aboutsummaryrefslogblamecommitdiffstats
path: root/include/os/linux/module.c
blob: 807df2cadfbc6d1d76008021e5dfabdea232c72b (plain) (tree)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
























































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































                                                                                
/*
 * GK20A Graphics
 *
 * Copyright (c) 2011-2020, NVIDIA CORPORATION.  All rights reserved.
 *
 * This program is free software; you can redistribute it and/or modify it
 * under the terms and conditions of the GNU General Public License,
 * version 2, as published by the Free Software Foundation.
 *
 * This program is distributed in the hope it will be useful, but WITHOUT
 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
 * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
 * more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program.  If not, see <http://www.gnu.org/licenses/>.
 */

#include <linux/module.h>
#include <linux/of.h>
#include <linux/of_device.h>
#include <linux/of_platform.h>
#include <linux/of_address.h>
#include <linux/interrupt.h>
#include <linux/pm_runtime.h>
#include <linux/reset.h>
#include <linux/reboot.h>
#include <linux/notifier.h>
#include <linux/platform/tegra/common.h>
#include <linux/pci.h>

#include <uapi/linux/nvgpu.h>
#include <dt-bindings/soc/gm20b-fuse.h>
#include <dt-bindings/soc/gp10b-fuse.h>
#include <dt-bindings/soc/gv11b-fuse.h>

#include <soc/tegra/fuse.h>

#include <nvgpu/hal_init.h>
#include <nvgpu/dma.h>
#include <nvgpu/kmem.h>
#include <nvgpu/nvgpu_common.h>
#include <nvgpu/soc.h>
#include <nvgpu/enabled.h>
#include <nvgpu/debug.h>
#include <nvgpu/ctxsw_trace.h>
#include <nvgpu/vidmem.h>
#include <nvgpu/sim.h>
#include <nvgpu/clk_arb.h>
#include <nvgpu/timers.h>
#include <nvgpu/channel.h>

#include "platform_gk20a.h"
#include "sysfs.h"
#include "vgpu/vgpu_linux.h"
#include "scale.h"
#include "pci.h"
#include "module.h"
#include "module_usermode.h"
#include "intr.h"
#include "ioctl.h"
#include "ioctl_ctrl.h"

#include "os_linux.h"
#include "os_ops.h"
#include "ctxsw_trace.h"
#include "driver_common.h"
#include "channel.h"
#include "debug_pmgr.h"

#ifdef CONFIG_NVGPU_SUPPORT_CDE
#include "cde.h"
#endif

#define CLASS_NAME "nvidia-gpu"
/* TODO: Change to e.g. "nvidia-gpu%s" once we have symlinks in place. */

#define GK20A_WAIT_FOR_IDLE_MS	2000

#define CREATE_TRACE_POINTS
#include <trace/events/gk20a.h>

static int nvgpu_kernel_shutdown_notification(struct notifier_block *nb,
					unsigned long event, void *unused)
{
	struct gk20a *g = container_of(nb, struct gk20a, nvgpu_reboot_nb);

	__nvgpu_set_enabled(g, NVGPU_KERNEL_IS_DYING, true);
	return NOTIFY_DONE;
}

struct device_node *nvgpu_get_node(struct gk20a *g)
{
	struct device *dev = dev_from_gk20a(g);

	if (dev_is_pci(dev)) {
		struct pci_bus *bus = to_pci_dev(dev)->bus;

		while (!pci_is_root_bus(bus))
			bus = bus->parent;

		return bus->bridge->parent->of_node;
	}

	return dev->of_node;
}

void gk20a_busy_noresume(struct gk20a *g)
{
	pm_runtime_get_noresume(dev_from_gk20a(g));
}

/*
 * Check if the device can go busy.
 */
static int nvgpu_can_busy(struct gk20a *g)
{
	/* Can't do anything if the system is rebooting/shutting down. */
	if (nvgpu_is_enabled(g, NVGPU_KERNEL_IS_DYING))
		return 0;

	/* Can't do anything if the driver is restarting. */
	if (nvgpu_is_enabled(g, NVGPU_DRIVER_IS_DYING))
		return 0;

	return 1;
}

int gk20a_busy(struct gk20a *g)
{
	struct nvgpu_os_linux *l = nvgpu_os_linux_from_gk20a(g);
	int ret = 0;
	struct device *dev;

	if (!g)
		return -ENODEV;

	atomic_inc(&g->usage_count.atomic_var);

	down_read(&l->busy_lock);

	if (!nvgpu_can_busy(g)) {
		ret = -ENODEV;
		atomic_dec(&g->usage_count.atomic_var);
		goto fail;
	}

	dev = dev_from_gk20a(g);

	if (pm_runtime_enabled(dev)) {
		/* Increment usage count and attempt to resume device */
		ret = pm_runtime_get_sync(dev);
		if (ret < 0) {
			/* Mark suspended so runtime pm will retry later */
			pm_runtime_set_suspended(dev);
			pm_runtime_put_noidle(dev);
			atomic_dec(&g->usage_count.atomic_var);
			goto fail;
		}
	} else {
		ret = gk20a_gpu_is_virtual(dev) ?
			vgpu_pm_finalize_poweron(dev) :
			gk20a_pm_finalize_poweron(dev);
		if (ret) {
			atomic_dec(&g->usage_count.atomic_var);
			goto fail;
		}
	}

fail:
	up_read(&l->busy_lock);

	return ret < 0 ? ret : 0;
}

void gk20a_idle_nosuspend(struct gk20a *g)
{
	pm_runtime_put_noidle(dev_from_gk20a(g));
}

void gk20a_idle(struct gk20a *g)
{
	struct device *dev;

	atomic_dec(&g->usage_count.atomic_var);

	dev = dev_from_gk20a(g);

	if (!(dev && nvgpu_can_busy(g)))
		return;

	if (pm_runtime_enabled(dev)) {
		pm_runtime_mark_last_busy(dev);
		pm_runtime_put_sync_autosuspend(dev);
	}
}

/*
 * Undoes gk20a_lockout_registers().
 */
static int gk20a_restore_registers(struct gk20a *g)
{
	struct nvgpu_os_linux *l = nvgpu_os_linux_from_gk20a(g);

	l->regs = l->regs_saved;
	l->bar1 = l->bar1_saved;

	nvgpu_restore_usermode_registers(g);

	return 0;
}

int nvgpu_finalize_poweron_linux(struct nvgpu_os_linux *l)
{
	struct gk20a *g = &l->g;
	int err;

	if (l->init_done)
		return 0;

	err = nvgpu_init_channel_support_linux(l);
	if (err) {
		nvgpu_err(g, "failed to init linux channel support");
		return err;
	}

	if (l->ops.clk.init_debugfs) {
		err = l->ops.clk.init_debugfs(g);
		if (err) {
			nvgpu_err(g, "failed to init linux clk debugfs");
			return err;
		}
	}

	if (l->ops.therm.init_debugfs) {
		err = l->ops.therm.init_debugfs(g);
		if (err) {
			nvgpu_err(g, "failed to init linux therm debugfs");
			return err;
		}
	}

	if (l->ops.fecs_trace.init_debugfs) {
		err = l->ops.fecs_trace.init_debugfs(g);
		if (err) {
			nvgpu_err(g, "failed to init linux fecs trace debugfs");
			return err;
		}
	}

	err = nvgpu_pmgr_init_debugfs_linux(l);
	if (err) {
		nvgpu_err(g, "failed to init linux pmgr debugfs");
		return err;
	}

	l->init_done = true;

	return 0;
}

bool gk20a_check_poweron(struct gk20a *g)
{
	bool ret;

	nvgpu_mutex_acquire(&g->power_lock);
	ret = g->power_on;
	nvgpu_mutex_release(&g->power_lock);

	return ret;
}

int gk20a_pm_finalize_poweron(struct device *dev)
{
	struct gk20a *g = get_gk20a(dev);
	struct nvgpu_os_linux *l = nvgpu_os_linux_from_gk20a(g);
	struct gk20a_platform *platform = gk20a_get_platform(dev);
	int err = 0;

	nvgpu_log_fn(g, " ");

	nvgpu_mutex_acquire(&g->power_lock);

	if (g->power_on)
		goto done;

	trace_gk20a_finalize_poweron(dev_name(dev));

	/* Increment platform power refcount */
	if (platform->busy) {
		err = platform->busy(dev);
		if (err < 0) {
			nvgpu_err(g, "failed to poweron platform dependency");
			goto done;
		}
	}

	err = gk20a_restore_registers(g);
	if (err)
		goto done;

	nvgpu_restore_usermode_for_poweron(g);

	/* Enable interrupt workqueue */
	if (!l->nonstall_work_queue) {
		l->nonstall_work_queue = alloc_workqueue("%s",
						WQ_HIGHPRI, 1, "mc_nonstall");
		INIT_WORK(&l->nonstall_fn_work, nvgpu_intr_nonstall_cb);
	}

	err = nvgpu_detect_chip(g);
	if (err)
		goto done;

	if (g->sim) {
		if (g->sim->sim_init_late)
			g->sim->sim_init_late(g);
	}

	err = gk20a_finalize_poweron(g);
	if (err)
		goto done;

	err = nvgpu_init_os_linux_ops(l);
	if (err)
		goto done;

	err = nvgpu_finalize_poweron_linux(l);
	if (err)
		goto done;

	nvgpu_init_mm_ce_context(g);

	nvgpu_vidmem_thread_unpause(&g->mm);

	/* Initialise scaling: it will initialize scaling drive only once */
	if (IS_ENABLED(CONFIG_GK20A_DEVFREQ) &&
			nvgpu_platform_is_silicon(g)) {
		gk20a_scale_init(dev);
		if (platform->initscale)
			platform->initscale(dev);
	}

	trace_gk20a_finalize_poweron_done(dev_name(dev));

	enable_irq(g->irq_stall);
	if (g->irq_stall != g->irq_nonstall)
		enable_irq(g->irq_nonstall);
	g->irqs_enabled = 1;

	gk20a_scale_resume(dev_from_gk20a(g));

#ifdef CONFIG_NVGPU_SUPPORT_CDE
	if (platform->has_cde)
		gk20a_init_cde_support(l);
#endif

	err = gk20a_sched_ctrl_init(g);
	if (err) {
		nvgpu_err(g, "failed to init sched control");
		goto done;
	}

	g->sw_ready = true;

done:
	if (err)
		g->power_on = false;

	nvgpu_mutex_release(&g->power_lock);
	return err;
}

/*
 * Locks out the driver from accessing GPU registers. This prevents access to
 * thse registers after the GPU has been clock or power gated. This should help
 * find annoying bugs where register reads and writes are silently dropped
 * after the GPU has been turned off. On older chips these reads and writes can
 * also lock the entire CPU up.
 */
static int gk20a_lockout_registers(struct gk20a *g)
{
	struct nvgpu_os_linux *l = nvgpu_os_linux_from_gk20a(g);

	l->regs = NULL;
	l->bar1 = NULL;

	nvgpu_lockout_usermode_registers(g);

	return 0;
}

static int gk20a_pm_prepare_poweroff(struct device *dev)
{
	struct gk20a *g = get_gk20a(dev);
#ifdef CONFIG_NVGPU_SUPPORT_CDE
	struct nvgpu_os_linux *l = nvgpu_os_linux_from_gk20a(g);
#endif
	struct gk20a_platform *platform = gk20a_get_platform(dev);
	bool irqs_enabled;
	int ret = 0;

	nvgpu_log_fn(g, " ");

	nvgpu_mutex_acquire(&g->power_lock);

	if (!g->power_on)
		goto done;

	/* disable IRQs and wait for completion */
	irqs_enabled = g->irqs_enabled;
	if (irqs_enabled) {
		disable_irq(g->irq_stall);
		if (g->irq_stall != g->irq_nonstall)
			disable_irq(g->irq_nonstall);
		g->irqs_enabled = 0;
	}

	gk20a_scale_suspend(dev);

#ifdef CONFIG_NVGPU_SUPPORT_CDE
	gk20a_cde_suspend(l);
#endif

	ret = gk20a_prepare_poweroff(g);
	if (ret)
		goto error;

	/* Decrement platform power refcount */
	if (platform->idle)
		platform->idle(dev);

	/* Stop CPU from accessing the GPU registers. */
	gk20a_lockout_registers(g);

	nvgpu_hide_usermode_for_poweroff(g);
	nvgpu_mutex_release(&g->power_lock);
	return 0;

error:
	/* re-enabled IRQs if previously enabled */
	if (irqs_enabled) {
		enable_irq(g->irq_stall);
		if (g->irq_stall != g->irq_nonstall)
			enable_irq(g->irq_nonstall);
		g->irqs_enabled = 1;
	}

	gk20a_scale_resume(dev);
done:
	nvgpu_mutex_release(&g->power_lock);

	return ret;
}

static struct of_device_id tegra_gk20a_of_match[] = {
#ifdef CONFIG_TEGRA_GK20A
	{ .compatible = "nvidia,tegra210-gm20b",
		.data = &gm20b_tegra_platform },
	{ .compatible = "nvidia,tegra186-gp10b",
		.data = &gp10b_tegra_platform },
	{ .compatible = "nvidia,gv11b",
		.data = &gv11b_tegra_platform },
#ifdef CONFIG_TEGRA_GR_VIRTUALIZATION
	{ .compatible = "nvidia,gv11b-vgpu",
		.data = &gv11b_vgpu_tegra_platform},
#endif
#ifdef CONFIG_TEGRA_GR_VIRTUALIZATION
	{ .compatible = "nvidia,tegra124-gk20a-vgpu",
		.data = &vgpu_tegra_platform },
#endif
#endif

	{ },
};
MODULE_DEVICE_TABLE(of, tegra_gk20a_of_match);

#ifdef CONFIG_PM
/**
 * __gk20a_do_idle() - force the GPU to idle and railgate
 *
 * In success, this call MUST be balanced by caller with __gk20a_do_unidle()
 *
 * Acquires two locks : &l->busy_lock and &platform->railgate_lock
 * In success, we hold these locks and return
 * In failure, we release these locks and return
 */
int __gk20a_do_idle(struct gk20a *g, bool force_reset)
{
	struct nvgpu_os_linux *l = nvgpu_os_linux_from_gk20a(g);
	struct device *dev = dev_from_gk20a(g);
	struct gk20a_platform *platform = dev_get_drvdata(dev);
	struct nvgpu_timeout timeout;
	int ref_cnt;
	int target_ref_cnt = 0;
	bool is_railgated;
	int err = 0;

	/*
	 * Hold back deterministic submits and changes to deterministic
	 * channels - this must be outside the power busy locks.
	 */
	gk20a_channel_deterministic_idle(g);

	/* acquire busy lock to block other busy() calls */
	down_write(&l->busy_lock);

	/* acquire railgate lock to prevent unrailgate in midst of do_idle() */
	nvgpu_mutex_acquire(&platform->railgate_lock);

	/* check if it is already railgated ? */
	if (platform->is_railgated(dev))
		return 0;

	/*
	 * release railgate_lock, prevent suspend by incrementing usage counter,
	 * re-acquire railgate_lock
	 */
	nvgpu_mutex_release(&platform->railgate_lock);
	pm_runtime_get_sync(dev);

	/*
	 * One refcount taken in this API
	 * If User disables rail gating, we take one more
	 * extra refcount
	 */
	if (nvgpu_is_enabled(g, NVGPU_CAN_RAILGATE))
		target_ref_cnt = 1;
	else
		target_ref_cnt = 2;
	nvgpu_mutex_acquire(&platform->railgate_lock);

	nvgpu_timeout_init(g, &timeout, GK20A_WAIT_FOR_IDLE_MS,
			   NVGPU_TIMER_CPU_TIMER);

	/* check and wait until GPU is idle (with a timeout) */
	do {
		nvgpu_usleep_range(1000, 1100);
		ref_cnt = atomic_read(&dev->power.usage_count);
	} while (ref_cnt != target_ref_cnt && !nvgpu_timeout_expired(&timeout));

	if (ref_cnt != target_ref_cnt) {
		nvgpu_err(g, "failed to idle - refcount %d != target_ref_cnt",
			ref_cnt);
		goto fail_drop_usage_count;
	}

	/* check if global force_reset flag is set */
	force_reset |= platform->force_reset_in_do_idle;

	nvgpu_timeout_init(g, &timeout, GK20A_WAIT_FOR_IDLE_MS,
			   NVGPU_TIMER_CPU_TIMER);

	if (nvgpu_is_enabled(g, NVGPU_CAN_RAILGATE) && !force_reset) {
		/*
		 * Case 1 : GPU railgate is supported
		 *
		 * if GPU is now idle, we will have only one ref count,
		 * drop this ref which will rail gate the GPU
		 */
		pm_runtime_put_sync(dev);

		/* add sufficient delay to allow GPU to rail gate */
		nvgpu_msleep(g->railgate_delay);

		/* check in loop if GPU is railgated or not */
		do {
			nvgpu_usleep_range(1000, 1100);
			is_railgated = platform->is_railgated(dev);
		} while (!is_railgated && !nvgpu_timeout_expired(&timeout));

		if (is_railgated) {
			return 0;
		} else {
			nvgpu_err(g, "failed to idle in timeout");
			goto fail_timeout;
		}
	} else {
		/*
		 * Case 2 : GPU railgate is not supported or we explicitly
		 * do not want to depend on runtime PM
		 *
		 * if GPU is now idle, call prepare_poweroff() to save the
		 * state and then do explicit railgate
		 *
		 * __gk20a_do_unidle() needs to unrailgate, call
		 * finalize_poweron(), and then call pm_runtime_put_sync()
		 * to balance the GPU usage counter
		 */

		/* Save the GPU state */
		err = gk20a_pm_prepare_poweroff(dev);
		if (err)
			goto fail_drop_usage_count;

		/* railgate GPU */
		platform->railgate(dev);

		nvgpu_udelay(10);

		g->forced_reset = true;
		return 0;
	}

fail_drop_usage_count:
	pm_runtime_put_noidle(dev);
fail_timeout:
	nvgpu_mutex_release(&platform->railgate_lock);
	up_write(&l->busy_lock);
	gk20a_channel_deterministic_unidle(g);
	return -EBUSY;
}

/**
 * gk20a_do_idle() - wrap up for __gk20a_do_idle() to be called
 * from outside of GPU driver
 *
 * In success, this call MUST be balanced by caller with gk20a_do_unidle()
 */
static int gk20a_do_idle(void *_g)
{
	struct gk20a *g = (struct gk20a *)_g;

	return __gk20a_do_idle(g, true);
}

/**
 * __gk20a_do_unidle() - unblock all the tasks blocked by __gk20a_do_idle()
 */
int __gk20a_do_unidle(struct gk20a *g)
{
	struct nvgpu_os_linux *l = nvgpu_os_linux_from_gk20a(g);
	struct device *dev = dev_from_gk20a(g);
	struct gk20a_platform *platform = dev_get_drvdata(dev);
	int err;

	if (g->forced_reset) {
		/*
		 * If we did a forced-reset/railgate
		 * then unrailgate the GPU here first
		 */
		platform->unrailgate(dev);

		/* restore the GPU state */
		err = gk20a_pm_finalize_poweron(dev);
		if (err)
			return err;

		/* balance GPU usage counter */
		pm_runtime_put_sync(dev);

		g->forced_reset = false;
	}

	/* release the lock and open up all other busy() calls */
	nvgpu_mutex_release(&platform->railgate_lock);
	up_write(&l->busy_lock);

	gk20a_channel_deterministic_unidle(g);

	return 0;
}

/**
 * gk20a_do_unidle() - wrap up for __gk20a_do_unidle()
 */
static int gk20a_do_unidle(void *_g)
{
	struct gk20a *g = (struct gk20a *)_g;

	return __gk20a_do_unidle(g);
}
#endif

void __iomem *nvgpu_devm_ioremap_resource(struct platform_device *dev, int i,
					  struct resource **out)
{
	struct resource *r = platform_get_resource(dev, IORESOURCE_MEM, i);

	if (!r)
		return NULL;
	if (out)
		*out = r;
	return devm_ioremap_resource(&dev->dev, r);
}

void __iomem *nvgpu_devm_ioremap(struct device *dev, resource_size_t offset,
				 resource_size_t size)
{
	return devm_ioremap(dev, offset, size);
}

u64 nvgpu_resource_addr(struct platform_device *dev, int i)
{
	struct resource *r = platform_get_resource(dev, IORESOURCE_MEM, i);

	if (!r)
		return 0;

	return r->start;
}

static irqreturn_t gk20a_intr_isr_stall(int irq, void *dev_id)
{
	struct gk20a *g = dev_id;

	return nvgpu_intr_stall(g);
}

static irqreturn_t gk20a_intr_isr_nonstall(int irq, void *dev_id)
{
	struct gk20a *g = dev_id;

	return nvgpu_intr_nonstall(g);
}

static irqreturn_t gk20a_intr_thread_stall(int irq, void *dev_id)
{
	struct gk20a *g = dev_id;

	return nvgpu_intr_thread_stall(g);
}

void gk20a_remove_support(struct gk20a *g)
{
	struct nvgpu_os_linux *l = nvgpu_os_linux_from_gk20a(g);
	struct sim_nvgpu_linux *sim_linux;

	tegra_unregister_idle_unidle(gk20a_do_idle);

	nvgpu_kfree(g, g->dbg_regops_tmp_buf);

	nvgpu_remove_channel_support_linux(l);

	if (g->pmu.remove_support)
		g->pmu.remove_support(&g->pmu);

	if (g->acr.remove_support != NULL) {
		g->acr.remove_support(&g->acr);
	}

	if (g->gr.remove_support)
		g->gr.remove_support(&g->gr);

	if (g->mm.remove_ce_support)
		g->mm.remove_ce_support(&g->mm);

	if (g->fifo.remove_support)
		g->fifo.remove_support(&g->fifo);

	if (g->mm.remove_support)
		g->mm.remove_support(&g->mm);

	if (g->sim) {
		sim_linux = container_of(g->sim, struct sim_nvgpu_linux, sim);
		if (g->sim->remove_support)
			g->sim->remove_support(g);
		if (sim_linux->remove_support_linux)
			sim_linux->remove_support_linux(g);
	}

	nvgpu_remove_usermode_support(g);

	nvgpu_free_enabled_flags(g);

	gk20a_lockout_registers(g);
}

static int gk20a_init_support(struct platform_device *pdev)
{
	struct device *dev = &pdev->dev;
	struct gk20a *g = get_gk20a(dev);
	struct nvgpu_os_linux *l = nvgpu_os_linux_from_gk20a(g);
	int err = -ENOMEM;

	tegra_register_idle_unidle(gk20a_do_idle, gk20a_do_unidle, g);

	l->regs = nvgpu_devm_ioremap_resource(pdev,
					      GK20A_BAR0_IORESOURCE_MEM,
					      &l->reg_mem);
	if (IS_ERR(l->regs)) {
		nvgpu_err(g, "failed to remap gk20a registers");
		err = PTR_ERR(l->regs);
		goto fail;
	}

	l->regs_bus_addr = nvgpu_resource_addr(pdev,
			GK20A_BAR0_IORESOURCE_MEM);
	if (!l->regs_bus_addr) {
		nvgpu_err(g, "failed to read register bus offset");
		err = -ENODEV;
		goto fail;
	}

	l->bar1 = nvgpu_devm_ioremap_resource(pdev,
					      GK20A_BAR1_IORESOURCE_MEM,
					      &l->bar1_mem);
	if (IS_ERR(l->bar1)) {
		nvgpu_err(g, "failed to remap gk20a bar1");
		err = PTR_ERR(l->bar1);
		goto fail;
	}

	err = nvgpu_init_sim_support_linux(g, pdev);
	if (err)
		goto fail;
	err = nvgpu_init_sim_support(g);
	if (err)
		goto fail_sim;

	nvgpu_init_usermode_support(g);
	return 0;

fail_sim:
	nvgpu_remove_sim_support_linux(g);
fail:
	if (l->regs)
		l->regs = NULL;

	if (l->bar1)
		l->bar1 = NULL;

	return err;
}

static int gk20a_pm_railgate(struct device *dev)
{
	struct gk20a_platform *platform = dev_get_drvdata(dev);
	int ret = 0;
	struct gk20a *g = get_gk20a(dev);

	/* return early if platform didn't implement railgate */
	if (!platform->railgate)
		return 0;

	/* if platform is already railgated, then just return */
	if (platform->is_railgated && platform->is_railgated(dev))
		return ret;

#ifdef CONFIG_DEBUG_FS
	g->pstats.last_rail_gate_start = jiffies;

	if (g->pstats.railgating_cycle_count >= 1)
		g->pstats.total_rail_ungate_time_ms =
			g->pstats.total_rail_ungate_time_ms +
			jiffies_to_msecs(g->pstats.last_rail_gate_start -
					g->pstats.last_rail_ungate_complete);
#endif

	ret = platform->railgate(dev);
	if (ret) {
		nvgpu_err(g, "failed to railgate platform, err=%d", ret);
		return ret;
	}

#ifdef CONFIG_DEBUG_FS
	g->pstats.last_rail_gate_complete = jiffies;
#endif
	ret = tegra_fuse_clock_disable();
	if (ret)
		nvgpu_err(g, "failed to disable tegra fuse clock, err=%d", ret);

	return ret;
}

static int gk20a_pm_unrailgate(struct device *dev)
{
	struct gk20a_platform *platform = dev_get_drvdata(dev);
	int ret = 0;
	struct gk20a *g = get_gk20a(dev);

	/* return early if platform didn't implement unrailgate */
	if (!platform->unrailgate)
		return 0;

	ret = tegra_fuse_clock_enable();
	if (ret) {
		nvgpu_err(g, "failed to enable tegra fuse clock, err=%d", ret);
		return ret;
	}
#ifdef CONFIG_DEBUG_FS
	g->pstats.last_rail_ungate_start = jiffies;
	if (g->pstats.railgating_cycle_count >= 1)
		g->pstats.total_rail_gate_time_ms =
			g->pstats.total_rail_gate_time_ms +
			jiffies_to_msecs(g->pstats.last_rail_ungate_start -
				g->pstats.last_rail_gate_complete);

	g->pstats.railgating_cycle_count++;
#endif

	trace_gk20a_pm_unrailgate(dev_name(dev));

	nvgpu_mutex_acquire(&platform->railgate_lock);
	ret = platform->unrailgate(dev);
	nvgpu_mutex_release(&platform->railgate_lock);

#ifdef CONFIG_DEBUG_FS
	g->pstats.last_rail_ungate_complete = jiffies;
#endif

	return ret;
}

/*
 * Remove association of the driver with OS interrupt handler
 */
void nvgpu_free_irq(struct gk20a *g)
{
	struct device *dev = dev_from_gk20a(g);

	devm_free_irq(dev, g->irq_stall, g);
	if (g->irq_stall != g->irq_nonstall)
		devm_free_irq(dev, g->irq_nonstall, g);
}

/*
 * Idle the GPU in preparation of shutdown/remove.
 * gk20a_driver_start_unload() does not idle the GPU, but instead changes the SW
 * state to prevent further activity on the driver SW side.
 * On driver removal quiesce() should be called after start_unload()
 */
int nvgpu_quiesce(struct gk20a *g)
{
	int err;
	struct device *dev = dev_from_gk20a(g);

	if (g->power_on) {
		err = gk20a_wait_for_idle(g);
		if (err) {
			nvgpu_err(g, "failed to idle GPU, err=%d", err);
			return err;
		}

		err = gk20a_fifo_disable_all_engine_activity(g, true);
		if (err) {
			nvgpu_err(g,
				"failed to disable engine activity, err=%d",
				err);
		return err;
		}

		err = gk20a_fifo_wait_engine_idle(g);
		if (err) {
			nvgpu_err(g, "failed to idle engines, err=%d",
				err);
			return err;
		}
	}

	if (gk20a_gpu_is_virtual(dev))
		err = vgpu_pm_prepare_poweroff(dev);
	else
		err = gk20a_pm_prepare_poweroff(dev);

	if (err)
		nvgpu_err(g, "failed to prepare for poweroff, err=%d",
			err);

	return err;
}

static void gk20a_pm_shutdown(struct platform_device *pdev)
{
	struct gk20a_platform *platform = platform_get_drvdata(pdev);
	struct gk20a *g = platform->g;
	int err;

	nvgpu_info(g, "shutting down");

	/* vgpu has nothing to clean up currently */
	if (gk20a_gpu_is_virtual(&pdev->dev))
		return;

	if (!g->power_on)
		goto finish;

	gk20a_driver_start_unload(g);

	/* If GPU is already railgated,
	 * just prevent more requests, and return */
	if (platform->is_railgated && platform->is_railgated(&pdev->dev)) {
		__pm_runtime_disable(&pdev->dev, false);
		nvgpu_info(g, "already railgated, shut down complete");
		return;
	}

	/* Prevent more requests by disabling Runtime PM */
	__pm_runtime_disable(&pdev->dev, false);

	err = nvgpu_quiesce(g);
	if (err)
		goto finish;

	err = gk20a_pm_railgate(&pdev->dev);
	if (err)
		nvgpu_err(g, "failed to railgate, err=%d", err);

finish:
	nvgpu_info(g, "shut down complete");
}

#ifdef CONFIG_PM
static int gk20a_pm_runtime_resume(struct device *dev)
{
	int err = 0;

	err = gk20a_pm_unrailgate(dev);
	if (err)
		goto fail;

	if (gk20a_gpu_is_virtual(dev))
		err = vgpu_pm_finalize_poweron(dev);
	else
		err = gk20a_pm_finalize_poweron(dev);
	if (err)
		goto fail_poweron;

	return 0;

fail_poweron:
	gk20a_pm_railgate(dev);
fail:
	return err;
}

static int gk20a_pm_runtime_suspend(struct device *dev)
{
	int err = 0;
	struct gk20a *g = get_gk20a(dev);

	if (!g)
		return 0;

	if (gk20a_gpu_is_virtual(dev))
		err = vgpu_pm_prepare_poweroff(dev);
	else
		err = gk20a_pm_prepare_poweroff(dev);
	if (err) {
		nvgpu_err(g, "failed to power off, err=%d", err);
		goto fail;
	}

	err = gk20a_pm_railgate(dev);
	if (err)
		goto fail;

	return 0;

fail:
	gk20a_pm_finalize_poweron(dev);
	pm_runtime_mark_last_busy(dev);
	return err;
}

static int gk20a_pm_suspend(struct device *dev)
{
	struct gk20a_platform *platform = dev_get_drvdata(dev);
	struct gk20a *g = get_gk20a(dev);
	int ret = 0;
	int usage_count;
	struct nvgpu_timeout timeout;

	if (!g->power_on) {
		if (platform->suspend)
			ret = platform->suspend(dev);

		if (ret)
			return ret;

		if (!pm_runtime_enabled(dev))
			ret = gk20a_pm_railgate(dev);

		return ret;
	}

	nvgpu_timeout_init(g, &timeout, GK20A_WAIT_FOR_IDLE_MS,
			   NVGPU_TIMER_CPU_TIMER);
	/*
	 * Hold back deterministic submits and changes to deterministic
	 * channels - this must be outside the power busy locks.
	 */
	gk20a_channel_deterministic_idle(g);

	/* check and wait until GPU is idle (with a timeout) */
	do {
		nvgpu_usleep_range(1000, 1100);
		usage_count = nvgpu_atomic_read(&g->usage_count);
	} while (usage_count != 0 && !nvgpu_timeout_expired(&timeout));

	if (usage_count != 0) {
		nvgpu_err(g, "failed to idle - usage_count %d", usage_count);
		ret = -EINVAL;
		goto fail_idle;
	}

	ret = gk20a_pm_runtime_suspend(dev);
	if (ret)
		goto fail_idle;

	if (platform->suspend)
		ret = platform->suspend(dev);
	if (ret)
		goto fail_suspend;

	g->suspended = true;

	return 0;

fail_suspend:
	gk20a_pm_runtime_resume(dev);
fail_idle:
	gk20a_channel_deterministic_unidle(g);
	return ret;
}

static int gk20a_pm_resume(struct device *dev)
{
	struct gk20a_platform *platform = dev_get_drvdata(dev);
	struct gk20a *g = get_gk20a(dev);
	int ret = 0;

	if (!g->suspended) {
		if (platform->resume)
			ret = platform->resume(dev);
		if (ret)
			return ret;

		if (!pm_runtime_enabled(dev))
			ret = gk20a_pm_unrailgate(dev);

		return ret;
	}

	if (platform->resume)
		ret = platform->resume(dev);
	if (ret)
		return ret;

	ret = gk20a_pm_runtime_resume(dev);
	if (ret)
		return ret;

	g->suspended = false;

	gk20a_channel_deterministic_unidle(g);

	return ret;
}

static const struct dev_pm_ops gk20a_pm_ops = {
	.runtime_resume = gk20a_pm_runtime_resume,
	.runtime_suspend = gk20a_pm_runtime_suspend,
	.resume = gk20a_pm_resume,
	.suspend = gk20a_pm_suspend,
};
#endif

static int gk20a_pm_init(struct device *dev)
{
	struct gk20a *g = get_gk20a(dev);
	int err = 0;

	nvgpu_log_fn(g, " ");

	/*
	 * Initialise pm runtime. For railgate disable
	 * case, set autosuspend delay to negative which
	 * will suspend runtime pm
	 */
	if (g->railgate_delay && nvgpu_is_enabled(g, NVGPU_CAN_RAILGATE))
		pm_runtime_set_autosuspend_delay(dev,
				 g->railgate_delay);
	else
		pm_runtime_set_autosuspend_delay(dev, -1);

	pm_runtime_use_autosuspend(dev);
	pm_runtime_enable(dev);

	return err;
}

static int gk20a_pm_deinit(struct device *dev)
{
	pm_runtime_dont_use_autosuspend(dev);
	pm_runtime_disable(dev);
	return 0;
}

/*
 * Start the process for unloading the driver. Set NVGPU_DRIVER_IS_DYING.
 */
void gk20a_driver_start_unload(struct gk20a *g)
{
	struct nvgpu_os_linux *l = nvgpu_os_linux_from_gk20a(g);

	nvgpu_log(g, gpu_dbg_shutdown, "Driver is now going down!\n");

	down_write(&l->busy_lock);
	__nvgpu_set_enabled(g, NVGPU_DRIVER_IS_DYING, true);
	/* GR SW ready needs to be invalidated at this time with the busy lock
	 * held to prevent a racing condition on the gr/mm code */
	g->gr.sw_ready = false;
	g->sw_ready = false;
	up_write(&l->busy_lock);

	if (g->is_virtual)
		return;

	gk20a_wait_for_idle(g);

	nvgpu_wait_for_deferred_interrupts(g);

	if (l->nonstall_work_queue) {
		cancel_work_sync(&l->nonstall_fn_work);
		destroy_workqueue(l->nonstall_work_queue);
		l->nonstall_work_queue = NULL;
	}
}

static inline void set_gk20a(struct platform_device *pdev, struct gk20a *gk20a)
{
	gk20a_get_platform(&pdev->dev)->g = gk20a;
}

static int nvgpu_read_fuse_overrides(struct gk20a *g)
{
	struct device_node *np = nvgpu_get_node(g);
	struct gk20a_platform *platform = dev_get_drvdata(dev_from_gk20a(g));
	u32 *fuses;
	int count, i;

	if (!np) /* may be pcie device */
		return 0;

	count = of_property_count_elems_of_size(np, "fuse-overrides", 8);
	if (count <= 0)
		return count;

	fuses = nvgpu_kmalloc(g, sizeof(u32) * count * 2);
	if (!fuses)
		return -ENOMEM;
	of_property_read_u32_array(np, "fuse-overrides", fuses, count * 2);
	for (i = 0; i < count; i++) {
		u32 fuse, value;

		fuse = fuses[2 * i];
		value = fuses[2 * i + 1];
		switch (fuse) {
		case GM20B_FUSE_OPT_TPC_DISABLE:
			g->tpc_fs_mask_user = ~value;
			break;
		case GP10B_FUSE_OPT_ECC_EN:
			g->gr.fecs_feature_override_ecc_val = value;
			break;
		case GV11B_FUSE_OPT_TPC_DISABLE:
			if (platform->set_tpc_pg_mask != NULL)
				platform->set_tpc_pg_mask(dev_from_gk20a(g),
								value);
			break;
		default:
			nvgpu_err(g, "ignore unknown fuse override %08x", fuse);
			break;
		}
	}

	nvgpu_kfree(g, fuses);

	return 0;
}

static int gk20a_probe(struct platform_device *dev)
{
	struct nvgpu_os_linux *l = NULL;
	struct gk20a *gk20a;
	int err;
	struct gk20a_platform *platform = NULL;
	struct device_node *np;

	if (dev->dev.of_node) {
		const struct of_device_id *match;

		match = of_match_device(tegra_gk20a_of_match, &dev->dev);
		if (match)
			platform = (struct gk20a_platform *)match->data;
	} else
		platform = (struct gk20a_platform *)dev->dev.platform_data;

	if (!platform) {
		dev_err(&dev->dev, "no platform data\n");
		return -ENODATA;
	}

	platform_set_drvdata(dev, platform);

	if (gk20a_gpu_is_virtual(&dev->dev))
		return vgpu_probe(dev);

	l = kzalloc(sizeof(*l), GFP_KERNEL);
	if (!l) {
		dev_err(&dev->dev, "couldn't allocate gk20a support");
		return -ENOMEM;
	}

	hash_init(l->ecc_sysfs_stats_htable);

	gk20a = &l->g;

	nvgpu_log_fn(gk20a, " ");

	nvgpu_init_gk20a(gk20a);
	set_gk20a(dev, gk20a);
	l->dev = &dev->dev;
	gk20a->log_mask = NVGPU_DEFAULT_DBG_MASK;

	nvgpu_kmem_init(gk20a);

	err = nvgpu_init_enabled_flags(gk20a);
	if (err)
		goto return_err;

	np = nvgpu_get_node(gk20a);
	if (of_dma_is_coherent(np)) {
		__nvgpu_set_enabled(gk20a, NVGPU_USE_COHERENT_SYSMEM, true);
		__nvgpu_set_enabled(gk20a, NVGPU_SUPPORT_IO_COHERENCE, true);
	}

	if (nvgpu_platform_is_simulation(gk20a))
		__nvgpu_set_enabled(gk20a, NVGPU_IS_FMODEL, true);

	gk20a->irq_stall = platform_get_irq(dev, 0);
	gk20a->irq_nonstall = platform_get_irq(dev, 1);
	if (gk20a->irq_stall < 0 || gk20a->irq_nonstall < 0) {
		err = -ENXIO;
		goto return_err;
	}

	err = devm_request_threaded_irq(&dev->dev,
			gk20a->irq_stall,
			gk20a_intr_isr_stall,
			gk20a_intr_thread_stall,
			0, "gk20a_stall", gk20a);
	if (err) {
		dev_err(&dev->dev,
			"failed to request stall intr irq @ %d\n",
				gk20a->irq_stall);
		goto return_err;
	}
	err = devm_request_irq(&dev->dev,
			gk20a->irq_nonstall,
			gk20a_intr_isr_nonstall,
			0, "gk20a_nonstall", gk20a);
	if (err) {
		dev_err(&dev->dev,
			"failed to request non-stall intr irq @ %d\n",
				gk20a->irq_nonstall);
		goto return_err;
	}
	disable_irq(gk20a->irq_stall);
	if (gk20a->irq_stall != gk20a->irq_nonstall)
		disable_irq(gk20a->irq_nonstall);

	err = gk20a_init_support(dev);
	if (err)
		goto return_err;

	err = nvgpu_read_fuse_overrides(gk20a);

#ifdef CONFIG_RESET_CONTROLLER
	platform->reset_control = devm_reset_control_get(&dev->dev, NULL);
	if (IS_ERR(platform->reset_control))
		platform->reset_control = NULL;
#endif

	err = nvgpu_probe(gk20a, "gpu.0", INTERFACE_NAME, &nvgpu_class);
	if (err)
		goto return_err;

	err = gk20a_pm_init(&dev->dev);
	if (err) {
		dev_err(&dev->dev, "pm init failed");
		goto return_err;
	}

	gk20a->nvgpu_reboot_nb.notifier_call =
		nvgpu_kernel_shutdown_notification;
	err = register_reboot_notifier(&gk20a->nvgpu_reboot_nb);
	if (err)
		goto return_err;

	return 0;

return_err:
	nvgpu_free_enabled_flags(gk20a);

	/*
	 * Last since the above allocs may use data structures in here.
	 */
	nvgpu_kmem_fini(gk20a, NVGPU_KMEM_FINI_FORCE_CLEANUP);

	kfree(l);

	return err;
}

int nvgpu_remove(struct device *dev, struct class *class)
{
	struct gk20a *g = get_gk20a(dev);
#ifdef CONFIG_NVGPU_SUPPORT_CDE
	struct nvgpu_os_linux *l = nvgpu_os_linux_from_gk20a(g);
#endif
	struct gk20a_platform *platform = gk20a_get_platform(dev);
	int err;

	nvgpu_log_fn(g, " ");

	err = nvgpu_quiesce(g);
	WARN(err, "gpu failed to idle during driver removal");

	if (nvgpu_mem_is_valid(&g->syncpt_mem))
		nvgpu_dma_free(g, &g->syncpt_mem);

#ifdef CONFIG_NVGPU_SUPPORT_CDE
	if (platform->has_cde)
		gk20a_cde_destroy(l);
#endif

#ifdef CONFIG_GK20A_CTXSW_TRACE
	gk20a_ctxsw_trace_cleanup(g);
#endif

	gk20a_sched_ctrl_cleanup(g);

	if (IS_ENABLED(CONFIG_GK20A_DEVFREQ))
		gk20a_scale_exit(dev);

	nvgpu_clk_arb_cleanup_arbiter(g);

	gk20a_user_deinit(dev, class);

	gk20a_debug_deinit(g);

	nvgpu_remove_sysfs(dev);

	if (platform->secure_buffer.destroy)
		platform->secure_buffer.destroy(g,
				&platform->secure_buffer);

	if (platform->remove)
		platform->remove(dev);

	nvgpu_mutex_destroy(&g->clk_arb_enable_lock);

	nvgpu_log_fn(g, "removed");

	return err;
}

static int __exit gk20a_remove(struct platform_device *pdev)
{
	int err;
	struct device *dev = &pdev->dev;
	struct gk20a *g = get_gk20a(dev);

	if (gk20a_gpu_is_virtual(dev))
		return vgpu_remove(pdev);

	err = nvgpu_remove(dev, &nvgpu_class);

	unregister_reboot_notifier(&g->nvgpu_reboot_nb);

	set_gk20a(pdev, NULL);

	gk20a_put(g);

	gk20a_pm_deinit(dev);

	return err;
}

static struct platform_driver gk20a_driver = {
	.probe = gk20a_probe,
	.remove = __exit_p(gk20a_remove),
	.shutdown = gk20a_pm_shutdown,
	.driver = {
		.owner = THIS_MODULE,
		.name = "gk20a",
		.probe_type = PROBE_PREFER_ASYNCHRONOUS,
#ifdef CONFIG_OF
		.of_match_table = tegra_gk20a_of_match,
#endif
#ifdef CONFIG_PM
		.pm = &gk20a_pm_ops,
#endif
		.suppress_bind_attrs = true,
	}
};

struct class nvgpu_class = {
	.owner = THIS_MODULE,
	.name = CLASS_NAME,
};

static int __init gk20a_init(void)
{

	int ret;

	ret = class_register(&nvgpu_class);
	if (ret)
		return ret;

	ret = nvgpu_pci_init();
	if (ret)
		return ret;

	return platform_driver_register(&gk20a_driver);
}

static void __exit gk20a_exit(void)
{
	nvgpu_pci_exit();
	platform_driver_unregister(&gk20a_driver);
	class_unregister(&nvgpu_class);
}

MODULE_LICENSE("GPL v2");
module_init(gk20a_init);
module_exit(gk20a_exit);