1 // SPDX-License-Identifier: MIT
2 /*
3  * Copyright 2014 Advanced Micro Devices, Inc.
4  * All Rights Reserved.
5  *
6  * Permission is hereby granted, free of charge, to any person obtaining a
7  * copy of this software and associated documentation files (the
8  * "Software"), to deal in the Software without restriction, including
9  * without limitation the rights to use, copy, modify, merge, publish,
10  * distribute, sub license, and/or sell copies of the Software, and to
11  * permit persons to whom the Software is furnished to do so, subject to
12  * the following conditions:
13  *
14  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
15  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
16  * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
17  * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
18  * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
19  * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
20  * USE OR OTHER DEALINGS IN THE SOFTWARE.
21  *
22  * The above copyright notice and this permission notice (including the
23  * next paragraph) shall be included in all copies or substantial portions
24  * of the Software.
25  *
26  */
27 /*
28  * Authors:
29  *    Christian König <christian.koenig@amd.com>
30  */
31 
32 #include <linux/dma-fence-chain.h>
33 
34 #include "amdgpu.h"
35 #include "amdgpu_trace.h"
36 #include "amdgpu_amdkfd.h"
37 
38 struct amdgpu_sync_entry {
39 	struct hlist_node	node;
40 	struct dma_fence	*fence;
41 };
42 
43 static struct kmem_cache *amdgpu_sync_slab;
44 
45 /**
46  * amdgpu_sync_create - zero init sync object
47  *
48  * @sync: sync object to initialize
49  *
50  * Just clear the sync object for now.
51  */
amdgpu_sync_create(struct amdgpu_sync * sync)52 void amdgpu_sync_create(struct amdgpu_sync *sync)
53 {
54 	hash_init(sync->fences);
55 }
56 
57 /**
58  * amdgpu_sync_same_dev - test if fence belong to us
59  *
60  * @adev: amdgpu device to use for the test
61  * @f: fence to test
62  *
63  * Test if the fence was issued by us.
64  */
amdgpu_sync_same_dev(struct amdgpu_device * adev,struct dma_fence * f)65 static bool amdgpu_sync_same_dev(struct amdgpu_device *adev,
66 				 struct dma_fence *f)
67 {
68 	struct drm_sched_fence *s_fence = to_drm_sched_fence(f);
69 
70 	if (s_fence) {
71 		struct amdgpu_ring *ring;
72 
73 		ring = container_of(s_fence->sched, struct amdgpu_ring, sched);
74 		return ring->adev == adev;
75 	}
76 
77 	return false;
78 }
79 
80 /**
81  * amdgpu_sync_get_owner - extract the owner of a fence
82  *
83  * @f: fence get the owner from
84  *
85  * Extract who originally created the fence.
86  */
amdgpu_sync_get_owner(struct dma_fence * f)87 static void *amdgpu_sync_get_owner(struct dma_fence *f)
88 {
89 	struct drm_sched_fence *s_fence;
90 	struct amdgpu_amdkfd_fence *kfd_fence;
91 
92 	if (!f)
93 		return AMDGPU_FENCE_OWNER_UNDEFINED;
94 
95 	s_fence = to_drm_sched_fence(f);
96 	if (s_fence)
97 		return s_fence->owner;
98 
99 	kfd_fence = to_amdgpu_amdkfd_fence(f);
100 	if (kfd_fence)
101 		return AMDGPU_FENCE_OWNER_KFD;
102 
103 	return AMDGPU_FENCE_OWNER_UNDEFINED;
104 }
105 
106 /**
107  * amdgpu_sync_keep_later - Keep the later fence
108  *
109  * @keep: existing fence to test
110  * @fence: new fence
111  *
112  * Either keep the existing fence or the new one, depending which one is later.
113  */
amdgpu_sync_keep_later(struct dma_fence ** keep,struct dma_fence * fence)114 static void amdgpu_sync_keep_later(struct dma_fence **keep,
115 				   struct dma_fence *fence)
116 {
117 	if (*keep && dma_fence_is_later(*keep, fence))
118 		return;
119 
120 	dma_fence_put(*keep);
121 	*keep = dma_fence_get(fence);
122 }
123 
124 /**
125  * amdgpu_sync_add_later - add the fence to the hash
126  *
127  * @sync: sync object to add the fence to
128  * @f: fence to add
129  *
130  * Tries to add the fence to an existing hash entry. Returns true when an entry
131  * was found, false otherwise.
132  */
amdgpu_sync_add_later(struct amdgpu_sync * sync,struct dma_fence * f)133 static bool amdgpu_sync_add_later(struct amdgpu_sync *sync, struct dma_fence *f)
134 {
135 	struct amdgpu_sync_entry *e;
136 
137 	hash_for_each_possible(sync->fences, e, node, f->context) {
138 		if (unlikely(e->fence->context != f->context))
139 			continue;
140 
141 		amdgpu_sync_keep_later(&e->fence, f);
142 		return true;
143 	}
144 	return false;
145 }
146 
147 /**
148  * amdgpu_sync_fence - remember to sync to this fence
149  *
150  * @sync: sync object to add fence to
151  * @f: fence to sync to
152  *
153  * Add the fence to the sync object.
154  */
amdgpu_sync_fence(struct amdgpu_sync * sync,struct dma_fence * f)155 int amdgpu_sync_fence(struct amdgpu_sync *sync, struct dma_fence *f)
156 {
157 	struct amdgpu_sync_entry *e;
158 
159 	if (!f)
160 		return 0;
161 
162 	if (amdgpu_sync_add_later(sync, f))
163 		return 0;
164 
165 	e = kmem_cache_alloc(amdgpu_sync_slab, GFP_KERNEL);
166 	if (!e)
167 		return -ENOMEM;
168 
169 	hash_add(sync->fences, &e->node, f->context);
170 	e->fence = dma_fence_get(f);
171 	return 0;
172 }
173 
174 /* Determine based on the owner and mode if we should sync to a fence or not */
amdgpu_sync_test_fence(struct amdgpu_device * adev,enum amdgpu_sync_mode mode,void * owner,struct dma_fence * f)175 static bool amdgpu_sync_test_fence(struct amdgpu_device *adev,
176 				   enum amdgpu_sync_mode mode,
177 				   void *owner, struct dma_fence *f)
178 {
179 	void *fence_owner = amdgpu_sync_get_owner(f);
180 
181 	/* Always sync to moves, no matter what */
182 	if (fence_owner == AMDGPU_FENCE_OWNER_UNDEFINED)
183 		return true;
184 
185 	/* We only want to trigger KFD eviction fences on
186 	 * evict or move jobs. Skip KFD fences otherwise.
187 	 */
188 	if (fence_owner == AMDGPU_FENCE_OWNER_KFD &&
189 	    owner != AMDGPU_FENCE_OWNER_UNDEFINED)
190 		return false;
191 
192 	/* Never sync to VM updates either. */
193 	if (fence_owner == AMDGPU_FENCE_OWNER_VM &&
194 	    owner != AMDGPU_FENCE_OWNER_UNDEFINED)
195 		return false;
196 
197 	/* Ignore fences depending on the sync mode */
198 	switch (mode) {
199 	case AMDGPU_SYNC_ALWAYS:
200 		return true;
201 
202 	case AMDGPU_SYNC_NE_OWNER:
203 		if (amdgpu_sync_same_dev(adev, f) &&
204 		    fence_owner == owner)
205 			return false;
206 		break;
207 
208 	case AMDGPU_SYNC_EQ_OWNER:
209 		if (amdgpu_sync_same_dev(adev, f) &&
210 		    fence_owner != owner)
211 			return false;
212 		break;
213 
214 	case AMDGPU_SYNC_EXPLICIT:
215 		return false;
216 	}
217 
218 	WARN(debug_evictions && fence_owner == AMDGPU_FENCE_OWNER_KFD,
219 	     "Adding eviction fence to sync obj");
220 	return true;
221 }
222 
223 /**
224  * amdgpu_sync_resv - sync to a reservation object
225  *
226  * @adev: amdgpu device
227  * @sync: sync object to add fences from reservation object to
228  * @resv: reservation object with embedded fence
229  * @mode: how owner affects which fences we sync to
230  * @owner: owner of the planned job submission
231  *
232  * Sync to the fence
233  */
amdgpu_sync_resv(struct amdgpu_device * adev,struct amdgpu_sync * sync,struct dma_resv * resv,enum amdgpu_sync_mode mode,void * owner)234 int amdgpu_sync_resv(struct amdgpu_device *adev, struct amdgpu_sync *sync,
235 		     struct dma_resv *resv, enum amdgpu_sync_mode mode,
236 		     void *owner)
237 {
238 	struct dma_resv_iter cursor;
239 	struct dma_fence *f;
240 	int r;
241 
242 	if (resv == NULL)
243 		return -EINVAL;
244 
245 	/* TODO: Use DMA_RESV_USAGE_READ here */
246 	dma_resv_for_each_fence(&cursor, resv, DMA_RESV_USAGE_BOOKKEEP, f) {
247 		dma_fence_chain_for_each(f, f) {
248 			struct dma_fence *tmp = dma_fence_chain_contained(f);
249 
250 			if (amdgpu_sync_test_fence(adev, mode, owner, tmp)) {
251 				r = amdgpu_sync_fence(sync, f);
252 				dma_fence_put(f);
253 				if (r)
254 					return r;
255 				break;
256 			}
257 		}
258 	}
259 	return 0;
260 }
261 
262 /* Free the entry back to the slab */
amdgpu_sync_entry_free(struct amdgpu_sync_entry * e)263 static void amdgpu_sync_entry_free(struct amdgpu_sync_entry *e)
264 {
265 	hash_del(&e->node);
266 	dma_fence_put(e->fence);
267 	kmem_cache_free(amdgpu_sync_slab, e);
268 }
269 
270 /**
271  * amdgpu_sync_peek_fence - get the next fence not signaled yet
272  *
273  * @sync: the sync object
274  * @ring: optional ring to use for test
275  *
276  * Returns the next fence not signaled yet without removing it from the sync
277  * object.
278  */
amdgpu_sync_peek_fence(struct amdgpu_sync * sync,struct amdgpu_ring * ring)279 struct dma_fence *amdgpu_sync_peek_fence(struct amdgpu_sync *sync,
280 					 struct amdgpu_ring *ring)
281 {
282 	struct amdgpu_sync_entry *e;
283 	struct hlist_node *tmp;
284 	int i;
285 
286 	hash_for_each_safe(sync->fences, i, tmp, e, node) {
287 		struct dma_fence *f = e->fence;
288 		struct drm_sched_fence *s_fence = to_drm_sched_fence(f);
289 
290 		if (dma_fence_is_signaled(f)) {
291 			amdgpu_sync_entry_free(e);
292 			continue;
293 		}
294 		if (ring && s_fence) {
295 			/* For fences from the same ring it is sufficient
296 			 * when they are scheduled.
297 			 */
298 			if (s_fence->sched == &ring->sched) {
299 				if (dma_fence_is_signaled(&s_fence->scheduled))
300 					continue;
301 
302 				return &s_fence->scheduled;
303 			}
304 		}
305 
306 		return f;
307 	}
308 
309 	return NULL;
310 }
311 
312 /**
313  * amdgpu_sync_get_fence - get the next fence from the sync object
314  *
315  * @sync: sync object to use
316  *
317  * Get and removes the next fence from the sync object not signaled yet.
318  */
amdgpu_sync_get_fence(struct amdgpu_sync * sync)319 struct dma_fence *amdgpu_sync_get_fence(struct amdgpu_sync *sync)
320 {
321 	struct amdgpu_sync_entry *e;
322 	struct hlist_node *tmp;
323 	struct dma_fence *f;
324 	int i;
325 
326 	hash_for_each_safe(sync->fences, i, tmp, e, node) {
327 
328 		f = e->fence;
329 
330 		hash_del(&e->node);
331 		kmem_cache_free(amdgpu_sync_slab, e);
332 
333 		if (!dma_fence_is_signaled(f))
334 			return f;
335 
336 		dma_fence_put(f);
337 	}
338 	return NULL;
339 }
340 
341 /**
342  * amdgpu_sync_clone - clone a sync object
343  *
344  * @source: sync object to clone
345  * @clone: pointer to destination sync object
346  *
347  * Adds references to all unsignaled fences in @source to @clone. Also
348  * removes signaled fences from @source while at it.
349  */
amdgpu_sync_clone(struct amdgpu_sync * source,struct amdgpu_sync * clone)350 int amdgpu_sync_clone(struct amdgpu_sync *source, struct amdgpu_sync *clone)
351 {
352 	struct amdgpu_sync_entry *e;
353 	struct hlist_node *tmp;
354 	struct dma_fence *f;
355 	int i, r;
356 
357 	hash_for_each_safe(source->fences, i, tmp, e, node) {
358 		f = e->fence;
359 		if (!dma_fence_is_signaled(f)) {
360 			r = amdgpu_sync_fence(clone, f);
361 			if (r)
362 				return r;
363 		} else {
364 			amdgpu_sync_entry_free(e);
365 		}
366 	}
367 
368 	return 0;
369 }
370 
371 /**
372  * amdgpu_sync_push_to_job - push fences into job
373  * @sync: sync object to get the fences from
374  * @job: job to push the fences into
375  *
376  * Add all unsignaled fences from sync to job.
377  */
amdgpu_sync_push_to_job(struct amdgpu_sync * sync,struct amdgpu_job * job)378 int amdgpu_sync_push_to_job(struct amdgpu_sync *sync, struct amdgpu_job *job)
379 {
380 	struct amdgpu_sync_entry *e;
381 	struct hlist_node *tmp;
382 	struct dma_fence *f;
383 	int i, r;
384 
385 	hash_for_each_safe(sync->fences, i, tmp, e, node) {
386 		f = e->fence;
387 		if (dma_fence_is_signaled(f)) {
388 			amdgpu_sync_entry_free(e);
389 			continue;
390 		}
391 
392 		dma_fence_get(f);
393 		r = drm_sched_job_add_dependency(&job->base, f);
394 		if (r) {
395 			dma_fence_put(f);
396 			return r;
397 		}
398 	}
399 	return 0;
400 }
401 
amdgpu_sync_wait(struct amdgpu_sync * sync,bool intr)402 int amdgpu_sync_wait(struct amdgpu_sync *sync, bool intr)
403 {
404 	struct amdgpu_sync_entry *e;
405 	struct hlist_node *tmp;
406 	int i, r;
407 
408 	hash_for_each_safe(sync->fences, i, tmp, e, node) {
409 		r = dma_fence_wait(e->fence, intr);
410 		if (r)
411 			return r;
412 
413 		amdgpu_sync_entry_free(e);
414 	}
415 
416 	return 0;
417 }
418 
419 /**
420  * amdgpu_sync_free - free the sync object
421  *
422  * @sync: sync object to use
423  *
424  * Free the sync object.
425  */
amdgpu_sync_free(struct amdgpu_sync * sync)426 void amdgpu_sync_free(struct amdgpu_sync *sync)
427 {
428 	struct amdgpu_sync_entry *e;
429 	struct hlist_node *tmp;
430 	unsigned int i;
431 
432 	hash_for_each_safe(sync->fences, i, tmp, e, node)
433 		amdgpu_sync_entry_free(e);
434 }
435 
436 /**
437  * amdgpu_sync_init - init sync object subsystem
438  *
439  * Allocate the slab allocator.
440  */
amdgpu_sync_init(void)441 int amdgpu_sync_init(void)
442 {
443 	amdgpu_sync_slab = kmem_cache_create(
444 		"amdgpu_sync", sizeof(struct amdgpu_sync_entry), 0,
445 		SLAB_HWCACHE_ALIGN, NULL);
446 	if (!amdgpu_sync_slab)
447 		return -ENOMEM;
448 
449 	return 0;
450 }
451 
452 /**
453  * amdgpu_sync_fini - fini sync object subsystem
454  *
455  * Free the slab allocator.
456  */
amdgpu_sync_fini(void)457 void amdgpu_sync_fini(void)
458 {
459 	kmem_cache_destroy(amdgpu_sync_slab);
460 }
461