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
|
/* $NetBSD: spec_vnops.c,v 1.218 2023/04/22 15:32:49 riastradh Exp $ */
/*-
* Copyright (c) 2008 The NetBSD Foundation, Inc.
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
* ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
* TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
* PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
* BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
* CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
* SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
* INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
* CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
* ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
* POSSIBILITY OF SUCH DAMAGE.
*/
/*
* Copyright (c) 1989, 1993
* The Regents of the University of California. All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. Neither the name of the University nor the names of its contributors
* may be used to endorse or promote products derived from this software
* without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*
* @(#)spec_vnops.c 8.15 (Berkeley) 7/14/95
*/
#include <sys/cdefs.h>
__KERNEL_RCSID(0, "$NetBSD: spec_vnops.c,v 1.218 2023/04/22 15:32:49 riastradh Exp $");
#ifdef _KERNEL_OPT
#include "opt_ddb.h"
#endif
#include <sys/param.h>
#include <sys/proc.h>
#include <sys/systm.h>
#include <sys/kernel.h>
#include <sys/conf.h>
#include <sys/buf.h>
#include <sys/mount.h>
#include <sys/namei.h>
#include <sys/vnode_impl.h>
#include <sys/stat.h>
#include <sys/errno.h>
#include <sys/ioctl.h>
#include <sys/poll.h>
#include <sys/file.h>
#include <sys/disklabel.h>
#include <sys/disk.h>
#include <sys/lockf.h>
#include <sys/tty.h>
#include <sys/kauth.h>
#include <sys/fstrans.h>
#include <sys/module.h>
#include <sys/atomic.h>
#include <miscfs/genfs/genfs.h>
#include <miscfs/specfs/specdev.h>
#ifdef DDB
#include <ddb/ddb.h>
#endif
/*
* Lock order:
*
* vnode lock
* -> device_lock
* -> struct vnode::v_interlock
*/
/* symbolic sleep message strings for devices */
const char devopn[] = "devopn";
const char devio[] = "devio";
const char devwait[] = "devwait";
const char devin[] = "devin";
const char devout[] = "devout";
const char devioc[] = "devioc";
const char devcls[] = "devcls";
#define SPECHSZ 64
#if ((SPECHSZ&(SPECHSZ-1)) == 0)
#define SPECHASH(rdev) (((rdev>>5)+(rdev))&(SPECHSZ-1))
#else
#define SPECHASH(rdev) (((unsigned)((rdev>>5)+(rdev)))%SPECHSZ)
#endif
static vnode_t *specfs_hash[SPECHSZ];
extern struct mount *dead_rootmount;
/*
* This vnode operations vector is used for special device nodes
* created from whole cloth by the kernel. For the ops vector for
* vnodes built from special devices found in a filesystem, see (e.g)
* ffs_specop_entries[] in ffs_vnops.c or the equivalent for other
* filesystems.
*/
int (**spec_vnodeop_p)(void *);
const struct vnodeopv_entry_desc spec_vnodeop_entries[] = {
{ &vop_default_desc, vn_default_error },
{ &vop_parsepath_desc, genfs_parsepath }, /* parsepath */
{ &vop_lookup_desc, spec_lookup }, /* lookup */
{ &vop_create_desc, genfs_badop }, /* create */
{ &vop_mknod_desc, genfs_badop }, /* mknod */
{ &vop_open_desc, spec_open }, /* open */
{ &vop_close_desc, spec_close }, /* close */
{ &vop_access_desc, genfs_ebadf }, /* access */
{ &vop_accessx_desc, genfs_ebadf }, /* accessx */
{ &vop_getattr_desc, genfs_ebadf }, /* getattr */
{ &vop_setattr_desc, genfs_ebadf }, /* setattr */
{ &vop_read_desc, spec_read }, /* read */
{ &vop_write_desc, spec_write }, /* write */
{ &vop_fallocate_desc, genfs_eopnotsupp }, /* fallocate */
{ &vop_fdiscard_desc, spec_fdiscard }, /* fdiscard */
{ &vop_fcntl_desc, genfs_fcntl }, /* fcntl */
{ &vop_ioctl_desc, spec_ioctl }, /* ioctl */
{ &vop_poll_desc, spec_poll }, /* poll */
{ &vop_kqfilter_desc, spec_kqfilter }, /* kqfilter */
{ &vop_revoke_desc, genfs_revoke }, /* revoke */
{ &vop_mmap_desc, spec_mmap }, /* mmap */
{ &vop_fsync_desc, spec_fsync }, /* fsync */
{ &vop_seek_desc, spec_seek }, /* seek */
{ &vop_remove_desc, genfs_badop }, /* remove */
{ &vop_link_desc, genfs_badop }, /* link */
{ &vop_rename_desc, genfs_badop }, /* rename */
{ &vop_mkdir_desc, genfs_badop }, /* mkdir */
{ &vop_rmdir_desc, genfs_badop }, /* rmdir */
{ &vop_symlink_desc, genfs_badop }, /* symlink */
{ &vop_readdir_desc, genfs_badop }, /* readdir */
{ &vop_readlink_desc, genfs_badop }, /* readlink */
{ &vop_abortop_desc, genfs_badop }, /* abortop */
{ &vop_inactive_desc, spec_inactive }, /* inactive */
{ &vop_reclaim_desc, spec_reclaim }, /* reclaim */
{ &vop_lock_desc, genfs_lock }, /* lock */
{ &vop_unlock_desc, genfs_unlock }, /* unlock */
{ &vop_bmap_desc, spec_bmap }, /* bmap */
{ &vop_strategy_desc, spec_strategy }, /* strategy */
{ &vop_print_desc, spec_print }, /* print */
{ &vop_islocked_desc, genfs_islocked }, /* islocked */
{ &vop_pathconf_desc, spec_pathconf }, /* pathconf */
{ &vop_advlock_desc, spec_advlock }, /* advlock */
{ &vop_bwrite_desc, vn_bwrite }, /* bwrite */
{ &vop_getpages_desc, genfs_getpages }, /* getpages */
{ &vop_putpages_desc, genfs_putpages }, /* putpages */
{ NULL, NULL }
};
const struct vnodeopv_desc spec_vnodeop_opv_desc =
{ &spec_vnodeop_p, spec_vnodeop_entries };
static kauth_listener_t rawio_listener;
static struct kcondvar specfs_iocv;
/*
* Returns true if vnode is /dev/mem or /dev/kmem.
*/
bool
iskmemvp(struct vnode *vp)
{
return ((vp->v_type == VCHR) && iskmemdev(vp->v_rdev));
}
/*
* Returns true if dev is /dev/mem or /dev/kmem.
*/
int
iskmemdev(dev_t dev)
{
/* mem_no is emitted by config(8) to generated devsw.c */
extern const int mem_no;
/* minor 14 is /dev/io on i386 with COMPAT_10 */
return (major(dev) == mem_no && (minor(dev) < 2 || minor(dev) == 14));
}
static int
rawio_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie,
void *arg0, void *arg1, void *arg2, void *arg3)
{
int result;
result = KAUTH_RESULT_DEFER;
if ((action != KAUTH_DEVICE_RAWIO_SPEC) &&
(action != KAUTH_DEVICE_RAWIO_PASSTHRU))
return result;
/* Access is mandated by permissions. */
result = KAUTH_RESULT_ALLOW;
return result;
}
void
spec_init(void)
{
rawio_listener = kauth_listen_scope(KAUTH_SCOPE_DEVICE,
rawio_listener_cb, NULL);
cv_init(&specfs_iocv, "specio");
}
/*
* spec_io_enter(vp, &sn, &dev)
*
* Enter an operation that may not hold vp's vnode lock or an
* fstrans on vp's mount. Until spec_io_exit, the vnode will not
* be revoked.
*
* On success, set sn to the specnode pointer and dev to the dev_t
* number and return zero. Caller must later call spec_io_exit
* when done.
*
* On failure, return ENXIO -- the device has been revoked and no
* longer exists.
*/
static int
spec_io_enter(struct vnode *vp, struct specnode **snp, dev_t *devp)
{
dev_t dev;
struct specnode *sn;
unsigned iocnt;
int error = 0;
mutex_enter(vp->v_interlock);
/*
* Extract all the info we need from the vnode, unless the
* vnode has already been reclaimed. This can happen if the
* underlying device has been removed and all the device nodes
* for it have been revoked. The caller may not hold a vnode
* lock or fstrans to prevent this from happening before it has
* had an opportunity to notice the vnode is dead.
*/
if (vdead_check(vp, VDEAD_NOWAIT) != 0 ||
(sn = vp->v_specnode) == NULL ||
(dev = vp->v_rdev) == NODEV) {
error = ENXIO;
goto out;
}
/*
* Notify spec_close that we are doing an I/O operation which
* may not be not bracketed by fstrans(9) and thus is not
* blocked by vfs suspension.
*
* We could hold this reference with psref(9) instead, but we
* already have to take the interlock for vdead_check, so
* there's not much more cost here to another atomic operation.
*/
do {
iocnt = atomic_load_relaxed(&sn->sn_dev->sd_iocnt);
if (__predict_false(iocnt == UINT_MAX)) {
/*
* The I/O count is limited by the number of
* LWPs (which will never overflow this) --
* unless one driver uses another driver via
* specfs, which is rather unusual, but which
* could happen via pud(4) userspace drivers.
* We could use a 64-bit count, but can't use
* atomics for that on all platforms.
* (Probably better to switch to psref or
* localcount instead.)
*/
error = EBUSY;
goto out;
}
} while (atomic_cas_uint(&sn->sn_dev->sd_iocnt, iocnt, iocnt + 1)
!= iocnt);
/* Success! */
*snp = sn;
*devp = dev;
error = 0;
out: mutex_exit(vp->v_interlock);
return error;
}
/*
* spec_io_exit(vp, sn)
*
* Exit an operation entered with a successful spec_io_enter --
* allow concurrent spec_node_revoke to proceed. The argument sn
* must match the struct specnode pointer returned by spec_io_exit
* for vp.
*/
static void
spec_io_exit(struct vnode *vp, struct specnode *sn)
{
struct specdev *sd = sn->sn_dev;
unsigned iocnt;
KASSERT(vp->v_specnode == sn);
/*
* We are done. Notify spec_close if appropriate. The
* transition of 1 -> 0 must happen under device_lock so
* spec_close doesn't miss a wakeup.
*/
do {
iocnt = atomic_load_relaxed(&sd->sd_iocnt);
KASSERT(iocnt > 0);
if (iocnt == 1) {
mutex_enter(&device_lock);
if (atomic_dec_uint_nv(&sd->sd_iocnt) == 0)
cv_broadcast(&specfs_iocv);
mutex_exit(&device_lock);
break;
}
} while (atomic_cas_uint(&sd->sd_iocnt, iocnt, iocnt - 1) != iocnt);
}
/*
* spec_io_drain(sd)
*
* Wait for all existing spec_io_enter/exit sections to complete.
* Caller must ensure spec_io_enter will fail at this point.
*/
static void
spec_io_drain(struct specdev *sd)
{
/*
* I/O at the same time as closing is unlikely -- it often
* indicates an application bug.
*/
if (__predict_true(atomic_load_relaxed(&sd->sd_iocnt) == 0))
return;
mutex_enter(&device_lock);
while (atomic_load_relaxed(&sd->sd_iocnt) > 0)
cv_wait(&specfs_iocv, &device_lock);
mutex_exit(&device_lock);
}
/*
* Initialize a vnode that represents a device.
*/
void
spec_node_init(vnode_t *vp, dev_t rdev)
{
specnode_t *sn;
specdev_t *sd;
vnode_t *vp2;
vnode_t **vpp;
KASSERT(vp->v_type == VBLK || vp->v_type == VCHR);
KASSERT(vp->v_specnode == NULL);
/*
* Search the hash table for this device. If known, add a
* reference to the device structure. If not known, create
* a new entry to represent the device. In all cases add
* the vnode to the hash table.
*/
sn = kmem_alloc(sizeof(*sn), KM_SLEEP);
sd = kmem_alloc(sizeof(*sd), KM_SLEEP);
mutex_enter(&device_lock);
vpp = &specfs_hash[SPECHASH(rdev)];
for (vp2 = *vpp; vp2 != NULL; vp2 = vp2->v_specnext) {
KASSERT(vp2->v_specnode != NULL);
if (rdev == vp2->v_rdev && vp->v_type == vp2->v_type) {
break;
}
}
if (vp2 == NULL) {
/* No existing record, create a new one. */
sd->sd_mountpoint = NULL;
sd->sd_lockf = NULL;
sd->sd_refcnt = 1;
sd->sd_opencnt = 0;
sd->sd_bdevvp = NULL;
sd->sd_iocnt = 0;
sd->sd_opened = false;
sd->sd_closing = false;
sn->sn_dev = sd;
sd = NULL;
} else {
/* Use the existing record. */
sn->sn_dev = vp2->v_specnode->sn_dev;
sn->sn_dev->sd_refcnt++;
}
/* Insert vnode into the hash chain. */
sn->sn_opencnt = 0;
sn->sn_rdev = rdev;
sn->sn_gone = false;
vp->v_specnode = sn;
vp->v_specnext = *vpp;
*vpp = vp;
mutex_exit(&device_lock);
/* Free the record we allocated if unused. */
if (sd != NULL) {
kmem_free(sd, sizeof(*sd));
}
}
/*
* Lookup a vnode by device number and return it referenced.
*/
int
spec_node_lookup_by_dev(enum vtype type, dev_t dev, int flags, vnode_t **vpp)
{
int error;
vnode_t *vp;
top: mutex_enter(&device_lock);
for (vp = specfs_hash[SPECHASH(dev)]; vp; vp = vp->v_specnext) {
if (type == vp->v_type && dev == vp->v_rdev) {
mutex_enter(vp->v_interlock);
/* If clean or being cleaned, then ignore it. */
if (vdead_check(vp, VDEAD_NOWAIT) == 0)
break;
if ((flags & VDEAD_NOWAIT) == 0) {
mutex_exit(&device_lock);
/*
* It may be being revoked as we speak,
* and the caller wants to wait until
* all revocation has completed. Let
* vcache_vget wait for it to finish
* dying; as a side effect, vcache_vget
* releases vp->v_interlock. Note that
* vcache_vget cannot succeed at this
* point because vdead_check already
* failed.
*/
error = vcache_vget(vp);
KASSERT(error);
goto top;
}
mutex_exit(vp->v_interlock);
}
}
KASSERT(vp == NULL || mutex_owned(vp->v_interlock));
if (vp == NULL) {
mutex_exit(&device_lock);
return ENOENT;
}
/*
* If it is an opened block device return the opened vnode.
*/
if (type == VBLK && vp->v_specnode->sn_dev->sd_bdevvp != NULL) {
mutex_exit(vp->v_interlock);
vp = vp->v_specnode->sn_dev->sd_bdevvp;
mutex_enter(vp->v_interlock);
}
mutex_exit(&device_lock);
error = vcache_vget(vp);
if (error)
return error;
*vpp = vp;
return 0;
}
/*
* Lookup a vnode by file system mounted on and return it referenced.
*/
int
spec_node_lookup_by_mount(struct mount *mp, vnode_t **vpp)
{
int i, error;
vnode_t *vp, *vq;
mutex_enter(&device_lock);
for (i = 0, vq = NULL; i < SPECHSZ && vq == NULL; i++) {
for (vp = specfs_hash[i]; vp; vp = vp->v_specnext) {
if (vp->v_type != VBLK)
continue;
vq = vp->v_specnode->sn_dev->sd_bdevvp;
if (vq != NULL &&
vq->v_specnode->sn_dev->sd_mountpoint == mp)
break;
vq = NULL;
}
}
if (vq == NULL) {
mutex_exit(&device_lock);
return ENOENT;
}
mutex_enter(vq->v_interlock);
mutex_exit(&device_lock);
error = vcache_vget(vq);
if (error)
return error;
*vpp = vq;
return 0;
}
/*
* Get the file system mounted on this block device.
*
* XXX Caller should hold the vnode lock -- shared or exclusive -- so
* that this can't changed, and the vnode can't be revoked while we
* examine it. But not all callers do, and they're scattered through a
* lot of file systems, so we can't assert this yet.
*/
struct mount *
spec_node_getmountedfs(vnode_t *devvp)
{
struct mount *mp;
KASSERT(devvp->v_type == VBLK);
mp = devvp->v_specnode->sn_dev->sd_mountpoint;
return mp;
}
/*
* Set the file system mounted on this block device.
*
* XXX Caller should hold the vnode lock exclusively so this can't be
* changed or assumed by spec_node_getmountedfs while we change it, and
* the vnode can't be revoked while we handle it. But not all callers
* do, and they're scattered through a lot of file systems, so we can't
* assert this yet. Instead, for now, we'll take an I/O reference so
* at least the ioctl doesn't race with revoke/detach.
*
* If you do change this to assert an exclusive vnode lock, you must
* also do vdead_check before trying bdev_ioctl, because the vnode may
* have been revoked by the time the caller locked it, and this is
* _not_ a vop -- calls to spec_node_setmountedfs don't go through
* v_op, so revoking the vnode doesn't prevent further calls.
*
* XXX Caller should additionally have the vnode open, at least if mp
* is nonnull, but I'm not sure all callers do that -- need to audit.
* Currently udf closes the vnode before clearing the mount.
*/
void
spec_node_setmountedfs(vnode_t *devvp, struct mount *mp)
{
struct dkwedge_info dkw;
struct specnode *sn;
dev_t dev;
int error;
KASSERT(devvp->v_type == VBLK);
error = spec_io_enter(devvp, &sn, &dev);
if (error)
return;
KASSERT(sn->sn_dev->sd_mountpoint == NULL || mp == NULL);
sn->sn_dev->sd_mountpoint = mp;
if (mp == NULL)
goto out;
error = bdev_ioctl(dev, DIOCGWEDGEINFO, &dkw, FREAD, curlwp);
if (error)
goto out;
strlcpy(mp->mnt_stat.f_mntfromlabel, dkw.dkw_wname,
sizeof(mp->mnt_stat.f_mntfromlabel));
out: spec_io_exit(devvp, sn);
}
/*
* A vnode representing a special device is going away. Close
* the device if the vnode holds it open.
*/
void
spec_node_revoke(vnode_t *vp)
{
specnode_t *sn;
specdev_t *sd;
struct vnode **vpp;
KASSERT(VOP_ISLOCKED(vp) == LK_EXCLUSIVE);
sn = vp->v_specnode;
sd = sn->sn_dev;
KASSERT(vp->v_type == VBLK || vp->v_type == VCHR);
KASSERT(vp->v_specnode != NULL);
KASSERT(sn->sn_gone == false);
mutex_enter(&device_lock);
KASSERTMSG(sn->sn_opencnt <= sd->sd_opencnt,
"sn_opencnt=%u > sd_opencnt=%u",
sn->sn_opencnt, sd->sd_opencnt);
sn->sn_gone = true;
if (sn->sn_opencnt != 0) {
sd->sd_opencnt -= (sn->sn_opencnt - 1);
sn->sn_opencnt = 1;
mutex_exit(&device_lock);
VOP_CLOSE(vp, FNONBLOCK, NOCRED);
mutex_enter(&device_lock);
KASSERT(sn->sn_opencnt == 0);
}
/*
* We may have revoked the vnode in this thread while another
* thread was in the middle of spec_close, in the window when
* spec_close releases the vnode lock to call .d_close for the
* last close. In that case, wait for the concurrent
* spec_close to complete.
*/
while (sd->sd_closing)
cv_wait(&specfs_iocv, &device_lock);
/*
* Remove from the hash so lookups stop returning this
* specnode. We will dissociate it from the specdev -- and
* possibly free the specdev -- in spec_node_destroy.
*/
KASSERT(sn->sn_gone);
KASSERT(sn->sn_opencnt == 0);
for (vpp = &specfs_hash[SPECHASH(vp->v_rdev)];;
vpp = &(*vpp)->v_specnext) {
if (*vpp == vp) {
*vpp = vp->v_specnext;
vp->v_specnext = NULL;
break;
}
}
mutex_exit(&device_lock);
}
/*
* A vnode representing a special device is being recycled.
* Destroy the specfs component.
*/
void
spec_node_destroy(vnode_t *vp)
{
specnode_t *sn;
specdev_t *sd;
int refcnt;
sn = vp->v_specnode;
sd = sn->sn_dev;
KASSERT(vp->v_type == VBLK || vp->v_type == VCHR);
KASSERT(vp->v_specnode != NULL);
KASSERT(sn->sn_opencnt == 0);
mutex_enter(&device_lock);
sn = vp->v_specnode;
vp->v_specnode = NULL;
refcnt = sd->sd_refcnt--;
KASSERT(refcnt > 0);
mutex_exit(&device_lock);
/* If the device is no longer in use, destroy our record. */
if (refcnt == 1) {
KASSERT(sd->sd_iocnt == 0);
KASSERT(sd->sd_opencnt == 0);
KASSERT(sd->sd_bdevvp == NULL);
kmem_free(sd, sizeof(*sd));
}
kmem_free(sn, sizeof(*sn));
}
/*
* Trivial lookup routine that always fails.
*/
int
spec_lookup(void *v)
{
struct vop_lookup_v2_args /* {
struct vnode *a_dvp;
struct vnode **a_vpp;
struct componentname *a_cnp;
} */ *ap = v;
*ap->a_vpp = NULL;
return ENOTDIR;
}
typedef int (*spec_ioctl_t)(dev_t, u_long, void *, int, struct lwp *);
/*
* Open a special file.
*/
/* ARGSUSED */
int
spec_open(void *v)
{
struct vop_open_args /* {
struct vnode *a_vp;
int a_mode;
kauth_cred_t a_cred;
} */ *ap = v;
struct lwp *l = curlwp;
struct vnode *vp = ap->a_vp;
dev_t dev, dev1;
int error;
enum kauth_device_req req;
specnode_t *sn, *sn1;
specdev_t *sd;
spec_ioctl_t ioctl;
u_int gen = 0;
const char *name = NULL;
bool needclose = false;
struct partinfo pi;
KASSERT(VOP_ISLOCKED(vp) == LK_EXCLUSIVE);
KASSERTMSG(vp->v_type == VBLK || vp->v_type == VCHR, "type=%d",
vp->v_type);
dev = vp->v_rdev;
sn = vp->v_specnode;
sd = sn->sn_dev;
/*
* Don't allow open if fs is mounted -nodev.
*/
if (vp->v_mount && (vp->v_mount->mnt_flag & MNT_NODEV))
return ENXIO;
switch (ap->a_mode & (FREAD | FWRITE)) {
case FREAD | FWRITE:
req = KAUTH_REQ_DEVICE_RAWIO_SPEC_RW;
break;
case FWRITE:
req = KAUTH_REQ_DEVICE_RAWIO_SPEC_WRITE;
break;
default:
req = KAUTH_REQ_DEVICE_RAWIO_SPEC_READ;
break;
}
error = kauth_authorize_device_spec(ap->a_cred, req, vp);
if (error)
return error;
/*
* Acquire an open reference -- as long as we hold onto it, and
* the vnode isn't revoked, it can't be closed, and the vnode
* can't be revoked until we release the vnode lock.
*/
mutex_enter(&device_lock);
KASSERT(!sn->sn_gone);
switch (vp->v_type) {
case VCHR:
/*
* Character devices can accept opens from multiple
* vnodes. But first, wait for any close to finish.
* Wait under the vnode lock so we don't have to worry
* about the vnode being revoked while we wait.
*/
while (sd->sd_closing) {
error = cv_wait_sig(&specfs_iocv, &device_lock);
if (error)
break;
}
if (error)
break;
sd->sd_opencnt++;
sn->sn_opencnt++;
KASSERTMSG(sn->sn_opencnt <= sd->sd_opencnt,
"sn_opencnt=%u > sd_opencnt=%u",
sn->sn_opencnt, sd->sd_opencnt);
break;
case VBLK:
/*
* For block devices, permit only one open. The buffer
* cache cannot remain self-consistent with multiple
* vnodes holding a block device open.
*
* Treat zero opencnt with non-NULL mountpoint as open.
* This may happen after forced detach of a mounted device.
*
* Also treat sd_closing, meaning there is a concurrent
* close in progress, as still open.
*/
if (sd->sd_opencnt != 0 ||
sd->sd_mountpoint != NULL ||
sd->sd_closing) {
error = EBUSY;
break;
}
KASSERTMSG(sn->sn_opencnt == 0, "sn_opencnt=%u",
sn->sn_opencnt);
sn->sn_opencnt = 1;
sd->sd_opencnt = 1;
sd->sd_bdevvp = vp;
break;
default:
panic("invalid specfs vnode type: %d", vp->v_type);
}
mutex_exit(&device_lock);
if (error)
return error;
/*
* Set VV_ISTTY if this is a tty cdev.
*
* XXX This does the wrong thing if the module has to be
* autoloaded. We should maybe set this after autoloading
* modules and calling .d_open successfully, except (a) we need
* the vnode lock to touch it, and (b) once we acquire the
* vnode lock again, the vnode may have been revoked, and
* deadfs's dead_read needs VV_ISTTY to be already set in order
* to return the right answer. So this needs some additional
* synchronization to be made to work correctly with tty driver
* module autoload. For now, let's just hope it doesn't cause
* too much trouble for a tty from an autoloaded driver module
* to fail with EIO instead of returning EOF.
*/
if (vp->v_type == VCHR) {
if (cdev_type(dev) == D_TTY)
vp->v_vflag |= VV_ISTTY;
}
/*
* Because opening the device may block indefinitely, e.g. when
* opening a tty, and loading a module may cross into many
* other subsystems, we must not hold the vnode lock while
* calling .d_open, so release it now and reacquire it when
* done.
*
* Take an I/O reference so that any concurrent spec_close via
* spec_node_revoke will wait for us to finish calling .d_open.
* The vnode can't be dead at this point because we have it
* locked. Note that if revoked, the driver must interrupt
* .d_open before spec_close starts waiting for I/O to drain so
* this doesn't deadlock.
*/
VOP_UNLOCK(vp);
error = spec_io_enter(vp, &sn1, &dev1);
if (error) {
vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
return error;
}
KASSERT(sn1 == sn);
KASSERT(dev1 == dev);
/*
* Open the device. If .d_open returns ENXIO (device not
* configured), the driver may not be loaded, so try
* autoloading a module and then try .d_open again if anything
* got loaded.
*/
switch (vp->v_type) {
case VCHR:
do {
const struct cdevsw *cdev;
gen = module_gen;
error = cdev_open(dev, ap->a_mode, S_IFCHR, l);
if (error != ENXIO)
break;
/* Check if we already have a valid driver */
mutex_enter(&device_lock);
cdev = cdevsw_lookup(dev);
mutex_exit(&device_lock);
if (cdev != NULL)
break;
/* Get device name from devsw_conv array */
if ((name = cdevsw_getname(major(dev))) == NULL)
break;
/* Try to autoload device module */
(void)module_autoload(name, MODULE_CLASS_DRIVER);
} while (gen != module_gen);
break;
case VBLK:
do {
const struct bdevsw *bdev;
gen = module_gen;
error = bdev_open(dev, ap->a_mode, S_IFBLK, l);
if (error != ENXIO)
break;
/* Check if we already have a valid driver */
mutex_enter(&device_lock);
bdev = bdevsw_lookup(dev);
mutex_exit(&device_lock);
if (bdev != NULL)
break;
/* Get device name from devsw_conv array */
if ((name = bdevsw_getname(major(dev))) == NULL)
break;
/* Try to autoload device module */
(void)module_autoload(name, MODULE_CLASS_DRIVER);
} while (gen != module_gen);
break;
default:
__unreachable();
}
/*
* Release the I/O reference now that we have called .d_open,
* and reacquire the vnode lock. At this point, the device may
* have been revoked, so we must tread carefully. However, sn
* and sd remain valid pointers until we drop our reference.
*/
spec_io_exit(vp, sn);
vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
KASSERT(vp->v_specnode == sn);
/*
* If it has been revoked since we released the vnode lock and
* reacquired it, then spec_node_revoke has closed it, and we
* must fail with EBADF.
*
* Otherwise, if opening it failed, back out and release the
* open reference. If it was ever successfully opened and we
* got the last reference this way, it's now our job to close
* it. This might happen in the following scenario:
*
* Thread 1 Thread 2
* VOP_OPEN
* ...
* .d_open -> 0 (success)
* acquire vnode lock
* do stuff VOP_OPEN
* release vnode lock ...
* .d_open -> EBUSY
* VOP_CLOSE
* acquire vnode lock
* --sd_opencnt != 0
* => no .d_close
* release vnode lock
* acquire vnode lock
* --sd_opencnt == 0
*
* We can't resolve this by making spec_close wait for .d_open
* to complete before examining sd_opencnt, because .d_open can
* hang indefinitely, e.g. for a tty.
*/
mutex_enter(&device_lock);
if (sn->sn_gone) {
if (error == 0)
error = EBADF;
} else if (error == 0) {
/*
* Device has not been revoked, so our opencnt can't
* have gone away at this point -- transition to
* sn_gone=true happens before transition to
* sn_opencnt=0 in spec_node_revoke.
*/
KASSERT(sd->sd_opencnt);
KASSERT(sn->sn_opencnt);
KASSERTMSG(sn->sn_opencnt <= sd->sd_opencnt,
"sn_opencnt=%u > sd_opencnt=%u",
sn->sn_opencnt, sd->sd_opencnt);
KASSERT(!sd->sd_closing);
sd->sd_opened = true;
} else if (sd->sd_opencnt == 1 && sd->sd_opened) {
/*
* We're the last reference to a _previous_ open even
* though this one failed, so we have to close it.
* Don't decrement the reference count here --
* spec_close will do that.
*/
KASSERT(sn->sn_opencnt == 1);
needclose = true;
} else {
KASSERT(sd->sd_opencnt);
KASSERT(sn->sn_opencnt);
KASSERTMSG(sn->sn_opencnt <= sd->sd_opencnt,
"sn_opencnt=%u > sd_opencnt=%u",
sn->sn_opencnt, sd->sd_opencnt);
sd->sd_opencnt--;
sn->sn_opencnt--;
if (vp->v_type == VBLK)
sd->sd_bdevvp = NULL;
}
mutex_exit(&device_lock);
/*
* If this open failed, but the device was previously opened,
* and another thread concurrently closed the vnode while we
* were in the middle of reopening it, the other thread will
* see sd_opencnt > 0 and thus decide not to call .d_close --
* it is now our responsibility to do so.
*
* XXX The flags passed to VOP_CLOSE here are wrong, but
* drivers can't rely on FREAD|FWRITE anyway -- e.g., consider
* a device opened by thread 0 with O_READ, then opened by
* thread 1 with O_WRITE, then closed by thread 0, and finally
* closed by thread 1; the last .d_close call will have FWRITE
* but not FREAD. We should just eliminate the FREAD/FWRITE
* parameter to .d_close altogether.
*/
if (needclose) {
KASSERT(error);
VOP_CLOSE(vp, FNONBLOCK, NOCRED);
}
/* If anything went wrong, we're done. */
if (error)
return error;
/*
* For disk devices, automagically set the vnode size to the
* partition size, if we can. This applies to block devices
* and character devices alike -- every block device must have
* a corresponding character device. And if the module is
* loaded it will remain loaded until we're done here (it is
* forbidden to devsw_detach until closed). So it is safe to
* query cdev_type unconditionally here.
*/
if (cdev_type(dev) == D_DISK) {
ioctl = vp->v_type == VCHR ? cdev_ioctl : bdev_ioctl;
if ((*ioctl)(dev, DIOCGPARTINFO, &pi, FREAD, curlwp) == 0)
uvm_vnp_setsize(vp,
(voff_t)pi.pi_secsize * pi.pi_size);
}
/* Success! */
return 0;
}
/*
* Vnode op for read
*/
/* ARGSUSED */
int
spec_read(void *v)
{
struct vop_read_args /* {
struct vnode *a_vp;
struct uio *a_uio;
int a_ioflag;
kauth_cred_t a_cred;
} */ *ap = v;
struct vnode *vp = ap->a_vp;
struct uio *uio = ap->a_uio;
struct lwp *l = curlwp;
struct specnode *sn;
dev_t dev;
struct buf *bp;
daddr_t bn;
int bsize, bscale;
struct partinfo pi;
int n, on;
int error = 0;
int i, nra;
daddr_t lastbn, *rablks;
int *rasizes;
int nrablks, ratogo;
KASSERT(uio->uio_rw == UIO_READ);
KASSERTMSG((VMSPACE_IS_KERNEL_P(uio->uio_vmspace) ||
uio->uio_vmspace == curproc->p_vmspace),
"vmspace belongs to neither kernel nor curproc");
if (uio->uio_resid == 0)
return 0;
switch (vp->v_type) {
case VCHR:
/*
* Release the lock while we sleep -- possibly
* indefinitely, if this is, e.g., a tty -- in
* cdev_read, so we don't hold up everything else that
* might want access to the vnode.
*
* But before we issue the read, take an I/O reference
* to the specnode so close will know when we're done
* reading. Note that the moment we release the lock,
* the vnode's identity may change; hence spec_io_enter
* may fail, and the caller may have a dead vnode on
* their hands, if the file system on which vp lived
* has been unmounted.
*/
VOP_UNLOCK(vp);
error = spec_io_enter(vp, &sn, &dev);
if (error)
goto out;
error = cdev_read(dev, uio, ap->a_ioflag);
spec_io_exit(vp, sn);
out: /* XXX What if the caller held an exclusive lock? */
vn_lock(vp, LK_SHARED | LK_RETRY);
return error;
case VBLK:
KASSERT(vp == vp->v_specnode->sn_dev->sd_bdevvp);
if (uio->uio_offset < 0)
return EINVAL;
if (bdev_ioctl(vp->v_rdev, DIOCGPARTINFO, &pi, FREAD, l) == 0)
bsize = imin(imax(pi.pi_bsize, DEV_BSIZE), MAXBSIZE);
else
bsize = BLKDEV_IOSIZE;
bscale = bsize >> DEV_BSHIFT;
nra = uimax(16 * MAXPHYS / bsize - 1, 511);
rablks = kmem_alloc(nra * sizeof(*rablks), KM_SLEEP);
rasizes = kmem_alloc(nra * sizeof(*rasizes), KM_SLEEP);
lastbn = ((uio->uio_offset + uio->uio_resid - 1) >> DEV_BSHIFT)
&~ (bscale - 1);
nrablks = ratogo = 0;
do {
bn = (uio->uio_offset >> DEV_BSHIFT) &~ (bscale - 1);
on = uio->uio_offset % bsize;
n = uimin((unsigned)(bsize - on), uio->uio_resid);
if (ratogo == 0) {
nrablks = uimin((lastbn - bn) / bscale, nra);
ratogo = nrablks;
for (i = 0; i < nrablks; ++i) {
rablks[i] = bn + (i+1) * bscale;
rasizes[i] = bsize;
}
error = breadn(vp, bn, bsize,
rablks, rasizes, nrablks,
0, &bp);
} else {
if (ratogo > 0)
--ratogo;
error = bread(vp, bn, bsize, 0, &bp);
}
if (error)
break;
n = uimin(n, bsize - bp->b_resid);
error = uiomove((char *)bp->b_data + on, n, uio);
brelse(bp, 0);
} while (error == 0 && uio->uio_resid > 0 && n != 0);
kmem_free(rablks, nra * sizeof(*rablks));
kmem_free(rasizes, nra * sizeof(*rasizes));
return error;
default:
panic("spec_read type");
}
/* NOTREACHED */
}
/*
* Vnode op for write
*/
/* ARGSUSED */
int
spec_write(void *v)
{
struct vop_write_args /* {
struct vnode *a_vp;
struct uio *a_uio;
int a_ioflag;
kauth_cred_t a_cred;
} */ *ap = v;
struct vnode *vp = ap->a_vp;
struct uio *uio = ap->a_uio;
struct lwp *l = curlwp;
struct specnode *sn;
dev_t dev;
struct buf *bp;
daddr_t bn;
int bsize, bscale;
struct partinfo pi;
int n, on;
int error = 0;
KASSERT(uio->uio_rw == UIO_WRITE);
KASSERTMSG((VMSPACE_IS_KERNEL_P(uio->uio_vmspace) ||
uio->uio_vmspace == curproc->p_vmspace),
"vmspace belongs to neither kernel nor curproc");
switch (vp->v_type) {
case VCHR:
/*
* Release the lock while we sleep -- possibly
* indefinitely, if this is, e.g., a tty -- in
* cdev_write, so we don't hold up everything else that
* might want access to the vnode.
*
* But before we issue the write, take an I/O reference
* to the specnode so close will know when we're done
* writing. Note that the moment we release the lock,
* the vnode's identity may change; hence spec_io_enter
* may fail, and the caller may have a dead vnode on
* their hands, if the file system on which vp lived
* has been unmounted.
*/
VOP_UNLOCK(vp);
error = spec_io_enter(vp, &sn, &dev);
if (error)
goto out;
error = cdev_write(dev, uio, ap->a_ioflag);
spec_io_exit(vp, sn);
out: vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
return error;
case VBLK:
KASSERT(vp == vp->v_specnode->sn_dev->sd_bdevvp);
if (uio->uio_resid == 0)
return 0;
if (uio->uio_offset < 0)
return EINVAL;
if (bdev_ioctl(vp->v_rdev, DIOCGPARTINFO, &pi, FREAD, l) == 0)
bsize = imin(imax(pi.pi_bsize, DEV_BSIZE), MAXBSIZE);
else
bsize = BLKDEV_IOSIZE;
bscale = bsize >> DEV_BSHIFT;
do {
bn = (uio->uio_offset >> DEV_BSHIFT) &~ (bscale - 1);
on = uio->uio_offset % bsize;
n = uimin((unsigned)(bsize - on), uio->uio_resid);
if (n == bsize)
bp = getblk(vp, bn, bsize, 0, 0);
else
error = bread(vp, bn, bsize, B_MODIFY, &bp);
if (error) {
return error;
}
n = uimin(n, bsize - bp->b_resid);
error = uiomove((char *)bp->b_data + on, n, uio);
if (error)
brelse(bp, 0);
else {
if (n + on == bsize)
bawrite(bp);
else
bdwrite(bp);
error = bp->b_error;
}
} while (error == 0 && uio->uio_resid > 0 && n != 0);
return error;
default:
panic("spec_write type");
}
/* NOTREACHED */
}
/*
* fdiscard, which on disk devices becomes TRIM.
*/
int
spec_fdiscard(void *v)
{
struct vop_fdiscard_args /* {
struct vnode *a_vp;
off_t a_pos;
off_t a_len;
} */ *ap = v;
struct vnode *vp = ap->a_vp;
dev_t dev;
KASSERT(VOP_ISLOCKED(vp) == LK_EXCLUSIVE);
dev = vp->v_rdev;
switch (vp->v_type) {
case VCHR:
#if 0 /* This is not stored for character devices. */
KASSERT(vp == vp->v_specnode->sn_dev->sd_cdevvp);
#endif
return cdev_discard(dev, ap->a_pos, ap->a_len);
case VBLK:
KASSERT(vp == vp->v_specnode->sn_dev->sd_bdevvp);
return bdev_discard(dev, ap->a_pos, ap->a_len);
default:
panic("spec_fdiscard: not a device\n");
}
}
/*
* Device ioctl operation.
*/
/* ARGSUSED */
int
spec_ioctl(void *v)
{
struct vop_ioctl_args /* {
struct vnode *a_vp;
u_long a_command;
void *a_data;
int a_fflag;
kauth_cred_t a_cred;
} */ *ap = v;
struct vnode *vp = ap->a_vp;
struct specnode *sn;
dev_t dev;
int error;
error = spec_io_enter(vp, &sn, &dev);
if (error)
return error;
switch (vp->v_type) {
case VCHR:
error = cdev_ioctl(dev, ap->a_command, ap->a_data,
ap->a_fflag, curlwp);
break;
case VBLK:
KASSERT(vp == vp->v_specnode->sn_dev->sd_bdevvp);
error = bdev_ioctl(dev, ap->a_command, ap->a_data,
ap->a_fflag, curlwp);
break;
default:
panic("spec_ioctl");
/* NOTREACHED */
}
spec_io_exit(vp, sn);
return error;
}
/* ARGSUSED */
int
spec_poll(void *v)
{
struct vop_poll_args /* {
struct vnode *a_vp;
int a_events;
} */ *ap = v;
struct vnode *vp = ap->a_vp;
struct specnode *sn;
dev_t dev;
int revents;
if (spec_io_enter(vp, &sn, &dev) != 0)
return POLLERR;
switch (vp->v_type) {
case VCHR:
revents = cdev_poll(dev, ap->a_events, curlwp);
break;
default:
revents = genfs_poll(v);
break;
}
spec_io_exit(vp, sn);
return revents;
}
/* ARGSUSED */
int
spec_kqfilter(void *v)
{
struct vop_kqfilter_args /* {
struct vnode *a_vp;
struct proc *a_kn;
} */ *ap = v;
struct vnode *vp = ap->a_vp;
struct specnode *sn;
dev_t dev;
int error;
error = spec_io_enter(vp, &sn, &dev);
if (error)
return error;
switch (vp->v_type) {
case VCHR:
error = cdev_kqfilter(dev, ap->a_kn);
break;
default:
/*
* Block devices don't support kqfilter, and refuse it
* for any other files (like those vflush()ed) too.
*/
error = EOPNOTSUPP;
break;
}
spec_io_exit(vp, sn);
return error;
}
/*
* Allow mapping of only D_DISK. This is called only for VBLK.
*/
int
spec_mmap(void *v)
{
struct vop_mmap_args /* {
struct vnode *a_vp;
vm_prot_t a_prot;
kauth_cred_t a_cred;
} */ *ap = v;
struct vnode *vp = ap->a_vp;
struct specnode *sn;
dev_t dev;
int error;
KASSERT(vp->v_type == VBLK);
error = spec_io_enter(vp, &sn, &dev);
if (error)
return error;
error = bdev_type(dev) == D_DISK ? 0 : EINVAL;
spec_io_exit(vp, sn);
return 0;
}
/*
* Synch buffers associated with a block device
*/
/* ARGSUSED */
int
spec_fsync(void *v)
{
struct vop_fsync_args /* {
struct vnode *a_vp;
kauth_cred_t a_cred;
int a_flags;
off_t offlo;
off_t offhi;
} */ *ap = v;
struct vnode *vp = ap->a_vp;
struct mount *mp;
int error;
if (vp->v_type == VBLK) {
if ((mp = spec_node_getmountedfs(vp)) != NULL) {
error = VFS_FSYNC(mp, vp, ap->a_flags);
if (error != EOPNOTSUPP)
return error;
}
return vflushbuf(vp, ap->a_flags);
}
return 0;
}
/*
* Just call the device strategy routine
*/
int
spec_strategy(void *v)
{
struct vop_strategy_args /* {
struct vnode *a_vp;
struct buf *a_bp;
} */ *ap = v;
struct vnode *vp = ap->a_vp;
struct buf *bp = ap->a_bp;
struct specnode *sn = NULL;
dev_t dev;
int error;
error = spec_io_enter(vp, &sn, &dev);
if (error)
goto out;
bp->b_dev = dev;
if (!(bp->b_flags & B_READ)) {
#ifdef DIAGNOSTIC
if (bp->b_vp && bp->b_vp->v_type == VBLK) {
struct mount *mp = spec_node_getmountedfs(bp->b_vp);
if (mp && (mp->mnt_flag & MNT_RDONLY)) {
printf("%s blk %"PRId64" written while ro!\n",
mp->mnt_stat.f_mntonname, bp->b_blkno);
#ifdef DDB
db_stacktrace();
#endif
}
}
#endif /* DIAGNOSTIC */
error = fscow_run(bp, false);
if (error)
goto out;
}
bdev_strategy(bp);
error = 0;
out: if (sn)
spec_io_exit(vp, sn);
if (error) {
bp->b_error = error;
bp->b_resid = bp->b_bcount;
biodone(bp);
}
return error;
}
int
spec_inactive(void *v)
{
struct vop_inactive_v2_args /* {
struct vnode *a_vp;
struct bool *a_recycle;
} */ *ap = v;
KASSERT(ap->a_vp->v_mount == dead_rootmount);
*ap->a_recycle = true;
return 0;
}
int
spec_reclaim(void *v)
{
struct vop_reclaim_v2_args /* {
struct vnode *a_vp;
} */ *ap = v;
struct vnode *vp = ap->a_vp;
KASSERT(vp->v_specnode->sn_opencnt == 0);
VOP_UNLOCK(vp);
KASSERT(vp->v_mount == dead_rootmount);
return 0;
}
/*
* This is a noop, simply returning what one has been given.
*/
int
spec_bmap(void *v)
{
struct vop_bmap_args /* {
struct vnode *a_vp;
daddr_t a_bn;
struct vnode **a_vpp;
daddr_t *a_bnp;
int *a_runp;
} */ *ap = v;
if (ap->a_vpp != NULL)
*ap->a_vpp = ap->a_vp;
if (ap->a_bnp != NULL)
*ap->a_bnp = ap->a_bn;
if (ap->a_runp != NULL)
*ap->a_runp = (MAXBSIZE >> DEV_BSHIFT) - 1;
return 0;
}
/*
* Device close routine
*/
/* ARGSUSED */
int
spec_close(void *v)
{
struct vop_close_args /* {
struct vnode *a_vp;
int a_fflag;
kauth_cred_t a_cred;
} */ *ap = v;
struct vnode *vp = ap->a_vp;
struct session *sess;
dev_t dev;
int flags = ap->a_fflag;
int mode, error, count;
specnode_t *sn;
specdev_t *sd;
KASSERT(VOP_ISLOCKED(vp) == LK_EXCLUSIVE);
mutex_enter(vp->v_interlock);
sn = vp->v_specnode;
dev = vp->v_rdev;
sd = sn->sn_dev;
/*
* If we're going away soon, make this non-blocking.
* Also ensures that we won't wedge in vn_lock below.
*/
if (vdead_check(vp, VDEAD_NOWAIT) != 0)
flags |= FNONBLOCK;
mutex_exit(vp->v_interlock);
switch (vp->v_type) {
case VCHR:
/*
* Hack: a tty device that is a controlling terminal
* has a reference from the session structure. We
* cannot easily tell that a character device is a
* controlling terminal, unless it is the closing
* process' controlling terminal. In that case, if the
* open count is 1 release the reference from the
* session. Also, remove the link from the tty back to
* the session and pgrp.
*
* XXX V. fishy.
*/
mutex_enter(&proc_lock);
sess = curlwp->l_proc->p_session;
if (sn->sn_opencnt == 1 && vp == sess->s_ttyvp) {
mutex_spin_enter(&tty_lock);
sess->s_ttyvp = NULL;
if (sess->s_ttyp->t_session != NULL) {
sess->s_ttyp->t_pgrp = NULL;
sess->s_ttyp->t_session = NULL;
mutex_spin_exit(&tty_lock);
/* Releases proc_lock. */
proc_sessrele(sess);
} else {
mutex_spin_exit(&tty_lock);
if (sess->s_ttyp->t_pgrp != NULL)
panic("spec_close: spurious pgrp ref");
mutex_exit(&proc_lock);
}
vrele(vp);
} else
mutex_exit(&proc_lock);
/*
* If the vnode is locked, then we are in the midst
* of forcably closing the device, otherwise we only
* close on last reference.
*/
mode = S_IFCHR;
break;
case VBLK:
KASSERT(vp == vp->v_specnode->sn_dev->sd_bdevvp);
/*
* On last close of a block device (that isn't mounted)
* we must invalidate any in core blocks, so that
* we can, for instance, change floppy disks.
*/
error = vinvalbuf(vp, V_SAVE, ap->a_cred, curlwp, 0, 0);
if (error)
return error;
/*
* We do not want to really close the device if it
* is still in use unless we are trying to close it
* forcibly. Since every use (buffer, vnode, swap, cmap)
* holds a reference to the vnode, and because we mark
* any other vnodes that alias this device, when the
* sum of the reference counts on all the aliased
* vnodes descends to one, we are on last close.
*/
mode = S_IFBLK;
break;
default:
panic("spec_close: not special");
}
/*
* Decrement the open reference count of this node and the
* device. For block devices, the open reference count must be
* 1 at this point. If the device's open reference count goes
* to zero, we're the last one out so get the lights.
*
* We may find --sd->sd_opencnt gives zero, and yet
* sd->sd_opened is false. This happens if the vnode is
* revoked at the same time as it is being opened, which can
* happen when opening a tty blocks indefinitely. In that
* case, we still must call close -- it is the job of close to
* interrupt the open. Either way, the device will be no
* longer opened, so we have to clear sd->sd_opened; subsequent
* opens will have responsibility for issuing close.
*
* This has the side effect that the sequence of opens might
* happen out of order -- we might end up doing open, open,
* close, close, instead of open, close, open, close. This is
* unavoidable with the current devsw API, where open is
* allowed to block and close must be able to run concurrently
* to interrupt it. It is the driver's responsibility to
* ensure that close is idempotent so that this works. Drivers
* requiring per-open state and exact 1:1 correspondence
* between open and close can use fd_clone.
*/
mutex_enter(&device_lock);
KASSERT(sn->sn_opencnt);
KASSERT(sd->sd_opencnt);
KASSERTMSG(sn->sn_opencnt <= sd->sd_opencnt,
"sn_opencnt=%u > sd_opencnt=%u",
sn->sn_opencnt, sd->sd_opencnt);
sn->sn_opencnt--;
count = --sd->sd_opencnt;
if (vp->v_type == VBLK) {
KASSERTMSG(count == 0, "block device with %u opens",
count + 1);
sd->sd_bdevvp = NULL;
}
if (count == 0) {
KASSERTMSG(sn->sn_opencnt == 0, "sn_opencnt=%u",
sn->sn_opencnt);
KASSERT(!sd->sd_closing);
sd->sd_opened = false;
sd->sd_closing = true;
}
mutex_exit(&device_lock);
if (count != 0)
return 0;
/*
* If we're able to block, release the vnode lock & reacquire. We
* might end up sleeping for someone else who wants our queues. They
* won't get them if we hold the vnode locked.
*/
if (!(flags & FNONBLOCK))
VOP_UNLOCK(vp);
/*
* If we can cancel all outstanding I/O, then wait for it to
* drain before we call .d_close. Drivers that split up
* .d_cancel and .d_close this way need not have any internal
* mechanism for waiting in .d_close for I/O to drain.
*/
if (vp->v_type == VBLK)
error = bdev_cancel(dev, flags, mode, curlwp);
else
error = cdev_cancel(dev, flags, mode, curlwp);
if (error == 0)
spec_io_drain(sd);
else
KASSERTMSG(error == ENODEV, "cancel dev=0x%lx failed with %d",
(unsigned long)dev, error);
if (vp->v_type == VBLK)
error = bdev_close(dev, flags, mode, curlwp);
else
error = cdev_close(dev, flags, mode, curlwp);
/*
* Wait for all other devsw operations to drain. After this
* point, no bdev/cdev_* can be active for this specdev.
*/
spec_io_drain(sd);
/*
* Wake any spec_open calls waiting for close to finish -- do
* this before reacquiring the vnode lock, because spec_open
* holds the vnode lock while waiting, so doing this after
* reacquiring the lock would deadlock.
*/
mutex_enter(&device_lock);
KASSERT(!sd->sd_opened);
KASSERT(sd->sd_closing);
sd->sd_closing = false;
cv_broadcast(&specfs_iocv);
mutex_exit(&device_lock);
if (!(flags & FNONBLOCK))
vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
return error;
}
/*
* Print out the contents of a special device vnode.
*/
int
spec_print(void *v)
{
struct vop_print_args /* {
struct vnode *a_vp;
} */ *ap = v;
printf("dev %llu, %llu\n", (unsigned long long)major(ap->a_vp->v_rdev),
(unsigned long long)minor(ap->a_vp->v_rdev));
return 0;
}
/*
* Return POSIX pathconf information applicable to special devices.
*/
int
spec_pathconf(void *v)
{
struct vop_pathconf_args /* {
struct vnode *a_vp;
int a_name;
register_t *a_retval;
} */ *ap = v;
switch (ap->a_name) {
case _PC_LINK_MAX:
*ap->a_retval = LINK_MAX;
return 0;
case _PC_MAX_CANON:
*ap->a_retval = MAX_CANON;
return 0;
case _PC_MAX_INPUT:
*ap->a_retval = MAX_INPUT;
return 0;
case _PC_PIPE_BUF:
*ap->a_retval = PIPE_BUF;
return 0;
case _PC_CHOWN_RESTRICTED:
*ap->a_retval = 1;
return 0;
case _PC_VDISABLE:
*ap->a_retval = _POSIX_VDISABLE;
return 0;
case _PC_SYNC_IO:
*ap->a_retval = 1;
return 0;
default:
return genfs_pathconf(ap);
}
/* NOTREACHED */
}
/*
* Advisory record locking support.
*/
int
spec_advlock(void *v)
{
struct vop_advlock_args /* {
struct vnode *a_vp;
void *a_id;
int a_op;
struct flock *a_fl;
int a_flags;
} */ *ap = v;
struct vnode *vp = ap->a_vp;
return lf_advlock(ap, &vp->v_speclockf, (off_t)0);
}
|