summaryrefslogtreecommitdiff
path: root/drivers/gpu/drm/i915/gem/i915_gem_domain.c
blob: 7c90a63c273d8d52e2148f45bee13e94d4e27e6e (plain)
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
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
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
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
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
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
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
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
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
423
424
425
426
427
428
429
430
431
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
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
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
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
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
629
630
631
632
633
634
635
636
637
638
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
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
/*
 * SPDX-License-Identifier: MIT
 *
 * Copyright © 2014-2016 Intel Corporation
 */

#include "display/intel_frontbuffer.h"

#include "i915_drv.h"
#include "i915_gem_clflush.h"
#include "i915_gem_gtt.h"
#include "i915_gem_ioctls.h"
#include "i915_gem_object.h"
#include "i915_vma.h"
#include "i915_gem_lmem.h"
#include "i915_gem_mman.h"

static void __i915_gem_object_flush_for_display(struct drm_i915_gem_object *obj)
{
	/*
	 * We manually flush the CPU domain so that we can override and
	 * force the flush for the display, and perform it asyncrhonously.
	 */
	i915_gem_object_flush_write_domain(obj, ~I915_GEM_DOMAIN_CPU);
	if (obj->cache_dirty)
		i915_gem_clflush_object(obj, I915_CLFLUSH_FORCE);
	obj->write_domain = 0;
}

void i915_gem_object_flush_if_display(struct drm_i915_gem_object *obj)
{
	if (!i915_gem_object_is_framebuffer(obj))
		return;

	i915_gem_object_lock(obj, NULL);
	__i915_gem_object_flush_for_display(obj);
	i915_gem_object_unlock(obj);
}

void i915_gem_object_flush_if_display_locked(struct drm_i915_gem_object *obj)
{
	if (i915_gem_object_is_framebuffer(obj))
		__i915_gem_object_flush_for_display(obj);
}

/**
 * Moves a single object to the WC read, and possibly write domain.
 * @obj: object to act on
 * @write: ask for write access or read only
 *
 * This function returns when the move is complete, including waiting on
 * flushes to occur.
 */
int
i915_gem_object_set_to_wc_domain(struct drm_i915_gem_object *obj, bool write)
{
	int ret;

	assert_object_held(obj);

	ret = i915_gem_object_wait(obj,
				   I915_WAIT_INTERRUPTIBLE |
				   (write ? I915_WAIT_ALL : 0),
				   MAX_SCHEDULE_TIMEOUT);
	if (ret)
		return ret;

	if (obj->write_domain == I915_GEM_DOMAIN_WC)
		return 0;

	/* Flush and acquire obj->pages so that we are coherent through
	 * direct access in memory with previous cached writes through
	 * shmemfs and that our cache domain tracking remains valid.
	 * For example, if the obj->filp was moved to swap without us
	 * being notified and releasing the pages, we would mistakenly
	 * continue to assume that the obj remained out of the CPU cached
	 * domain.
	 */
	ret = i915_gem_object_pin_pages(obj);
	if (ret)
		return ret;

	i915_gem_object_flush_write_domain(obj, ~I915_GEM_DOMAIN_WC);

	/* Serialise direct access to this object with the barriers for
	 * coherent writes from the GPU, by effectively invalidating the
	 * WC domain upon first access.
	 */
	if ((obj->read_domains & I915_GEM_DOMAIN_WC) == 0)
		mb();

	/* It should now be out of any other write domains, and we can update
	 * the domain values for our changes.
	 */
	GEM_BUG_ON((obj->write_domain & ~I915_GEM_DOMAIN_WC) != 0);
	obj->read_domains |= I915_GEM_DOMAIN_WC;
	if (write) {
		obj->read_domains = I915_GEM_DOMAIN_WC;
		obj->write_domain = I915_GEM_DOMAIN_WC;
		obj->mm.dirty = true;
	}

	i915_gem_object_unpin_pages(obj);
	return 0;
}

/**
 * Moves a single object to the GTT read, and possibly write domain.
 * @obj: object to act on
 * @write: ask for write access or read only
 *
 * This function returns when the move is complete, including waiting on
 * flushes to occur.
 */
int
i915_gem_object_set_to_gtt_domain(struct drm_i915_gem_object *obj, bool write)
{
	int ret;

	assert_object_held(obj);

	ret = i915_gem_object_wait(obj,
				   I915_WAIT_INTERRUPTIBLE |
				   (write ? I915_WAIT_ALL : 0),
				   MAX_SCHEDULE_TIMEOUT);
	if (ret)
		return ret;

	if (obj->write_domain == I915_GEM_DOMAIN_GTT)
		return 0;

	/* Flush and acquire obj->pages so that we are coherent through
	 * direct access in memory with previous cached writes through
	 * shmemfs and that our cache domain tracking remains valid.
	 * For example, if the obj->filp was moved to swap without us
	 * being notified and releasing the pages, we would mistakenly
	 * continue to assume that the obj remained out of the CPU cached
	 * domain.
	 */
	ret = i915_gem_object_pin_pages(obj);
	if (ret)
		return ret;

	i915_gem_object_flush_write_domain(obj, ~I915_GEM_DOMAIN_GTT);

	/* Serialise direct access to this object with the barriers for
	 * coherent writes from the GPU, by effectively invalidating the
	 * GTT domain upon first access.
	 */
	if ((obj->read_domains & I915_GEM_DOMAIN_GTT) == 0)
		mb();

	/* It should now be out of any other write domains, and we can update
	 * the domain values for our changes.
	 */
	GEM_BUG_ON((obj->write_domain & ~I915_GEM_DOMAIN_GTT) != 0);
	obj->read_domains |= I915_GEM_DOMAIN_GTT;
	if (write) {
		struct i915_vma *vma;

		obj->read_domains = I915_GEM_DOMAIN_GTT;
		obj->write_domain = I915_GEM_DOMAIN_GTT;
		obj->mm.dirty = true;

		spin_lock(&obj->vma.lock);
		for_each_ggtt_vma(vma, obj)
			if (i915_vma_is_bound(vma, I915_VMA_GLOBAL_BIND))
				i915_vma_set_ggtt_write(vma);
		spin_unlock(&obj->vma.lock);
	}

	i915_gem_object_unpin_pages(obj);
	return 0;
}

/**
 * Changes the cache-level of an object across all VMA.
 * @obj: object to act on
 * @cache_level: new cache level to set for the object
 *
 * After this function returns, the object will be in the new cache-level
 * across all GTT and the contents of the backing storage will be coherent,
 * with respect to the new cache-level. In order to keep the backing storage
 * coherent for all users, we only allow a single cache level to be set
 * globally on the object and prevent it from being changed whilst the
 * hardware is reading from the object. That is if the object is currently
 * on the scanout it will be set to uncached (or equivalent display
 * cache coherency) and all non-MOCS GPU access will also be uncached so
 * that all direct access to the scanout remains coherent.
 */
int i915_gem_object_set_cache_level(struct drm_i915_gem_object *obj,
				    enum i915_cache_level cache_level)
{
	int ret;

	if (obj->cache_level == cache_level)
		return 0;

	ret = i915_gem_object_wait(obj,
				   I915_WAIT_INTERRUPTIBLE |
				   I915_WAIT_ALL,
				   MAX_SCHEDULE_TIMEOUT);
	if (ret)
		return ret;

	/* Always invalidate stale cachelines */
	if (obj->cache_level != cache_level) {
		i915_gem_object_set_cache_coherency(obj, cache_level);
		obj->cache_dirty = true;
	}

	/* The cache-level will be applied when each vma is rebound. */
	return i915_gem_object_unbind(obj,
				      I915_GEM_OBJECT_UNBIND_ACTIVE |
				      I915_GEM_OBJECT_UNBIND_BARRIER);
}

int i915_gem_get_caching_ioctl(struct drm_device *dev, void *data,
			       struct drm_file *file)
{
	struct drm_i915_gem_caching *args = data;
	struct drm_i915_gem_object *obj;
	int err = 0;

	rcu_read_lock();
	obj = i915_gem_object_lookup_rcu(file, args->handle);
	if (!obj) {
		err = -ENOENT;
		goto out;
	}

	switch (obj->cache_level) {
	case I915_CACHE_LLC:
	case I915_CACHE_L3_LLC:
		args->caching = I915_CACHING_CACHED;
		break;

	case I915_CACHE_WT:
		args->caching = I915_CACHING_DISPLAY;
		break;

	default:
		args->caching = I915_CACHING_NONE;
		break;
	}
out:
	rcu_read_unlock();
	return err;
}

int i915_gem_set_caching_ioctl(struct drm_device *dev, void *data,
			       struct drm_file *file)
{
	struct drm_i915_private *i915 = to_i915(dev);
	struct drm_i915_gem_caching *args = data;
	struct drm_i915_gem_object *obj;
	enum i915_cache_level level;
	int ret = 0;

	switch (args->caching) {
	case I915_CACHING_NONE:
		level = I915_CACHE_NONE;
		break;
	case I915_CACHING_CACHED:
		/*
		 * Due to a HW issue on BXT A stepping, GPU stores via a
		 * snooped mapping may leave stale data in a corresponding CPU
		 * cacheline, whereas normally such cachelines would get
		 * invalidated.
		 */
		if (!HAS_LLC(i915) && !HAS_SNOOP(i915))
			return -ENODEV;

		level = I915_CACHE_LLC;
		break;
	case I915_CACHING_DISPLAY:
		level = HAS_WT(i915) ? I915_CACHE_WT : I915_CACHE_NONE;
		break;
	default:
		return -EINVAL;
	}

	obj = i915_gem_object_lookup(file, args->handle);
	if (!obj)
		return -ENOENT;

	/*
	 * The caching mode of proxy object is handled by its generator, and
	 * not allowed to be changed by userspace.
	 */
	if (i915_gem_object_is_proxy(obj)) {
		ret = -ENXIO;
		goto out;
	}

	ret = i915_gem_object_lock_interruptible(obj, NULL);
	if (ret)
		goto out;

	ret = i915_gem_object_set_cache_level(obj, level);
	i915_gem_object_unlock(obj);

out:
	i915_gem_object_put(obj);
	return ret;
}

/*
 * Prepare buffer for display plane (scanout, cursors, etc). Can be called from
 * an uninterruptible phase (modesetting) and allows any flushes to be pipelined
 * (for pageflips). We only flush the caches while preparing the buffer for
 * display, the callers are responsible for frontbuffer flush.
 */
struct i915_vma *
i915_gem_object_pin_to_display_plane(struct drm_i915_gem_object *obj,
				     u32 alignment,
				     const struct i915_ggtt_view *view,
				     unsigned int flags)
{
	struct drm_i915_private *i915 = to_i915(obj->base.dev);
	struct i915_gem_ww_ctx ww;
	struct i915_vma *vma;
	int ret;

	/* Frame buffer must be in LMEM (no migration yet) */
	if (HAS_LMEM(i915) && !i915_gem_object_is_lmem(obj))
		return ERR_PTR(-EINVAL);

	i915_gem_ww_ctx_init(&ww, true);
retry:
	ret = i915_gem_object_lock(obj, &ww);
	if (ret)
		goto err;
	/*
	 * The display engine is not coherent with the LLC cache on gen6.  As
	 * a result, we make sure that the pinning that is about to occur is
	 * done with uncached PTEs. This is lowest common denominator for all
	 * chipsets.
	 *
	 * However for gen6+, we could do better by using the GFDT bit instead
	 * of uncaching, which would allow us to flush all the LLC-cached data
	 * with that bit in the PTE to main memory with just one PIPE_CONTROL.
	 */
	ret = i915_gem_object_set_cache_level(obj,
					      HAS_WT(i915) ?
					      I915_CACHE_WT : I915_CACHE_NONE);
	if (ret)
		goto err;

	/*
	 * As the user may map the buffer once pinned in the display plane
	 * (e.g. libkms for the bootup splash), we have to ensure that we
	 * always use map_and_fenceable for all scanout buffers. However,
	 * it may simply be too big to fit into mappable, in which case
	 * put it anyway and hope that userspace can cope (but always first
	 * try to preserve the existing ABI).
	 */
	vma = ERR_PTR(-ENOSPC);
	if ((flags & PIN_MAPPABLE) == 0 &&
	    (!view || view->type == I915_GGTT_VIEW_NORMAL))
		vma = i915_gem_object_ggtt_pin_ww(obj, &ww, view, 0, alignment,
						  flags | PIN_MAPPABLE |
						  PIN_NONBLOCK);
	if (IS_ERR(vma) && vma != ERR_PTR(-EDEADLK))
		vma = i915_gem_object_ggtt_pin_ww(obj, &ww, view, 0,
						  alignment, flags);
	if (IS_ERR(vma)) {
		ret = PTR_ERR(vma);
		goto err;
	}

	vma->display_alignment = max_t(u64, vma->display_alignment, alignment);

	i915_gem_object_flush_if_display_locked(obj);

err:
	if (ret == -EDEADLK) {
		ret = i915_gem_ww_ctx_backoff(&ww);
		if (!ret)
			goto retry;
	}
	i915_gem_ww_ctx_fini(&ww);

	if (ret)
		return ERR_PTR(ret);

	return vma;
}

static void i915_gem_object_bump_inactive_ggtt(struct drm_i915_gem_object *obj)
{
	struct drm_i915_private *i915 = to_i915(obj->base.dev);
	struct i915_vma *vma;

	if (list_empty(&obj->vma.list))
		return;

	mutex_lock(&i915->ggtt.vm.mutex);
	spin_lock(&obj->vma.lock);
	for_each_ggtt_vma(vma, obj) {
		if (!drm_mm_node_allocated(&vma->node))
			continue;

		GEM_BUG_ON(vma->vm != &i915->ggtt.vm);
		list_move_tail(&vma->vm_link, &vma->vm->bound_list);
	}
	spin_unlock(&obj->vma.lock);
	mutex_unlock(&i915->ggtt.vm.mutex);

	if (i915_gem_object_is_shrinkable(obj)) {
		unsigned long flags;

		spin_lock_irqsave(&i915->mm.obj_lock, flags);

		if (obj->mm.madv == I915_MADV_WILLNEED &&
		    !atomic_read(&obj->mm.shrink_pin))
			list_move_tail(&obj->mm.link, &i915->mm.shrink_list);

		spin_unlock_irqrestore(&i915->mm.obj_lock, flags);
	}
}

void
i915_gem_object_unpin_from_display_plane(struct i915_vma *vma)
{
	/* Bump the LRU to try and avoid premature eviction whilst flipping  */
	i915_gem_object_bump_inactive_ggtt(vma->obj);

	i915_vma_unpin(vma);
}

/**
 * Moves a single object to the CPU read, and possibly write domain.
 * @obj: object to act on
 * @write: requesting write or read-only access
 *
 * This function returns when the move is complete, including waiting on
 * flushes to occur.
 */
int
i915_gem_object_set_to_cpu_domain(struct drm_i915_gem_object *obj, bool write)
{
	int ret;

	assert_object_held(obj);

	ret = i915_gem_object_wait(obj,
				   I915_WAIT_INTERRUPTIBLE |
				   (write ? I915_WAIT_ALL : 0),
				   MAX_SCHEDULE_TIMEOUT);
	if (ret)
		return ret;

	i915_gem_object_flush_write_domain(obj, ~I915_GEM_DOMAIN_CPU);

	/* Flush the CPU cache if it's still invalid. */
	if ((obj->read_domains & I915_GEM_DOMAIN_CPU) == 0) {
		i915_gem_clflush_object(obj, I915_CLFLUSH_SYNC);
		obj->read_domains |= I915_GEM_DOMAIN_CPU;
	}

	/* It should now be out of any other write domains, and we can update
	 * the domain values for our changes.
	 */
	GEM_BUG_ON(obj->write_domain & ~I915_GEM_DOMAIN_CPU);

	/* If we're writing through the CPU, then the GPU read domains will
	 * need to be invalidated at next use.
	 */
	if (write)
		__start_cpu_write(obj);

	return 0;
}

/**
 * Called when user space prepares to use an object with the CPU, either
 * through the mmap ioctl's mapping or a GTT mapping.
 * @dev: drm device
 * @data: ioctl data blob
 * @file: drm file
 */
int
i915_gem_set_domain_ioctl(struct drm_device *dev, void *data,
			  struct drm_file *file)
{
	struct drm_i915_gem_set_domain *args = data;
	struct drm_i915_gem_object *obj;
	u32 read_domains = args->read_domains;
	u32 write_domain = args->write_domain;
	int err;

	/* Only handle setting domains to types used by the CPU. */
	if ((write_domain | read_domains) & I915_GEM_GPU_DOMAINS)
		return -EINVAL;

	/*
	 * Having something in the write domain implies it's in the read
	 * domain, and only that read domain.  Enforce that in the request.
	 */
	if (write_domain && read_domains != write_domain)
		return -EINVAL;

	if (!read_domains)
		return 0;

	obj = i915_gem_object_lookup(file, args->handle);
	if (!obj)
		return -ENOENT;

	/*
	 * Already in the desired write domain? Nothing for us to do!
	 *
	 * We apply a little bit of cunning here to catch a broader set of
	 * no-ops. If obj->write_domain is set, we must be in the same
	 * obj->read_domains, and only that domain. Therefore, if that
	 * obj->write_domain matches the request read_domains, we are
	 * already in the same read/write domain and can skip the operation,
	 * without having to further check the requested write_domain.
	 */
	if (READ_ONCE(obj->write_domain) == read_domains) {
		err = 0;
		goto out;
	}

	/*
	 * Try to flush the object off the GPU without holding the lock.
	 * We will repeat the flush holding the lock in the normal manner
	 * to catch cases where we are gazumped.
	 */
	err = i915_gem_object_wait(obj,
				   I915_WAIT_INTERRUPTIBLE |
				   I915_WAIT_PRIORITY |
				   (write_domain ? I915_WAIT_ALL : 0),
				   MAX_SCHEDULE_TIMEOUT);
	if (err)
		goto out;

	/*
	 * Proxy objects do not control access to the backing storage, ergo
	 * they cannot be used as a means to manipulate the cache domain
	 * tracking for that backing storage. The proxy object is always
	 * considered to be outside of any cache domain.
	 */
	if (i915_gem_object_is_proxy(obj)) {
		err = -ENXIO;
		goto out;
	}

	/*
	 * Flush and acquire obj->pages so that we are coherent through
	 * direct access in memory with previous cached writes through
	 * shmemfs and that our cache domain tracking remains valid.
	 * For example, if the obj->filp was moved to swap without us
	 * being notified and releasing the pages, we would mistakenly
	 * continue to assume that the obj remained out of the CPU cached
	 * domain.
	 */
	err = i915_gem_object_pin_pages(obj);
	if (err)
		goto out;

	err = i915_gem_object_lock_interruptible(obj, NULL);
	if (err)
		goto out_unpin;

	if (read_domains & I915_GEM_DOMAIN_WC)
		err = i915_gem_object_set_to_wc_domain(obj, write_domain);
	else if (read_domains & I915_GEM_DOMAIN_GTT)
		err = i915_gem_object_set_to_gtt_domain(obj, write_domain);
	else
		err = i915_gem_object_set_to_cpu_domain(obj, write_domain);

	/* And bump the LRU for this access */
	i915_gem_object_bump_inactive_ggtt(obj);

	i915_gem_object_unlock(obj);

	if (write_domain)
		i915_gem_object_invalidate_frontbuffer(obj, ORIGIN_CPU);

out_unpin:
	i915_gem_object_unpin_pages(obj);
out:
	i915_gem_object_put(obj);
	return err;
}

/*
 * Pins the specified object's pages and synchronizes the object with
 * GPU accesses. Sets needs_clflush to non-zero if the caller should
 * flush the object from the CPU cache.
 */
int i915_gem_object_prepare_read(struct drm_i915_gem_object *obj,
				 unsigned int *needs_clflush)
{
	int ret;

	*needs_clflush = 0;
	if (!i915_gem_object_has_struct_page(obj))
		return -ENODEV;

	assert_object_held(obj);

	ret = i915_gem_object_wait(obj,
				   I915_WAIT_INTERRUPTIBLE,
				   MAX_SCHEDULE_TIMEOUT);
	if (ret)
		return ret;

	ret = i915_gem_object_pin_pages(obj);
	if (ret)
		return ret;

	if (obj->cache_coherent & I915_BO_CACHE_COHERENT_FOR_READ ||
	    !static_cpu_has(X86_FEATURE_CLFLUSH)) {
		ret = i915_gem_object_set_to_cpu_domain(obj, false);
		if (ret)
			goto err_unpin;
		else
			goto out;
	}

	i915_gem_object_flush_write_domain(obj, ~I915_GEM_DOMAIN_CPU);

	/* If we're not in the cpu read domain, set ourself into the gtt
	 * read domain and manually flush cachelines (if required). This
	 * optimizes for the case when the gpu will dirty the data
	 * anyway again before the next pread happens.
	 */
	if (!obj->cache_dirty &&
	    !(obj->read_domains & I915_GEM_DOMAIN_CPU))
		*needs_clflush = CLFLUSH_BEFORE;

out:
	/* return with the pages pinned */
	return 0;

err_unpin:
	i915_gem_object_unpin_pages(obj);
	return ret;
}

int i915_gem_object_prepare_write(struct drm_i915_gem_object *obj,
				  unsigned int *needs_clflush)
{
	int ret;

	*needs_clflush = 0;
	if (!i915_gem_object_has_struct_page(obj))
		return -ENODEV;

	assert_object_held(obj);

	ret = i915_gem_object_wait(obj,
				   I915_WAIT_INTERRUPTIBLE |
				   I915_WAIT_ALL,
				   MAX_SCHEDULE_TIMEOUT);
	if (ret)
		return ret;

	ret = i915_gem_object_pin_pages(obj);
	if (ret)
		return ret;

	if (obj->cache_coherent & I915_BO_CACHE_COHERENT_FOR_WRITE ||
	    !static_cpu_has(X86_FEATURE_CLFLUSH)) {
		ret = i915_gem_object_set_to_cpu_domain(obj, true);
		if (ret)
			goto err_unpin;
		else
			goto out;
	}

	i915_gem_object_flush_write_domain(obj, ~I915_GEM_DOMAIN_CPU);

	/* If we're not in the cpu write domain, set ourself into the
	 * gtt write domain and manually flush cachelines (as required).
	 * This optimizes for the case when the gpu will use the data
	 * right away and we therefore have to clflush anyway.
	 */
	if (!obj->cache_dirty) {
		*needs_clflush |= CLFLUSH_AFTER;

		/*
		 * Same trick applies to invalidate partially written
		 * cachelines read before writing.
		 */
		if (!(obj->read_domains & I915_GEM_DOMAIN_CPU))
			*needs_clflush |= CLFLUSH_BEFORE;
	}

out:
	i915_gem_object_invalidate_frontbuffer(obj, ORIGIN_CPU);
	obj->mm.dirty = true;
	/* return with the pages pinned */
	return 0;

err_unpin:
	i915_gem_object_unpin_pages(obj);
	return ret;
}