summaryrefslogtreecommitdiff
path: root/libdrm/intel/intel_bufmgr_fake.c
blob: 4c28146739850b53cc58cffd5b499437e45b4ca4 (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
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
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
/**************************************************************************
 * 
 * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas.
 * 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 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 NON-INFRINGEMENT.
 * IN NO EVENT SHALL TUNGSTEN GRAPHICS 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.
 * 
 **************************************************************************/

/* Originally a fake version of the buffer manager so that we can
 * prototype the changes in a driver fairly quickly, has been fleshed
 * out to a fully functional interim solution.
 *
 * Basically wraps the old style memory management in the new
 * programming interface, but is more expressive and avoids many of
 * the bugs in the old texture manager.
 */

#ifdef HAVE_CONFIG_H
#include "config.h"
#endif

#include <stdlib.h>
#include <string.h>
#include <assert.h>
#include <errno.h>
#include <xf86drm.h>
#include <pthread.h>
#include "intel_bufmgr.h"
#include "intel_bufmgr_priv.h"
#include "drm.h"
#include "i915_drm.h"
#include "mm.h"
#include "libdrm_lists.h"

#define ALIGN(value, alignment)  ((value + alignment - 1) & ~(alignment - 1))

#define DBG(...) do {					\
   if (bufmgr_fake->bufmgr.debug)			\
      drmMsg(__VA_ARGS__);				\
} while (0)

/* Internal flags:
 */
#define BM_NO_BACKING_STORE			0x00000001
#define BM_NO_FENCE_SUBDATA			0x00000002
#define BM_PINNED				0x00000004

/* Wrapper around mm.c's mem_block, which understands that you must
 * wait for fences to expire before memory can be freed.  This is
 * specific to our use of memcpy for uploads - an upload that was
 * processed through the command queue wouldn't need to care about
 * fences.
 */
#define MAX_RELOCS 4096

struct fake_buffer_reloc
{
   /** Buffer object that the relocation points at. */
   dri_bo *target_buf;
   /** Offset of the relocation entry within reloc_buf. */
   uint32_t offset;
   /** Cached value of the offset when we last performed this relocation. */
   uint32_t last_target_offset;
   /** Value added to target_buf's offset to get the relocation entry. */
   uint32_t delta;
   /** Cache domains the target buffer is read into. */
   uint32_t read_domains;
   /** Cache domain the target buffer will have dirty cachelines in. */
   uint32_t write_domain;
};

struct block {
   struct block *next, *prev;
   struct mem_block *mem;	/* BM_MEM_AGP */

   /**
    * Marks that the block is currently in the aperture and has yet to be
    * fenced.
    */
   unsigned on_hardware:1;
   /**
    * Marks that the block is currently fenced (being used by rendering) and
    * can't be freed until @fence is passed.
    */
   unsigned fenced:1;

   /** Fence cookie for the block. */
   unsigned fence; /* Split to read_fence, write_fence */

   dri_bo *bo;
   void *virtual;
};

typedef struct _bufmgr_fake {
   dri_bufmgr bufmgr;

   pthread_mutex_t lock;

   unsigned long low_offset;
   unsigned long size;
   void *virtual;

   struct mem_block *heap;

   unsigned buf_nr;		/* for generating ids */

   /**
    * List of blocks which are currently in the GART but haven't been
    * fenced yet.
    */
   struct block on_hardware;
   /**
    * List of blocks which are in the GART and have an active fence on them.
    */
   struct block fenced;
   /**
    * List of blocks which have an expired fence and are ready to be evicted.
    */
   struct block lru;

   unsigned int last_fence;

   unsigned fail:1;
   unsigned need_fence:1;
   int thrashing;

   /**
    * Driver callback to emit a fence, returning the cookie.
    *
    * This allows the driver to hook in a replacement for the DRM usage in
    * bufmgr_fake.
    *
    * Currently, this also requires that a write flush be emitted before
    * emitting the fence, but this should change.
    */
   unsigned int (*fence_emit)(void *private);
   /** Driver callback to wait for a fence cookie to have passed. */
   void (*fence_wait)(unsigned int fence, void *private);
   void *fence_priv;

   /**
    * Driver callback to execute a buffer.
    *
    * This allows the driver to hook in a replacement for the DRM usage in
    * bufmgr_fake.
    */
   int (*exec)(dri_bo *bo, unsigned int used, void *priv);
   void *exec_priv;

   /** Driver-supplied argument to driver callbacks */
   void *driver_priv;
   /* Pointer to kernel-updated sarea data for the last completed user irq */
   volatile int *last_dispatch;

   int fd;

   int debug;

   int performed_rendering;
} dri_bufmgr_fake;

typedef struct _dri_bo_fake {
   dri_bo bo;

   unsigned id;			/* debug only */
   const char *name;

   unsigned dirty:1;
   /** has the card written to this buffer - we make need to copy it back */
   unsigned card_dirty:1;
   unsigned int refcount;
   /* Flags may consist of any of the DRM_BO flags, plus
    * DRM_BO_NO_BACKING_STORE and BM_NO_FENCE_SUBDATA, which are the first two
    * driver private flags.
    */
   uint64_t flags;
   /** Cache domains the target buffer is read into. */
   uint32_t read_domains;
   /** Cache domain the target buffer will have dirty cachelines in. */
   uint32_t write_domain;

   unsigned int alignment;
   int is_static, validated;
   unsigned int map_count;

   /** relocation list */
   struct fake_buffer_reloc *relocs;
   int nr_relocs;
   /**
    * Total size of the target_bos of this buffer.
    *
    * Used for estimation in check_aperture.
    */
   unsigned int child_size;

   struct block *block;
   void *backing_store;
   void (*invalidate_cb)(dri_bo *bo, void *ptr);
   void *invalidate_ptr;
} dri_bo_fake;

static int clear_fenced(dri_bufmgr_fake *bufmgr_fake,
			unsigned int fence_cookie);

#define MAXFENCE 0x7fffffff

static int FENCE_LTE( unsigned a, unsigned b )
{
   if (a == b)
      return 1;

   if (a < b && b - a < (1<<24))
      return 1;

   if (a > b && MAXFENCE - a + b < (1<<24))
      return 1;

   return 0;
}

void intel_bufmgr_fake_set_fence_callback(dri_bufmgr *bufmgr,
					  unsigned int (*emit)(void *priv),
					  void (*wait)(unsigned int fence,
						       void *priv),
					  void *priv)
{
   dri_bufmgr_fake *bufmgr_fake = (dri_bufmgr_fake *)bufmgr;

   bufmgr_fake->fence_emit = emit;
   bufmgr_fake->fence_wait = wait;
   bufmgr_fake->fence_priv = priv;
}

static unsigned int
_fence_emit_internal(dri_bufmgr_fake *bufmgr_fake)
{
   struct drm_i915_irq_emit ie;
   int ret, seq = 1;

   if (bufmgr_fake->fence_emit != NULL)
      return bufmgr_fake->fence_emit(bufmgr_fake->fence_priv);

   ie.irq_seq = &seq;
   ret = drmCommandWriteRead(bufmgr_fake->fd, DRM_I915_IRQ_EMIT,
			     &ie, sizeof(ie));
   if (ret) {
      drmMsg("%s: drm_i915_irq_emit: %d\n", __FUNCTION__, ret);
      abort();
   }

   DBG("emit 0x%08x\n", seq);
   bufmgr_fake->last_fence = seq;
   return bufmgr_fake->last_fence;
}

static void
_fence_wait_internal(dri_bufmgr_fake *bufmgr_fake, int seq)
{
   struct drm_i915_irq_wait iw;
   int hw_seq, busy_count = 0;
   int ret;
   int kernel_lied;

   if (bufmgr_fake->fence_wait != NULL) {
      bufmgr_fake->fence_wait(seq, bufmgr_fake->fence_priv);
      return;
   }

   DBG("wait 0x%08x\n", iw.irq_seq);

   iw.irq_seq = seq;

   /* The kernel IRQ_WAIT implementation is all sorts of broken.
    * 1) It returns 1 to 0x7fffffff instead of using the full 32-bit unsigned
    *    range.
    * 2) It returns 0 if hw_seq >= seq, not seq - hw_seq < 0 on the 32-bit
    *    signed range.
    * 3) It waits if seq < hw_seq, not seq - hw_seq > 0 on the 32-bit
    *    signed range.
    * 4) It returns -EBUSY in 3 seconds even if the hardware is still
    *    successfully chewing through buffers.
    *
    * Assume that in userland we treat sequence numbers as ints, which makes
    * some of the comparisons convenient, since the sequence numbers are
    * all postive signed integers.
    *
    * From this we get several cases we need to handle.  Here's a timeline.
    * 0x2   0x7                                         0x7ffffff8   0x7ffffffd
    *   |    |                                                   |    |
    * -------------------------------------------------------------------
    *
    * A) Normal wait for hw to catch up
    * hw_seq seq
    *   |    |
    * -------------------------------------------------------------------
    * seq - hw_seq = 5.  If we call IRQ_WAIT, it will wait for hw to catch up.
    *
    * B) Normal wait for a sequence number that's already passed.
    * seq    hw_seq
    *   |    |
    * -------------------------------------------------------------------
    * seq - hw_seq = -5.  If we call IRQ_WAIT, it returns 0 quickly.
    *
    * C) Hardware has already wrapped around ahead of us
    * hw_seq                                                         seq
    *   |                                                             |
    * -------------------------------------------------------------------
    * seq - hw_seq = 0x80000000 - 5.  If we called IRQ_WAIT, it would wait
    * for hw_seq >= seq, which may never occur.  Thus, we want to catch this
    * in userland and return 0.
    *
    * D) We've wrapped around ahead of the hardware.
    * seq                                                           hw_seq
    *   |                                                             |
    * -------------------------------------------------------------------
    * seq - hw_seq = -(0x80000000 - 5).  If we called IRQ_WAIT, it would return
    * 0 quickly because hw_seq >= seq, even though the hardware isn't caught up.
    * Thus, we need to catch this early return in userland and bother the
    * kernel until the hardware really does catch up.
    *
    * E) Hardware might wrap after we test in userland.
    *                                                         hw_seq  seq
    *                                                            |    |
    * -------------------------------------------------------------------
    * seq - hw_seq = 5.  If we call IRQ_WAIT, it will likely see seq >= hw_seq
    * and wait.  However, suppose hw_seq wraps before we make it into the
    * kernel.  The kernel sees hw_seq >= seq and waits for 3 seconds then
    * returns -EBUSY.  This is case C).  We should catch this and then return
    * successfully.
    *
    * F) Hardware might take a long time on a buffer.
    * hw_seq seq
    *   |    |
    * -------------------------------------------------------------------
    * seq - hw_seq = 5.  If we call IRQ_WAIT, if sequence 2 through 5 take too
    * long, it will return -EBUSY.  Batchbuffers in the gltestperf demo were
    * seen to take up to 7 seconds.  We should catch early -EBUSY return
    * and keep trying.
    */

   do {
      /* Keep a copy of last_dispatch so that if the wait -EBUSYs because the
       * hardware didn't catch up in 3 seconds, we can see if it at least made
       * progress and retry.
       */
      hw_seq = *bufmgr_fake->last_dispatch;

      /* Catch case C */
      if (seq - hw_seq > 0x40000000)
	 return;

      ret = drmCommandWrite(bufmgr_fake->fd, DRM_I915_IRQ_WAIT,
			    &iw, sizeof(iw));
      /* Catch case D */
      kernel_lied = (ret == 0) && (seq - *bufmgr_fake->last_dispatch <
				   -0x40000000);

      /* Catch case E */
      if (ret == -EBUSY && (seq - *bufmgr_fake->last_dispatch > 0x40000000))
	 ret = 0;

      /* Catch case F: Allow up to 15 seconds chewing on one buffer. */
      if ((ret == -EBUSY) && (hw_seq != *bufmgr_fake->last_dispatch))
	 busy_count = 0;
      else
	 busy_count++;
   } while (kernel_lied || ret == -EAGAIN || ret == -EINTR ||
	    (ret == -EBUSY && busy_count < 5));

   if (ret != 0) {
      drmMsg("%s:%d: Error waiting for fence: %s.\n", __FILE__, __LINE__,
	     strerror(-ret));
      abort();
   }
   clear_fenced(bufmgr_fake, seq);
}

static int
_fence_test(dri_bufmgr_fake *bufmgr_fake, unsigned fence)
{
   /* Slight problem with wrap-around:
    */
   return fence == 0 || FENCE_LTE(fence, bufmgr_fake->last_fence);
}

/**
 * Allocate a memory manager block for the buffer.
 */
static int
alloc_block(dri_bo *bo)
{
   dri_bo_fake *bo_fake = (dri_bo_fake *)bo;
   dri_bufmgr_fake *bufmgr_fake= (dri_bufmgr_fake *)bo->bufmgr;
   struct block *block = (struct block *)calloc(sizeof *block, 1);
   unsigned int align_log2 = ffs(bo_fake->alignment) - 1;
   unsigned int sz;

   if (!block)
      return 1;

   sz = (bo->size + bo_fake->alignment - 1) & ~(bo_fake->alignment - 1);

   block->mem = mmAllocMem(bufmgr_fake->heap, sz, align_log2, 0);
   if (!block->mem) {
      free(block);
      return 0;
   }

   DRMINITLISTHEAD(block);

   /* Insert at head or at tail???   
    */
   DRMLISTADDTAIL(block, &bufmgr_fake->lru);

   block->virtual = (uint8_t *)bufmgr_fake->virtual +
      block->mem->ofs - bufmgr_fake->low_offset;
   block->bo = bo;

   bo_fake->block = block;

   return 1;
}

/* Release the card storage associated with buf:
 */
static void free_block(dri_bufmgr_fake *bufmgr_fake, struct block *block)
{
   dri_bo_fake *bo_fake;
   DBG("free block %p %08x %d %d\n", block, block->mem->ofs, block->on_hardware, block->fenced);

   if (!block)
      return;

   bo_fake = (dri_bo_fake *)block->bo;
   if (!(bo_fake->flags & BM_NO_BACKING_STORE) && (bo_fake->card_dirty == 1)) {
     memcpy(bo_fake->backing_store, block->virtual, block->bo->size);
     bo_fake->card_dirty = 1;
     bo_fake->dirty = 1;
   }

   if (block->on_hardware) {
      block->bo = NULL;
   }
   else if (block->fenced) {
      block->bo = NULL;
   }
   else {
      DBG("    - free immediately\n");
      DRMLISTDEL(block);

      mmFreeMem(block->mem);
      free(block);
   }
}

static void
alloc_backing_store(dri_bo *bo)
{
   dri_bufmgr_fake *bufmgr_fake = (dri_bufmgr_fake *)bo->bufmgr;
   dri_bo_fake *bo_fake = (dri_bo_fake *)bo;
   assert(!bo_fake->backing_store);
   assert(!(bo_fake->flags & (BM_PINNED|BM_NO_BACKING_STORE)));

   bo_fake->backing_store = malloc(bo->size);

   DBG("alloc_backing - buf %d %p %d\n", bo_fake->id, bo_fake->backing_store, bo->size);
   assert(bo_fake->backing_store);
}

static void
free_backing_store(dri_bo *bo)
{
   dri_bo_fake *bo_fake = (dri_bo_fake *)bo;

   if (bo_fake->backing_store) {
      assert(!(bo_fake->flags & (BM_PINNED|BM_NO_BACKING_STORE)));
      free(bo_fake->backing_store);
      bo_fake->backing_store = NULL;
   }
}

static void
set_dirty(dri_bo *bo)
{
   dri_bufmgr_fake *bufmgr_fake = (dri_bufmgr_fake *)bo->bufmgr;
   dri_bo_fake *bo_fake = (dri_bo_fake *)bo;

   if (bo_fake->flags & BM_NO_BACKING_STORE && bo_fake->invalidate_cb != NULL)
      bo_fake->invalidate_cb(bo, bo_fake->invalidate_ptr);

   assert(!(bo_fake->flags & BM_PINNED));

   DBG("set_dirty - buf %d\n", bo_fake->id);
   bo_fake->dirty = 1;
}

static int
evict_lru(dri_bufmgr_fake *bufmgr_fake, unsigned int max_fence)
{
   struct block *block, *tmp;

   DBG("%s\n", __FUNCTION__);

   DRMLISTFOREACHSAFE(block, tmp, &bufmgr_fake->lru) {
      dri_bo_fake *bo_fake = (dri_bo_fake *)block->bo;

      if (bo_fake != NULL && (bo_fake->flags & BM_NO_FENCE_SUBDATA))
	 continue;

      if (block->fence && max_fence && !FENCE_LTE(block->fence, max_fence))
	 return 0;

      set_dirty(&bo_fake->bo);
      bo_fake->block = NULL;

      free_block(bufmgr_fake, block);
      return 1;
   }

   return 0;
}

static int
evict_mru(dri_bufmgr_fake *bufmgr_fake)
{
   struct block *block, *tmp;

   DBG("%s\n", __FUNCTION__);

   DRMLISTFOREACHSAFEREVERSE(block, tmp, &bufmgr_fake->lru) {
      dri_bo_fake *bo_fake = (dri_bo_fake *)block->bo;

      if (bo_fake && (bo_fake->flags & BM_NO_FENCE_SUBDATA))
	 continue;

      set_dirty(&bo_fake->bo);
      bo_fake->block = NULL;

      free_block(bufmgr_fake, block);
      return 1;
   }

   return 0;
}

/**
 * Removes all objects from the fenced list older than the given fence.
 */
static int clear_fenced(dri_bufmgr_fake *bufmgr_fake,
			unsigned int fence_cookie)
{
   struct block *block, *tmp;
   int ret = 0;

   DRMLISTFOREACHSAFE(block, tmp, &bufmgr_fake->fenced) {
      assert(block->fenced);

      if (_fence_test(bufmgr_fake, block->fence)) {

	 block->fenced = 0;

	 if (!block->bo) {
	    DBG("delayed free: offset %x sz %x\n",
		block->mem->ofs, block->mem->size);
	    DRMLISTDEL(block);
	    mmFreeMem(block->mem);
	    free(block);
	 }
	 else {
	    DBG("return to lru: offset %x sz %x\n",
		block->mem->ofs, block->mem->size);
	    DRMLISTDEL(block);
	    DRMLISTADDTAIL(block, &bufmgr_fake->lru);
	 }

	 ret = 1;
      }
      else {
	 /* Blocks are ordered by fence, so if one fails, all from
	  * here will fail also:
	  */
	DBG("fence not passed: offset %x sz %x %d %d \n",
	    block->mem->ofs, block->mem->size, block->fence, bufmgr_fake->last_fence);
	 break;
      }
   }

   DBG("%s: %d\n", __FUNCTION__, ret);
   return ret;
}

static void fence_blocks(dri_bufmgr_fake *bufmgr_fake, unsigned fence)
{
   struct block *block, *tmp;

   DRMLISTFOREACHSAFE(block, tmp, &bufmgr_fake->on_hardware) {
      DBG("Fence block %p (sz 0x%x ofs %x buf %p) with fence %d\n", block,
	  block->mem->size, block->mem->ofs, block->bo, fence);
      block->fence = fence;

      block->on_hardware = 0;
      block->fenced = 1;

      /* Move to tail of pending list here
       */
      DRMLISTDEL(block);
      DRMLISTADDTAIL(block, &bufmgr_fake->fenced);
   }

   assert(DRMLISTEMPTY(&bufmgr_fake->on_hardware));
}

static int evict_and_alloc_block(dri_bo *bo)
{
   dri_bufmgr_fake *bufmgr_fake = (dri_bufmgr_fake *)bo->bufmgr;
   dri_bo_fake *bo_fake = (dri_bo_fake *)bo;

   assert(bo_fake->block == NULL);

   /* Search for already free memory:
    */
   if (alloc_block(bo))
      return 1;

   /* If we're not thrashing, allow lru eviction to dig deeper into
    * recently used textures.  We'll probably be thrashing soon:
    */
   if (!bufmgr_fake->thrashing) {
      while (evict_lru(bufmgr_fake, 0))
	 if (alloc_block(bo))
	    return 1;
   }

   /* Keep thrashing counter alive?
    */
   if (bufmgr_fake->thrashing)
      bufmgr_fake->thrashing = 20;

   /* Wait on any already pending fences - here we are waiting for any
    * freed memory that has been submitted to hardware and fenced to
    * become available:
    */
   while (!DRMLISTEMPTY(&bufmgr_fake->fenced)) {
      uint32_t fence = bufmgr_fake->fenced.next->fence;
      _fence_wait_internal(bufmgr_fake, fence);

      if (alloc_block(bo))
	 return 1;
   }

   if (!DRMLISTEMPTY(&bufmgr_fake->on_hardware)) {
      while (!DRMLISTEMPTY(&bufmgr_fake->fenced)) {
	 uint32_t fence = bufmgr_fake->fenced.next->fence;
	 _fence_wait_internal(bufmgr_fake, fence);
      }

      if (!bufmgr_fake->thrashing) {
	 DBG("thrashing\n");
      }
      bufmgr_fake->thrashing = 20;

      if (alloc_block(bo))
	 return 1;
   }

   while (evict_mru(bufmgr_fake))
      if (alloc_block(bo))
	 return 1;

   DBG("%s 0x%x bytes failed\n", __FUNCTION__, bo->size);

   return 0;
}

/***********************************************************************
 * Public functions
 */

/**
 * Wait for hardware idle by emitting a fence and waiting for it.
 */
static void
dri_bufmgr_fake_wait_idle(dri_bufmgr_fake *bufmgr_fake)
{
   unsigned int cookie;

   cookie = _fence_emit_internal(bufmgr_fake);
   _fence_wait_internal(bufmgr_fake, cookie);
}

/**
 * Wait for rendering to a buffer to complete.
 *
 * It is assumed that the bathcbuffer which performed the rendering included
 * the necessary flushing.
 */
static void
dri_fake_bo_wait_rendering(dri_bo *bo)
{
   dri_bufmgr_fake *bufmgr_fake = (dri_bufmgr_fake *)bo->bufmgr;
   dri_bo_fake *bo_fake = (dri_bo_fake *)bo;

   pthread_mutex_lock(&bufmgr_fake->lock);

   if (bo_fake->block == NULL || !bo_fake->block->fenced) {
      pthread_mutex_unlock(&bufmgr_fake->lock);
      return;
   }

   _fence_wait_internal(bufmgr_fake, bo_fake->block->fence);

   pthread_mutex_unlock(&bufmgr_fake->lock);
}

/* Specifically ignore texture memory sharing.
 *  -- just evict everything
 *  -- and wait for idle
 */
void
intel_bufmgr_fake_contended_lock_take(dri_bufmgr *bufmgr)
{
   dri_bufmgr_fake *bufmgr_fake = (dri_bufmgr_fake *)bufmgr;
   struct block *block, *tmp;

   pthread_mutex_lock(&bufmgr_fake->lock);

   bufmgr_fake->need_fence = 1;
   bufmgr_fake->fail = 0;

   /* Wait for hardware idle.  We don't know where acceleration has been
    * happening, so we'll need to wait anyway before letting anything get
    * put on the card again.
    */
   dri_bufmgr_fake_wait_idle(bufmgr_fake);

   /* Check that we hadn't released the lock without having fenced the last
    * set of buffers.
    */
   assert(DRMLISTEMPTY(&bufmgr_fake->fenced));
   assert(DRMLISTEMPTY(&bufmgr_fake->on_hardware));

   DRMLISTFOREACHSAFE(block, tmp, &bufmgr_fake->lru) {
      assert(_fence_test(bufmgr_fake, block->fence));
      set_dirty(block->bo);
   }

   pthread_mutex_unlock(&bufmgr_fake->lock);
}

static dri_bo *
dri_fake_bo_alloc(dri_bufmgr *bufmgr, const char *name,
		  unsigned long size, unsigned int alignment)
{
   dri_bufmgr_fake *bufmgr_fake;
   dri_bo_fake *bo_fake;

   bufmgr_fake = (dri_bufmgr_fake *)bufmgr;

   assert(size != 0);

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

   bo_fake->bo.size = size;
   bo_fake->bo.offset = -1;
   bo_fake->bo.virtual = NULL;
   bo_fake->bo.bufmgr = bufmgr;
   bo_fake->refcount = 1;

   /* Alignment must be a power of two */
   assert((alignment & (alignment - 1)) == 0);
   if (alignment == 0)
      alignment = 1;
   bo_fake->alignment = alignment;
   bo_fake->id = ++bufmgr_fake->buf_nr;
   bo_fake->name = name;
   bo_fake->flags = 0;
   bo_fake->is_static = 0;

   DBG("drm_bo_alloc: (buf %d: %s, %d kb)\n", bo_fake->id, bo_fake->name,
       bo_fake->bo.size / 1024);

   return &bo_fake->bo;
}

dri_bo *
intel_bo_fake_alloc_static(dri_bufmgr *bufmgr, const char *name,
			   unsigned long offset, unsigned long size,
			   void *virtual)
{
   dri_bufmgr_fake *bufmgr_fake;
   dri_bo_fake *bo_fake;

   bufmgr_fake = (dri_bufmgr_fake *)bufmgr;

   assert(size != 0);

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

   bo_fake->bo.size = size;
   bo_fake->bo.offset = offset;
   bo_fake->bo.virtual = virtual;
   bo_fake->bo.bufmgr = bufmgr;
   bo_fake->refcount = 1;
   bo_fake->id = ++bufmgr_fake->buf_nr;
   bo_fake->name = name;
   bo_fake->flags = BM_PINNED;
   bo_fake->is_static = 1;

   DBG("drm_bo_alloc_static: (buf %d: %s, %d kb)\n", bo_fake->id, bo_fake->name,
       bo_fake->bo.size / 1024);

   return &bo_fake->bo;
}

static void
dri_fake_bo_reference(dri_bo *bo)
{
   dri_bufmgr_fake *bufmgr_fake = (dri_bufmgr_fake *)bo->bufmgr;
   dri_bo_fake *bo_fake = (dri_bo_fake *)bo;

   pthread_mutex_lock(&bufmgr_fake->lock);
   bo_fake->refcount++;
   pthread_mutex_unlock(&bufmgr_fake->lock);
}

static void
dri_fake_bo_reference_locked(dri_bo *bo)
{
   dri_bo_fake *bo_fake = (dri_bo_fake *)bo;

   bo_fake->refcount++;
}

static void
dri_fake_bo_unreference_locked(dri_bo *bo)
{
   dri_bufmgr_fake *bufmgr_fake = (dri_bufmgr_fake *)bo->bufmgr;
   dri_bo_fake *bo_fake = (dri_bo_fake *)bo;
   int i;

   if (--bo_fake->refcount == 0) {
      assert(bo_fake->map_count == 0);
      /* No remaining references, so free it */
      if (bo_fake->block)
	 free_block(bufmgr_fake, bo_fake->block);
      free_backing_store(bo);

      for (i = 0; i < bo_fake->nr_relocs; i++)
	 dri_fake_bo_unreference_locked(bo_fake->relocs[i].target_buf);

      DBG("drm_bo_unreference: free buf %d %s\n", bo_fake->id, bo_fake->name);

      free(bo_fake->relocs);
      free(bo);
   }
}

static void
dri_fake_bo_unreference(dri_bo *bo)
{
   dri_bufmgr_fake *bufmgr_fake = (dri_bufmgr_fake *)bo->bufmgr;

   pthread_mutex_lock(&bufmgr_fake->lock);
   dri_fake_bo_unreference_locked(bo);
   pthread_mutex_unlock(&bufmgr_fake->lock);
}

/**
 * Set the buffer as not requiring backing store, and instead get the callback
 * invoked whenever it would be set dirty.
 */
void intel_bo_fake_disable_backing_store(dri_bo *bo,
					 void (*invalidate_cb)(dri_bo *bo,
							       void *ptr),
					 void *ptr)
{
   dri_bufmgr_fake *bufmgr_fake = (dri_bufmgr_fake *)bo->bufmgr;
   dri_bo_fake *bo_fake = (dri_bo_fake *)bo;

   pthread_mutex_lock(&bufmgr_fake->lock);

   if (bo_fake->backing_store)
      free_backing_store(bo);

   bo_fake->flags |= BM_NO_BACKING_STORE;

   DBG("disable_backing_store set buf %d dirty\n", bo_fake->id);
   bo_fake->dirty = 1;
   bo_fake->invalidate_cb = invalidate_cb;
   bo_fake->invalidate_ptr = ptr;

   /* Note that it is invalid right from the start.  Also note
    * invalidate_cb is called with the bufmgr locked, so cannot
    * itself make bufmgr calls.
    */
   if (invalidate_cb != NULL)
      invalidate_cb(bo, ptr);

   pthread_mutex_unlock(&bufmgr_fake->lock);
}

/**
 * Map a buffer into bo->virtual, allocating either card memory space (If
 * BM_NO_BACKING_STORE or BM_PINNED) or backing store, as necessary.
 */
static int
dri_fake_bo_map_locked(dri_bo *bo, int write_enable)
{
   dri_bufmgr_fake *bufmgr_fake = (dri_bufmgr_fake *)bo->bufmgr;
   dri_bo_fake *bo_fake = (dri_bo_fake *)bo;

   /* Static buffers are always mapped. */
   if (bo_fake->is_static)
      return 0;

   /* Allow recursive mapping.  Mesa may recursively map buffers with
    * nested display loops, and it is used internally in bufmgr_fake
    * for relocation.
    */
   if (bo_fake->map_count++ != 0)
      return 0;

   {
      DBG("drm_bo_map: (buf %d: %s, %d kb)\n", bo_fake->id, bo_fake->name,
	  bo_fake->bo.size / 1024);

      if (bo->virtual != NULL) {
	 drmMsg("%s: already mapped\n", __FUNCTION__);
	 abort();
      }
      else if (bo_fake->flags & (BM_NO_BACKING_STORE|BM_PINNED)) {

	 if (!bo_fake->block && !evict_and_alloc_block(bo)) {
	    DBG("%s: alloc failed\n", __FUNCTION__);
	    bufmgr_fake->fail = 1;
	    return 1;
	 }
	 else {
	    assert(bo_fake->block);
	    bo_fake->dirty = 0;

	    if (!(bo_fake->flags & BM_NO_FENCE_SUBDATA) &&
		bo_fake->block->fenced) {
	       dri_fake_bo_wait_rendering(bo);
	    }

	    bo->virtual = bo_fake->block->virtual;
	 }
      }
      else {
	 if (write_enable)
	    set_dirty(bo);

	 if (bo_fake->backing_store == 0)
	    alloc_backing_store(bo);

	 bo->virtual = bo_fake->backing_store;
      }
   }

   return 0;
}

static int
dri_fake_bo_map(dri_bo *bo, int write_enable)
{
   dri_bufmgr_fake *bufmgr_fake = (dri_bufmgr_fake *)bo->bufmgr;
   int ret;

   pthread_mutex_lock(&bufmgr_fake->lock);
   ret = dri_fake_bo_map_locked(bo, write_enable);
   pthread_mutex_unlock(&bufmgr_fake->lock);

   return ret;
}

static int
dri_fake_bo_unmap_locked(dri_bo *bo)
{
   dri_bufmgr_fake *bufmgr_fake = (dri_bufmgr_fake *)bo->bufmgr;
   dri_bo_fake *bo_fake = (dri_bo_fake *)bo;

   /* Static buffers are always mapped. */
   if (bo_fake->is_static)
      return 0;

   assert(bo_fake->map_count != 0);
   if (--bo_fake->map_count != 0)
      return 0;

   DBG("drm_bo_unmap: (buf %d: %s, %d kb)\n", bo_fake->id, bo_fake->name,
       bo_fake->bo.size / 1024);

   bo->virtual = NULL;

   return 0;
}

static int
dri_fake_bo_unmap(dri_bo *bo)
{
   dri_bufmgr_fake *bufmgr_fake = (dri_bufmgr_fake *)bo->bufmgr;
   int ret;

   pthread_mutex_lock(&bufmgr_fake->lock);
   ret = dri_fake_bo_unmap_locked(bo);
   pthread_mutex_unlock(&bufmgr_fake->lock);

   return ret;
}

static void
dri_fake_kick_all(dri_bufmgr_fake *bufmgr_fake)
{
   struct block *block, *tmp;

   pthread_mutex_lock(&bufmgr_fake->lock);

   bufmgr_fake->performed_rendering = 0;
   /* okay for ever BO that is on the HW kick it off.
      seriously not afraid of the POLICE right now */
   DRMLISTFOREACHSAFE(block, tmp, &bufmgr_fake->on_hardware) {
      dri_bo_fake *bo_fake = (dri_bo_fake *)block->bo;

      block->on_hardware = 0;
      free_block(bufmgr_fake, block);
      bo_fake->block = NULL;
      bo_fake->validated = 0;
      if (!(bo_fake->flags & BM_NO_BACKING_STORE))
         bo_fake->dirty = 1;
   }

   pthread_mutex_unlock(&bufmgr_fake->lock);
}

static int
dri_fake_bo_validate(dri_bo *bo)
{
   dri_bufmgr_fake *bufmgr_fake;
   dri_bo_fake *bo_fake = (dri_bo_fake *)bo;

   bufmgr_fake = (dri_bufmgr_fake *)bo->bufmgr;

   DBG("drm_bo_validate: (buf %d: %s, %d kb)\n", bo_fake->id, bo_fake->name,
       bo_fake->bo.size / 1024);

   /* Sanity check: Buffers should be unmapped before being validated.
    * This is not so much of a problem for bufmgr_fake, but TTM refuses,
    * and the problem is harder to debug there.
    */
   assert(bo_fake->map_count == 0);

   if (bo_fake->is_static) {
      /* Add it to the needs-fence list */
      bufmgr_fake->need_fence = 1;
      return 0;
   }

   /* Allocate the card memory */
   if (!bo_fake->block && !evict_and_alloc_block(bo)) {
      bufmgr_fake->fail = 1;
      DBG("Failed to validate buf %d:%s\n", bo_fake->id, bo_fake->name);
      return -1;
   }

   assert(bo_fake->block);
   assert(bo_fake->block->bo == &bo_fake->bo);

   bo->offset = bo_fake->block->mem->ofs;

   /* Upload the buffer contents if necessary */
   if (bo_fake->dirty) {
      DBG("Upload dirty buf %d:%s, sz %d offset 0x%x\n", bo_fake->id,
	  bo_fake->name, bo->size, bo_fake->block->mem->ofs);

      assert(!(bo_fake->flags &
	       (BM_NO_BACKING_STORE|BM_PINNED)));

      /* Actually, should be able to just wait for a fence on the memory,
       * which we would be tracking when we free it.  Waiting for idle is
       * a sufficiently large hammer for now.
       */
      dri_bufmgr_fake_wait_idle(bufmgr_fake);

      /* we may never have mapped this BO so it might not have any backing
       * store if this happens it should be rare, but 0 the card memory
       * in any case */
      if (bo_fake->backing_store)
         memcpy(bo_fake->block->virtual, bo_fake->backing_store, bo->size);
      else
         memset(bo_fake->block->virtual, 0, bo->size);

      bo_fake->dirty = 0;
   }

   bo_fake->block->fenced = 0;
   bo_fake->block->on_hardware = 1;
   DRMLISTDEL(bo_fake->block);
   DRMLISTADDTAIL(bo_fake->block, &bufmgr_fake->on_hardware);

   bo_fake->validated = 1;
   bufmgr_fake->need_fence = 1;

   return 0;
}

static void
dri_fake_fence_validated(dri_bufmgr *bufmgr)
{
   dri_bufmgr_fake *bufmgr_fake = (dri_bufmgr_fake *)bufmgr;
   unsigned int cookie;

   cookie = _fence_emit_internal(bufmgr_fake);
   fence_blocks(bufmgr_fake, cookie);

   DBG("drm_fence_validated: 0x%08x cookie\n", cookie);
}

static void
dri_fake_destroy(dri_bufmgr *bufmgr)
{
   dri_bufmgr_fake *bufmgr_fake = (dri_bufmgr_fake *)bufmgr;

   pthread_mutex_destroy(&bufmgr_fake->lock);
   mmDestroy(bufmgr_fake->heap);
   free(bufmgr);
}

static int
dri_fake_emit_reloc(dri_bo *reloc_buf,
		    uint32_t read_domains, uint32_t write_domain,
		    uint32_t delta, uint32_t offset, dri_bo *target_buf)
{
   dri_bufmgr_fake *bufmgr_fake = (dri_bufmgr_fake *)reloc_buf->bufmgr;
   struct fake_buffer_reloc *r;
   dri_bo_fake *reloc_fake = (dri_bo_fake *)reloc_buf;
   dri_bo_fake *target_fake = (dri_bo_fake *)target_buf;
   int i;

   pthread_mutex_lock(&bufmgr_fake->lock);

   assert(reloc_buf);
   assert(target_buf);

   if (reloc_fake->relocs == NULL) {
      reloc_fake->relocs = malloc(sizeof(struct fake_buffer_reloc) *
				  MAX_RELOCS);
   }

   r = &reloc_fake->relocs[reloc_fake->nr_relocs++];

   assert(reloc_fake->nr_relocs <= MAX_RELOCS);

   dri_fake_bo_reference_locked(target_buf);

   if (!target_fake->is_static)
      reloc_fake->child_size += ALIGN(target_buf->size, target_fake->alignment);

   r->target_buf = target_buf;
   r->offset = offset;
   r->last_target_offset = target_buf->offset;
   r->delta = delta;
   r->read_domains = read_domains;
   r->write_domain = write_domain;

   if (bufmgr_fake->debug) {
      /* Check that a conflicting relocation hasn't already been emitted. */
      for (i = 0; i < reloc_fake->nr_relocs - 1; i++) {
	 struct fake_buffer_reloc *r2 = &reloc_fake->relocs[i];

	 assert(r->offset != r2->offset);
      }
   }

   pthread_mutex_unlock(&bufmgr_fake->lock);

   return 0;
}

/**
 * Incorporates the validation flags associated with each relocation into
 * the combined validation flags for the buffer on this batchbuffer submission.
 */
static void
dri_fake_calculate_domains(dri_bo *bo)
{
   dri_bo_fake *bo_fake = (dri_bo_fake *)bo;
   int i;

   for (i = 0; i < bo_fake->nr_relocs; i++) {
      struct fake_buffer_reloc *r = &bo_fake->relocs[i];
      dri_bo_fake *target_fake = (dri_bo_fake *)r->target_buf;

      /* Do the same for the tree of buffers we depend on */
      dri_fake_calculate_domains(r->target_buf);

      target_fake->read_domains |= r->read_domains;
      if (target_fake->write_domain != 0)
	 target_fake->write_domain = r->write_domain;
   }
}


static int
dri_fake_reloc_and_validate_buffer(dri_bo *bo)
{
   dri_bufmgr_fake *bufmgr_fake = (dri_bufmgr_fake *)bo->bufmgr;
   dri_bo_fake *bo_fake = (dri_bo_fake *)bo;
   int i, ret;

   assert(bo_fake->map_count == 0);

   for (i = 0; i < bo_fake->nr_relocs; i++) {
      struct fake_buffer_reloc *r = &bo_fake->relocs[i];
      dri_bo_fake *target_fake = (dri_bo_fake *)r->target_buf;
      uint32_t reloc_data;

      /* Validate the target buffer if that hasn't been done. */
      if (!target_fake->validated) {
         ret = dri_fake_reloc_and_validate_buffer(r->target_buf);
         if (ret != 0) {
            if (bo->virtual != NULL)
                dri_fake_bo_unmap_locked(bo);
            return ret;
         }
      }

      /* Calculate the value of the relocation entry. */
      if (r->target_buf->offset != r->last_target_offset) {
	 reloc_data = r->target_buf->offset + r->delta;

	 if (bo->virtual == NULL)
	    dri_fake_bo_map_locked(bo, 1);

	 *(uint32_t *)((uint8_t *)bo->virtual + r->offset) = reloc_data;

	 r->last_target_offset = r->target_buf->offset;
      }
   }

   if (bo->virtual != NULL)
      dri_fake_bo_unmap_locked(bo);

   if (bo_fake->write_domain != 0) {
      if (!(bo_fake->flags & (BM_NO_BACKING_STORE|BM_PINNED))) {
         if (bo_fake->backing_store == 0)
            alloc_backing_store(bo);

         bo_fake->card_dirty = 1;
      }
      bufmgr_fake->performed_rendering = 1;
   }

   return dri_fake_bo_validate(bo);
}

static void
dri_bo_fake_post_submit(dri_bo *bo)
{
   dri_bufmgr_fake *bufmgr_fake = (dri_bufmgr_fake *)bo->bufmgr;
   dri_bo_fake *bo_fake = (dri_bo_fake *)bo;
   int i;

   for (i = 0; i < bo_fake->nr_relocs; i++) {
      struct fake_buffer_reloc *r = &bo_fake->relocs[i];
      dri_bo_fake *target_fake = (dri_bo_fake *)r->target_buf;

      if (target_fake->validated)
	 dri_bo_fake_post_submit(r->target_buf);

      DBG("%s@0x%08x + 0x%08x -> %s@0x%08x + 0x%08x\n",
	  bo_fake->name, (uint32_t)bo->offset, r->offset,
	  target_fake->name, (uint32_t)r->target_buf->offset, r->delta);
   }

   assert(bo_fake->map_count == 0);
   bo_fake->validated = 0;
   bo_fake->read_domains = 0;
   bo_fake->write_domain = 0;
}


void intel_bufmgr_fake_set_exec_callback(dri_bufmgr *bufmgr,
					 int (*exec)(dri_bo *bo,
						     unsigned int used,
						     void *priv),
					 void *priv)
{
   dri_bufmgr_fake *bufmgr_fake = (dri_bufmgr_fake *)bufmgr;

   bufmgr_fake->exec = exec;
   bufmgr_fake->exec_priv = priv;
}

static int
dri_fake_bo_exec(dri_bo *bo, int used,
		 drm_clip_rect_t *cliprects, int num_cliprects,
		 int DR4)
{
   dri_bufmgr_fake *bufmgr_fake = (dri_bufmgr_fake *)bo->bufmgr;
   dri_bo_fake *batch_fake = (dri_bo_fake *)bo;
   struct drm_i915_batchbuffer batch;
   int ret;
   int retry_count = 0;

   pthread_mutex_lock(&bufmgr_fake->lock);

   bufmgr_fake->performed_rendering = 0;

   dri_fake_calculate_domains(bo);

   batch_fake->read_domains = I915_GEM_DOMAIN_COMMAND;

   /* we've ran out of RAM so blow the whole lot away and retry */
 restart:
   ret = dri_fake_reloc_and_validate_buffer(bo);
   if (bufmgr_fake->fail == 1) {
      if (retry_count == 0) {
         retry_count++;
         dri_fake_kick_all(bufmgr_fake);
         bufmgr_fake->fail = 0;
         goto restart;
      } else /* dump out the memory here */
         mmDumpMemInfo(bufmgr_fake->heap);
   }

   assert(ret == 0);

   if (bufmgr_fake->exec != NULL) {
      int ret = bufmgr_fake->exec(bo, used, bufmgr_fake->exec_priv);
      if (ret != 0)
	 return ret;
   } else {
      batch.start = bo->offset;
      batch.used = used;
      batch.cliprects = cliprects;
      batch.num_cliprects = num_cliprects;
      batch.DR1 = 0;
      batch.DR4 = DR4;

      if (drmCommandWrite(bufmgr_fake->fd, DRM_I915_BATCHBUFFER, &batch,
			  sizeof(batch))) {
	 drmMsg("DRM_I915_BATCHBUFFER: %d\n", -errno);
	 return -errno;
      }
   }

   dri_fake_fence_validated(bo->bufmgr);

   dri_bo_fake_post_submit(bo);

   pthread_mutex_unlock(&bufmgr_fake->lock);

   return 0;
}

/**
 * Return an error if the list of BOs will exceed the aperture size.
 *
 * This is a rough guess and likely to fail, as during the validate sequence we
 * may place a buffer in an inopportune spot early on and then fail to fit
 * a set smaller than the aperture.
 */
static int
dri_fake_check_aperture_space(dri_bo **bo_array, int count)
{
   dri_bufmgr_fake *bufmgr_fake = (dri_bufmgr_fake *)bo_array[0]->bufmgr;
   unsigned int sz = 0;
   int i;

   for (i = 0; i < count; i++) {
      dri_bo_fake *bo_fake = (dri_bo_fake *)bo_array[i];

      if (bo_fake == NULL)
	 continue;

      if (!bo_fake->is_static)
	 sz += ALIGN(bo_array[i]->size, bo_fake->alignment);
      sz += bo_fake->child_size;
   }

   if (sz > bufmgr_fake->size) {
      DBG("check_space: overflowed bufmgr size, %dkb vs %dkb\n",
	  sz / 1024, bufmgr_fake->size / 1024);
      return -1;
   }

   DBG("drm_check_space: sz %dkb vs bufgr %dkb\n", sz / 1024 ,
       bufmgr_fake->size / 1024);
   return 0;
}

/**
 * Evicts all buffers, waiting for fences to pass and copying contents out
 * as necessary.
 *
 * Used by the X Server on LeaveVT, when the card memory is no longer our
 * own.
 */
void
intel_bufmgr_fake_evict_all(dri_bufmgr *bufmgr)
{
   dri_bufmgr_fake *bufmgr_fake = (dri_bufmgr_fake *)bufmgr;
   struct block *block, *tmp;

   pthread_mutex_lock(&bufmgr_fake->lock);

   bufmgr_fake->need_fence = 1;
   bufmgr_fake->fail = 0;

   /* Wait for hardware idle.  We don't know where acceleration has been
    * happening, so we'll need to wait anyway before letting anything get
    * put on the card again.
    */
   dri_bufmgr_fake_wait_idle(bufmgr_fake);

   /* Check that we hadn't released the lock without having fenced the last
    * set of buffers.
    */
   assert(DRMLISTEMPTY(&bufmgr_fake->fenced));
   assert(DRMLISTEMPTY(&bufmgr_fake->on_hardware));

   DRMLISTFOREACHSAFE(block, tmp, &bufmgr_fake->lru) {
      /* Releases the memory, and memcpys dirty contents out if necessary. */
      free_block(bufmgr_fake, block);
   }

   pthread_mutex_unlock(&bufmgr_fake->lock);
}
void intel_bufmgr_fake_set_last_dispatch(dri_bufmgr *bufmgr,
					 volatile unsigned int *last_dispatch)
{
   dri_bufmgr_fake *bufmgr_fake = (dri_bufmgr_fake *)bufmgr;

   bufmgr_fake->last_dispatch = (volatile int *)last_dispatch;
}

dri_bufmgr *
intel_bufmgr_fake_init(int fd,
		       unsigned long low_offset, void *low_virtual,
		       unsigned long size,
		       volatile unsigned int *last_dispatch)
{
   dri_bufmgr_fake *bufmgr_fake;

   bufmgr_fake = calloc(1, sizeof(*bufmgr_fake));

   if (pthread_mutex_init(&bufmgr_fake->lock, NULL) != 0) {
      free(bufmgr_fake);
      return NULL;
   }

   /* Initialize allocator */
   DRMINITLISTHEAD(&bufmgr_fake->fenced);
   DRMINITLISTHEAD(&bufmgr_fake->on_hardware);
   DRMINITLISTHEAD(&bufmgr_fake->lru);

   bufmgr_fake->low_offset = low_offset;
   bufmgr_fake->virtual = low_virtual;
   bufmgr_fake->size = size;
   bufmgr_fake->heap = mmInit(low_offset, size);

   /* Hook in methods */
   bufmgr_fake->bufmgr.bo_alloc = dri_fake_bo_alloc;
   bufmgr_fake->bufmgr.bo_reference = dri_fake_bo_reference;
   bufmgr_fake->bufmgr.bo_unreference = dri_fake_bo_unreference;
   bufmgr_fake->bufmgr.bo_map = dri_fake_bo_map;
   bufmgr_fake->bufmgr.bo_unmap = dri_fake_bo_unmap;
   bufmgr_fake->bufmgr.bo_wait_rendering = dri_fake_bo_wait_rendering;
   bufmgr_fake->bufmgr.bo_emit_reloc = dri_fake_emit_reloc;
   bufmgr_fake->bufmgr.destroy = dri_fake_destroy;
   bufmgr_fake->bufmgr.bo_exec = dri_fake_bo_exec;
   bufmgr_fake->bufmgr.check_aperture_space = dri_fake_check_aperture_space;
   bufmgr_fake->bufmgr.debug = 0;

   bufmgr_fake->fd = fd;
   bufmgr_fake->last_dispatch = (volatile int *)last_dispatch;

   return &bufmgr_fake->bufmgr;
}