no_wait_gpu)
1101 1102
{
	int ret;
1103
	uint32_t new_flags;
1104

1105
	lockdep_assert_held(&bo->resv->lock.base);
1106 1107 1108
	/*
	 * Check whether we need to move buffer.
	 */
1109
	if (!ttm_bo_mem_compat(placement, &bo->mem, &new_flags)) {
1110 1111
		ret = ttm_bo_move_buffer(bo, placement, interruptible,
					 no_wait_gpu);
1112
		if (ret)
1113
			return ret;
1114 1115 1116 1117 1118
	} else {
		/*
		 * Use the access and other non-mapping-related flag bits from
		 * the compatible memory placement flags to the active flags
		 */
1119
		ttm_flag_masked(&bo->mem.placement, new_flags,
1120
				~TTM_PL_MASK_MEMTYPE);
1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131
	}
	/*
	 * We might need to add a TTM.
	 */
	if (bo->mem.mem_type == TTM_PL_SYSTEM && bo->ttm == NULL) {
		ret = ttm_bo_add_ttm(bo, true);
		if (ret)
			return ret;
	}
	return 0;
}
1132
EXPORT_SYMBOL(ttm_bo_validate);
1133

1134 1135 1136 1137 1138 1139 1140
int ttm_bo_init(struct ttm_bo_device *bdev,
		struct ttm_buffer_object *bo,
		unsigned long size,
		enum ttm_bo_type type,
		struct ttm_placement *placement,
		uint32_t page_alignment,
		bool interruptible,
J
Jan Engelhardt 已提交
1141
		struct file *persistent_swap_storage,
1142
		size_t acc_size,
1143
		struct sg_table *sg,
1144
		struct reservation_object *resv,
1145
		void (*destroy) (struct ttm_buffer_object *))
1146
{
1147
	int ret = 0;
1148
	unsigned long num_pages;
1149
	struct ttm_mem_global *mem_glob = bdev->glob->mem_glob;
1150
	bool locked;
1151 1152 1153

	ret = ttm_mem_global_alloc(mem_glob, acc_size, false, false);
	if (ret) {
J
Joe Perches 已提交
1154
		pr_err("Out of kernel memory\n");
1155 1156 1157 1158 1159 1160
		if (destroy)
			(*destroy)(bo);
		else
			kfree(bo);
		return -ENOMEM;
	}
1161 1162 1163

	num_pages = (size + PAGE_SIZE - 1) >> PAGE_SHIFT;
	if (num_pages == 0) {
J
Joe Perches 已提交
1164
		pr_err("Illegal buffer object size\n");
1165 1166 1167 1168
		if (destroy)
			(*destroy)(bo);
		else
			kfree(bo);
1169
		ttm_mem_global_free(mem_glob, acc_size);
1170 1171 1172 1173 1174 1175 1176 1177 1178 1179
		return -EINVAL;
	}
	bo->destroy = destroy;

	kref_init(&bo->kref);
	kref_init(&bo->list_kref);
	atomic_set(&bo->cpu_writers, 0);
	INIT_LIST_HEAD(&bo->lru);
	INIT_LIST_HEAD(&bo->ddestroy);
	INIT_LIST_HEAD(&bo->swap);
1180
	INIT_LIST_HEAD(&bo->io_reserve_lru);
1181
	mutex_init(&bo->wu_mutex);
1182
	bo->bdev = bdev;
1183
	bo->glob = bdev->glob;
1184 1185
	bo->type = type;
	bo->num_pages = num_pages;
1186
	bo->mem.size = num_pages << PAGE_SHIFT;
1187 1188 1189 1190
	bo->mem.mem_type = TTM_PL_SYSTEM;
	bo->mem.num_pages = bo->num_pages;
	bo->mem.mm_node = NULL;
	bo->mem.page_alignment = page_alignment;
1191 1192
	bo->mem.bus.io_reserved_vm = false;
	bo->mem.bus.io_reserved_count = 0;
1193
	bo->moving = NULL;
1194
	bo->mem.placement = (TTM_PL_FLAG_SYSTEM | TTM_PL_FLAG_CACHED);
J
Jan Engelhardt 已提交
1195
	bo->persistent_swap_storage = persistent_swap_storage;
1196
	bo->acc_size = acc_size;
1197
	bo->sg = sg;
1198 1199 1200 1201 1202 1203 1204
	if (resv) {
		bo->resv = resv;
		lockdep_assert_held(&bo->resv->lock.base);
	} else {
		bo->resv = &bo->ttm_resv;
		reservation_object_init(&bo->ttm_resv);
	}
1205
	atomic_inc(&bo->glob->bo_count);
1206
	drm_vma_node_reset(&bo->vma_node);
1207
	bo->priority = 0;
1208 1209 1210 1211 1212

	/*
	 * For ttm_bo_type_device buffers, allocate
	 * address space from the device.
	 */
1213 1214
	if (bo->type == ttm_bo_type_device ||
	    bo->type == ttm_bo_type_sg)
1215 1216
		ret = drm_vma_offset_add(&bdev->vma_manager, &bo->vma_node,
					 bo->mem.num_pages);
1217

1218 1219 1220 1221 1222 1223 1224
	/* passed reservation objects should already be locked,
	 * since otherwise lockdep will be angered in radeon.
	 */
	if (!resv) {
		locked = ww_mutex_trylock(&bo->resv->lock);
		WARN_ON(!locked);
	}
1225

1226 1227
	if (likely(!ret))
		ret = ttm_bo_validate(bo, placement, interruptible, false);
1228

1229
	if (!resv) {
1230
		ttm_bo_unreserve(bo);
1231

1232 1233 1234 1235 1236 1237
	} else if (!(bo->mem.placement & TTM_PL_FLAG_NO_EVICT)) {
		spin_lock(&bo->glob->lru_lock);
		ttm_bo_add_to_lru(bo);
		spin_unlock(&bo->glob->lru_lock);
	}

1238 1239
	if (unlikely(ret))
		ttm_bo_unref(&bo);
1240 1241 1242

	return ret;
}
1243
EXPORT_SYMBOL(ttm_bo_init);
1244

1245 1246 1247
size_t ttm_bo_acc_size(struct ttm_bo_device *bdev,
		       unsigned long bo_size,
		       unsigned struct_size)
1248
{
1249 1250
	unsigned npages = (PAGE_ALIGN(bo_size)) >> PAGE_SHIFT;
	size_t size = 0;
1251

1252
	size += ttm_round_pot(struct_size);
F
Felix Kuehling 已提交
1253
	size += ttm_round_pot(npages * sizeof(void *));
1254 1255
	size += ttm_round_pot(sizeof(struct ttm_tt));
	return size;
1256
}
1257 1258 1259 1260 1261 1262 1263 1264 1265 1266
EXPORT_SYMBOL(ttm_bo_acc_size);

size_t ttm_bo_dma_acc_size(struct ttm_bo_device *bdev,
			   unsigned long bo_size,
			   unsigned struct_size)
{
	unsigned npages = (PAGE_ALIGN(bo_size)) >> PAGE_SHIFT;
	size_t size = 0;

	size += ttm_round_pot(struct_size);
F
Felix Kuehling 已提交
1267
	size += ttm_round_pot(npages * (2*sizeof(void *) + sizeof(dma_addr_t)));
1268 1269 1270 1271
	size += ttm_round_pot(sizeof(struct ttm_dma_tt));
	return size;
}
EXPORT_SYMBOL(ttm_bo_dma_acc_size);
1272

1273 1274 1275 1276 1277 1278
int ttm_bo_create(struct ttm_bo_device *bdev,
			unsigned long size,
			enum ttm_bo_type type,
			struct ttm_placement *placement,
			uint32_t page_alignment,
			bool interruptible,
J
Jan Engelhardt 已提交
1279
			struct file *persistent_swap_storage,
1280
			struct ttm_buffer_object **p_bo)
1281 1282
{
	struct ttm_buffer_object *bo;
1283
	size_t acc_size;
1284
	int ret;
1285 1286

	bo = kzalloc(sizeof(*bo), GFP_KERNEL);
1287
	if (unlikely(bo == NULL))
1288 1289
		return -ENOMEM;

1290
	acc_size = ttm_bo_acc_size(bdev, size, sizeof(struct ttm_buffer_object));
1291
	ret = ttm_bo_init(bdev, bo, size, type, placement, page_alignment,
1292
			  interruptible, persistent_swap_storage, acc_size,
1293
			  NULL, NULL, NULL);
1294 1295 1296 1297 1298
	if (likely(ret == 0))
		*p_bo = bo;

	return ret;
}
T
Thomas Hellstrom 已提交
1299
EXPORT_SYMBOL(ttm_bo_create);
1300 1301

static int ttm_bo_force_list_clean(struct ttm_bo_device *bdev,
1302
				   unsigned mem_type)
1303
{
1304
	struct ttm_mem_type_manager *man = &bdev->man[mem_type];
1305
	struct ttm_bo_global *glob = bdev->glob;
1306
	struct dma_fence *fence;
1307
	int ret;
1308
	unsigned i;
1309 1310 1311 1312 1313

	/*
	 * Can't use standard list traversal since we're unlocking.
	 */

1314
	spin_lock(&glob->lru_lock);
1315 1316 1317 1318 1319 1320 1321 1322
	for (i = 0; i < TTM_MAX_BO_PRIORITY; ++i) {
		while (!list_empty(&man->lru[i])) {
			spin_unlock(&glob->lru_lock);
			ret = ttm_mem_evict_first(bdev, mem_type, NULL, false, false);
			if (ret)
				return ret;
			spin_lock(&glob->lru_lock);
		}
1323
	}
1324
	spin_unlock(&glob->lru_lock);
1325 1326

	spin_lock(&man->move_lock);
1327
	fence = dma_fence_get(man->move);
1328 1329 1330
	spin_unlock(&man->move_lock);

	if (fence) {
1331 1332
		ret = dma_fence_wait(fence, false);
		dma_fence_put(fence);
1333 1334
		if (ret)
			return ret;
1335 1336
	}

1337 1338 1339 1340 1341
	return 0;
}

int ttm_bo_clean_mm(struct ttm_bo_device *bdev, unsigned mem_type)
{
R
Roel Kluin 已提交
1342
	struct ttm_mem_type_manager *man;
1343 1344 1345
	int ret = -EINVAL;

	if (mem_type >= TTM_NUM_MEM_TYPES) {
J
Joe Perches 已提交
1346
		pr_err("Illegal memory type %d\n", mem_type);
1347 1348
		return ret;
	}
R
Roel Kluin 已提交
1349
	man = &bdev->man[mem_type];
1350 1351

	if (!man->has_type) {
J
Joe Perches 已提交
1352 1353
		pr_err("Trying to take down uninitialized memory manager type %u\n",
		       mem_type);
1354 1355
		return ret;
	}
1356
	dma_fence_put(man->move);
1357 1358 1359 1360 1361 1362

	man->use_type = false;
	man->has_type = false;

	ret = 0;
	if (mem_type > 0) {
1363 1364 1365 1366 1367
		ret = ttm_bo_force_list_clean(bdev, mem_type);
		if (ret) {
			pr_err("Cleanup eviction failed\n");
			return ret;
		}
1368

1369
		ret = (*man->func->takedown)(man);
1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380
	}

	return ret;
}
EXPORT_SYMBOL(ttm_bo_clean_mm);

int ttm_bo_evict_mm(struct ttm_bo_device *bdev, unsigned mem_type)
{
	struct ttm_mem_type_manager *man = &bdev->man[mem_type];

	if (mem_type == 0 || mem_type >= TTM_NUM_MEM_TYPES) {
J
Joe Perches 已提交
1381
		pr_err("Illegal memory manager memory type %u\n", mem_type);
1382 1383 1384 1385
		return -EINVAL;
	}

	if (!man->has_type) {
J
Joe Perches 已提交
1386
		pr_err("Memory type %u has not been initialized\n", mem_type);
1387 1388 1389
		return 0;
	}

1390
	return ttm_bo_force_list_clean(bdev, mem_type);
1391 1392 1393 1394
}
EXPORT_SYMBOL(ttm_bo_evict_mm);

int ttm_bo_init_mm(struct ttm_bo_device *bdev, unsigned type,
1395
			unsigned long p_size)
1396 1397 1398
{
	int ret = -EINVAL;
	struct ttm_mem_type_manager *man;
1399
	unsigned i;
1400

1401
	BUG_ON(type >= TTM_NUM_MEM_TYPES);
1402
	man = &bdev->man[type];
1403
	BUG_ON(man->has_type);
1404 1405 1406
	man->io_reserve_fastpath = true;
	man->use_io_reserve_lru = false;
	mutex_init(&man->io_reserve_mutex);
1407
	spin_lock_init(&man->move_lock);
1408
	INIT_LIST_HEAD(&man->io_reserve_lru);
1409 1410 1411 1412

	ret = bdev->driver->init_mem_type(bdev, type, man);
	if (ret)
		return ret;
1413
	man->bdev = bdev;
1414 1415 1416

	ret = 0;
	if (type != TTM_PL_SYSTEM) {
1417
		ret = (*man->func->init)(man, p_size);
1418 1419 1420 1421 1422 1423 1424
		if (ret)
			return ret;
	}
	man->has_type = true;
	man->use_type = true;
	man->size = p_size;

1425 1426
	for (i = 0; i < TTM_MAX_BO_PRIORITY; ++i)
		INIT_LIST_HEAD(&man->lru[i]);
1427
	man->move = NULL;
1428 1429 1430 1431 1432

	return 0;
}
EXPORT_SYMBOL(ttm_bo_init_mm);

1433 1434 1435 1436 1437 1438 1439 1440 1441 1442
static void ttm_bo_global_kobj_release(struct kobject *kobj)
{
	struct ttm_bo_global *glob =
		container_of(kobj, struct ttm_bo_global, kobj);

	ttm_mem_unregister_shrink(glob->mem_glob, &glob->shrink);
	__free_page(glob->dummy_read_page);
	kfree(glob);
}

1443
void ttm_bo_global_release(struct drm_global_reference *ref)
1444 1445 1446 1447 1448 1449 1450 1451
{
	struct ttm_bo_global *glob = ref->object;

	kobject_del(&glob->kobj);
	kobject_put(&glob->kobj);
}
EXPORT_SYMBOL(ttm_bo_global_release);

1452
int ttm_bo_global_init(struct drm_global_reference *ref)
1453 1454 1455 1456 1457
{
	struct ttm_bo_global_ref *bo_ref =
		container_of(ref, struct ttm_bo_global_ref, ref);
	struct ttm_bo_global *glob = ref->object;
	int ret;
1458
	unsigned i;
1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469

	mutex_init(&glob->device_list_mutex);
	spin_lock_init(&glob->lru_lock);
	glob->mem_glob = bo_ref->mem_glob;
	glob->dummy_read_page = alloc_page(__GFP_ZERO | GFP_DMA32);

	if (unlikely(glob->dummy_read_page == NULL)) {
		ret = -ENOMEM;
		goto out_no_drp;
	}

1470 1471
	for (i = 0; i < TTM_MAX_BO_PRIORITY; ++i)
		INIT_LIST_HEAD(&glob->swap_lru[i]);
1472 1473 1474 1475 1476
	INIT_LIST_HEAD(&glob->device_list);

	ttm_mem_init_shrink(&glob->shrink, ttm_bo_swapout);
	ret = ttm_mem_register_shrink(glob->mem_glob, &glob->shrink);
	if (unlikely(ret != 0)) {
J
Joe Perches 已提交
1477
		pr_err("Could not register buffer object swapout\n");
1478 1479 1480 1481 1482
		goto out_no_shrink;
	}

	atomic_set(&glob->bo_count, 0);

1483 1484
	ret = kobject_init_and_add(
		&glob->kobj, &ttm_bo_glob_kobj_type, ttm_get_kobj(), "buffer_objects");
1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496
	if (unlikely(ret != 0))
		kobject_put(&glob->kobj);
	return ret;
out_no_shrink:
	__free_page(glob->dummy_read_page);
out_no_drp:
	kfree(glob);
	return ret;
}
EXPORT_SYMBOL(ttm_bo_global_init);


1497 1498 1499 1500 1501
int ttm_bo_device_release(struct ttm_bo_device *bdev)
{
	int ret = 0;
	unsigned i = TTM_NUM_MEM_TYPES;
	struct ttm_mem_type_manager *man;
1502
	struct ttm_bo_global *glob = bdev->glob;
1503 1504 1505 1506 1507 1508 1509

	while (i--) {
		man = &bdev->man[i];
		if (man->has_type) {
			man->use_type = false;
			if ((i != TTM_PL_SYSTEM) && ttm_bo_clean_mm(bdev, i)) {
				ret = -EBUSY;
J
Joe Perches 已提交
1510 1511
				pr_err("DRM memory manager type %d is not clean\n",
				       i);
1512 1513 1514 1515 1516
			}
			man->has_type = false;
		}
	}

1517 1518 1519 1520
	mutex_lock(&glob->device_list_mutex);
	list_del(&bdev->device_list);
	mutex_unlock(&glob->device_list_mutex);

1521
	cancel_delayed_work_sync(&bdev->wq);
1522 1523 1524 1525

	while (ttm_bo_delayed_delete(bdev, true))
		;

1526
	spin_lock(&glob->lru_lock);
1527 1528 1529
	if (list_empty(&bdev->ddestroy))
		TTM_DEBUG("Delayed destroy list was clean\n");

1530 1531 1532
	for (i = 0; i < TTM_MAX_BO_PRIORITY; ++i)
		if (list_empty(&bdev->man[0].lru[0]))
			TTM_DEBUG("Swap list %d was clean\n", i);
1533
	spin_unlock(&glob->lru_lock);
1534

1535
	drm_vma_offset_manager_destroy(&bdev->vma_manager);
1536 1537 1538 1539 1540 1541

	return ret;
}
EXPORT_SYMBOL(ttm_bo_device_release);

int ttm_bo_device_init(struct ttm_bo_device *bdev,
1542 1543
		       struct ttm_bo_global *glob,
		       struct ttm_bo_driver *driver,
1544
		       struct address_space *mapping,
D
Dave Airlie 已提交
1545
		       uint64_t file_page_offset,
D
Dave Airlie 已提交
1546
		       bool need_dma32)
1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557
{
	int ret = -EINVAL;

	bdev->driver = driver;

	memset(bdev->man, 0, sizeof(bdev->man));

	/*
	 * Initialize the system memory buffer type.
	 * Other types need to be driver / IOCTL initialized.
	 */
1558
	ret = ttm_bo_init_mm(bdev, TTM_PL_SYSTEM, 0);
1559
	if (unlikely(ret != 0))
1560
		goto out_no_sys;
1561

1562 1563
	drm_vma_offset_manager_init(&bdev->vma_manager, file_page_offset,
				    0x10000000);
1564 1565
	INIT_DELAYED_WORK(&bdev->wq, ttm_bo_delayed_workqueue);
	INIT_LIST_HEAD(&bdev->ddestroy);
1566
	bdev->dev_mapping = mapping;
1567
	bdev->glob = glob;
D
Dave Airlie 已提交
1568
	bdev->need_dma32 = need_dma32;
1569 1570 1571
	mutex_lock(&glob->device_list_mutex);
	list_add_tail(&bdev->device_list, &glob->device_list);
	mutex_unlock(&glob->device_list_mutex);
1572 1573

	return 0;
1574
out_no_sys:
1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599
	return ret;
}
EXPORT_SYMBOL(ttm_bo_device_init);

/*
 * buffer object vm functions.
 */

bool ttm_mem_reg_is_pci(struct ttm_bo_device *bdev, struct ttm_mem_reg *mem)
{
	struct ttm_mem_type_manager *man = &bdev->man[mem->mem_type];

	if (!(man->flags & TTM_MEMTYPE_FLAG_FIXED)) {
		if (mem->mem_type == TTM_PL_SYSTEM)
			return false;

		if (man->flags & TTM_MEMTYPE_FLAG_CMA)
			return false;

		if (mem->placement & TTM_PL_FLAG_CACHED)
			return false;
	}
	return true;
}

1600
void ttm_bo_unmap_virtual_locked(struct ttm_buffer_object *bo)
1601 1602 1603
{
	struct ttm_bo_device *bdev = bo->bdev;

1604
	drm_vma_node_unmap(&bo->vma_node, bdev->dev_mapping);
1605
	ttm_mem_io_free_vm(bo);
1606
}
1607 1608 1609 1610 1611 1612 1613 1614 1615

void ttm_bo_unmap_virtual(struct ttm_buffer_object *bo)
{
	struct ttm_bo_device *bdev = bo->bdev;
	struct ttm_mem_type_manager *man = &bdev->man[bo->mem.mem_type];

	ttm_mem_io_lock(man, false);
	ttm_bo_unmap_virtual_locked(bo);
	ttm_mem_io_unlock(man);
1616
}
1617 1618


1619
EXPORT_SYMBOL(ttm_bo_unmap_virtual);
1620 1621

int ttm_bo_wait(struct ttm_buffer_object *bo,
1622
		bool interruptible, bool no_wait)
1623
{
C
Christian König 已提交
1624 1625 1626 1627 1628 1629 1630 1631
	long timeout = 15 * HZ;

	if (no_wait) {
		if (reservation_object_test_signaled_rcu(bo->resv, true))
			return 0;
		else
			return -EBUSY;
	}
1632

C
Christian König 已提交
1633 1634
	timeout = reservation_object_wait_timeout_rcu(bo->resv, true,
						      interruptible, timeout);
1635 1636 1637 1638 1639 1640
	if (timeout < 0)
		return timeout;

	if (timeout == 0)
		return -EBUSY;

C
Christian König 已提交
1641
	reservation_object_add_excl_fence(bo->resv, NULL);
1642
	return 0;
1643 1644 1645 1646 1647 1648 1649 1650
}
EXPORT_SYMBOL(ttm_bo_wait);

int ttm_bo_synccpu_write_grab(struct ttm_buffer_object *bo, bool no_wait)
{
	int ret = 0;

	/*
1651
	 * Using ttm_bo_reserve makes sure the lru lists are updated.
1652 1653
	 */

1654
	ret = ttm_bo_reserve(bo, true, no_wait, NULL);
1655 1656
	if (unlikely(ret != 0))
		return ret;
1657
	ret = ttm_bo_wait(bo, true, no_wait);
1658 1659 1660 1661 1662
	if (likely(ret == 0))
		atomic_inc(&bo->cpu_writers);
	ttm_bo_unreserve(bo);
	return ret;
}
1663
EXPORT_SYMBOL(ttm_bo_synccpu_write_grab);
1664 1665 1666

void ttm_bo_synccpu_write_release(struct ttm_buffer_object *bo)
{
1667
	atomic_dec(&bo->cpu_writers);
1668
}
1669
EXPORT_SYMBOL(ttm_bo_synccpu_write_release);
1670 1671 1672 1673 1674 1675 1676 1677

/**
 * A buffer object shrink method that tries to swap out the first
 * buffer object on the bo_global::swap_lru list.
 */

static int ttm_bo_swapout(struct ttm_mem_shrink *shrink)
{
1678 1679
	struct ttm_bo_global *glob =
	    container_of(shrink, struct ttm_bo_global, shrink);
1680 1681 1682 1683
	struct ttm_buffer_object *bo;
	int ret = -EBUSY;
	int put_count;
	uint32_t swap_placement = (TTM_PL_FLAG_CACHED | TTM_PL_FLAG_SYSTEM);
1684
	unsigned i;
1685

1686
	spin_lock(&glob->lru_lock);
1687 1688 1689 1690 1691 1692
	for (i = 0; i < TTM_MAX_BO_PRIORITY; ++i) {
		list_for_each_entry(bo, &glob->swap_lru[i], swap) {
			ret = __ttm_bo_reserve(bo, false, true, NULL);
			if (!ret)
				break;
		}
1693 1694 1695
		if (!ret)
			break;
	}
1696

1697 1698 1699 1700
	if (ret) {
		spin_unlock(&glob->lru_lock);
		return ret;
	}
1701

1702
	kref_get(&bo->list_kref);
1703

1704 1705 1706 1707
	if (!list_empty(&bo->ddestroy)) {
		ret = ttm_bo_cleanup_refs_and_unlock(bo, false, false);
		kref_put(&bo->list_kref, ttm_bo_release_list);
		return ret;
1708 1709 1710
	}

	put_count = ttm_bo_del_from_lru(bo);
1711
	spin_unlock(&glob->lru_lock);
1712

1713
	ttm_bo_list_ref_sub(bo, put_count, true);
1714 1715

	/**
1716
	 * Move to system cached
1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727
	 */

	if ((bo->mem.placement & swap_placement) != swap_placement) {
		struct ttm_mem_reg evict_mem;

		evict_mem = bo->mem;
		evict_mem.mm_node = NULL;
		evict_mem.placement = TTM_PL_FLAG_SYSTEM | TTM_PL_FLAG_CACHED;
		evict_mem.mem_type = TTM_PL_SYSTEM;

		ret = ttm_bo_handle_move_mem(bo, &evict_mem, true,
1728
					     false, false);
1729 1730 1731 1732
		if (unlikely(ret != 0))
			goto out;
	}

1733 1734 1735 1736 1737 1738 1739 1740
	/**
	 * Make sure BO is idle.
	 */

	ret = ttm_bo_wait(bo, false, false);
	if (unlikely(ret != 0))
		goto out;

1741 1742 1743 1744 1745 1746 1747
	ttm_bo_unmap_virtual(bo);

	/**
	 * Swap out. Buffer will be swapped in again as soon as
	 * anyone tries to access a ttm page.
	 */

1748 1749 1750
	if (bo->bdev->driver->swap_notify)
		bo->bdev->driver->swap_notify(bo);

J
Jan Engelhardt 已提交
1751
	ret = ttm_tt_swapout(bo->ttm, bo->persistent_swap_storage);
1752 1753 1754 1755 1756 1757 1758 1759
out:

	/**
	 *
	 * Unreserve without putting on LRU to avoid swapping out an
	 * already swapped buffer.
	 */

1760
	__ttm_bo_unreserve(bo);
1761 1762 1763 1764 1765 1766
	kref_put(&bo->list_kref, ttm_bo_release_list);
	return ret;
}

void ttm_bo_swapout_all(struct ttm_bo_device *bdev)
{
1767
	while (ttm_bo_swapout(&bdev->glob->shrink) == 0)
1768 1769
		;
}
1770
EXPORT_SYMBOL(ttm_bo_swapout_all);
1771 1772 1773 1774 1775 1776 1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790 1791 1792 1793

/**
 * ttm_bo_wait_unreserved - interruptible wait for a buffer object to become
 * unreserved
 *
 * @bo: Pointer to buffer
 */
int ttm_bo_wait_unreserved(struct ttm_buffer_object *bo)
{
	int ret;

	/*
	 * In the absense of a wait_unlocked API,
	 * Use the bo::wu_mutex to avoid triggering livelocks due to
	 * concurrent use of this function. Note that this use of
	 * bo::wu_mutex can go away if we change locking order to
	 * mmap_sem -> bo::reserve.
	 */
	ret = mutex_lock_interruptible(&bo->wu_mutex);
	if (unlikely(ret != 0))
		return -ERESTARTSYS;
	if (!ww_mutex_is_locked(&bo->resv->lock))
		goto out_unlock;
1794
	ret = __ttm_bo_reserve(bo, true, false, NULL);
1795 1796
	if (unlikely(ret != 0))
		goto out_unlock;
1797
	__ttm_bo_unreserve(bo);
1798 1799 1800 1801 1802

out_unlock:
	mutex_unlock(&bo->wu_mutex);
	return ret;
}