Training courses

Kernel and Embedded Linux

Bootlin training courses

Embedded Linux, kernel,
Yocto Project, Buildroot, real-time,
graphics, boot time, debugging...

Bootlin logo

Elixir Cross Referencer

  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
// SPDX-License-Identifier: GPL-2.0+
/* Copyright (C) 2014-2018 Broadcom */

#include <linux/device.h>
#include <linux/dma-mapping.h>
#include <linux/io.h>
#include <linux/module.h>
#include <linux/platform_device.h>
#include <linux/pm_runtime.h>
#include <linux/reset.h>
#include <linux/sched/signal.h>
#include <linux/uaccess.h>

#include <drm/drm_syncobj.h>
#include <uapi/drm/v3d_drm.h>

#include "v3d_drv.h"
#include "v3d_regs.h"
#include "v3d_trace.h"

static void
v3d_init_core(struct v3d_dev *v3d, int core)
{
	/* Set OVRTMUOUT, which means that the texture sampler uniform
	 * configuration's tmu output type field is used, instead of
	 * using the hardware default behavior based on the texture
	 * type.  If you want the default behavior, you can still put
	 * "2" in the indirect texture state's output_type field.
	 */
	if (v3d->ver < 40)
		V3D_CORE_WRITE(core, V3D_CTL_MISCCFG, V3D_MISCCFG_OVRTMUOUT);

	/* Whenever we flush the L2T cache, we always want to flush
	 * the whole thing.
	 */
	V3D_CORE_WRITE(core, V3D_CTL_L2TFLSTA, 0);
	V3D_CORE_WRITE(core, V3D_CTL_L2TFLEND, ~0);
}

/* Sets invariant state for the HW. */
static void
v3d_init_hw_state(struct v3d_dev *v3d)
{
	v3d_init_core(v3d, 0);
}

static void
v3d_idle_axi(struct v3d_dev *v3d, int core)
{
	V3D_CORE_WRITE(core, V3D_GMP_CFG, V3D_GMP_CFG_STOP_REQ);

	if (wait_for((V3D_CORE_READ(core, V3D_GMP_STATUS) &
		      (V3D_GMP_STATUS_RD_COUNT_MASK |
		       V3D_GMP_STATUS_WR_COUNT_MASK |
		       V3D_GMP_STATUS_CFG_BUSY)) == 0, 100)) {
		DRM_ERROR("Failed to wait for safe GMP shutdown\n");
	}
}

static void
v3d_idle_gca(struct v3d_dev *v3d)
{
	if (v3d->ver >= 41)
		return;

	V3D_GCA_WRITE(V3D_GCA_SAFE_SHUTDOWN, V3D_GCA_SAFE_SHUTDOWN_EN);

	if (wait_for((V3D_GCA_READ(V3D_GCA_SAFE_SHUTDOWN_ACK) &
		      V3D_GCA_SAFE_SHUTDOWN_ACK_ACKED) ==
		     V3D_GCA_SAFE_SHUTDOWN_ACK_ACKED, 100)) {
		DRM_ERROR("Failed to wait for safe GCA shutdown\n");
	}
}

static void
v3d_reset_by_bridge(struct v3d_dev *v3d)
{
	int version = V3D_BRIDGE_READ(V3D_TOP_GR_BRIDGE_REVISION);

	if (V3D_GET_FIELD(version, V3D_TOP_GR_BRIDGE_MAJOR) == 2) {
		V3D_BRIDGE_WRITE(V3D_TOP_GR_BRIDGE_SW_INIT_0,
				 V3D_TOP_GR_BRIDGE_SW_INIT_0_V3D_CLK_108_SW_INIT);
		V3D_BRIDGE_WRITE(V3D_TOP_GR_BRIDGE_SW_INIT_0, 0);

		/* GFXH-1383: The SW_INIT may cause a stray write to address 0
		 * of the unit, so reset it to its power-on value here.
		 */
		V3D_WRITE(V3D_HUB_AXICFG, V3D_HUB_AXICFG_MAX_LEN_MASK);
	} else {
		WARN_ON_ONCE(V3D_GET_FIELD(version,
					   V3D_TOP_GR_BRIDGE_MAJOR) != 7);
		V3D_BRIDGE_WRITE(V3D_TOP_GR_BRIDGE_SW_INIT_1,
				 V3D_TOP_GR_BRIDGE_SW_INIT_1_V3D_CLK_108_SW_INIT);
		V3D_BRIDGE_WRITE(V3D_TOP_GR_BRIDGE_SW_INIT_1, 0);
	}
}

static void
v3d_reset_v3d(struct v3d_dev *v3d)
{
	if (v3d->reset)
		reset_control_reset(v3d->reset);
	else
		v3d_reset_by_bridge(v3d);

	v3d_init_hw_state(v3d);
}

void
v3d_reset(struct v3d_dev *v3d)
{
	struct drm_device *dev = &v3d->drm;

	DRM_DEV_ERROR(dev->dev, "Resetting GPU for hang.\n");
	DRM_DEV_ERROR(dev->dev, "V3D_ERR_STAT: 0x%08x\n",
		      V3D_CORE_READ(0, V3D_ERR_STAT));
	trace_v3d_reset_begin(dev);

	/* XXX: only needed for safe powerdown, not reset. */
	if (false)
		v3d_idle_axi(v3d, 0);

	v3d_idle_gca(v3d);
	v3d_reset_v3d(v3d);

	v3d_mmu_set_page_table(v3d);
	v3d_irq_reset(v3d);

	trace_v3d_reset_end(dev);
}

static void
v3d_flush_l3(struct v3d_dev *v3d)
{
	if (v3d->ver < 41) {
		u32 gca_ctrl = V3D_GCA_READ(V3D_GCA_CACHE_CTRL);

		V3D_GCA_WRITE(V3D_GCA_CACHE_CTRL,
			      gca_ctrl | V3D_GCA_CACHE_CTRL_FLUSH);

		if (v3d->ver < 33) {
			V3D_GCA_WRITE(V3D_GCA_CACHE_CTRL,
				      gca_ctrl & ~V3D_GCA_CACHE_CTRL_FLUSH);
		}
	}
}

/* Invalidates the (read-only) L2C cache.  This was the L2 cache for
 * uniforms and instructions on V3D 3.2.
 */
static void
v3d_invalidate_l2c(struct v3d_dev *v3d, int core)
{
	if (v3d->ver > 32)
		return;

	V3D_CORE_WRITE(core, V3D_CTL_L2CACTL,
		       V3D_L2CACTL_L2CCLR |
		       V3D_L2CACTL_L2CENA);
}

/* Invalidates texture L2 cachelines */
static void
v3d_flush_l2t(struct v3d_dev *v3d, int core)
{
	/* While there is a busy bit (V3D_L2TCACTL_L2TFLS), we don't
	 * need to wait for completion before dispatching the job --
	 * L2T accesses will be stalled until the flush has completed.
	 * However, we do need to make sure we don't try to trigger a
	 * new flush while the L2_CLEAN queue is trying to
	 * synchronously clean after a job.
	 */
	mutex_lock(&v3d->cache_clean_lock);
	V3D_CORE_WRITE(core, V3D_CTL_L2TCACTL,
		       V3D_L2TCACTL_L2TFLS |
		       V3D_SET_FIELD(V3D_L2TCACTL_FLM_FLUSH, V3D_L2TCACTL_FLM));
	mutex_unlock(&v3d->cache_clean_lock);
}

/* Cleans texture L1 and L2 cachelines (writing back dirty data).
 *
 * For cleaning, which happens from the CACHE_CLEAN queue after CSD has
 * executed, we need to make sure that the clean is done before
 * signaling job completion.  So, we synchronously wait before
 * returning, and we make sure that L2 invalidates don't happen in the
 * meantime to confuse our are-we-done checks.
 */
void
v3d_clean_caches(struct v3d_dev *v3d)
{
	struct drm_device *dev = &v3d->drm;
	int core = 0;

	trace_v3d_cache_clean_begin(dev);

	V3D_CORE_WRITE(core, V3D_CTL_L2TCACTL, V3D_L2TCACTL_TMUWCF);
	if (wait_for(!(V3D_CORE_READ(core, V3D_CTL_L2TCACTL) &
		       V3D_L2TCACTL_L2TFLS), 100)) {
		DRM_ERROR("Timeout waiting for L1T write combiner flush\n");
	}

	mutex_lock(&v3d->cache_clean_lock);
	V3D_CORE_WRITE(core, V3D_CTL_L2TCACTL,
		       V3D_L2TCACTL_L2TFLS |
		       V3D_SET_FIELD(V3D_L2TCACTL_FLM_CLEAN, V3D_L2TCACTL_FLM));

	if (wait_for(!(V3D_CORE_READ(core, V3D_CTL_L2TCACTL) &
		       V3D_L2TCACTL_L2TFLS), 100)) {
		DRM_ERROR("Timeout waiting for L2T clean\n");
	}

	mutex_unlock(&v3d->cache_clean_lock);

	trace_v3d_cache_clean_end(dev);
}

/* Invalidates the slice caches.  These are read-only caches. */
static void
v3d_invalidate_slices(struct v3d_dev *v3d, int core)
{
	V3D_CORE_WRITE(core, V3D_CTL_SLCACTL,
		       V3D_SET_FIELD(0xf, V3D_SLCACTL_TVCCS) |
		       V3D_SET_FIELD(0xf, V3D_SLCACTL_TDCCS) |
		       V3D_SET_FIELD(0xf, V3D_SLCACTL_UCC) |
		       V3D_SET_FIELD(0xf, V3D_SLCACTL_ICC));
}

void
v3d_invalidate_caches(struct v3d_dev *v3d)
{
	/* Invalidate the caches from the outside in.  That way if
	 * another CL's concurrent use of nearby memory were to pull
	 * an invalidated cacheline back in, we wouldn't leave stale
	 * data in the inner cache.
	 */
	v3d_flush_l3(v3d);
	v3d_invalidate_l2c(v3d, 0);
	v3d_flush_l2t(v3d, 0);
	v3d_invalidate_slices(v3d, 0);
}

/* 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 v3d, so we don't attach dma-buf fences to them.
 */
static int
v3d_lock_bo_reservations(struct v3d_job *job,
			 struct ww_acquire_ctx *acquire_ctx)
{
	int i, ret;

	ret = drm_gem_lock_reservations(job->bo, job->bo_count, acquire_ctx);
	if (ret)
		return ret;

	for (i = 0; i < job->bo_count; i++) {
		ret = drm_gem_fence_array_add_implicit(&job->deps,
						       job->bo[i], true);
		if (ret) {
			drm_gem_unlock_reservations(job->bo, job->bo_count,
						    acquire_ctx);
			return ret;
		}
	}

	return 0;
}

/**
 * v3d_lookup_bos() - Sets up job->bo[] with the GEM objects
 * referenced by the job.
 * @dev: DRM device
 * @file_priv: DRM file for this fd
 * @job: 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 v3d_exec_cleanup() time.
 */
static int
v3d_lookup_bos(struct drm_device *dev,
	       struct drm_file *file_priv,
	       struct v3d_job *job,
	       u64 bo_handles,
	       u32 bo_count)
{
	u32 *handles;
	int ret = 0;
	int i;

	job->bo_count = bo_count;

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

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

	handles = kvmalloc_array(job->bo_count, sizeof(u32), GFP_KERNEL);
	if (!handles) {
		ret = -ENOMEM;
		DRM_DEBUG("Failed to allocate incoming GEM handles\n");
		goto fail;
	}

	if (copy_from_user(handles,
			   (void __user *)(uintptr_t)bo_handles,
			   job->bo_count * sizeof(u32))) {
		ret = -EFAULT;
		DRM_DEBUG("Failed to copy in GEM handles\n");
		goto fail;
	}

	spin_lock(&file_priv->table_lock);
	for (i = 0; i < job->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 = -ENOENT;
			spin_unlock(&file_priv->table_lock);
			goto fail;
		}
		drm_gem_object_get(bo);
		job->bo[i] = bo;
	}
	spin_unlock(&file_priv->table_lock);

fail:
	kvfree(handles);
	return ret;
}

static void
v3d_job_free(struct kref *ref)
{
	struct v3d_job *job = container_of(ref, struct v3d_job, refcount);
	unsigned long index;
	struct dma_fence *fence;
	int i;

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

	xa_for_each(&job->deps, index, fence) {
		dma_fence_put(fence);
	}
	xa_destroy(&job->deps);

	dma_fence_put(job->irq_fence);
	dma_fence_put(job->done_fence);

	pm_runtime_mark_last_busy(job->v3d->dev);
	pm_runtime_put_autosuspend(job->v3d->dev);

	kfree(job);
}

static void
v3d_render_job_free(struct kref *ref)
{
	struct v3d_render_job *job = container_of(ref, struct v3d_render_job,
						  base.refcount);
	struct v3d_bo *bo, *save;

	list_for_each_entry_safe(bo, save, &job->unref_list, unref_head) {
		drm_gem_object_put_unlocked(&bo->base.base);
	}

	v3d_job_free(ref);
}

void v3d_job_put(struct v3d_job *job)
{
	kref_put(&job->refcount, job->free);
}

int
v3d_wait_bo_ioctl(struct drm_device *dev, void *data,
		  struct drm_file *file_priv)
{
	int ret;
	struct drm_v3d_wait_bo *args = data;
	ktime_t start = ktime_get();
	u64 delta_ns;
	unsigned long timeout_jiffies =
		nsecs_to_jiffies_timeout(args->timeout_ns);

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

	ret = drm_gem_dma_resv_wait(file_priv, args->handle,
					      true, timeout_jiffies);

	/* Decrement the user's timeout, in case we got interrupted
	 * such that the ioctl will be restarted.
	 */
	delta_ns = ktime_to_ns(ktime_sub(ktime_get(), start));
	if (delta_ns < args->timeout_ns)
		args->timeout_ns -= delta_ns;
	else
		args->timeout_ns = 0;

	/* Asked to wait beyond the jiffie/scheduler precision? */
	if (ret == -ETIME && args->timeout_ns)
		ret = -EAGAIN;

	return ret;
}

static int
v3d_job_init(struct v3d_dev *v3d, struct drm_file *file_priv,
	     struct v3d_job *job, void (*free)(struct kref *ref),
	     u32 in_sync)
{
	struct dma_fence *in_fence = NULL;
	int ret;

	job->v3d = v3d;
	job->free = free;

	ret = pm_runtime_get_sync(v3d->dev);
	if (ret < 0)
		return ret;

	xa_init_flags(&job->deps, XA_FLAGS_ALLOC);

	ret = drm_syncobj_find_fence(file_priv, in_sync, 0, 0, &in_fence);
	if (ret == -EINVAL)
		goto fail;

	ret = drm_gem_fence_array_add(&job->deps, in_fence);
	if (ret)
		goto fail;

	kref_init(&job->refcount);

	return 0;
fail:
	xa_destroy(&job->deps);
	pm_runtime_put_autosuspend(v3d->dev);
	return ret;
}

static int
v3d_push_job(struct v3d_file_priv *v3d_priv,
	     struct v3d_job *job, enum v3d_queue queue)
{
	int ret;

	ret = drm_sched_job_init(&job->base, &v3d_priv->sched_entity[queue],
				 v3d_priv);
	if (ret)
		return ret;

	job->done_fence = dma_fence_get(&job->base.s_fence->finished);

	/* put by scheduler job completion */
	kref_get(&job->refcount);

	drm_sched_entity_push_job(&job->base, &v3d_priv->sched_entity[queue]);

	return 0;
}

static void
v3d_attach_fences_and_unlock_reservation(struct drm_file *file_priv,
					 struct v3d_job *job,
					 struct ww_acquire_ctx *acquire_ctx,
					 u32 out_sync,
					 struct dma_fence *done_fence)
{
	struct drm_syncobj *sync_out;
	int i;

	for (i = 0; i < job->bo_count; i++) {
		/* XXX: Use shared fences for read-only objects. */
		dma_resv_add_excl_fence(job->bo[i]->resv,
						  job->done_fence);
	}

	drm_gem_unlock_reservations(job->bo, job->bo_count, acquire_ctx);

	/* Update the return sync object for the job */
	sync_out = drm_syncobj_find(file_priv, out_sync);
	if (sync_out) {
		drm_syncobj_replace_fence(sync_out, done_fence);
		drm_syncobj_put(sync_out);
	}
}

/**
 * v3d_submit_cl_ioctl() - Submits a job (frame) to the V3D.
 * @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
v3d_submit_cl_ioctl(struct drm_device *dev, void *data,
		    struct drm_file *file_priv)
{
	struct v3d_dev *v3d = to_v3d_dev(dev);
	struct v3d_file_priv *v3d_priv = file_priv->driver_priv;
	struct drm_v3d_submit_cl *args = data;
	struct v3d_bin_job *bin = NULL;
	struct v3d_render_job *render;
	struct ww_acquire_ctx acquire_ctx;
	int ret = 0;

	trace_v3d_submit_cl_ioctl(&v3d->drm, args->rcl_start, args->rcl_end);

	if (args->pad != 0) {
		DRM_INFO("pad must be zero: %d\n", args->pad);
		return -EINVAL;
	}

	render = kcalloc(1, sizeof(*render), GFP_KERNEL);
	if (!render)
		return -ENOMEM;

	render->start = args->rcl_start;
	render->end = args->rcl_end;
	INIT_LIST_HEAD(&render->unref_list);

	ret = v3d_job_init(v3d, file_priv, &render->base,
			   v3d_render_job_free, args->in_sync_rcl);
	if (ret) {
		kfree(render);
		return ret;
	}

	if (args->bcl_start != args->bcl_end) {
		bin = kcalloc(1, sizeof(*bin), GFP_KERNEL);
		if (!bin) {
			v3d_job_put(&render->base);
			return -ENOMEM;
		}

		ret = v3d_job_init(v3d, file_priv, &bin->base,
				   v3d_job_free, args->in_sync_bcl);
		if (ret) {
			v3d_job_put(&render->base);
			kfree(bin);
			return ret;
		}

		bin->start = args->bcl_start;
		bin->end = args->bcl_end;
		bin->qma = args->qma;
		bin->qms = args->qms;
		bin->qts = args->qts;
		bin->render = render;
	}

	ret = v3d_lookup_bos(dev, file_priv, &render->base,
			     args->bo_handles, args->bo_handle_count);
	if (ret)
		goto fail;

	ret = v3d_lock_bo_reservations(&render->base, &acquire_ctx);
	if (ret)
		goto fail;

	mutex_lock(&v3d->sched_lock);
	if (bin) {
		ret = v3d_push_job(v3d_priv, &bin->base, V3D_BIN);
		if (ret)
			goto fail_unreserve;

		ret = drm_gem_fence_array_add(&render->base.deps,
					      dma_fence_get(bin->base.done_fence));
		if (ret)
			goto fail_unreserve;
	}

	ret = v3d_push_job(v3d_priv, &render->base, V3D_RENDER);
	if (ret)
		goto fail_unreserve;
	mutex_unlock(&v3d->sched_lock);

	v3d_attach_fences_and_unlock_reservation(file_priv,
						 &render->base,
						 &acquire_ctx,
						 args->out_sync,
						 render->base.done_fence);

	if (bin)
		v3d_job_put(&bin->base);
	v3d_job_put(&render->base);

	return 0;

fail_unreserve:
	mutex_unlock(&v3d->sched_lock);
	drm_gem_unlock_reservations(render->base.bo,
				    render->base.bo_count, &acquire_ctx);
fail:
	if (bin)
		v3d_job_put(&bin->base);
	v3d_job_put(&render->base);

	return ret;
}

/**
 * v3d_submit_tfu_ioctl() - Submits a TFU (texture formatting) job to the V3D.
 * @dev: DRM device
 * @data: ioctl argument
 * @file_priv: DRM file for this fd
 *
 * Userspace provides the register setup for the TFU, which we don't
 * need to validate since the TFU is behind the MMU.
 */
int
v3d_submit_tfu_ioctl(struct drm_device *dev, void *data,
		     struct drm_file *file_priv)
{
	struct v3d_dev *v3d = to_v3d_dev(dev);
	struct v3d_file_priv *v3d_priv = file_priv->driver_priv;
	struct drm_v3d_submit_tfu *args = data;
	struct v3d_tfu_job *job;
	struct ww_acquire_ctx acquire_ctx;
	int ret = 0;

	trace_v3d_submit_tfu_ioctl(&v3d->drm, args->iia);

	job = kcalloc(1, sizeof(*job), GFP_KERNEL);
	if (!job)
		return -ENOMEM;

	ret = v3d_job_init(v3d, file_priv, &job->base,
			   v3d_job_free, args->in_sync);
	if (ret) {
		kfree(job);
		return ret;
	}

	job->base.bo = kcalloc(ARRAY_SIZE(args->bo_handles),
			       sizeof(*job->base.bo), GFP_KERNEL);
	if (!job->base.bo) {
		v3d_job_put(&job->base);
		return -ENOMEM;
	}

	job->args = *args;

	spin_lock(&file_priv->table_lock);
	for (job->base.bo_count = 0;
	     job->base.bo_count < ARRAY_SIZE(args->bo_handles);
	     job->base.bo_count++) {
		struct drm_gem_object *bo;

		if (!args->bo_handles[job->base.bo_count])
			break;

		bo = idr_find(&file_priv->object_idr,
			      args->bo_handles[job->base.bo_count]);
		if (!bo) {
			DRM_DEBUG("Failed to look up GEM BO %d: %d\n",
				  job->base.bo_count,
				  args->bo_handles[job->base.bo_count]);
			ret = -ENOENT;
			spin_unlock(&file_priv->table_lock);
			goto fail;
		}
		drm_gem_object_get(bo);
		job->base.bo[job->base.bo_count] = bo;
	}
	spin_unlock(&file_priv->table_lock);

	ret = v3d_lock_bo_reservations(&job->base, &acquire_ctx);
	if (ret)
		goto fail;

	mutex_lock(&v3d->sched_lock);
	ret = v3d_push_job(v3d_priv, &job->base, V3D_TFU);
	if (ret)
		goto fail_unreserve;
	mutex_unlock(&v3d->sched_lock);

	v3d_attach_fences_and_unlock_reservation(file_priv,
						 &job->base, &acquire_ctx,
						 args->out_sync,
						 job->base.done_fence);

	v3d_job_put(&job->base);

	return 0;

fail_unreserve:
	mutex_unlock(&v3d->sched_lock);
	drm_gem_unlock_reservations(job->base.bo, job->base.bo_count,
				    &acquire_ctx);
fail:
	v3d_job_put(&job->base);

	return ret;
}

/**
 * v3d_submit_csd_ioctl() - Submits a CSD (texture formatting) job to the V3D.
 * @dev: DRM device
 * @data: ioctl argument
 * @file_priv: DRM file for this fd
 *
 * Userspace provides the register setup for the CSD, which we don't
 * need to validate since the CSD is behind the MMU.
 */
int
v3d_submit_csd_ioctl(struct drm_device *dev, void *data,
		     struct drm_file *file_priv)
{
	struct v3d_dev *v3d = to_v3d_dev(dev);
	struct v3d_file_priv *v3d_priv = file_priv->driver_priv;
	struct drm_v3d_submit_csd *args = data;
	struct v3d_csd_job *job;
	struct v3d_job *clean_job;
	struct ww_acquire_ctx acquire_ctx;
	int ret;

	trace_v3d_submit_csd_ioctl(&v3d->drm, args->cfg[5], args->cfg[6]);

	if (!v3d_has_csd(v3d)) {
		DRM_DEBUG("Attempting CSD submit on non-CSD hardware\n");
		return -EINVAL;
	}

	job = kcalloc(1, sizeof(*job), GFP_KERNEL);
	if (!job)
		return -ENOMEM;

	ret = v3d_job_init(v3d, file_priv, &job->base,
			   v3d_job_free, args->in_sync);
	if (ret) {
		kfree(job);
		return ret;
	}

	clean_job = kcalloc(1, sizeof(*clean_job), GFP_KERNEL);
	if (!clean_job) {
		v3d_job_put(&job->base);
		kfree(job);
		return -ENOMEM;
	}

	ret = v3d_job_init(v3d, file_priv, clean_job, v3d_job_free, 0);
	if (ret) {
		v3d_job_put(&job->base);
		kfree(clean_job);
		return ret;
	}

	job->args = *args;

	ret = v3d_lookup_bos(dev, file_priv, clean_job,
			     args->bo_handles, args->bo_handle_count);
	if (ret)
		goto fail;

	ret = v3d_lock_bo_reservations(clean_job, &acquire_ctx);
	if (ret)
		goto fail;

	mutex_lock(&v3d->sched_lock);
	ret = v3d_push_job(v3d_priv, &job->base, V3D_CSD);
	if (ret)
		goto fail_unreserve;

	ret = drm_gem_fence_array_add(&clean_job->deps,
				      dma_fence_get(job->base.done_fence));
	if (ret)
		goto fail_unreserve;

	ret = v3d_push_job(v3d_priv, clean_job, V3D_CACHE_CLEAN);
	if (ret)
		goto fail_unreserve;
	mutex_unlock(&v3d->sched_lock);

	v3d_attach_fences_and_unlock_reservation(file_priv,
						 clean_job,
						 &acquire_ctx,
						 args->out_sync,
						 clean_job->done_fence);

	v3d_job_put(&job->base);
	v3d_job_put(clean_job);

	return 0;

fail_unreserve:
	mutex_unlock(&v3d->sched_lock);
	drm_gem_unlock_reservations(clean_job->bo, clean_job->bo_count,
				    &acquire_ctx);
fail:
	v3d_job_put(&job->base);
	v3d_job_put(clean_job);

	return ret;
}

int
v3d_gem_init(struct drm_device *dev)
{
	struct v3d_dev *v3d = to_v3d_dev(dev);
	u32 pt_size = 4096 * 1024;
	int ret, i;

	for (i = 0; i < V3D_MAX_QUEUES; i++)
		v3d->queue[i].fence_context = dma_fence_context_alloc(1);

	spin_lock_init(&v3d->mm_lock);
	spin_lock_init(&v3d->job_lock);
	mutex_init(&v3d->bo_lock);
	mutex_init(&v3d->reset_lock);
	mutex_init(&v3d->sched_lock);
	mutex_init(&v3d->cache_clean_lock);

	/* Note: We don't allocate address 0.  Various bits of HW
	 * treat 0 as special, such as the occlusion query counters
	 * where 0 means "disabled".
	 */
	drm_mm_init(&v3d->mm, 1, pt_size / sizeof(u32) - 1);

	v3d->pt = dma_alloc_wc(v3d->dev, pt_size,
			       &v3d->pt_paddr,
			       GFP_KERNEL | __GFP_NOWARN | __GFP_ZERO);
	if (!v3d->pt) {
		drm_mm_takedown(&v3d->mm);
		dev_err(v3d->dev,
			"Failed to allocate page tables. "
			"Please ensure you have CMA enabled.\n");
		return -ENOMEM;
	}

	v3d_init_hw_state(v3d);
	v3d_mmu_set_page_table(v3d);

	ret = v3d_sched_init(v3d);
	if (ret) {
		drm_mm_takedown(&v3d->mm);
		dma_free_coherent(v3d->dev, 4096 * 1024, (void *)v3d->pt,
				  v3d->pt_paddr);
	}

	return 0;
}

void
v3d_gem_destroy(struct drm_device *dev)
{
	struct v3d_dev *v3d = to_v3d_dev(dev);

	v3d_sched_fini(v3d);

	/* Waiting for jobs to finish would need to be done before
	 * unregistering V3D.
	 */
	WARN_ON(v3d->bin_job);
	WARN_ON(v3d->render_job);

	drm_mm_takedown(&v3d->mm);

	dma_free_coherent(v3d->dev, 4096 * 1024, (void *)v3d->pt, v3d->pt_paddr);
}