summaryrefslogtreecommitdiffstats
path: root/include/uapi/linux/nvgpu.h
blob: b14610bd35f99d8cbf80200cce79536a5c27b428 (plain) (blame)
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
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809
1810
1811
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
1839
1840
1841
1842
1843
1844
1845
1846
1847
1848
1849
1850
1851
1852
1853
1854
1855
1856
1857
1858
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
1906
1907
1908
1909
1910
1911
1912
1913
1914
1915
1916
1917
1918
1919
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
2025
2026
2027
2028
2029
2030
2031
2032
2033
2034
2035
2036
2037
2038
2039
2040
2041
2042
2043
2044
2045
2046
2047
2048
2049
2050
2051
2052
2053
2054
2055
2056
2057
2058
2059
2060
2061
2062
2063
2064
2065
2066
2067
2068
2069
2070
2071
2072
2073
2074
2075
2076
2077
2078
2079
2080
2081
2082
2083
2084
2085
2086
2087
2088
2089
2090
2091
2092
2093
2094
2095
2096
2097
2098
2099
2100
2101
2102
2103
2104
2105
2106
2107
2108
2109
2110
2111
2112
2113
2114
2115
2116
2117
2118
2119
2120
2121
2122
2123
2124
2125
2126
2127
2128
2129
2130
2131
2132
2133
2134
2135
2136
2137
2138
2139
2140
2141
2142
2143
2144
2145
2146
2147
2148
2149
2150
2151
2152
2153
2154
2155
2156
2157
2158
2159
2160
2161
2162
2163
2164
2165
2166
2167
2168
2169
2170
2171
2172
2173
2174
2175
2176
2177
2178
2179
2180
2181
2182
2183
2184
2185
2186
2187
2188
2189
2190
2191
2192
2193
2194
2195
2196
2197
2198
2199
2200
2201
2202
2203
2204
2205
2206
2207
2208
2209
2210
2211
2212
2213
2214
2215
2216
2217
2218
2219
2220
2221
2222
2223
2224
2225
2226
2227
2228
2229
2230
2231
2232
2233
2234
2235
2236
2237
2238
2239
2240
2241
2242
2243
2244
2245
2246
2247
2248
2249
2250
2251
2252
2253
2254
2255
/*
 * NVGPU Public Interface Header
 *
 * Copyright (c) 2011-2018, 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.
 */

#ifndef _UAPI__LINUX_NVGPU_IOCTL_H
#define _UAPI__LINUX_NVGPU_IOCTL_H

#include <linux/ioctl.h>
#include <linux/types.h>

#if !defined(__KERNEL__)
#define __user
#define __packed __attribute__((packed))
#endif

/*
 * /dev/nvhost-ctrl-gpu device
 *
 * Opening a '/dev/nvhost-ctrl-gpu' device node creates a way to send
 * ctrl ioctl to gpu driver.
 *
 * /dev/nvhost-gpu is for channel (context specific) operations. We use
 * /dev/nvhost-ctrl-gpu for global (context independent) operations on
 * gpu device.
 */

#define NVGPU_GPU_IOCTL_MAGIC 'G'

/* return zcull ctx size */
struct nvgpu_gpu_zcull_get_ctx_size_args {
	__u32 size;
} __packed;

/* return zcull info */
struct nvgpu_gpu_zcull_get_info_args {
	__u32 width_align_pixels;
	__u32 height_align_pixels;
	__u32 pixel_squares_by_aliquots;
	__u32 aliquot_total;
	__u32 region_byte_multiplier;
	__u32 region_header_size;
	__u32 subregion_header_size;
	__u32 subregion_width_align_pixels;
	__u32 subregion_height_align_pixels;
	__u32 subregion_count;
};

#define NVGPU_ZBC_COLOR_VALUE_SIZE	4
#define NVGPU_ZBC_TYPE_INVALID		0
#define NVGPU_ZBC_TYPE_COLOR		1
#define NVGPU_ZBC_TYPE_DEPTH		2

struct nvgpu_gpu_zbc_set_table_args {
	__u32 color_ds[NVGPU_ZBC_COLOR_VALUE_SIZE];
	__u32 color_l2[NVGPU_ZBC_COLOR_VALUE_SIZE];
	__u32 depth;
	__u32 format;
	__u32 type;	/* color or depth */
} __packed;

struct nvgpu_gpu_zbc_query_table_args {
	__u32 color_ds[NVGPU_ZBC_COLOR_VALUE_SIZE];
	__u32 color_l2[NVGPU_ZBC_COLOR_VALUE_SIZE];
	__u32 depth;
	__u32 ref_cnt;
	__u32 format;
	__u32 type;		/* color or depth */
	__u32 index_size;	/* [out] size, [in] index */
} __packed;


/* This contains the minimal set by which the userspace can
   determine all the properties of the GPU */
#define NVGPU_GPU_ARCH_GK100	0x000000E0
#define NVGPU_GPU_ARCH_GM200	0x00000120
#define NVGPU_GPU_ARCH_GP100	0x00000130
#define NVGPU_GPU_ARCH_GV110	0x00000150
#define NVGPU_GPU_ARCH_GV100	0x00000140

#define NVGPU_GPU_IMPL_GK20A	0x0000000A
#define NVGPU_GPU_IMPL_GM204	0x00000004
#define NVGPU_GPU_IMPL_GM206	0x00000006
#define NVGPU_GPU_IMPL_GM20B	0x0000000B
#define NVGPU_GPU_IMPL_GM20B_B	0x0000000E
#define NVGPU_GPU_IMPL_GP104	0x00000004
#define NVGPU_GPU_IMPL_GP106	0x00000006
#define NVGPU_GPU_IMPL_GP10B	0x0000000B
#define NVGPU_GPU_IMPL_GV11B	0x0000000B
#define NVGPU_GPU_IMPL_GV100	0x00000000

#define NVGPU_GPU_BUS_TYPE_NONE         0
#define NVGPU_GPU_BUS_TYPE_AXI         32

#define NVGPU_GPU_FLAGS_HAS_SYNCPOINTS			(1ULL << 0)
/* MAP_BUFFER_EX with partial mappings */
#define NVGPU_GPU_FLAGS_SUPPORT_PARTIAL_MAPPINGS	(1ULL << 1)
/* MAP_BUFFER_EX with sparse allocations */
#define NVGPU_GPU_FLAGS_SUPPORT_SPARSE_ALLOCS		(1ULL << 2)
/* sync fence FDs are available in, e.g., submit_gpfifo */
#define NVGPU_GPU_FLAGS_SUPPORT_SYNC_FENCE_FDS		(1ULL << 3)
/* NVGPU_IOCTL_CHANNEL_CYCLE_STATS is available */
#define NVGPU_GPU_FLAGS_SUPPORT_CYCLE_STATS		(1ULL << 4)
/* NVGPU_IOCTL_CHANNEL_CYCLE_STATS_SNAPSHOT is available */
#define NVGPU_GPU_FLAGS_SUPPORT_CYCLE_STATS_SNAPSHOT	(1ULL << 6)
/* User-space managed address spaces support */
#define NVGPU_GPU_FLAGS_SUPPORT_USERSPACE_MANAGED_AS	(1ULL << 7)
/* Both gpu driver and device support TSG */
#define NVGPU_GPU_FLAGS_SUPPORT_TSG			(1ULL << 8)
/* Clock control support */
#define NVGPU_GPU_FLAGS_SUPPORT_CLOCK_CONTROLS		(1ULL << 9)
/* NVGPU_GPU_IOCTL_GET_VOLTAGE is available */
#define NVGPU_GPU_FLAGS_SUPPORT_GET_VOLTAGE		(1ULL << 10)
/* NVGPU_GPU_IOCTL_GET_CURRENT is available */
#define NVGPU_GPU_FLAGS_SUPPORT_GET_CURRENT		(1ULL << 11)
/* NVGPU_GPU_IOCTL_GET_POWER is available */
#define NVGPU_GPU_FLAGS_SUPPORT_GET_POWER		(1ULL << 12)
/* NVGPU_GPU_IOCTL_GET_TEMPERATURE is available */
#define NVGPU_GPU_FLAGS_SUPPORT_GET_TEMPERATURE		(1ULL << 13)
/* NVGPU_GPU_IOCTL_SET_THERM_ALERT_LIMIT is available */
#define NVGPU_GPU_FLAGS_SUPPORT_SET_THERM_ALERT_LIMIT	(1ULL << 14)
/* NVGPU_GPU_IOCTL_GET_EVENT_FD is available */
#define NVGPU_GPU_FLAGS_SUPPORT_DEVICE_EVENTS		(1ULL << 15)
/* FECS context switch tracing is available */
#define NVGPU_GPU_FLAGS_SUPPORT_FECS_CTXSW_TRACE	(1ULL << 16)
/* NVGPU_AS_IOCTL_MAP_BUFFER_COMPBITS is available */
#define NVGPU_GPU_FLAGS_SUPPORT_MAP_COMPBITS		(1ULL << 17)
/* Fast deterministic submits with no job tracking are supported */
#define NVGPU_GPU_FLAGS_SUPPORT_DETERMINISTIC_SUBMIT_NO_JOBTRACKING (1ULL << 18)
/* Deterministic submits are supported even with job tracking */
#define NVGPU_GPU_FLAGS_SUPPORT_DETERMINISTIC_SUBMIT_FULL (1ULL << 19)
/* IO coherence support is available */
#define NVGPU_GPU_FLAGS_SUPPORT_IO_COHERENCE		(1ULL << 20)
/* NVGPU_IOCTL_CHANNEL_RESCHEDULE_RUNLIST is available */
#define NVGPU_GPU_FLAGS_SUPPORT_RESCHEDULE_RUNLIST	(1ULL << 21)
/*  subcontexts are available */
#define NVGPU_GPU_FLAGS_SUPPORT_TSG_SUBCONTEXTS         (1ULL << 22)
/* Direct PTE kind control is supported (map_buffer_ex) */
#define NVGPU_GPU_FLAGS_SUPPORT_MAP_DIRECT_KIND_CTRL	(1ULL << 23)
/* NVGPU_GPU_IOCTL_SET_DETERMINISTIC_OPTS is available */
#define NVGPU_GPU_FLAGS_SUPPORT_DETERMINISTIC_OPTS	(1ULL << 24)
/* SCG support is available */
#define NVGPU_GPU_FLAGS_SUPPORT_SCG			(1ULL << 25)
/* GPU_VA address of a syncpoint is supported */
#define NVGPU_GPU_FLAGS_SUPPORT_SYNCPOINT_ADDRESS	(1ULL << 26)
/* VPR is supported */
#define NVGPU_GPU_FLAGS_SUPPORT_VPR			(1ULL << 27)
/* Allocating per-channel syncpoint in user space is supported */
#define NVGPU_GPU_FLAGS_SUPPORT_USER_SYNCPOINT		(1ULL << 28)
/* SM LRF ECC is enabled */
#define NVGPU_GPU_FLAGS_ECC_ENABLED_SM_LRF	(1ULL << 60)
/* SM SHM ECC is enabled */
#define NVGPU_GPU_FLAGS_ECC_ENABLED_SM_SHM	(1ULL << 61)
/* TEX ECC is enabled */
#define NVGPU_GPU_FLAGS_ECC_ENABLED_TEX		(1ULL << 62)
/* L2 ECC is enabled */
#define NVGPU_GPU_FLAGS_ECC_ENABLED_LTC		(1ULL << 63)
/* All types of ECC are enabled */
#define NVGPU_GPU_FLAGS_ALL_ECC_ENABLED	\
				(NVGPU_GPU_FLAGS_ECC_ENABLED_SM_LRF |	\
				NVGPU_GPU_FLAGS_ECC_ENABLED_SM_SHM |	\
				NVGPU_GPU_FLAGS_ECC_ENABLED_TEX    |	\
				NVGPU_GPU_FLAGS_ECC_ENABLED_LTC)

struct nvgpu_gpu_characteristics {
	__u32 arch;
	__u32 impl;
	__u32 rev;
	__u32 num_gpc;

	__u64 L2_cache_size;               /* bytes */
	__u64 on_board_video_memory_size;  /* bytes */

	__u32 num_tpc_per_gpc; /* the architectural maximum */
	__u32 bus_type;

	__u32 big_page_size; /* the default big page size */
	__u32 compression_page_size;

	__u32 pde_coverage_bit_count;

	/* bit N set ==> big page size 2^N is available in
	   NVGPU_GPU_IOCTL_ALLOC_AS. The default big page size is
	   always available regardless of this field. */
	__u32 available_big_page_sizes;

	__u64 flags;

	__u32 twod_class;
	__u32 threed_class;
	__u32 compute_class;
	__u32 gpfifo_class;
	__u32 inline_to_memory_class;
	__u32 dma_copy_class;

	__u32 gpc_mask; /* enabled GPCs */

	__u32 sm_arch_sm_version; /* sm version */
	__u32 sm_arch_spa_version; /* sm instruction set */
	__u32 sm_arch_warp_count;

	/* IOCTL interface levels by service. -1 if not supported */
	__s16 gpu_ioctl_nr_last;
	__s16 tsg_ioctl_nr_last;
	__s16 dbg_gpu_ioctl_nr_last;
	__s16 ioctl_channel_nr_last;
	__s16 as_ioctl_nr_last;

	__u8 gpu_va_bit_count;
	__u8 reserved;

	__u32 max_fbps_count;
	__u32 fbp_en_mask;
	__u32 max_ltc_per_fbp;
	__u32 max_lts_per_ltc;
	__u32 max_tex_per_tpc;
	__u32 max_gpc_count;
	/* mask of Rop_L2 for each FBP */
	__u32 rop_l2_en_mask_DEPRECATED[2];


	__u8 chipname[8];

	__u64 gr_compbit_store_base_hw;
	__u32 gr_gobs_per_comptagline_per_slice;
	__u32 num_ltc;
	__u32 lts_per_ltc;
	__u32 cbc_cache_line_size;
	__u32 cbc_comptags_per_line;

	/* MAP_BUFFER_BATCH: the upper limit for num_unmaps and
	 * num_maps */
	__u32 map_buffer_batch_limit;

	__u64 max_freq;

	/* supported preemption modes */
	__u32 graphics_preemption_mode_flags; /* NVGPU_GRAPHICS_PREEMPTION_MODE_* */
	__u32 compute_preemption_mode_flags; /* NVGPU_COMPUTE_PREEMPTION_MODE_* */
	/* default preemption modes */
	__u32 default_graphics_preempt_mode; /* NVGPU_GRAPHICS_PREEMPTION_MODE_* */
	__u32 default_compute_preempt_mode; /* NVGPU_COMPUTE_PREEMPTION_MODE_* */

	__u64 local_video_memory_size; /* in bytes, non-zero only for dGPUs */

	/* These are meaningful only for PCI devices */
	__u16 pci_vendor_id, pci_device_id;
	__u16 pci_subsystem_vendor_id, pci_subsystem_device_id;
	__u16 pci_class;
	__u8  pci_revision;
	__u8  vbios_oem_version;
	__u32 vbios_version;

	/* NVGPU_DBG_GPU_IOCTL_REG_OPS: the upper limit for the number
	 * of regops */
	__u32 reg_ops_limit;
	__u32 reserved1;

	__s16 event_ioctl_nr_last;
	__u16 pad;

	__u32 max_css_buffer_size;

	/* Notes:
	   - This struct can be safely appended with new fields. However, always
	     keep the structure size multiple of 8 and make sure that the binary
	     layout does not change between 32-bit and 64-bit architectures.
	   - If the last field is reserved/padding, it is not
	     generally safe to repurpose the field in future revisions.
	*/
};

struct nvgpu_gpu_get_characteristics {
	/* [in]  size reserved by the user space. Can be 0.
	   [out] full buffer size by kernel */
	__u64 gpu_characteristics_buf_size;

	/* [in]  address of nvgpu_gpu_characteristics buffer. Filled with field
	   values by exactly MIN(buf_size_in, buf_size_out) bytes. Ignored, if
	   buf_size_in is zero.  */
	__u64 gpu_characteristics_buf_addr;
};

#define NVGPU_GPU_COMPBITS_NONE		0
#define NVGPU_GPU_COMPBITS_GPU		(1 << 0)
#define NVGPU_GPU_COMPBITS_CDEH		(1 << 1)
#define NVGPU_GPU_COMPBITS_CDEV		(1 << 2)

struct nvgpu_gpu_prepare_compressible_read_args {
	__u32 handle;			/* in, dmabuf fd */
	union {
		__u32 request_compbits;	/* in */
		__u32 valid_compbits;	/* out */
	};
	__u64 offset;			/* in, within handle */
	__u64 compbits_hoffset;		/* in, within handle */
	__u64 compbits_voffset;		/* in, within handle */
	__u32 width;			/* in, in pixels */
	__u32 height;			/* in, in pixels */
	__u32 block_height_log2;	/* in */
	__u32 submit_flags;		/* in (NVGPU_SUBMIT_GPFIFO_FLAGS_) */
	union {
		struct {
			__u32 syncpt_id;
			__u32 syncpt_value;
		};
		__s32 fd;
	} fence;			/* in/out */
	__u32 zbc_color;		/* out */
	__u32 reserved;		/* must be zero */
	__u64 scatterbuffer_offset;	/* in, within handle */
	__u32 reserved2[2];		/* must be zero */
};

struct nvgpu_gpu_mark_compressible_write_args {
	__u32 handle;			/* in, dmabuf fd */
	__u32 valid_compbits;		/* in */
	__u64 offset;			/* in, within handle */
	__u32 zbc_color;		/* in */
	__u32 reserved[3];		/* must be zero */
};

struct nvgpu_alloc_as_args {
	__u32 big_page_size;
	__s32 as_fd;

/*
 * The GPU address space will be managed by the userspace. This has
 * the following changes in functionality:
 *   1. All non-fixed-offset user mappings are rejected (i.e.,
 *      fixed-offset only)
 *   2. Address space does not need to be allocated for fixed-offset
 *      mappings, except to mark sparse address space areas.
 *   3. Maps and unmaps are immediate. In particular, mapping ref
 *      increments at kickoffs and decrements at job completion are
 *      bypassed.
 */
#define NVGPU_GPU_IOCTL_ALLOC_AS_FLAGS_USERSPACE_MANAGED (1 << 0)
	__u32 flags;

	__u32 reserved;			/* must be zero */
};

struct nvgpu_gpu_open_tsg_args {
	__u32 tsg_fd;			/* out, tsg fd */
	__u32 reserved;			/* must be zero */
};

struct nvgpu_gpu_get_tpc_masks_args {
	/* [in]  TPC mask buffer size reserved by userspace. Should be
		 at least sizeof(__u32) * fls(gpc_mask) to receive TPC
		 mask for each GPC.
	   [out] full kernel buffer size
	*/
	__u32 mask_buf_size;
	__u32 reserved;

	/* [in]  pointer to TPC mask buffer. It will receive one
		 32-bit TPC mask per GPC or 0 if GPC is not enabled or
		 not present. This parameter is ignored if
		 mask_buf_size is 0. */
	__u64 mask_buf_addr;
};

struct nvgpu_gpu_open_channel_args {
	union {
		__s32 channel_fd; /* deprecated: use out.channel_fd instead */
		struct {
			 /* runlist_id is the runlist for the
			  * channel. Basically, the runlist specifies the target
			  * engine(s) for which the channel is
			  * opened. Runlist_id -1 is synonym for the primary
			  * graphics runlist. */
			__s32 runlist_id;
		} in;
		struct {
			__s32 channel_fd;
		} out;
	};
};

/* L2 cache writeback, optionally invalidate clean lines and flush fb */
struct nvgpu_gpu_l2_fb_args {
	__u32 l2_flush:1;
	__u32 l2_invalidate:1;
	__u32 fb_flush:1;
	__u32 reserved;
} __packed;

struct nvgpu_gpu_inval_icache_args {
	int channel_fd;
	__u32 reserved;
} __packed;

struct nvgpu_gpu_mmu_debug_mode_args {
	__u32 state;
	__u32 reserved;
} __packed;

struct nvgpu_gpu_sm_debug_mode_args {
	int channel_fd;
	__u32 enable;
	__u64 sms;
} __packed;

struct warpstate {
	__u64 valid_warps[2];
	__u64 trapped_warps[2];
	__u64 paused_warps[2];
};

struct nvgpu_gpu_wait_pause_args {
	__u64 pwarpstate;
};

struct nvgpu_gpu_tpc_exception_en_status_args {
	__u64 tpc_exception_en_sm_mask;
};

struct nvgpu_gpu_num_vsms {
	__u32 num_vsms;
	__u32 reserved;
};

struct nvgpu_gpu_vsms_mapping_entry {
	__u8 gpc_index;
	__u8 tpc_index;
};

struct nvgpu_gpu_vsms_mapping {
	__u64 vsms_map_buf_addr;
};

struct nvgpu_gpu_get_buffer_info_args {
	union {
		struct {
			__u32 dmabuf_fd;   /* dma-buf fd */
		} in;
		struct {
			__u64 id;          /* Unique within live
					    * buffers */
			__u64 length;      /* Allocated length of the
					    * buffer */
			__u64 reserved0;
			__u64 reserved1;
		} out;
	};
};

#define NVGPU_GPU_GET_CPU_TIME_CORRELATION_INFO_MAX_COUNT		16
#define NVGPU_GPU_GET_CPU_TIME_CORRELATION_INFO_SRC_ID_TSC		1

struct nvgpu_gpu_get_cpu_time_correlation_sample {
	/* gpu timestamp value */
	__u64 cpu_timestamp;
	/* raw GPU counter (PTIMER) value */
	__u64 gpu_timestamp;
};

struct nvgpu_gpu_get_cpu_time_correlation_info_args {
	/* timestamp pairs */
	struct nvgpu_gpu_get_cpu_time_correlation_sample samples[
		NVGPU_GPU_GET_CPU_TIME_CORRELATION_INFO_MAX_COUNT];
	/* number of pairs to read */
	__u32 count;
	/* cpu clock source id */
	__u32 source_id;
};

struct nvgpu_gpu_get_gpu_time_args {
	/* raw GPU counter (PTIMER) value */
	__u64 gpu_timestamp;

	/* reserved for future extensions */
	__u64 reserved;
};

struct nvgpu_gpu_get_engine_info_item {

#define NVGPU_GPU_ENGINE_ID_GR 0
#define NVGPU_GPU_ENGINE_ID_GR_COPY 1
#define NVGPU_GPU_ENGINE_ID_ASYNC_COPY 2
	__u32 engine_id;

	__u32 engine_instance;

	/* runlist id for opening channels to the engine, or -1 if
	 * channels are not supported */
	__s32 runlist_id;

	__u32 reserved;
};

struct nvgpu_gpu_get_engine_info_args {
	/* [in]  Buffer size reserved by userspace.
	 *
	 * [out] Full kernel buffer size. Multiple of sizeof(struct
	 *       nvgpu_gpu_get_engine_info_item)
	*/
	__u32 engine_info_buf_size;
	__u32 reserved;
	__u64 engine_info_buf_addr;
};

#define NVGPU_GPU_ALLOC_VIDMEM_FLAG_CONTIGUOUS		(1U << 0)

/* CPU access and coherency flags (3 bits). Use CPU access with care,
 * BAR resources are scarce. */
#define NVGPU_GPU_ALLOC_VIDMEM_FLAG_CPU_NOT_MAPPABLE	(0U << 1)
#define NVGPU_GPU_ALLOC_VIDMEM_FLAG_CPU_WRITE_COMBINE	(1U << 1)
#define NVGPU_GPU_ALLOC_VIDMEM_FLAG_CPU_CACHED		(2U << 1)
#define NVGPU_GPU_ALLOC_VIDMEM_FLAG_CPU_MASK		(7U << 1)

#define NVGPU_GPU_ALLOC_VIDMEM_FLAG_VPR			(1U << 4)

/* Allocation of device-specific local video memory. Returns dmabuf fd
 * on success. */
struct nvgpu_gpu_alloc_vidmem_args {
	union {
		struct {
			/* Size for allocation. Must be a multiple of
			 * small page size. */
			__u64 size;

			/* NVGPU_GPU_ALLOC_VIDMEM_FLAG_* */
			__u32 flags;

			/* Informational mem tag for resource usage
			 * tracking. */
			__u16 memtag;

			__u16 reserved0;

			/* GPU-visible physical memory alignment in
			 * bytes.
			 *
			 * Alignment must be a power of two. Minimum
			 * alignment is the small page size, which 0
			 * also denotes.
			 *
			 * For contiguous and non-contiguous
			 * allocations, the start address of the
			 * physical memory allocation will be aligned
			 * by this value.
			 *
			 * For non-contiguous allocations, memory is
			 * internally allocated in round_up(size /
			 * alignment) contiguous blocks. The start
			 * address of each block is aligned by the
			 * alignment value. If the size is not a
			 * multiple of alignment (which is ok), the
			 * last allocation block size is (size %
			 * alignment).
			 *
			 * By specifying the big page size here and
			 * allocation size that is a multiple of big
			 * pages, it will be guaranteed that the
			 * allocated buffer is big page size mappable.
			 */
			__u32 alignment;

			__u32 reserved1[3];
		} in;

		struct {
			__s32 dmabuf_fd;
		} out;
	};
};

/* Memory clock */
#define NVGPU_GPU_CLK_DOMAIN_MCLK                                (0)
/* Main graphics core clock */
#define NVGPU_GPU_CLK_DOMAIN_GPCCLK	                         (1)

struct nvgpu_gpu_clk_range {

	/* Flags (not currently used) */
	__u32 flags;

	/* NVGPU_GPU_CLK_DOMAIN_* */
	__u32 clk_domain;
	__u64 min_hz;
	__u64 max_hz;
};

/* Request on specific clock domains */
#define NVGPU_GPU_CLK_FLAG_SPECIFIC_DOMAINS		(1UL << 0)

struct nvgpu_gpu_clk_range_args {

	/* Flags. If NVGPU_GPU_CLK_FLAG_SPECIFIC_DOMAINS the request will
	   apply only to domains specified in clock entries. In this case
	   caller must set clock domain in each entry. Otherwise, the
	   ioctl will return all clock domains.
	*/
	__u32 flags;

	__u16 pad0;

	/* in/out: Number of entries in clk_range_entries buffer. If zero,
	   NVGPU_GPU_IOCTL_CLK_GET_RANGE will return 0 and
	   num_entries will be set to number of clock domains.
	 */
	__u16 num_entries;

	/* in: Pointer to clock range entries in the caller's address space.
	   size must be >= max_entries * sizeof(struct nvgpu_gpu_clk_range)
	 */
	__u64 clk_range_entries;
};

struct nvgpu_gpu_clk_vf_point {
	__u64 freq_hz;
};

struct nvgpu_gpu_clk_vf_points_args {

	/* in: Flags (not currently used) */
	__u32 flags;

	/* in: NVGPU_GPU_CLK_DOMAIN_* */
	__u32 clk_domain;

	/* in/out: max number of nvgpu_gpu_clk_vf_point entries in
	   clk_vf_point_entries.  If max_entries is zero,
	   NVGPU_GPU_IOCTL_CLK_GET_VF_POINTS will return 0 and max_entries will
	   be set to the max number of VF entries for this clock domain. If
	   there are more entries than max_entries, then ioctl will return
	   -EINVAL.
	*/
	__u16 max_entries;

	/* out: Number of nvgpu_gpu_clk_vf_point entries returned in
	   clk_vf_point_entries. Number of entries might vary depending on
	   thermal conditions.
	*/
	__u16 num_entries;

	__u32 reserved;

	/* in: Pointer to clock VF point entries in the caller's address space.
	   size must be >= max_entries * sizeof(struct nvgpu_gpu_clk_vf_point).
	 */
	__u64 clk_vf_point_entries;
};

/* Target clock requested by application*/
#define NVGPU_GPU_CLK_TYPE_TARGET	1
/* Actual clock frequency for the domain.
   May deviate from desired target frequency due to PLL constraints. */
#define NVGPU_GPU_CLK_TYPE_ACTUAL	2
/* Effective clock, measured from hardware */
#define NVGPU_GPU_CLK_TYPE_EFFECTIVE	3

struct nvgpu_gpu_clk_info {

	/* Flags (not currently used) */
	__u16 flags;

	/* in: When NVGPU_GPU_CLK_FLAG_SPECIFIC_DOMAINS set, indicates
	   the type of clock info to be returned for this entry. It is
	   allowed to have several entries with different clock types in
	   the same request (for instance query both target and actual
	   clocks for a given clock domain). This field is ignored for a
	   SET operation. */
	__u16 clk_type;

	/* NVGPU_GPU_CLK_DOMAIN_xxx */
	__u32 clk_domain;

	__u64 freq_hz;
};

struct nvgpu_gpu_clk_get_info_args {

	/* Flags. If NVGPU_GPU_CLK_FLAG_SPECIFIC_DOMAINS the request will
	   apply only to domains specified in clock entries. In this case
	   caller must set clock domain in each entry. Otherwise, the
	   ioctl will return all clock domains.
	*/
	__u32 flags;

	/* in: indicates which type of clock info to be returned (see
	   NVGPU_GPU_CLK_TYPE_xxx). If NVGPU_GPU_CLK_FLAG_SPECIFIC_DOMAINS
	   is defined, clk_type is specified in each clock info entry instead.
	 */
	__u16 clk_type;

	/* in/out: Number of clock info entries contained in clk_info_entries.
	   If zero, NVGPU_GPU_IOCTL_CLK_GET_INFO will return 0 and
	   num_entries will be set to number of clock domains. Also,
	   last_req_nr will be updated, which allows checking if a given
	   request has completed. If there are more entries than max_entries,
	   then ioctl will return -EINVAL.
	 */
	__u16 num_entries;

	/* in: Pointer to nvgpu_gpu_clk_info entries in the caller's address
	   space. Buffer size must be at least:
		num_entries * sizeof(struct nvgpu_gpu_clk_info)
	   If NVGPU_GPU_CLK_FLAG_SPECIFIC_DOMAINS is set, caller should set
	   clk_domain to be queried in  each entry. With this flag,
	   clk_info_entries passed to an NVGPU_GPU_IOCTL_CLK_SET_INFO,
	   can be re-used on completion for a NVGPU_GPU_IOCTL_CLK_GET_INFO.
	   This allows checking actual_mhz.
	 */
	__u64 clk_info_entries;

};

struct nvgpu_gpu_clk_set_info_args {

	/* in: Flags (not currently used). */
	__u32 flags;

	__u16 pad0;

	/* Number of clock info entries contained in clk_info_entries.
	   Must be > 0.
	 */
	__u16 num_entries;

	/* Pointer to clock info entries in the caller's address space. Buffer
	   size must be at least
		num_entries * sizeof(struct nvgpu_gpu_clk_info)
	 */
	__u64 clk_info_entries;

	/* out: File descriptor for request completion. Application can poll
	   this file descriptor to determine when the request has completed.
	   The fd must be closed afterwards.
	 */
	__s32 completion_fd;
};

struct nvgpu_gpu_get_event_fd_args {

	/* in: Flags (not currently used). */
	__u32 flags;

	/* out: File descriptor for events, e.g. clock update.
	 * On successful polling of this event_fd, application is
	 * expected to read status (nvgpu_gpu_event_info),
	 * which provides detailed event information
	 * For a poll operation, alarms will be reported with POLLPRI,
	 * and GPU shutdown will be reported with POLLHUP.
	 */
	__s32 event_fd;
};

struct nvgpu_gpu_get_memory_state_args {
	/*
	 * Current free space for this device; may change even when any
	 * kernel-managed metadata (e.g., page tables or channels) is allocated
	 * or freed. For an idle gpu, an allocation of this size would succeed.
	 */
	__u64 total_free_bytes;

	/* For future use; must be set to 0. */
	__u64 reserved[4];
};

struct nvgpu_gpu_get_fbp_l2_masks_args {
	/* [in]  L2 mask buffer size reserved by userspace. Should be
		 at least sizeof(__u32) * fls(fbp_en_mask) to receive LTC
		 mask for each FBP.
	   [out] full kernel buffer size
	*/
	__u32 mask_buf_size;
	__u32 reserved;

	/* [in]  pointer to L2 mask buffer. It will receive one
		 32-bit L2 mask per FBP or 0 if FBP is not enabled or
		 not present. This parameter is ignored if
		 mask_buf_size is 0. */
	__u64 mask_buf_addr;
};

#define NVGPU_GPU_VOLTAGE_CORE		1
#define NVGPU_GPU_VOLTAGE_SRAM		2
#define NVGPU_GPU_VOLTAGE_BUS		3	/* input to regulator */

struct nvgpu_gpu_get_voltage_args {
	__u64 reserved;
	__u32 which;		/* in: NVGPU_GPU_VOLTAGE_* */
	__u32 voltage;		/* uV */
};

struct nvgpu_gpu_get_current_args {
	__u32 reserved[3];
	__u32 currnt;		/* mA */
};

struct nvgpu_gpu_get_power_args {
	__u32 reserved[3];
	__u32 power;		/* mW */
};

struct nvgpu_gpu_get_temperature_args {
	__u32 reserved[3];
	/* Temperature in signed fixed point format SFXP24.8
	 *    Celsius = temp_f24_8 / 256.
	 */
	__s32 temp_f24_8;
};

struct nvgpu_gpu_set_therm_alert_limit_args {
	__u32 reserved[3];
	/* Temperature in signed fixed point format SFXP24.8
	 *    Celsius = temp_f24_8 / 256.
	 */
	__s32 temp_f24_8;
};

/*
 * Adjust options of deterministic channels in channel batches.
 *
 * This supports only one option currently: relax railgate blocking by
 * "disabling" the channel.
 *
 * Open deterministic channels do not allow the GPU to railgate by default. It
 * may be preferable to hold preopened channel contexts open and idle and still
 * railgate the GPU, taking the channels back into use dynamically in userspace
 * as an optimization. This ioctl allows to drop or reacquire the requirement
 * to hold GPU power on for individual channels. If allow_railgate is set on a
 * channel, no work can be submitted to it.
 *
 * num_channels is updated to signify how many channels were updated
 * successfully. It can be used to test which was the first update to fail.
 */
struct nvgpu_gpu_set_deterministic_opts_args {
	__u32 num_channels; /* in/out */
/*
 * Set or unset the railgating reference held by deterministic channels. If
 * the channel status is already the same as the flag, this is a no-op. Both
 * of these flags cannot be set at the same time. If none are set, the state
 * is left as is.
 */
#define NVGPU_GPU_SET_DETERMINISTIC_OPTS_FLAGS_ALLOW_RAILGATING    (1 << 0)
#define NVGPU_GPU_SET_DETERMINISTIC_OPTS_FLAGS_DISALLOW_RAILGATING (1 << 1)
	__u32 flags;        /* in */
	/*
	 * This is a pointer to an array of size num_channels.
	 *
	 * The channels have to be valid fds and be previously set as
	 * deterministic.
	 */
	__u64 channels; /* in */
};

#define NVGPU_GPU_IOCTL_ZCULL_GET_CTX_SIZE \
	_IOR(NVGPU_GPU_IOCTL_MAGIC, 1, struct nvgpu_gpu_zcull_get_ctx_size_args)
#define NVGPU_GPU_IOCTL_ZCULL_GET_INFO \
	_IOR(NVGPU_GPU_IOCTL_MAGIC, 2, struct nvgpu_gpu_zcull_get_info_args)
#define NVGPU_GPU_IOCTL_ZBC_SET_TABLE	\
	_IOW(NVGPU_GPU_IOCTL_MAGIC, 3, struct nvgpu_gpu_zbc_set_table_args)
#define NVGPU_GPU_IOCTL_ZBC_QUERY_TABLE	\
	_IOWR(NVGPU_GPU_IOCTL_MAGIC, 4, struct nvgpu_gpu_zbc_query_table_args)
#define NVGPU_GPU_IOCTL_GET_CHARACTERISTICS   \
	_IOWR(NVGPU_GPU_IOCTL_MAGIC, 5, struct nvgpu_gpu_get_characteristics)
#define NVGPU_GPU_IOCTL_PREPARE_COMPRESSIBLE_READ \
	_IOWR(NVGPU_GPU_IOCTL_MAGIC, 6, struct nvgpu_gpu_prepare_compressible_read_args)
#define NVGPU_GPU_IOCTL_MARK_COMPRESSIBLE_WRITE \
	_IOWR(NVGPU_GPU_IOCTL_MAGIC, 7, struct nvgpu_gpu_mark_compressible_write_args)
#define NVGPU_GPU_IOCTL_ALLOC_AS \
	_IOWR(NVGPU_GPU_IOCTL_MAGIC, 8, struct nvgpu_alloc_as_args)
#define NVGPU_GPU_IOCTL_OPEN_TSG \
	_IOWR(NVGPU_GPU_IOCTL_MAGIC, 9, struct nvgpu_gpu_open_tsg_args)
#define NVGPU_GPU_IOCTL_GET_TPC_MASKS \
	_IOWR(NVGPU_GPU_IOCTL_MAGIC, 10, struct nvgpu_gpu_get_tpc_masks_args)
#define NVGPU_GPU_IOCTL_OPEN_CHANNEL \
	_IOWR(NVGPU_GPU_IOCTL_MAGIC, 11, struct nvgpu_gpu_open_channel_args)
#define NVGPU_GPU_IOCTL_FLUSH_L2 \
	_IOWR(NVGPU_GPU_IOCTL_MAGIC, 12, struct nvgpu_gpu_l2_fb_args)
#define NVGPU_GPU_IOCTL_INVAL_ICACHE \
	_IOWR(NVGPU_GPU_IOCTL_MAGIC, 13, struct nvgpu_gpu_inval_icache_args)
#define NVGPU_GPU_IOCTL_SET_MMUDEBUG_MODE \
	_IOWR(NVGPU_GPU_IOCTL_MAGIC, 14, struct nvgpu_gpu_mmu_debug_mode_args)
#define NVGPU_GPU_IOCTL_SET_SM_DEBUG_MODE \
	_IOWR(NVGPU_GPU_IOCTL_MAGIC, 15, struct nvgpu_gpu_sm_debug_mode_args)
#define NVGPU_GPU_IOCTL_WAIT_FOR_PAUSE \
	_IOWR(NVGPU_GPU_IOCTL_MAGIC, 16, struct nvgpu_gpu_wait_pause_args)
#define NVGPU_GPU_IOCTL_GET_TPC_EXCEPTION_EN_STATUS \
	_IOWR(NVGPU_GPU_IOCTL_MAGIC, 17, struct nvgpu_gpu_tpc_exception_en_status_args)
#define NVGPU_GPU_IOCTL_NUM_VSMS \
	_IOWR(NVGPU_GPU_IOCTL_MAGIC, 18, struct nvgpu_gpu_num_vsms)
#define NVGPU_GPU_IOCTL_VSMS_MAPPING \
	_IOWR(NVGPU_GPU_IOCTL_MAGIC, 19, struct nvgpu_gpu_vsms_mapping)
#define NVGPU_GPU_IOCTL_RESUME_FROM_PAUSE \
	_IO(NVGPU_GPU_IOCTL_MAGIC, 21)
#define NVGPU_GPU_IOCTL_TRIGGER_SUSPEND \
	_IO(NVGPU_GPU_IOCTL_MAGIC, 22)
#define NVGPU_GPU_IOCTL_CLEAR_SM_ERRORS \
	_IO(NVGPU_GPU_IOCTL_MAGIC, 23)
#define NVGPU_GPU_IOCTL_GET_CPU_TIME_CORRELATION_INFO \
	_IOWR(NVGPU_GPU_IOCTL_MAGIC, 24, \
			struct nvgpu_gpu_get_cpu_time_correlation_info_args)
#define NVGPU_GPU_IOCTL_GET_GPU_TIME \
	_IOWR(NVGPU_GPU_IOCTL_MAGIC, 25, \
			struct nvgpu_gpu_get_gpu_time_args)
#define NVGPU_GPU_IOCTL_GET_ENGINE_INFO \
	_IOWR(NVGPU_GPU_IOCTL_MAGIC, 26, \
			struct nvgpu_gpu_get_engine_info_args)
#define NVGPU_GPU_IOCTL_ALLOC_VIDMEM \
	_IOWR(NVGPU_GPU_IOCTL_MAGIC, 27, \
			struct nvgpu_gpu_alloc_vidmem_args)
#define NVGPU_GPU_IOCTL_CLK_GET_RANGE \
	_IOWR(NVGPU_GPU_IOCTL_MAGIC, 28, struct nvgpu_gpu_clk_range_args)
#define NVGPU_GPU_IOCTL_CLK_GET_VF_POINTS \
	_IOWR(NVGPU_GPU_IOCTL_MAGIC, 29, struct nvgpu_gpu_clk_vf_points_args)
#define NVGPU_GPU_IOCTL_CLK_GET_INFO \
	_IOWR(NVGPU_GPU_IOCTL_MAGIC, 30, struct nvgpu_gpu_clk_get_info_args)
#define NVGPU_GPU_IOCTL_CLK_SET_INFO \
	_IOWR(NVGPU_GPU_IOCTL_MAGIC, 31, struct nvgpu_gpu_clk_set_info_args)
#define NVGPU_GPU_IOCTL_GET_EVENT_FD \
	_IOWR(NVGPU_GPU_IOCTL_MAGIC, 32, struct nvgpu_gpu_get_event_fd_args)
#define NVGPU_GPU_IOCTL_GET_MEMORY_STATE \
	_IOWR(NVGPU_GPU_IOCTL_MAGIC, 33, \
			struct nvgpu_gpu_get_memory_state_args)
#define NVGPU_GPU_IOCTL_GET_VOLTAGE \
	_IOWR(NVGPU_GPU_IOCTL_MAGIC, 34, struct nvgpu_gpu_get_voltage_args)
#define NVGPU_GPU_IOCTL_GET_CURRENT \
	_IOWR(NVGPU_GPU_IOCTL_MAGIC, 35, struct nvgpu_gpu_get_current_args)
#define NVGPU_GPU_IOCTL_GET_POWER \
	_IOWR(NVGPU_GPU_IOCTL_MAGIC, 36, struct nvgpu_gpu_get_power_args)
#define NVGPU_GPU_IOCTL_GET_TEMPERATURE \
	_IOWR(NVGPU_GPU_IOCTL_MAGIC, 37, struct nvgpu_gpu_get_temperature_args)
#define NVGPU_GPU_IOCTL_GET_FBP_L2_MASKS \
	_IOWR(NVGPU_GPU_IOCTL_MAGIC, 38, struct nvgpu_gpu_get_fbp_l2_masks_args)
#define NVGPU_GPU_IOCTL_SET_THERM_ALERT_LIMIT \
		_IOWR(NVGPU_GPU_IOCTL_MAGIC, 39, \
			struct nvgpu_gpu_set_therm_alert_limit_args)
#define NVGPU_GPU_IOCTL_SET_DETERMINISTIC_OPTS \
	_IOWR(NVGPU_GPU_IOCTL_MAGIC, 40, \
			struct nvgpu_gpu_set_deterministic_opts_args)
#define NVGPU_GPU_IOCTL_LAST		\
	_IOC_NR(NVGPU_GPU_IOCTL_SET_DETERMINISTIC_OPTS)
#define NVGPU_GPU_IOCTL_MAX_ARG_SIZE	\
	sizeof(struct nvgpu_gpu_get_cpu_time_correlation_info_args)

/*
 * Event session
 *
 * NVGPU_GPU_IOCTL_GET_EVENT_FD opens an event session.
 * Below ioctls can be used on these sessions fds.
 */
#define NVGPU_EVENT_IOCTL_MAGIC	'E'

/* Normal events (POLLIN) */
/* Event associated to a VF update */
#define NVGPU_GPU_EVENT_VF_UPDATE				0

/* Recoverable alarms (POLLPRI) */
/* Alarm when target frequency on any session is not possible */
#define NVGPU_GPU_EVENT_ALARM_TARGET_VF_NOT_POSSIBLE		1
/* Alarm when target frequency on current session is not possible */
#define NVGPU_GPU_EVENT_ALARM_LOCAL_TARGET_VF_NOT_POSSIBLE	2
/* Alarm when Clock Arbiter failed */
#define NVGPU_GPU_EVENT_ALARM_CLOCK_ARBITER_FAILED		3
/* Alarm when VF table update failed */
#define NVGPU_GPU_EVENT_ALARM_VF_TABLE_UPDATE_FAILED		4
/* Alarm on thermal condition */
#define NVGPU_GPU_EVENT_ALARM_THERMAL_ABOVE_THRESHOLD		5
/* Alarm on power condition */
#define NVGPU_GPU_EVENT_ALARM_POWER_ABOVE_THRESHOLD		6

/* Non recoverable alarm (POLLHUP) */
/* Alarm on GPU shutdown/fall from bus */
#define NVGPU_GPU_EVENT_ALARM_GPU_LOST				7

#define NVGPU_GPU_EVENT_LAST	NVGPU_GPU_EVENT_ALARM_GPU_LOST

struct nvgpu_gpu_event_info {
	__u32 event_id;		/* NVGPU_GPU_EVENT_* */
	__u32 reserved;
	__u64 timestamp;	/* CPU timestamp (in nanoseconds) */
};

struct nvgpu_gpu_set_event_filter_args {

	/* in: Flags (not currently used). */
	__u32 flags;

	/* in: Size of event filter in 32-bit words */
	__u32 size;

	/* in: Address of buffer containing bit mask of events.
	 * Bit #n is set if event #n should be monitored.
	 */
	__u64 buffer;
};

#define NVGPU_EVENT_IOCTL_SET_FILTER \
	_IOW(NVGPU_EVENT_IOCTL_MAGIC, 1, struct nvgpu_gpu_set_event_filter_args)
#define NVGPU_EVENT_IOCTL_LAST		\
	_IOC_NR(NVGPU_EVENT_IOCTL_SET_FILTER)
#define NVGPU_EVENT_IOCTL_MAX_ARG_SIZE	\
	sizeof(struct nvgpu_gpu_set_event_filter_args)

/*
 * /dev/nvhost-tsg-gpu device
 *
 * Opening a '/dev/nvhost-tsg-gpu' device node creates a way to
 * bind/unbind a channel to/from TSG group
 */

#define NVGPU_TSG_IOCTL_MAGIC 'T'

struct nvgpu_tsg_bind_channel_ex_args {
	/* in: channel fd */
	__s32 channel_fd;

	/* in: VEID in Volta */
	__u32 subcontext_id;
	__u32 num_active_tpcs;
	__u8 tpc_pg_enabled;
	__u8 reserved[11];
};

#define NVGPU_TSG_IOCTL_BIND_CHANNEL \
	_IOW(NVGPU_TSG_IOCTL_MAGIC, 1, int)
#define NVGPU_TSG_IOCTL_UNBIND_CHANNEL \
	_IOW(NVGPU_TSG_IOCTL_MAGIC, 2, int)
#define NVGPU_IOCTL_TSG_ENABLE \
	_IO(NVGPU_TSG_IOCTL_MAGIC, 3)
#define NVGPU_IOCTL_TSG_DISABLE \
	_IO(NVGPU_TSG_IOCTL_MAGIC, 4)
#define NVGPU_IOCTL_TSG_PREEMPT \
	_IO(NVGPU_TSG_IOCTL_MAGIC, 5)
#define NVGPU_IOCTL_TSG_EVENT_ID_CTRL \
	_IOWR(NVGPU_TSG_IOCTL_MAGIC, 7, struct nvgpu_event_id_ctrl_args)
#define NVGPU_IOCTL_TSG_SET_RUNLIST_INTERLEAVE \
	_IOW(NVGPU_TSG_IOCTL_MAGIC, 8, struct nvgpu_runlist_interleave_args)
#define NVGPU_IOCTL_TSG_SET_TIMESLICE \
	_IOW(NVGPU_TSG_IOCTL_MAGIC, 9, struct nvgpu_timeslice_args)
#define NVGPU_IOCTL_TSG_GET_TIMESLICE \
	_IOR(NVGPU_TSG_IOCTL_MAGIC, 10, struct nvgpu_timeslice_args)
#define NVGPU_TSG_IOCTL_BIND_CHANNEL_EX \
	_IOWR(NVGPU_TSG_IOCTL_MAGIC, 11, struct nvgpu_tsg_bind_channel_ex_args)
#define NVGPU_TSG_IOCTL_MAX_ARG_SIZE	\
		sizeof(struct nvgpu_tsg_bind_channel_ex_args)
#define NVGPU_TSG_IOCTL_LAST		\
	_IOC_NR(NVGPU_TSG_IOCTL_BIND_CHANNEL_EX)

/*
 * /dev/nvhost-dbg-gpu device
 *
 * Opening a '/dev/nvhost-dbg-gpu' device node creates a new debugger
 * session.  nvgpu channels (for the same module) can then be bound to such a
 * session.
 *
 * One nvgpu channel can also be bound to multiple debug sessions
 *
 * As long as there is an open device file to the session, or any bound
 * nvgpu channels it will be valid.  Once all references to the session
 * are removed the session is deleted.
 *
 */

#define NVGPU_DBG_GPU_IOCTL_MAGIC 'D'

/*
 * Binding/attaching a debugger session to an nvgpu channel
 *
 * The 'channel_fd' given here is the fd used to allocate the
 * gpu channel context.
 *
 */
struct nvgpu_dbg_gpu_bind_channel_args {
	__u32 channel_fd; /* in */
	__u32 _pad0[1];
};

#define NVGPU_DBG_GPU_IOCTL_BIND_CHANNEL				\
	_IOWR(NVGPU_DBG_GPU_IOCTL_MAGIC, 1, struct nvgpu_dbg_gpu_bind_channel_args)

/*
 * Register operations
 * All operations are targeted towards first channel
 * attached to debug session
 */
/* valid op values */
#define NVGPU_DBG_GPU_REG_OP_READ_32                             (0x00000000)
#define NVGPU_DBG_GPU_REG_OP_WRITE_32                            (0x00000001)
#define NVGPU_DBG_GPU_REG_OP_READ_64                             (0x00000002)
#define NVGPU_DBG_GPU_REG_OP_WRITE_64                            (0x00000003)
/* note: 8b ops are unsupported */
#define NVGPU_DBG_GPU_REG_OP_READ_08                             (0x00000004)
#define NVGPU_DBG_GPU_REG_OP_WRITE_08                            (0x00000005)

/* valid type values */
#define NVGPU_DBG_GPU_REG_OP_TYPE_GLOBAL                         (0x00000000)
#define NVGPU_DBG_GPU_REG_OP_TYPE_GR_CTX                         (0x00000001)
#define NVGPU_DBG_GPU_REG_OP_TYPE_GR_CTX_TPC                     (0x00000002)
#define NVGPU_DBG_GPU_REG_OP_TYPE_GR_CTX_SM                      (0x00000004)
#define NVGPU_DBG_GPU_REG_OP_TYPE_GR_CTX_CROP                    (0x00000008)
#define NVGPU_DBG_GPU_REG_OP_TYPE_GR_CTX_ZROP                    (0x00000010)
/*#define NVGPU_DBG_GPU_REG_OP_TYPE_FB                           (0x00000020)*/
#define NVGPU_DBG_GPU_REG_OP_TYPE_GR_CTX_QUAD                    (0x00000040)

/* valid status values */
#define NVGPU_DBG_GPU_REG_OP_STATUS_SUCCESS                      (0x00000000)
#define NVGPU_DBG_GPU_REG_OP_STATUS_INVALID_OP                   (0x00000001)
#define NVGPU_DBG_GPU_REG_OP_STATUS_INVALID_TYPE                 (0x00000002)
#define NVGPU_DBG_GPU_REG_OP_STATUS_INVALID_OFFSET               (0x00000004)
#define NVGPU_DBG_GPU_REG_OP_STATUS_UNSUPPORTED_OP               (0x00000008)
#define NVGPU_DBG_GPU_REG_OP_STATUS_INVALID_MASK                 (0x00000010)

struct nvgpu_dbg_gpu_reg_op {
	__u8    op;
	__u8    type;
	__u8    status;
	__u8    quad;
	__u32   group_mask;
	__u32   sub_group_mask;
	__u32   offset;
	__u32   value_lo;
	__u32   value_hi;
	__u32   and_n_mask_lo;
	__u32   and_n_mask_hi;
};

struct nvgpu_dbg_gpu_exec_reg_ops_args {
	__u64 ops; /* pointer to nvgpu_reg_op operations */
	__u32 num_ops;
	__u32 _pad0[1];
};

#define NVGPU_DBG_GPU_IOCTL_REG_OPS					\
	_IOWR(NVGPU_DBG_GPU_IOCTL_MAGIC, 2, struct nvgpu_dbg_gpu_exec_reg_ops_args)

/* Enable/disable/clear event notifications */
struct nvgpu_dbg_gpu_events_ctrl_args {
	__u32 cmd; /* in */
	__u32 _pad0[1];
};

/* valid event ctrl values */
#define NVGPU_DBG_GPU_EVENTS_CTRL_CMD_DISABLE                    (0x00000000)
#define NVGPU_DBG_GPU_EVENTS_CTRL_CMD_ENABLE                     (0x00000001)
#define NVGPU_DBG_GPU_EVENTS_CTRL_CMD_CLEAR                      (0x00000002)

#define NVGPU_DBG_GPU_IOCTL_EVENTS_CTRL				\
	_IOWR(NVGPU_DBG_GPU_IOCTL_MAGIC, 3, struct nvgpu_dbg_gpu_events_ctrl_args)


/* Powergate/Unpowergate control */

#define NVGPU_DBG_GPU_POWERGATE_MODE_ENABLE                                 1
#define NVGPU_DBG_GPU_POWERGATE_MODE_DISABLE                                2

struct nvgpu_dbg_gpu_powergate_args {
	__u32 mode;
} __packed;

#define NVGPU_DBG_GPU_IOCTL_POWERGATE					\
	_IOWR(NVGPU_DBG_GPU_IOCTL_MAGIC, 4, struct nvgpu_dbg_gpu_powergate_args)


/* SMPC Context Switch Mode */
#define NVGPU_DBG_GPU_SMPC_CTXSW_MODE_NO_CTXSW               (0x00000000)
#define NVGPU_DBG_GPU_SMPC_CTXSW_MODE_CTXSW                  (0x00000001)

struct nvgpu_dbg_gpu_smpc_ctxsw_mode_args {
	__u32 mode;
} __packed;

#define NVGPU_DBG_GPU_IOCTL_SMPC_CTXSW_MODE \
	_IOWR(NVGPU_DBG_GPU_IOCTL_MAGIC, 5, struct nvgpu_dbg_gpu_smpc_ctxsw_mode_args)

/* Suspend /Resume SM control */
#define NVGPU_DBG_GPU_SUSPEND_ALL_SMS 1
#define NVGPU_DBG_GPU_RESUME_ALL_SMS  2

struct nvgpu_dbg_gpu_suspend_resume_all_sms_args {
	__u32 mode;
} __packed;

#define NVGPU_DBG_GPU_IOCTL_SUSPEND_RESUME_ALL_SMS			\
	_IOWR(NVGPU_DBG_GPU_IOCTL_MAGIC, 6, struct nvgpu_dbg_gpu_suspend_resume_all_sms_args)

struct nvgpu_dbg_gpu_perfbuf_map_args {
	__u32 dmabuf_fd;	/* in */
	__u32 reserved;
	__u64 mapping_size;	/* in, size of mapped buffer region */
	__u64 offset;		/* out, virtual address of the mapping */
};

struct nvgpu_dbg_gpu_perfbuf_unmap_args {
	__u64 offset;
};

#define NVGPU_DBG_GPU_IOCTL_PERFBUF_MAP \
	_IOWR(NVGPU_DBG_GPU_IOCTL_MAGIC, 7, struct nvgpu_dbg_gpu_perfbuf_map_args)
#define NVGPU_DBG_GPU_IOCTL_PERFBUF_UNMAP \
	_IOWR(NVGPU_DBG_GPU_IOCTL_MAGIC, 8, struct nvgpu_dbg_gpu_perfbuf_unmap_args)

/* Enable/disable PC Sampling */
struct nvgpu_dbg_gpu_pc_sampling_args {
	__u32 enable;
	__u32 _pad0[1];
};

#define NVGPU_DBG_GPU_IOCTL_PC_SAMPLING_DISABLE	0
#define NVGPU_DBG_GPU_IOCTL_PC_SAMPLING_ENABLE	1

#define NVGPU_DBG_GPU_IOCTL_PC_SAMPLING \
	_IOW(NVGPU_DBG_GPU_IOCTL_MAGIC,  9, struct nvgpu_dbg_gpu_pc_sampling_args)

/* Enable/Disable timeouts */
#define NVGPU_DBG_GPU_IOCTL_TIMEOUT_ENABLE                                   1
#define NVGPU_DBG_GPU_IOCTL_TIMEOUT_DISABLE                                  0

struct nvgpu_dbg_gpu_timeout_args {
	__u32 enable;
	__u32 padding;
};

#define NVGPU_DBG_GPU_IOCTL_TIMEOUT \
	_IOW(NVGPU_DBG_GPU_IOCTL_MAGIC, 10, struct nvgpu_dbg_gpu_timeout_args)

#define NVGPU_DBG_GPU_IOCTL_GET_TIMEOUT \
	_IOR(NVGPU_DBG_GPU_IOCTL_MAGIC, 11, struct nvgpu_dbg_gpu_timeout_args)


struct nvgpu_dbg_gpu_set_next_stop_trigger_type_args {
	__u32 broadcast;
	__u32 reserved;
};

#define NVGPU_DBG_GPU_IOCTL_SET_NEXT_STOP_TRIGGER_TYPE			\
	_IOWR(NVGPU_DBG_GPU_IOCTL_MAGIC, 12, struct nvgpu_dbg_gpu_set_next_stop_trigger_type_args)


/* PM Context Switch Mode */
#define NVGPU_DBG_GPU_HWPM_CTXSW_MODE_NO_CTXSW               (0x00000000)
#define NVGPU_DBG_GPU_HWPM_CTXSW_MODE_CTXSW                  (0x00000001)

struct nvgpu_dbg_gpu_hwpm_ctxsw_mode_args {
	__u32 mode;
	__u32 reserved;
};

#define NVGPU_DBG_GPU_IOCTL_HWPM_CTXSW_MODE \
	_IOWR(NVGPU_DBG_GPU_IOCTL_MAGIC, 13, struct nvgpu_dbg_gpu_hwpm_ctxsw_mode_args)


struct nvgpu_dbg_gpu_sm_error_state_record {
	__u32 hww_global_esr;
	__u32 hww_warp_esr;
	__u64 hww_warp_esr_pc;
	__u32 hww_global_esr_report_mask;
	__u32 hww_warp_esr_report_mask;

	/*
	 * Notes
	 * - This struct can be safely appended with new fields. However, always
	 *   keep the structure size multiple of 8 and make sure that the binary
	 *   layout does not change between 32-bit and 64-bit architectures.
	 */
};

struct nvgpu_dbg_gpu_read_single_sm_error_state_args {
	__u32 sm_id;
	__u32 padding;
	__u64 sm_error_state_record_mem;
	__u64 sm_error_state_record_size;
};

#define NVGPU_DBG_GPU_IOCTL_READ_SINGLE_SM_ERROR_STATE			\
	_IOWR(NVGPU_DBG_GPU_IOCTL_MAGIC, 14, struct nvgpu_dbg_gpu_read_single_sm_error_state_args)


struct nvgpu_dbg_gpu_clear_single_sm_error_state_args {
	__u32 sm_id;
	__u32 padding;
};

#define NVGPU_DBG_GPU_IOCTL_CLEAR_SINGLE_SM_ERROR_STATE			\
	_IOW(NVGPU_DBG_GPU_IOCTL_MAGIC, 15, struct nvgpu_dbg_gpu_clear_single_sm_error_state_args)


struct nvgpu_dbg_gpu_write_single_sm_error_state_args {
	__u32 sm_id;
	__u32 padding;
	__u64 sm_error_state_record_mem;
	__u64 sm_error_state_record_size;
};

#define NVGPU_DBG_GPU_IOCTL_WRITE_SINGLE_SM_ERROR_STATE			\
	_IOW(NVGPU_DBG_GPU_IOCTL_MAGIC, 16, struct nvgpu_dbg_gpu_write_single_sm_error_state_args)

/*
 * Unbinding/detaching a debugger session from a nvgpu channel
 *
 * The 'channel_fd' given here is the fd used to allocate the
 * gpu channel context.
 */
struct nvgpu_dbg_gpu_unbind_channel_args {
	__u32 channel_fd; /* in */
	__u32 _pad0[1];
};

#define NVGPU_DBG_GPU_IOCTL_UNBIND_CHANNEL				\
	_IOW(NVGPU_DBG_GPU_IOCTL_MAGIC, 17, struct nvgpu_dbg_gpu_unbind_channel_args)


#define NVGPU_DBG_GPU_SUSPEND_ALL_CONTEXTS	1
#define NVGPU_DBG_GPU_RESUME_ALL_CONTEXTS	2

struct nvgpu_dbg_gpu_suspend_resume_contexts_args {
	__u32 action;
	__u32 is_resident_context;
	__s32 resident_context_fd;
	__u32 padding;
};

#define NVGPU_DBG_GPU_IOCTL_SUSPEND_RESUME_CONTEXTS			\
	_IOWR(NVGPU_DBG_GPU_IOCTL_MAGIC, 18, struct nvgpu_dbg_gpu_suspend_resume_contexts_args)


#define NVGPU_DBG_GPU_IOCTL_ACCESS_FB_MEMORY_CMD_READ	1
#define NVGPU_DBG_GPU_IOCTL_ACCESS_FB_MEMORY_CMD_WRITE	2

struct nvgpu_dbg_gpu_access_fb_memory_args {
	__u32 cmd;       /* in: either read or write */

	__s32 dmabuf_fd; /* in: dmabuf fd of the buffer in FB */
	__u64 offset;    /* in: offset within buffer in FB, should be 4B aligned */

	__u64 buffer;    /* in/out: temp buffer to read/write from */
	__u64 size;      /* in: size of the buffer i.e. size of read/write in bytes, should be 4B aligned */
};

#define NVGPU_DBG_GPU_IOCTL_ACCESS_FB_MEMORY	\
	_IOWR(NVGPU_DBG_GPU_IOCTL_MAGIC, 19, struct nvgpu_dbg_gpu_access_fb_memory_args)

struct nvgpu_dbg_gpu_profiler_obj_mgt_args {
	__u32 profiler_handle;
	__u32 reserved;
};

#define NVGPU_DBG_GPU_IOCTL_PROFILER_ALLOCATE	\
	_IOWR(NVGPU_DBG_GPU_IOCTL_MAGIC, 20, struct nvgpu_dbg_gpu_profiler_obj_mgt_args)

#define NVGPU_DBG_GPU_IOCTL_PROFILER_FREE	\
	_IOWR(NVGPU_DBG_GPU_IOCTL_MAGIC, 21, struct nvgpu_dbg_gpu_profiler_obj_mgt_args)

struct nvgpu_dbg_gpu_profiler_reserve_args {
	__u32 profiler_handle;
	__u32 acquire;
};

#define NVGPU_DBG_GPU_IOCTL_PROFILER_RESERVE			\
	_IOWR(NVGPU_DBG_GPU_IOCTL_MAGIC, 22, struct nvgpu_dbg_gpu_profiler_reserve_args)

#define NVGPU_DBG_GPU_IOCTL_LAST		\
	_IOC_NR(NVGPU_DBG_GPU_IOCTL_PROFILER_RESERVE)

#define NVGPU_DBG_GPU_IOCTL_MAX_ARG_SIZE		\
	sizeof(struct nvgpu_dbg_gpu_access_fb_memory_args)

/*
 * /dev/nvhost-gpu device
 */

#define NVGPU_IOCTL_MAGIC 'H'
#define NVGPU_NO_TIMEOUT ((u32)~0)
#define NVGPU_TIMEOUT_FLAG_DISABLE_DUMP		0

/* this is also the hardware memory format */
struct nvgpu_gpfifo {
	__u32 entry0; /* first word of gpfifo entry */
	__u32 entry1; /* second word of gpfifo entry */
};

struct nvgpu_get_param_args {
	__u32 value;
} __packed;

struct nvgpu_channel_open_args {
	union {
		__s32 channel_fd; /* deprecated: use out.channel_fd instead */
		struct {
			 /* runlist_id is the runlist for the
			  * channel. Basically, the runlist specifies the target
			  * engine(s) for which the channel is
			  * opened. Runlist_id -1 is synonym for the primary
			  * graphics runlist. */
			__s32 runlist_id;
		} in;
		struct {
			__s32 channel_fd;
		} out;
	};
};

struct nvgpu_set_nvmap_fd_args {
	__u32 fd;
} __packed;

#define NVGPU_ALLOC_OBJ_FLAGS_LOCKBOOST_ZERO	(1 << 0)
/* Flags in nvgpu_alloc_obj_ctx_args.flags */
#define NVGPU_ALLOC_OBJ_FLAGS_GFXP		(1 << 1)
#define NVGPU_ALLOC_OBJ_FLAGS_CILP		(1 << 2)

struct nvgpu_alloc_obj_ctx_args {
	__u32 class_num; /* kepler3d, 2d, compute, etc       */
	__u32 flags;     /* input, output */
	__u64 obj_id;    /* output, used to free later       */
};

struct nvgpu_alloc_gpfifo_args {
	__u32 num_entries;
#define NVGPU_ALLOC_GPFIFO_FLAGS_VPR_ENABLED	(1 << 0) /* set owner channel of this gpfifo as a vpr channel */
/*
 * this flag is used in struct nvgpu_alloc_gpfifo_args
 * to enable re-playable faults for that channel
 */
#define NVGPU_ALLOC_GPFIFO_FLAGS_REPLAYABLE_FAULTS_ENABLE   (1 << 2)
	__u32 flags;
};

struct nvgpu_alloc_gpfifo_ex_args {
	__u32 num_entries;
	__u32 num_inflight_jobs;
/* Set owner channel of this gpfifo as a vpr channel. */
#define NVGPU_ALLOC_GPFIFO_EX_FLAGS_VPR_ENABLED		(1 << 0)
/*
 * Channel shall exhibit deterministic behavior in the submit path.
 *
 * NOTE: as an exception, VPR resize may still cause the GPU to reset at any
 * time, which is not deterministic behavior. If this is not acceptable, the
 * user has to make sure that VPR resize does not occur.
 *
 * With this flag, any submits with in-kernel job tracking also require that
 * num_inflight_jobs is nonzero, and additionally that
 * NVGPU_GPU_FLAGS_SUPPORT_DETERMINISTIC_SUBMIT_FULL is found in gpu
 * characteristics.flags.
 *
 * Note that fast submits (with no in-kernel job tracking) are also
 * deterministic and are supported if the characteristics flags contain
 * NVGPU_GPU_FLAGS_SUPPORT_DETERMINISTIC_SUBMIT_NO_JOBTRACKING; this flag or
 * num_inflight_jobs are not necessary in that case.
 */
#define NVGPU_ALLOC_GPFIFO_EX_FLAGS_DETERMINISTIC	(1 << 1)
	__u32 flags;
	__u32 reserved[5];
};

struct nvgpu_fence {
	__u32 id;        /* syncpoint id or sync fence fd */
	__u32 value;     /* syncpoint value (discarded when using sync fence) */
};

/* insert a wait on the fence before submitting gpfifo */
#define NVGPU_SUBMIT_GPFIFO_FLAGS_FENCE_WAIT	(1 << 0)
/* insert a fence update after submitting gpfifo and
   return the new fence for others to wait on */
#define NVGPU_SUBMIT_GPFIFO_FLAGS_FENCE_GET	(1 << 1)
/* choose between different gpfifo entry formats */
#define NVGPU_SUBMIT_GPFIFO_FLAGS_HW_FORMAT	(1 << 2)
/* interpret fence as a sync fence fd instead of raw syncpoint fence */
#define NVGPU_SUBMIT_GPFIFO_FLAGS_SYNC_FENCE	(1 << 3)
/* suppress WFI before fence trigger */
#define NVGPU_SUBMIT_GPFIFO_FLAGS_SUPPRESS_WFI	(1 << 4)
/* skip buffer refcounting during submit */
#define NVGPU_SUBMIT_GPFIFO_FLAGS_SKIP_BUFFER_REFCOUNTING	(1 << 5)

struct nvgpu_submit_gpfifo_args {
	__u64 gpfifo;
	__u32 num_entries;
	__u32 flags;
	struct nvgpu_fence fence;
};

struct nvgpu_wait_args {
#define NVGPU_WAIT_TYPE_NOTIFIER	0x0
#define NVGPU_WAIT_TYPE_SEMAPHORE	0x1
	__u32 type;
	__u32 timeout;
	union {
		struct {
			/* handle and offset for notifier memory */
			__u32 dmabuf_fd;
			__u32 offset;
			__u32 padding1;
			__u32 padding2;
		} notifier;
		struct {
			/* handle and offset for semaphore memory */
			__u32 dmabuf_fd;
			__u32 offset;
			/* semaphore payload to wait for */
			__u32 payload;
			__u32 padding;
		} semaphore;
	} condition; /* determined by type field */
};

/* cycle stats support */
struct nvgpu_cycle_stats_args {
	__u32 dmabuf_fd;
} __packed;

struct nvgpu_set_timeout_args {
	__u32 timeout;
} __packed;

struct nvgpu_set_timeout_ex_args {
	__u32 timeout;
	__u32 flags;
};

#define NVGPU_ZCULL_MODE_GLOBAL		0
#define NVGPU_ZCULL_MODE_NO_CTXSW		1
#define NVGPU_ZCULL_MODE_SEPARATE_BUFFER	2
#define NVGPU_ZCULL_MODE_PART_OF_REGULAR_BUF	3

struct nvgpu_zcull_bind_args {
	__u64 gpu_va;
	__u32 mode;
	__u32 padding;
};

struct nvgpu_set_error_notifier {
	__u64 offset;
	__u64 size;
	__u32 mem;
	__u32 padding;
};

struct nvgpu_notification {
	struct {			/* 0000- */
		__u32 nanoseconds[2];	/* nanoseconds since Jan. 1, 1970 */
	} time_stamp;			/* -0007 */
	__u32 info32;	/* info returned depends on method 0008-000b */
#define	NVGPU_CHANNEL_FIFO_ERROR_IDLE_TIMEOUT		8
#define	NVGPU_CHANNEL_GR_ERROR_SW_METHOD		12
#define	NVGPU_CHANNEL_GR_ERROR_SW_NOTIFY		13
#define	NVGPU_CHANNEL_GR_EXCEPTION			13
#define	NVGPU_CHANNEL_GR_SEMAPHORE_TIMEOUT		24
#define	NVGPU_CHANNEL_GR_ILLEGAL_NOTIFY			25
#define	NVGPU_CHANNEL_FIFO_ERROR_MMU_ERR_FLT		31
#define	NVGPU_CHANNEL_PBDMA_ERROR			32
#define NVGPU_CHANNEL_FECS_ERR_UNIMP_FIRMWARE_METHOD	37
#define	NVGPU_CHANNEL_RESETCHANNEL_VERIF_ERROR		43
#define	NVGPU_CHANNEL_PBDMA_PUSHBUFFER_CRC_MISMATCH	80
	__u16 info16;	/* info returned depends on method 000c-000d */
	__u16 status;	/* user sets bit 15, NV sets status 000e-000f */
#define	NVGPU_CHANNEL_SUBMIT_TIMEOUT		1
};

/* cycle stats snapshot buffer support for mode E */
struct nvgpu_cycle_stats_snapshot_args {
	__u32 cmd;		/* in: command to handle     */
	__u32 dmabuf_fd;	/* in: dma buffer handler    */
	__u32 extra;		/* in/out: extra payload e.g.*/
				/*    counter/start perfmon  */
	__u32 pad0[1];
};

/* valid commands to control cycle stats shared buffer */
#define NVGPU_IOCTL_CHANNEL_CYCLE_STATS_SNAPSHOT_CMD_FLUSH   0
#define NVGPU_IOCTL_CHANNEL_CYCLE_STATS_SNAPSHOT_CMD_ATTACH  1
#define NVGPU_IOCTL_CHANNEL_CYCLE_STATS_SNAPSHOT_CMD_DETACH  2

/* configure watchdog per-channel */
struct nvgpu_channel_wdt_args {
	__u32 wdt_status;
	__u32 timeout_ms;
};
#define NVGPU_IOCTL_CHANNEL_DISABLE_WDT		  (1 << 0)
#define NVGPU_IOCTL_CHANNEL_ENABLE_WDT		  (1 << 1)
#define NVGPU_IOCTL_CHANNEL_WDT_FLAG_SET_TIMEOUT  (1 << 2)
#define NVGPU_IOCTL_CHANNEL_WDT_FLAG_DISABLE_DUMP (1 << 3)

/*
 * Interleaving channels in a runlist is an approach to improve
 * GPU scheduling by allowing certain channels to appear multiple
 * times on the runlist. The number of times a channel appears is
 * governed by the following levels:
 *
 * low (L)   : appears once
 * medium (M): if L, appears L times
 *             else, appears once
 * high (H)  : if L, appears (M + 1) x L times
 *             else if M, appears M times
 *             else, appears once
 */
struct nvgpu_runlist_interleave_args {
	__u32 level;
	__u32 reserved;
};
#define NVGPU_RUNLIST_INTERLEAVE_LEVEL_LOW	0
#define NVGPU_RUNLIST_INTERLEAVE_LEVEL_MEDIUM	1
#define NVGPU_RUNLIST_INTERLEAVE_LEVEL_HIGH	2
#define NVGPU_RUNLIST_INTERLEAVE_NUM_LEVELS	3

/* controls how long a channel occupies an engine uninterrupted */
struct nvgpu_timeslice_args {
	__u32 timeslice_us;
	__u32 reserved;
};

struct nvgpu_event_id_ctrl_args {
	__u32 cmd; /* in */
	__u32 event_id; /* in */
	__s32 event_fd; /* out */
	__u32 padding;
};
#define NVGPU_IOCTL_CHANNEL_EVENT_ID_BPT_INT		0
#define NVGPU_IOCTL_CHANNEL_EVENT_ID_BPT_PAUSE		1
#define NVGPU_IOCTL_CHANNEL_EVENT_ID_BLOCKING_SYNC	2
#define NVGPU_IOCTL_CHANNEL_EVENT_ID_CILP_PREEMPTION_STARTED	3
#define NVGPU_IOCTL_CHANNEL_EVENT_ID_CILP_PREEMPTION_COMPLETE	4
#define NVGPU_IOCTL_CHANNEL_EVENT_ID_GR_SEMAPHORE_WRITE_AWAKEN	5
#define NVGPU_IOCTL_CHANNEL_EVENT_ID_MAX		6

#define NVGPU_IOCTL_CHANNEL_EVENT_ID_CMD_ENABLE		1

struct nvgpu_preemption_mode_args {
/* only one should be enabled at a time */
#define NVGPU_GRAPHICS_PREEMPTION_MODE_WFI              (1 << 0)
#define NVGPU_GRAPHICS_PREEMPTION_MODE_GFXP		(1 << 1)
	__u32 graphics_preempt_mode; /* in */

/* only one should be enabled at a time */
#define NVGPU_COMPUTE_PREEMPTION_MODE_WFI               (1 << 0)
#define NVGPU_COMPUTE_PREEMPTION_MODE_CTA               (1 << 1)
#define NVGPU_COMPUTE_PREEMPTION_MODE_CILP		(1 << 2)

	__u32 compute_preempt_mode; /* in */
};

struct nvgpu_boosted_ctx_args {
#define NVGPU_BOOSTED_CTX_MODE_NORMAL			(0U)
#define NVGPU_BOOSTED_CTX_MODE_BOOSTED_EXECUTION	(1U)
	__u32 boost;
	__u32 padding;
};

struct nvgpu_get_user_syncpoint_args {
	__u64 gpu_va;		/* out */
	__u32 syncpoint_id;	/* out */
	__u32 syncpoint_max;	/* out */
};

struct nvgpu_reschedule_runlist_args {
#define NVGPU_RESCHEDULE_RUNLIST_PREEMPT_NEXT           (1 << 0)
	__u32 flags;
};

#define NVGPU_IOCTL_CHANNEL_SET_NVMAP_FD	\
	_IOW(NVGPU_IOCTL_MAGIC, 5, struct nvgpu_set_nvmap_fd_args)
#define NVGPU_IOCTL_CHANNEL_SET_TIMEOUT	\
	_IOW(NVGPU_IOCTL_MAGIC, 11, struct nvgpu_set_timeout_args)
#define NVGPU_IOCTL_CHANNEL_GET_TIMEDOUT	\
	_IOR(NVGPU_IOCTL_MAGIC, 12, struct nvgpu_get_param_args)
#define NVGPU_IOCTL_CHANNEL_SET_TIMEOUT_EX	\
	_IOWR(NVGPU_IOCTL_MAGIC, 18, struct nvgpu_set_timeout_ex_args)
#define NVGPU_IOCTL_CHANNEL_ALLOC_GPFIFO	\
	_IOW(NVGPU_IOCTL_MAGIC,  100, struct nvgpu_alloc_gpfifo_args)
#define NVGPU_IOCTL_CHANNEL_WAIT		\
	_IOWR(NVGPU_IOCTL_MAGIC, 102, struct nvgpu_wait_args)
#define NVGPU_IOCTL_CHANNEL_CYCLE_STATS	\
	_IOWR(NVGPU_IOCTL_MAGIC, 106, struct nvgpu_cycle_stats_args)
#define NVGPU_IOCTL_CHANNEL_SUBMIT_GPFIFO	\
	_IOWR(NVGPU_IOCTL_MAGIC, 107, struct nvgpu_submit_gpfifo_args)
#define NVGPU_IOCTL_CHANNEL_ALLOC_OBJ_CTX	\
	_IOWR(NVGPU_IOCTL_MAGIC, 108, struct nvgpu_alloc_obj_ctx_args)
#define NVGPU_IOCTL_CHANNEL_ZCULL_BIND		\
	_IOWR(NVGPU_IOCTL_MAGIC, 110, struct nvgpu_zcull_bind_args)
#define NVGPU_IOCTL_CHANNEL_SET_ERROR_NOTIFIER  \
	_IOWR(NVGPU_IOCTL_MAGIC, 111, struct nvgpu_set_error_notifier)
#define NVGPU_IOCTL_CHANNEL_OPEN	\
	_IOR(NVGPU_IOCTL_MAGIC,  112, struct nvgpu_channel_open_args)
#define NVGPU_IOCTL_CHANNEL_ENABLE	\
	_IO(NVGPU_IOCTL_MAGIC,  113)
#define NVGPU_IOCTL_CHANNEL_DISABLE	\
	_IO(NVGPU_IOCTL_MAGIC,  114)
#define NVGPU_IOCTL_CHANNEL_PREEMPT	\
	_IO(NVGPU_IOCTL_MAGIC,  115)
#define NVGPU_IOCTL_CHANNEL_FORCE_RESET	\
	_IO(NVGPU_IOCTL_MAGIC,  116)
#define NVGPU_IOCTL_CHANNEL_EVENT_ID_CTRL \
	_IOWR(NVGPU_IOCTL_MAGIC, 117, struct nvgpu_event_id_ctrl_args)
#define NVGPU_IOCTL_CHANNEL_CYCLE_STATS_SNAPSHOT	\
	_IOWR(NVGPU_IOCTL_MAGIC, 118, struct nvgpu_cycle_stats_snapshot_args)
#define NVGPU_IOCTL_CHANNEL_WDT \
	_IOW(NVGPU_IOCTL_MAGIC, 119, struct nvgpu_channel_wdt_args)
#define NVGPU_IOCTL_CHANNEL_SET_RUNLIST_INTERLEAVE \
	_IOW(NVGPU_IOCTL_MAGIC, 120, struct nvgpu_runlist_interleave_args)
#define NVGPU_IOCTL_CHANNEL_SET_TIMESLICE \
	_IOW(NVGPU_IOCTL_MAGIC, 121, struct nvgpu_timeslice_args)
#define NVGPU_IOCTL_CHANNEL_SET_PREEMPTION_MODE \
	_IOW(NVGPU_IOCTL_MAGIC, 122, struct nvgpu_preemption_mode_args)
#define NVGPU_IOCTL_CHANNEL_ALLOC_GPFIFO_EX	\
	_IOW(NVGPU_IOCTL_MAGIC, 123, struct nvgpu_alloc_gpfifo_ex_args)
#define NVGPU_IOCTL_CHANNEL_SET_BOOSTED_CTX	\
	_IOW(NVGPU_IOCTL_MAGIC, 124, struct nvgpu_boosted_ctx_args)
#define NVGPU_IOCTL_CHANNEL_GET_TIMESLICE \
	_IOW(NVGPU_IOCTL_MAGIC, 125, struct nvgpu_timeslice_args)
#define NVGPU_IOCTL_CHANNEL_GET_USER_SYNCPOINT \
	_IOR(NVGPU_IOCTL_MAGIC, 126, struct nvgpu_get_user_syncpoint_args)
#define NVGPU_IOCTL_CHANNEL_RESCHEDULE_RUNLIST	\
	_IOW(NVGPU_IOCTL_MAGIC, 127, struct nvgpu_reschedule_runlist_args)

#define NVGPU_IOCTL_CHANNEL_LAST	\
	_IOC_NR(NVGPU_IOCTL_CHANNEL_RESCHEDULE_RUNLIST)
#define NVGPU_IOCTL_CHANNEL_MAX_ARG_SIZE sizeof(struct nvgpu_alloc_gpfifo_ex_args)

/*
 * /dev/nvhost-as-gpu device
 *
 * Opening a '/dev/nvhost-as-gpu' device node creates a new address
 * space.  nvgpu channels (for the same module) can then be bound to such an
 * address space to define the addresses it has access to.
 *
 * Once a nvgpu channel has been bound to an address space it cannot be
 * unbound.  There is no support for allowing an nvgpu channel to change from
 * one address space to another (or from one to none).
 *
 * As long as there is an open device file to the address space, or any bound
 * nvgpu channels it will be valid.  Once all references to the address space
 * are removed the address space is deleted.
 *
 */

#define NVGPU_AS_IOCTL_MAGIC 'A'

/*
 * Allocating an address space range:
 *
 * Address ranges created with this ioctl are reserved for later use with
 * fixed-address buffer mappings.
 *
 * If _FLAGS_FIXED_OFFSET is specified then the new range starts at the 'offset'
 * given.  Otherwise the address returned is chosen to be a multiple of 'align.'
 *
 */
struct nvgpu32_as_alloc_space_args {
	__u32 pages;     /* in, pages */
	__u32 page_size; /* in, bytes */
	__u32 flags;     /* in */
#define NVGPU_AS_ALLOC_SPACE_FLAGS_FIXED_OFFSET 0x1
#define NVGPU_AS_ALLOC_SPACE_FLAGS_SPARSE 0x2
	union {
		__u64 offset; /* inout, byte address valid iff _FIXED_OFFSET */
		__u64 align;  /* in, alignment multiple (0:={1 or n/a}) */
	} o_a;
};

struct nvgpu_as_alloc_space_args {
	__u32 pages;     /* in, pages */
	__u32 page_size; /* in, bytes */
	__u32 flags;     /* in */
	__u32 padding;     /* in */
	union {
		__u64 offset; /* inout, byte address valid iff _FIXED_OFFSET */
		__u64 align;  /* in, alignment multiple (0:={1 or n/a}) */
	} o_a;
};

/*
 * Releasing an address space range:
 *
 * The previously allocated region starting at 'offset' is freed.  If there are
 * any buffers currently mapped inside the region the ioctl will fail.
 */
struct nvgpu_as_free_space_args {
	__u64 offset; /* in, byte address */
	__u32 pages;     /* in, pages */
	__u32 page_size; /* in, bytes */
};

/*
 * Binding a nvgpu channel to an address space:
 *
 * A channel must be bound to an address space before allocating a gpfifo
 * in nvgpu.  The 'channel_fd' given here is the fd used to allocate the
 * channel.  Once a channel has been bound to an address space it cannot
 * be unbound (except for when the channel is destroyed).
 */
struct nvgpu_as_bind_channel_args {
	__u32 channel_fd; /* in */
} __packed;

/*
 * Mapping nvmap buffers into an address space:
 *
 * The start address is the 'offset' given if _FIXED_OFFSET is specified.
 * Otherwise the address returned is a multiple of 'align.'
 *
 * If 'page_size' is set to 0 the nvmap buffer's allocation alignment/sizing
 * will be used to determine the page size (largest possible).  The page size
 * chosen will be returned back to the caller in the 'page_size' parameter in
 * that case.
 */
#define NVGPU_AS_MAP_BUFFER_FLAGS_FIXED_OFFSET	    (1 << 0)
#define NVGPU_AS_MAP_BUFFER_FLAGS_CACHEABLE	    (1 << 2)
#define NVGPU_AS_MAP_BUFFER_FLAGS_IO_COHERENT	    (1 << 4)
#define NVGPU_AS_MAP_BUFFER_FLAGS_UNMAPPED_PTE	    (1 << 5)
#define NVGPU_AS_MAP_BUFFER_FLAGS_MAPPABLE_COMPBITS (1 << 6)
#define NVGPU_AS_MAP_BUFFER_FLAGS_L3_ALLOC          (1 << 7)
#define NVGPU_AS_MAP_BUFFER_FLAGS_DIRECT_KIND_CTRL  (1 << 8)

/*
 * Mapping dmabuf fds into an address space:
 *
 * The caller requests a mapping to a particular page 'kind'.
 *
 * If 'page_size' is set to 0 the dmabuf's alignment/sizing will be used to
 * determine the page size (largest possible).  The page size chosen will be
 * returned back to the caller in the 'page_size' parameter in that case.
 */
struct nvgpu_as_map_buffer_ex_args {
	/* NVGPU_AS_MAP_BUFFER_FLAGS_DIRECT_KIND_CTRL must be set */
	__u32 flags;		/* in/out */

	/*
	 * - If both compr_kind and incompr_kind are set
	 *   (i.e., value is other than NV_KIND_INVALID),
	 *   kernel attempts to use compr_kind first.
	 *
	 * - If compr_kind is set, kernel attempts to allocate
	 *   comptags for the buffer. If successful,
	 *   compr_kind is used as the PTE kind.
	 *
	 * - If incompr_kind is set, kernel uses incompr_kind as the
	 *   PTE kind, if compr_kind cannot be used. Comptags are not
	 *   allocated.
	 *
	 * - If neither compr_kind or incompr_kind is set, the
	 *   map call will fail.
	 */
#define NV_KIND_INVALID -1
	__s16 compr_kind;
	__s16 incompr_kind;

	__u32 dmabuf_fd;	/* in */
	__u32 page_size;	/* inout, 0:= best fit to buffer */

	__u64 buffer_offset;	/* in, offset of mapped buffer region */
	__u64 mapping_size;	/* in, size of mapped buffer region */

	__u64 offset;		/* in/out, we use this address if flag
				 * FIXED_OFFSET is set. This will fail
				 * if space is not properly allocated. The
				 * actual virtual address to which we mapped
				 * the buffer is returned in this field. */
};

/*
 * Get info about buffer compbits. Requires that buffer is mapped with
 * NVGPU_AS_MAP_BUFFER_FLAGS_MAPPABLE_COMPBITS.
 *
 * The compbits for a mappable buffer are organized in a mappable
 * window to the compbits store. In case the window contains comptags
 * for more than one buffer, the buffer comptag line index may differ
 * from the window comptag line index.
 */
struct nvgpu_as_get_buffer_compbits_info_args {

	/* in: address of an existing buffer mapping */
	__u64 mapping_gva;

	/* out: size of compbits mapping window (bytes) */
	__u64 compbits_win_size;

	/* out: comptag line index of the window start */
	__u32 compbits_win_ctagline;

	/* out: comptag line index of the buffer mapping */
	__u32 mapping_ctagline;

/* Buffer uses compbits */
#define NVGPU_AS_GET_BUFFER_COMPBITS_INFO_FLAGS_HAS_COMPBITS    (1 << 0)

/* Buffer compbits are mappable */
#define NVGPU_AS_GET_BUFFER_COMPBITS_INFO_FLAGS_MAPPABLE        (1 << 1)

/* Buffer IOVA addresses are discontiguous */
#define NVGPU_AS_GET_BUFFER_COMPBITS_INFO_FLAGS_DISCONTIG_IOVA  (1 << 2)

	/* out */
	__u32 flags;

	__u32 reserved1;
};

/*
 * Map compbits of a mapped buffer to the GPU address space. The
 * compbits mapping is automatically unmapped when the buffer is
 * unmapped.
 *
 * The compbits mapping always uses small pages, it is read-only, and
 * is GPU cacheable. The mapping is a window to the compbits
 * store. The window may not be exactly the size of the cache lines
 * for the buffer mapping.
 */
struct nvgpu_as_map_buffer_compbits_args {

	/* in: address of an existing buffer mapping */
	__u64 mapping_gva;

	/* in: gva to the mapped compbits store window when
	 * FIXED_OFFSET is set. Otherwise, ignored and should be be 0.
	 *
	 * For FIXED_OFFSET mapping:
	 * - If compbits are already mapped compbits_win_gva
	 *   must match with the previously mapped gva.
	 * - The user must have allocated enough GVA space for the
	 *   mapping window (see compbits_win_size in
	 *   nvgpu_as_get_buffer_compbits_info_args)
	 *
	 * out: gva to the mapped compbits store window */
	__u64 compbits_win_gva;

	/* in: reserved, must be 0
	   out: physical or IOMMU address for mapping */
	union {
		/* contiguous iova addresses */
		__u64 mapping_iova;

		/* buffer to receive discontiguous iova addresses (reserved) */
		__u64 mapping_iova_buf_addr;
	};

	/* in: Buffer size (in bytes) for discontiguous iova
	 * addresses. Reserved, must be 0. */
	__u64 mapping_iova_buf_size;

#define NVGPU_AS_MAP_BUFFER_COMPBITS_FLAGS_FIXED_OFFSET        (1 << 0)
	__u32 flags;
	__u32 reserved1;
};

/*
 * Unmapping a buffer:
 *
 * To unmap a previously mapped buffer set 'offset' to the offset returned in
 * the mapping call.  This includes where a buffer has been mapped into a fixed
 * offset of a previously allocated address space range.
 */
struct nvgpu_as_unmap_buffer_args {
	__u64 offset; /* in, byte address */
};


struct nvgpu_as_va_region {
	__u64 offset;
	__u32 page_size;
	__u32 reserved;
	__u64 pages;
};

struct nvgpu_as_get_va_regions_args {
	__u64 buf_addr; /* Pointer to array of nvgpu_as_va_region:s.
			 * Ignored if buf_size is 0 */
	__u32 buf_size; /* in:  userspace buf size (in bytes)
			   out: kernel buf size    (in bytes) */
	__u32 reserved;
};

struct nvgpu_as_map_buffer_batch_args {
	__u64 unmaps; /* ptr to array of nvgpu_as_unmap_buffer_args */
	__u64 maps;   /* ptr to array of nvgpu_as_map_buffer_ex_args */
	__u32 num_unmaps; /* in: number of unmaps
			   * out: on error, number of successful unmaps */
	__u32 num_maps;   /* in: number of maps
			   * out: on error, number of successful maps */
	__u64 reserved;
};

struct nvgpu_as_get_sync_ro_map_args {
	__u64 base_gpuva;
	__u32 sync_size;
	__u32 padding;
};

#define NVGPU_AS_IOCTL_BIND_CHANNEL \
	_IOWR(NVGPU_AS_IOCTL_MAGIC, 1, struct nvgpu_as_bind_channel_args)
#define NVGPU32_AS_IOCTL_ALLOC_SPACE \
	_IOWR(NVGPU_AS_IOCTL_MAGIC, 2, struct nvgpu32_as_alloc_space_args)
#define NVGPU_AS_IOCTL_FREE_SPACE \
	_IOWR(NVGPU_AS_IOCTL_MAGIC, 3, struct nvgpu_as_free_space_args)
#define NVGPU_AS_IOCTL_MAP_BUFFER \
	_IOWR(NVGPU_AS_IOCTL_MAGIC, 4, struct nvgpu_as_map_buffer_args)
#define NVGPU_AS_IOCTL_UNMAP_BUFFER \
	_IOWR(NVGPU_AS_IOCTL_MAGIC, 5, struct nvgpu_as_unmap_buffer_args)
#define NVGPU_AS_IOCTL_ALLOC_SPACE \
	_IOWR(NVGPU_AS_IOCTL_MAGIC, 6, struct nvgpu_as_alloc_space_args)
#define NVGPU_AS_IOCTL_MAP_BUFFER_EX \
	_IOWR(NVGPU_AS_IOCTL_MAGIC, 7, struct nvgpu_as_map_buffer_ex_args)
#define NVGPU_AS_IOCTL_GET_VA_REGIONS \
	_IOWR(NVGPU_AS_IOCTL_MAGIC, 8, struct nvgpu_as_get_va_regions_args)
#define NVGPU_AS_IOCTL_GET_BUFFER_COMPBITS_INFO \
	_IOWR(NVGPU_AS_IOCTL_MAGIC, 9, struct nvgpu_as_get_buffer_compbits_info_args)
#define NVGPU_AS_IOCTL_MAP_BUFFER_COMPBITS \
	_IOWR(NVGPU_AS_IOCTL_MAGIC, 10, struct nvgpu_as_map_buffer_compbits_args)
#define NVGPU_AS_IOCTL_MAP_BUFFER_BATCH	\
	_IOWR(NVGPU_AS_IOCTL_MAGIC, 11, struct nvgpu_as_map_buffer_batch_args)
#define NVGPU_AS_IOCTL_GET_SYNC_RO_MAP	\
	_IOR(NVGPU_AS_IOCTL_MAGIC,  12, struct nvgpu_as_get_sync_ro_map_args)

#define NVGPU_AS_IOCTL_LAST            \
	_IOC_NR(NVGPU_AS_IOCTL_GET_SYNC_RO_MAP)
#define NVGPU_AS_IOCTL_MAX_ARG_SIZE	\
	sizeof(struct nvgpu_as_map_buffer_ex_args)


/*
 * /dev/nvhost-ctxsw-gpu device
 *
 * Opening a '/dev/nvhost-ctxsw-gpu' device node creates a way to trace
 * context switches on GR engine
 */

#define NVGPU_CTXSW_IOCTL_MAGIC 'C'

#define NVGPU_CTXSW_TAG_SOF			0x00
#define NVGPU_CTXSW_TAG_CTXSW_REQ_BY_HOST	0x01
#define NVGPU_CTXSW_TAG_FE_ACK			0x02
#define NVGPU_CTXSW_TAG_FE_ACK_WFI		0x0a
#define NVGPU_CTXSW_TAG_FE_ACK_GFXP		0x0b
#define NVGPU_CTXSW_TAG_FE_ACK_CTAP		0x0c
#define NVGPU_CTXSW_TAG_FE_ACK_CILP		0x0d
#define NVGPU_CTXSW_TAG_SAVE_END		0x03
#define NVGPU_CTXSW_TAG_RESTORE_START		0x04
#define NVGPU_CTXSW_TAG_CONTEXT_START		0x05
#define NVGPU_CTXSW_TAG_ENGINE_RESET		0xfe
#define NVGPU_CTXSW_TAG_INVALID_TIMESTAMP	0xff
#define NVGPU_CTXSW_TAG_LAST			\
	NVGPU_CTXSW_TAG_INVALID_TIMESTAMP

struct nvgpu_ctxsw_trace_entry {
	__u8 tag;
	__u8 vmid;
	__u16 seqno;		/* sequence number to detect drops */
	__u32 context_id;	/* context_id as allocated by FECS */
	__u64 pid;		/* 64-bit is max bits of different OS pid */
	__u64 timestamp;	/* 64-bit time */
};

#define NVGPU_CTXSW_RING_HEADER_MAGIC 0x7000fade
#define NVGPU_CTXSW_RING_HEADER_VERSION 0

struct nvgpu_ctxsw_ring_header {
	__u32 magic;
	__u32 version;
	__u32 num_ents;
	__u32 ent_size;
	volatile __u32 drop_count;	/* excluding filtered out events */
	volatile __u32 write_seqno;
	volatile __u32 write_idx;
	volatile __u32 read_idx;
};

struct nvgpu_ctxsw_ring_setup_args {
	__u32 size;	/* [in/out] size of ring buffer in bytes (including
			   header). will be rounded page size. this parameter
			   is updated with actual allocated size. */
};

#define NVGPU_CTXSW_FILTER_SIZE	(NVGPU_CTXSW_TAG_LAST + 1)
#define NVGPU_CTXSW_FILTER_SET(n, p) \
	((p)->tag_bits[(n) / 64] |=  (1 << ((n) & 63)))
#define NVGPU_CTXSW_FILTER_CLR(n, p) \
	((p)->tag_bits[(n) / 64] &= ~(1 << ((n) & 63)))
#define NVGPU_CTXSW_FILTER_ISSET(n, p) \
	((p)->tag_bits[(n) / 64] &   (1 << ((n) & 63)))
#define NVGPU_CTXSW_FILTER_CLR_ALL(p)    memset((void *)(p), 0, sizeof(*(p)))
#define NVGPU_CTXSW_FILTER_SET_ALL(p)    memset((void *)(p), ~0, sizeof(*(p)))

struct nvgpu_ctxsw_trace_filter {
	__u64 tag_bits[(NVGPU_CTXSW_FILTER_SIZE + 63) / 64];
};

struct nvgpu_ctxsw_trace_filter_args {
	struct nvgpu_ctxsw_trace_filter filter;
};

#define NVGPU_CTXSW_IOCTL_TRACE_ENABLE \
	_IO(NVGPU_CTXSW_IOCTL_MAGIC, 1)
#define NVGPU_CTXSW_IOCTL_TRACE_DISABLE \
	_IO(NVGPU_CTXSW_IOCTL_MAGIC, 2)
#define NVGPU_CTXSW_IOCTL_RING_SETUP \
	_IOWR(NVGPU_CTXSW_IOCTL_MAGIC, 3, struct nvgpu_ctxsw_ring_setup_args)
#define NVGPU_CTXSW_IOCTL_SET_FILTER \
	_IOW(NVGPU_CTXSW_IOCTL_MAGIC, 4, struct nvgpu_ctxsw_trace_filter_args)
#define NVGPU_CTXSW_IOCTL_GET_FILTER \
	_IOR(NVGPU_CTXSW_IOCTL_MAGIC, 5, struct nvgpu_ctxsw_trace_filter_args)
#define NVGPU_CTXSW_IOCTL_POLL \
	_IO(NVGPU_CTXSW_IOCTL_MAGIC, 6)

#define NVGPU_CTXSW_IOCTL_LAST            \
	_IOC_NR(NVGPU_CTXSW_IOCTL_POLL)

#define NVGPU_CTXSW_IOCTL_MAX_ARG_SIZE	\
	sizeof(struct nvgpu_ctxsw_trace_filter_args)

/*
 * /dev/nvhost-sched-gpu device
 *
 * Opening a '/dev/nvhost-sched-gpu' device node creates a way to control
 * GPU scheduling parameters.
 */

#define NVGPU_SCHED_IOCTL_MAGIC 'S'

/*
 * When the app manager receives a NVGPU_SCHED_STATUS_TSG_OPEN notification,
 * it is expected to query the list of recently opened TSGs using
 * NVGPU_SCHED_IOCTL_GET_RECENT_TSGS. The kernel driver maintains a bitmap
 * of recently opened TSGs. When the app manager queries the list, it
 * atomically clears the bitmap. This way, at each invocation of
 * NVGPU_SCHED_IOCTL_GET_RECENT_TSGS, app manager only receives the list of
 * TSGs that have been opened since last invocation.
 *
 * If the app manager needs to re-synchronize with the driver, it can use
 * NVGPU_SCHED_IOCTL_GET_TSGS to retrieve the complete list of TSGs. The
 * recent TSG bitmap will be cleared in that case too.
 */
struct nvgpu_sched_get_tsgs_args {
	/* in: size of buffer in bytes */
	/* out: actual size of size of TSG bitmap. if user-provided size is too
	 * small, ioctl will return -ENOSPC, and update this field, allowing
	 * application to discover required number of bytes and allocate
	 * a buffer accordingly.
	 */
	__u32 size;

	/* in: address of 64-bit aligned buffer */
	/* out: buffer contains a TSG bitmap.
	 * Bit #n will be set in the bitmap if TSG #n is present.
	 * When using NVGPU_SCHED_IOCTL_GET_RECENT_TSGS, the first time you use
	 * this command, it will return the opened TSGs and subsequent calls
	 * will only return the delta (ie. each invocation clears bitmap)
	 */
	__u64 buffer;
};

struct nvgpu_sched_get_tsgs_by_pid_args {
	/* in: process id for which we want to retrieve TSGs */
	__u64 pid;

	/* in: size of buffer in bytes */
	/* out: actual size of size of TSG bitmap. if user-provided size is too
	 * small, ioctl will return -ENOSPC, and update this field, allowing
	 * application to discover required number of bytes and allocate
	 * a buffer accordingly.
	 */
	__u32 size;

	/* in: address of 64-bit aligned buffer */
	/* out: buffer contains a TSG bitmap. */
	__u64 buffer;
};

struct nvgpu_sched_tsg_get_params_args {
	__u32 tsgid;		/* in: TSG identifier */
	__u32 timeslice;	/* out: timeslice in usecs */
	__u32 runlist_interleave;
	__u32 graphics_preempt_mode;
	__u32 compute_preempt_mode;
	__u64 pid;		/* out: process identifier of TSG owner */
};

struct nvgpu_sched_tsg_timeslice_args {
	__u32 tsgid;                    /* in: TSG identifier */
	__u32 timeslice;                /* in: timeslice in usecs */
};

struct nvgpu_sched_tsg_runlist_interleave_args {
	__u32 tsgid;			/* in: TSG identifier */

		/* in: see NVGPU_RUNLIST_INTERLEAVE_LEVEL_ */
	__u32 runlist_interleave;
};

struct nvgpu_sched_api_version_args {
	__u32 version;
};

struct nvgpu_sched_tsg_refcount_args {
	__u32 tsgid;                    /* in: TSG identifier */
};

#define NVGPU_SCHED_IOCTL_GET_TSGS					\
	_IOWR(NVGPU_SCHED_IOCTL_MAGIC, 1,				\
		struct nvgpu_sched_get_tsgs_args)
#define NVGPU_SCHED_IOCTL_GET_RECENT_TSGS				\
	_IOWR(NVGPU_SCHED_IOCTL_MAGIC, 2,				\
		struct nvgpu_sched_get_tsgs_args)
#define NVGPU_SCHED_IOCTL_GET_TSGS_BY_PID				\
	_IOWR(NVGPU_SCHED_IOCTL_MAGIC, 3,				\
		struct nvgpu_sched_get_tsgs_by_pid_args)
#define NVGPU_SCHED_IOCTL_TSG_GET_PARAMS				\
	_IOWR(NVGPU_SCHED_IOCTL_MAGIC, 4,				\
		struct nvgpu_sched_tsg_get_params_args)
#define NVGPU_SCHED_IOCTL_TSG_SET_TIMESLICE				\
	_IOW(NVGPU_SCHED_IOCTL_MAGIC, 5,				\
		struct nvgpu_sched_tsg_timeslice_args)
#define NVGPU_SCHED_IOCTL_TSG_SET_RUNLIST_INTERLEAVE			\
	_IOW(NVGPU_SCHED_IOCTL_MAGIC, 6,				\
		struct nvgpu_sched_tsg_runlist_interleave_args)
#define NVGPU_SCHED_IOCTL_LOCK_CONTROL					\
	_IO(NVGPU_SCHED_IOCTL_MAGIC, 7)
#define NVGPU_SCHED_IOCTL_UNLOCK_CONTROL				\
	_IO(NVGPU_SCHED_IOCTL_MAGIC, 8)
#define NVGPU_SCHED_IOCTL_GET_API_VERSION				\
	_IOR(NVGPU_SCHED_IOCTL_MAGIC, 9,				\
	    struct nvgpu_sched_api_version_args)
#define NVGPU_SCHED_IOCTL_GET_TSG					\
	_IOW(NVGPU_SCHED_IOCTL_MAGIC, 10,				\
		struct nvgpu_sched_tsg_refcount_args)
#define NVGPU_SCHED_IOCTL_PUT_TSG					\
	_IOW(NVGPU_SCHED_IOCTL_MAGIC, 11,				\
		struct nvgpu_sched_tsg_refcount_args)
#define NVGPU_SCHED_IOCTL_LAST						\
	_IOC_NR(NVGPU_SCHED_IOCTL_PUT_TSG)

#define NVGPU_SCHED_IOCTL_MAX_ARG_SIZE					\
	sizeof(struct nvgpu_sched_tsg_get_params_args)


#define NVGPU_SCHED_SET(n, bitmap)	\
	(((__u64 *)(bitmap))[(n) / 64] |=  (1ULL << (((__u64)n) & 63)))
#define NVGPU_SCHED_CLR(n, bitmap)	\
	(((__u64 *)(bitmap))[(n) / 64] &= ~(1ULL << (((__u64)n) & 63)))
#define NVGPU_SCHED_ISSET(n, bitmap)	\
	(((__u64 *)(bitmap))[(n) / 64] & (1ULL << (((__u64)n) & 63)))

#define NVGPU_SCHED_STATUS_TSG_OPEN	(1ULL << 0)

struct nvgpu_sched_event_arg {
	__u64 reserved;
	__u64 status;
};

#define NVGPU_SCHED_API_VERSION		1

#endif