summaryrefslogtreecommitdiff
path: root/drivers/gpu/drm/i915/gem/i915_gem_pages.c
blob: 29f4c28507458f09e2139b29d1dcf919003e3dea (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
/*
 * SPDX-License-Identifier: MIT
 *
 * Copyright © 2014-2016 Intel Corporation
 */

#include "i915_drv.h"
#include "i915_gem_object.h"
#include "i915_scatterlist.h"
#include "i915_gem_lmem.h"

void __i915_gem_object_set_pages(struct drm_i915_gem_object *obj,
				 struct sg_table *pages,
				 unsigned int sg_page_sizes)
{
	struct drm_i915_private *i915 = to_i915(obj->base.dev);
	unsigned long supported = INTEL_INFO(i915)->page_sizes;
	int i;

	lockdep_assert_held(&obj->mm.lock);

	if (i915_gem_object_is_volatile(obj))
		obj->mm.madv = I915_MADV_DONTNEED;

	/* Make the pages coherent with the GPU (flushing any swapin). */
	if (obj->cache_dirty) {
		obj->write_domain = 0;
		if (i915_gem_object_has_struct_page(obj))
			drm_clflush_sg(pages);
		obj->cache_dirty = false;
	}

	obj->mm.get_page.sg_pos = pages->sgl;
	obj->mm.get_page.sg_idx = 0;

	obj->mm.pages = pages;

	if (i915_gem_object_is_tiled(obj) &&
	    i915->quirks & QUIRK_PIN_SWIZZLED_PAGES) {
		GEM_BUG_ON(obj->mm.quirked);
		__i915_gem_object_pin_pages(obj);
		obj->mm.quirked = true;
	}

	GEM_BUG_ON(!sg_page_sizes);
	obj->mm.page_sizes.phys = sg_page_sizes;

	/*
	 * Calculate the supported page-sizes which fit into the given
	 * sg_page_sizes. This will give us the page-sizes which we may be able
	 * to use opportunistically when later inserting into the GTT. For
	 * example if phys=2G, then in theory we should be able to use 1G, 2M,
	 * 64K or 4K pages, although in practice this will depend on a number of
	 * other factors.
	 */
	obj->mm.page_sizes.sg = 0;
	for_each_set_bit(i, &supported, ilog2(I915_GTT_MAX_PAGE_SIZE) + 1) {
		if (obj->mm.page_sizes.phys & ~0u << i)
			obj->mm.page_sizes.sg |= BIT(i);
	}
	GEM_BUG_ON(!HAS_PAGE_SIZES(i915, obj->mm.page_sizes.sg));

	if (i915_gem_object_is_shrinkable(obj)) {
		struct list_head *list;
		unsigned long flags;

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

		i915->mm.shrink_count++;
		i915->mm.shrink_memory += obj->base.size;

		if (obj->mm.madv != I915_MADV_WILLNEED)
			list = &i915->mm.purge_list;
		else
			list = &i915->mm.shrink_list;
		list_add_tail(&obj->mm.link, list);

		atomic_set(&obj->mm.shrink_pin, 0);
		spin_unlock_irqrestore(&i915->mm.obj_lock, flags);
	}
}

int ____i915_gem_object_get_pages(struct drm_i915_gem_object *obj)
{
	int err;

	if (unlikely(obj->mm.madv != I915_MADV_WILLNEED)) {
		DRM_DEBUG("Attempting to obtain a purgeable object\n");
		return -EFAULT;
	}

	err = obj->ops->get_pages(obj);
	GEM_BUG_ON(!err && !i915_gem_object_has_pages(obj));

	return err;
}

/* Ensure that the associated pages are gathered from the backing storage
 * and pinned into our object. i915_gem_object_pin_pages() may be called
 * multiple times before they are released by a single call to
 * i915_gem_object_unpin_pages() - once the pages are no longer referenced
 * either as a result of memory pressure (reaping pages under the shrinker)
 * or as the object is itself released.
 */
int __i915_gem_object_get_pages(struct drm_i915_gem_object *obj)
{
	int err;

	err = mutex_lock_interruptible(&obj->mm.lock);
	if (err)
		return err;

	if (unlikely(!i915_gem_object_has_pages(obj))) {
		GEM_BUG_ON(i915_gem_object_has_pinned_pages(obj));

		err = ____i915_gem_object_get_pages(obj);
		if (err)
			goto unlock;

		smp_mb__before_atomic();
	}
	atomic_inc(&obj->mm.pages_pin_count);

unlock:
	mutex_unlock(&obj->mm.lock);
	return err;
}

/* Immediately discard the backing storage */
void i915_gem_object_truncate(struct drm_i915_gem_object *obj)
{
	drm_gem_free_mmap_offset(&obj->base);
	if (obj->ops->truncate)
		obj->ops->truncate(obj);
}

/* Try to discard unwanted pages */
void i915_gem_object_writeback(struct drm_i915_gem_object *obj)
{
	lockdep_assert_held(&obj->mm.lock);
	GEM_BUG_ON(i915_gem_object_has_pages(obj));

	if (obj->ops->writeback)
		obj->ops->writeback(obj);
}

static void __i915_gem_object_reset_page_iter(struct drm_i915_gem_object *obj)
{
	struct radix_tree_iter iter;
	void __rcu **slot;

	rcu_read_lock();
	radix_tree_for_each_slot(slot, &obj->mm.get_page.radix, &iter, 0)
		radix_tree_delete(&obj->mm.get_page.radix, iter.index);
	rcu_read_unlock();
}

static void unmap_object(struct drm_i915_gem_object *obj, void *ptr)
{
	if (i915_gem_object_is_lmem(obj))
		io_mapping_unmap((void __force __iomem *)ptr);
	else if (is_vmalloc_addr(ptr))
		vunmap(ptr);
	else
		kunmap(kmap_to_page(ptr));
}

struct sg_table *
__i915_gem_object_unset_pages(struct drm_i915_gem_object *obj)
{
	struct sg_table *pages;

	pages = fetch_and_zero(&obj->mm.pages);
	if (IS_ERR_OR_NULL(pages))
		return pages;

	if (i915_gem_object_is_volatile(obj))
		obj->mm.madv = I915_MADV_WILLNEED;

	i915_gem_object_make_unshrinkable(obj);

	if (obj->mm.mapping) {
		unmap_object(obj, page_mask_bits(obj->mm.mapping));
		obj->mm.mapping = NULL;
	}

	__i915_gem_object_reset_page_iter(obj);
	obj->mm.page_sizes.phys = obj->mm.page_sizes.sg = 0;

	return pages;
}

int __i915_gem_object_put_pages(struct drm_i915_gem_object *obj,
				enum i915_mm_subclass subclass)
{
	struct sg_table *pages;
	int err;

	if (i915_gem_object_has_pinned_pages(obj))
		return -EBUSY;

	GEM_BUG_ON(atomic_read(&obj->bind_count));

	/* May be called by shrinker from within get_pages() (on another bo) */
	mutex_lock_nested(&obj->mm.lock, subclass);
	if (unlikely(atomic_read(&obj->mm.pages_pin_count))) {
		err = -EBUSY;
		goto unlock;
	}

	/*
	 * ->put_pages might need to allocate memory for the bit17 swizzle
	 * array, hence protect them from being reaped by removing them from gtt
	 * lists early.
	 */
	pages = __i915_gem_object_unset_pages(obj);

	/*
	 * XXX Temporary hijinx to avoid updating all backends to handle
	 * NULL pages. In the future, when we have more asynchronous
	 * get_pages backends we should be better able to handle the
	 * cancellation of the async task in a more uniform manner.
	 */
	if (!pages && !i915_gem_object_needs_async_cancel(obj))
		pages = ERR_PTR(-EINVAL);

	if (!IS_ERR(pages))
		obj->ops->put_pages(obj, pages);

	err = 0;
unlock:
	mutex_unlock(&obj->mm.lock);

	return err;
}

/* The 'mapping' part of i915_gem_object_pin_map() below */
static void *i915_gem_object_map(struct drm_i915_gem_object *obj,
				 enum i915_map_type type)
{
	unsigned long n_pages = obj->base.size >> PAGE_SHIFT;
	struct sg_table *sgt = obj->mm.pages;
	struct sgt_iter sgt_iter;
	struct page *page;
	struct page *stack_pages[32];
	struct page **pages = stack_pages;
	unsigned long i = 0;
	pgprot_t pgprot;
	void *addr;

	if (i915_gem_object_is_lmem(obj)) {
		void __iomem *io;

		if (type != I915_MAP_WC)
			return NULL;

		io = i915_gem_object_lmem_io_map(obj, 0, obj->base.size);
		return (void __force *)io;
	}

	/* A single page can always be kmapped */
	if (n_pages == 1 && type == I915_MAP_WB)
		return kmap(sg_page(sgt->sgl));

	if (n_pages > ARRAY_SIZE(stack_pages)) {
		/* Too big for stack -- allocate temporary array instead */
		pages = kvmalloc_array(n_pages, sizeof(*pages), GFP_KERNEL);
		if (!pages)
			return NULL;
	}

	for_each_sgt_page(page, sgt_iter, sgt)
		pages[i++] = page;

	/* Check that we have the expected number of pages */
	GEM_BUG_ON(i != n_pages);

	switch (type) {
	default:
		MISSING_CASE(type);
		/* fallthrough - to use PAGE_KERNEL anyway */
	case I915_MAP_WB:
		pgprot = PAGE_KERNEL;
		break;
	case I915_MAP_WC:
		pgprot = pgprot_writecombine(PAGE_KERNEL_IO);
		break;
	}
	addr = vmap(pages, n_pages, 0, pgprot);

	if (pages != stack_pages)
		kvfree(pages);

	return addr;
}

/* get, pin, and map the pages of the object into kernel space */
void *i915_gem_object_pin_map(struct drm_i915_gem_object *obj,
			      enum i915_map_type type)
{
	enum i915_map_type has_type;
	unsigned int flags;
	bool pinned;
	void *ptr;
	int err;

	flags = I915_GEM_OBJECT_HAS_STRUCT_PAGE | I915_GEM_OBJECT_HAS_IOMEM;
	if (!i915_gem_object_type_has(obj, flags))
		return ERR_PTR(-ENXIO);

	err = mutex_lock_interruptible(&obj->mm.lock);
	if (err)
		return ERR_PTR(err);

	pinned = !(type & I915_MAP_OVERRIDE);
	type &= ~I915_MAP_OVERRIDE;

	if (!atomic_inc_not_zero(&obj->mm.pages_pin_count)) {
		if (unlikely(!i915_gem_object_has_pages(obj))) {
			GEM_BUG_ON(i915_gem_object_has_pinned_pages(obj));

			err = ____i915_gem_object_get_pages(obj);
			if (err)
				goto err_unlock;

			smp_mb__before_atomic();
		}
		atomic_inc(&obj->mm.pages_pin_count);
		pinned = false;
	}
	GEM_BUG_ON(!i915_gem_object_has_pages(obj));

	ptr = page_unpack_bits(obj->mm.mapping, &has_type);
	if (ptr && has_type != type) {
		if (pinned) {
			err = -EBUSY;
			goto err_unpin;
		}

		unmap_object(obj, ptr);

		ptr = obj->mm.mapping = NULL;
	}

	if (!ptr) {
		ptr = i915_gem_object_map(obj, type);
		if (!ptr) {
			err = -ENOMEM;
			goto err_unpin;
		}

		obj->mm.mapping = page_pack_bits(ptr, type);
	}

out_unlock:
	mutex_unlock(&obj->mm.lock);
	return ptr;

err_unpin:
	atomic_dec(&obj->mm.pages_pin_count);
err_unlock:
	ptr = ERR_PTR(err);
	goto out_unlock;
}

void __i915_gem_object_flush_map(struct drm_i915_gem_object *obj,
				 unsigned long offset,
				 unsigned long size)
{
	enum i915_map_type has_type;
	void *ptr;

	GEM_BUG_ON(!i915_gem_object_has_pinned_pages(obj));
	GEM_BUG_ON(range_overflows_t(typeof(obj->base.size),
				     offset, size, obj->base.size));

	obj->mm.dirty = true;

	if (obj->cache_coherent & I915_BO_CACHE_COHERENT_FOR_WRITE)
		return;

	ptr = page_unpack_bits(obj->mm.mapping, &has_type);
	if (has_type == I915_MAP_WC)
		return;

	drm_clflush_virt_range(ptr + offset, size);
	if (size == obj->base.size) {
		obj->write_domain &= ~I915_GEM_DOMAIN_CPU;
		obj->cache_dirty = false;
	}
}

struct scatterlist *
i915_gem_object_get_sg(struct drm_i915_gem_object *obj,
		       unsigned int n,
		       unsigned int *offset)
{
	struct i915_gem_object_page_iter *iter = &obj->mm.get_page;
	struct scatterlist *sg;
	unsigned int idx, count;

	might_sleep();
	GEM_BUG_ON(n >= obj->base.size >> PAGE_SHIFT);
	GEM_BUG_ON(!i915_gem_object_has_pinned_pages(obj));

	/* As we iterate forward through the sg, we record each entry in a
	 * radixtree for quick repeated (backwards) lookups. If we have seen
	 * this index previously, we will have an entry for it.
	 *
	 * Initial lookup is O(N), but this is amortized to O(1) for
	 * sequential page access (where each new request is consecutive
	 * to the previous one). Repeated lookups are O(lg(obj->base.size)),
	 * i.e. O(1) with a large constant!
	 */
	if (n < READ_ONCE(iter->sg_idx))
		goto lookup;

	mutex_lock(&iter->lock);

	/* We prefer to reuse the last sg so that repeated lookup of this
	 * (or the subsequent) sg are fast - comparing against the last
	 * sg is faster than going through the radixtree.
	 */

	sg = iter->sg_pos;
	idx = iter->sg_idx;
	count = __sg_page_count(sg);

	while (idx + count <= n) {
		void *entry;
		unsigned long i;
		int ret;

		/* If we cannot allocate and insert this entry, or the
		 * individual pages from this range, cancel updating the
		 * sg_idx so that on this lookup we are forced to linearly
		 * scan onwards, but on future lookups we will try the
		 * insertion again (in which case we need to be careful of
		 * the error return reporting that we have already inserted
		 * this index).
		 */
		ret = radix_tree_insert(&iter->radix, idx, sg);
		if (ret && ret != -EEXIST)
			goto scan;

		entry = xa_mk_value(idx);
		for (i = 1; i < count; i++) {
			ret = radix_tree_insert(&iter->radix, idx + i, entry);
			if (ret && ret != -EEXIST)
				goto scan;
		}

		idx += count;
		sg = ____sg_next(sg);
		count = __sg_page_count(sg);
	}

scan:
	iter->sg_pos = sg;
	iter->sg_idx = idx;

	mutex_unlock(&iter->lock);

	if (unlikely(n < idx)) /* insertion completed by another thread */
		goto lookup;

	/* In case we failed to insert the entry into the radixtree, we need
	 * to look beyond the current sg.
	 */
	while (idx + count <= n) {
		idx += count;
		sg = ____sg_next(sg);
		count = __sg_page_count(sg);
	}

	*offset = n - idx;
	return sg;

lookup:
	rcu_read_lock();

	sg = radix_tree_lookup(&iter->radix, n);
	GEM_BUG_ON(!sg);

	/* If this index is in the middle of multi-page sg entry,
	 * the radix tree will contain a value entry that points
	 * to the start of that range. We will return the pointer to
	 * the base page and the offset of this page within the
	 * sg entry's range.
	 */
	*offset = 0;
	if (unlikely(xa_is_value(sg))) {
		unsigned long base = xa_to_value(sg);

		sg = radix_tree_lookup(&iter->radix, base);
		GEM_BUG_ON(!sg);

		*offset = n - base;
	}

	rcu_read_unlock();

	return sg;
}

struct page *
i915_gem_object_get_page(struct drm_i915_gem_object *obj, unsigned int n)
{
	struct scatterlist *sg;
	unsigned int offset;

	GEM_BUG_ON(!i915_gem_object_has_struct_page(obj));

	sg = i915_gem_object_get_sg(obj, n, &offset);
	return nth_page(sg_page(sg), offset);
}

/* Like i915_gem_object_get_page(), but mark the returned page dirty */
struct page *
i915_gem_object_get_dirty_page(struct drm_i915_gem_object *obj,
			       unsigned int n)
{
	struct page *page;

	page = i915_gem_object_get_page(obj, n);
	if (!obj->mm.dirty)
		set_page_dirty(page);

	return page;
}

dma_addr_t
i915_gem_object_get_dma_address_len(struct drm_i915_gem_object *obj,
				    unsigned long n,
				    unsigned int *len)
{
	struct scatterlist *sg;
	unsigned int offset;

	sg = i915_gem_object_get_sg(obj, n, &offset);

	if (len)
		*len = sg_dma_len(sg) - (offset << PAGE_SHIFT);

	return sg_dma_address(sg) + (offset << PAGE_SHIFT);
}

dma_addr_t
i915_gem_object_get_dma_address(struct drm_i915_gem_object *obj,
				unsigned long n)
{
	return i915_gem_object_get_dma_address_len(obj, n, NULL);
}