summaryrefslogtreecommitdiffstats
path: root/drivers/gpu/drm/radeon/radeon_fence.c
blob: df09ca7c488949896f68ff1d78ad7e152f0fd9a5 (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
/*
 * Copyright 2009 Jerome Glisse.
 * All Rights Reserved.
 *
 * Permission is hereby granted, free of charge, to any person obtaining a
 * copy of this software and associated documentation files (the
 * "Software"), to deal in the Software without restriction, including
 * without limitation the rights to use, copy, modify, merge, publish,
 * distribute, sub license, and/or sell copies of the Software, and to
 * permit persons to whom the Software is furnished to do so, subject to
 * the following conditions:
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
 * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
 * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
 * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
 * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
 * USE OR OTHER DEALINGS IN THE SOFTWARE.
 *
 * The above copyright notice and this permission notice (including the
 * next paragraph) shall be included in all copies or substantial portions
 * of the Software.
 *
 */
/*
 * Authors:
 *    Jerome Glisse <glisse@freedesktop.org>
 *    Dave Airlie
 */
#include <linux/seq_file.h>
#include <linux/atomic.h>
#include <linux/wait.h>
#include <linux/kref.h>
#include <linux/slab.h>
#include <linux/firmware.h>
#include <drm/drmP.h>
#include "radeon_reg.h"
#include "radeon.h"
#include "radeon_trace.h"

/*
 * Fences
 * Fences mark an event in the GPUs pipeline and are used
 * for GPU/CPU synchronization.  When the fence is written,
 * it is expected that all buffers associated with that fence
 * are no longer in use by the associated ring on the GPU and
 * that the the relevant GPU caches have been flushed.  Whether
 * we use a scratch register or memory location depends on the asic
 * and whether writeback is enabled.
 */

/**
 * radeon_fence_write - write a fence value
 *
 * @rdev: radeon_device pointer
 * @seq: sequence number to write
 * @ring: ring index the fence is associated with
 *
 * Writes a fence value to memory or a scratch register (all asics).
 */
static void radeon_fence_write(struct radeon_device *rdev, u32 seq, int ring)
{
	struct radeon_fence_driver *drv = &rdev->fence_drv[ring];
	if (likely(rdev->wb.enabled || !drv->scratch_reg)) {
		if (drv->cpu_addr) {
			*drv->cpu_addr = cpu_to_le32(seq);
		}
	} else {
		WREG32(drv->scratch_reg, seq);
	}
}

/**
 * radeon_fence_read - read a fence value
 *
 * @rdev: radeon_device pointer
 * @ring: ring index the fence is associated with
 *
 * Reads a fence value from memory or a scratch register (all asics).
 * Returns the value of the fence read from memory or register.
 */
static u32 radeon_fence_read(struct radeon_device *rdev, int ring)
{
	struct radeon_fence_driver *drv = &rdev->fence_drv[ring];
	u32 seq = 0;

	if (likely(rdev->wb.enabled || !drv->scratch_reg)) {
		if (drv->cpu_addr) {
			seq = le32_to_cpu(*drv->cpu_addr);
		} else {
			seq = lower_32_bits(atomic64_read(&drv->last_seq));
		}
	} else {
		seq = RREG32(drv->scratch_reg);
	}
	return seq;
}

/**
 * radeon_fence_schedule_check - schedule lockup check
 *
 * @rdev: radeon_device pointer
 * @ring: ring index we should work with
 *
 * Queues a delayed work item to check for lockups.
 */
static void radeon_fence_schedule_check(struct radeon_device *rdev, int ring)
{
	/*
	 * Do not reset the timer here with mod_delayed_work,
	 * this can livelock in an interaction with TTM delayed destroy.
	 */
	queue_delayed_work(system_power_efficient_wq,
			   &rdev->fence_drv[ring].lockup_work,
			   RADEON_FENCE_JIFFIES_TIMEOUT);
}

/**
 * radeon_fence_emit - emit a fence on the requested ring
 *
 * @rdev: radeon_device pointer
 * @fence: radeon fence object
 * @ring: ring index the fence is associated with
 *
 * Emits a fence command on the requested ring (all asics).
 * Returns 0 on success, -ENOMEM on failure.
 */
int radeon_fence_emit(struct radeon_device *rdev,
		      struct radeon_fence **fence,
		      int ring)
{
	u64 seq = ++rdev->fence_drv[ring].sync_seq[ring];

	/* we are protected by the ring emission mutex */
	*fence = kmalloc(sizeof(struct radeon_fence), GFP_KERNEL);
	if ((*fence) == NULL) {
		return -ENOMEM;
	}
	(*fence)->rdev = rdev;
	(*fence)->seq = seq;
	(*fence)->ring = ring;
	(*fence)->is_vm_update = false;
	fence_init(&(*fence)->base, &radeon_fence_ops,
		   &rdev->fence_queue.lock, rdev->fence_context + ring, seq);
	radeon_fence_ring_emit(rdev, ring, *fence);
	trace_radeon_fence_emit(rdev->ddev, ring, (*fence)->seq);
	radeon_fence_schedule_check(rdev, ring);
	return 0;
}

/**
 * radeon_fence_check_signaled - callback from fence_queue
 *
 * this function is called with fence_queue lock held, which is also used
 * for the fence locking itself, so unlocked variants are used for
 * fence_signal, and remove_wait_queue.
 */
static int radeon_fence_check_signaled(wait_queue_t *wait, unsigned mode, int flags, void *key)
{
	struct radeon_fence *fence;
	u64 seq;

	fence = container_of(wait, struct radeon_fence, fence_wake);

	/*
	 * We cannot use radeon_fence_process here because we're already
	 * in the waitqueue, in a call from wake_up_all.
	 */
	seq = atomic64_read(&fence->rdev->fence_drv[fence->ring].last_seq);
	if (seq >= fence->seq) {
		int ret = fence_signal_locked(&fence->base);

		if (!ret)
			FENCE_TRACE(&fence->base, "signaled from irq context\n");
		else
			FENCE_TRACE(&fence->base, "was already signaled\n");

		radeon_irq_kms_sw_irq_put(fence->rdev, fence->ring);
		__remove_wait_queue(&fence->rdev->fence_queue, &fence->fence_wake);
		fence_put(&fence->base);
	} else
		FENCE_TRACE(&fence->base, "pending\n");
	return 0;
}

/**
 * radeon_fence_activity - check for fence activity
 *
 * @rdev: radeon_device pointer
 * @ring: ring index the fence is associated with
 *
 * Checks the current fence value and calculates the last
 * signalled fence value. Returns true if activity occured
 * on the ring, and the fence_queue should be waken up.
 */
static bool radeon_fence_activity(struct radeon_device *rdev, int ring)
{
	uint64_t seq, last_seq, last_emitted;
	unsigned count_loop = 0;
	bool wake = false;

	/* Note there is a scenario here for an infinite loop but it's
	 * very unlikely to happen. For it to happen, the current polling
	 * process need to be interrupted by another process and another
	 * process needs to update the last_seq btw the atomic read and
	 * xchg of the current process.
	 *
	 * More over for this to go in infinite loop there need to be
	 * continuously new fence signaled ie radeon_fence_read needs
	 * to return a different value each time for both the currently
	 * polling process and the other process that xchg the last_seq
	 * btw atomic read and xchg of the current process. And the
	 * value the other process set as last seq must be higher than
	 * the seq value we just read. Which means that current process
	 * need to be interrupted after radeon_fence_read and before
	 * atomic xchg.
	 *
	 * To be even more safe we count the number of time we loop and
	 * we bail after 10 loop just accepting the fact that we might
	 * have temporarly set the last_seq not to the true real last
	 * seq but to an older one.
	 */
	last_seq = atomic64_read(&rdev->fence_drv[ring].last_seq);
	do {
		last_emitted = rdev->fence_drv[ring].sync_seq[ring];
		seq = radeon_fence_read(rdev, ring);
		seq |= last_seq & 0xffffffff00000000LL;
		if (seq < last_seq) {
			seq &= 0xffffffff;
			seq |= last_emitted & 0xffffffff00000000LL;
		}

		if (seq <= last_seq || seq > last_emitted) {
			break;
		}
		/* If we loop over we don't want to return without
		 * checking if a fence is signaled as it means that the
		 * seq we just read is different from the previous on.
		 */
		wake = true;
		last_seq = seq;
		if ((count_loop++) > 10) {
			/* We looped over too many time leave with the
			 * fact that we might have set an older fence
			 * seq then the current real last seq as signaled
			 * by the hw.
			 */
			break;
		}
	} while (atomic64_xchg(&rdev->fence_drv[ring].last_seq, seq) > seq);

	if (seq < last_emitted)
		radeon_fence_schedule_check(rdev, ring);

	return wake;
}

/**
 * radeon_fence_check_lockup - check for hardware lockup
 *
 * @work: delayed work item
 *
 * Checks for fence activity and if there is none probe
 * the hardware if a lockup occured.
 */
static void radeon_fence_check_lockup(struct work_struct *work)
{
	struct radeon_fence_driver *fence_drv;
	struct radeon_device *rdev;
	int ring;

	fence_drv = container_of(work, struct radeon_fence_driver,
				 lockup_work.work);
	rdev = fence_drv->rdev;
	ring = fence_drv - &rdev->fence_drv[0];

	if (!down_read_trylock(&rdev->exclusive_lock)) {
		/* just reschedule the check if a reset is going on */
		radeon_fence_schedule_check(rdev, ring);
		return;
	}

	if (fence_drv->delayed_irq && rdev->ddev->irq_enabled) {
		unsigned long irqflags;

		fence_drv->delayed_irq = false;
		spin_lock_irqsave(&rdev->irq.lock, irqflags);
		radeon_irq_set(rdev);
		spin_unlock_irqrestore(&rdev->irq.lock, irqflags);
	}

	if (radeon_fence_activity(rdev, ring))
		wake_up_all(&rdev->fence_queue);

	else if (radeon_ring_is_lockup(rdev, ring, &rdev->ring[ring])) {

		/* good news we believe it's a lockup */
		dev_warn(rdev->dev, "GPU lockup (current fence id "
			 "0x%016llx last fence id 0x%016llx on ring %d)\n",
			 (uint64_t)atomic64_read(&fence_drv->last_seq),
			 fence_drv->sync_seq[ring], ring);

		/* remember that we need an reset */
		rdev->needs_reset = true;
		wake_up_all(&rdev->fence_queue);
	}
	up_read(&rdev->exclusive_lock);
}

/**
 * radeon_fence_process - process a fence
 *
 * @rdev: radeon_device pointer
 * @ring: ring index the fence is associated with
 *
 * Checks the current fence value and wakes the fence queue
 * if the sequence number has increased (all asics).
 */
void radeon_fence_process(struct radeon_device *rdev, int ring)
{
	if (radeon_fence_activity(rdev, ring))
		wake_up_all(&rdev->fence_queue);
}

/**
 * radeon_fence_seq_signaled - check if a fence sequence number has signaled
 *
 * @rdev: radeon device pointer
 * @seq: sequence number
 * @ring: ring index the fence is associated with
 *
 * Check if the last signaled fence sequnce number is >= the requested
 * sequence number (all asics).
 * Returns true if the fence has signaled (current fence value
 * is >= requested value) or false if it has not (current fence
 * value is < the requested value.  Helper function for
 * radeon_fence_signaled().
 */
static bool radeon_fence_seq_signaled(struct radeon_device *rdev,
				      u64 seq, unsigned ring)
{
	if (atomic64_read(&rdev->fence_drv[ring].last_seq) >= seq) {
		return true;
	}
	/* poll new last sequence at least once */
	radeon_fence_process(rdev, ring);
	if (atomic64_read(&rdev->fence_drv[ring].last_seq) >= seq) {
		return true;
	}
	return false;
}

static bool radeon_fence_is_signaled(struct fence *f)
{
	struct radeon_fence *fence = to_radeon_fence(f);
	struct radeon_device *rdev = fence->rdev;
	unsigned ring = fence->ring;
	u64 seq = fence->seq;

	if (atomic64_read(&rdev->fence_drv[ring].last_seq) >= seq) {
		return true;
	}

	if (down_read_trylock(&rdev->exclusive_lock)) {
		radeon_fence_process(rdev, ring);
		up_read(&rdev->exclusive_lock);

		if (atomic64_read(&rdev->fence_drv[ring].last_seq) >= seq) {
			return true;
		}
	}
	return false;
}

/**
 * radeon_fence_enable_signaling - enable signalling on fence
 * @fence: fence
 *
 * This function is called with fence_queue lock held, and adds a callback
 * to fence_queue that checks if this fence is signaled, and if so it
 * signals the fence and removes itself.
 */
static bool radeon_fence_enable_signaling(struct fence *f)
{
	struct radeon_fence *fence = to_radeon_fence(f);
	struct radeon_device *rdev = fence->rdev;

	if (atomic64_read(&rdev->fence_drv[fence->ring].last_seq) >= fence->seq)
		return false;

	if (down_read_trylock(&rdev->exclusive_lock)) {
		radeon_irq_kms_sw_irq_get(rdev, fence->ring);

		if (radeon_fence_activity(rdev, fence->ring))
			wake_up_all_locked(&rdev->fence_queue);

		/* did fence get signaled after we enabled the sw irq? */
		if (atomic64_read(&rdev->fence_drv[fence->ring].last_seq) >= fence->seq) {
			radeon_irq_kms_sw_irq_put(rdev, fence->ring);
			up_read(&rdev->exclusive_lock);
			return false;
		}

		up_read(&rdev->exclusive_lock);
	} else {
		/* we're probably in a lockup, lets not fiddle too much */
		if (radeon_irq_kms_sw_irq_get_delayed(rdev, fence->ring))
			rdev->fence_drv[fence->ring].delayed_irq = true;
		radeon_fence_schedule_check(rdev, fence->ring);
	}

	fence->fence_wake.flags = 0;
	fence->fence_wake.private = NULL;
	fence->fence_wake.func = radeon_fence_check_signaled;
	__add_wait_queue(&rdev->fence_queue, &fence->fence_wake);
	fence_get(f);

	FENCE_TRACE(&fence->base, "armed on ring %i!\n", fence->ring);
	return true;
}

/**
 * radeon_fence_signaled - check if a fence has signaled
 *
 * @fence: radeon fence object
 *
 * Check if the requested fence has signaled (all asics).
 * Returns true if the fence has signaled or false if it has not.
 */
bool radeon_fence_signaled(struct radeon_fence *fence)
{
	if (!fence)
		return true;

	if (radeon_fence_seq_signaled(fence->rdev, fence->seq, fence->ring)) {
		int ret;

		ret = fence_signal(&fence->base);
		if (!ret)
			FENCE_TRACE(&fence->base, "signaled from radeon_fence_signaled\n");
		return true;
	}
	return false;
}

/**
 * radeon_fence_any_seq_signaled - check if any sequence number is signaled
 *
 * @rdev: radeon device pointer
 * @seq: sequence numbers
 *
 * Check if the last signaled fence sequnce number is >= the requested
 * sequence number (all asics).
 * Returns true if any has signaled (current value is >= requested value)
 * or false if it has not. Helper function for radeon_fence_wait_seq.
 */
static bool radeon_fence_any_seq_signaled(struct radeon_device *rdev, u64 *seq)
{
	unsigned i;

	for (i = 0; i < RADEON_NUM_RINGS; ++i) {
		if (seq[i] && radeon_fence_seq_signaled(rdev, seq[i], i))
			return true;
	}
	return false;
}

/**
 * radeon_fence_wait_seq_timeout - wait for a specific sequence numbers
 *
 * @rdev: radeon device pointer
 * @target_seq: sequence number(s) we want to wait for
 * @intr: use interruptable sleep
 * @timeout: maximum time to wait, or MAX_SCHEDULE_TIMEOUT for infinite wait
 *
 * Wait for the requested sequence number(s) to be written by any ring
 * (all asics).  Sequnce number array is indexed by ring id.
 * @intr selects whether to use interruptable (true) or non-interruptable
 * (false) sleep when waiting for the sequence number.  Helper function
 * for radeon_fence_wait_*().
 * Returns remaining time if the sequence number has passed, 0 when
 * the wait timeout, or an error for all other cases.
 * -EDEADLK is returned when a GPU lockup has been detected.
 */
static long radeon_fence_wait_seq_timeout(struct radeon_device *rdev,
					  u64 *target_seq, bool intr,
					  long timeout)
{
	long r;
	int i;

	if (radeon_fence_any_seq_signaled(rdev, target_seq))
		return timeout;

	/* enable IRQs and tracing */
	for (i = 0; i < RADEON_NUM_RINGS; ++i) {
		if (!target_seq[i])
			continue;

		trace_radeon_fence_wait_begin(rdev->ddev, i, target_seq[i]);
		radeon_irq_kms_sw_irq_get(rdev, i);
	}

	if (intr) {
		r = wait_event_interruptible_timeout(rdev->fence_queue, (
			radeon_fence_any_seq_signaled(rdev, target_seq)
			 || rdev->needs_reset), timeout);
	} else {
		r = wait_event_timeout(rdev->fence_queue, (
			radeon_fence_any_seq_signaled(rdev, target_seq)
			 || rdev->needs_reset), timeout);
	}

	if (rdev->needs_reset)
		r = -EDEADLK;

	for (i = 0; i < RADEON_NUM_RINGS; ++i) {
		if (!target_seq[i])
			continue;

		radeon_irq_kms_sw_irq_put(rdev, i);
		trace_radeon_fence_wait_end(rdev->ddev, i, target_seq[i]);
	}

	return r;
}

/**
 * radeon_fence_wait - wait for a fence to signal
 *
 * @fence: radeon fence object
 * @intr: use interruptible sleep
 *
 * Wait for the requested fence to signal (all asics).
 * @intr selects whether to use interruptable (true) or non-interruptable
 * (false) sleep when waiting for the fence.
 * Returns 0 if the fence has passed, error for all other cases.
 */
int radeon_fence_wait(struct radeon_fence *fence, bool intr)
{
	uint64_t seq[RADEON_NUM_RINGS] = {};
	long r;

	/*
	 * This function should not be called on !radeon fences.
	 * If this is the case, it would mean this function can
	 * also be called on radeon fences belonging to another card.
	 * exclusive_lock is not held in that case.
	 */
	if (WARN_ON_ONCE(!to_radeon_fence(&fence->base)))
		return fence_wait(&fence->base, intr);

	seq[fence->ring] = fence->seq;
	r = radeon_fence_wait_seq_timeout(fence->rdev, seq, intr, MAX_SCHEDULE_TIMEOUT);
	if (r < 0) {
		return r;
	}

	r = fence_signal(&fence->base);
	if (!r)
		FENCE_TRACE(&fence->base, "signaled from fence_wait\n");
	return 0;
}

/**
 * radeon_fence_wait_any - wait for a fence to signal on any ring
 *
 * @rdev: radeon device pointer
 * @fences: radeon fence object(s)
 * @intr: use interruptable sleep
 *
 * Wait for any requested fence to signal (all asics).  Fence
 * array is indexed by ring id.  @intr selects whether to use
 * interruptable (true) or non-interruptable (false) sleep when
 * waiting for the fences. Used by the suballocator.
 * Returns 0 if any fence has passed, error for all other cases.
 */
int radeon_fence_wait_any(struct radeon_device *rdev,
			  struct radeon_fence **fences,
			  bool intr)
{
	uint64_t seq[RADEON_NUM_RINGS];
	unsigned i, num_rings = 0;
	long r;

	for (i = 0; i < RADEON_NUM_RINGS; ++i) {
		seq[i] = 0;

		if (!fences[i]) {
			continue;
		}

		seq[i] = fences[i]->seq;
		++num_rings;
	}

	/* nothing to wait for ? */
	if (num_rings == 0)
		return -ENOENT;

	r = radeon_fence_wait_seq_timeout(rdev, seq, intr, MAX_SCHEDULE_TIMEOUT);
	if (r < 0) {
		return r;
	}
	return 0;
}

/**
 * radeon_fence_wait_next - wait for the next fence to signal
 *
 * @rdev: radeon device pointer
 * @ring: ring index the fence is associated with
 *
 * Wait for the next fence on the requested ring to signal (all asics).
 * Returns 0 if the next fence has passed, error for all other cases.
 * Caller must hold ring lock.
 */
int radeon_fence_wait_next(struct radeon_device *rdev, int ring)
{
	uint64_t seq[RADEON_NUM_RINGS] = {};
	long r;

	seq[ring] = atomic64_read(&rdev->fence_drv[ring].last_seq) + 1ULL;
	if (seq[ring] >= rdev->fence_drv[ring].sync_seq[ring]) {
		/* nothing to wait for, last_seq is
		   already the last emited fence */
		return -ENOENT;
	}
	r = radeon_fence_wait_seq_timeout(rdev, seq, false, MAX_SCHEDULE_TIMEOUT);
	if (r < 0)
		return r;
	return 0;
}

/**
 * radeon_fence_wait_empty - wait for all fences to signal
 *
 * @rdev: radeon device pointer
 * @ring: ring index the fence is associated with
 *
 * Wait for all fences on the requested ring to signal (all asics).
 * Returns 0 if the fences have passed, error for all other cases.
 * Caller must hold ring lock.
 */
int radeon_fence_wait_empty(struct radeon_device *rdev, int ring)
{
	uint64_t seq[RADEON_NUM_RINGS] = {};
	long r;

	seq[ring] = rdev->fence_drv[ring].sync_seq[ring];
	if (!seq[ring])
		return 0;

	r = radeon_fence_wait_seq_timeout(rdev, seq, false, MAX_SCHEDULE_TIMEOUT);
	if (r < 0) {
		if (r == -EDEADLK)
			return -EDEADLK;

		dev_err(rdev->dev, "error waiting for ring[%d] to become idle (%ld)\n",
			ring, r);
	}
	return 0;
}

/**
 * radeon_fence_ref - take a ref on a fence
 *
 * @fence: radeon fence object
 *
 * Take a reference on a fence (all asics).
 * Returns the fence.
 */
struct radeon_fence *radeon_fence_ref(struct radeon_fence *fence)
{
	fence_get(&fence->base);
	return fence;
}

/**
 * radeon_fence_unref - remove a ref on a fence
 *
 * @fence: radeon fence object
 *
 * Remove a reference on a fence (all asics).
 */
void radeon_fence_unref(struct radeon_fence **fence)
{
	struct radeon_fence *tmp = *fence;

	*fence = NULL;
	if (tmp) {
		fence_put(&tmp->base);
	}
}

/**
 * radeon_fence_count_emitted - get the count of emitted fences
 *
 * @rdev: radeon device pointer
 * @ring: ring index the fence is associated with
 *
 * Get the number of fences emitted on the requested ring (all asics).
 * Returns the number of emitted fences on the ring.  Used by the
 * dynpm code to ring track activity.
 */
unsigned radeon_fence_count_emitted(struct radeon_device *rdev, int ring)
{
	uint64_t emitted;

	/* We are not protected by ring lock when reading the last sequence
	 * but it's ok to report slightly wrong fence count here.
	 */
	radeon_fence_process(rdev, ring);
	emitted = rdev->fence_drv[ring].sync_seq[ring]
		- atomic64_read(&rdev->fence_drv[ring].last_seq);
	/* to avoid 32bits warp around */
	if (emitted > 0x10000000) {
		emitted = 0x10000000;
	}
	return (unsigned)emitted;
}

/**
 * radeon_fence_need_sync - do we need a semaphore
 *
 * @fence: radeon fence object
 * @dst_ring: which ring to check against
 *
 * Check if the fence needs to be synced against another ring
 * (all asics).  If so, we need to emit a semaphore.
 * Returns true if we need to sync with another ring, false if
 * not.
 */
bool radeon_fence_need_sync(struct radeon_fence *fence, int dst_ring)
{
	struct radeon_fence_driver *fdrv;

	if (!fence) {
		return false;
	}

	if (fence->ring == dst_ring) {
		return false;
	}

	/* we are protected by the ring mutex */
	fdrv = &fence->rdev->fence_drv[dst_ring];
	if (fence->seq <= fdrv->sync_seq[fence->ring]) {
		return false;
	}

	return true;
}

/**
 * radeon_fence_note_sync - record the sync point
 *
 * @fence: radeon fence object
 * @dst_ring: which ring to check against
 *
 * Note the sequence number at which point the fence will
 * be synced with the requested ring (all asics).
 */
void radeon_fence_note_sync(struct radeon_fence *fence, int dst_ring)
{
	struct radeon_fence_driver *dst, *src;
	unsigned i;

	if (!fence) {
		return;
	}

	if (fence->ring == dst_ring) {
		return;
	}

	/* we are protected by the ring mutex */
	src = &fence->rdev->fence_drv[fence->ring];
	dst = &fence->rdev->fence_drv[dst_ring];
	for (i = 0; i < RADEON_NUM_RINGS; ++i) {
		if (i == dst_ring) {
			continue;
		}
		dst->sync_seq[i] = max(dst->sync_seq[i], src->sync_seq[i]);
	}
}

/**
 * radeon_fence_driver_start_ring - make the fence driver
 * ready for use on the requested ring.
 *
 * @rdev: radeon device pointer
 * @ring: ring index to start the fence driver on
 *
 * Make the fence driver ready for processing (all asics).
 * Not all asics have all rings, so each asic will only
 * start the fence driver on the rings it has.
 * Returns 0 for success, errors for failure.
 */
int radeon_fence_driver_start_ring(struct radeon_device *rdev, int ring)
{
	uint64_t index;
	int r;

	radeon_scratch_free(rdev, rdev->fence_drv[ring].scratch_reg);
	if (rdev->wb.use_event || !radeon_ring_supports_scratch_reg(rdev, &rdev->ring[ring])) {
		rdev->fence_drv[ring].scratch_reg = 0;
		if (ring != R600_RING_TYPE_UVD_INDEX) {
			index = R600_WB_EVENT_OFFSET + ring * 4;
			rdev->fence_drv[ring].cpu_addr = &rdev->wb.wb[index/4];
			rdev->fence_drv[ring].gpu_addr = rdev->wb.gpu_addr +
							 index;

		} else {
			/* put fence directly behind firmware */
			index = ALIGN(rdev->uvd_fw->size, 8);
			rdev->fence_drv[ring].cpu_addr = rdev->uvd.cpu_addr + index;
			rdev->fence_drv[ring].gpu_addr = rdev->uvd.gpu_addr + index;
		}

	} else {
		r = radeon_scratch_get(rdev, &rdev->fence_drv[ring].scratch_reg);
		if (r) {
			dev_err(rdev->dev, "fence failed to get scratch register\n");
			return r;
		}
		index = RADEON_WB_SCRATCH_OFFSET +
			rdev->fence_drv[ring].scratch_reg -
			rdev->scratch.reg_base;
		rdev->fence_drv[ring].cpu_addr = &rdev->wb.wb[index/4];
		rdev->fence_drv[ring].gpu_addr = rdev->wb.gpu_addr + index;
	}
	radeon_fence_write(rdev, atomic64_read(&rdev->fence_drv[ring].last_seq), ring);
	rdev->fence_drv[ring].initialized = true;
	dev_info(rdev->dev, "fence driver on ring %d use gpu addr 0x%016llx and cpu addr 0x%p\n",
		 ring, rdev->fence_drv[ring].gpu_addr, rdev->fence_drv[ring].cpu_addr);
	return 0;
}

/**
 * radeon_fence_driver_init_ring - init the fence driver
 * for the requested ring.
 *
 * @rdev: radeon device pointer
 * @ring: ring index to start the fence driver on
 *
 * Init the fence driver for the requested ring (all asics).
 * Helper function for radeon_fence_driver_init().
 */
static void radeon_fence_driver_init_ring(struct radeon_device *rdev, int ring)
{
	int i;

	rdev->fence_drv[ring].scratch_reg = -1;
	rdev->fence_drv[ring].cpu_addr = NULL;
	rdev->fence_drv[ring].gpu_addr = 0;
	for (i = 0; i < RADEON_NUM_RINGS; ++i)
		rdev->fence_drv[ring].sync_seq[i] = 0;
	atomic64_set(&rdev->fence_drv[ring].last_seq, 0);
	rdev->fence_drv[ring].initialized = false;
	INIT_DELAYED_WORK(&rdev->fence_drv[ring].lockup_work,
			  radeon_fence_check_lockup);
	rdev->fence_drv[ring].rdev = rdev;
}

/**
 * radeon_fence_driver_init - init the fence driver
 * for all possible rings.
 *
 * @rdev: radeon device pointer
 *
 * Init the fence driver for all possible rings (all asics).
 * Not all asics have all rings, so each asic will only
 * start the fence driver on the rings it has using
 * radeon_fence_driver_start_ring().
 * Returns 0 for success.
 */
int radeon_fence_driver_init(struct radeon_device *rdev)
{
	int ring;

	init_waitqueue_head(&rdev->fence_queue);
	for (ring = 0; ring < RADEON_NUM_RINGS; ring++) {
		radeon_fence_driver_init_ring(rdev, ring);
	}
	if (radeon_debugfs_fence_init(rdev)) {
		dev_err(rdev->dev, "fence debugfs file creation failed\n");
	}
	return 0;
}

/**
 * radeon_fence_driver_fini - tear down the fence driver
 * for all possible rings.
 *
 * @rdev: radeon device pointer
 *
 * Tear down the fence driver for all possible rings (all asics).
 */
void radeon_fence_driver_fini(struct radeon_device *rdev)
{
	int ring, r;

	mutex_lock(&rdev->ring_lock);
	for (ring = 0; ring < RADEON_NUM_RINGS; ring++) {
		if (!rdev->fence_drv[ring].initialized)
			continue;
		r = radeon_fence_wait_empty(rdev, ring);
		if (r) {
			/* no need to trigger GPU reset as we are unloading */
			radeon_fence_driver_force_completion(rdev, ring);
		}
		cancel_delayed_work_sync(&rdev->fence_drv[ring].lockup_work);
		wake_up_all(&rdev->fence_queue);
		radeon_scratch_free(rdev, rdev->fence_drv[ring].scratch_reg);
		rdev->fence_drv[ring].initialized = false;
	}
	mutex_unlock(&rdev->ring_lock);
}

/**
 * radeon_fence_driver_force_completion - force all fence waiter to complete
 *
 * @rdev: radeon device pointer
 * @ring: the ring to complete
 *
 * In case of GPU reset failure make sure no process keep waiting on fence
 * that will never complete.
 */
void radeon_fence_driver_force_completion(struct radeon_device *rdev, int ring)
{
	if (rdev->fence_drv[ring].initialized) {
		radeon_fence_write(rdev, rdev->fence_drv[ring].sync_seq[ring], ring);
		cancel_delayed_work_sync(&rdev->fence_drv[ring].lockup_work);
	}
}


/*
 * Fence debugfs
 */
#if defined(CONFIG_DEBUG_FS)
static int radeon_debugfs_fence_info(struct seq_file *m, void *data)
{
	struct drm_info_node *node = (struct drm_info_node *)m->private;
	struct drm_device *dev = node->minor->dev;
	struct radeon_device *rdev = dev->dev_private;
	int i, j;

	for (i = 0; i < RADEON_NUM_RINGS; ++i) {
		if (!rdev->fence_drv[i].initialized)
			continue;

		radeon_fence_process(rdev, i);

		seq_printf(m, "--- ring %d ---\n", i);
		seq_printf(m, "Last signaled fence 0x%016llx\n",
			   (unsigned long long)atomic64_read(&rdev->fence_drv[i].last_seq));
		seq_printf(m, "Last emitted        0x%016llx\n",
			   rdev->fence_drv[i].sync_seq[i]);

		for (j = 0; j < RADEON_NUM_RINGS; ++j) {
			if (i != j && rdev->fence_drv[j].initialized)
				seq_printf(m, "Last sync to ring %d 0x%016llx\n",
					   j, rdev->fence_drv[i].sync_seq[j]);
		}
	}
	return 0;
}

/**
 * radeon_debugfs_gpu_reset - manually trigger a gpu reset
 *
 * Manually trigger a gpu reset at the next fence wait.
 */
static int radeon_debugfs_gpu_reset(struct seq_file *m, void *data)
{
	struct drm_info_node *node = (struct drm_info_node *) m->private;
	struct drm_device *dev = node->minor->dev;
	struct radeon_device *rdev = dev->dev_private;

	down_read(&rdev->exclusive_lock);
	seq_printf(m, "%d\n", rdev->needs_reset);
	rdev->needs_reset = true;
	wake_up_all(&rdev->fence_queue);
	up_read(&rdev->exclusive_lock);

	return 0;
}

static struct drm_info_list radeon_debugfs_fence_list[] = {
	{"radeon_fence_info", &radeon_debugfs_fence_info, 0, NULL},
	{"radeon_gpu_reset", &radeon_debugfs_gpu_reset, 0, NULL}
};
#endif

int radeon_debugfs_fence_init(struct radeon_device *rdev)
{
#if defined(CONFIG_DEBUG_FS)
	return radeon_debugfs_add_files(rdev, radeon_debugfs_fence_list, 2);
#else
	return 0;
#endif
}

static const char *radeon_fence_get_driver_name(struct fence *fence)
{
	return "radeon";
}

static const char *radeon_fence_get_timeline_name(struct fence *f)
{
	struct radeon_fence *fence = to_radeon_fence(f);
	switch (fence->ring) {
	case RADEON_RING_TYPE_GFX_INDEX: return "radeon.gfx";
	case CAYMAN_RING_TYPE_CP1_INDEX: return "radeon.cp1";
	case CAYMAN_RING_TYPE_CP2_INDEX: return "radeon.cp2";
	case R600_RING_TYPE_DMA_INDEX: return "radeon.dma";
	case CAYMAN_RING_TYPE_DMA1_INDEX: return "radeon.dma1";
	case R600_RING_TYPE_UVD_INDEX: return "radeon.uvd";
	case TN_RING_TYPE_VCE1_INDEX: return "radeon.vce1";
	case TN_RING_TYPE_VCE2_INDEX: return "radeon.vce2";
	default: WARN_ON_ONCE(1); return "radeon.unk";
	}
}

static inline bool radeon_test_signaled(struct radeon_fence *fence)
{
	return test_bit(FENCE_FLAG_SIGNALED_BIT, &fence->base.flags);
}

struct radeon_wait_cb {
	struct fence_cb base;
	struct task_struct *task;
};

static void
radeon_fence_wait_cb(struct fence *fence, struct fence_cb *cb)
{
	struct radeon_wait_cb *wait =
		container_of(cb, struct radeon_wait_cb, base);

	wake_up_process(wait->task);
}

static signed long radeon_fence_default_wait(struct fence *f, bool intr,
					     signed long t)
{
	struct radeon_fence *fence = to_radeon_fence(f);
	struct radeon_device *rdev = fence->rdev;
	struct radeon_wait_cb cb;

	cb.task = current;

	if (fence_add_callback(f, &cb.base, radeon_fence_wait_cb))
		return t;

	while (t > 0) {
		if (intr)
			set_current_state(TASK_INTERRUPTIBLE);
		else
			set_current_state(TASK_UNINTERRUPTIBLE);

		/*
		 * radeon_test_signaled must be called after
		 * set_current_state to prevent a race with wake_up_process
		 */
		if (radeon_test_signaled(fence))
			break;

		if (rdev->needs_reset) {
			t = -EDEADLK;
			break;
		}

		t = schedule_timeout(t);

		if (t > 0 && intr && signal_pending(current))
			t = -ERESTARTSYS;
	}

	__set_current_state(TASK_RUNNING);
	fence_remove_callback(f, &cb.base);

	return t;
}

const struct fence_ops radeon_fence_ops = {
	.get_driver_name = radeon_fence_get_driver_name,
	.get_timeline_name = radeon_fence_get_timeline_name,
	.enable_signaling = radeon_fence_enable_signaling,
	.signaled = radeon_fence_is_signaled,
	.wait = radeon_fence_default_wait,
	.release = NULL,
};