aboutsummaryrefslogtreecommitdiff
path: root/drivers/char/drm/savage_state.c
blob: 1ca1e9cb5a3396be02753a0d584d90e269f5b73c (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
/* savage_state.c -- State and drawing support for Savage
 *
 * Copyright 2004  Felix Kuehling
 * 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 FELIX KUEHLING 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.
 */
#include "drmP.h"
#include "savage_drm.h"
#include "savage_drv.h"

void savage_emit_clip_rect_s3d(drm_savage_private_t * dev_priv,
			       const drm_clip_rect_t * pbox)
{
	uint32_t scstart = dev_priv->state.s3d.new_scstart;
	uint32_t scend = dev_priv->state.s3d.new_scend;
	scstart = (scstart & ~SAVAGE_SCISSOR_MASK_S3D) |
	    ((uint32_t) pbox->x1 & 0x000007ff) |
	    (((uint32_t) pbox->y1 << 16) & 0x07ff0000);
	scend = (scend & ~SAVAGE_SCISSOR_MASK_S3D) |
	    (((uint32_t) pbox->x2 - 1) & 0x000007ff) |
	    ((((uint32_t) pbox->y2 - 1) << 16) & 0x07ff0000);
	if (scstart != dev_priv->state.s3d.scstart ||
	    scend != dev_priv->state.s3d.scend) {
		DMA_LOCALS;
		BEGIN_DMA(4);
		DMA_WRITE(BCI_CMD_WAIT | BCI_CMD_WAIT_3D);
		DMA_SET_REGISTERS(SAVAGE_SCSTART_S3D, 2);
		DMA_WRITE(scstart);
		DMA_WRITE(scend);
		dev_priv->state.s3d.scstart = scstart;
		dev_priv->state.s3d.scend = scend;
		dev_priv->waiting = 1;
		DMA_COMMIT();
	}
}

void savage_emit_clip_rect_s4(drm_savage_private_t * dev_priv,
			      const drm_clip_rect_t * pbox)
{
	uint32_t drawctrl0 = dev_priv->state.s4.new_drawctrl0;
	uint32_t drawctrl1 = dev_priv->state.s4.new_drawctrl1;
	drawctrl0 = (drawctrl0 & ~SAVAGE_SCISSOR_MASK_S4) |
	    ((uint32_t) pbox->x1 & 0x000007ff) |
	    (((uint32_t) pbox->y1 << 12) & 0x00fff000);
	drawctrl1 = (drawctrl1 & ~SAVAGE_SCISSOR_MASK_S4) |
	    (((uint32_t) pbox->x2 - 1) & 0x000007ff) |
	    ((((uint32_t) pbox->y2 - 1) << 12) & 0x00fff000);
	if (drawctrl0 != dev_priv->state.s4.drawctrl0 ||
	    drawctrl1 != dev_priv->state.s4.drawctrl1) {
		DMA_LOCALS;
		BEGIN_DMA(4);
		DMA_WRITE(BCI_CMD_WAIT | BCI_CMD_WAIT_3D);
		DMA_SET_REGISTERS(SAVAGE_DRAWCTRL0_S4, 2);
		DMA_WRITE(drawctrl0);
		DMA_WRITE(drawctrl1);
		dev_priv->state.s4.drawctrl0 = drawctrl0;
		dev_priv->state.s4.drawctrl1 = drawctrl1;
		dev_priv->waiting = 1;
		DMA_COMMIT();
	}
}

static int savage_verify_texaddr(drm_savage_private_t * dev_priv, int unit,
				 uint32_t addr)
{
	if ((addr & 6) != 2) {	/* reserved bits */
		DRM_ERROR("bad texAddr%d %08x (reserved bits)\n", unit, addr);
		return DRM_ERR(EINVAL);
	}
	if (!(addr & 1)) {	/* local */
		addr &= ~7;
		if (addr < dev_priv->texture_offset ||
		    addr >= dev_priv->texture_offset + dev_priv->texture_size) {
			DRM_ERROR
			    ("bad texAddr%d %08x (local addr out of range)\n",
			     unit, addr);
			return DRM_ERR(EINVAL);
		}
	} else {		/* AGP */
		if (!dev_priv->agp_textures) {
			DRM_ERROR("bad texAddr%d %08x (AGP not available)\n",
				  unit, addr);
			return DRM_ERR(EINVAL);
		}
		addr &= ~7;
		if (addr < dev_priv->agp_textures->offset ||
		    addr >= (dev_priv->agp_textures->offset +
			     dev_priv->agp_textures->size)) {
			DRM_ERROR
			    ("bad texAddr%d %08x (AGP addr out of range)\n",
			     unit, addr);
			return DRM_ERR(EINVAL);
		}
	}
	return 0;
}

#define SAVE_STATE(reg,where)			\
	if(start <= reg && start+count > reg)	\
		dev_priv->state.where = regs[reg - start]
#define SAVE_STATE_MASK(reg,where,mask) do {			\
	if(start <= reg && start+count > reg) {			\
		uint32_t tmp;					\
		tmp = regs[reg - start];			\
		dev_priv->state.where = (tmp & (mask)) |	\
			(dev_priv->state.where & ~(mask));	\
	}							\
} while (0)

static int savage_verify_state_s3d(drm_savage_private_t * dev_priv,
				   unsigned int start, unsigned int count,
				   const uint32_t *regs)
{
	if (start < SAVAGE_TEXPALADDR_S3D ||
	    start + count - 1 > SAVAGE_DESTTEXRWWATERMARK_S3D) {
		DRM_ERROR("invalid register range (0x%04x-0x%04x)\n",
			  start, start + count - 1);
		return DRM_ERR(EINVAL);
	}

	SAVE_STATE_MASK(SAVAGE_SCSTART_S3D, s3d.new_scstart,
			~SAVAGE_SCISSOR_MASK_S3D);
	SAVE_STATE_MASK(SAVAGE_SCEND_S3D, s3d.new_scend,
			~SAVAGE_SCISSOR_MASK_S3D);

	/* if any texture regs were changed ... */
	if (start <= SAVAGE_TEXCTRL_S3D &&
	    start + count > SAVAGE_TEXPALADDR_S3D) {
		/* ... check texture state */
		SAVE_STATE(SAVAGE_TEXCTRL_S3D, s3d.texctrl);
		SAVE_STATE(SAVAGE_TEXADDR_S3D, s3d.texaddr);
		if (dev_priv->state.s3d.texctrl & SAVAGE_TEXCTRL_TEXEN_MASK)
			return savage_verify_texaddr(dev_priv, 0,
						dev_priv->state.s3d.texaddr);
	}

	return 0;
}

static int savage_verify_state_s4(drm_savage_private_t * dev_priv,
				  unsigned int start, unsigned int count,
				  const uint32_t *regs)
{
	int ret = 0;

	if (start < SAVAGE_DRAWLOCALCTRL_S4 ||
	    start + count - 1 > SAVAGE_TEXBLENDCOLOR_S4) {
		DRM_ERROR("invalid register range (0x%04x-0x%04x)\n",
			  start, start + count - 1);
		return DRM_ERR(EINVAL);
	}

	SAVE_STATE_MASK(SAVAGE_DRAWCTRL0_S4, s4.new_drawctrl0,
			~SAVAGE_SCISSOR_MASK_S4);
	SAVE_STATE_MASK(SAVAGE_DRAWCTRL1_S4, s4.new_drawctrl1,
			~SAVAGE_SCISSOR_MASK_S4);

	/* if any texture regs were changed ... */
	if (start <= SAVAGE_TEXDESCR_S4 &&
	    start + count > SAVAGE_TEXPALADDR_S4) {
		/* ... check texture state */
		SAVE_STATE(SAVAGE_TEXDESCR_S4, s4.texdescr);
		SAVE_STATE(SAVAGE_TEXADDR0_S4, s4.texaddr0);
		SAVE_STATE(SAVAGE_TEXADDR1_S4, s4.texaddr1);
		if (dev_priv->state.s4.texdescr & SAVAGE_TEXDESCR_TEX0EN_MASK)
			ret |= savage_verify_texaddr(dev_priv, 0,
						dev_priv->state.s4.texaddr0);
		if (dev_priv->state.s4.texdescr & SAVAGE_TEXDESCR_TEX1EN_MASK)
			ret |= savage_verify_texaddr(dev_priv, 1,
						dev_priv->state.s4.texaddr1);
	}

	return ret;
}

#undef SAVE_STATE
#undef SAVE_STATE_MASK

static int savage_dispatch_state(drm_savage_private_t * dev_priv,
				 const drm_savage_cmd_header_t * cmd_header,
				 const uint32_t *regs)
{
	unsigned int count = cmd_header->state.count;
	unsigned int start = cmd_header->state.start;
	unsigned int count2 = 0;
	unsigned int bci_size;
	int ret;
	DMA_LOCALS;

	if (!count)
		return 0;

	if (S3_SAVAGE3D_SERIES(dev_priv->chipset)) {
		ret = savage_verify_state_s3d(dev_priv, start, count, regs);
		if (ret != 0)
			return ret;
		/* scissor regs are emitted in savage_dispatch_draw */
		if (start < SAVAGE_SCSTART_S3D) {
			if (start + count > SAVAGE_SCEND_S3D + 1)
				count2 = count - (SAVAGE_SCEND_S3D + 1 - start);
			if (start + count > SAVAGE_SCSTART_S3D)
				count = SAVAGE_SCSTART_S3D - start;
		} else if (start <= SAVAGE_SCEND_S3D) {
			if (start + count > SAVAGE_SCEND_S3D + 1) {
				count -= SAVAGE_SCEND_S3D + 1 - start;
				start = SAVAGE_SCEND_S3D + 1;
			} else
				return 0;
		}
	} else {
		ret = savage_verify_state_s4(dev_priv, start, count, regs);
		if (ret != 0)
			return ret;
		/* scissor regs are emitted in savage_dispatch_draw */
		if (start < SAVAGE_DRAWCTRL0_S4) {
			if (start + count > SAVAGE_DRAWCTRL1_S4 + 1)
				count2 = count -
					 (SAVAGE_DRAWCTRL1_S4 + 1 - start);
			if (start + count > SAVAGE_DRAWCTRL0_S4)
				count = SAVAGE_DRAWCTRL0_S4 - start;
		} else if (start <= SAVAGE_DRAWCTRL1_S4) {
			if (start + count > SAVAGE_DRAWCTRL1_S4 + 1) {
				count -= SAVAGE_DRAWCTRL1_S4 + 1 - start;
				start = SAVAGE_DRAWCTRL1_S4 + 1;
			} else
				return 0;
		}
	}

	bci_size = count + (count + 254) / 255 + count2 + (count2 + 254) / 255;

	if (cmd_header->state.global) {
		BEGIN_DMA(bci_size + 1);
		DMA_WRITE(BCI_CMD_WAIT | BCI_CMD_WAIT_3D);
		dev_priv->waiting = 1;
	} else {
		BEGIN_DMA(bci_size);
	}

	do {
		while (count > 0) {
			unsigned int n = count < 255 ? count : 255;
			DMA_SET_REGISTERS(start, n);
			DMA_COPY(regs, n);
			count -= n;
			start += n;
			regs += n;
		}
		start += 2;
		regs += 2;
		count = count2;
		count2 = 0;
	} while (count);

	DMA_COMMIT();

	return 0;
}

static int savage_dispatch_dma_prim(drm_savage_private_t * dev_priv,
				    const drm_savage_cmd_header_t * cmd_header,
				    const drm_buf_t * dmabuf)
{
	unsigned char reorder = 0;
	unsigned int prim = cmd_header->prim.prim;
	unsigned int skip = cmd_header->prim.skip;
	unsigned int n = cmd_header->prim.count;
	unsigned int start = cmd_header->prim.start;
	unsigned int i;
	BCI_LOCALS;

	if (!dmabuf) {
		DRM_ERROR("called without dma buffers!\n");
		return DRM_ERR(EINVAL);
	}

	if (!n)
		return 0;

	switch (prim) {
	case SAVAGE_PRIM_TRILIST_201:
		reorder = 1;
		prim = SAVAGE_PRIM_TRILIST;
	case SAVAGE_PRIM_TRILIST:
		if (n % 3 != 0) {
			DRM_ERROR("wrong number of vertices %u in TRILIST\n",
				  n);
			return DRM_ERR(EINVAL);
		}
		break;
	case SAVAGE_PRIM_TRISTRIP:
	case SAVAGE_PRIM_TRIFAN:
		if (n < 3) {
			DRM_ERROR
			    ("wrong number of vertices %u in TRIFAN/STRIP\n",
			     n);
			return DRM_ERR(EINVAL);
		}
		break;
	default:
		DRM_ERROR("invalid primitive type %u\n", prim);
		return DRM_ERR(EINVAL);
	}

	if (S3_SAVAGE3D_SERIES(dev_priv->chipset)) {
		if (skip != 0) {
			DRM_ERROR("invalid skip flags 0x%04x for DMA\n", skip);
			return DRM_ERR(EINVAL);
		}
	} else {
		unsigned int size = 10 - (skip & 1) - (skip >> 1 & 1) -
		    (skip >> 2 & 1) - (skip >> 3 & 1) - (skip >> 4 & 1) -
		    (skip >> 5 & 1) - (skip >> 6 & 1) - (skip >> 7 & 1);
		if (skip > SAVAGE_SKIP_ALL_S4 || size != 8) {
			DRM_ERROR("invalid skip flags 0x%04x for DMA\n", skip);
			return DRM_ERR(EINVAL);
		}
		if (reorder) {
			DRM_ERROR("TRILIST_201 used on Savage4 hardware\n");
			return DRM_ERR(EINVAL);
		}
	}

	if (start + n > dmabuf->total / 32) {
		DRM_ERROR("vertex indices (%u-%u) out of range (0-%u)\n",
			  start, start + n - 1, dmabuf->total / 32);
		return DRM_ERR(EINVAL);
	}

	/* Vertex DMA doesn't work with command DMA at the same time,
	 * so we use BCI_... to submit commands here. Flush buffered
	 * faked DMA first. */
	DMA_FLUSH();

	if (dmabuf->bus_address != dev_priv->state.common.vbaddr) {
		BEGIN_BCI(2);
		BCI_SET_REGISTERS(SAVAGE_VERTBUFADDR, 1);
		BCI_WRITE(dmabuf->bus_address | dev_priv->dma_type);
		dev_priv->state.common.vbaddr = dmabuf->bus_address;
	}
	if (S3_SAVAGE3D_SERIES(dev_priv->chipset) && dev_priv->waiting) {
		/* Workaround for what looks like a hardware bug. If a
		 * WAIT_3D_IDLE was emitted some time before the
		 * indexed drawing command then the engine will lock
		 * up. There are two known workarounds:
		 * WAIT_IDLE_EMPTY or emit at least 63 NOPs. */
		BEGIN_BCI(63);
		for (i = 0; i < 63; ++i)
			BCI_WRITE(BCI_CMD_WAIT);
		dev_priv->waiting = 0;
	}

	prim <<= 25;
	while (n != 0) {
		/* Can emit up to 255 indices (85 triangles) at once. */
		unsigned int count = n > 255 ? 255 : n;
		if (reorder) {
			/* Need to reorder indices for correct flat
			 * shading while preserving the clock sense
			 * for correct culling. Only on Savage3D. */
			int reorder[3] = { -1, -1, -1 };
			reorder[start % 3] = 2;

			BEGIN_BCI((count + 1 + 1) / 2);
			BCI_DRAW_INDICES_S3D(count, prim, start + 2);

			for (i = start + 1; i + 1 < start + count; i += 2)
				BCI_WRITE((i + reorder[i % 3]) |
					  ((i + 1 +
					    reorder[(i + 1) % 3]) << 16));
			if (i < start + count)
				BCI_WRITE(i + reorder[i % 3]);
		} else if (S3_SAVAGE3D_SERIES(dev_priv->chipset)) {
			BEGIN_BCI((count + 1 + 1) / 2);
			BCI_DRAW_INDICES_S3D(count, prim, start);

			for (i = start + 1; i + 1 < start + count; i += 2)
				BCI_WRITE(i | ((i + 1) << 16));
			if (i < start + count)
				BCI_WRITE(i);
		} else {
			BEGIN_BCI((count + 2 + 1) / 2);
			BCI_DRAW_INDICES_S4(count, prim, skip);

			for (i = start; i + 1 < start + count; i += 2)
				BCI_WRITE(i | ((i + 1) << 16));
			if (i < start + count)
				BCI_WRITE(i);
		}

		start += count;
		n -= count;

		prim |= BCI_CMD_DRAW_CONT;
	}

	return 0;
}

static int savage_dispatch_vb_prim(drm_savage_private_t * dev_priv,
				   const drm_savage_cmd_header_t * cmd_header,
				   const uint32_t *vtxbuf, unsigned int vb_size,
				   unsigned int vb_stride)
{
	unsigned char reorder = 0;
	unsigned int prim = cmd_header->prim.prim;
	unsigned int skip = cmd_header->prim.skip;
	unsigned int n = cmd_header->prim.count;
	unsigned int start = cmd_header->prim.start;
	unsigned int vtx_size;
	unsigned int i;
	DMA_LOCALS;

	if (!n)
		return 0;

	switch (prim) {
	case SAVAGE_PRIM_TRILIST_201:
		reorder = 1;
		prim = SAVAGE_PRIM_TRILIST;
	case SAVAGE_PRIM_TRILIST:
		if (n % 3 != 0) {
			DRM_ERROR("wrong number of vertices %u in TRILIST\n",
				  n);
			return DRM_ERR(EINVAL);
		}
		break;
	case SAVAGE_PRIM_TRISTRIP:
	case SAVAGE_PRIM_TRIFAN:
		if (n < 3) {
			DRM_ERROR
			    ("wrong number of vertices %u in TRIFAN/STRIP\n",
			     n);
			return DRM_ERR(EINVAL);
		}
		break;
	default:
		DRM_ERROR("invalid primitive type %u\n", prim);
		return DRM_ERR(EINVAL);
	}

	if (S3_SAVAGE3D_SERIES(dev_priv->chipset)) {
		if (skip > SAVAGE_SKIP_ALL_S3D) {
			DRM_ERROR("invalid skip flags 0x%04x\n", skip);
			return DRM_ERR(EINVAL);
		}
		vtx_size = 8;	/* full vertex */
	} else {
		if (skip > SAVAGE_SKIP_ALL_S4) {
			DRM_ERROR("invalid skip flags 0x%04x\n", skip);
			return DRM_ERR(EINVAL);
		}
		vtx_size = 10;	/* full vertex */
	}

	vtx_size -= (skip & 1) + (skip >> 1 & 1) +
	    (skip >> 2 & 1) + (skip >> 3 & 1) + (skip >> 4 & 1) +
	    (skip >> 5 & 1) + (skip >> 6 & 1) + (skip >> 7 & 1);

	if (vtx_size > vb_stride) {
		DRM_ERROR("vertex size greater than vb stride (%u > %u)\n",
			  vtx_size, vb_stride);
		return DRM_ERR(EINVAL);
	}

	if (start + n > vb_size / (vb_stride * 4)) {
		DRM_ERROR("vertex indices (%u-%u) out of range (0-%u)\n",
			  start, start + n - 1, vb_size / (vb_stride * 4));
		return DRM_ERR(EINVAL);
	}

	prim <<= 25;
	while (n != 0) {
		/* Can emit up to 255 vertices (85 triangles) at once. */
		unsigned int count = n > 255 ? 255 : n;
		if (reorder) {
			/* Need to reorder vertices for correct flat
			 * shading while preserving the clock sense
			 * for correct culling. Only on Savage3D. */
			int reorder[3] = { -1, -1, -1 };
			reorder[start % 3] = 2;

			BEGIN_DMA(count * vtx_size + 1);
			DMA_DRAW_PRIMITIVE(count, prim, skip);

			for (i = start; i < start + count; ++i) {
				unsigned int j = i + reorder[i % 3];
				DMA_COPY(&vtxbuf[vb_stride * j], vtx_size);
			}

			DMA_COMMIT();
		} else {
			BEGIN_DMA(count * vtx_size + 1);
			DMA_DRAW_PRIMITIVE(count, prim, skip);

			if (vb_stride == vtx_size) {
				DMA_COPY(&vtxbuf[vb_stride * start], 
					 vtx_size * count);
			} else {
				for (i = start; i < start + count; ++i) {
					DMA_COPY(&vtxbuf [vb_stride * i],
						 vtx_size);
				}
			}

			DMA_COMMIT();
		}

		start += count;
		n -= count;

		prim |= BCI_CMD_DRAW_CONT;
	}

	return 0;
}

static int savage_dispatch_dma_idx(drm_savage_private_t * dev_priv,
				   const drm_savage_cmd_header_t * cmd_header,
				   const uint16_t *idx,
				   const drm_buf_t * dmabuf)
{
	unsigned char reorder = 0;
	unsigned int prim = cmd_header->idx.prim;
	unsigned int skip = cmd_header->idx.skip;
	unsigned int n = cmd_header->idx.count;
	unsigned int i;
	BCI_LOCALS;

	if (!dmabuf) {
		DRM_ERROR("called without dma buffers!\n");
		return DRM_ERR(EINVAL);
	}

	if (!n)
		return 0;

	switch (prim) {
	case SAVAGE_PRIM_TRILIST_201:
		reorder = 1;
		prim = SAVAGE_PRIM_TRILIST;
	case SAVAGE_PRIM_TRILIST:
		if (n % 3 != 0) {
			DRM_ERROR("wrong number of indices %u in TRILIST\n", n);
			return DRM_ERR(EINVAL);
		}
		break;
	case SAVAGE_PRIM_TRISTRIP:
	case SAVAGE_PRIM_TRIFAN:
		if (n < 3) {
			DRM_ERROR
			    ("wrong number of indices %u in TRIFAN/STRIP\n", n);
			return DRM_ERR(EINVAL);
		}
		break;
	default:
		DRM_ERROR("invalid primitive type %u\n", prim);
		return DRM_ERR(EINVAL);
	}

	if (S3_SAVAGE3D_SERIES(dev_priv->chipset)) {
		if (skip != 0) {
			DRM_ERROR("invalid skip flags 0x%04x for DMA\n", skip);
			return DRM_ERR(EINVAL);
		}
	} else {
		unsigned int size = 10 - (skip & 1) - (skip >> 1 & 1) -
		    (skip >> 2 & 1) - (skip >> 3 & 1) - (skip >> 4 & 1) -
		    (skip >> 5 & 1) - (skip >> 6 & 1) - (skip >> 7 & 1);
		if (skip > SAVAGE_SKIP_ALL_S4 || size != 8) {
			DRM_ERROR("invalid skip flags 0x%04x for DMA\n", skip);
			return DRM_ERR(EINVAL);
		}
		if (reorder) {
			DRM_ERROR("TRILIST_201 used on Savage4 hardware\n");
			return DRM_ERR(EINVAL);
		}
	}

	/* Vertex DMA doesn't work with command DMA at the same time,
	 * so we use BCI_... to submit commands here. Flush buffered
	 * faked DMA first. */
	DMA_FLUSH();

	if (dmabuf->bus_address != dev_priv->state.common.vbaddr) {
		BEGIN_BCI(2);
		BCI_SET_REGISTERS(SAVAGE_VERTBUFADDR, 1);
		BCI_WRITE(dmabuf->bus_address | dev_priv->dma_type);
		dev_priv->state.common.vbaddr = dmabuf->bus_address;
	}
	if (S3_SAVAGE3D_SERIES(dev_priv->chipset) && dev_priv->waiting) {
		/* Workaround for what looks like a hardware bug. If a
		 * WAIT_3D_IDLE was emitted some time before the
		 * indexed drawing command then the engine will lock
		 * up. There are two known workarounds:
		 * WAIT_IDLE_EMPTY or emit at least 63 NOPs. */
		BEGIN_BCI(63);
		for (i = 0; i < 63; ++i)
			BCI_WRITE(BCI_CMD_WAIT);
		dev_priv->waiting = 0;
	}

	prim <<= 25;
	while (n != 0) {
		/* Can emit up to 255 indices (85 triangles) at once. */
		unsigned int count = n > 255 ? 255 : n;

		/* check indices */
		for (i = 0; i < count; ++i) {
			if (idx[i] > dmabuf->total / 32) {
				DRM_ERROR("idx[%u]=%u out of range (0-%u)\n",
					  i, idx[i], dmabuf->total / 32);
				return DRM_ERR(EINVAL);
			}
		}

		if (reorder) {
			/* Need to reorder indices for correct flat
			 * shading while preserving the clock sense
			 * for correct culling. Only on Savage3D. */
			int reorder[3] = { 2, -1, -1 };

			BEGIN_BCI((count + 1 + 1) / 2);
			BCI_DRAW_INDICES_S3D(count, prim, idx[2]);

			for (i = 1; i + 1 < count; i += 2)
				BCI_WRITE(idx[i + reorder[i % 3]] |
					  (idx[i + 1 +
					   reorder[(i + 1) % 3]] << 16));
			if (i < count)
				BCI_WRITE(idx[i + reorder[i % 3]]);
		} else if (S3_SAVAGE3D_SERIES(dev_priv->chipset)) {
			BEGIN_BCI((count + 1 + 1) / 2);
			BCI_DRAW_INDICES_S3D(count, prim, idx[0]);

			for (i = 1; i + 1 < count; i += 2)
				BCI_WRITE(idx[i] | (idx[i + 1] << 16));
			if (i < count)
				BCI_WRITE(idx[i]);
		} else {
			BEGIN_BCI((count + 2 + 1) / 2);
			BCI_DRAW_INDICES_S4(count, prim, skip);

			for (i = 0; i + 1 < count; i += 2)
				BCI_WRITE(idx[i] | (idx[i + 1] << 16));
			if (i < count)
				BCI_WRITE(idx[i]);
		}

		idx += count;
		n -= count;

		prim |= BCI_CMD_DRAW_CONT;
	}

	return 0;
}

static int savage_dispatch_vb_idx(drm_savage_private_t * dev_priv,
				  const drm_savage_cmd_header_t * cmd_header,
				  const uint16_t *idx,
				  const uint32_t *vtxbuf,
				  unsigned int vb_size, unsigned int vb_stride)
{
	unsigned char reorder = 0;
	unsigned int prim = cmd_header->idx.prim;
	unsigned int skip = cmd_header->idx.skip;
	unsigned int n = cmd_header->idx.count;
	unsigned int vtx_size;
	unsigned int i;
	DMA_LOCALS;

	if (!n)
		return 0;

	switch (prim) {
	case SAVAGE_PRIM_TRILIST_201:
		reorder = 1;
		prim = SAVAGE_PRIM_TRILIST;
	case SAVAGE_PRIM_TRILIST:
		if (n % 3 != 0) {
			DRM_ERROR("wrong number of indices %u in TRILIST\n", n);
			return DRM_ERR(EINVAL);
		}
		break;
	case SAVAGE_PRIM_TRISTRIP:
	case SAVAGE_PRIM_TRIFAN:
		if (n < 3) {
			DRM_ERROR
			    ("wrong number of indices %u in TRIFAN/STRIP\n", n);
			return DRM_ERR(EINVAL);
		}
		break;
	default:
		DRM_ERROR("invalid primitive type %u\n", prim);
		return DRM_ERR(EINVAL);
	}

	if (S3_SAVAGE3D_SERIES(dev_priv->chipset)) {
		if (skip > SAVAGE_SKIP_ALL_S3D) {
			DRM_ERROR("invalid skip flags 0x%04x\n", skip);
			return DRM_ERR(EINVAL);
		}
		vtx_size = 8;	/* full vertex */
	} else {
		if (skip > SAVAGE_SKIP_ALL_S4) {
			DRM_ERROR("invalid skip flags 0x%04x\n", skip);
			return DRM_ERR(EINVAL);
		}
		vtx_size = 10;	/* full vertex */
	}

	vtx_size -= (skip & 1) + (skip >> 1 & 1) +
	    (skip >> 2 & 1) + (skip >> 3 & 1) + (skip >> 4 & 1) +
	    (skip >> 5 & 1) + (skip >> 6 & 1) + (skip >> 7 & 1);

	if (vtx_size > vb_stride) {
		DRM_ERROR("vertex size greater than vb stride (%u > %u)\n",
			  vtx_size, vb_stride);
		return DRM_ERR(EINVAL);
	}

	prim <<= 25;
	while (n != 0) {
		/* Can emit up to 255 vertices (85 triangles) at once. */
		unsigned int count = n > 255 ? 255 : n;
		
		/* Check indices */
		for (i = 0; i < count; ++i) {
			if (idx[i] > vb_size / (vb_stride * 4)) {
				DRM_ERROR("idx[%u]=%u out of range (0-%u)\n",
					  i, idx[i], vb_size / (vb_stride * 4));
				return DRM_ERR(EINVAL);
			}
		}

		if (reorder) {
			/* Need to reorder vertices for correct flat
			 * shading while preserving the clock sense
			 * for correct culling. Only on Savage3D. */
			int reorder[3] = { 2, -1, -1 };

			BEGIN_DMA(count * vtx_size + 1);
			DMA_DRAW_PRIMITIVE(count, prim, skip);

			for (i = 0; i < count; ++i) {
				unsigned int j = idx[i + reorder[i % 3]];
				DMA_COPY(&vtxbuf[vb_stride * j], vtx_size);
			}

			DMA_COMMIT();
		} else {
			BEGIN_DMA(count * vtx_size + 1);
			DMA_DRAW_PRIMITIVE(count, prim, skip);

			for (i = 0; i < count; ++i) {
				unsigned int j = idx[i];
				DMA_COPY(&vtxbuf[vb_stride * j], vtx_size);
			}

			DMA_COMMIT();
		}

		idx += count;
		n -= count;

		prim |= BCI_CMD_DRAW_CONT;
	}

	return 0;
}

static int savage_dispatch_clear(drm_savage_private_t * dev_priv,
				 const drm_savage_cmd_header_t * cmd_header,
				 const drm_savage_cmd_header_t *data,
				 unsigned int nbox,
				 const drm_clip_rect_t *boxes)
{
	unsigned int flags = cmd_header->clear0.flags;
	unsigned int clear_cmd;
	unsigned int i, nbufs;
	DMA_LOCALS;

	if (nbox == 0)
		return 0;

	clear_cmd = BCI_CMD_RECT | BCI_CMD_RECT_XP | BCI_CMD_RECT_YP |
	    BCI_CMD_SEND_COLOR | BCI_CMD_DEST_PBD_NEW;
	BCI_CMD_SET_ROP(clear_cmd, 0xCC);

	nbufs = ((flags & SAVAGE_FRONT) ? 1 : 0) +
	    ((flags & SAVAGE_BACK) ? 1 : 0) + ((flags & SAVAGE_DEPTH) ? 1 : 0);
	if (nbufs == 0)
		return 0;

	if (data->clear1.mask != 0xffffffff) {
		/* set mask */
		BEGIN_DMA(2);
		DMA_SET_REGISTERS(SAVAGE_BITPLANEWTMASK, 1);
		DMA_WRITE(data->clear1.mask);
		DMA_COMMIT();
	}
	for (i = 0; i < nbox; ++i) {
		unsigned int x, y, w, h;
		unsigned int buf;
		x = boxes[i].x1, y = boxes[i].y1;
		w = boxes[i].x2 - boxes[i].x1;
		h = boxes[i].y2 - boxes[i].y1;
		BEGIN_DMA(nbufs * 6);
		for (buf = SAVAGE_FRONT; buf <= SAVAGE_DEPTH; buf <<= 1) {
			if (!(flags & buf))
				continue;
			DMA_WRITE(clear_cmd);
			switch (buf) {
			case SAVAGE_FRONT:
				DMA_WRITE(dev_priv->front_offset);
				DMA_WRITE(dev_priv->front_bd);
				break;
			case SAVAGE_BACK:
				DMA_WRITE(dev_priv->back_offset);
				DMA_WRITE(dev_priv->back_bd);
				break;
			case SAVAGE_DEPTH:
				DMA_WRITE(dev_priv->depth_offset);
				DMA_WRITE(dev_priv->depth_bd);
				break;
			}
			DMA_WRITE(data->clear1.value);
			DMA_WRITE(BCI_X_Y(x, y));
			DMA_WRITE(BCI_W_H(w, h));
		}
		DMA_COMMIT();
	}
	if (data->clear1.mask != 0xffffffff) {
		/* reset mask */
		BEGIN_DMA(2);
		DMA_SET_REGISTERS(SAVAGE_BITPLANEWTMASK, 1);
		DMA_WRITE(0xffffffff);
		DMA_COMMIT();
	}

	return 0;
}

static int savage_dispatch_swap(drm_savage_private_t * dev_priv,
				unsigned int nbox, const drm_clip_rect_t *boxes)
{
	unsigned int swap_cmd;
	unsigned int i;
	DMA_LOCALS;

	if (nbox == 0)
		return 0;

	swap_cmd = BCI_CMD_RECT | BCI_CMD_RECT_XP | BCI_CMD_RECT_YP |
	    BCI_CMD_SRC_PBD_COLOR_NEW | BCI_CMD_DEST_GBD;
	BCI_CMD_SET_ROP(swap_cmd, 0xCC);

	for (i = 0; i < nbox; ++i) {
		BEGIN_DMA(6);
		DMA_WRITE(swap_cmd);
		DMA_WRITE(dev_priv->back_offset);
		DMA_WRITE(dev_priv->back_bd);
		DMA_WRITE(BCI_X_Y(boxes[i].x1, boxes[i].y1));
		DMA_WRITE(BCI_X_Y(boxes[i].x1, boxes[i].y1));
		DMA_WRITE(BCI_W_H(boxes[i].x2 - boxes[i].x1,
				  boxes[i].y2 - boxes[i].y1));
		DMA_COMMIT();
	}

	return 0;
}

static int savage_dispatch_draw(drm_savage_private_t * dev_priv,
				const drm_savage_cmd_header_t *start,
				const drm_savage_cmd_header_t *end,
				const drm_buf_t * dmabuf,
				const unsigned int *vtxbuf,
				unsigned int vb_size, unsigned int vb_stride,
				unsigned int nbox,
				const drm_clip_rect_t *boxes)
{
	unsigned int i, j;
	int ret;

	for (i = 0; i < nbox; ++i) {
		const drm_savage_cmd_header_t *cmdbuf;
		dev_priv->emit_clip_rect(dev_priv, &boxes[i]);

		cmdbuf = start;
		while (cmdbuf < end) {
			drm_savage_cmd_header_t cmd_header;
			cmd_header = *cmdbuf;
			cmdbuf++;
			switch (cmd_header.cmd.cmd) {
			case SAVAGE_CMD_DMA_PRIM:
				ret = savage_dispatch_dma_prim(
					dev_priv, &cmd_header, dmabuf);
				break;
			case SAVAGE_CMD_VB_PRIM:
				ret = savage_dispatch_vb_prim(
					dev_priv, &cmd_header,
					vtxbuf, vb_size, vb_stride);
				break;
			case SAVAGE_CMD_DMA_IDX:
				j = (cmd_header.idx.count + 3) / 4;
				/* j was check in savage_bci_cmdbuf */
				ret = savage_dispatch_dma_idx(dev_priv,
					&cmd_header, (const uint16_t *)cmdbuf,
					dmabuf);
				cmdbuf += j;
				break;
			case SAVAGE_CMD_VB_IDX:
				j = (cmd_header.idx.count + 3) / 4;
				/* j was check in savage_bci_cmdbuf */
				ret = savage_dispatch_vb_idx(dev_priv,
					&cmd_header, (const uint16_t *)cmdbuf,
					(const uint32_t *)vtxbuf, vb_size, 
					vb_stride);
				cmdbuf += j;
				break;
			default:
				/* What's the best return code? EFAULT? */
				DRM_ERROR("IMPLEMENTATION ERROR: "
					  "non-drawing-command %d\n",
					  cmd_header.cmd.cmd);
				return DRM_ERR(EINVAL);
			}

			if (ret != 0)
				return ret;
		}
	}

	return 0;
}

int savage_bci_cmdbuf(DRM_IOCTL_ARGS)
{
	DRM_DEVICE;
	drm_savage_private_t *dev_priv = dev->dev_private;
	drm_device_dma_t *dma = dev->dma;
	drm_buf_t *dmabuf;
	drm_savage_cmdbuf_t cmdbuf;
	drm_savage_cmd_header_t *kcmd_addr = NULL;
	drm_savage_cmd_header_t *first_draw_cmd;
	unsigned int *kvb_addr = NULL;
	drm_clip_rect_t *kbox_addr = NULL;
	unsigned int i, j;
	int ret = 0;

	DRM_DEBUG("\n");

	LOCK_TEST_WITH_RETURN(dev, filp);

	DRM_COPY_FROM_USER_IOCTL(cmdbuf, (drm_savage_cmdbuf_t __user *) data,
				 sizeof(cmdbuf));

	if (dma && dma->buflist) {
		if (cmdbuf.dma_idx > dma->buf_count) {
			DRM_ERROR
			    ("vertex buffer index %u out of range (0-%u)\n",
			     cmdbuf.dma_idx, dma->buf_count - 1);
			return DRM_ERR(EINVAL);
		}
		dmabuf = dma->buflist[cmdbuf.dma_idx];
	} else {
		dmabuf = NULL;
	}

	/* Copy the user buffers into kernel temporary areas.  This hasn't been
	 * a performance loss compared to VERIFYAREA_READ/
	 * COPY_FROM_USER_UNCHECKED when done in other drivers, and is correct
	 * for locking on FreeBSD.
	 */
	if (cmdbuf.size) {
		kcmd_addr = drm_alloc(cmdbuf.size * 8, DRM_MEM_DRIVER);
		if (kcmd_addr == NULL)
			return DRM_ERR(ENOMEM);

		if (DRM_COPY_FROM_USER(kcmd_addr, cmdbuf.cmd_addr,
				       cmdbuf.size * 8))
		{
			drm_free(kcmd_addr, cmdbuf.size * 8, DRM_MEM_DRIVER);
			return DRM_ERR(EFAULT);
		}
		cmdbuf.cmd_addr = kcmd_addr;
	}
	if (cmdbuf.vb_size) {
		kvb_addr = drm_alloc(cmdbuf.vb_size, DRM_MEM_DRIVER);
		if (kvb_addr == NULL) {
			ret = DRM_ERR(ENOMEM);
			goto done;
		}

		if (DRM_COPY_FROM_USER(kvb_addr, cmdbuf.vb_addr,
				       cmdbuf.vb_size)) {
			ret = DRM_ERR(EFAULT);
			goto done;
		}
		cmdbuf.vb_addr = kvb_addr;
	}
	if (cmdbuf.nbox) {
		kbox_addr = drm_alloc(cmdbuf.nbox * sizeof(drm_clip_rect_t),
				       DRM_MEM_DRIVER);
		if (kbox_addr == NULL) {
			ret = DRM_ERR(ENOMEM);
			goto done;
		}

		if (DRM_COPY_FROM_USER(kbox_addr, cmdbuf.box_addr,
				       cmdbuf.nbox * sizeof(drm_clip_rect_t))) {
			ret = DRM_ERR(EFAULT);
			goto done;
		}
	cmdbuf.box_addr = kbox_addr;
	}

	/* Make sure writes to DMA buffers are finished before sending
	 * DMA commands to the graphics hardware. */
	DRM_MEMORYBARRIER();

	/* Coming from user space. Don't know if the Xserver has
	 * emitted wait commands. Assuming the worst. */
	dev_priv->waiting = 1;

	i = 0;
	first_draw_cmd = NULL;
	while (i < cmdbuf.size) {
		drm_savage_cmd_header_t cmd_header;
		cmd_header = *(drm_savage_cmd_header_t *)cmdbuf.cmd_addr;
		cmdbuf.cmd_addr++;
		i++;

		/* Group drawing commands with same state to minimize
		 * iterations over clip rects. */
		j = 0;
		switch (cmd_header.cmd.cmd) {
		case SAVAGE_CMD_DMA_IDX:
		case SAVAGE_CMD_VB_IDX:
			j = (cmd_header.idx.count + 3) / 4;
			if (i + j > cmdbuf.size) {
				DRM_ERROR("indexed drawing command extends "
					  "beyond end of command buffer\n");
				DMA_FLUSH();
				return DRM_ERR(EINVAL);
			}
			/* fall through */
		case SAVAGE_CMD_DMA_PRIM:
		case SAVAGE_CMD_VB_PRIM:
			if (!first_draw_cmd)
				first_draw_cmd = cmdbuf.cmd_addr - 1;
			cmdbuf.cmd_addr += j;
			i += j;
			break;
		default:
			if (first_draw_cmd) {
				ret = savage_dispatch_draw(
				      dev_priv, first_draw_cmd,
				      cmdbuf.cmd_addr - 1,
				      dmabuf, cmdbuf.vb_addr, cmdbuf.vb_size,
				      cmdbuf.vb_stride,
				      cmdbuf.nbox, cmdbuf.box_addr);
				if (ret != 0)
					return ret;
				first_draw_cmd = NULL;
			}
		}
		if (first_draw_cmd)
			continue;

		switch (cmd_header.cmd.cmd) {
		case SAVAGE_CMD_STATE:
			j = (cmd_header.state.count + 1) / 2;
			if (i + j > cmdbuf.size) {
				DRM_ERROR("command SAVAGE_CMD_STATE extends "
					  "beyond end of command buffer\n");
				DMA_FLUSH();
				ret = DRM_ERR(EINVAL);
				goto done;
			}
			ret = savage_dispatch_state(dev_priv, &cmd_header,
				(const uint32_t *)cmdbuf.cmd_addr);
			cmdbuf.cmd_addr += j;
			i += j;
			break;
		case SAVAGE_CMD_CLEAR:
			if (i + 1 > cmdbuf.size) {
				DRM_ERROR("command SAVAGE_CMD_CLEAR extends "
					  "beyond end of command buffer\n");
				DMA_FLUSH();
				ret = DRM_ERR(EINVAL);
				goto done;
			}
			ret = savage_dispatch_clear(dev_priv, &cmd_header,
						    cmdbuf.cmd_addr,
						    cmdbuf.nbox, cmdbuf.box_addr);
			cmdbuf.cmd_addr++;
			i++;
			break;
		case SAVAGE_CMD_SWAP:
			ret = savage_dispatch_swap(dev_priv, cmdbuf.nbox,
						   cmdbuf.box_addr);
			break;
		default:
			DRM_ERROR("invalid command 0x%x\n", cmd_header.cmd.cmd);
			DMA_FLUSH();
			ret = DRM_ERR(EINVAL);
			goto done;
		}

		if (ret != 0) {
			DMA_FLUSH();
			goto done;
		}
	}

	if (first_draw_cmd) {
		ret = savage_dispatch_draw (
			dev_priv, first_draw_cmd, cmdbuf.cmd_addr, dmabuf,
			cmdbuf.vb_addr, cmdbuf.vb_size, cmdbuf.vb_stride,
			cmdbuf.nbox, cmdbuf.box_addr);
		if (ret != 0) {
			DMA_FLUSH();
			goto done;
		}
	}

	DMA_FLUSH();

	if (dmabuf && cmdbuf.discard) {
		drm_savage_buf_priv_t *buf_priv = dmabuf->dev_private;
		uint16_t event;
		event = savage_bci_emit_event(dev_priv, SAVAGE_WAIT_3D);
		SET_AGE(&buf_priv->age, event, dev_priv->event_wrap);
		savage_freelist_put(dev, dmabuf);
	}

done:
	/* If we didn't need to allocate them, these'll be NULL */
	drm_free(kcmd_addr, cmdbuf.size * 8, DRM_MEM_DRIVER);
	drm_free(kvb_addr, cmdbuf.vb_size, DRM_MEM_DRIVER);
	drm_free(kbox_addr, cmdbuf.nbox * sizeof(drm_clip_rect_t),
		 DRM_MEM_DRIVER);

	return ret;
}