summaryrefslogtreecommitdiff
path: root/src/mesa/drivers/dri/intel/intel_bufmgr_gem.c
blob: 07f782ca3a8a90e5b4bceb3e15903b14f54411fc (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
/**************************************************************************
 *
 * 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>
 */

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

#include "errno.h"
#include "mtypes.h"
#include "dri_bufmgr.h"
#include "string.h"
#include "imports.h"

#include "i915_drm.h"

#include "intel_bufmgr_gem.h"

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

struct intel_validate_entry {
    dri_bo *bo;
    struct drm_i915_op_arg bo_arg;
};

struct dri_gem_bo_bucket_entry {
   uint32_t gem_handle;
   struct dri_gem_bo_bucket_entry *next;
};

struct dri_gem_bo_bucket {
   struct dri_gem_bo_bucket_entry *head;
   struct dri_gem_bo_bucket_entry **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 INTEL_GEM_BO_BUCKETS	16
typedef struct _dri_bufmgr_gem {
    dri_bufmgr bufmgr;

    int fd;

    uint32_t max_relocs;

    struct drm_i915_gem_validate_entry *validate_array;
    dri_bo **validate_bo;
    int validate_array_size;
    int validate_count;

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

    struct drm_i915_gem_execbuffer exec_arg;
} dri_bufmgr_gem;

typedef struct _dri_bo_gem {
    dri_bo bo;

    int refcount;
    unsigned int map_count;
    uint32_t gem_handle;
    const char *name;

    /**
     * Index of the buffer within the validation list while preparing a
     * batchbuffer execution.
     */
    int validate_index;

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

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

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

   return log2;
}

static struct dri_gem_bo_bucket *
dri_gem_bo_bucket_for_size(dri_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 >= INTEL_GEM_BO_BUCKETS)
	return NULL;

    return &bufmgr_gem->cache_bucket[i];
}


static void dri_gem_dump_validation_list(dri_bufmgr_gem *bufmgr_gem)
{
    int i, j;

    for (i = 0; i < bufmgr_gem->validate_count; i++) {
	dri_bo *bo = bufmgr_gem->validate_bo[i];
	dri_bo_gem *bo_gem = (dri_bo_gem *)bo;

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

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

	    DBG("%2d: %s@0x%08llx -> %s@0x%08lx + 0x%08x\n",
		i,
		bo_gem->name, bo_gem->relocs[j].offset,
		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
intel_add_validate_buffer(dri_bo *bo)
{
    dri_bufmgr_gem *bufmgr_gem = (dri_bufmgr_gem *)bo->bufmgr;
    dri_bo_gem *bo_gem = (dri_bo_gem *)bo;
    int index;

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

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

	if (new_size == 0)
	    new_size = 5;

	bufmgr_gem->validate_array =
	    realloc(bufmgr_gem->validate_array,
		    sizeof(*bufmgr_gem->validate_array) * new_size);
	bufmgr_gem->validate_bo =
	    realloc(bufmgr_gem->validate_bo,
		    sizeof(*bufmgr_gem->validate_bo) * new_size);
	bufmgr_gem->validate_array_size = new_size;
    }

    index = bufmgr_gem->validate_count;
    /* Fill in array entry */
    bufmgr_gem->validate_array[index].buffer_handle = bo_gem->gem_handle;
    bufmgr_gem->validate_array[index].relocation_count = bo_gem->reloc_count;
    bufmgr_gem->validate_array[index].relocs_ptr = (uintptr_t)bo_gem->relocs;
    bufmgr_gem->validate_bo[index] = bo;
    dri_bo_reference(bo);
    bufmgr_gem->validate_count++;
}


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

static int
intel_setup_reloc_list(dri_bo *bo)
{
    dri_bo_gem *bo_gem = (dri_bo_gem *)bo;
    dri_bufmgr_gem *bufmgr_gem = (dri_bufmgr_gem *)bo->bufmgr;

    bo_gem->relocs = calloc(bufmgr_gem->max_relocs,
			    sizeof(struct drm_i915_gem_relocation_entry));
    bo_gem->reloc_target_bo = calloc(1, sizeof(dri_bo *));

    return 0;
}

static dri_bo *
dri_gem_alloc(dri_bufmgr *bufmgr, const char *name,
	      unsigned long size, unsigned int alignment,
	      uint64_t location_mask)
{
    dri_bufmgr_gem *bufmgr_gem = (dri_bufmgr_gem *)bufmgr;
    dri_bo_gem *bo_gem;
    unsigned int page_size = getpagesize();
    int ret;
    struct dri_gem_bo_bucket *bucket;
    GLboolean alloc_from_cache = GL_FALSE;

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

    /* Round the allocated size up to a power of two number of pages. */
    bo_gem->bo.size = 1 << logbase2(size);
    if (bo_gem->bo.size < page_size)
	bo_gem->bo.size = page_size;
    bucket = dri_gem_bo_bucket_for_size(bufmgr_gem, bo_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_gem->bo.size = size;
	if (bo_gem->bo.size < page_size)
	    bo_gem->bo.size = page_size;
    }

    /* Get a buffer out of the cache if available */
    if (bucket != NULL && bucket->num_entries > 0) {
	struct dri_gem_bo_bucket_entry *entry = bucket->head;
#if 0
	int busy;

	/* XXX */
	/* Check if the buffer is still in flight.  If not, reuse it. */
	ret = drmBOBusy(bufmgr_gem->fd, &entry->drm_bo, &busy);
	alloc_from_cache = (ret == 0 && busy == 0);
#else
	alloc_from_cache = 0;
#endif

	if (alloc_from_cache) {
	    bucket->head = entry->next;
	    if (entry->next == NULL)
		bucket->tail = &bucket->head;
	    bucket->num_entries--;

	    bo_gem->gem_handle = entry->gem_handle;
	    free(entry);
	}
    }

    if (!alloc_from_cache) {
	struct drm_gem_alloc alloc;

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

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

    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;

    DBG("bo_create: %p (%s) %ldb\n", &bo_gem->bo, bo_gem->name, size);

    return &bo_gem->bo;
}

/* Our GEM backend doesn't allow creation of static buffers, as that requires
 * privelege for the non-fake case, and the lock in the fake case where we were
 * working around the X Server not creating buffers and passing handles to us.
 */
static dri_bo *
dri_gem_alloc_static(dri_bufmgr *bufmgr, const char *name,
		     unsigned long offset, unsigned long size, void *virtual,
		     uint64_t location_mask)
{
    return NULL;
}

/**
 * Returns a dri_bo wrapping the given buffer object handle.
 *
 * This can be used when one application needs to pass a buffer object
 * to another.
 */
dri_bo *
intel_gem_bo_create_from_handle(dri_bufmgr *bufmgr, const char *name,
			      unsigned int handle)
{
    dri_bufmgr_gem *bufmgr_gem = (dri_bufmgr_gem *)bufmgr;
    dri_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(-ret));
	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;

    DBG("bo_create_from_handle: %p %08x (%s)\n",
	&bo_gem->bo, handle, bo_gem->name);

    return &bo_gem->bo;
}

static void
dri_gem_bo_reference(dri_bo *bo)
{
    dri_bo_gem *bo_gem = (dri_bo_gem *)bo;

    bo_gem->refcount++;
}

static void
dri_gem_bo_unreference(dri_bo *bo)
{
    dri_bufmgr_gem *bufmgr_gem = (dri_bufmgr_gem *)bo->bufmgr;
    dri_bo_gem *bo_gem = (dri_bo_gem *)bo;

    if (!bo)
	return;

    if (--bo_gem->refcount == 0) {
	struct dri_gem_bo_bucket *bucket;
	int ret;

	assert(bo_gem->map_count == 0);

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

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

	bucket = dri_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)))
	{
	    struct dri_gem_bo_bucket_entry *entry;

	    entry = calloc(1, sizeof(*entry));
	    entry->gem_handle = bo_gem->gem_handle;

	    entry->next = NULL;
	    *bucket->tail = entry;
	    bucket->tail = &entry->next;
	    bucket->num_entries++;
	} else {
	    struct drm_gem_unreference unref;

	    /* Decrement the kernel refcount for the buffer. */
	    unref.handle = bo_gem->gem_handle;
	    ret = ioctl(bufmgr_gem->fd, DRM_IOCTL_GEM_UNREFERENCE, &unref);
	    if (ret != 0) {
	       fprintf(stderr, "DRM_IOCTL_GEM_UNREFERENCE failed (%s): %s\n",
		       bo_gem->name, strerror(-ret));
	    }
	}

	DBG("bo_unreference final: %p (%s)\n", &bo_gem->bo, bo_gem->name);

	free(bo);
	return;
    }
}

static int
dri_gem_bo_map(dri_bo *bo, GLboolean write_enable)
{
    dri_bufmgr_gem *bufmgr_gem;
    dri_bo_gem *bo_gem = (dri_bo_gem *)bo;
    int ret;

    bufmgr_gem = (dri_bufmgr_gem *)bo->bufmgr;

    /* Allow recursive mapping. Mesa may recursively map buffers with
     * nested display loops.
     */
    if (bo_gem->map_count++ != 0)
	return 0;

    assert(bo->virtual == NULL);

    DBG("bo_map: %p (%s)\n", &bo_gem->bo, bo_gem->name);

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

	mmap_arg.handle = bo_gem->gem_handle;
	ret = ioctl(bufmgr_gem->fd, DRM_IOCTL_GEM_MMAP, &mmap_arg);
	if (ret != 0) {
	    fprintf(stderr, "%s:%d: Error mapping buffer %s: %s .\n",
		    __FILE__, __LINE__, bo_gem->name, strerror(-ret));
	}
	bo_gem->virtual = (void *)(uintptr_t)mmap_arg.addr_ptr;
    }

    /* XXX Synchronization with hardware */

    bo->virtual = bo_gem->virtual;

    return 0;
}

static int
dri_gem_bo_unmap(dri_bo *bo)
{
    dri_bufmgr_gem *bufmgr_gem;
    dri_bo_gem *bo_gem = (dri_bo_gem *)bo;

    if (bo == NULL)
	return 0;

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

    bufmgr_gem = (dri_bufmgr_gem *)bo->bufmgr;

    assert(bo->virtual != NULL);

    DBG("bo_unmap: %p (%s)\n", &bo_gem->bo, bo_gem->name);

    munmap(bo_gem->virtual, bo->size);
    bo_gem->virtual = NULL;
    bo->virtual = NULL;

    return 0;
}

static void
dri_bufmgr_gem_destroy(dri_bufmgr *bufmgr)
{
    dri_bufmgr_gem *bufmgr_gem = (dri_bufmgr_gem *)bufmgr;
    int i;

    free(bufmgr_gem->validate_array);
    free(bufmgr_gem->validate_bo);

    /* Free any cached buffer objects we were going to reuse */
    for (i = 0; i < INTEL_GEM_BO_BUCKETS; i++) {
	struct dri_gem_bo_bucket *bucket = &bufmgr_gem->cache_bucket[i];
	struct dri_gem_bo_bucket_entry *entry;

	while ((entry = bucket->head) != NULL) {
	    struct drm_gem_unreference unref;
	    int ret;

	    bucket->head = entry->next;
	    if (entry->next == NULL)
		bucket->tail = &bucket->head;
	    bucket->num_entries--;

	    /* Decrement the kernel refcount for the buffer. */
	    unref.handle = entry->gem_handle;
	    ret = ioctl(bufmgr_gem->fd, DRM_IOCTL_GEM_UNREFERENCE, &unref);
	    if (ret != 0) {
	       fprintf(stderr, "DRM_IOCTL_GEM_UNREFERENCE failed: %s\n",
		       strerror(-ret));
	    }

	    free(entry);
	}
    }

    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
dri_gem_emit_reloc(dri_bo *bo, uint64_t flags, GLuint delta,
		   GLuint offset, dri_bo *target_bo)
{
    dri_bufmgr_gem *bufmgr_gem = (dri_bufmgr_gem *)bo->bufmgr;
    dri_bo_gem *bo_gem = (dri_bo_gem *)bo;
    dri_bo_gem *target_bo_gem = (dri_bo_gem *)target_bo;

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

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

    bo_gem->relocs[bo_gem->reloc_count].offset = offset;
    bo_gem->relocs[bo_gem->reloc_count].delta = delta;
    bo_gem->relocs[bo_gem->reloc_count].target_handle =
	target_bo_gem->gem_handle;

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

    bo_gem->reloc_count++;
    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
dri_gem_bo_process_reloc(dri_bo *bo)
{
    dri_bo_gem *bo_gem = (dri_bo_gem *)bo;
    int i;

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

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

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

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

static void *
dri_gem_process_reloc(dri_bo *batch_buf)
{
    dri_bufmgr_gem *bufmgr_gem = (dri_bufmgr_gem *)batch_buf->bufmgr;

    /* Update indices and set up the validate list. */
    dri_gem_bo_process_reloc(batch_buf);

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

    bufmgr_gem->exec_arg.buffers_ptr = (uintptr_t)bufmgr_gem->validate_array;
    bufmgr_gem->exec_arg.buffer_count = bufmgr_gem->validate_count;
    bufmgr_gem->exec_arg.batch_start_offset = bufmgr_gem->validate_count;

    return &bufmgr_gem->exec_arg;
}

static void
intel_update_buffer_offsets (dri_bufmgr_gem *bufmgr_gem)
{
    int i;

    for (i = 0; i < bufmgr_gem->validate_count; i++) {
	dri_bo *bo = bufmgr_gem->validate_bo[i];
	dri_bo_gem *bo_gem = (dri_bo_gem *)bo;

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

static void
dri_gem_post_submit(dri_bo *batch_buf)
{
    dri_bufmgr_gem *bufmgr_gem = (dri_bufmgr_gem *)batch_buf->bufmgr;
    int i;

    intel_update_buffer_offsets (bufmgr_gem);

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

    for (i = 0; i < bufmgr_gem->validate_count; i++) {
	dri_bo *bo = bufmgr_gem->validate_bo[i];
	dri_bo_gem *bo_gem = (dri_bo_gem *)bo;

	/* Disconnect the buffer from the validate list */
	bo_gem->validate_index = -1;
	dri_bo_unreference(bo);
	bufmgr_gem->validate_bo[i] = NULL;
    }
    bufmgr_gem->validate_count = 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
intel_gem_enable_bo_reuse(dri_bufmgr *bufmgr)
{
    /*
    dri_bufmgr_gem *bufmgr_gem = (dri_bufmgr_gem *)bufmgr;
    int i;

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

/*
 *
 */
static int
dri_gem_check_aperture_space(dri_bo *bo)
{
    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.
 */
dri_bufmgr *
intel_bufmgr_gem_init(int fd, int batch_size)
{
    dri_bufmgr_gem *bufmgr_gem;
    int i;

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

    /* 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 = dri_gem_alloc;
    bufmgr_gem->bufmgr.bo_alloc_static = dri_gem_alloc_static;
    bufmgr_gem->bufmgr.bo_reference = dri_gem_bo_reference;
    bufmgr_gem->bufmgr.bo_unreference = dri_gem_bo_unreference;
    bufmgr_gem->bufmgr.bo_map = dri_gem_bo_map;
    bufmgr_gem->bufmgr.bo_unmap = dri_gem_bo_unmap;
    bufmgr_gem->bufmgr.destroy = dri_bufmgr_gem_destroy;
    bufmgr_gem->bufmgr.emit_reloc = dri_gem_emit_reloc;
    bufmgr_gem->bufmgr.process_relocs = dri_gem_process_reloc;
    bufmgr_gem->bufmgr.post_submit = dri_gem_post_submit;
    bufmgr_gem->bufmgr.debug = GL_FALSE;
    bufmgr_gem->bufmgr.check_aperture_space = dri_gem_check_aperture_space;
    /* Initialize the linked lists for BO reuse cache. */
    for (i = 0; i < INTEL_GEM_BO_BUCKETS; i++)
	bufmgr_gem->cache_bucket[i].tail = &bufmgr_gem->cache_bucket[i].head;

    return &bufmgr_gem->bufmgr;
}