summaryrefslogtreecommitdiff
path: root/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c
blob: 2d6f5ec77a687a0a905bd15e222b3215d2273c29 (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
/*
 * 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 <drm/drmP.h>
#include "amdgpu.h"
#include "amdgpu_trace.h"
#include "amdgpu_amdkfd.h"

struct amdgpu_sync_entry {
	struct hlist_node	node;
	struct dma_fence	*fence;
	bool	explicit;
};

static struct kmem_cache *amdgpu_sync_slab;

/**
 * amdgpu_sync_create - zero init sync object
 *
 * @sync: sync object to initialize
 *
 * Just clear the sync object for now.
 */
void amdgpu_sync_create(struct amdgpu_sync *sync)
{
	hash_init(sync->fences);
	sync->last_vm_update = NULL;
}

/**
 * amdgpu_sync_same_dev - test if fence belong to us
 *
 * @adev: amdgpu device to use for the test
 * @f: fence to test
 *
 * Test if the fence was issued by us.
 */
static bool amdgpu_sync_same_dev(struct amdgpu_device *adev,
				 struct dma_fence *f)
{
	struct drm_sched_fence *s_fence = to_drm_sched_fence(f);

	if (s_fence) {
		struct amdgpu_ring *ring;

		ring = container_of(s_fence->sched, struct amdgpu_ring, sched);
		return ring->adev == adev;
	}

	return false;
}

/**
 * amdgpu_sync_get_owner - extract the owner of a fence
 *
 * @fence: fence get the owner from
 *
 * Extract who originally created the fence.
 */
static void *amdgpu_sync_get_owner(struct dma_fence *f)
{
	struct drm_sched_fence *s_fence;
	struct amdgpu_amdkfd_fence *kfd_fence;

	if (!f)
		return AMDGPU_FENCE_OWNER_UNDEFINED;

	s_fence = to_drm_sched_fence(f);
	if (s_fence)
		return s_fence->owner;

	kfd_fence = to_amdgpu_amdkfd_fence(f);
	if (kfd_fence)
		return AMDGPU_FENCE_OWNER_KFD;

	return AMDGPU_FENCE_OWNER_UNDEFINED;
}

/**
 * amdgpu_sync_keep_later - Keep the later fence
 *
 * @keep: existing fence to test
 * @fence: new fence
 *
 * Either keep the existing fence or the new one, depending which one is later.
 */
static void amdgpu_sync_keep_later(struct dma_fence **keep,
				   struct dma_fence *fence)
{
	if (*keep && dma_fence_is_later(*keep, fence))
		return;

	dma_fence_put(*keep);
	*keep = dma_fence_get(fence);
}

/**
 * amdgpu_sync_add_later - add the fence to the hash
 *
 * @sync: sync object to add the fence to
 * @f: fence to add
 *
 * Tries to add the fence to an existing hash entry. Returns true when an entry
 * was found, false otherwise.
 */
static bool amdgpu_sync_add_later(struct amdgpu_sync *sync, struct dma_fence *f, bool explicit)
{
	struct amdgpu_sync_entry *e;

	hash_for_each_possible(sync->fences, e, node, f->context) {
		if (unlikely(e->fence->context != f->context))
			continue;

		amdgpu_sync_keep_later(&e->fence, f);

		/* Preserve eplicit flag to not loose pipe line sync */
		e->explicit |= explicit;

		return true;
	}
	return false;
}

/**
 * amdgpu_sync_fence - remember to sync to this fence
 *
 * @sync: sync object to add fence to
 * @fence: fence to sync to
 *
 */
int amdgpu_sync_fence(struct amdgpu_device *adev, struct amdgpu_sync *sync,
		      struct dma_fence *f, bool explicit)
{
	struct amdgpu_sync_entry *e;

	if (!f)
		return 0;
	if (amdgpu_sync_same_dev(adev, f) &&
	    amdgpu_sync_get_owner(f) == AMDGPU_FENCE_OWNER_VM)
		amdgpu_sync_keep_later(&sync->last_vm_update, f);

	if (amdgpu_sync_add_later(sync, f, explicit))
		return 0;

	e = kmem_cache_alloc(amdgpu_sync_slab, GFP_KERNEL);
	if (!e)
		return -ENOMEM;

	e->explicit = explicit;

	hash_add(sync->fences, &e->node, f->context);
	e->fence = dma_fence_get(f);
	return 0;
}

/**
 * amdgpu_sync_resv - sync to a reservation object
 *
 * @sync: sync object to add fences from reservation object to
 * @resv: reservation object with embedded fence
 * @explicit_sync: true if we should only sync to the exclusive fence
 *
 * Sync to the fence
 */
int amdgpu_sync_resv(struct amdgpu_device *adev,
		     struct amdgpu_sync *sync,
		     struct reservation_object *resv,
		     void *owner, bool explicit_sync)
{
	struct reservation_object_list *flist;
	struct dma_fence *f;
	void *fence_owner;
	unsigned i;
	int r = 0;

	if (resv == NULL)
		return -EINVAL;

	/* always sync to the exclusive fence */
	f = reservation_object_get_excl(resv);
	r = amdgpu_sync_fence(adev, sync, f, false);

	flist = reservation_object_get_list(resv);
	if (!flist || r)
		return r;

	for (i = 0; i < flist->shared_count; ++i) {
		f = rcu_dereference_protected(flist->shared[i],
					      reservation_object_held(resv));
		/* We only want to trigger KFD eviction fences on
		 * evict or move jobs. Skip KFD fences otherwise.
		 */
		fence_owner = amdgpu_sync_get_owner(f);
		if (fence_owner == AMDGPU_FENCE_OWNER_KFD &&
		    owner != AMDGPU_FENCE_OWNER_UNDEFINED)
			continue;

		if (amdgpu_sync_same_dev(adev, f)) {
			/* VM updates are only interesting
			 * for other VM updates and moves.
			 */
			if ((owner != AMDGPU_FENCE_OWNER_UNDEFINED) &&
			    (fence_owner != AMDGPU_FENCE_OWNER_UNDEFINED) &&
			    ((owner == AMDGPU_FENCE_OWNER_VM) !=
			     (fence_owner == AMDGPU_FENCE_OWNER_VM)))
				continue;

			/* Ignore fence from the same owner and explicit one as
			 * long as it isn't undefined.
			 */
			if (owner != AMDGPU_FENCE_OWNER_UNDEFINED &&
			    (fence_owner == owner || explicit_sync))
				continue;
		}

		r = amdgpu_sync_fence(adev, sync, f, false);
		if (r)
			break;
	}
	return r;
}

/**
 * amdgpu_sync_peek_fence - get the next fence not signaled yet
 *
 * @sync: the sync object
 * @ring: optional ring to use for test
 *
 * Returns the next fence not signaled yet without removing it from the sync
 * object.
 */
struct dma_fence *amdgpu_sync_peek_fence(struct amdgpu_sync *sync,
					 struct amdgpu_ring *ring)
{
	struct amdgpu_sync_entry *e;
	struct hlist_node *tmp;
	int i;

	hash_for_each_safe(sync->fences, i, tmp, e, node) {
		struct dma_fence *f = e->fence;
		struct drm_sched_fence *s_fence = to_drm_sched_fence(f);

		if (dma_fence_is_signaled(f)) {
			hash_del(&e->node);
			dma_fence_put(f);
			kmem_cache_free(amdgpu_sync_slab, e);
			continue;
		}
		if (ring && s_fence) {
			/* For fences from the same ring it is sufficient
			 * when they are scheduled.
			 */
			if (s_fence->sched == &ring->sched) {
				if (dma_fence_is_signaled(&s_fence->scheduled))
					continue;

				return &s_fence->scheduled;
			}
		}

		return f;
	}

	return NULL;
}

/**
 * amdgpu_sync_get_fence - get the next fence from the sync object
 *
 * @sync: sync object to use
 * @explicit: true if the next fence is explicit
 *
 * Get and removes the next fence from the sync object not signaled yet.
 */
struct dma_fence *amdgpu_sync_get_fence(struct amdgpu_sync *sync, bool *explicit)
{
	struct amdgpu_sync_entry *e;
	struct hlist_node *tmp;
	struct dma_fence *f;
	int i;
	hash_for_each_safe(sync->fences, i, tmp, e, node) {

		f = e->fence;
		if (explicit)
			*explicit = e->explicit;

		hash_del(&e->node);
		kmem_cache_free(amdgpu_sync_slab, e);

		if (!dma_fence_is_signaled(f))
			return f;

		dma_fence_put(f);
	}
	return NULL;
}

/**
 * amdgpu_sync_clone - clone a sync object
 *
 * @source: sync object to clone
 * @clone: pointer to destination sync object
 *
 * Adds references to all unsignaled fences in @source to @clone. Also
 * removes signaled fences from @source while at it.
 */
int amdgpu_sync_clone(struct amdgpu_sync *source, struct amdgpu_sync *clone)
{
	struct amdgpu_sync_entry *e;
	struct hlist_node *tmp;
	struct dma_fence *f;
	int i, r;

	hash_for_each_safe(source->fences, i, tmp, e, node) {
		f = e->fence;
		if (!dma_fence_is_signaled(f)) {
			r = amdgpu_sync_fence(NULL, clone, f, e->explicit);
			if (r)
				return r;
		} else {
			hash_del(&e->node);
			dma_fence_put(f);
			kmem_cache_free(amdgpu_sync_slab, e);
		}
	}

	dma_fence_put(clone->last_vm_update);
	clone->last_vm_update = dma_fence_get(source->last_vm_update);

	return 0;
}

int amdgpu_sync_wait(struct amdgpu_sync *sync, bool intr)
{
	struct amdgpu_sync_entry *e;
	struct hlist_node *tmp;
	int i, r;

	hash_for_each_safe(sync->fences, i, tmp, e, node) {
		r = dma_fence_wait(e->fence, intr);
		if (r)
			return r;

		hash_del(&e->node);
		dma_fence_put(e->fence);
		kmem_cache_free(amdgpu_sync_slab, e);
	}

	return 0;
}

/**
 * amdgpu_sync_free - free the sync object
 *
 * @sync: sync object to use
 *
 * Free the sync object.
 */
void amdgpu_sync_free(struct amdgpu_sync *sync)
{
	struct amdgpu_sync_entry *e;
	struct hlist_node *tmp;
	unsigned i;

	hash_for_each_safe(sync->fences, i, tmp, e, node) {
		hash_del(&e->node);
		dma_fence_put(e->fence);
		kmem_cache_free(amdgpu_sync_slab, e);
	}

	dma_fence_put(sync->last_vm_update);
}

/**
 * amdgpu_sync_init - init sync object subsystem
 *
 * Allocate the slab allocator.
 */
int amdgpu_sync_init(void)
{
	amdgpu_sync_slab = kmem_cache_create(
		"amdgpu_sync", sizeof(struct amdgpu_sync_entry), 0,
		SLAB_HWCACHE_ALIGN, NULL);
	if (!amdgpu_sync_slab)
		return -ENOMEM;

	return 0;
}

/**
 * amdgpu_sync_fini - fini sync object subsystem
 *
 * Free the slab allocator.
 */
void amdgpu_sync_fini(void)
{
	kmem_cache_destroy(amdgpu_sync_slab);
}