aboutsummaryrefslogtreecommitdiffstats
blob: f7ad75716ef46a12086a93f8ddf072fdd08fe0ad (plain) (blame)
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
/**************************************************************************
 *
 * Copyright © 2007 Red Hat Inc.
 * Copyright © 2007 Intel Corporation
 * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA
 * 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: Thomas Hellström <thomas-at-tungstengraphics-dot-com>
 *          Keith Whitwell <keithw-at-tungstengraphics-dot-com>
 *	    Eric Anholt <eric@anholt.net>
 *	    Dave Airlie <airlied@linux.ie>
 */

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

#include <xf86drm.h>
#include <fcntl.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <unistd.h>
#include <assert.h>
#include <pthread.h>
#include <sys/ioctl.h>
#include <sys/mman.h>
#include <sys/stat.h>
#include <sys/types.h>

#include "errno.h"
#include "intel_bufmgr.h"
#include "intel_bufmgr_priv.h"
#include "string.h"

#include "i915_drm.h"

#define DBG(...) do {					\
   if (bufmgr_gem->bufmgr.debug)			\
      fprintf(stderr, __VA_ARGS__);			\
} while (0)

typedef struct _drm_intel_bo_gem drm_intel_bo_gem;

struct drm_intel_gem_bo_bucket {
   drm_intel_bo_gem *head, **tail;
   /**
    * Limit on the number of entries in this bucket.
    *
    * 0 means that this caching at this bucket size is disabled.
    * -1 means that there is no limit to caching at this size.
    */
   int max_entries;
   int num_entries;
};

/* Arbitrarily chosen, 16 means that the maximum size we'll cache for reuse
 * is 1 << 16 pages, or 256MB.
 */
#define DRM_INTEL_GEM_BO_BUCKETS	16
typedef struct _drm_intel_bufmgr_gem {
    drm_intel_bufmgr bufmgr;

    int fd;

    int max_relocs;

    pthread_mutex_t lock;

    struct drm_i915_gem_exec_object *exec_objects;
    drm_intel_bo **exec_bos;
    int exec_size;
    int exec_count;

    /** Array of lists of cached gem objects of power-of-two sizes */
    struct drm_intel_gem_bo_bucket cache_bucket[DRM_INTEL_GEM_BO_BUCKETS];

    uint64_t gtt_size;
} drm_intel_bufmgr_gem;

struct _drm_intel_bo_gem {
    drm_intel_bo bo;

    int refcount;
    /** Boolean whether the mmap ioctl has been called for this buffer yet. */
    int mapped;
    uint32_t gem_handle;
    const char *name;

    /**
     * Kenel-assigned global name for this object
     */
    unsigned int global_name;
    
    /**
     * Index of the buffer within the validation list while preparing a
     * batchbuffer execution.
     */
    int validate_index;

    /**
     * Boolean whether we've started swrast
     * Set when the buffer has been mapped
     * Cleared when the buffer is unmapped
     */
    int swrast;

    /** Array passed to the DRM containing relocation information. */
    struct drm_i915_gem_relocation_entry *relocs;
    /** Array of bos corresponding to relocs[i].target_handle */
    drm_intel_bo **reloc_target_bo;
    /** Number of entries in relocs */
    int reloc_count;
    /** Mapped address for the buffer */
    void *virtual;

    /** free list */
    drm_intel_bo_gem *next;

    /**
     * Boolean of whether this BO and its children have been included in
     * the current drm_intel_bufmgr_check_aperture_space() total.
     */
    char included_in_check_aperture;

    /**
     * Boolean of whether this buffer has been used as a relocation
     * target and had its size accounted for, and thus can't have any
     * further relocations added to it.
     */
     char used_as_reloc_target;

    /**
     * Size in bytes of this buffer and its relocation descendents.
     *
     * Used to avoid costly tree walking in drm_intel_bufmgr_check_aperture in
     * the common case.
     */
    int reloc_tree_size;
};

static void drm_intel_gem_bo_reference_locked(drm_intel_bo *bo);

static int
logbase2(int n)
{
   int i = 1;
   int log2 = 0;

   while (n > i) {
      i *= 2;
      log2++;
   }

   return log2;
}

static struct drm_intel_gem_bo_bucket *
drm_intel_gem_bo_bucket_for_size(drm_intel_bufmgr_gem *bufmgr_gem,
				 unsigned long size)
{
    int i;

    /* We only do buckets in power of two increments */
    if ((size & (size - 1)) != 0)
	return NULL;

    /* We should only see sizes rounded to pages. */
    assert((size % 4096) == 0);

    /* We always allocate in units of pages */
    i = ffs(size / 4096) - 1;
    if (i >= DRM_INTEL_GEM_BO_BUCKETS)
	return NULL;

    return &bufmgr_gem->cache_bucket[i];
}


static void drm_intel_gem_dump_validation_list(drm_intel_bufmgr_gem *bufmgr_gem)
{
    int i, j;

    for (i = 0; i < bufmgr_gem->exec_count; i++) {
	drm_intel_bo *bo = bufmgr_gem->exec_bos[i];
	drm_intel_bo_gem *bo_gem = (drm_intel_bo_gem *)bo;

	if (bo_gem->relocs == NULL) {
	    DBG("%2d: %d (%s)\n", i, bo_gem->gem_handle, bo_gem->name);
	    continue;
	}

	for (j = 0; j < bo_gem->reloc_count; j++) {
	    drm_intel_bo *target_bo = bo_gem->reloc_target_bo[j];
	    drm_intel_bo_gem *target_gem = (drm_intel_bo_gem *)target_bo;

	    DBG("%2d: %d (%s)@0x%08llx -> %d (%s)@0x%08lx + 0x%08x\n",
		i,
		bo_gem->gem_handle, bo_gem->name,
		(unsigned long long)bo_gem->relocs[j].offset,
		target_gem->gem_handle, target_gem->name, target_bo->offset,
		bo_gem->relocs[j].delta);
	}
    }
}

/**
 * Adds the given buffer to the list of buffers to be validated (moved into the
 * appropriate memory type) with the next batch submission.
 *
 * If a buffer is validated multiple times in a batch submission, it ends up
 * with the intersection of the memory type flags and the union of the
 * access flags.
 */
static void
drm_intel_add_validate_buffer(drm_intel_bo *bo)
{
    drm_intel_bufmgr_gem *bufmgr_gem = (drm_intel_bufmgr_gem *)bo->bufmgr;
    drm_intel_bo_gem *bo_gem = (drm_intel_bo_gem *)bo;
    int index;

    if (bo_gem->validate_index != -1)
	return;

    /* Extend the array of validation entries as necessary. */
    if (bufmgr_gem->exec_count == bufmgr_gem->exec_size) {
	int new_size = bufmgr_gem->exec_size * 2;

	if (new_size == 0)
	    new_size = 5;

	bufmgr_gem->exec_objects =
	    realloc(bufmgr_gem->exec_objects,
		    sizeof(*bufmgr_gem->exec_objects) * new_size);
	bufmgr_gem->exec_bos =
	    realloc(bufmgr_gem->exec_bos,
		    sizeof(*bufmgr_gem->exec_bos) * new_size);
	bufmgr_gem->exec_size = new_size;
    }

    index = bufmgr_gem->exec_count;
    bo_gem->validate_index = index;
    /* Fill in array entry */
    bufmgr_gem->exec_objects[index].handle = bo_gem->gem_handle;
    bufmgr_gem->exec_objects[index].relocation_count = bo_gem->reloc_count;
    bufmgr_gem->exec_objects[index].relocs_ptr = (uintptr_t)bo_gem->relocs;
    bufmgr_gem->exec_objects[index].alignment = 0;
    bufmgr_gem->exec_objects[index].offset = 0;
    bufmgr_gem->exec_bos[index] = bo;
    drm_intel_gem_bo_reference_locked(bo);
    bufmgr_gem->exec_count++;
}


#define RELOC_BUF_SIZE(x) ((I915_RELOC_HEADER + x * I915_RELOC0_STRIDE) * \
	sizeof(uint32_t))

static int
drm_intel_setup_reloc_list(drm_intel_bo *bo)
{
    drm_intel_bo_gem *bo_gem = (drm_intel_bo_gem *)bo;
    drm_intel_bufmgr_gem *bufmgr_gem = (drm_intel_bufmgr_gem *)bo->bufmgr;

    bo_gem->relocs = malloc(bufmgr_gem->max_relocs *
			    sizeof(struct drm_i915_gem_relocation_entry));
    bo_gem->reloc_target_bo = malloc(bufmgr_gem->max_relocs *
				     sizeof(drm_intel_bo *));

    return 0;
}

static drm_intel_bo *
drm_intel_gem_bo_alloc(drm_intel_bufmgr *bufmgr, const char *name,
		   unsigned long size, unsigned int alignment)
{
    drm_intel_bufmgr_gem *bufmgr_gem = (drm_intel_bufmgr_gem *)bufmgr;
    drm_intel_bo_gem *bo_gem;
    unsigned int page_size = getpagesize();
    int ret;
    struct drm_intel_gem_bo_bucket *bucket;
    int alloc_from_cache = 0;
    unsigned long bo_size;

    /* Round the allocated size up to a power of two number of pages. */
    bo_size = 1 << logbase2(size);
    if (bo_size < page_size)
	bo_size = page_size;
    bucket = drm_intel_gem_bo_bucket_for_size(bufmgr_gem, bo_size);

    /* If we don't have caching at this size, don't actually round the
     * allocation up.
     */
    if (bucket == NULL || bucket->max_entries == 0) {
	bo_size = size;
	if (bo_size < page_size)
	    bo_size = page_size;
    }

    pthread_mutex_lock(&bufmgr_gem->lock);
    /* Get a buffer out of the cache if available */
    if (bucket != NULL && bucket->num_entries > 0) {
	struct drm_i915_gem_busy busy;
	
	bo_gem = bucket->head;
        busy.handle = bo_gem->gem_handle;

        ret = ioctl(bufmgr_gem->fd, DRM_IOCTL_I915_GEM_BUSY, &busy);
        alloc_from_cache = (ret == 0 && busy.busy == 0);

	if (alloc_from_cache) {
	    bucket->head = bo_gem->next;
	    if (bo_gem->next == NULL)
		bucket->tail = &bucket->head;
	    bucket->num_entries--;
	}
    }
    pthread_mutex_unlock(&bufmgr_gem->lock);

    if (!alloc_from_cache) {
	struct drm_i915_gem_create create;

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

	bo_gem->bo.size = bo_size;
	memset(&create, 0, sizeof(create));
	create.size = bo_size;

	ret = ioctl(bufmgr_gem->fd, DRM_IOCTL_I915_GEM_CREATE, &create);
	bo_gem->gem_handle = create.handle;
	bo_gem->bo.handle = bo_gem->gem_handle;
	if (ret != 0) {
	    free(bo_gem);
	    return NULL;
	}
	bo_gem->bo.bufmgr = bufmgr;
    }

    bo_gem->name = name;
    bo_gem->refcount = 1;
    bo_gem->validate_index = -1;
    bo_gem->reloc_tree_size = bo_gem->bo.size;
    bo_gem->used_as_reloc_target = 0;

    DBG("bo_create: buf %d (%s) %ldb\n",
	bo_gem->gem_handle, bo_gem->name, size);

    return &bo_gem->bo;
}

/**
 * Returns a drm_intel_bo wrapping the given buffer object handle.
 *
 * This can be used when one application needs to pass a buffer object
 * to another.
 */
drm_intel_bo *
drm_intel_bo_gem_create_from_name(drm_intel_bufmgr *bufmgr, const char *name,
				  unsigned int handle)
{
    drm_intel_bufmgr_gem *bufmgr_gem = (drm_intel_bufmgr_gem *)bufmgr;
    drm_intel_bo_gem *bo_gem;
    int ret;
    struct drm_gem_open open_arg;

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

    memset(&open_arg, 0, sizeof(open_arg));
    open_arg.name = handle;
    ret = ioctl(bufmgr_gem->fd, DRM_IOCTL_GEM_OPEN, &open_arg);
    if (ret != 0) {
	fprintf(stderr, "Couldn't reference %s handle 0x%08x: %s\n",
	       name, handle, strerror(errno));
	free(bo_gem);
	return NULL;
    }
    bo_gem->bo.size = open_arg.size;
    bo_gem->bo.offset = 0;
    bo_gem->bo.virtual = NULL;
    bo_gem->bo.bufmgr = bufmgr;
    bo_gem->name = name;
    bo_gem->refcount = 1;
    bo_gem->validate_index = -1;
    bo_gem->gem_handle = open_arg.handle;
    bo_gem->global_name = handle;

    DBG("bo_create_from_handle: %d (%s)\n", handle, bo_gem->name);

    return &bo_gem->bo;
}

static void
drm_intel_gem_bo_reference(drm_intel_bo *bo)
{
    drm_intel_bufmgr_gem *bufmgr_gem = (drm_intel_bufmgr_gem *)bo->bufmgr;
    drm_intel_bo_gem *bo_gem = (drm_intel_bo_gem *)bo;

    pthread_mutex_lock(&bufmgr_gem->lock);
    bo_gem->refcount++;
    pthread_mutex_unlock(&bufmgr_gem->lock);
}

static void
dri_gem_bo_reference_locked(dri_bo *bo)
{
    drm_intel_bo_gem *bo_gem = (drm_intel_bo_gem *)bo;

    bo_gem->refcount++;
}

static void
drm_intel_gem_bo_reference_locked(drm_intel_bo *bo)
{
    drm_intel_bo_gem *bo_gem = (drm_intel_bo_gem *)bo;

    bo_gem->refcount++;
}

static void
drm_intel_gem_bo_free(drm_intel_bo *bo)
{
    drm_intel_bufmgr_gem *bufmgr_gem = (drm_intel_bufmgr_gem *)bo->bufmgr;
    drm_intel_bo_gem *bo_gem = (drm_intel_bo_gem *)bo;
    struct drm_gem_close close;
    int ret;

    if (bo_gem->mapped)
	munmap (bo_gem->virtual, bo_gem->bo.size);

    /* Close this object */
    close.handle = bo_gem->gem_handle;
    ret = ioctl(bufmgr_gem->fd, DRM_IOCTL_GEM_CLOSE, &close);
    if (ret != 0) {
	fprintf(stderr,
		"DRM_IOCTL_GEM_CLOSE %d failed (%s): %s\n",
		bo_gem->gem_handle, bo_gem->name, strerror(errno));
    }
    free(bo);
}

static void
drm_intel_gem_bo_unreference_locked(drm_intel_bo *bo)
{
    drm_intel_bufmgr_gem *bufmgr_gem = (drm_intel_bufmgr_gem *)bo->bufmgr;
    drm_intel_bo_gem *bo_gem = (drm_intel_bo_gem *)bo;

    if (--bo_gem->refcount == 0) {
	struct drm_intel_gem_bo_bucket *bucket;

	if (bo_gem->relocs != NULL) {
	    int i;

	    /* Unreference all the target buffers */
	    for (i = 0; i < bo_gem->reloc_count; i++)
		 drm_intel_gem_bo_unreference_locked(bo_gem->reloc_target_bo[i]);
	    free(bo_gem->reloc_target_bo);
	    free(bo_gem->relocs);
	}

	DBG("bo_unreference final: %d (%s)\n",
	    bo_gem->gem_handle, bo_gem->name);

	bucket = drm_intel_gem_bo_bucket_for_size(bufmgr_gem, bo->size);
	/* Put the buffer into our internal cache for reuse if we can. */
	if (bucket != NULL &&
	    (bucket->max_entries == -1 ||
	     (bucket->max_entries > 0 &&
	      bucket->num_entries < bucket->max_entries)))
	{
	    bo_gem->name = 0;
	    bo_gem->validate_index = -1;
	    bo_gem->relocs = NULL;
	    bo_gem->reloc_target_bo = NULL;
	    bo_gem->reloc_count = 0;

	    bo_gem->next = NULL;
	    *bucket->tail = bo_gem;
	    bucket->tail = &bo_gem->next;
	    bucket->num_entries++;
	} else {
	    drm_intel_gem_bo_free(bo);
	}
    }
}

static void
drm_intel_gem_bo_unreference(drm_intel_bo *bo)
{
    drm_intel_bufmgr_gem *bufmgr_gem = (drm_intel_bufmgr_gem *)bo->bufmgr;

    pthread_mutex_lock(&bufmgr_gem->lock);
    drm_intel_gem_bo_unreference_locked(bo);
    pthread_mutex_unlock(&bufmgr_gem->lock);
}

static int
drm_intel_gem_bo_map(drm_intel_bo *bo, int write_enable)
{
    drm_intel_bufmgr_gem *bufmgr_gem = (drm_intel_bufmgr_gem *)bo->bufmgr;
    drm_intel_bo_gem *bo_gem = (drm_intel_bo_gem *)bo;
    struct drm_i915_gem_set_domain set_domain;
    int ret;

    pthread_mutex_lock(&bufmgr_gem->lock);

    /* Allow recursive mapping. Mesa may recursively map buffers with
     * nested display loops.
     */
    if (!bo_gem->mapped) {
    
	assert(bo->virtual == NULL);
    
	DBG("bo_map: %d (%s)\n", bo_gem->gem_handle, bo_gem->name);
    
	if (bo_gem->virtual == NULL) {
	    struct drm_i915_gem_mmap mmap_arg;
    
	    memset(&mmap_arg, 0, sizeof(mmap_arg));
	    mmap_arg.handle = bo_gem->gem_handle;
	    mmap_arg.offset = 0;
	    mmap_arg.size = bo->size;
	    ret = ioctl(bufmgr_gem->fd, DRM_IOCTL_I915_GEM_MMAP, &mmap_arg);
	    if (ret != 0) {
		fprintf(stderr, "%s:%d: Error mapping buffer %d (%s): %s .\n",
			__FILE__, __LINE__,
			bo_gem->gem_handle, bo_gem->name, strerror(errno));
	    }
	    bo_gem->virtual = (void *)(uintptr_t)mmap_arg.addr_ptr;
	}
	bo->virtual = bo_gem->virtual;
	bo_gem->swrast = 0;
	bo_gem->mapped = 1;
	DBG("bo_map: %d (%s) -> %p\n", bo_gem->gem_handle, bo_gem->name, bo_gem->virtual);
    }

    if (!bo_gem->swrast) {
	set_domain.handle = bo_gem->gem_handle;
	set_domain.read_domains = I915_GEM_DOMAIN_CPU;
	if (write_enable)
	    set_domain.write_domain = I915_GEM_DOMAIN_CPU;
	else
	    set_domain.write_domain = 0;
	do {
	    ret = ioctl(bufmgr_gem->fd, DRM_IOCTL_I915_GEM_SET_DOMAIN,
			&set_domain);
	} while (ret == -1 && errno == EINTR);
	if (ret != 0) {
	    fprintf (stderr, "%s:%d: Error setting swrast %d: %s\n",
		     __FILE__, __LINE__, bo_gem->gem_handle, strerror (errno));
	}
	bo_gem->swrast = 1;
    }

    pthread_mutex_unlock(&bufmgr_gem->lock);

    return 0;
}

int
drm_intel_gem_bo_map_gtt(drm_intel_bo *bo)
{
    drm_intel_bufmgr_gem *bufmgr_gem = (drm_intel_bufmgr_gem *)bo->bufmgr;
    drm_intel_bo_gem *bo_gem = (drm_intel_bo_gem *)bo;
    struct drm_i915_gem_set_domain set_domain;
    int ret;

    pthread_mutex_lock(&bufmgr_gem->lock);

    /* Allow recursive mapping. Mesa may recursively map buffers with
     * nested display loops.
     */
    if (!bo_gem->mapped) {

	assert(bo->virtual == NULL);

	DBG("bo_map_gtt: %d (%s)\n", bo_gem->gem_handle, bo_gem->name);

	if (bo_gem->virtual == NULL) {
		struct drm_i915_gem_mmap_gtt mmap_arg;

		memset(&mmap_arg, 0, sizeof(mmap_arg));
		mmap_arg.handle = bo_gem->gem_handle;

		/* Get the fake offset back... */
		ret = ioctl(bufmgr_gem->fd, DRM_IOCTL_I915_GEM_MMAP_GTT,
			    &mmap_arg);
		if (ret != 0) {
			fprintf(stderr,
				"%s:%d: Error preparing buffer map %d (%s): %s .\n",
				__FILE__, __LINE__,
				bo_gem->gem_handle, bo_gem->name,
				strerror(errno));
			pthread_mutex_unlock(&bufmgr_gem->lock);
			return ret;
		}

		/* and mmap it */
		bo_gem->virtual = mmap(0, bo->size, PROT_READ | PROT_WRITE,
				       MAP_SHARED, bufmgr_gem->fd,
				       mmap_arg.offset);
		if (bo_gem->virtual == MAP_FAILED) {
			fprintf(stderr,
				"%s:%d: Error mapping buffer %d (%s): %s .\n",
				__FILE__, __LINE__,
				bo_gem->gem_handle, bo_gem->name,
				strerror(errno));
			pthread_mutex_unlock(&bufmgr_gem->lock);
			return errno;
		}
	}

	bo->virtual = bo_gem->virtual;
	bo_gem->mapped = 1;
	DBG("bo_map: %d (%s) -> %p\n", bo_gem->gem_handle, bo_gem->name,
	    bo_gem->virtual);
    }

    /* Now move it to the GTT domain so that the CPU caches are flushed */
    set_domain.handle = bo_gem->gem_handle;
    set_domain.read_domains = I915_GEM_DOMAIN_GTT;
    set_domain.write_domain = I915_GEM_DOMAIN_GTT;
    do {
	    ret = ioctl(bufmgr_gem->fd, DRM_IOCTL_I915_GEM_SET_DOMAIN,
			&set_domain);
    } while (ret == -1 && errno == EINTR);

    if (ret != 0) {
	    fprintf (stderr, "%s:%d: Error setting swrast %d: %s\n",
		     __FILE__, __LINE__, bo_gem->gem_handle, strerror (errno));
    }

    pthread_mutex_unlock(&bufmgr_gem->lock);

    return 0;
}

static int
drm_intel_gem_bo_unmap(drm_intel_bo *bo)
{
    drm_intel_bufmgr_gem *bufmgr_gem = (drm_intel_bufmgr_gem *)bo->bufmgr;
    drm_intel_bo_gem *bo_gem = (drm_intel_bo_gem *)bo;
    struct drm_i915_gem_sw_finish sw_finish;
    int ret;

    if (bo == NULL)
	return 0;

    assert(bo_gem->mapped);

    pthread_mutex_lock(&bufmgr_gem->lock);
    if (bo_gem->swrast) {
	sw_finish.handle = bo_gem->gem_handle;
	do {
	    ret = ioctl(bufmgr_gem->fd, DRM_IOCTL_I915_GEM_SW_FINISH,
			&sw_finish);
	} while (ret == -1 && errno == EINTR);
	bo_gem->swrast = 0;
    }
    pthread_mutex_unlock(&bufmgr_gem->lock);
    return 0;
}

static int
drm_intel_gem_bo_subdata (drm_intel_bo *bo, unsigned long offset,
			  unsigned long size, const void *data)
{
    drm_intel_bufmgr_gem *bufmgr_gem = (drm_intel_bufmgr_gem *)bo->bufmgr;
    drm_intel_bo_gem *bo_gem = (drm_intel_bo_gem *)bo;
    struct drm_i915_gem_pwrite pwrite;
    int ret;

    memset (&pwrite, 0, sizeof (pwrite));
    pwrite.handle = bo_gem->gem_handle;
    pwrite.offset = offset;
    pwrite.size = size;
    pwrite.data_ptr = (uint64_t) (uintptr_t) data;
    do {
	ret = ioctl (bufmgr_gem->fd, DRM_IOCTL_I915_GEM_PWRITE, &pwrite);
    } while (ret == -1 && errno == EINTR);
    if (ret != 0) {
	fprintf (stderr, "%s:%d: Error writing data to buffer %d: (%d %d) %s .\n",
		 __FILE__, __LINE__,
		 bo_gem->gem_handle, (int) offset, (int) size,
		 strerror (errno));
    }
    return 0;
}

static int
drm_intel_gem_bo_get_subdata (drm_intel_bo *bo, unsigned long offset,
			      unsigned long size, void *data)
{
    drm_intel_bufmgr_gem *bufmgr_gem = (drm_intel_bufmgr_gem *)bo->bufmgr;
    drm_intel_bo_gem *bo_gem = (drm_intel_bo_gem *)bo;
    struct drm_i915_gem_pread pread;
    int ret;

    memset (&pread, 0, sizeof (pread));
    pread.handle = bo_gem->gem_handle;
    pread.offset = offset;
    pread.size = size;
    pread.data_ptr = (uint64_t) (uintptr_t) data;
    do {
	ret = ioctl (bufmgr_gem->fd, DRM_IOCTL_I915_GEM_PREAD, &pread);
    } while (ret == -1 && errno == EINTR);
    if (ret != 0) {
	fprintf (stderr, "%s:%d: Error reading data from buffer %d: (%d %d) %s .\n",
		 __FILE__, __LINE__,
		 bo_gem->gem_handle, (int) offset, (int) size,
		 strerror (errno));
    }
    return 0;
}

/** Waits for all GPU rendering to the object to have completed. */
static void
drm_intel_gem_bo_wait_rendering(drm_intel_bo *bo)
{
    return drm_intel_gem_bo_start_gtt_access(bo, 0);
}

/**
 * Sets the object to the GTT read and possibly write domain, used by the X
 * 2D driver in the absence of kernel support to do drm_intel_gem_bo_map_gtt().
 *
 * In combination with drm_intel_gem_bo_pin() and manual fence management, we
 * can do tiled pixmaps this way.
 */
void
drm_intel_gem_bo_start_gtt_access(drm_intel_bo *bo, int write_enable)
{
    drm_intel_bufmgr_gem *bufmgr_gem = (drm_intel_bufmgr_gem *)bo->bufmgr;
    drm_intel_bo_gem *bo_gem = (drm_intel_bo_gem *)bo;
    struct drm_i915_gem_set_domain set_domain;
    int ret;

    set_domain.handle = bo_gem->gem_handle;
    set_domain.read_domains = I915_GEM_DOMAIN_GTT;
    set_domain.write_domain = write_enable ? I915_GEM_DOMAIN_GTT : 0;
    do {
	ret = ioctl(bufmgr_gem->fd, DRM_IOCTL_I915_GEM_SET_DOMAIN, &set_domain);
    } while (ret == -1 && errno == EINTR);
    if (ret != 0) {
	fprintf (stderr, "%s:%d: Error setting memory domains %d (%08x %08x): %s .\n",
		 __FILE__, __LINE__,
		 bo_gem->gem_handle, set_domain.read_domains, set_domain.write_domain,
		 strerror (errno));
    }
}

static void
drm_intel_bufmgr_gem_destroy(drm_intel_bufmgr *bufmgr)
{
    drm_intel_bufmgr_gem *bufmgr_gem = (drm_intel_bufmgr_gem *)bufmgr;
    int i;

    free(bufmgr_gem->exec_objects);
    free(bufmgr_gem->exec_bos);

    pthread_mutex_destroy(&bufmgr_gem->lock);

    /* Free any cached buffer objects we were going to reuse */
    for (i = 0; i < DRM_INTEL_GEM_BO_BUCKETS; i++) {
	struct drm_intel_gem_bo_bucket *bucket = &bufmgr_gem->cache_bucket[i];
	drm_intel_bo_gem *bo_gem;

	while ((bo_gem = bucket->head) != NULL) {
	    bucket->head = bo_gem->next;
	    if (bo_gem->next == NULL)
		bucket->tail = &bucket->head;
	    bucket->num_entries--;

	    drm_intel_gem_bo_free(&bo_gem->bo);
	}
    }

    free(bufmgr);
}

/**
 * Adds the target buffer to the validation list and adds the relocation
 * to the reloc_buffer's relocation list.
 *
 * The relocation entry at the given offset must already contain the
 * precomputed relocation value, because the kernel will optimize out
 * the relocation entry write when the buffer hasn't moved from the
 * last known offset in target_bo.
 */
static int
drm_intel_gem_bo_emit_reloc(drm_intel_bo *bo, uint32_t offset,
			    drm_intel_bo *target_bo, uint32_t target_offset,
			    uint32_t read_domains, uint32_t write_domain)
{
    drm_intel_bufmgr_gem *bufmgr_gem = (drm_intel_bufmgr_gem *)bo->bufmgr;
    drm_intel_bo_gem *bo_gem = (drm_intel_bo_gem *)bo;
    drm_intel_bo_gem *target_bo_gem = (drm_intel_bo_gem *)target_bo;

    pthread_mutex_lock(&bufmgr_gem->lock);

    /* Create a new relocation list if needed */
    if (bo_gem->relocs == NULL)
	drm_intel_setup_reloc_list(bo);

    /* Check overflow */
    assert(bo_gem->reloc_count < bufmgr_gem->max_relocs);

    /* Check args */
    assert (offset <= bo->size - 4);
    assert ((write_domain & (write_domain-1)) == 0);

    /* Make sure that we're not adding a reloc to something whose size has
     * already been accounted for.
     */
    assert(!bo_gem->used_as_reloc_target);
    bo_gem->reloc_tree_size += target_bo_gem->reloc_tree_size;

    /* Flag the target to disallow further relocations in it. */
    target_bo_gem->used_as_reloc_target = 1;

    bo_gem->relocs[bo_gem->reloc_count].offset = offset;
    bo_gem->relocs[bo_gem->reloc_count].delta = target_offset;
    bo_gem->relocs[bo_gem->reloc_count].target_handle =
	target_bo_gem->gem_handle;
    bo_gem->relocs[bo_gem->reloc_count].read_domains = read_domains;
    bo_gem->relocs[bo_gem->reloc_count].write_domain = write_domain;
    bo_gem->relocs[bo_gem->reloc_count].presumed_offset = target_bo->offset;

    bo_gem->reloc_target_bo[bo_gem->reloc_count] = target_bo;
    drm_intel_gem_bo_reference_locked(target_bo);

    bo_gem->reloc_count++;

    pthread_mutex_unlock(&bufmgr_gem->lock);

    return 0;
}

/**
 * Walk the tree of relocations rooted at BO and accumulate the list of
 * validations to be performed and update the relocation buffers with
 * index values into the validation list.
 */
static void
drm_intel_gem_bo_process_reloc(drm_intel_bo *bo)
{
    drm_intel_bo_gem *bo_gem = (drm_intel_bo_gem *)bo;
    int i;

    if (bo_gem->relocs == NULL)
	return;

    for (i = 0; i < bo_gem->reloc_count; i++) {
	drm_intel_bo *target_bo = bo_gem->reloc_target_bo[i];

	/* Continue walking the tree depth-first. */
	drm_intel_gem_bo_process_reloc(target_bo);

	/* Add the target to the validate list */
	drm_intel_add_validate_buffer(target_bo);
    }
}

static void
drm_intel_update_buffer_offsets (drm_intel_bufmgr_gem *bufmgr_gem)
{
    int i;

    for (i = 0; i < bufmgr_gem->exec_count; i++) {
	drm_intel_bo *bo = bufmgr_gem->exec_bos[i];
	drm_intel_bo_gem *bo_gem = (drm_intel_bo_gem *)bo;

	/* Update the buffer offset */
	if (bufmgr_gem->exec_objects[i].offset != bo->offset) {
	    DBG("BO %d (%s) migrated: 0x%08lx -> 0x%08llx\n",
		bo_gem->gem_handle, bo_gem->name, bo->offset,
		(unsigned long long)bufmgr_gem->exec_objects[i].offset);
	    bo->offset = bufmgr_gem->exec_objects[i].offset;
	}
    }
}

static int
drm_intel_gem_bo_exec(drm_intel_bo *bo, int used,
		      drm_clip_rect_t *cliprects, int num_cliprects,
		      int DR4)
{
    drm_intel_bufmgr_gem *bufmgr_gem = (drm_intel_bufmgr_gem *)bo->bufmgr;
    struct drm_i915_gem_execbuffer execbuf;
    int ret, i;

    pthread_mutex_lock(&bufmgr_gem->lock);
    /* Update indices and set up the validate list. */
    drm_intel_gem_bo_process_reloc(bo);

    /* Add the batch buffer to the validation list.  There are no relocations
     * pointing to it.
     */
    drm_intel_add_validate_buffer(bo);

    execbuf.buffers_ptr = (uintptr_t)bufmgr_gem->exec_objects;
    execbuf.buffer_count = bufmgr_gem->exec_count;
    execbuf.batch_start_offset = 0;
    execbuf.batch_len = used;
    execbuf.cliprects_ptr = (uintptr_t)cliprects;
    execbuf.num_cliprects = num_cliprects;
    execbuf.DR1 = 0;
    execbuf.DR4 = DR4;

    do {
	ret = ioctl(bufmgr_gem->fd, DRM_IOCTL_I915_GEM_EXECBUFFER, &execbuf);
    } while (ret != 0 && errno == EAGAIN);

    drm_intel_update_buffer_offsets (bufmgr_gem);

    if (bufmgr_gem->bufmgr.debug)
	drm_intel_gem_dump_validation_list(bufmgr_gem);

    for (i = 0; i < bufmgr_gem->exec_count; i++) {
	drm_intel_bo *bo = bufmgr_gem->exec_bos[i];
	drm_intel_bo_gem *bo_gem = (drm_intel_bo_gem *)bo;

	/* Need to call swrast on next bo_map */
	bo_gem->swrast = 0;

	/* Disconnect the buffer from the validate list */
	bo_gem->validate_index = -1;
	drm_intel_gem_bo_unreference_locked(bo);
	bufmgr_gem->exec_bos[i] = NULL;
    }
    bufmgr_gem->exec_count = 0;
    pthread_mutex_unlock(&bufmgr_gem->lock);

    return 0;
}

static int
drm_intel_gem_bo_pin(drm_intel_bo *bo, uint32_t alignment)
{
    drm_intel_bufmgr_gem *bufmgr_gem = (drm_intel_bufmgr_gem *)bo->bufmgr;
    drm_intel_bo_gem *bo_gem = (drm_intel_bo_gem *)bo;
    struct drm_i915_gem_pin pin;
    int ret;

    pin.handle = bo_gem->gem_handle;
    pin.alignment = alignment;

    ret = ioctl(bufmgr_gem->fd, DRM_IOCTL_I915_GEM_PIN, &pin);
    if (ret != 0)
	return -errno;

    bo->offset = pin.offset;
    return 0;
}

static int
drm_intel_gem_bo_unpin(drm_intel_bo *bo)
{
    drm_intel_bufmgr_gem *bufmgr_gem = (drm_intel_bufmgr_gem *)bo->bufmgr;
    drm_intel_bo_gem *bo_gem = (drm_intel_bo_gem *)bo;
    struct drm_i915_gem_unpin unpin;
    int ret;

    unpin.handle = bo_gem->gem_handle;

    ret = ioctl(bufmgr_gem->fd, DRM_IOCTL_I915_GEM_UNPIN, &unpin);
    if (ret != 0)
	return -errno;

    return 0;
}

static int
drm_intel_gem_bo_set_tiling(drm_intel_bo *bo, uint32_t *tiling_mode,
			    uint32_t stride)
{
    drm_intel_bufmgr_gem *bufmgr_gem = (drm_intel_bufmgr_gem *)bo->bufmgr;
    drm_intel_bo_gem *bo_gem = (drm_intel_bo_gem *)bo;
    struct drm_i915_gem_set_tiling set_tiling;
    int ret;

    set_tiling.handle = bo_gem->gem_handle;
    set_tiling.tiling_mode = *tiling_mode;
    set_tiling.stride = stride;

    ret = ioctl(bufmgr_gem->fd, DRM_IOCTL_I915_GEM_SET_TILING, &set_tiling);
    if (ret != 0) {
	*tiling_mode = I915_TILING_NONE;
	return -errno;
    }

    *tiling_mode = set_tiling.tiling_mode;
    return 0;
}

static int
drm_intel_gem_bo_get_tiling(drm_intel_bo *bo, uint32_t *tiling_mode,
			    uint32_t *swizzle_mode)
{
    drm_intel_bufmgr_gem *bufmgr_gem = (drm_intel_bufmgr_gem *)bo->bufmgr;
    drm_intel_bo_gem *bo_gem = (drm_intel_bo_gem *)bo;
    struct drm_i915_gem_get_tiling get_tiling;
    int ret;

    get_tiling.handle = bo_gem->gem_handle;

    ret = ioctl(bufmgr_gem->fd, DRM_IOCTL_I915_GEM_GET_TILING, &get_tiling);
    if (ret != 0) {
	*tiling_mode = I915_TILING_NONE;
	*swizzle_mode = I915_BIT_6_SWIZZLE_NONE;
	return -errno;
    }

    *tiling_mode = get_tiling.tiling_mode;
    *swizzle_mode = get_tiling.swizzle_mode;
    return 0;
}

static int
drm_intel_gem_bo_flink(drm_intel_bo *bo, uint32_t *name)
{
    drm_intel_bufmgr_gem *bufmgr_gem = (drm_intel_bufmgr_gem *)bo->bufmgr;
    drm_intel_bo_gem *bo_gem = (drm_intel_bo_gem *)bo;
    struct drm_gem_flink flink;
    int ret;

    if (!bo_gem->global_name) {
	flink.handle = bo_gem->gem_handle;
    
	ret = ioctl(bufmgr_gem->fd, DRM_IOCTL_GEM_FLINK, &flink);
	if (ret != 0)
	    return -errno;
	bo_gem->global_name = flink.name;
    }
    
    *name = bo_gem->global_name;
    return 0;
}

/**
 * Enables unlimited caching of buffer objects for reuse.
 *
 * This is potentially very memory expensive, as the cache at each bucket
 * size is only bounded by how many buffers of that size we've managed to have
 * in flight at once.
 */
void
drm_intel_bufmgr_gem_enable_reuse(drm_intel_bufmgr *bufmgr)
{
    drm_intel_bufmgr_gem *bufmgr_gem = (drm_intel_bufmgr_gem *)bufmgr;
    int i;

    for (i = 0; i < DRM_INTEL_GEM_BO_BUCKETS; i++) {
	bufmgr_gem->cache_bucket[i].max_entries = -1;
    }
}

/**
 * Return the additional aperture space required by the tree of buffer objects
 * rooted at bo.
 */
static int
drm_intel_gem_bo_get_aperture_space(drm_intel_bo *bo)
{
    drm_intel_bo_gem *bo_gem = (drm_intel_bo_gem *)bo;
    int i;
    int total = 0;

    if (bo == NULL || bo_gem->included_in_check_aperture)
	return 0;

    total += bo->size;
    bo_gem->included_in_check_aperture = 1;

    for (i = 0; i < bo_gem->reloc_count; i++)
	total += drm_intel_gem_bo_get_aperture_space(bo_gem->reloc_target_bo[i]);

    return total;
}

/**
 * Clear the flag set by drm_intel_gem_bo_get_aperture_space() so we're ready
 * for the next drm_intel_bufmgr_check_aperture_space() call.
 */
static void
drm_intel_gem_bo_clear_aperture_space_flag(drm_intel_bo *bo)
{
    drm_intel_bo_gem *bo_gem = (drm_intel_bo_gem *)bo;
    int i;

    if (bo == NULL || !bo_gem->included_in_check_aperture)
	return;

    bo_gem->included_in_check_aperture = 0;

    for (i = 0; i < bo_gem->reloc_count; i++)
	drm_intel_gem_bo_clear_aperture_space_flag(bo_gem->reloc_target_bo[i]);
}

/**
 * Return -1 if the batchbuffer should be flushed before attempting to
 * emit rendering referencing the buffers pointed to by bo_array.
 *
 * This is required because if we try to emit a batchbuffer with relocations
 * to a tree of buffers that won't simultaneously fit in the aperture,
 * the rendering will return an error at a point where the software is not
 * prepared to recover from it.
 *
 * However, we also want to emit the batchbuffer significantly before we reach
 * the limit, as a series of batchbuffers each of which references buffers
 * covering almost all of the aperture means that at each emit we end up
 * waiting to evict a buffer from the last rendering, and we get synchronous
 * performance.  By emitting smaller batchbuffers, we eat some CPU overhead to
 * get better parallelism.
 */
static int
drm_intel_gem_check_aperture_space(drm_intel_bo **bo_array, int count)
{
    drm_intel_bufmgr_gem *bufmgr_gem = (drm_intel_bufmgr_gem *)bo_array[0]->bufmgr;
    unsigned int total = 0;
    unsigned int threshold = bufmgr_gem->gtt_size * 3 / 4;
    int i;

    for (i = 0; i < count; i++) {
	drm_intel_bo_gem *bo_gem = (drm_intel_bo_gem *)bo_array[i];
	if (bo_gem != NULL)
		total += bo_gem->reloc_tree_size;
    }

    if (total > threshold) {
	total = 0;
	for (i = 0; i < count; i++)
	    total += drm_intel_gem_bo_get_aperture_space(bo_array[i]);

	for (i = 0; i < count; i++)
	    drm_intel_gem_bo_clear_aperture_space_flag(bo_array[i]);
    }

    if (total > bufmgr_gem->gtt_size * 3 / 4) {
	DBG("check_space: overflowed available aperture, %dkb vs %dkb\n",
	    total / 1024, (int)bufmgr_gem->gtt_size / 1024);
	return -1;
    } else {
	DBG("drm_check_space: total %dkb vs bufgr %dkb\n", total / 1024 ,
	    (int)bufmgr_gem->gtt_size / 1024);
	return 0;
    }
}

/**
 * Initializes the GEM buffer manager, which uses the kernel to allocate, map,
 * and manage map buffer objections.
 *
 * \param fd File descriptor of the opened DRM device.
 */
drm_intel_bufmgr *
drm_intel_bufmgr_gem_init(int fd, int batch_size)
{
    drm_intel_bufmgr_gem *bufmgr_gem;
    struct drm_i915_gem_get_aperture aperture;
    int ret, i;

    bufmgr_gem = calloc(1, sizeof(*bufmgr_gem));
    bufmgr_gem->fd = fd;

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

    ret = ioctl(bufmgr_gem->fd, DRM_IOCTL_I915_GEM_GET_APERTURE, &aperture);

    if (ret == 0)
	bufmgr_gem->gtt_size = aperture.aper_available_size;
    else {
	fprintf(stderr, "DRM_IOCTL_I915_GEM_APERTURE failed: %s\n",
		strerror(errno));
	bufmgr_gem->gtt_size = 128 * 1024 * 1024;
	fprintf(stderr, "Assuming %dkB available aperture size.\n"
		"May lead to reduced performance or incorrect rendering.\n",
		(int)bufmgr_gem->gtt_size / 1024);
    }

    /* Let's go with one relocation per every 2 dwords (but round down a bit
     * since a power of two will mean an extra page allocation for the reloc
     * buffer).
     *
     * Every 4 was too few for the blender benchmark.
     */
    bufmgr_gem->max_relocs = batch_size / sizeof(uint32_t) / 2 - 2;

    bufmgr_gem->bufmgr.bo_alloc = drm_intel_gem_bo_alloc;
    bufmgr_gem->bufmgr.bo_reference = drm_intel_gem_bo_reference;
    bufmgr_gem->bufmgr.bo_unreference = drm_intel_gem_bo_unreference;
    bufmgr_gem->bufmgr.bo_map = drm_intel_gem_bo_map;
    bufmgr_gem->bufmgr.bo_unmap = drm_intel_gem_bo_unmap;
    bufmgr_gem->bufmgr.bo_subdata = drm_intel_gem_bo_subdata;
    bufmgr_gem->bufmgr.bo_get_subdata = drm_intel_gem_bo_get_subdata;
    bufmgr_gem->bufmgr.bo_wait_rendering = drm_intel_gem_bo_wait_rendering;
    bufmgr_gem->bufmgr.bo_emit_reloc = drm_intel_gem_bo_emit_reloc;
    bufmgr_gem->bufmgr.bo_pin = drm_intel_gem_bo_pin;
    bufmgr_gem->bufmgr.bo_unpin = drm_intel_gem_bo_unpin;
    bufmgr_gem->bufmgr.bo_get_tiling = drm_intel_gem_bo_get_tiling;
    bufmgr_gem->bufmgr.bo_set_tiling = drm_intel_gem_bo_set_tiling;
    bufmgr_gem->bufmgr.bo_flink = drm_intel_gem_bo_flink;
    bufmgr_gem->bufmgr.bo_exec = drm_intel_gem_bo_exec;
    bufmgr_gem->bufmgr.destroy = drm_intel_bufmgr_gem_destroy;
    bufmgr_gem->bufmgr.debug = 0;
    bufmgr_gem->bufmgr.check_aperture_space = drm_intel_gem_check_aperture_space;
    /* Initialize the linked lists for BO reuse cache. */
    for (i = 0; i < DRM_INTEL_GEM_BO_BUCKETS; i++)
	bufmgr_gem->cache_bucket[i].tail = &bufmgr_gem->cache_bucket[i].head;

    return &bufmgr_gem->bufmgr;
}