summaryrefslogtreecommitdiffstats
path: root/drivers/gpu/drm/i915/gem/i915_gem_userptr.c
blob: ba14b18d65f380cc75fe2fa7a52b4eff84886e5c (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
/*
 * SPDX-License-Identifier: MIT
 *
 * Copyright © 2012-2014 Intel Corporation
 *
  * Based on amdgpu_mn, which bears the following notice:
 *
 * Copyright 2014 Advanced Micro Devices, Inc.
 * 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 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.
 *
 * The above copyright notice and this permission notice (including the
 * next paragraph) shall be included in all copies or substantial portions
 * of the Software.
 *
 */
/*
 * Authors:
 *    Christian König <christian.koenig@amd.com>
 */

#include <linux/mmu_context.h>
#include <linux/mempolicy.h>
#include <linux/swap.h>
#include <linux/sched/mm.h>

#include "i915_drv.h"
#include "i915_gem_ioctls.h"
#include "i915_gem_object.h"
#include "i915_gem_userptr.h"
#include "i915_scatterlist.h"

#ifdef CONFIG_MMU_NOTIFIER

/**
 * i915_gem_userptr_invalidate - callback to notify about mm change
 *
 * @mni: the range (mm) is about to update
 * @range: details on the invalidation
 * @cur_seq: Value to pass to mmu_interval_set_seq()
 *
 * Block for operations on BOs to finish and mark pages as accessed and
 * potentially dirty.
 */
static bool i915_gem_userptr_invalidate(struct mmu_interval_notifier *mni,
					const struct mmu_notifier_range *range,
					unsigned long cur_seq)
{
	struct drm_i915_gem_object *obj = container_of(mni, struct drm_i915_gem_object, userptr.notifier);
	struct drm_i915_private *i915 = to_i915(obj->base.dev);
	long r;

	if (!mmu_notifier_range_blockable(range))
		return false;

	write_lock(&i915->mm.notifier_lock);

	mmu_interval_set_seq(mni, cur_seq);

	write_unlock(&i915->mm.notifier_lock);

	/*
	 * We don't wait when the process is exiting. This is valid
	 * because the object will be cleaned up anyway.
	 *
	 * This is also temporarily required as a hack, because we
	 * cannot currently force non-consistent batch buffers to preempt
	 * and reschedule by waiting on it, hanging processes on exit.
	 */
	if (current->flags & PF_EXITING)
		return true;

	/* we will unbind on next submission, still have userptr pins */
	r = dma_resv_wait_timeout(obj->base.resv, DMA_RESV_USAGE_BOOKKEEP, false,
				  MAX_SCHEDULE_TIMEOUT);
	if (r <= 0)
		drm_err(&i915->drm, "(%ld) failed to wait for idle\n", r);

	return true;
}

static const struct mmu_interval_notifier_ops i915_gem_userptr_notifier_ops = {
	.invalidate = i915_gem_userptr_invalidate,
};

static int
i915_gem_userptr_init__mmu_notifier(struct drm_i915_gem_object *obj)
{
	return mmu_interval_notifier_insert(&obj->userptr.notifier, current->mm,
					    obj->userptr.ptr, obj->base.size,
					    &i915_gem_userptr_notifier_ops);
}

static void i915_gem_object_userptr_drop_ref(struct drm_i915_gem_object *obj)
{
	struct page **pvec = NULL;

	assert_object_held_shared(obj);

	if (!--obj->userptr.page_ref) {
		pvec = obj->userptr.pvec;
		obj->userptr.pvec = NULL;
	}
	GEM_BUG_ON(obj->userptr.page_ref < 0);

	if (pvec) {
		const unsigned long num_pages = obj->base.size >> PAGE_SHIFT;

		unpin_user_pages(pvec, num_pages);
		kvfree(pvec);
	}
}

static int i915_gem_userptr_get_pages(struct drm_i915_gem_object *obj)
{
	const unsigned long num_pages = obj->base.size >> PAGE_SHIFT;
	unsigned int max_segment = i915_sg_segment_size(obj->base.dev->dev);
	struct sg_table *st;
	unsigned int sg_page_sizes;
	struct page **pvec;
	int ret;

	st = kmalloc(sizeof(*st), GFP_KERNEL);
	if (!st)
		return -ENOMEM;

	if (!obj->userptr.page_ref) {
		ret = -EAGAIN;
		goto err_free;
	}

	obj->userptr.page_ref++;
	pvec = obj->userptr.pvec;

alloc_table:
	ret = sg_alloc_table_from_pages_segment(st, pvec, num_pages, 0,
						num_pages << PAGE_SHIFT,
						max_segment, GFP_KERNEL);
	if (ret)
		goto err;

	ret = i915_gem_gtt_prepare_pages(obj, st);
	if (ret) {
		sg_free_table(st);

		if (max_segment > PAGE_SIZE) {
			max_segment = PAGE_SIZE;
			goto alloc_table;
		}

		goto err;
	}

	WARN_ON_ONCE(!(obj->cache_coherent & I915_BO_CACHE_COHERENT_FOR_WRITE));
	if (i915_gem_object_can_bypass_llc(obj))
		obj->cache_dirty = true;

	sg_page_sizes = i915_sg_dma_sizes(st->sgl);
	__i915_gem_object_set_pages(obj, st, sg_page_sizes);

	return 0;

err:
	i915_gem_object_userptr_drop_ref(obj);
err_free:
	kfree(st);
	return ret;
}

static void
i915_gem_userptr_put_pages(struct drm_i915_gem_object *obj,
			   struct sg_table *pages)
{
	struct sgt_iter sgt_iter;
	struct page *page;

	if (!pages)
		return;

	__i915_gem_object_release_shmem(obj, pages, true);
	i915_gem_gtt_finish_pages(obj, pages);

	/*
	 * We always mark objects as dirty when they are used by the GPU,
	 * just in case. However, if we set the vma as being read-only we know
	 * that the object will never have been written to.
	 */
	if (i915_gem_object_is_readonly(obj))
		obj->mm.dirty = false;

	for_each_sgt_page(page, sgt_iter, pages) {
		if (obj->mm.dirty && trylock_page(page)) {
			/*
			 * As this may not be anonymous memory (e.g. shmem)
			 * but exist on a real mapping, we have to lock
			 * the page in order to dirty it -- holding
			 * the page reference is not sufficient to
			 * prevent the inode from being truncated.
			 * Play safe and take the lock.
			 *
			 * However...!
			 *
			 * The mmu-notifier can be invalidated for a
			 * migrate_folio, that is alreadying holding the lock
			 * on the folio. Such a try_to_unmap() will result
			 * in us calling put_pages() and so recursively try
			 * to lock the page. We avoid that deadlock with
			 * a trylock_page() and in exchange we risk missing
			 * some page dirtying.
			 */
			set_page_dirty(page);
			unlock_page(page);
		}

		mark_page_accessed(page);
	}
	obj->mm.dirty = false;

	sg_free_table(pages);
	kfree(pages);

	i915_gem_object_userptr_drop_ref(obj);
}

static int i915_gem_object_userptr_unbind(struct drm_i915_gem_object *obj)
{
	struct sg_table *pages;
	int err;

	err = i915_gem_object_unbind(obj, I915_GEM_OBJECT_UNBIND_ACTIVE);
	if (err)
		return err;

	if (GEM_WARN_ON(i915_gem_object_has_pinned_pages(obj)))
		return -EBUSY;

	assert_object_held(obj);

	pages = __i915_gem_object_unset_pages(obj);
	if (!IS_ERR_OR_NULL(pages))
		i915_gem_userptr_put_pages(obj, pages);

	return err;
}

int i915_gem_object_userptr_submit_init(struct drm_i915_gem_object *obj)
{
	const unsigned long num_pages = obj->base.size >> PAGE_SHIFT;
	struct page **pvec;
	unsigned int gup_flags = 0;
	unsigned long notifier_seq;
	int pinned, ret;

	if (obj->userptr.notifier.mm != current->mm)
		return -EFAULT;

	notifier_seq = mmu_interval_read_begin(&obj->userptr.notifier);

	ret = i915_gem_object_lock_interruptible(obj, NULL);
	if (ret)
		return ret;

	if (notifier_seq == obj->userptr.notifier_seq && obj->userptr.pvec) {
		i915_gem_object_unlock(obj);
		return 0;
	}

	ret = i915_gem_object_userptr_unbind(obj);
	i915_gem_object_unlock(obj);
	if (ret)
		return ret;

	pvec = kvmalloc_array(num_pages, sizeof(struct page *), GFP_KERNEL);
	if (!pvec)
		return -ENOMEM;

	if (!i915_gem_object_is_readonly(obj))
		gup_flags |= FOLL_WRITE;

	pinned = ret = 0;
	while (pinned < num_pages) {
		ret = pin_user_pages_fast(obj->userptr.ptr + pinned * PAGE_SIZE,
					  num_pages - pinned, gup_flags,
					  &pvec[pinned]);
		if (ret < 0)
			goto out;

		pinned += ret;
	}
	ret = 0;

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

	if (mmu_interval_read_retry(&obj->userptr.notifier,
		!obj->userptr.page_ref ? notifier_seq :
		obj->userptr.notifier_seq)) {
		ret = -EAGAIN;
		goto out_unlock;
	}

	if (!obj->userptr.page_ref++) {
		obj->userptr.pvec = pvec;
		obj->userptr.notifier_seq = notifier_seq;
		pvec = NULL;
		ret = ____i915_gem_object_get_pages(obj);
	}

	obj->userptr.page_ref--;

out_unlock:
	i915_gem_object_unlock(obj);

out:
	if (pvec) {
		unpin_user_pages(pvec, pinned);
		kvfree(pvec);
	}

	return ret;
}

int i915_gem_object_userptr_submit_done(struct drm_i915_gem_object *obj)
{
	if (mmu_interval_read_retry(&obj->userptr.notifier,
				    obj->userptr.notifier_seq)) {
		/* We collided with the mmu notifier, need to retry */

		return -EAGAIN;
	}

	return 0;
}

int i915_gem_object_userptr_validate(struct drm_i915_gem_object *obj)
{
	int err;

	err = i915_gem_object_userptr_submit_init(obj);
	if (err)
		return err;

	err = i915_gem_object_lock_interruptible(obj, NULL);
	if (!err) {
		/*
		 * Since we only check validity, not use the pages,
		 * it doesn't matter if we collide with the mmu notifier,
		 * and -EAGAIN handling is not required.
		 */
		err = i915_gem_object_pin_pages(obj);
		if (!err)
			i915_gem_object_unpin_pages(obj);

		i915_gem_object_unlock(obj);
	}

	return err;
}

static void
i915_gem_userptr_release(struct drm_i915_gem_object *obj)
{
	GEM_WARN_ON(obj->userptr.page_ref);

	mmu_interval_notifier_remove(&obj->userptr.notifier);
	obj->userptr.notifier.mm = NULL;
}

static int
i915_gem_userptr_dmabuf_export(struct drm_i915_gem_object *obj)
{
	drm_dbg(obj->base.dev, "Exporting userptr no longer allowed\n");

	return -EINVAL;
}

static int
i915_gem_userptr_pwrite(struct drm_i915_gem_object *obj,
			const struct drm_i915_gem_pwrite *args)
{
	drm_dbg(obj->base.dev, "pwrite to userptr no longer allowed\n");

	return -EINVAL;
}

static int
i915_gem_userptr_pread(struct drm_i915_gem_object *obj,
		       const struct drm_i915_gem_pread *args)
{
	drm_dbg(obj->base.dev, "pread from userptr no longer allowed\n");

	return -EINVAL;
}

static const struct drm_i915_gem_object_ops i915_gem_userptr_ops = {
	.name = "i915_gem_object_userptr",
	.flags = I915_GEM_OBJECT_IS_SHRINKABLE |
		 I915_GEM_OBJECT_NO_MMAP |
		 I915_GEM_OBJECT_IS_PROXY,
	.get_pages = i915_gem_userptr_get_pages,
	.put_pages = i915_gem_userptr_put_pages,
	.dmabuf_export = i915_gem_userptr_dmabuf_export,
	.pwrite = i915_gem_userptr_pwrite,
	.pread = i915_gem_userptr_pread,
	.release = i915_gem_userptr_release,
};

#endif

static int
probe_range(struct mm_struct *mm, unsigned long addr, unsigned long len)
{
	VMA_ITERATOR(vmi, mm, addr);
	struct vm_area_struct *vma;
	unsigned long end = addr + len;

	mmap_read_lock(mm);
	for_each_vma_range(vmi, vma, end) {
		/* Check for holes, note that we also update the addr below */
		if (vma->vm_start > addr)
			break;

		if (vma->vm_flags & (VM_PFNMAP | VM_MIXEDMAP))
			break;

		addr = vma->vm_end;
	}
	mmap_read_unlock(mm);

	if (vma || addr < end)
		return -EFAULT;
	return 0;
}

/*
 * Creates a new mm object that wraps some normal memory from the process
 * context - user memory.
 *
 * We impose several restrictions upon the memory being mapped
 * into the GPU.
 * 1. It must be page aligned (both start/end addresses, i.e ptr and size).
 * 2. It must be normal system memory, not a pointer into another map of IO
 *    space (e.g. it must not be a GTT mmapping of another object).
 * 3. We only allow a bo as large as we could in theory map into the GTT,
 *    that is we limit the size to the total size of the GTT.
 * 4. The bo is marked as being snoopable. The backing pages are left
 *    accessible directly by the CPU, but reads and writes by the GPU may
 *    incur the cost of a snoop (unless you have an LLC architecture).
 *
 * Synchronisation between multiple users and the GPU is left to userspace
 * through the normal set-domain-ioctl. The kernel will enforce that the
 * GPU relinquishes the VMA before it is returned back to the system
 * i.e. upon free(), munmap() or process termination. However, the userspace
 * malloc() library may not immediately relinquish the VMA after free() and
 * instead reuse it whilst the GPU is still reading and writing to the VMA.
 * Caveat emptor.
 *
 * Also note, that the object created here is not currently a "first class"
 * object, in that several ioctls are banned. These are the CPU access
 * ioctls: mmap(), pwrite and pread. In practice, you are expected to use
 * direct access via your pointer rather than use those ioctls. Another
 * restriction is that we do not allow userptr surfaces to be pinned to the
 * hardware and so we reject any attempt to create a framebuffer out of a
 * userptr.
 *
 * If you think this is a good interface to use to pass GPU memory between
 * drivers, please use dma-buf instead. In fact, wherever possible use
 * dma-buf instead.
 */
int
i915_gem_userptr_ioctl(struct drm_device *dev,
		       void *data,
		       struct drm_file *file)
{
	static struct lock_class_key __maybe_unused lock_class;
	struct drm_i915_private *dev_priv = to_i915(dev);
	struct drm_i915_gem_userptr *args = data;
	struct drm_i915_gem_object __maybe_unused *obj;
	int __maybe_unused ret;
	u32 __maybe_unused handle;

	if (!HAS_LLC(dev_priv) && !HAS_SNOOP(dev_priv)) {
		/* We cannot support coherent userptr objects on hw without
		 * LLC and broken snooping.
		 */
		return -ENODEV;
	}

	if (args->flags & ~(I915_USERPTR_READ_ONLY |
			    I915_USERPTR_UNSYNCHRONIZED |
			    I915_USERPTR_PROBE))
		return -EINVAL;

	if (i915_gem_object_size_2big(args->user_size))
		return -E2BIG;

	if (!args->user_size)
		return -EINVAL;

	if (offset_in_page(args->user_ptr | args->user_size))
		return -EINVAL;

	if (!access_ok((char __user *)(unsigned long)args->user_ptr, args->user_size))
		return -EFAULT;

	if (args->flags & I915_USERPTR_UNSYNCHRONIZED)
		return -ENODEV;

	if (args->flags & I915_USERPTR_READ_ONLY) {
		/*
		 * On almost all of the older hw, we cannot tell the GPU that
		 * a page is readonly.
		 */
		if (!to_gt(dev_priv)->vm->has_read_only)
			return -ENODEV;
	}

	if (args->flags & I915_USERPTR_PROBE) {
		/*
		 * Check that the range pointed to represents real struct
		 * pages and not iomappings (at this moment in time!)
		 */
		ret = probe_range(current->mm, args->user_ptr, args->user_size);
		if (ret)
			return ret;
	}

#ifdef CONFIG_MMU_NOTIFIER
	obj = i915_gem_object_alloc();
	if (obj == NULL)
		return -ENOMEM;

	drm_gem_private_object_init(dev, &obj->base, args->user_size);
	i915_gem_object_init(obj, &i915_gem_userptr_ops, &lock_class,
			     I915_BO_ALLOC_USER);
	obj->mem_flags = I915_BO_FLAG_STRUCT_PAGE;
	obj->read_domains = I915_GEM_DOMAIN_CPU;
	obj->write_domain = I915_GEM_DOMAIN_CPU;
	i915_gem_object_set_cache_coherency(obj, I915_CACHE_LLC);

	obj->userptr.ptr = args->user_ptr;
	obj->userptr.notifier_seq = ULONG_MAX;
	if (args->flags & I915_USERPTR_READ_ONLY)
		i915_gem_object_set_readonly(obj);

	/* And keep a pointer to the current->mm for resolving the user pages
	 * at binding. This means that we need to hook into the mmu_notifier
	 * in order to detect if the mmu is destroyed.
	 */
	ret = i915_gem_userptr_init__mmu_notifier(obj);
	if (ret == 0)
		ret = drm_gem_handle_create(file, &obj->base, &handle);

	/* drop reference from allocate - handle holds it now */
	i915_gem_object_put(obj);
	if (ret)
		return ret;

	args->handle = handle;
	return 0;
#else
	return -ENODEV;
#endif
}

int i915_gem_init_userptr(struct drm_i915_private *dev_priv)
{
#ifdef CONFIG_MMU_NOTIFIER
	rwlock_init(&dev_priv->mm.notifier_lock);
#endif

	return 0;
}

void i915_gem_cleanup_userptr(struct drm_i915_private *dev_priv)
{
}