summaryrefslogtreecommitdiff
path: root/drivers/gpu/drm/vc4/vc4_gem.c
blob: d0c6bfb68c4ee9d88c2a026e5908915d331733db (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
/*
 * Copyright © 2014 Broadcom
 *
 * 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 (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 NONINFRINGEMENT.  IN NO EVENT SHALL
 * THE AUTHORS OR COPYRIGHT HOLDERS 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/module.h>
#include <linux/platform_device.h>
#include <linux/pm_runtime.h>
#include <linux/device.h>
#include <linux/io.h>
#include <linux/sched/signal.h>

#include "uapi/drm/vc4_drm.h"
#include "vc4_drv.h"
#include "vc4_regs.h"
#include "vc4_trace.h"

static void
vc4_queue_hangcheck(struct drm_device *dev)
{
	struct vc4_dev *vc4 = to_vc4_dev(dev);

	mod_timer(&vc4->hangcheck.timer,
		  round_jiffies_up(jiffies + msecs_to_jiffies(100)));
}

struct vc4_hang_state {
	struct drm_vc4_get_hang_state user_state;

	u32 bo_count;
	struct drm_gem_object **bo;
};

static void
vc4_free_hang_state(struct drm_device *dev, struct vc4_hang_state *state)
{
	unsigned int i;

	for (i = 0; i < state->user_state.bo_count; i++)
		drm_gem_object_put_unlocked(state->bo[i]);

	kfree(state);
}

int
vc4_get_hang_state_ioctl(struct drm_device *dev, void *data,
			 struct drm_file *file_priv)
{
	struct drm_vc4_get_hang_state *get_state = data;
	struct drm_vc4_get_hang_state_bo *bo_state;
	struct vc4_hang_state *kernel_state;
	struct drm_vc4_get_hang_state *state;
	struct vc4_dev *vc4 = to_vc4_dev(dev);
	unsigned long irqflags;
	u32 i;
	int ret = 0;

	spin_lock_irqsave(&vc4->job_lock, irqflags);
	kernel_state = vc4->hang_state;
	if (!kernel_state) {
		spin_unlock_irqrestore(&vc4->job_lock, irqflags);
		return -ENOENT;
	}
	state = &kernel_state->user_state;

	/* If the user's array isn't big enough, just return the
	 * required array size.
	 */
	if (get_state->bo_count < state->bo_count) {
		get_state->bo_count = state->bo_count;
		spin_unlock_irqrestore(&vc4->job_lock, irqflags);
		return 0;
	}

	vc4->hang_state = NULL;
	spin_unlock_irqrestore(&vc4->job_lock, irqflags);

	/* Save the user's BO pointer, so we don't stomp it with the memcpy. */
	state->bo = get_state->bo;
	memcpy(get_state, state, sizeof(*state));

	bo_state = kcalloc(state->bo_count, sizeof(*bo_state), GFP_KERNEL);
	if (!bo_state) {
		ret = -ENOMEM;
		goto err_free;
	}

	for (i = 0; i < state->bo_count; i++) {
		struct vc4_bo *vc4_bo = to_vc4_bo(kernel_state->bo[i]);
		u32 handle;

		ret = drm_gem_handle_create(file_priv, kernel_state->bo[i],
					    &handle);

		if (ret) {
			state->bo_count = i;
			goto err_delete_handle;
		}
		bo_state[i].handle = handle;
		bo_state[i].paddr = vc4_bo->base.paddr;
		bo_state[i].size = vc4_bo->base.base.size;
	}

	if (copy_to_user(u64_to_user_ptr(get_state->bo),
			 bo_state,
			 state->bo_count * sizeof(*bo_state)))
		ret = -EFAULT;

err_delete_handle:
	if (ret) {
		for (i = 0; i < state->bo_count; i++)
			drm_gem_handle_delete(file_priv, bo_state[i].handle);
	}

err_free:
	vc4_free_hang_state(dev, kernel_state);
	kfree(bo_state);

	return ret;
}

static void
vc4_save_hang_state(struct drm_device *dev)
{
	struct vc4_dev *vc4 = to_vc4_dev(dev);
	struct drm_vc4_get_hang_state *state;
	struct vc4_hang_state *kernel_state;
	struct vc4_exec_info *exec[2];
	struct vc4_bo *bo;
	unsigned long irqflags;
	unsigned int i, j, unref_list_count, prev_idx;

	kernel_state = kcalloc(1, sizeof(*kernel_state), GFP_KERNEL);
	if (!kernel_state)
		return;

	state = &kernel_state->user_state;

	spin_lock_irqsave(&vc4->job_lock, irqflags);
	exec[0] = vc4_first_bin_job(vc4);
	exec[1] = vc4_first_render_job(vc4);
	if (!exec[0] && !exec[1]) {
		spin_unlock_irqrestore(&vc4->job_lock, irqflags);
		return;
	}

	/* Get the bos from both binner and renderer into hang state. */
	state->bo_count = 0;
	for (i = 0; i < 2; i++) {
		if (!exec[i])
			continue;

		unref_list_count = 0;
		list_for_each_entry(bo, &exec[i]->unref_list, unref_head)
			unref_list_count++;
		state->bo_count += exec[i]->bo_count + unref_list_count;
	}

	kernel_state->bo = kcalloc(state->bo_count,
				   sizeof(*kernel_state->bo), GFP_ATOMIC);

	if (!kernel_state->bo) {
		spin_unlock_irqrestore(&vc4->job_lock, irqflags);
		return;
	}

	prev_idx = 0;
	for (i = 0; i < 2; i++) {
		if (!exec[i])
			continue;

		for (j = 0; j < exec[i]->bo_count; j++) {
			drm_gem_object_get(&exec[i]->bo[j]->base);
			kernel_state->bo[j + prev_idx] = &exec[i]->bo[j]->base;
		}

		list_for_each_entry(bo, &exec[i]->unref_list, unref_head) {
			drm_gem_object_get(&bo->base.base);
			kernel_state->bo[j + prev_idx] = &bo->base.base;
			j++;
		}
		prev_idx = j + 1;
	}

	if (exec[0])
		state->start_bin = exec[0]->ct0ca;
	if (exec[1])
		state->start_render = exec[1]->ct1ca;

	spin_unlock_irqrestore(&vc4->job_lock, irqflags);

	state->ct0ca = V3D_READ(V3D_CTNCA(0));
	state->ct0ea = V3D_READ(V3D_CTNEA(0));

	state->ct1ca = V3D_READ(V3D_CTNCA(1));
	state->ct1ea = V3D_READ(V3D_CTNEA(1));

	state->ct0cs = V3D_READ(V3D_CTNCS(0));
	state->ct1cs = V3D_READ(V3D_CTNCS(1));

	state->ct0ra0 = V3D_READ(V3D_CT00RA0);
	state->ct1ra0 = V3D_READ(V3D_CT01RA0);

	state->bpca = V3D_READ(V3D_BPCA);
	state->bpcs = V3D_READ(V3D_BPCS);
	state->bpoa = V3D_READ(V3D_BPOA);
	state->bpos = V3D_READ(V3D_BPOS);

	state->vpmbase = V3D_READ(V3D_VPMBASE);

	state->dbge = V3D_READ(V3D_DBGE);
	state->fdbgo = V3D_READ(V3D_FDBGO);
	state->fdbgb = V3D_READ(V3D_FDBGB);
	state->fdbgr = V3D_READ(V3D_FDBGR);
	state->fdbgs = V3D_READ(V3D_FDBGS);
	state->errstat = V3D_READ(V3D_ERRSTAT);

	spin_lock_irqsave(&vc4->job_lock, irqflags);
	if (vc4->hang_state) {
		spin_unlock_irqrestore(&vc4->job_lock, irqflags);
		vc4_free_hang_state(dev, kernel_state);
	} else {
		vc4->hang_state = kernel_state;
		spin_unlock_irqrestore(&vc4->job_lock, irqflags);
	}
}

static void
vc4_reset(struct drm_device *dev)
{
	struct vc4_dev *vc4 = to_vc4_dev(dev);

	DRM_INFO("Resetting GPU.\n");

	mutex_lock(&vc4->power_lock);
	if (vc4->power_refcount) {
		/* Power the device off and back on the by dropping the
		 * reference on runtime PM.
		 */
		pm_runtime_put_sync_suspend(&vc4->v3d->pdev->dev);
		pm_runtime_get_sync(&vc4->v3d->pdev->dev);
	}
	mutex_unlock(&vc4->power_lock);

	vc4_irq_reset(dev);

	/* Rearm the hangcheck -- another job might have been waiting
	 * for our hung one to get kicked off, and vc4_irq_reset()
	 * would have started it.
	 */
	vc4_queue_hangcheck(dev);
}

static void
vc4_reset_work(struct work_struct *work)
{
	struct vc4_dev *vc4 =
		container_of(work, struct vc4_dev, hangcheck.reset_work);

	vc4_save_hang_state(vc4->dev);

	vc4_reset(vc4->dev);
}

static void
vc4_hangcheck_elapsed(unsigned long data)
{
	struct drm_device *dev = (struct drm_device *)data;
	struct vc4_dev *vc4 = to_vc4_dev(dev);
	uint32_t ct0ca, ct1ca;
	unsigned long irqflags;
	struct vc4_exec_info *bin_exec, *render_exec;

	spin_lock_irqsave(&vc4->job_lock, irqflags);

	bin_exec = vc4_first_bin_job(vc4);
	render_exec = vc4_first_render_job(vc4);

	/* If idle, we can stop watching for hangs. */
	if (!bin_exec && !render_exec) {
		spin_unlock_irqrestore(&vc4->job_lock, irqflags);
		return;
	}

	ct0ca = V3D_READ(V3D_CTNCA(0));
	ct1ca = V3D_READ(V3D_CTNCA(1));

	/* If we've made any progress in execution, rearm the timer
	 * and wait.
	 */
	if ((bin_exec && ct0ca != bin_exec->last_ct0ca) ||
	    (render_exec && ct1ca != render_exec->last_ct1ca)) {
		if (bin_exec)
			bin_exec->last_ct0ca = ct0ca;
		if (render_exec)
			render_exec->last_ct1ca = ct1ca;
		spin_unlock_irqrestore(&vc4->job_lock, irqflags);
		vc4_queue_hangcheck(dev);
		return;
	}

	spin_unlock_irqrestore(&vc4->job_lock, irqflags);

	/* We've gone too long with no progress, reset.  This has to
	 * be done from a work struct, since resetting can sleep and
	 * this timer hook isn't allowed to.
	 */
	schedule_work(&vc4->hangcheck.reset_work);
}

static void
submit_cl(struct drm_device *dev, uint32_t thread, uint32_t start, uint32_t end)
{
	struct vc4_dev *vc4 = to_vc4_dev(dev);

	/* Set the current and end address of the control list.
	 * Writing the end register is what starts the job.
	 */
	V3D_WRITE(V3D_CTNCA(thread), start);
	V3D_WRITE(V3D_CTNEA(thread), end);
}

int
vc4_wait_for_seqno(struct drm_device *dev, uint64_t seqno, uint64_t timeout_ns,
		   bool interruptible)
{
	struct vc4_dev *vc4 = to_vc4_dev(dev);
	int ret = 0;
	unsigned long timeout_expire;
	DEFINE_WAIT(wait);

	if (vc4->finished_seqno >= seqno)
		return 0;

	if (timeout_ns == 0)
		return -ETIME;

	timeout_expire = jiffies + nsecs_to_jiffies(timeout_ns);

	trace_vc4_wait_for_seqno_begin(dev, seqno, timeout_ns);
	for (;;) {
		prepare_to_wait(&vc4->job_wait_queue, &wait,
				interruptible ? TASK_INTERRUPTIBLE :
				TASK_UNINTERRUPTIBLE);

		if (interruptible && signal_pending(current)) {
			ret = -ERESTARTSYS;
			break;
		}

		if (vc4->finished_seqno >= seqno)
			break;

		if (timeout_ns != ~0ull) {
			if (time_after_eq(jiffies, timeout_expire)) {
				ret = -ETIME;
				break;
			}
			schedule_timeout(timeout_expire - jiffies);
		} else {
			schedule();
		}
	}

	finish_wait(&vc4->job_wait_queue, &wait);
	trace_vc4_wait_for_seqno_end(dev, seqno);

	return ret;
}

static void
vc4_flush_caches(struct drm_device *dev)
{
	struct vc4_dev *vc4 = to_vc4_dev(dev);

	/* Flush the GPU L2 caches.  These caches sit on top of system
	 * L3 (the 128kb or so shared with the CPU), and are
	 * non-allocating in the L3.
	 */
	V3D_WRITE(V3D_L2CACTL,
		  V3D_L2CACTL_L2CCLR);

	V3D_WRITE(V3D_SLCACTL,
		  VC4_SET_FIELD(0xf, V3D_SLCACTL_T1CC) |
		  VC4_SET_FIELD(0xf, V3D_SLCACTL_T0CC) |
		  VC4_SET_FIELD(0xf, V3D_SLCACTL_UCC) |
		  VC4_SET_FIELD(0xf, V3D_SLCACTL_ICC));
}

/* Sets the registers for the next job to be actually be executed in
 * the hardware.
 *
 * The job_lock should be held during this.
 */
void
vc4_submit_next_bin_job(struct drm_device *dev)
{
	struct vc4_dev *vc4 = to_vc4_dev(dev);
	struct vc4_exec_info *exec;

again:
	exec = vc4_first_bin_job(vc4);
	if (!exec)
		return;

	vc4_flush_caches(dev);

	/* Either put the job in the binner if it uses the binner, or
	 * immediately move it to the to-be-rendered queue.
	 */
	if (exec->ct0ca != exec->ct0ea) {
		submit_cl(dev, 0, exec->ct0ca, exec->ct0ea);
	} else {
		vc4_move_job_to_render(dev, exec);
		goto again;
	}
}

void
vc4_submit_next_render_job(struct drm_device *dev)
{
	struct vc4_dev *vc4 = to_vc4_dev(dev);
	struct vc4_exec_info *exec = vc4_first_render_job(vc4);

	if (!exec)
		return;

	submit_cl(dev, 1, exec->ct1ca, exec->ct1ea);
}

void
vc4_move_job_to_render(struct drm_device *dev, struct vc4_exec_info *exec)
{
	struct vc4_dev *vc4 = to_vc4_dev(dev);
	bool was_empty = list_empty(&vc4->render_job_list);

	list_move_tail(&exec->head, &vc4->render_job_list);
	if (was_empty)
		vc4_submit_next_render_job(dev);
}

static void
vc4_update_bo_seqnos(struct vc4_exec_info *exec, uint64_t seqno)
{
	struct vc4_bo *bo;
	unsigned i;

	for (i = 0; i < exec->bo_count; i++) {
		bo = to_vc4_bo(&exec->bo[i]->base);
		bo->seqno = seqno;

		reservation_object_add_shared_fence(bo->resv, exec->fence);
	}

	list_for_each_entry(bo, &exec->unref_list, unref_head) {
		bo->seqno = seqno;
	}

	for (i = 0; i < exec->rcl_write_bo_count; i++) {
		bo = to_vc4_bo(&exec->rcl_write_bo[i]->base);
		bo->write_seqno = seqno;

		reservation_object_add_excl_fence(bo->resv, exec->fence);
	}
}

static void
vc4_unlock_bo_reservations(struct drm_device *dev,
			   struct vc4_exec_info *exec,
			   struct ww_acquire_ctx *acquire_ctx)
{
	int i;

	for (i = 0; i < exec->bo_count; i++) {
		struct vc4_bo *bo = to_vc4_bo(&exec->bo[i]->base);

		ww_mutex_unlock(&bo->resv->lock);
	}

	ww_acquire_fini(acquire_ctx);
}

/* Takes the reservation lock on all the BOs being referenced, so that
 * at queue submit time we can update the reservations.
 *
 * We don't lock the RCL the tile alloc/state BOs, or overflow memory
 * (all of which are on exec->unref_list).  They're entirely private
 * to vc4, so we don't attach dma-buf fences to them.
 */
static int
vc4_lock_bo_reservations(struct drm_device *dev,
			 struct vc4_exec_info *exec,
			 struct ww_acquire_ctx *acquire_ctx)
{
	int contended_lock = -1;
	int i, ret;
	struct vc4_bo *bo;

	ww_acquire_init(acquire_ctx, &reservation_ww_class);

retry:
	if (contended_lock != -1) {
		bo = to_vc4_bo(&exec->bo[contended_lock]->base);
		ret = ww_mutex_lock_slow_interruptible(&bo->resv->lock,
						       acquire_ctx);
		if (ret) {
			ww_acquire_done(acquire_ctx);
			return ret;
		}
	}

	for (i = 0; i < exec->bo_count; i++) {
		if (i == contended_lock)
			continue;

		bo = to_vc4_bo(&exec->bo[i]->base);

		ret = ww_mutex_lock_interruptible(&bo->resv->lock, acquire_ctx);
		if (ret) {
			int j;

			for (j = 0; j < i; j++) {
				bo = to_vc4_bo(&exec->bo[j]->base);
				ww_mutex_unlock(&bo->resv->lock);
			}

			if (contended_lock != -1 && contended_lock >= i) {
				bo = to_vc4_bo(&exec->bo[contended_lock]->base);

				ww_mutex_unlock(&bo->resv->lock);
			}

			if (ret == -EDEADLK) {
				contended_lock = i;
				goto retry;
			}

			ww_acquire_done(acquire_ctx);
			return ret;
		}
	}

	ww_acquire_done(acquire_ctx);

	/* Reserve space for our shared (read-only) fence references,
	 * before we commit the CL to the hardware.
	 */
	for (i = 0; i < exec->bo_count; i++) {
		bo = to_vc4_bo(&exec->bo[i]->base);

		ret = reservation_object_reserve_shared(bo->resv);
		if (ret) {
			vc4_unlock_bo_reservations(dev, exec, acquire_ctx);
			return ret;
		}
	}

	return 0;
}

/* Queues a struct vc4_exec_info for execution.  If no job is
 * currently executing, then submits it.
 *
 * Unlike most GPUs, our hardware only handles one command list at a
 * time.  To queue multiple jobs at once, we'd need to edit the
 * previous command list to have a jump to the new one at the end, and
 * then bump the end address.  That's a change for a later date,
 * though.
 */
static int
vc4_queue_submit(struct drm_device *dev, struct vc4_exec_info *exec,
		 struct ww_acquire_ctx *acquire_ctx)
{
	struct vc4_dev *vc4 = to_vc4_dev(dev);
	uint64_t seqno;
	unsigned long irqflags;
	struct vc4_fence *fence;

	fence = kzalloc(sizeof(*fence), GFP_KERNEL);
	if (!fence)
		return -ENOMEM;
	fence->dev = dev;

	spin_lock_irqsave(&vc4->job_lock, irqflags);

	seqno = ++vc4->emit_seqno;
	exec->seqno = seqno;

	dma_fence_init(&fence->base, &vc4_fence_ops, &vc4->job_lock,
		       vc4->dma_fence_context, exec->seqno);
	fence->seqno = exec->seqno;
	exec->fence = &fence->base;

	vc4_update_bo_seqnos(exec, seqno);

	vc4_unlock_bo_reservations(dev, exec, acquire_ctx);

	list_add_tail(&exec->head, &vc4->bin_job_list);

	/* If no job was executing, kick ours off.  Otherwise, it'll
	 * get started when the previous job's flush done interrupt
	 * occurs.
	 */
	if (vc4_first_bin_job(vc4) == exec) {
		vc4_submit_next_bin_job(dev);
		vc4_queue_hangcheck(dev);
	}

	spin_unlock_irqrestore(&vc4->job_lock, irqflags);

	return 0;
}

/**
 * vc4_cl_lookup_bos() - Sets up exec->bo[] with the GEM objects
 * referenced by the job.
 * @dev: DRM device
 * @file_priv: DRM file for this fd
 * @exec: V3D job being set up
 *
 * The command validator needs to reference BOs by their index within
 * the submitted job's BO list.  This does the validation of the job's
 * BO list and reference counting for the lifetime of the job.
 *
 * Note that this function doesn't need to unreference the BOs on
 * failure, because that will happen at vc4_complete_exec() time.
 */
static int
vc4_cl_lookup_bos(struct drm_device *dev,
		  struct drm_file *file_priv,
		  struct vc4_exec_info *exec)
{
	struct drm_vc4_submit_cl *args = exec->args;
	uint32_t *handles;
	int ret = 0;
	int i;

	exec->bo_count = args->bo_handle_count;

	if (!exec->bo_count) {
		/* See comment on bo_index for why we have to check
		 * this.
		 */
		DRM_DEBUG("Rendering requires BOs to validate\n");
		return -EINVAL;
	}

	exec->bo = kvmalloc_array(exec->bo_count,
				    sizeof(struct drm_gem_cma_object *),
				    GFP_KERNEL | __GFP_ZERO);
	if (!exec->bo) {
		DRM_ERROR("Failed to allocate validated BO pointers\n");
		return -ENOMEM;
	}

	handles = kvmalloc_array(exec->bo_count, sizeof(uint32_t), GFP_KERNEL);
	if (!handles) {
		ret = -ENOMEM;
		DRM_ERROR("Failed to allocate incoming GEM handles\n");
		goto fail;
	}

	if (copy_from_user(handles, u64_to_user_ptr(args->bo_handles),
			   exec->bo_count * sizeof(uint32_t))) {
		ret = -EFAULT;
		DRM_ERROR("Failed to copy in GEM handles\n");
		goto fail;
	}

	spin_lock(&file_priv->table_lock);
	for (i = 0; i < exec->bo_count; i++) {
		struct drm_gem_object *bo = idr_find(&file_priv->object_idr,
						     handles[i]);
		if (!bo) {
			DRM_DEBUG("Failed to look up GEM BO %d: %d\n",
				  i, handles[i]);
			ret = -EINVAL;
			spin_unlock(&file_priv->table_lock);
			goto fail;
		}
		drm_gem_object_get(bo);
		exec->bo[i] = (struct drm_gem_cma_object *)bo;
	}
	spin_unlock(&file_priv->table_lock);

fail:
	kvfree(handles);
	return ret;
}

static int
vc4_get_bcl(struct drm_device *dev, struct vc4_exec_info *exec)
{
	struct drm_vc4_submit_cl *args = exec->args;
	void *temp = NULL;
	void *bin;
	int ret = 0;
	uint32_t bin_offset = 0;
	uint32_t shader_rec_offset = roundup(bin_offset + args->bin_cl_size,
					     16);
	uint32_t uniforms_offset = shader_rec_offset + args->shader_rec_size;
	uint32_t exec_size = uniforms_offset + args->uniforms_size;
	uint32_t temp_size = exec_size + (sizeof(struct vc4_shader_state) *
					  args->shader_rec_count);
	struct vc4_bo *bo;

	if (shader_rec_offset < args->bin_cl_size ||
	    uniforms_offset < shader_rec_offset ||
	    exec_size < uniforms_offset ||
	    args->shader_rec_count >= (UINT_MAX /
					  sizeof(struct vc4_shader_state)) ||
	    temp_size < exec_size) {
		DRM_DEBUG("overflow in exec arguments\n");
		ret = -EINVAL;
		goto fail;
	}

	/* Allocate space where we'll store the copied in user command lists
	 * and shader records.
	 *
	 * We don't just copy directly into the BOs because we need to
	 * read the contents back for validation, and I think the
	 * bo->vaddr is uncached access.
	 */
	temp = kvmalloc_array(temp_size, 1, GFP_KERNEL);
	if (!temp) {
		DRM_ERROR("Failed to allocate storage for copying "
			  "in bin/render CLs.\n");
		ret = -ENOMEM;
		goto fail;
	}
	bin = temp + bin_offset;
	exec->shader_rec_u = temp + shader_rec_offset;
	exec->uniforms_u = temp + uniforms_offset;
	exec->shader_state = temp + exec_size;
	exec->shader_state_size = args->shader_rec_count;

	if (copy_from_user(bin,
			   u64_to_user_ptr(args->bin_cl),
			   args->bin_cl_size)) {
		ret = -EFAULT;
		goto fail;
	}

	if (copy_from_user(exec->shader_rec_u,
			   u64_to_user_ptr(args->shader_rec),
			   args->shader_rec_size)) {
		ret = -EFAULT;
		goto fail;
	}

	if (copy_from_user(exec->uniforms_u,
			   u64_to_user_ptr(args->uniforms),
			   args->uniforms_size)) {
		ret = -EFAULT;
		goto fail;
	}

	bo = vc4_bo_create(dev, exec_size, true, VC4_BO_TYPE_BCL);
	if (IS_ERR(bo)) {
		DRM_ERROR("Couldn't allocate BO for binning\n");
		ret = PTR_ERR(bo);
		goto fail;
	}
	exec->exec_bo = &bo->base;

	list_add_tail(&to_vc4_bo(&exec->exec_bo->base)->unref_head,
		      &exec->unref_list);

	exec->ct0ca = exec->exec_bo->paddr + bin_offset;

	exec->bin_u = bin;

	exec->shader_rec_v = exec->exec_bo->vaddr + shader_rec_offset;
	exec->shader_rec_p = exec->exec_bo->paddr + shader_rec_offset;
	exec->shader_rec_size = args->shader_rec_size;

	exec->uniforms_v = exec->exec_bo->vaddr + uniforms_offset;
	exec->uniforms_p = exec->exec_bo->paddr + uniforms_offset;
	exec->uniforms_size = args->uniforms_size;

	ret = vc4_validate_bin_cl(dev,
				  exec->exec_bo->vaddr + bin_offset,
				  bin,
				  exec);
	if (ret)
		goto fail;

	ret = vc4_validate_shader_recs(dev, exec);
	if (ret)
		goto fail;

	/* Block waiting on any previous rendering into the CS's VBO,
	 * IB, or textures, so that pixels are actually written by the
	 * time we try to read them.
	 */
	ret = vc4_wait_for_seqno(dev, exec->bin_dep_seqno, ~0ull, true);

fail:
	kvfree(temp);
	return ret;
}

static void
vc4_complete_exec(struct drm_device *dev, struct vc4_exec_info *exec)
{
	struct vc4_dev *vc4 = to_vc4_dev(dev);
	unsigned long irqflags;
	unsigned i;

	/* If we got force-completed because of GPU reset rather than
	 * through our IRQ handler, signal the fence now.
	 */
	if (exec->fence)
		dma_fence_signal(exec->fence);

	if (exec->bo) {
		for (i = 0; i < exec->bo_count; i++)
			drm_gem_object_put_unlocked(&exec->bo[i]->base);
		kvfree(exec->bo);
	}

	while (!list_empty(&exec->unref_list)) {
		struct vc4_bo *bo = list_first_entry(&exec->unref_list,
						     struct vc4_bo, unref_head);
		list_del(&bo->unref_head);
		drm_gem_object_put_unlocked(&bo->base.base);
	}

	/* Free up the allocation of any bin slots we used. */
	spin_lock_irqsave(&vc4->job_lock, irqflags);
	vc4->bin_alloc_used &= ~exec->bin_slots;
	spin_unlock_irqrestore(&vc4->job_lock, irqflags);

	mutex_lock(&vc4->power_lock);
	if (--vc4->power_refcount == 0) {
		pm_runtime_mark_last_busy(&vc4->v3d->pdev->dev);
		pm_runtime_put_autosuspend(&vc4->v3d->pdev->dev);
	}
	mutex_unlock(&vc4->power_lock);

	kfree(exec);
}

void
vc4_job_handle_completed(struct vc4_dev *vc4)
{
	unsigned long irqflags;
	struct vc4_seqno_cb *cb, *cb_temp;

	spin_lock_irqsave(&vc4->job_lock, irqflags);
	while (!list_empty(&vc4->job_done_list)) {
		struct vc4_exec_info *exec =
			list_first_entry(&vc4->job_done_list,
					 struct vc4_exec_info, head);
		list_del(&exec->head);

		spin_unlock_irqrestore(&vc4->job_lock, irqflags);
		vc4_complete_exec(vc4->dev, exec);
		spin_lock_irqsave(&vc4->job_lock, irqflags);
	}

	list_for_each_entry_safe(cb, cb_temp, &vc4->seqno_cb_list, work.entry) {
		if (cb->seqno <= vc4->finished_seqno) {
			list_del_init(&cb->work.entry);
			schedule_work(&cb->work);
		}
	}

	spin_unlock_irqrestore(&vc4->job_lock, irqflags);
}

static void vc4_seqno_cb_work(struct work_struct *work)
{
	struct vc4_seqno_cb *cb = container_of(work, struct vc4_seqno_cb, work);

	cb->func(cb);
}

int vc4_queue_seqno_cb(struct drm_device *dev,
		       struct vc4_seqno_cb *cb, uint64_t seqno,
		       void (*func)(struct vc4_seqno_cb *cb))
{
	struct vc4_dev *vc4 = to_vc4_dev(dev);
	int ret = 0;
	unsigned long irqflags;

	cb->func = func;
	INIT_WORK(&cb->work, vc4_seqno_cb_work);

	spin_lock_irqsave(&vc4->job_lock, irqflags);
	if (seqno > vc4->finished_seqno) {
		cb->seqno = seqno;
		list_add_tail(&cb->work.entry, &vc4->seqno_cb_list);
	} else {
		schedule_work(&cb->work);
	}
	spin_unlock_irqrestore(&vc4->job_lock, irqflags);

	return ret;
}

/* Scheduled when any job has been completed, this walks the list of
 * jobs that had completed and unrefs their BOs and frees their exec
 * structs.
 */
static void
vc4_job_done_work(struct work_struct *work)
{
	struct vc4_dev *vc4 =
		container_of(work, struct vc4_dev, job_done_work);

	vc4_job_handle_completed(vc4);
}

static int
vc4_wait_for_seqno_ioctl_helper(struct drm_device *dev,
				uint64_t seqno,
				uint64_t *timeout_ns)
{
	unsigned long start = jiffies;
	int ret = vc4_wait_for_seqno(dev, seqno, *timeout_ns, true);

	if ((ret == -EINTR || ret == -ERESTARTSYS) && *timeout_ns != ~0ull) {
		uint64_t delta = jiffies_to_nsecs(jiffies - start);

		if (*timeout_ns >= delta)
			*timeout_ns -= delta;
	}

	return ret;
}

int
vc4_wait_seqno_ioctl(struct drm_device *dev, void *data,
		     struct drm_file *file_priv)
{
	struct drm_vc4_wait_seqno *args = data;

	return vc4_wait_for_seqno_ioctl_helper(dev, args->seqno,
					       &args->timeout_ns);
}

int
vc4_wait_bo_ioctl(struct drm_device *dev, void *data,
		  struct drm_file *file_priv)
{
	int ret;
	struct drm_vc4_wait_bo *args = data;
	struct drm_gem_object *gem_obj;
	struct vc4_bo *bo;

	if (args->pad != 0)
		return -EINVAL;

	gem_obj = drm_gem_object_lookup(file_priv, args->handle);
	if (!gem_obj) {
		DRM_DEBUG("Failed to look up GEM BO %d\n", args->handle);
		return -EINVAL;
	}
	bo = to_vc4_bo(gem_obj);

	ret = vc4_wait_for_seqno_ioctl_helper(dev, bo->seqno,
					      &args->timeout_ns);

	drm_gem_object_put_unlocked(gem_obj);
	return ret;
}

/**
 * vc4_submit_cl_ioctl() - Submits a job (frame) to the VC4.
 * @dev: DRM device
 * @data: ioctl argument
 * @file_priv: DRM file for this fd
 *
 * This is the main entrypoint for userspace to submit a 3D frame to
 * the GPU.  Userspace provides the binner command list (if
 * applicable), and the kernel sets up the render command list to draw
 * to the framebuffer described in the ioctl, using the command lists
 * that the 3D engine's binner will produce.
 */
int
vc4_submit_cl_ioctl(struct drm_device *dev, void *data,
		    struct drm_file *file_priv)
{
	struct vc4_dev *vc4 = to_vc4_dev(dev);
	struct drm_vc4_submit_cl *args = data;
	struct vc4_exec_info *exec;
	struct ww_acquire_ctx acquire_ctx;
	int ret = 0;

	if ((args->flags & ~(VC4_SUBMIT_CL_USE_CLEAR_COLOR |
			     VC4_SUBMIT_CL_FIXED_RCL_ORDER |
			     VC4_SUBMIT_CL_RCL_ORDER_INCREASING_X |
			     VC4_SUBMIT_CL_RCL_ORDER_INCREASING_Y)) != 0) {
		DRM_DEBUG("Unknown flags: 0x%02x\n", args->flags);
		return -EINVAL;
	}

	exec = kcalloc(1, sizeof(*exec), GFP_KERNEL);
	if (!exec) {
		DRM_ERROR("malloc failure on exec struct\n");
		return -ENOMEM;
	}

	mutex_lock(&vc4->power_lock);
	if (vc4->power_refcount++ == 0) {
		ret = pm_runtime_get_sync(&vc4->v3d->pdev->dev);
		if (ret < 0) {
			mutex_unlock(&vc4->power_lock);
			vc4->power_refcount--;
			kfree(exec);
			return ret;
		}
	}
	mutex_unlock(&vc4->power_lock);

	exec->args = args;
	INIT_LIST_HEAD(&exec->unref_list);

	ret = vc4_cl_lookup_bos(dev, file_priv, exec);
	if (ret)
		goto fail;

	if (exec->args->bin_cl_size != 0) {
		ret = vc4_get_bcl(dev, exec);
		if (ret)
			goto fail;
	} else {
		exec->ct0ca = 0;
		exec->ct0ea = 0;
	}

	ret = vc4_get_rcl(dev, exec);
	if (ret)
		goto fail;

	ret = vc4_lock_bo_reservations(dev, exec, &acquire_ctx);
	if (ret)
		goto fail;

	/* Clear this out of the struct we'll be putting in the queue,
	 * since it's part of our stack.
	 */
	exec->args = NULL;

	ret = vc4_queue_submit(dev, exec, &acquire_ctx);
	if (ret)
		goto fail;

	/* Return the seqno for our job. */
	args->seqno = vc4->emit_seqno;

	return 0;

fail:
	vc4_complete_exec(vc4->dev, exec);

	return ret;
}

void
vc4_gem_init(struct drm_device *dev)
{
	struct vc4_dev *vc4 = to_vc4_dev(dev);

	vc4->dma_fence_context = dma_fence_context_alloc(1);

	INIT_LIST_HEAD(&vc4->bin_job_list);
	INIT_LIST_HEAD(&vc4->render_job_list);
	INIT_LIST_HEAD(&vc4->job_done_list);
	INIT_LIST_HEAD(&vc4->seqno_cb_list);
	spin_lock_init(&vc4->job_lock);

	INIT_WORK(&vc4->hangcheck.reset_work, vc4_reset_work);
	setup_timer(&vc4->hangcheck.timer,
		    vc4_hangcheck_elapsed,
		    (unsigned long)dev);

	INIT_WORK(&vc4->job_done_work, vc4_job_done_work);

	mutex_init(&vc4->power_lock);
}

void
vc4_gem_destroy(struct drm_device *dev)
{
	struct vc4_dev *vc4 = to_vc4_dev(dev);

	/* Waiting for exec to finish would need to be done before
	 * unregistering V3D.
	 */
	WARN_ON(vc4->emit_seqno != vc4->finished_seqno);

	/* V3D should already have disabled its interrupt and cleared
	 * the overflow allocation registers.  Now free the object.
	 */
	if (vc4->bin_bo) {
		drm_gem_object_put_unlocked(&vc4->bin_bo->base.base);
		vc4->bin_bo = NULL;
	}

	if (vc4->hang_state)
		vc4_free_hang_state(dev, vc4->hang_state);
}