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
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
|
/*
* Copyright © 2008 Jérôme Glisse
* Copyright © 2010 Marek Olšák <maraeo@gmail.com>
* 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:
* Marek Olšák <maraeo@gmail.com>
*
* Based on work from libdrm_radeon by:
* Aapo Tahkola <aet@rasterburn.org>
* Nicolai Haehnle <prefect_@gmx.net>
* Jérôme Glisse <glisse@freedesktop.org>
*/
/*
This file replaces libdrm's radeon_cs_gem with our own implemention.
It's optimized specifically for Radeon DRM.
Reloc writes and space checking are faster and simpler than their
counterparts in libdrm (the time complexity of all the functions
is O(1) in nearly all scenarios, thanks to hashing).
It works like this:
cs_add_reloc(cs, buf, read_domain, write_domain) adds a new relocation and
also adds the size of 'buf' to the used_gart and used_vram winsys variables
based on the domains, which are simply or'd for the accounting purposes.
The adding is skipped if the reloc is already present in the list, but it
accounts any newly-referenced domains.
cs_validate is then called, which just checks:
used_vram/gart < vram/gart_size * 0.8
The 0.8 number allows for some memory fragmentation. If the validation
fails, the pipe driver flushes CS and tries do the validation again,
i.e. it validates only that one operation. If it fails again, it drops
the operation on the floor and prints some nasty message to stderr.
(done in the pipe driver)
cs_write_reloc(cs, buf) just writes a reloc that has been added using
cs_add_reloc. The read_domain and write_domain parameters have been removed,
because we already specify them in cs_add_reloc.
*/
#include "radeon_drm_cs.h"
#include "util/u_memory.h"
#include "os/os_time.h"
#include <stdio.h>
#include <stdlib.h>
#include <stdint.h>
#include <xf86drm.h>
/*
* this are copy from radeon_drm, once an updated libdrm is released
* we should bump configure.ac requirement for it and remove the following
* field
*/
#ifndef RADEON_CHUNK_ID_FLAGS
#define RADEON_CHUNK_ID_FLAGS 0x03
/* The first dword of RADEON_CHUNK_ID_FLAGS is a uint32 of these flags: */
#define RADEON_CS_KEEP_TILING_FLAGS 0x01
#endif
#ifndef RADEON_CS_USE_VM
#define RADEON_CS_USE_VM 0x02
/* The second dword of RADEON_CHUNK_ID_FLAGS is a uint32 that sets the ring type */
#define RADEON_CS_RING_GFX 0
#define RADEON_CS_RING_COMPUTE 1
#endif
#ifndef RADEON_CS_RING_DMA
#define RADEON_CS_RING_DMA 2
#endif
#ifndef RADEON_CS_RING_UVD
#define RADEON_CS_RING_UVD 3
#endif
#ifndef RADEON_CS_RING_VCE
#define RADEON_CS_RING_VCE 4
#endif
#ifndef RADEON_CS_END_OF_FRAME
#define RADEON_CS_END_OF_FRAME 0x04
#endif
#define RELOC_DWORDS (sizeof(struct drm_radeon_cs_reloc) / sizeof(uint32_t))
static boolean radeon_init_cs_context(struct radeon_cs_context *csc,
struct radeon_drm_winsys *ws)
{
csc->fd = ws->fd;
csc->nrelocs = 512;
csc->relocs_bo = (struct radeon_bo**)
CALLOC(1, csc->nrelocs * sizeof(struct radeon_bo*));
if (!csc->relocs_bo) {
return FALSE;
}
csc->relocs = (struct drm_radeon_cs_reloc*)
CALLOC(1, csc->nrelocs * sizeof(struct drm_radeon_cs_reloc));
if (!csc->relocs) {
FREE(csc->relocs_bo);
return FALSE;
}
csc->chunks[0].chunk_id = RADEON_CHUNK_ID_IB;
csc->chunks[0].length_dw = 0;
csc->chunks[0].chunk_data = (uint64_t)(uintptr_t)csc->buf;
csc->chunks[1].chunk_id = RADEON_CHUNK_ID_RELOCS;
csc->chunks[1].length_dw = 0;
csc->chunks[1].chunk_data = (uint64_t)(uintptr_t)csc->relocs;
csc->chunks[2].chunk_id = RADEON_CHUNK_ID_FLAGS;
csc->chunks[2].length_dw = 2;
csc->chunks[2].chunk_data = (uint64_t)(uintptr_t)&csc->flags;
csc->chunk_array[0] = (uint64_t)(uintptr_t)&csc->chunks[0];
csc->chunk_array[1] = (uint64_t)(uintptr_t)&csc->chunks[1];
csc->chunk_array[2] = (uint64_t)(uintptr_t)&csc->chunks[2];
csc->cs.chunks = (uint64_t)(uintptr_t)csc->chunk_array;
return TRUE;
}
static void radeon_cs_context_cleanup(struct radeon_cs_context *csc)
{
unsigned i;
for (i = 0; i < csc->crelocs; i++) {
p_atomic_dec(&csc->relocs_bo[i]->num_cs_references);
radeon_bo_reference(&csc->relocs_bo[i], NULL);
}
csc->crelocs = 0;
csc->validated_crelocs = 0;
csc->chunks[0].length_dw = 0;
csc->chunks[1].length_dw = 0;
csc->used_gart = 0;
csc->used_vram = 0;
memset(csc->is_handle_added, 0, sizeof(csc->is_handle_added));
}
static void radeon_destroy_cs_context(struct radeon_cs_context *csc)
{
radeon_cs_context_cleanup(csc);
FREE(csc->relocs_bo);
FREE(csc->relocs);
}
static struct radeon_winsys_cs *radeon_drm_cs_create(struct radeon_winsys *rws,
enum ring_type ring_type,
struct radeon_winsys_cs_handle *trace_buf)
{
struct radeon_drm_winsys *ws = radeon_drm_winsys(rws);
struct radeon_drm_cs *cs;
cs = CALLOC_STRUCT(radeon_drm_cs);
if (!cs) {
return NULL;
}
pipe_semaphore_init(&cs->flush_completed, 1);
cs->ws = ws;
cs->trace_buf = (struct radeon_bo*)trace_buf;
if (!radeon_init_cs_context(&cs->csc1, cs->ws)) {
FREE(cs);
return NULL;
}
if (!radeon_init_cs_context(&cs->csc2, cs->ws)) {
radeon_destroy_cs_context(&cs->csc1);
FREE(cs);
return NULL;
}
/* Set the first command buffer as current. */
cs->csc = &cs->csc1;
cs->cst = &cs->csc2;
cs->base.buf = cs->csc->buf;
cs->base.ring_type = ring_type;
p_atomic_inc(&ws->num_cs);
return &cs->base;
}
#define OUT_CS(cs, value) (cs)->buf[(cs)->cdw++] = (value)
static INLINE void update_reloc(struct drm_radeon_cs_reloc *reloc,
enum radeon_bo_domain rd,
enum radeon_bo_domain wd,
unsigned priority,
enum radeon_bo_domain *added_domains)
{
*added_domains = (rd | wd) & ~(reloc->read_domains | reloc->write_domain);
reloc->read_domains |= rd;
reloc->write_domain |= wd;
reloc->flags = MAX2(reloc->flags, priority);
}
int radeon_get_reloc(struct radeon_cs_context *csc, struct radeon_bo *bo)
{
struct drm_radeon_cs_reloc *reloc;
unsigned i;
unsigned hash = bo->handle & (sizeof(csc->is_handle_added)-1);
if (csc->is_handle_added[hash]) {
i = csc->reloc_indices_hashlist[hash];
reloc = &csc->relocs[i];
if (reloc->handle == bo->handle) {
return i;
}
/* Hash collision, look for the BO in the list of relocs linearly. */
for (i = csc->crelocs; i != 0;) {
--i;
reloc = &csc->relocs[i];
if (reloc->handle == bo->handle) {
/* Put this reloc in the hash list.
* This will prevent additional hash collisions if there are
* several consecutive get_reloc calls for the same buffer.
*
* Example: Assuming buffers A,B,C collide in the hash list,
* the following sequence of relocs:
* AAAAAAAAAAABBBBBBBBBBBBBBCCCCCCCC
* will collide here: ^ and here: ^,
* meaning that we should get very few collisions in the end. */
csc->reloc_indices_hashlist[hash] = i;
/*printf("write_reloc collision, hash: %i, handle: %i\n", hash, bo->handle);*/
return i;
}
}
}
return -1;
}
static unsigned radeon_add_reloc(struct radeon_drm_cs *cs,
struct radeon_bo *bo,
enum radeon_bo_usage usage,
enum radeon_bo_domain domains,
unsigned priority,
enum radeon_bo_domain *added_domains)
{
struct radeon_cs_context *csc = cs->csc;
struct drm_radeon_cs_reloc *reloc;
unsigned hash = bo->handle & (sizeof(csc->is_handle_added)-1);
enum radeon_bo_domain rd = usage & RADEON_USAGE_READ ? domains : 0;
enum radeon_bo_domain wd = usage & RADEON_USAGE_WRITE ? domains : 0;
bool update_hash = TRUE;
int i;
priority = MIN2(priority, 15);
*added_domains = 0;
if (csc->is_handle_added[hash]) {
i = csc->reloc_indices_hashlist[hash];
reloc = &csc->relocs[i];
if (reloc->handle != bo->handle) {
/* Hash collision, look for the BO in the list of relocs linearly. */
for (i = csc->crelocs - 1; i >= 0; i--) {
reloc = &csc->relocs[i];
if (reloc->handle == bo->handle) {
/*printf("write_reloc collision, hash: %i, handle: %i\n", hash, bo->handle);*/
break;
}
}
}
if (i >= 0) {
update_reloc(reloc, rd, wd, priority, added_domains);
/* For async DMA, every add_reloc call must add a buffer to the list
* no matter how many duplicates there are. This is due to the fact
* the DMA CS checker doesn't use NOP packets for offset patching,
* but always uses the i-th buffer from the list to patch the i-th
* offset. If there are N offsets in a DMA CS, there must also be N
* buffers in the relocation list.
*
* This doesn't have to be done if virtual memory is enabled,
* because there is no offset patching with virtual memory.
*/
if (cs->base.ring_type != RING_DMA || cs->ws->info.r600_virtual_address) {
csc->reloc_indices_hashlist[hash] = i;
return i;
}
update_hash = FALSE;
}
}
/* New relocation, check if the backing array is large enough. */
if (csc->crelocs >= csc->nrelocs) {
uint32_t size;
csc->nrelocs += 10;
size = csc->nrelocs * sizeof(struct radeon_bo*);
csc->relocs_bo = realloc(csc->relocs_bo, size);
size = csc->nrelocs * sizeof(struct drm_radeon_cs_reloc);
csc->relocs = realloc(csc->relocs, size);
csc->chunks[1].chunk_data = (uint64_t)(uintptr_t)csc->relocs;
}
/* Initialize the new relocation. */
csc->relocs_bo[csc->crelocs] = NULL;
radeon_bo_reference(&csc->relocs_bo[csc->crelocs], bo);
p_atomic_inc(&bo->num_cs_references);
reloc = &csc->relocs[csc->crelocs];
reloc->handle = bo->handle;
reloc->read_domains = rd;
reloc->write_domain = wd;
reloc->flags = priority;
csc->is_handle_added[hash] = TRUE;
if (update_hash) {
csc->reloc_indices_hashlist[hash] = csc->crelocs;
}
csc->chunks[1].length_dw += RELOC_DWORDS;
*added_domains = rd | wd;
return csc->crelocs++;
}
static unsigned radeon_drm_cs_add_reloc(struct radeon_winsys_cs *rcs,
struct radeon_winsys_cs_handle *buf,
enum radeon_bo_usage usage,
enum radeon_bo_domain domains,
enum radeon_bo_priority priority)
{
struct radeon_drm_cs *cs = radeon_drm_cs(rcs);
struct radeon_bo *bo = (struct radeon_bo*)buf;
enum radeon_bo_domain added_domains;
unsigned index = radeon_add_reloc(cs, bo, usage, domains, priority, &added_domains);
if (added_domains & RADEON_DOMAIN_GTT)
cs->csc->used_gart += bo->base.size;
if (added_domains & RADEON_DOMAIN_VRAM)
cs->csc->used_vram += bo->base.size;
return index;
}
static boolean radeon_drm_cs_validate(struct radeon_winsys_cs *rcs)
{
struct radeon_drm_cs *cs = radeon_drm_cs(rcs);
boolean status =
cs->csc->used_gart < cs->ws->info.gart_size * 0.8 &&
cs->csc->used_vram < cs->ws->info.vram_size * 0.8;
if (status) {
cs->csc->validated_crelocs = cs->csc->crelocs;
} else {
/* Remove lately-added relocations. The validation failed with them
* and the CS is about to be flushed because of that. Keep only
* the already-validated relocations. */
unsigned i;
for (i = cs->csc->validated_crelocs; i < cs->csc->crelocs; i++) {
p_atomic_dec(&cs->csc->relocs_bo[i]->num_cs_references);
radeon_bo_reference(&cs->csc->relocs_bo[i], NULL);
}
cs->csc->crelocs = cs->csc->validated_crelocs;
/* Flush if there are any relocs. Clean up otherwise. */
if (cs->csc->crelocs) {
cs->flush_cs(cs->flush_data, RADEON_FLUSH_ASYNC);
} else {
radeon_cs_context_cleanup(cs->csc);
assert(cs->base.cdw == 0);
if (cs->base.cdw != 0) {
fprintf(stderr, "radeon: Unexpected error in %s.\n", __func__);
}
}
}
return status;
}
static boolean radeon_drm_cs_memory_below_limit(struct radeon_winsys_cs *rcs, uint64_t vram, uint64_t gtt)
{
struct radeon_drm_cs *cs = radeon_drm_cs(rcs);
boolean status =
(cs->csc->used_gart + gtt) < cs->ws->info.gart_size * 0.7 &&
(cs->csc->used_vram + vram) < cs->ws->info.vram_size * 0.7;
return status;
}
static void radeon_drm_cs_write_reloc(struct radeon_winsys_cs *rcs,
struct radeon_winsys_cs_handle *buf)
{
struct radeon_drm_cs *cs = radeon_drm_cs(rcs);
struct radeon_bo *bo = (struct radeon_bo*)buf;
unsigned index = radeon_get_reloc(cs->csc, bo);
if (index == -1) {
fprintf(stderr, "radeon: Cannot get a relocation in %s.\n", __func__);
return;
}
OUT_CS(&cs->base, 0xc0001000);
OUT_CS(&cs->base, index * RELOC_DWORDS);
}
void radeon_drm_cs_emit_ioctl_oneshot(struct radeon_drm_cs *cs, struct radeon_cs_context *csc)
{
unsigned i;
if (drmCommandWriteRead(csc->fd, DRM_RADEON_CS,
&csc->cs, sizeof(struct drm_radeon_cs))) {
if (debug_get_bool_option("RADEON_DUMP_CS", FALSE)) {
unsigned i;
fprintf(stderr, "radeon: The kernel rejected CS, dumping...\n");
for (i = 0; i < csc->chunks[0].length_dw; i++) {
fprintf(stderr, "0x%08X\n", csc->buf[i]);
}
} else {
fprintf(stderr, "radeon: The kernel rejected CS, "
"see dmesg for more information.\n");
}
}
if (cs->trace_buf) {
radeon_dump_cs_on_lockup(cs, csc);
}
for (i = 0; i < csc->crelocs; i++)
p_atomic_dec(&csc->relocs_bo[i]->num_active_ioctls);
radeon_cs_context_cleanup(csc);
}
/*
* Make sure previous submission of this cs are completed
*/
void radeon_drm_cs_sync_flush(struct radeon_winsys_cs *rcs)
{
struct radeon_drm_cs *cs = radeon_drm_cs(rcs);
/* Wait for any pending ioctl to complete. */
if (cs->ws->thread) {
pipe_semaphore_wait(&cs->flush_completed);
pipe_semaphore_signal(&cs->flush_completed);
}
}
DEBUG_GET_ONCE_BOOL_OPTION(noop, "RADEON_NOOP", FALSE)
static void radeon_drm_cs_flush(struct radeon_winsys_cs *rcs, unsigned flags, uint32_t cs_trace_id)
{
struct radeon_drm_cs *cs = radeon_drm_cs(rcs);
struct radeon_cs_context *tmp;
switch (cs->base.ring_type) {
case RING_DMA:
/* pad DMA ring to 8 DWs */
if (cs->ws->info.chip_class <= SI) {
while (rcs->cdw & 7)
OUT_CS(&cs->base, 0xf0000000); /* NOP packet */
} else {
while (rcs->cdw & 7)
OUT_CS(&cs->base, 0x00000000); /* NOP packet */
}
break;
case RING_GFX:
/* pad DMA ring to 8 DWs to meet CP fetch alignment requirements
* r6xx, requires at least 4 dw alignment to avoid a hw bug.
*/
if (cs->ws->info.chip_class <= SI) {
while (rcs->cdw & 7)
OUT_CS(&cs->base, 0x80000000); /* type2 nop packet */
} else {
while (rcs->cdw & 7)
OUT_CS(&cs->base, 0xffff1000); /* type3 nop packet */
}
break;
case RING_UVD:
while (rcs->cdw & 15)
OUT_CS(&cs->base, 0x80000000); /* type2 nop packet */
break;
default:
break;
}
if (rcs->cdw > RADEON_MAX_CMDBUF_DWORDS) {
fprintf(stderr, "radeon: command stream overflowed\n");
}
radeon_drm_cs_sync_flush(rcs);
/* Flip command streams. */
tmp = cs->csc;
cs->csc = cs->cst;
cs->cst = tmp;
cs->cst->cs_trace_id = cs_trace_id;
/* If the CS is not empty or overflowed, emit it in a separate thread. */
if (cs->base.cdw && cs->base.cdw <= RADEON_MAX_CMDBUF_DWORDS && !debug_get_option_noop()) {
unsigned i, crelocs = cs->cst->crelocs;
cs->cst->chunks[0].length_dw = cs->base.cdw;
for (i = 0; i < crelocs; i++) {
/* Update the number of active asynchronous CS ioctls for the buffer. */
p_atomic_inc(&cs->cst->relocs_bo[i]->num_active_ioctls);
}
switch (cs->base.ring_type) {
case RING_DMA:
cs->cst->flags[0] = 0;
cs->cst->flags[1] = RADEON_CS_RING_DMA;
cs->cst->cs.num_chunks = 3;
if (cs->ws->info.r600_virtual_address) {
cs->cst->flags[0] |= RADEON_CS_USE_VM;
}
break;
case RING_UVD:
cs->cst->flags[0] = 0;
cs->cst->flags[1] = RADEON_CS_RING_UVD;
cs->cst->cs.num_chunks = 3;
break;
case RING_VCE:
cs->cst->flags[0] = 0;
cs->cst->flags[1] = RADEON_CS_RING_VCE;
cs->cst->cs.num_chunks = 3;
break;
default:
case RING_GFX:
cs->cst->flags[0] = 0;
cs->cst->flags[1] = RADEON_CS_RING_GFX;
cs->cst->cs.num_chunks = 2;
if (flags & RADEON_FLUSH_KEEP_TILING_FLAGS) {
cs->cst->flags[0] |= RADEON_CS_KEEP_TILING_FLAGS;
cs->cst->cs.num_chunks = 3;
}
if (cs->ws->info.r600_virtual_address) {
cs->cst->flags[0] |= RADEON_CS_USE_VM;
cs->cst->cs.num_chunks = 3;
}
if (flags & RADEON_FLUSH_END_OF_FRAME) {
cs->cst->flags[0] |= RADEON_CS_END_OF_FRAME;
cs->cst->cs.num_chunks = 3;
}
if (flags & RADEON_FLUSH_COMPUTE) {
cs->cst->flags[1] = RADEON_CS_RING_COMPUTE;
cs->cst->cs.num_chunks = 3;
}
break;
}
if (cs->ws->thread) {
pipe_semaphore_wait(&cs->flush_completed);
radeon_drm_ws_queue_cs(cs->ws, cs);
if (!(flags & RADEON_FLUSH_ASYNC))
radeon_drm_cs_sync_flush(rcs);
} else {
radeon_drm_cs_emit_ioctl_oneshot(cs, cs->cst);
}
} else {
radeon_cs_context_cleanup(cs->cst);
}
/* Prepare a new CS. */
cs->base.buf = cs->csc->buf;
cs->base.cdw = 0;
}
static void radeon_drm_cs_destroy(struct radeon_winsys_cs *rcs)
{
struct radeon_drm_cs *cs = radeon_drm_cs(rcs);
radeon_drm_cs_sync_flush(rcs);
pipe_semaphore_destroy(&cs->flush_completed);
radeon_cs_context_cleanup(&cs->csc1);
radeon_cs_context_cleanup(&cs->csc2);
p_atomic_dec(&cs->ws->num_cs);
radeon_destroy_cs_context(&cs->csc1);
radeon_destroy_cs_context(&cs->csc2);
FREE(cs);
}
static void radeon_drm_cs_set_flush(struct radeon_winsys_cs *rcs,
void (*flush)(void *ctx, unsigned flags),
void *user)
{
struct radeon_drm_cs *cs = radeon_drm_cs(rcs);
cs->flush_cs = flush;
cs->flush_data = user;
}
static boolean radeon_bo_is_referenced(struct radeon_winsys_cs *rcs,
struct radeon_winsys_cs_handle *_buf,
enum radeon_bo_usage usage)
{
struct radeon_drm_cs *cs = radeon_drm_cs(rcs);
struct radeon_bo *bo = (struct radeon_bo*)_buf;
int index;
if (!bo->num_cs_references)
return FALSE;
index = radeon_get_reloc(cs->csc, bo);
if (index == -1)
return FALSE;
if ((usage & RADEON_USAGE_WRITE) && cs->csc->relocs[index].write_domain)
return TRUE;
if ((usage & RADEON_USAGE_READ) && cs->csc->relocs[index].read_domains)
return TRUE;
return FALSE;
}
/* FENCES */
static struct pipe_fence_handle *
radeon_cs_create_fence(struct radeon_winsys_cs *rcs)
{
struct radeon_drm_cs *cs = radeon_drm_cs(rcs);
struct pb_buffer *fence;
/* Create a fence, which is a dummy BO. */
fence = cs->ws->base.buffer_create(&cs->ws->base, 1, 1, TRUE,
RADEON_DOMAIN_GTT);
/* Add the fence as a dummy relocation. */
cs->ws->base.cs_add_reloc(rcs, cs->ws->base.buffer_get_cs_handle(fence),
RADEON_USAGE_READWRITE, RADEON_DOMAIN_GTT,
RADEON_PRIO_MIN);
return (struct pipe_fence_handle*)fence;
}
static bool radeon_fence_wait(struct radeon_winsys *ws,
struct pipe_fence_handle *fence,
uint64_t timeout)
{
struct pb_buffer *rfence = (struct pb_buffer*)fence;
if (timeout == 0)
return !ws->buffer_is_busy(rfence, RADEON_USAGE_READWRITE);
if (timeout != PIPE_TIMEOUT_INFINITE) {
int64_t start_time = os_time_get();
/* Convert to microseconds. */
timeout /= 1000;
/* Wait in a loop. */
while (ws->buffer_is_busy(rfence, RADEON_USAGE_READWRITE)) {
if (os_time_get() - start_time >= timeout) {
return FALSE;
}
os_time_sleep(10);
}
return TRUE;
}
ws->buffer_wait(rfence, RADEON_USAGE_READWRITE);
return TRUE;
}
static void radeon_fence_reference(struct pipe_fence_handle **dst,
struct pipe_fence_handle *src)
{
pb_reference((struct pb_buffer**)dst, (struct pb_buffer*)src);
}
void radeon_drm_cs_init_functions(struct radeon_drm_winsys *ws)
{
ws->base.cs_create = radeon_drm_cs_create;
ws->base.cs_destroy = radeon_drm_cs_destroy;
ws->base.cs_add_reloc = radeon_drm_cs_add_reloc;
ws->base.cs_validate = radeon_drm_cs_validate;
ws->base.cs_memory_below_limit = radeon_drm_cs_memory_below_limit;
ws->base.cs_write_reloc = radeon_drm_cs_write_reloc;
ws->base.cs_flush = radeon_drm_cs_flush;
ws->base.cs_set_flush_callback = radeon_drm_cs_set_flush;
ws->base.cs_is_buffer_referenced = radeon_bo_is_referenced;
ws->base.cs_sync_flush = radeon_drm_cs_sync_flush;
ws->base.cs_create_fence = radeon_cs_create_fence;
ws->base.fence_wait = radeon_fence_wait;
ws->base.fence_reference = radeon_fence_reference;
}
|