summaryrefslogtreecommitdiffstats
path: root/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
blob: e92bdc9a39d353ac5dfac9d51aefd32e4740377e (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
/*
 * Copyright 2014 Advanced Micro Devices, Inc.
 * Copyright 2008 Red Hat Inc.
 * Copyright 2009 Jerome Glisse.
 *
 * 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, sublicense,
 * 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 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 NONINFRINGEMENT.  IN NO EVENT SHALL
 * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) 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 <linux/firmware.h>
#include "amdgpu.h"
#include "amdgpu_gfx.h"
#include "amdgpu_rlc.h"
#include "amdgpu_ras.h"
#include "amdgpu_xcp.h"
#include "amdgpu_xgmi.h"

/* delay 0.1 second to enable gfx off feature */
#define GFX_OFF_DELAY_ENABLE         msecs_to_jiffies(100)

#define GFX_OFF_NO_DELAY 0

/*
 * GPU GFX IP block helpers function.
 */

int amdgpu_gfx_mec_queue_to_bit(struct amdgpu_device *adev, int mec,
				int pipe, int queue)
{
	int bit = 0;

	bit += mec * adev->gfx.mec.num_pipe_per_mec
		* adev->gfx.mec.num_queue_per_pipe;
	bit += pipe * adev->gfx.mec.num_queue_per_pipe;
	bit += queue;

	return bit;
}

void amdgpu_queue_mask_bit_to_mec_queue(struct amdgpu_device *adev, int bit,
				 int *mec, int *pipe, int *queue)
{
	*queue = bit % adev->gfx.mec.num_queue_per_pipe;
	*pipe = (bit / adev->gfx.mec.num_queue_per_pipe)
		% adev->gfx.mec.num_pipe_per_mec;
	*mec = (bit / adev->gfx.mec.num_queue_per_pipe)
	       / adev->gfx.mec.num_pipe_per_mec;

}

bool amdgpu_gfx_is_mec_queue_enabled(struct amdgpu_device *adev,
				     int xcc_id, int mec, int pipe, int queue)
{
	return test_bit(amdgpu_gfx_mec_queue_to_bit(adev, mec, pipe, queue),
			adev->gfx.mec_bitmap[xcc_id].queue_bitmap);
}

int amdgpu_gfx_me_queue_to_bit(struct amdgpu_device *adev,
			       int me, int pipe, int queue)
{
	int bit = 0;

	bit += me * adev->gfx.me.num_pipe_per_me
		* adev->gfx.me.num_queue_per_pipe;
	bit += pipe * adev->gfx.me.num_queue_per_pipe;
	bit += queue;

	return bit;
}

void amdgpu_gfx_bit_to_me_queue(struct amdgpu_device *adev, int bit,
				int *me, int *pipe, int *queue)
{
	*queue = bit % adev->gfx.me.num_queue_per_pipe;
	*pipe = (bit / adev->gfx.me.num_queue_per_pipe)
		% adev->gfx.me.num_pipe_per_me;
	*me = (bit / adev->gfx.me.num_queue_per_pipe)
		/ adev->gfx.me.num_pipe_per_me;
}

bool amdgpu_gfx_is_me_queue_enabled(struct amdgpu_device *adev,
				    int me, int pipe, int queue)
{
	return test_bit(amdgpu_gfx_me_queue_to_bit(adev, me, pipe, queue),
			adev->gfx.me.queue_bitmap);
}

/**
 * amdgpu_gfx_parse_disable_cu - Parse the disable_cu module parameter
 *
 * @mask: array in which the per-shader array disable masks will be stored
 * @max_se: number of SEs
 * @max_sh: number of SHs
 *
 * The bitmask of CUs to be disabled in the shader array determined by se and
 * sh is stored in mask[se * max_sh + sh].
 */
void amdgpu_gfx_parse_disable_cu(unsigned int *mask, unsigned int max_se, unsigned int max_sh)
{
	unsigned int se, sh, cu;
	const char *p;

	memset(mask, 0, sizeof(*mask) * max_se * max_sh);

	if (!amdgpu_disable_cu || !*amdgpu_disable_cu)
		return;

	p = amdgpu_disable_cu;
	for (;;) {
		char *next;
		int ret = sscanf(p, "%u.%u.%u", &se, &sh, &cu);

		if (ret < 3) {
			DRM_ERROR("amdgpu: could not parse disable_cu\n");
			return;
		}

		if (se < max_se && sh < max_sh && cu < 16) {
			DRM_INFO("amdgpu: disabling CU %u.%u.%u\n", se, sh, cu);
			mask[se * max_sh + sh] |= 1u << cu;
		} else {
			DRM_ERROR("amdgpu: disable_cu %u.%u.%u is out of range\n",
				  se, sh, cu);
		}

		next = strchr(p, ',');
		if (!next)
			break;
		p = next + 1;
	}
}

static bool amdgpu_gfx_is_graphics_multipipe_capable(struct amdgpu_device *adev)
{
	return amdgpu_async_gfx_ring && adev->gfx.me.num_pipe_per_me > 1;
}

static bool amdgpu_gfx_is_compute_multipipe_capable(struct amdgpu_device *adev)
{
	if (amdgpu_compute_multipipe != -1) {
		DRM_INFO("amdgpu: forcing compute pipe policy %d\n",
			 amdgpu_compute_multipipe);
		return amdgpu_compute_multipipe == 1;
	}

	if (amdgpu_ip_version(adev, GC_HWIP, 0) > IP_VERSION(9, 0, 0))
		return true;

	/* FIXME: spreading the queues across pipes causes perf regressions
	 * on POLARIS11 compute workloads */
	if (adev->asic_type == CHIP_POLARIS11)
		return false;

	return adev->gfx.mec.num_mec > 1;
}

bool amdgpu_gfx_is_high_priority_graphics_queue(struct amdgpu_device *adev,
						struct amdgpu_ring *ring)
{
	int queue = ring->queue;
	int pipe = ring->pipe;

	/* Policy: use pipe1 queue0 as high priority graphics queue if we
	 * have more than one gfx pipe.
	 */
	if (amdgpu_gfx_is_graphics_multipipe_capable(adev) &&
	    adev->gfx.num_gfx_rings > 1 && pipe == 1 && queue == 0) {
		int me = ring->me;
		int bit;

		bit = amdgpu_gfx_me_queue_to_bit(adev, me, pipe, queue);
		if (ring == &adev->gfx.gfx_ring[bit])
			return true;
	}

	return false;
}

bool amdgpu_gfx_is_high_priority_compute_queue(struct amdgpu_device *adev,
					       struct amdgpu_ring *ring)
{
	/* Policy: use 1st queue as high priority compute queue if we
	 * have more than one compute queue.
	 */
	if (adev->gfx.num_compute_rings > 1 &&
	    ring == &adev->gfx.compute_ring[0])
		return true;

	return false;
}

void amdgpu_gfx_compute_queue_acquire(struct amdgpu_device *adev)
{
	int i, j, queue, pipe;
	bool multipipe_policy = amdgpu_gfx_is_compute_multipipe_capable(adev);
	int max_queues_per_mec = min(adev->gfx.mec.num_pipe_per_mec *
				     adev->gfx.mec.num_queue_per_pipe,
				     adev->gfx.num_compute_rings);
	int num_xcc = adev->gfx.xcc_mask ? NUM_XCC(adev->gfx.xcc_mask) : 1;

	if (multipipe_policy) {
		/* policy: make queues evenly cross all pipes on MEC1 only
		 * for multiple xcc, just use the original policy for simplicity */
		for (j = 0; j < num_xcc; j++) {
			for (i = 0; i < max_queues_per_mec; i++) {
				pipe = i % adev->gfx.mec.num_pipe_per_mec;
				queue = (i / adev->gfx.mec.num_pipe_per_mec) %
					 adev->gfx.mec.num_queue_per_pipe;

				set_bit(pipe * adev->gfx.mec.num_queue_per_pipe + queue,
					adev->gfx.mec_bitmap[j].queue_bitmap);
			}
		}
	} else {
		/* policy: amdgpu owns all queues in the given pipe */
		for (j = 0; j < num_xcc; j++) {
			for (i = 0; i < max_queues_per_mec; ++i)
				set_bit(i, adev->gfx.mec_bitmap[j].queue_bitmap);
		}
	}

	for (j = 0; j < num_xcc; j++) {
		dev_dbg(adev->dev, "mec queue bitmap weight=%d\n",
			bitmap_weight(adev->gfx.mec_bitmap[j].queue_bitmap, AMDGPU_MAX_COMPUTE_QUEUES));
	}
}

void amdgpu_gfx_graphics_queue_acquire(struct amdgpu_device *adev)
{
	int i, queue, pipe;
	bool multipipe_policy = amdgpu_gfx_is_graphics_multipipe_capable(adev);
	int max_queues_per_me = adev->gfx.me.num_pipe_per_me *
					adev->gfx.me.num_queue_per_pipe;

	if (multipipe_policy) {
		/* policy: amdgpu owns the first queue per pipe at this stage
		 * will extend to mulitple queues per pipe later */
		for (i = 0; i < max_queues_per_me; i++) {
			pipe = i % adev->gfx.me.num_pipe_per_me;
			queue = (i / adev->gfx.me.num_pipe_per_me) %
				adev->gfx.me.num_queue_per_pipe;

			set_bit(pipe * adev->gfx.me.num_queue_per_pipe + queue,
				adev->gfx.me.queue_bitmap);
		}
	} else {
		for (i = 0; i < max_queues_per_me; ++i)
			set_bit(i, adev->gfx.me.queue_bitmap);
	}

	/* update the number of active graphics rings */
	adev->gfx.num_gfx_rings =
		bitmap_weight(adev->gfx.me.queue_bitmap, AMDGPU_MAX_GFX_QUEUES);
}

static int amdgpu_gfx_kiq_acquire(struct amdgpu_device *adev,
				  struct amdgpu_ring *ring, int xcc_id)
{
	int queue_bit;
	int mec, pipe, queue;

	queue_bit = adev->gfx.mec.num_mec
		    * adev->gfx.mec.num_pipe_per_mec
		    * adev->gfx.mec.num_queue_per_pipe;

	while (--queue_bit >= 0) {
		if (test_bit(queue_bit, adev->gfx.mec_bitmap[xcc_id].queue_bitmap))
			continue;

		amdgpu_queue_mask_bit_to_mec_queue(adev, queue_bit, &mec, &pipe, &queue);

		/*
		 * 1. Using pipes 2/3 from MEC 2 seems cause problems.
		 * 2. It must use queue id 0, because CGPG_IDLE/SAVE/LOAD/RUN
		 * only can be issued on queue 0.
		 */
		if ((mec == 1 && pipe > 1) || queue != 0)
			continue;

		ring->me = mec + 1;
		ring->pipe = pipe;
		ring->queue = queue;

		return 0;
	}

	dev_err(adev->dev, "Failed to find a queue for KIQ\n");
	return -EINVAL;
}

int amdgpu_gfx_kiq_init_ring(struct amdgpu_device *adev, int xcc_id)
{
	struct amdgpu_kiq *kiq = &adev->gfx.kiq[xcc_id];
	struct amdgpu_irq_src *irq = &kiq->irq;
	struct amdgpu_ring *ring = &kiq->ring;
	int r = 0;

	spin_lock_init(&kiq->ring_lock);

	ring->adev = NULL;
	ring->ring_obj = NULL;
	ring->use_doorbell = true;
	ring->xcc_id = xcc_id;
	ring->vm_hub = AMDGPU_GFXHUB(xcc_id);
	ring->doorbell_index =
		(adev->doorbell_index.kiq +
		 xcc_id * adev->doorbell_index.xcc_doorbell_range)
		<< 1;

	r = amdgpu_gfx_kiq_acquire(adev, ring, xcc_id);
	if (r)
		return r;

	ring->eop_gpu_addr = kiq->eop_gpu_addr;
	ring->no_scheduler = true;
	snprintf(ring->name, sizeof(ring->name), "kiq_%hhu.%hhu.%hhu.%hhu",
		 (unsigned char)xcc_id, (unsigned char)ring->me,
		 (unsigned char)ring->pipe, (unsigned char)ring->queue);
	r = amdgpu_ring_init(adev, ring, 1024, irq, AMDGPU_CP_KIQ_IRQ_DRIVER0,
			     AMDGPU_RING_PRIO_DEFAULT, NULL);
	if (r)
		dev_warn(adev->dev, "(%d) failed to init kiq ring\n", r);

	return r;
}

void amdgpu_gfx_kiq_free_ring(struct amdgpu_ring *ring)
{
	amdgpu_ring_fini(ring);
}

void amdgpu_gfx_kiq_fini(struct amdgpu_device *adev, int xcc_id)
{
	struct amdgpu_kiq *kiq = &adev->gfx.kiq[xcc_id];

	amdgpu_bo_free_kernel(&kiq->eop_obj, &kiq->eop_gpu_addr, NULL);
}

int amdgpu_gfx_kiq_init(struct amdgpu_device *adev,
			unsigned int hpd_size, int xcc_id)
{
	int r;
	u32 *hpd;
	struct amdgpu_kiq *kiq = &adev->gfx.kiq[xcc_id];

	r = amdgpu_bo_create_kernel(adev, hpd_size, PAGE_SIZE,
				    AMDGPU_GEM_DOMAIN_GTT, &kiq->eop_obj,
				    &kiq->eop_gpu_addr, (void **)&hpd);
	if (r) {
		dev_warn(adev->dev, "failed to create KIQ bo (%d).\n", r);
		return r;
	}

	memset(hpd, 0, hpd_size);

	r = amdgpu_bo_reserve(kiq->eop_obj, true);
	if (unlikely(r != 0))
		dev_warn(adev->dev, "(%d) reserve kiq eop bo failed\n", r);
	amdgpu_bo_kunmap(kiq->eop_obj);
	amdgpu_bo_unreserve(kiq->eop_obj);

	return 0;
}

/* create MQD for each compute/gfx queue */
int amdgpu_gfx_mqd_sw_init(struct amdgpu_device *adev,
			   unsigned int mqd_size, int xcc_id)
{
	int r, i, j;
	struct amdgpu_kiq *kiq = &adev->gfx.kiq[xcc_id];
	struct amdgpu_ring *ring = &kiq->ring;
	u32 domain = AMDGPU_GEM_DOMAIN_GTT;

#if !defined(CONFIG_ARM) && !defined(CONFIG_ARM64)
	/* Only enable on gfx10 and 11 for now to avoid changing behavior on older chips */
	if (amdgpu_ip_version(adev, GC_HWIP, 0) >= IP_VERSION(10, 0, 0))
		domain |= AMDGPU_GEM_DOMAIN_VRAM;
#endif

	/* create MQD for KIQ */
	if (!adev->enable_mes_kiq && !ring->mqd_obj) {
		/* originaly the KIQ MQD is put in GTT domain, but for SRIOV VRAM domain is a must
		 * otherwise hypervisor trigger SAVE_VF fail after driver unloaded which mean MQD
		 * deallocated and gart_unbind, to strict diverage we decide to use VRAM domain for
		 * KIQ MQD no matter SRIOV or Bare-metal
		 */
		r = amdgpu_bo_create_kernel(adev, mqd_size, PAGE_SIZE,
					    AMDGPU_GEM_DOMAIN_VRAM |
					    AMDGPU_GEM_DOMAIN_GTT,
					    &ring->mqd_obj,
					    &ring->mqd_gpu_addr,
					    &ring->mqd_ptr);
		if (r) {
			dev_warn(adev->dev, "failed to create ring mqd ob (%d)", r);
			return r;
		}

		/* prepare MQD backup */
		kiq->mqd_backup = kmalloc(mqd_size, GFP_KERNEL);
		if (!kiq->mqd_backup) {
			dev_warn(adev->dev,
				 "no memory to create MQD backup for ring %s\n", ring->name);
			return -ENOMEM;
		}
	}

	if (adev->asic_type >= CHIP_NAVI10 && amdgpu_async_gfx_ring) {
		/* create MQD for each KGQ */
		for (i = 0; i < adev->gfx.num_gfx_rings; i++) {
			ring = &adev->gfx.gfx_ring[i];
			if (!ring->mqd_obj) {
				r = amdgpu_bo_create_kernel(adev, mqd_size, PAGE_SIZE,
							    domain, &ring->mqd_obj,
							    &ring->mqd_gpu_addr, &ring->mqd_ptr);
				if (r) {
					dev_warn(adev->dev, "failed to create ring mqd bo (%d)", r);
					return r;
				}

				ring->mqd_size = mqd_size;
				/* prepare MQD backup */
				adev->gfx.me.mqd_backup[i] = kmalloc(mqd_size, GFP_KERNEL);
				if (!adev->gfx.me.mqd_backup[i]) {
					dev_warn(adev->dev, "no memory to create MQD backup for ring %s\n", ring->name);
					return -ENOMEM;
				}
			}
		}
	}

	/* create MQD for each KCQ */
	for (i = 0; i < adev->gfx.num_compute_rings; i++) {
		j = i + xcc_id * adev->gfx.num_compute_rings;
		ring = &adev->gfx.compute_ring[j];
		if (!ring->mqd_obj) {
			r = amdgpu_bo_create_kernel(adev, mqd_size, PAGE_SIZE,
						    domain, &ring->mqd_obj,
						    &ring->mqd_gpu_addr, &ring->mqd_ptr);
			if (r) {
				dev_warn(adev->dev, "failed to create ring mqd bo (%d)", r);
				return r;
			}

			ring->mqd_size = mqd_size;
			/* prepare MQD backup */
			adev->gfx.mec.mqd_backup[j] = kmalloc(mqd_size, GFP_KERNEL);
			if (!adev->gfx.mec.mqd_backup[j]) {
				dev_warn(adev->dev, "no memory to create MQD backup for ring %s\n", ring->name);
				return -ENOMEM;
			}
		}
	}

	return 0;
}

void amdgpu_gfx_mqd_sw_fini(struct amdgpu_device *adev, int xcc_id)
{
	struct amdgpu_ring *ring = NULL;
	int i, j;
	struct amdgpu_kiq *kiq = &adev->gfx.kiq[xcc_id];

	if (adev->asic_type >= CHIP_NAVI10 && amdgpu_async_gfx_ring) {
		for (i = 0; i < adev->gfx.num_gfx_rings; i++) {
			ring = &adev->gfx.gfx_ring[i];
			kfree(adev->gfx.me.mqd_backup[i]);
			amdgpu_bo_free_kernel(&ring->mqd_obj,
					      &ring->mqd_gpu_addr,
					      &ring->mqd_ptr);
		}
	}

	for (i = 0; i < adev->gfx.num_compute_rings; i++) {
		j = i + xcc_id * adev->gfx.num_compute_rings;
		ring = &adev->gfx.compute_ring[j];
		kfree(adev->gfx.mec.mqd_backup[j]);
		amdgpu_bo_free_kernel(&ring->mqd_obj,
				      &ring->mqd_gpu_addr,
				      &ring->mqd_ptr);
	}

	ring = &kiq->ring;
	kfree(kiq->mqd_backup);
	amdgpu_bo_free_kernel(&ring->mqd_obj,
			      &ring->mqd_gpu_addr,
			      &ring->mqd_ptr);
}

int amdgpu_gfx_disable_kcq(struct amdgpu_device *adev, int xcc_id)
{
	struct amdgpu_kiq *kiq = &adev->gfx.kiq[xcc_id];
	struct amdgpu_ring *kiq_ring = &kiq->ring;
	struct amdgpu_hive_info *hive;
	struct amdgpu_ras *ras;
	int hive_ras_recovery = 0;
	int i, r = 0;
	int j;

	if (!kiq->pmf || !kiq->pmf->kiq_unmap_queues)
		return -EINVAL;

	spin_lock(&kiq->ring_lock);
	if (amdgpu_ring_alloc(kiq_ring, kiq->pmf->unmap_queues_size *
					adev->gfx.num_compute_rings)) {
		spin_unlock(&kiq->ring_lock);
		return -ENOMEM;
	}

	for (i = 0; i < adev->gfx.num_compute_rings; i++) {
		j = i + xcc_id * adev->gfx.num_compute_rings;
		kiq->pmf->kiq_unmap_queues(kiq_ring,
					   &adev->gfx.compute_ring[j],
					   RESET_QUEUES, 0, 0);
	}

	/**
	 * This is workaround: only skip kiq_ring test
	 * during ras recovery in suspend stage for gfx9.4.3
	 */
	hive = amdgpu_get_xgmi_hive(adev);
	if (hive) {
		hive_ras_recovery = atomic_read(&hive->ras_recovery);
		amdgpu_put_xgmi_hive(hive);
	}

	ras = amdgpu_ras_get_context(adev);
	if ((amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 3)) &&
		ras && (atomic_read(&ras->in_recovery) || hive_ras_recovery)) {
		spin_unlock(&kiq->ring_lock);
		return 0;
	}

	if (kiq_ring->sched.ready && !adev->job_hang)
		r = amdgpu_ring_test_helper(kiq_ring);
	spin_unlock(&kiq->ring_lock);

	return r;
}

int amdgpu_gfx_disable_kgq(struct amdgpu_device *adev, int xcc_id)
{
	struct amdgpu_kiq *kiq = &adev->gfx.kiq[xcc_id];
	struct amdgpu_ring *kiq_ring = &kiq->ring;
	int i, r = 0;
	int j;

	if (!kiq->pmf || !kiq->pmf->kiq_unmap_queues)
		return -EINVAL;

	spin_lock(&kiq->ring_lock);
	if (amdgpu_gfx_is_master_xcc(adev, xcc_id)) {
		if (amdgpu_ring_alloc(kiq_ring, kiq->pmf->unmap_queues_size *
						adev->gfx.num_gfx_rings)) {
			spin_unlock(&kiq->ring_lock);
			return -ENOMEM;
		}

		for (i = 0; i < adev->gfx.num_gfx_rings; i++) {
			j = i + xcc_id * adev->gfx.num_gfx_rings;
			kiq->pmf->kiq_unmap_queues(kiq_ring,
						   &adev->gfx.gfx_ring[j],
						   PREEMPT_QUEUES, 0, 0);
		}
	}

	if (adev->gfx.kiq[0].ring.sched.ready && !adev->job_hang)
		r = amdgpu_ring_test_helper(kiq_ring);
	spin_unlock(&kiq->ring_lock);

	return r;
}

int amdgpu_queue_mask_bit_to_set_resource_bit(struct amdgpu_device *adev,
					int queue_bit)
{
	int mec, pipe, queue;
	int set_resource_bit = 0;

	amdgpu_queue_mask_bit_to_mec_queue(adev, queue_bit, &mec, &pipe, &queue);

	set_resource_bit = mec * 4 * 8 + pipe * 8 + queue;

	return set_resource_bit;
}

int amdgpu_gfx_enable_kcq(struct amdgpu_device *adev, int xcc_id)
{
	struct amdgpu_kiq *kiq = &adev->gfx.kiq[xcc_id];
	struct amdgpu_ring *kiq_ring = &kiq->ring;
	uint64_t queue_mask = 0;
	int r, i, j;

	if (!kiq->pmf || !kiq->pmf->kiq_map_queues || !kiq->pmf->kiq_set_resources)
		return -EINVAL;

	for (i = 0; i < AMDGPU_MAX_COMPUTE_QUEUES; ++i) {
		if (!test_bit(i, adev->gfx.mec_bitmap[xcc_id].queue_bitmap))
			continue;

		/* This situation may be hit in the future if a new HW
		 * generation exposes more than 64 queues. If so, the
		 * definition of queue_mask needs updating */
		if (WARN_ON(i > (sizeof(queue_mask)*8))) {
			DRM_ERROR("Invalid KCQ enabled: %d\n", i);
			break;
		}

		queue_mask |= (1ull << amdgpu_queue_mask_bit_to_set_resource_bit(adev, i));
	}

	DRM_INFO("kiq ring mec %d pipe %d q %d\n", kiq_ring->me, kiq_ring->pipe,
							kiq_ring->queue);
	amdgpu_device_flush_hdp(adev, NULL);

	spin_lock(&kiq->ring_lock);
	r = amdgpu_ring_alloc(kiq_ring, kiq->pmf->map_queues_size *
					adev->gfx.num_compute_rings +
					kiq->pmf->set_resources_size);
	if (r) {
		DRM_ERROR("Failed to lock KIQ (%d).\n", r);
		spin_unlock(&kiq->ring_lock);
		return r;
	}

	if (adev->enable_mes)
		queue_mask = ~0ULL;

	kiq->pmf->kiq_set_resources(kiq_ring, queue_mask);
	for (i = 0; i < adev->gfx.num_compute_rings; i++) {
		j = i + xcc_id * adev->gfx.num_compute_rings;
		kiq->pmf->kiq_map_queues(kiq_ring,
					 &adev->gfx.compute_ring[j]);
	}

	r = amdgpu_ring_test_helper(kiq_ring);
	spin_unlock(&kiq->ring_lock);
	if (r)
		DRM_ERROR("KCQ enable failed\n");

	return r;
}

int amdgpu_gfx_enable_kgq(struct amdgpu_device *adev, int xcc_id)
{
	struct amdgpu_kiq *kiq = &adev->gfx.kiq[xcc_id];
	struct amdgpu_ring *kiq_ring = &kiq->ring;
	int r, i, j;

	if (!kiq->pmf || !kiq->pmf->kiq_map_queues)
		return -EINVAL;

	amdgpu_device_flush_hdp(adev, NULL);

	spin_lock(&kiq->ring_lock);
	/* No need to map kcq on the slave */
	if (amdgpu_gfx_is_master_xcc(adev, xcc_id)) {
		r = amdgpu_ring_alloc(kiq_ring, kiq->pmf->map_queues_size *
						adev->gfx.num_gfx_rings);
		if (r) {
			DRM_ERROR("Failed to lock KIQ (%d).\n", r);
			spin_unlock(&kiq->ring_lock);
			return r;
		}

		for (i = 0; i < adev->gfx.num_gfx_rings; i++) {
			j = i + xcc_id * adev->gfx.num_gfx_rings;
			kiq->pmf->kiq_map_queues(kiq_ring,
						 &adev->gfx.gfx_ring[j]);
		}
	}

	r = amdgpu_ring_test_helper(kiq_ring);
	spin_unlock(&kiq->ring_lock);
	if (r)
		DRM_ERROR("KGQ enable failed\n");

	return r;
}

/* amdgpu_gfx_off_ctrl - Handle gfx off feature enable/disable
 *
 * @adev: amdgpu_device pointer
 * @bool enable true: enable gfx off feature, false: disable gfx off feature
 *
 * 1. gfx off feature will be enabled by gfx ip after gfx cg gp enabled.
 * 2. other client can send request to disable gfx off feature, the request should be honored.
 * 3. other client can cancel their request of disable gfx off feature
 * 4. other client should not send request to enable gfx off feature before disable gfx off feature.
 */

void amdgpu_gfx_off_ctrl(struct amdgpu_device *adev, bool enable)
{
	unsigned long delay = GFX_OFF_DELAY_ENABLE;

	if (!(adev->pm.pp_feature & PP_GFXOFF_MASK))
		return;

	mutex_lock(&adev->gfx.gfx_off_mutex);

	if (enable) {
		/* If the count is already 0, it means there's an imbalance bug somewhere.
		 * Note that the bug may be in a different caller than the one which triggers the
		 * WARN_ON_ONCE.
		 */
		if (WARN_ON_ONCE(adev->gfx.gfx_off_req_count == 0))
			goto unlock;

		adev->gfx.gfx_off_req_count--;

		if (adev->gfx.gfx_off_req_count == 0 &&
		    !adev->gfx.gfx_off_state) {
			/* If going to s2idle, no need to wait */
			if (adev->in_s0ix) {
				if (!amdgpu_dpm_set_powergating_by_smu(adev,
						AMD_IP_BLOCK_TYPE_GFX, true))
					adev->gfx.gfx_off_state = true;
			} else {
				schedule_delayed_work(&adev->gfx.gfx_off_delay_work,
					      delay);
			}
		}
	} else {
		if (adev->gfx.gfx_off_req_count == 0) {
			cancel_delayed_work_sync(&adev->gfx.gfx_off_delay_work);

			if (adev->gfx.gfx_off_state &&
			    !amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_GFX, false)) {
				adev->gfx.gfx_off_state = false;

				if (adev->gfx.funcs->init_spm_golden) {
					dev_dbg(adev->dev,
						"GFXOFF is disabled, re-init SPM golden settings\n");
					amdgpu_gfx_init_spm_golden(adev);
				}
			}
		}

		adev->gfx.gfx_off_req_count++;
	}

unlock:
	mutex_unlock(&adev->gfx.gfx_off_mutex);
}

int amdgpu_set_gfx_off_residency(struct amdgpu_device *adev, bool value)
{
	int r = 0;

	mutex_lock(&adev->gfx.gfx_off_mutex);

	r = amdgpu_dpm_set_residency_gfxoff(adev, value);

	mutex_unlock(&adev->gfx.gfx_off_mutex);

	return r;
}

int amdgpu_get_gfx_off_residency(struct amdgpu_device *adev, u32 *value)
{
	int r = 0;

	mutex_lock(&adev->gfx.gfx_off_mutex);

	r = amdgpu_dpm_get_residency_gfxoff(adev, value);

	mutex_unlock(&adev->gfx.gfx_off_mutex);

	return r;
}

int amdgpu_get_gfx_off_entrycount(struct amdgpu_device *adev, u64 *value)
{
	int r = 0;

	mutex_lock(&adev->gfx.gfx_off_mutex);

	r = amdgpu_dpm_get_entrycount_gfxoff(adev, value);

	mutex_unlock(&adev->gfx.gfx_off_mutex);

	return r;
}

int amdgpu_get_gfx_off_status(struct amdgpu_device *adev, uint32_t *value)
{

	int r = 0;

	mutex_lock(&adev->gfx.gfx_off_mutex);

	r = amdgpu_dpm_get_status_gfxoff(adev, value);

	mutex_unlock(&adev->gfx.gfx_off_mutex);

	return r;
}

int amdgpu_gfx_ras_late_init(struct amdgpu_device *adev, struct ras_common_if *ras_block)
{
	int r;

	if (amdgpu_ras_is_supported(adev, ras_block->block)) {
		if (!amdgpu_persistent_edc_harvesting_supported(adev))
			amdgpu_ras_reset_error_status(adev, AMDGPU_RAS_BLOCK__GFX);

		r = amdgpu_ras_block_late_init(adev, ras_block);
		if (r)
			return r;

		if (adev->gfx.cp_ecc_error_irq.funcs) {
			r = amdgpu_irq_get(adev, &adev->gfx.cp_ecc_error_irq, 0);
			if (r)
				goto late_fini;
		}
	} else {
		amdgpu_ras_feature_enable_on_boot(adev, ras_block, 0);
	}

	return 0;
late_fini:
	amdgpu_ras_block_late_fini(adev, ras_block);
	return r;
}

int amdgpu_gfx_ras_sw_init(struct amdgpu_device *adev)
{
	int err = 0;
	struct amdgpu_gfx_ras *ras = NULL;

	/* adev->gfx.ras is NULL, which means gfx does not
	 * support ras function, then do nothing here.
	 */
	if (!adev->gfx.ras)
		return 0;

	ras = adev->gfx.ras;

	err = amdgpu_ras_register_ras_block(adev, &ras->ras_block);
	if (err) {
		dev_err(adev->dev, "Failed to register gfx ras block!\n");
		return err;
	}

	strcpy(ras->ras_block.ras_comm.name, "gfx");
	ras->ras_block.ras_comm.block = AMDGPU_RAS_BLOCK__GFX;
	ras->ras_block.ras_comm.type = AMDGPU_RAS_ERROR__MULTI_UNCORRECTABLE;
	adev->gfx.ras_if = &ras->ras_block.ras_comm;

	/* If not define special ras_late_init function, use gfx default ras_late_init */
	if (!ras->ras_block.ras_late_init)
		ras->ras_block.ras_late_init = amdgpu_gfx_ras_late_init;

	/* If not defined special ras_cb function, use default ras_cb */
	if (!ras->ras_block.ras_cb)
		ras->ras_block.ras_cb = amdgpu_gfx_process_ras_data_cb;

	return 0;
}

int amdgpu_gfx_poison_consumption_handler(struct amdgpu_device *adev,
						struct amdgpu_iv_entry *entry)
{
	if (adev->gfx.ras && adev->gfx.ras->poison_consumption_handler)
		return adev->gfx.ras->poison_consumption_handler(adev, entry);

	return 0;
}

int amdgpu_gfx_process_ras_data_cb(struct amdgpu_device *adev,
		void *err_data,
		struct amdgpu_iv_entry *entry)
{
	/* TODO ue will trigger an interrupt.
	 *
	 * When “Full RAS” is enabled, the per-IP interrupt sources should
	 * be disabled and the driver should only look for the aggregated
	 * interrupt via sync flood
	 */
	if (!amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__GFX)) {
		kgd2kfd_set_sram_ecc_flag(adev->kfd.dev);
		if (adev->gfx.ras && adev->gfx.ras->ras_block.hw_ops &&
		    adev->gfx.ras->ras_block.hw_ops->query_ras_error_count)
			adev->gfx.ras->ras_block.hw_ops->query_ras_error_count(adev, err_data);
		amdgpu_ras_reset_gpu(adev);
	}
	return AMDGPU_RAS_SUCCESS;
}

int amdgpu_gfx_cp_ecc_error_irq(struct amdgpu_device *adev,
				  struct amdgpu_irq_src *source,
				  struct amdgpu_iv_entry *entry)
{
	struct ras_common_if *ras_if = adev->gfx.ras_if;
	struct ras_dispatch_if ih_data = {
		.entry = entry,
	};

	if (!ras_if)
		return 0;

	ih_data.head = *ras_if;

	DRM_ERROR("CP ECC ERROR IRQ\n");
	amdgpu_ras_interrupt_dispatch(adev, &ih_data);
	return 0;
}

void amdgpu_gfx_ras_error_func(struct amdgpu_device *adev,
		void *ras_error_status,
		void (*func)(struct amdgpu_device *adev, void *ras_error_status,
				int xcc_id))
{
	int i;
	int num_xcc = adev->gfx.xcc_mask ? NUM_XCC(adev->gfx.xcc_mask) : 1;
	uint32_t xcc_mask = GENMASK(num_xcc - 1, 0);
	struct ras_err_data *err_data = (struct ras_err_data *)ras_error_status;

	if (err_data) {
		err_data->ue_count = 0;
		err_data->ce_count = 0;
	}

	for_each_inst(i, xcc_mask)
		func(adev, ras_error_status, i);
}

uint32_t amdgpu_kiq_rreg(struct amdgpu_device *adev, uint32_t reg, uint32_t xcc_id)
{
	signed long r, cnt = 0;
	unsigned long flags;
	uint32_t seq, reg_val_offs = 0, value = 0;
	struct amdgpu_kiq *kiq = &adev->gfx.kiq[xcc_id];
	struct amdgpu_ring *ring = &kiq->ring;

	if (amdgpu_device_skip_hw_access(adev))
		return 0;

	if (adev->mes.ring.sched.ready)
		return amdgpu_mes_rreg(adev, reg);

	BUG_ON(!ring->funcs->emit_rreg);

	spin_lock_irqsave(&kiq->ring_lock, flags);
	if (amdgpu_device_wb_get(adev, &reg_val_offs)) {
		pr_err("critical bug! too many kiq readers\n");
		goto failed_unlock;
	}
	amdgpu_ring_alloc(ring, 32);
	amdgpu_ring_emit_rreg(ring, reg, reg_val_offs);
	r = amdgpu_fence_emit_polling(ring, &seq, MAX_KIQ_REG_WAIT);
	if (r)
		goto failed_undo;

	amdgpu_ring_commit(ring);
	spin_unlock_irqrestore(&kiq->ring_lock, flags);

	r = amdgpu_fence_wait_polling(ring, seq, MAX_KIQ_REG_WAIT);

	/* don't wait anymore for gpu reset case because this way may
	 * block gpu_recover() routine forever, e.g. this virt_kiq_rreg
	 * is triggered in TTM and ttm_bo_lock_delayed_workqueue() will
	 * never return if we keep waiting in virt_kiq_rreg, which cause
	 * gpu_recover() hang there.
	 *
	 * also don't wait anymore for IRQ context
	 * */
	if (r < 1 && (amdgpu_in_reset(adev) || in_interrupt()))
		goto failed_kiq_read;

	might_sleep();
	while (r < 1 && cnt++ < MAX_KIQ_REG_TRY) {
		msleep(MAX_KIQ_REG_BAILOUT_INTERVAL);
		r = amdgpu_fence_wait_polling(ring, seq, MAX_KIQ_REG_WAIT);
	}

	if (cnt > MAX_KIQ_REG_TRY)
		goto failed_kiq_read;

	mb();
	value = adev->wb.wb[reg_val_offs];
	amdgpu_device_wb_free(adev, reg_val_offs);
	return value;

failed_undo:
	amdgpu_ring_undo(ring);
failed_unlock:
	spin_unlock_irqrestore(&kiq->ring_lock, flags);
failed_kiq_read:
	if (reg_val_offs)
		amdgpu_device_wb_free(adev, reg_val_offs);
	dev_err(adev->dev, "failed to read reg:%x\n", reg);
	return ~0;
}

void amdgpu_kiq_wreg(struct amdgpu_device *adev, uint32_t reg, uint32_t v, uint32_t xcc_id)
{
	signed long r, cnt = 0;
	unsigned long flags;
	uint32_t seq;
	struct amdgpu_kiq *kiq = &adev->gfx.kiq[xcc_id];
	struct amdgpu_ring *ring = &kiq->ring;

	BUG_ON(!ring->funcs->emit_wreg);

	if (amdgpu_device_skip_hw_access(adev))
		return;

	if (adev->mes.ring.sched.ready) {
		amdgpu_mes_wreg(adev, reg, v);
		return;
	}

	spin_lock_irqsave(&kiq->ring_lock, flags);
	amdgpu_ring_alloc(ring, 32);
	amdgpu_ring_emit_wreg(ring, reg, v);
	r = amdgpu_fence_emit_polling(ring, &seq, MAX_KIQ_REG_WAIT);
	if (r)
		goto failed_undo;

	amdgpu_ring_commit(ring);
	spin_unlock_irqrestore(&kiq->ring_lock, flags);

	r = amdgpu_fence_wait_polling(ring, seq, MAX_KIQ_REG_WAIT);

	/* don't wait anymore for gpu reset case because this way may
	 * block gpu_recover() routine forever, e.g. this virt_kiq_rreg
	 * is triggered in TTM and ttm_bo_lock_delayed_workqueue() will
	 * never return if we keep waiting in virt_kiq_rreg, which cause
	 * gpu_recover() hang there.
	 *
	 * also don't wait anymore for IRQ context
	 * */
	if (r < 1 && (amdgpu_in_reset(adev) || in_interrupt()))
		goto failed_kiq_write;

	might_sleep();
	while (r < 1 && cnt++ < MAX_KIQ_REG_TRY) {

		msleep(MAX_KIQ_REG_BAILOUT_INTERVAL);
		r = amdgpu_fence_wait_polling(ring, seq, MAX_KIQ_REG_WAIT);
	}

	if (cnt > MAX_KIQ_REG_TRY)
		goto failed_kiq_write;

	return;

failed_undo:
	amdgpu_ring_undo(ring);
	spin_unlock_irqrestore(&kiq->ring_lock, flags);
failed_kiq_write:
	dev_err(adev->dev, "failed to write reg:%x\n", reg);
}

int amdgpu_gfx_get_num_kcq(struct amdgpu_device *adev)
{
	if (amdgpu_num_kcq == -1) {
		return 8;
	} else if (amdgpu_num_kcq > 8 || amdgpu_num_kcq < 0) {
		dev_warn(adev->dev, "set kernel compute queue number to 8 due to invalid parameter provided by user\n");
		return 8;
	}
	return amdgpu_num_kcq;
}

void amdgpu_gfx_cp_init_microcode(struct amdgpu_device *adev,
				  uint32_t ucode_id)
{
	const struct gfx_firmware_header_v1_0 *cp_hdr;
	const struct gfx_firmware_header_v2_0 *cp_hdr_v2_0;
	struct amdgpu_firmware_info *info = NULL;
	const struct firmware *ucode_fw;
	unsigned int fw_size;

	switch (ucode_id) {
	case AMDGPU_UCODE_ID_CP_PFP:
		cp_hdr = (const struct gfx_firmware_header_v1_0 *)
			adev->gfx.pfp_fw->data;
		adev->gfx.pfp_fw_version =
			le32_to_cpu(cp_hdr->header.ucode_version);
		adev->gfx.pfp_feature_version =
			le32_to_cpu(cp_hdr->ucode_feature_version);
		ucode_fw = adev->gfx.pfp_fw;
		fw_size = le32_to_cpu(cp_hdr->header.ucode_size_bytes);
		break;
	case AMDGPU_UCODE_ID_CP_RS64_PFP:
		cp_hdr_v2_0 = (const struct gfx_firmware_header_v2_0 *)
			adev->gfx.pfp_fw->data;
		adev->gfx.pfp_fw_version =
			le32_to_cpu(cp_hdr_v2_0->header.ucode_version);
		adev->gfx.pfp_feature_version =
			le32_to_cpu(cp_hdr_v2_0->ucode_feature_version);
		ucode_fw = adev->gfx.pfp_fw;
		fw_size = le32_to_cpu(cp_hdr_v2_0->ucode_size_bytes);
		break;
	case AMDGPU_UCODE_ID_CP_RS64_PFP_P0_STACK:
	case AMDGPU_UCODE_ID_CP_RS64_PFP_P1_STACK:
		cp_hdr_v2_0 = (const struct gfx_firmware_header_v2_0 *)
			adev->gfx.pfp_fw->data;
		ucode_fw = adev->gfx.pfp_fw;
		fw_size = le32_to_cpu(cp_hdr_v2_0->data_size_bytes);
		break;
	case AMDGPU_UCODE_ID_CP_ME:
		cp_hdr = (const struct gfx_firmware_header_v1_0 *)
			adev->gfx.me_fw->data;
		adev->gfx.me_fw_version =
			le32_to_cpu(cp_hdr->header.ucode_version);
		adev->gfx.me_feature_version =
			le32_to_cpu(cp_hdr->ucode_feature_version);
		ucode_fw = adev->gfx.me_fw;
		fw_size = le32_to_cpu(cp_hdr->header.ucode_size_bytes);
		break;
	case AMDGPU_UCODE_ID_CP_RS64_ME:
		cp_hdr_v2_0 = (const struct gfx_firmware_header_v2_0 *)
			adev->gfx.me_fw->data;
		adev->gfx.me_fw_version =
			le32_to_cpu(cp_hdr_v2_0->header.ucode_version);
		adev->gfx.me_feature_version =
			le32_to_cpu(cp_hdr_v2_0->ucode_feature_version);
		ucode_fw = adev->gfx.me_fw;
		fw_size = le32_to_cpu(cp_hdr_v2_0->ucode_size_bytes);
		break;
	case AMDGPU_UCODE_ID_CP_RS64_ME_P0_STACK:
	case AMDGPU_UCODE_ID_CP_RS64_ME_P1_STACK:
		cp_hdr_v2_0 = (const struct gfx_firmware_header_v2_0 *)
			adev->gfx.me_fw->data;
		ucode_fw = adev->gfx.me_fw;
		fw_size = le32_to_cpu(cp_hdr_v2_0->data_size_bytes);
		break;
	case AMDGPU_UCODE_ID_CP_CE:
		cp_hdr = (const struct gfx_firmware_header_v1_0 *)
			adev->gfx.ce_fw->data;
		adev->gfx.ce_fw_version =
			le32_to_cpu(cp_hdr->header.ucode_version);
		adev->gfx.ce_feature_version =
			le32_to_cpu(cp_hdr->ucode_feature_version);
		ucode_fw = adev->gfx.ce_fw;
		fw_size = le32_to_cpu(cp_hdr->header.ucode_size_bytes);
		break;
	case AMDGPU_UCODE_ID_CP_MEC1:
		cp_hdr = (const struct gfx_firmware_header_v1_0 *)
			adev->gfx.mec_fw->data;
		adev->gfx.mec_fw_version =
			le32_to_cpu(cp_hdr->header.ucode_version);
		adev->gfx.mec_feature_version =
			le32_to_cpu(cp_hdr->ucode_feature_version);
		ucode_fw = adev->gfx.mec_fw;
		fw_size = le32_to_cpu(cp_hdr->header.ucode_size_bytes) -
			  le32_to_cpu(cp_hdr->jt_size) * 4;
		break;
	case AMDGPU_UCODE_ID_CP_MEC1_JT:
		cp_hdr = (const struct gfx_firmware_header_v1_0 *)
			adev->gfx.mec_fw->data;
		ucode_fw = adev->gfx.mec_fw;
		fw_size = le32_to_cpu(cp_hdr->jt_size) * 4;
		break;
	case AMDGPU_UCODE_ID_CP_MEC2:
		cp_hdr = (const struct gfx_firmware_header_v1_0 *)
			adev->gfx.mec2_fw->data;
		adev->gfx.mec2_fw_version =
			le32_to_cpu(cp_hdr->header.ucode_version);
		adev->gfx.mec2_feature_version =
			le32_to_cpu(cp_hdr->ucode_feature_version);
		ucode_fw = adev->gfx.mec2_fw;
		fw_size = le32_to_cpu(cp_hdr->header.ucode_size_bytes) -
			  le32_to_cpu(cp_hdr->jt_size) * 4;
		break;
	case AMDGPU_UCODE_ID_CP_MEC2_JT:
		cp_hdr = (const struct gfx_firmware_header_v1_0 *)
			adev->gfx.mec2_fw->data;
		ucode_fw = adev->gfx.mec2_fw;
		fw_size = le32_to_cpu(cp_hdr->jt_size) * 4;
		break;
	case AMDGPU_UCODE_ID_CP_RS64_MEC:
		cp_hdr_v2_0 = (const struct gfx_firmware_header_v2_0 *)
			adev->gfx.mec_fw->data;
		adev->gfx.mec_fw_version =
			le32_to_cpu(cp_hdr_v2_0->header.ucode_version);
		adev->gfx.mec_feature_version =
			le32_to_cpu(cp_hdr_v2_0->ucode_feature_version);
		ucode_fw = adev->gfx.mec_fw;
		fw_size = le32_to_cpu(cp_hdr_v2_0->ucode_size_bytes);
		break;
	case AMDGPU_UCODE_ID_CP_RS64_MEC_P0_STACK:
	case AMDGPU_UCODE_ID_CP_RS64_MEC_P1_STACK:
	case AMDGPU_UCODE_ID_CP_RS64_MEC_P2_STACK:
	case AMDGPU_UCODE_ID_CP_RS64_MEC_P3_STACK:
		cp_hdr_v2_0 = (const struct gfx_firmware_header_v2_0 *)
			adev->gfx.mec_fw->data;
		ucode_fw = adev->gfx.mec_fw;
		fw_size = le32_to_cpu(cp_hdr_v2_0->data_size_bytes);
		break;
	default:
		dev_err(adev->dev, "Invalid ucode id %u\n", ucode_id);
		return;
	}

	if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) {
		info = &adev->firmware.ucode[ucode_id];
		info->ucode_id = ucode_id;
		info->fw = ucode_fw;
		adev->firmware.fw_size += ALIGN(fw_size, PAGE_SIZE);
	}
}

bool amdgpu_gfx_is_master_xcc(struct amdgpu_device *adev, int xcc_id)
{
	return !(xcc_id % (adev->gfx.num_xcc_per_xcp ?
			adev->gfx.num_xcc_per_xcp : 1));
}

static ssize_t amdgpu_gfx_get_current_compute_partition(struct device *dev,
						struct device_attribute *addr,
						char *buf)
{
	struct drm_device *ddev = dev_get_drvdata(dev);
	struct amdgpu_device *adev = drm_to_adev(ddev);
	int mode;

	mode = amdgpu_xcp_query_partition_mode(adev->xcp_mgr,
					       AMDGPU_XCP_FL_NONE);

	return sysfs_emit(buf, "%s\n", amdgpu_gfx_compute_mode_desc(mode));
}

static ssize_t amdgpu_gfx_set_compute_partition(struct device *dev,
						struct device_attribute *addr,
						const char *buf, size_t count)
{
	struct drm_device *ddev = dev_get_drvdata(dev);
	struct amdgpu_device *adev = drm_to_adev(ddev);
	enum amdgpu_gfx_partition mode;
	int ret = 0, num_xcc;

	num_xcc = NUM_XCC(adev->gfx.xcc_mask);
	if (num_xcc % 2 != 0)
		return -EINVAL;

	if (!strncasecmp("SPX", buf, strlen("SPX"))) {
		mode = AMDGPU_SPX_PARTITION_MODE;
	} else if (!strncasecmp("DPX", buf, strlen("DPX"))) {
		/*
		 * DPX mode needs AIDs to be in multiple of 2.
		 * Each AID connects 2 XCCs.
		 */
		if (num_xcc%4)
			return -EINVAL;
		mode = AMDGPU_DPX_PARTITION_MODE;
	} else if (!strncasecmp("TPX", buf, strlen("TPX"))) {
		if (num_xcc != 6)
			return -EINVAL;
		mode = AMDGPU_TPX_PARTITION_MODE;
	} else if (!strncasecmp("QPX", buf, strlen("QPX"))) {
		if (num_xcc != 8)
			return -EINVAL;
		mode = AMDGPU_QPX_PARTITION_MODE;
	} else if (!strncasecmp("CPX", buf, strlen("CPX"))) {
		mode = AMDGPU_CPX_PARTITION_MODE;
	} else {
		return -EINVAL;
	}

	ret = amdgpu_xcp_switch_partition_mode(adev->xcp_mgr, mode);

	if (ret)
		return ret;

	return count;
}

static ssize_t amdgpu_gfx_get_available_compute_partition(struct device *dev,
						struct device_attribute *addr,
						char *buf)
{
	struct drm_device *ddev = dev_get_drvdata(dev);
	struct amdgpu_device *adev = drm_to_adev(ddev);
	char *supported_partition;

	/* TBD */
	switch (NUM_XCC(adev->gfx.xcc_mask)) {
	case 8:
		supported_partition = "SPX, DPX, QPX, CPX";
		break;
	case 6:
		supported_partition = "SPX, TPX, CPX";
		break;
	case 4:
		supported_partition = "SPX, DPX, CPX";
		break;
	/* this seems only existing in emulation phase */
	case 2:
		supported_partition = "SPX, CPX";
		break;
	default:
		supported_partition = "Not supported";
		break;
	}

	return sysfs_emit(buf, "%s\n", supported_partition);
}

static DEVICE_ATTR(current_compute_partition, 0644,
		   amdgpu_gfx_get_current_compute_partition,
		   amdgpu_gfx_set_compute_partition);

static DEVICE_ATTR(available_compute_partition, 0444,
		   amdgpu_gfx_get_available_compute_partition, NULL);

int amdgpu_gfx_sysfs_init(struct amdgpu_device *adev)
{
	int r;

	r = device_create_file(adev->dev, &dev_attr_current_compute_partition);
	if (r)
		return r;

	r = device_create_file(adev->dev, &dev_attr_available_compute_partition);

	return r;
}

void amdgpu_gfx_sysfs_fini(struct amdgpu_device *adev)
{
	device_remove_file(adev->dev, &dev_attr_current_compute_partition);
	device_remove_file(adev->dev, &dev_attr_available_compute_partition);
}