summaryrefslogtreecommitdiff
path: root/src/mesa/drivers/dri/i965/brw_sync.c
blob: 639164887913e6956ba1cc1c95188c16354150c1 (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
/*
 * Copyright © 2008 Intel Corporation
 *
 * 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, sublicense,
 * 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 above copyright notice and this permission notice (including the next
 * paragraph) shall be included in all copies or substantial portions of the
 * Software.
 *
 * 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 NONINFRINGEMENT.  IN NO EVENT SHALL
 * THE AUTHORS OR COPYRIGHT HOLDERS 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.
 *
 * Authors:
 *    Eric Anholt <eric@anholt.net>
 *
 */

/**
 * \file
 * \brief Support for GL_ARB_sync and EGL_KHR_fence_sync.
 *
 * GL_ARB_sync is implemented by flushing the current batchbuffer and keeping a
 * reference on it.  We can then check for completion or wait for completion
 * using the normal buffer object mechanisms.  This does mean that if an
 * application is using many sync objects, it will emit small batchbuffers
 * which may end up being a significant overhead.  In other tests of removing
 * gratuitous batchbuffer syncs in Mesa, it hasn't appeared to be a significant
 * performance bottleneck, though.
 */

#include <libsync.h> /* Requires Android or libdrm-2.4.72 */

#include "main/imports.h"

#include "brw_context.h"
#include "intel_batchbuffer.h"

struct brw_fence {
   struct brw_context *brw;

   enum brw_fence_type {
      /** The fence waits for completion of brw_fence::batch_bo. */
      BRW_FENCE_TYPE_BO_WAIT,

      /** The fence waits for brw_fence::sync_fd to signal. */
      BRW_FENCE_TYPE_SYNC_FD,
   } type;

   union {
      struct brw_bo *batch_bo;

      /* This struct owns the fd. */
      int sync_fd;
   };

   mtx_t mutex;
   bool signalled;
};

struct brw_gl_sync {
   struct gl_sync_object gl;
   struct brw_fence fence;
};

static void
brw_fence_init(struct brw_context *brw, struct brw_fence *fence,
               enum brw_fence_type type)
{
   fence->brw = brw;
   fence->type = type;
   mtx_init(&fence->mutex, mtx_plain);

   switch (type) {
   case BRW_FENCE_TYPE_BO_WAIT:
      fence->batch_bo = NULL;
      break;
    case BRW_FENCE_TYPE_SYNC_FD:
      fence->sync_fd = -1;
      break;
   }
}

static void
brw_fence_finish(struct brw_fence *fence)
{
   switch (fence->type) {
   case BRW_FENCE_TYPE_BO_WAIT:
      if (fence->batch_bo)
         brw_bo_unreference(fence->batch_bo);
      break;
   case BRW_FENCE_TYPE_SYNC_FD:
      if (fence->sync_fd != -1)
         close(fence->sync_fd);
      break;
   }

   mtx_destroy(&fence->mutex);
}

static bool MUST_CHECK
brw_fence_insert_locked(struct brw_context *brw, struct brw_fence *fence)
{
   __DRIcontext *driContext = brw->driContext;
   __DRIdrawable *driDrawable = driContext->driDrawablePriv;

   /*
    * From KHR_fence_sync:
    *
    *   When the condition of the sync object is satisfied by the fence
    *   command, the sync is signaled by the associated client API context,
    *   causing any eglClientWaitSyncKHR commands (see below) blocking on
    *   <sync> to unblock. The only condition currently supported is
    *   EGL_SYNC_PRIOR_COMMANDS_COMPLETE_KHR, which is satisfied by
    *   completion of the fence command corresponding to the sync object,
    *   and all preceding commands in the associated client API context's
    *   command stream. The sync object will not be signaled until all
    *   effects from these commands on the client API's internal and
    *   framebuffer state are fully realized. No other state is affected by
    *   execution of the fence command.
    *
    * Note the emphasis there on ensuring that the framebuffer is fully
    * realised before the fence is signaled. We cannot just flush the batch,
    * but must also resolve the drawable first. The importance of this is,
    * for example, in creating a fence for a frame to be passed to a
    * remote compositor. Without us flushing the drawable explicitly, the
    * resolve will be in a following batch (when the client finally calls
    * SwapBuffers, or triggers a resolve via some other path) and so the
    * compositor may read the incomplete framebuffer instead.
    */
   if (driDrawable)
      intel_resolve_for_dri2_flush(brw, driDrawable);
   brw_emit_mi_flush(brw);

   switch (fence->type) {
   case BRW_FENCE_TYPE_BO_WAIT:
      assert(!fence->batch_bo);
      assert(!fence->signalled);

      fence->batch_bo = brw->batch.batch.bo;
      brw_bo_reference(fence->batch_bo);

      if (intel_batchbuffer_flush(brw) < 0) {
         brw_bo_unreference(fence->batch_bo);
         fence->batch_bo = NULL;
         return false;
      }
      break;
   case BRW_FENCE_TYPE_SYNC_FD:
      assert(!fence->signalled);

      if (fence->sync_fd == -1) {
         /* Create an out-fence that signals after all pending commands
          * complete.
          */
         if (intel_batchbuffer_flush_fence(brw, -1, &fence->sync_fd) < 0)
            return false;
         assert(fence->sync_fd != -1);
      } else {
         /* Wait on the in-fence before executing any subsequently submitted
          * commands.
          */
         if (intel_batchbuffer_flush(brw) < 0)
            return false;

         /* Emit a dummy batch just for the fence. */
         brw_emit_mi_flush(brw);
         if (intel_batchbuffer_flush_fence(brw, fence->sync_fd, NULL) < 0)
            return false;
      }
      break;
   }

   return true;
}

static bool MUST_CHECK
brw_fence_insert(struct brw_context *brw, struct brw_fence *fence)
{
   bool ret;

   mtx_lock(&fence->mutex);
   ret = brw_fence_insert_locked(brw, fence);
   mtx_unlock(&fence->mutex);

   return ret;
}

static bool
brw_fence_has_completed_locked(struct brw_fence *fence)
{
   if (fence->signalled)
      return true;

   switch (fence->type) {
   case BRW_FENCE_TYPE_BO_WAIT:
      if (!fence->batch_bo) {
         /* There may be no batch if intel_batchbuffer_flush() failed. */
         return false;
      }

      if (brw_bo_busy(fence->batch_bo))
         return false;

      brw_bo_unreference(fence->batch_bo);
      fence->batch_bo = NULL;
      fence->signalled = true;

      return true;

   case BRW_FENCE_TYPE_SYNC_FD:
      assert(fence->sync_fd != -1);

      if (sync_wait(fence->sync_fd, 0) == -1)
         return false;

      fence->signalled = true;

      return true;
   }

   return false;
}

static bool
brw_fence_has_completed(struct brw_fence *fence)
{
   bool ret;

   mtx_lock(&fence->mutex);
   ret = brw_fence_has_completed_locked(fence);
   mtx_unlock(&fence->mutex);

   return ret;
}

static bool
brw_fence_client_wait_locked(struct brw_context *brw, struct brw_fence *fence,
                             uint64_t timeout)
{
   int32_t timeout_i32;

   if (fence->signalled)
      return true;

   switch (fence->type) {
   case BRW_FENCE_TYPE_BO_WAIT:
      if (!fence->batch_bo) {
         /* There may be no batch if intel_batchbuffer_flush() failed. */
         return false;
      }

      /* DRM_IOCTL_I915_GEM_WAIT uses a signed 64 bit timeout and returns
       * immediately for timeouts <= 0.  The best we can do is to clamp the
       * timeout to INT64_MAX.  This limits the maximum timeout from 584 years to
       * 292 years - likely not a big deal.
       */
      if (timeout > INT64_MAX)
         timeout = INT64_MAX;

      if (brw_bo_wait(fence->batch_bo, timeout) != 0)
         return false;

      fence->signalled = true;
      brw_bo_unreference(fence->batch_bo);
      fence->batch_bo = NULL;

      return true;
   case BRW_FENCE_TYPE_SYNC_FD:
      if (fence->sync_fd == -1)
         return false;

      if (timeout > INT32_MAX)
         timeout_i32 = -1;
      else
         timeout_i32 = timeout;

      if (sync_wait(fence->sync_fd, timeout_i32) == -1)
         return false;

      fence->signalled = true;
      return true;
   }

   assert(!"bad enum brw_fence_type");
   return false;
}

/**
 * Return true if the function successfully signals or has already signalled.
 * (This matches the behavior expected from __DRI2fence::client_wait_sync).
 */
static bool
brw_fence_client_wait(struct brw_context *brw, struct brw_fence *fence,
                      uint64_t timeout)
{
   bool ret;

   mtx_lock(&fence->mutex);
   ret = brw_fence_client_wait_locked(brw, fence, timeout);
   mtx_unlock(&fence->mutex);

   return ret;
}

static void
brw_fence_server_wait(struct brw_context *brw, struct brw_fence *fence)
{
   switch (fence->type) {
   case BRW_FENCE_TYPE_BO_WAIT:
      /* We have nothing to do for WaitSync.  Our GL command stream is sequential,
       * so given that the sync object has already flushed the batchbuffer, any
       * batchbuffers coming after this waitsync will naturally not occur until
       * the previous one is done.
       */
      break;
   case BRW_FENCE_TYPE_SYNC_FD:
      assert(fence->sync_fd != -1);

      /* The user wants explicit synchronization, so give them what they want. */
      if (!brw_fence_insert(brw, fence)) {
         /* FIXME: There exists no way yet to report an error here. If an error
          * occurs, continue silently and hope for the best.
          */
      }
      break;
   }
}

static struct gl_sync_object *
brw_gl_new_sync(struct gl_context *ctx)
{
   struct brw_gl_sync *sync;

   sync = calloc(1, sizeof(*sync));
   if (!sync)
      return NULL;

   return &sync->gl;
}

static void
brw_gl_delete_sync(struct gl_context *ctx, struct gl_sync_object *_sync)
{
   struct brw_gl_sync *sync = (struct brw_gl_sync *) _sync;

   brw_fence_finish(&sync->fence);
   free(sync);
}

static void
brw_gl_fence_sync(struct gl_context *ctx, struct gl_sync_object *_sync,
                  GLenum condition, GLbitfield flags)
{
   struct brw_context *brw = brw_context(ctx);
   struct brw_gl_sync *sync = (struct brw_gl_sync *) _sync;

   /* brw_fence_insert_locked() assumes it must do a complete flush */
   assert(condition == GL_SYNC_GPU_COMMANDS_COMPLETE);

   brw_fence_init(brw, &sync->fence, BRW_FENCE_TYPE_BO_WAIT);

   if (!brw_fence_insert_locked(brw, &sync->fence)) {
      /* FIXME: There exists no way to report a GL error here. If an error
       * occurs, continue silently and hope for the best.
       */
   }
}

static void
brw_gl_client_wait_sync(struct gl_context *ctx, struct gl_sync_object *_sync,
                        GLbitfield flags, GLuint64 timeout)
{
   struct brw_context *brw = brw_context(ctx);
   struct brw_gl_sync *sync = (struct brw_gl_sync *) _sync;

   if (brw_fence_client_wait(brw, &sync->fence, timeout))
      sync->gl.StatusFlag = 1;
}

static void
brw_gl_server_wait_sync(struct gl_context *ctx, struct gl_sync_object *_sync,
                          GLbitfield flags, GLuint64 timeout)
{
   struct brw_context *brw = brw_context(ctx);
   struct brw_gl_sync *sync = (struct brw_gl_sync *) _sync;

   brw_fence_server_wait(brw, &sync->fence);
}

static void
brw_gl_check_sync(struct gl_context *ctx, struct gl_sync_object *_sync)
{
   struct brw_gl_sync *sync = (struct brw_gl_sync *) _sync;

   if (brw_fence_has_completed(&sync->fence))
      sync->gl.StatusFlag = 1;
}

void
brw_init_syncobj_functions(struct dd_function_table *functions)
{
   functions->NewSyncObject = brw_gl_new_sync;
   functions->DeleteSyncObject = brw_gl_delete_sync;
   functions->FenceSync = brw_gl_fence_sync;
   functions->CheckSync = brw_gl_check_sync;
   functions->ClientWaitSync = brw_gl_client_wait_sync;
   functions->ServerWaitSync = brw_gl_server_wait_sync;
}

static void *
brw_dri_create_fence(__DRIcontext *ctx)
{
   struct brw_context *brw = ctx->driverPrivate;
   struct brw_fence *fence;

   fence = calloc(1, sizeof(*fence));
   if (!fence)
      return NULL;

   brw_fence_init(brw, fence, BRW_FENCE_TYPE_BO_WAIT);

   if (!brw_fence_insert_locked(brw, fence)) {
      brw_fence_finish(fence);
      free(fence);
      return NULL;
   }

   return fence;
}

static void
brw_dri_destroy_fence(__DRIscreen *dri_screen, void *_fence)
{
   struct brw_fence *fence = _fence;

   brw_fence_finish(fence);
   free(fence);
}

static GLboolean
brw_dri_client_wait_sync(__DRIcontext *ctx, void *_fence, unsigned flags,
                         uint64_t timeout)
{
   struct brw_fence *fence = _fence;

   return brw_fence_client_wait(fence->brw, fence, timeout);
}

static void
brw_dri_server_wait_sync(__DRIcontext *ctx, void *_fence, unsigned flags)
{
   struct brw_fence *fence = _fence;

   /* We might be called here with a NULL fence as a result of WaitSyncKHR
    * on a EGL_KHR_reusable_sync fence. Nothing to do here in such case.
    */
   if (!fence)
      return;

   brw_fence_server_wait(fence->brw, fence);
}

static unsigned
brw_dri_get_capabilities(__DRIscreen *dri_screen)
{
   struct intel_screen *screen = dri_screen->driverPrivate;
   unsigned caps = 0;

   if (screen->has_exec_fence)
      caps |=  __DRI_FENCE_CAP_NATIVE_FD;

   return caps;
}

static void *
brw_dri_create_fence_fd(__DRIcontext *dri_ctx, int fd)
{
   struct brw_context *brw = dri_ctx->driverPrivate;
   struct brw_fence *fence;

   assert(brw->screen->has_exec_fence);

   fence = calloc(1, sizeof(*fence));
   if (!fence)
      return NULL;

   brw_fence_init(brw, fence, BRW_FENCE_TYPE_SYNC_FD);

   if (fd == -1) {
      /* Create an out-fence fd */
      if (!brw_fence_insert_locked(brw, fence))
         goto fail;
   } else {
      /* Import the sync fd as an in-fence. */
      fence->sync_fd = dup(fd);
   }

   assert(fence->sync_fd != -1);

   return fence;

fail:
   brw_fence_finish(fence);
   free(fence);
   return NULL;
}

static int
brw_dri_get_fence_fd_locked(struct brw_fence *fence)
{
   assert(fence->type == BRW_FENCE_TYPE_SYNC_FD);
   return dup(fence->sync_fd);
}

static int
brw_dri_get_fence_fd(__DRIscreen *dri_screen, void *_fence)
{
   struct brw_fence *fence = _fence;
   int fd;

   mtx_lock(&fence->mutex);
   fd = brw_dri_get_fence_fd_locked(fence);
   mtx_unlock(&fence->mutex);

   return fd;
}

const __DRI2fenceExtension intelFenceExtension = {
   .base = { __DRI2_FENCE, 2 },

   .create_fence = brw_dri_create_fence,
   .destroy_fence = brw_dri_destroy_fence,
   .client_wait_sync = brw_dri_client_wait_sync,
   .server_wait_sync = brw_dri_server_wait_sync,
   .get_fence_from_cl_event = NULL,
   .get_capabilities = brw_dri_get_capabilities,
   .create_fence_fd = brw_dri_create_fence_fd,
   .get_fence_fd = brw_dri_get_fence_fd,
};