1 /* $NetBSD: ffs_vfsops.c,v 1.384 2024/12/30 09:03:07 hannken Exp $ */
2
3 /*-
4 * Copyright (c) 2008, 2009 The NetBSD Foundation, Inc.
5 * All rights reserved.
6 *
7 * This code is derived from software contributed to The NetBSD Foundation
8 * by Wasabi Systems, Inc, and by Andrew Doran.
9 *
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
12 * are met:
13 * 1. Redistributions of source code must retain the above copyright
14 * notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 * notice, this list of conditions and the following disclaimer in the
17 * documentation and/or other materials provided with the distribution.
18 *
19 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
20 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
21 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
22 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
23 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
24 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
25 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
26 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
27 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
28 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
29 * POSSIBILITY OF SUCH DAMAGE.
30 */
31
32 /*
33 * Copyright (c) 1989, 1991, 1993, 1994
34 * The Regents of the University of California. All rights reserved.
35 *
36 * Redistribution and use in source and binary forms, with or without
37 * modification, are permitted provided that the following conditions
38 * are met:
39 * 1. Redistributions of source code must retain the above copyright
40 * notice, this list of conditions and the following disclaimer.
41 * 2. Redistributions in binary form must reproduce the above copyright
42 * notice, this list of conditions and the following disclaimer in the
43 * documentation and/or other materials provided with the distribution.
44 * 3. Neither the name of the University nor the names of its contributors
45 * may be used to endorse or promote products derived from this software
46 * without specific prior written permission.
47 *
48 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
49 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
50 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
51 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
52 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
53 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
54 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
55 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
56 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
57 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
58 * SUCH DAMAGE.
59 *
60 * @(#)ffs_vfsops.c 8.31 (Berkeley) 5/20/95
61 */
62
63 #include <sys/cdefs.h>
64 __KERNEL_RCSID(0, "$NetBSD: ffs_vfsops.c,v 1.384 2024/12/30 09:03:07 hannken Exp $");
65
66 #if defined(_KERNEL_OPT)
67 #include "opt_ffs.h"
68 #include "opt_quota.h"
69 #include "opt_wapbl.h"
70 #endif
71
72 #include <sys/param.h>
73 #include <sys/systm.h>
74 #include <sys/namei.h>
75 #include <sys/proc.h>
76 #include <sys/kernel.h>
77 #include <sys/vnode.h>
78 #include <sys/fstrans.h>
79 #include <sys/socket.h>
80 #include <sys/mount.h>
81 #include <sys/buf.h>
82 #include <sys/device.h>
83 #include <sys/disk.h>
84 #include <sys/file.h>
85 #include <sys/disklabel.h>
86 #include <sys/ioctl.h>
87 #include <sys/errno.h>
88 #include <sys/kmem.h>
89 #include <sys/pool.h>
90 #include <sys/lock.h>
91 #include <sys/sysctl.h>
92 #include <sys/conf.h>
93 #include <sys/kauth.h>
94 #include <sys/wapbl.h>
95 #include <sys/module.h>
96
97 #include <miscfs/genfs/genfs.h>
98 #include <miscfs/specfs/specdev.h>
99
100 #include <ufs/ufs/quota.h>
101 #include <ufs/ufs/ufsmount.h>
102 #include <ufs/ufs/inode.h>
103 #include <ufs/ufs/dir.h>
104 #include <ufs/ufs/ufs_extern.h>
105 #include <ufs/ufs/ufs_bswap.h>
106 #include <ufs/ufs/ufs_wapbl.h>
107
108 #include <ufs/ffs/fs.h>
109 #include <ufs/ffs/ffs_extern.h>
110
111 #ifdef WAPBL
112 MODULE(MODULE_CLASS_VFS, ffs, "ufs,wapbl");
113 #else
114 MODULE(MODULE_CLASS_VFS, ffs, "ufs");
115 #endif
116
117 static int ffs_vfs_fsync(vnode_t *, int);
118 static int ffs_superblock_validate(struct fs *);
119 static int ffs_is_appleufs(struct vnode *, struct fs *);
120
121 static int ffs_init_vnode(struct ufsmount *, struct vnode *, ino_t);
122 static void ffs_deinit_vnode(struct ufsmount *, struct vnode *);
123
124 static kauth_listener_t ffs_snapshot_listener;
125
126 /* how many times ffs_init() was called */
127 int ffs_initcount = 0;
128
129 #ifdef DEBUG_FFS_MOUNT
130 #define DPRINTF(_fmt, args...) printf("%s: " _fmt "\n", __func__, ##args)
131 #else
132 #define DPRINTF(_fmt, args...) do {} while (/*CONSTCOND*/0)
133 #endif
134
135 extern const struct vnodeopv_desc ffs_vnodeop_opv_desc;
136 extern const struct vnodeopv_desc ffs_specop_opv_desc;
137 extern const struct vnodeopv_desc ffs_fifoop_opv_desc;
138
139 const struct vnodeopv_desc * const ffs_vnodeopv_descs[] = {
140 &ffs_vnodeop_opv_desc,
141 &ffs_specop_opv_desc,
142 &ffs_fifoop_opv_desc,
143 NULL,
144 };
145
146 struct vfsops ffs_vfsops = {
147 .vfs_name = MOUNT_FFS,
148 .vfs_min_mount_data = sizeof (struct ufs_args),
149 .vfs_mount = ffs_mount,
150 .vfs_start = ufs_start,
151 .vfs_unmount = ffs_unmount,
152 .vfs_root = ufs_root,
153 .vfs_quotactl = ufs_quotactl,
154 .vfs_statvfs = ffs_statvfs,
155 .vfs_sync = ffs_sync,
156 .vfs_vget = ufs_vget,
157 .vfs_loadvnode = ffs_loadvnode,
158 .vfs_newvnode = ffs_newvnode,
159 .vfs_fhtovp = ffs_fhtovp,
160 .vfs_vptofh = ffs_vptofh,
161 .vfs_init = ffs_init,
162 .vfs_reinit = ffs_reinit,
163 .vfs_done = ffs_done,
164 .vfs_mountroot = ffs_mountroot,
165 .vfs_snapshot = ffs_snapshot,
166 .vfs_extattrctl = ffs_extattrctl,
167 .vfs_suspendctl = genfs_suspendctl,
168 .vfs_renamelock_enter = genfs_renamelock_enter,
169 .vfs_renamelock_exit = genfs_renamelock_exit,
170 .vfs_fsync = ffs_vfs_fsync,
171 .vfs_opv_descs = ffs_vnodeopv_descs
172 };
173
174 static const struct genfs_ops ffs_genfsops = {
175 .gop_size = ffs_gop_size,
176 .gop_alloc = ufs_gop_alloc,
177 .gop_write = genfs_gop_write,
178 .gop_markupdate = ufs_gop_markupdate,
179 .gop_putrange = genfs_gop_putrange,
180 };
181
182 static const struct ufs_ops ffs_ufsops = {
183 .uo_itimes = ffs_itimes,
184 .uo_update = ffs_update,
185 .uo_truncate = ffs_truncate,
186 .uo_balloc = ffs_balloc,
187 .uo_snapgone = ffs_snapgone,
188 .uo_bufrd = ffs_bufrd,
189 .uo_bufwr = ffs_bufwr,
190 };
191
192 static int
ffs_checkrange(struct mount * mp,ino_t ino)193 ffs_checkrange(struct mount *mp, ino_t ino)
194 {
195 struct fs *fs = VFSTOUFS(mp)->um_fs;
196
197 if (ino < UFS_ROOTINO || ino >= fs->fs_ncg * fs->fs_ipg) {
198 DPRINTF("out of range %" PRIu64 "\n", ino);
199 return ESTALE;
200 }
201
202 /*
203 * Need to check if inode is initialized because ffsv2 does
204 * lazy initialization and we can get here from nfs_fhtovp
205 */
206 if (fs->fs_magic != FS_UFS2_MAGIC)
207 return 0;
208
209 struct buf *bp;
210 int cg = ino_to_cg(fs, ino);
211 struct ufsmount *ump = VFSTOUFS(mp);
212
213 int error = bread(ump->um_devvp, FFS_FSBTODB(fs, cgtod(fs, cg)),
214 (int)fs->fs_cgsize, B_MODIFY, &bp);
215 if (error) {
216 DPRINTF("error %d reading cg %d ino %" PRIu64 "\n",
217 error, cg, ino);
218 return error;
219 }
220
221 const int needswap = UFS_FSNEEDSWAP(fs);
222
223 struct cg *cgp = (struct cg *)bp->b_data;
224 if (!cg_chkmagic(cgp, needswap)) {
225 brelse(bp, 0);
226 DPRINTF("bad cylinder group magic cg %d ino %" PRIu64 "\n",
227 cg, ino);
228 return ESTALE;
229 }
230
231 int32_t initediblk = ufs_rw32(cgp->cg_initediblk, needswap);
232 brelse(bp, 0);
233
234 if (cg * fs->fs_ipg + initediblk < ino) {
235 DPRINTF("cg=%d fs->fs_ipg=%d initediblk=%d ino=%" PRIu64 "\n",
236 cg, fs->fs_ipg, initediblk, ino);
237 return ESTALE;
238 }
239 return 0;
240 }
241
242 static int
ffs_snapshot_cb(kauth_cred_t cred,kauth_action_t action,void * cookie,void * arg0,void * arg1,void * arg2,void * arg3)243 ffs_snapshot_cb(kauth_cred_t cred, kauth_action_t action, void *cookie,
244 void *arg0, void *arg1, void *arg2, void *arg3)
245 {
246 vnode_t *vp = arg2;
247 int result = KAUTH_RESULT_DEFER;
248
249 if (action != KAUTH_SYSTEM_FS_SNAPSHOT)
250 return result;
251
252 if (VTOI(vp)->i_uid == kauth_cred_geteuid(cred))
253 result = KAUTH_RESULT_ALLOW;
254
255 return result;
256 }
257
258 SYSCTL_SETUP(ffs_sysctl_setup, "ffs sysctls")
259 {
260 #ifdef UFS_EXTATTR
261 extern int ufs_extattr_autocreate;
262 #endif
263 extern int ffs_log_changeopt;
264
265 sysctl_createv(clog, 0, NULL, NULL,
266 CTLFLAG_PERMANENT,
267 CTLTYPE_NODE, "ffs",
268 SYSCTL_DESCR("Berkeley Fast File System"),
269 NULL, 0, NULL, 0,
270 CTL_VFS, 1, CTL_EOL);
271 /*
272 * @@@ should we even bother with these first three?
273 */
274 sysctl_createv(clog, 0, NULL, NULL,
275 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
276 CTLTYPE_INT, "doclusterread", NULL,
277 sysctl_notavail, 0, NULL, 0,
278 CTL_VFS, 1, FFS_CLUSTERREAD, CTL_EOL);
279 sysctl_createv(clog, 0, NULL, NULL,
280 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
281 CTLTYPE_INT, "doclusterwrite", NULL,
282 sysctl_notavail, 0, NULL, 0,
283 CTL_VFS, 1, FFS_CLUSTERWRITE, CTL_EOL);
284 sysctl_createv(clog, 0, NULL, NULL,
285 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
286 CTLTYPE_INT, "doreallocblks", NULL,
287 sysctl_notavail, 0, NULL, 0,
288 CTL_VFS, 1, FFS_REALLOCBLKS, CTL_EOL);
289 #if 0
290 sysctl_createv(clog, 0, NULL, NULL,
291 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
292 CTLTYPE_INT, "doasyncfree",
293 SYSCTL_DESCR("Release dirty blocks asynchronously"),
294 NULL, 0, &doasyncfree, 0,
295 CTL_VFS, 1, FFS_ASYNCFREE, CTL_EOL);
296 #endif
297 sysctl_createv(clog, 0, NULL, NULL,
298 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
299 CTLTYPE_INT, "log_changeopt",
300 SYSCTL_DESCR("Log changes in optimization strategy"),
301 NULL, 0, &ffs_log_changeopt, 0,
302 CTL_VFS, 1, FFS_LOG_CHANGEOPT, CTL_EOL);
303 #ifdef UFS_EXTATTR
304 sysctl_createv(clog, 0, NULL, NULL,
305 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
306 CTLTYPE_INT, "extattr_autocreate",
307 SYSCTL_DESCR("Size of attribute for "
308 "backing file autocreation"),
309 NULL, 0, &ufs_extattr_autocreate, 0,
310 CTL_VFS, 1, FFS_EXTATTR_AUTOCREATE, CTL_EOL);
311
312 #endif /* UFS_EXTATTR */
313 }
314
315 static int
ffs_modcmd(modcmd_t cmd,void * arg)316 ffs_modcmd(modcmd_t cmd, void *arg)
317 {
318 int error;
319
320 #if 0
321 extern int doasyncfree;
322 #endif
323
324 switch (cmd) {
325 case MODULE_CMD_INIT:
326 error = vfs_attach(&ffs_vfsops);
327 if (error != 0)
328 break;
329
330 ffs_snapshot_listener = kauth_listen_scope(KAUTH_SCOPE_SYSTEM,
331 ffs_snapshot_cb, NULL);
332 if (ffs_snapshot_listener == NULL)
333 printf("ffs_modcmd: can't listen on system scope.\n");
334
335 break;
336 case MODULE_CMD_FINI:
337 error = vfs_detach(&ffs_vfsops);
338 if (error != 0)
339 break;
340 if (ffs_snapshot_listener != NULL)
341 kauth_unlisten_scope(ffs_snapshot_listener);
342 break;
343 default:
344 error = ENOTTY;
345 break;
346 }
347
348 return (error);
349 }
350
351 pool_cache_t ffs_inode_cache;
352 pool_cache_t ffs_dinode1_cache;
353 pool_cache_t ffs_dinode2_cache;
354
355 static void ffs_oldfscompat_read(struct fs *, struct ufsmount *, daddr_t);
356 static void ffs_oldfscompat_write(struct fs *, struct ufsmount *);
357
358 /*
359 * Called by main() when ffs is going to be mounted as root.
360 */
361
362 int
ffs_mountroot(void)363 ffs_mountroot(void)
364 {
365 struct fs *fs;
366 struct mount *mp;
367 struct lwp *l = curlwp; /* XXX */
368 struct ufsmount *ump;
369 int error;
370
371 if (device_class(root_device) != DV_DISK)
372 return (ENODEV);
373
374 if ((error = vfs_rootmountalloc(MOUNT_FFS, "root_device", &mp))) {
375 vrele(rootvp);
376 return (error);
377 }
378
379 /*
380 * We always need to be able to mount the root file system.
381 */
382 mp->mnt_flag |= MNT_FORCE;
383 if ((error = ffs_mountfs(rootvp, mp, l)) != 0) {
384 vfs_unbusy(mp);
385 vfs_rele(mp);
386 return (error);
387 }
388 mp->mnt_flag &= ~MNT_FORCE;
389 mountlist_append(mp);
390 ump = VFSTOUFS(mp);
391 fs = ump->um_fs;
392 memset(fs->fs_fsmnt, 0, sizeof(fs->fs_fsmnt));
393 (void)copystr(mp->mnt_stat.f_mntonname, fs->fs_fsmnt, MNAMELEN - 1, 0);
394 (void)ffs_statvfs(mp, &mp->mnt_stat);
395 vfs_unbusy(mp);
396 setrootfstime((time_t)fs->fs_time);
397 return (0);
398 }
399
400 static int
ffs_acls(struct mount * mp,int fs_flags)401 ffs_acls(struct mount *mp, int fs_flags)
402 {
403 struct ufsmount *ump;
404
405 ump = VFSTOUFS(mp);
406 if (ump->um_fstype == UFS2 && (ump->um_flags & UFS_EA) == 0 &&
407 ((mp->mnt_flag & (MNT_POSIX1EACLS | MNT_NFS4ACLS)) != 0 ||
408 (fs_flags & (FS_POSIX1EACLS | FS_NFS4ACLS)) != 0)) {
409 printf("%s: ACLs requested but not supported by this fs\n",
410 mp->mnt_stat.f_mntonname);
411 return EINVAL;
412 }
413
414 if ((fs_flags & FS_POSIX1EACLS) != 0) {
415 #ifdef UFS_ACL
416 if (mp->mnt_flag & MNT_NFS4ACLS)
417 printf("WARNING: %s: POSIX.1e ACLs flag on fs conflicts "
418 "with \"nfsv4acls\" mount option; option ignored\n",
419 mp->mnt_stat.f_mntonname);
420 mp->mnt_flag &= ~MNT_NFS4ACLS;
421 mp->mnt_flag |= MNT_POSIX1EACLS;
422 #else
423 printf("WARNING: %s: POSIX.1e ACLs flag on fs but no "
424 "ACLs support\n", mp->mnt_stat.f_mntonname);
425 #endif
426 }
427 if ((fs_flags & FS_NFS4ACLS) != 0) {
428 #ifdef UFS_ACL
429 if (mp->mnt_flag & MNT_POSIX1EACLS)
430 printf("WARNING: %s: NFSv4 ACLs flag on fs conflicts "
431 "with \"posix1eacls\" mount option; option ignored\n",
432 mp->mnt_stat.f_mntonname);
433 mp->mnt_flag &= ~MNT_POSIX1EACLS;
434 mp->mnt_flag |= MNT_NFS4ACLS;
435
436 #else
437 printf("WARNING: %s: NFSv4 ACLs flag on fs but no "
438 "ACLs support\n", mp->mnt_stat.f_mntonname);
439 #endif
440 }
441 if ((mp->mnt_flag & (MNT_NFS4ACLS | MNT_POSIX1EACLS))
442 == (MNT_NFS4ACLS | MNT_POSIX1EACLS))
443 {
444 printf("%s: \"posix1eacls\" and \"nfsv4acls\" options "
445 "are mutually exclusive\n",
446 mp->mnt_stat.f_mntonname);
447 return EINVAL;
448 }
449
450 if (mp->mnt_flag & (MNT_NFS4ACLS | MNT_POSIX1EACLS))
451 mp->mnt_iflag &= ~(IMNT_SHRLOOKUP|IMNT_NCLOOKUP);
452 else
453 mp->mnt_iflag |= IMNT_SHRLOOKUP|IMNT_NCLOOKUP;
454 return 0;
455 }
456
457 /*
458 * VFS Operations.
459 *
460 * mount system call
461 */
462 int
ffs_mount(struct mount * mp,const char * path,void * data,size_t * data_len)463 ffs_mount(struct mount *mp, const char *path, void *data, size_t *data_len)
464 {
465 struct lwp *l = curlwp;
466 struct vnode *devvp = NULL;
467 struct ufs_args *args = data;
468 struct ufsmount *ump = NULL;
469 struct fs *fs;
470 int error = 0, flags, update;
471 mode_t accessmode;
472
473 if (args == NULL) {
474 DPRINTF("NULL args");
475 return EINVAL;
476 }
477 if (*data_len < sizeof(*args)) {
478 DPRINTF("bad size args %zu != %zu", *data_len, sizeof(*args));
479 return EINVAL;
480 }
481
482 ump = VFSTOUFS(mp);
483 if ((mp->mnt_flag & (MNT_GETARGS|MNT_UPDATE)) && ump == NULL) {
484 DPRINTF("no ump");
485 return EIO;
486 }
487
488 if (mp->mnt_flag & MNT_GETARGS) {
489 args->fspec = NULL;
490 *data_len = sizeof *args;
491 return 0;
492 }
493
494 update = mp->mnt_flag & MNT_UPDATE;
495
496 /* Check arguments */
497 if (args->fspec == NULL) {
498 if (!update) {
499 /* New mounts must have a filename for the device */
500 DPRINTF("no filename for mount");
501 return EINVAL;
502 }
503 } else {
504 /*
505 * Look up the name and verify that it's sane.
506 */
507 error = namei_simple_user(args->fspec,
508 NSM_FOLLOW_NOEMULROOT, &devvp);
509 if (error != 0) {
510 DPRINTF("namei_simple_user returned %d", error);
511 return error;
512 }
513
514 /*
515 * Be sure this is a valid block device
516 */
517 if (devvp->v_type != VBLK) {
518 DPRINTF("non block device %d", devvp->v_type);
519 error = ENOTBLK;
520 goto fail;
521 }
522
523 if (bdevsw_lookup(devvp->v_rdev) == NULL) {
524 DPRINTF("can't find block device 0x%jx",
525 devvp->v_rdev);
526 error = ENXIO;
527 goto fail;
528 }
529
530 if (update) {
531 /*
532 * Be sure we're still naming the same device
533 * used for our initial mount
534 */
535 if (devvp != ump->um_devvp &&
536 devvp->v_rdev != ump->um_devvp->v_rdev) {
537 DPRINTF("wrong device 0x%jx != 0x%jx",
538 (uintmax_t)devvp->v_rdev,
539 (uintmax_t)ump->um_devvp->v_rdev);
540 error = EINVAL;
541 goto fail;
542 }
543 vrele(devvp);
544 devvp = NULL;
545 }
546 }
547
548 if (devvp == NULL) {
549 devvp = ump->um_devvp;
550 vref(devvp);
551 }
552
553 /*
554 * If mount by non-root, then verify that user has necessary
555 * permissions on the device.
556 *
557 * Permission to update a mount is checked higher, so here we presume
558 * updating the mount is okay (for example, as far as securelevel goes)
559 * which leaves us with the normal check.
560 */
561 accessmode = VREAD;
562 if (update ? (mp->mnt_iflag & IMNT_WANTRDWR) != 0 :
563 (mp->mnt_flag & MNT_RDONLY) == 0)
564 accessmode |= VWRITE;
565 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
566 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT,
567 KAUTH_REQ_SYSTEM_MOUNT_DEVICE, mp, devvp, KAUTH_ARG(accessmode));
568 VOP_UNLOCK(devvp);
569 if (error) {
570 DPRINTF("kauth returned %d", error);
571 goto fail;
572 }
573
574 #ifdef WAPBL
575 /* WAPBL can only be enabled on a r/w mount. */
576 if (((mp->mnt_flag & MNT_RDONLY) && !(mp->mnt_iflag & IMNT_WANTRDWR)) ||
577 (mp->mnt_iflag & IMNT_WANTRDONLY)) {
578 mp->mnt_flag &= ~MNT_LOG;
579 }
580 #else /* !WAPBL */
581 mp->mnt_flag &= ~MNT_LOG;
582 #endif /* !WAPBL */
583
584 error = set_statvfs_info(path, UIO_USERSPACE, args->fspec,
585 UIO_USERSPACE, mp->mnt_op->vfs_name, mp, l);
586 if (error)
587 goto fail;
588
589 if (!update) {
590 int xflags;
591
592 if (mp->mnt_flag & MNT_RDONLY)
593 xflags = FREAD;
594 else
595 xflags = FREAD | FWRITE;
596 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
597 error = VOP_OPEN(devvp, xflags, FSCRED);
598 VOP_UNLOCK(devvp);
599 if (error) {
600 DPRINTF("VOP_OPEN returned %d", error);
601 goto fail;
602 }
603 /* Need fstrans_start() for assertion in ufs_strategy(). */
604 if ((mp->mnt_flag & MNT_RDONLY) == 0)
605 fstrans_start(mp);
606 error = ffs_mountfs(devvp, mp, l);
607 if ((mp->mnt_flag & MNT_RDONLY) == 0)
608 fstrans_done(mp);
609 if (error) {
610 DPRINTF("ffs_mountfs returned %d", error);
611 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
612 (void)VOP_CLOSE(devvp, xflags, NOCRED);
613 VOP_UNLOCK(devvp);
614 goto fail;
615 }
616
617 ump = VFSTOUFS(mp);
618 fs = ump->um_fs;
619 } else {
620 /*
621 * Update the mount. The file system is suspended.
622 */
623 KASSERT(fstrans_is_owner(mp));
624
625 /*
626 * The initial mount got a reference on this
627 * device, so drop the one obtained via
628 * namei(), above.
629 */
630 vrele(devvp);
631
632 ump = VFSTOUFS(mp);
633 fs = ump->um_fs;
634 if (fs->fs_ronly == 0 && (mp->mnt_iflag & IMNT_WANTRDONLY)) {
635 /*
636 * Changing from r/w to r/o
637 */
638 flags = WRITECLOSE;
639 if (mp->mnt_flag & MNT_FORCE)
640 flags |= FORCECLOSE;
641 error = ffs_flushfiles(mp, flags, l);
642 if (error)
643 return error;
644
645 error = UFS_WAPBL_BEGIN(mp);
646 if (error) {
647 DPRINTF("wapbl %d", error);
648 return error;
649 }
650
651 if (ffs_cgupdate(ump, MNT_WAIT) == 0 &&
652 fs->fs_clean & FS_WASCLEAN) {
653 if (mp->mnt_flag & MNT_SOFTDEP)
654 fs->fs_flags &= ~FS_DOSOFTDEP;
655 fs->fs_clean = FS_ISCLEAN;
656 (void) ffs_sbupdate(ump, MNT_WAIT);
657 }
658
659 UFS_WAPBL_END(mp);
660 }
661
662 #ifdef WAPBL
663 if ((mp->mnt_flag & MNT_LOG) == 0) {
664 error = ffs_wapbl_stop(mp, mp->mnt_flag & MNT_FORCE);
665 if (error) {
666 DPRINTF("ffs_wapbl_stop returned %d", error);
667 return error;
668 }
669 }
670 #endif /* WAPBL */
671
672 if (fs->fs_ronly == 0 && (mp->mnt_iflag & IMNT_WANTRDONLY)) {
673 /*
674 * Finish change from r/w to r/o
675 */
676 fs->fs_ronly = 1;
677 fs->fs_fmod = 0;
678 }
679
680 error = ffs_acls(mp, fs->fs_flags);
681 if (error)
682 return error;
683 if (mp->mnt_flag & MNT_RELOAD) {
684 error = ffs_reload(mp, l->l_cred, l);
685 if (error) {
686 DPRINTF("ffs_reload returned %d", error);
687 return error;
688 }
689 }
690
691 if (fs->fs_ronly && (mp->mnt_iflag & IMNT_WANTRDWR)) {
692 /*
693 * Changing from read-only to read/write
694 */
695 #ifndef QUOTA2
696 if (fs->fs_flags & FS_DOQUOTA2) {
697 ump->um_flags |= UFS_QUOTA2;
698 uprintf("%s: options QUOTA2 not enabled%s\n",
699 mp->mnt_stat.f_mntonname,
700 (mp->mnt_flag & MNT_FORCE) ? "" :
701 ", not mounting");
702 DPRINTF("ffs_quota2 %d", EINVAL);
703 return EINVAL;
704 }
705 #endif
706 fs->fs_ronly = 0;
707 fs->fs_clean =
708 fs->fs_clean == FS_ISCLEAN ? FS_WASCLEAN : 0;
709 fs->fs_fmod = 1;
710 #ifdef WAPBL
711 if (fs->fs_flags & FS_DOWAPBL) {
712 const char *nm = mp->mnt_stat.f_mntonname;
713 if (!mp->mnt_wapbl_replay) {
714 printf("%s: log corrupted;"
715 " replay cancelled\n", nm);
716 return EFTYPE;
717 }
718 printf("%s: replaying log to disk\n", nm);
719 error = wapbl_replay_write(mp->mnt_wapbl_replay,
720 devvp);
721 if (error) {
722 DPRINTF("%s: wapbl_replay_write %d",
723 nm, error);
724 return error;
725 }
726 wapbl_replay_stop(mp->mnt_wapbl_replay);
727 fs->fs_clean = FS_WASCLEAN;
728 }
729 #endif /* WAPBL */
730 if (fs->fs_snapinum[0] != 0)
731 ffs_snapshot_mount(mp);
732 }
733
734 #ifdef WAPBL
735 error = ffs_wapbl_start(mp);
736 if (error) {
737 DPRINTF("ffs_wapbl_start returned %d", error);
738 return error;
739 }
740 #endif /* WAPBL */
741
742 #ifdef QUOTA2
743 if (!fs->fs_ronly) {
744 error = ffs_quota2_mount(mp);
745 if (error) {
746 DPRINTF("ffs_quota2_mount returned %d", error);
747 return error;
748 }
749 }
750 #endif
751
752 if ((mp->mnt_flag & MNT_DISCARD) && !(ump->um_discarddata))
753 ump->um_discarddata = ffs_discard_init(devvp, fs);
754
755 if (args->fspec == NULL)
756 return 0;
757 }
758
759 (void)strncpy(fs->fs_fsmnt, mp->mnt_stat.f_mntonname,
760 sizeof(fs->fs_fsmnt));
761
762 fs->fs_flags &= ~FS_DOSOFTDEP;
763
764 if ((fs->fs_ronly && (fs->fs_clean & FS_ISCLEAN) == 0) ||
765 (!fs->fs_ronly && (fs->fs_clean & FS_WASCLEAN) == 0)) {
766 printf("%s: file system not clean (fs_clean=%#x); "
767 "please fsck(8)\n", mp->mnt_stat.f_mntfromname,
768 fs->fs_clean);
769 }
770
771 if (UFS_WAPBL_BEGIN(mp) == 0) {
772 mutex_enter(&ump->um_lock);
773 if (fs->fs_fmod != 0) {
774 KASSERT(!fs->fs_ronly);
775
776 if (fs->fs_clean & FS_WASCLEAN)
777 fs->fs_time = time_second;
778 fs->fs_fmod = 0;
779 mutex_exit(&ump->um_lock);
780 (void) ffs_cgupdate(ump, MNT_WAIT);
781 } else {
782 mutex_exit(&ump->um_lock);
783 }
784 UFS_WAPBL_END(mp);
785 }
786 if ((mp->mnt_flag & MNT_SOFTDEP) != 0) {
787 printf("%s: `-o softdep' is no longer supported, "
788 "consider `-o log'\n", mp->mnt_stat.f_mntfromname);
789 mp->mnt_flag &= ~MNT_SOFTDEP;
790 }
791
792 return (error);
793
794 fail:
795 vrele(devvp);
796 return (error);
797 }
798
799 /*
800 * Reload all incore data for a filesystem (used after running fsck on
801 * the root filesystem and finding things to fix). The filesystem must
802 * be mounted read-only.
803 *
804 * Things to do to update the mount:
805 * 1) invalidate all cached meta-data.
806 * 2) re-read superblock from disk.
807 * 3) re-read summary information from disk.
808 * 4) invalidate all inactive vnodes.
809 * 5) invalidate all cached file data.
810 * 6) re-read inode data for all active vnodes.
811 */
812 int
ffs_reload(struct mount * mp,kauth_cred_t cred,struct lwp * l)813 ffs_reload(struct mount *mp, kauth_cred_t cred, struct lwp *l)
814 {
815 struct vnode *vp, *devvp;
816 struct inode *ip;
817 void *space;
818 struct buf *bp;
819 struct fs *fs, *newfs;
820 int i, bsize, blks, error;
821 int32_t *lp, fs_sbsize;
822 struct ufsmount *ump;
823 daddr_t sblockloc;
824 struct vnode_iterator *marker;
825
826 if ((mp->mnt_flag & MNT_RDONLY) == 0)
827 return (EINVAL);
828
829 ump = VFSTOUFS(mp);
830
831 /*
832 * Step 1: invalidate all cached meta-data.
833 */
834 devvp = ump->um_devvp;
835 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
836 error = vinvalbuf(devvp, 0, cred, l, 0, 0);
837 VOP_UNLOCK(devvp);
838 if (error)
839 panic("%s: dirty1", __func__);
840
841 /*
842 * Step 2: re-read superblock from disk. XXX: We don't handle
843 * possibility that superblock moved. Which implies that we don't
844 * want its size to change either.
845 */
846 fs = ump->um_fs;
847 fs_sbsize = fs->fs_sbsize;
848 error = bread(devvp, fs->fs_sblockloc / DEV_BSIZE, fs_sbsize,
849 0, &bp);
850 if (error)
851 return (error);
852 newfs = kmem_alloc(fs_sbsize, KM_SLEEP);
853 memcpy(newfs, bp->b_data, fs_sbsize);
854
855 #ifdef FFS_EI
856 if (ump->um_flags & UFS_NEEDSWAP) {
857 ffs_sb_swap((struct fs *)bp->b_data, newfs);
858 newfs->fs_flags |= FS_SWAPPED;
859 } else
860 #endif
861 newfs->fs_flags &= ~FS_SWAPPED;
862
863 brelse(bp, 0);
864
865 /* Allow converting from UFS2 to UFS2EA but not vice versa. */
866 if (newfs->fs_magic == FS_UFS2EA_MAGIC) {
867 ump->um_flags |= UFS_EA;
868 newfs->fs_magic = FS_UFS2_MAGIC;
869 } else {
870 if ((ump->um_flags & UFS_EA) != 0)
871 return EINVAL;
872 }
873
874 if ((newfs->fs_magic != FS_UFS1_MAGIC) &&
875 (newfs->fs_magic != FS_UFS2_MAGIC)) {
876 kmem_free(newfs, fs_sbsize);
877 return (EIO); /* XXX needs translation */
878 }
879 if (!ffs_superblock_validate(newfs)) {
880 kmem_free(newfs, fs_sbsize);
881 return (EINVAL);
882 }
883
884 /*
885 * The current implementation doesn't handle the possibility that
886 * these values may have changed.
887 */
888 if ((newfs->fs_sbsize != fs_sbsize) ||
889 (newfs->fs_cssize != fs->fs_cssize) ||
890 (newfs->fs_contigsumsize != fs->fs_contigsumsize) ||
891 (newfs->fs_ncg != fs->fs_ncg)) {
892 kmem_free(newfs, fs_sbsize);
893 return (EINVAL);
894 }
895
896 /* Store off old fs_sblockloc for fs_oldfscompat_read. */
897 sblockloc = fs->fs_sblockloc;
898 /*
899 * Copy pointer fields back into superblock before copying in XXX
900 * new superblock. These should really be in the ufsmount. XXX
901 * Note that important parameters (eg fs_ncg) are unchanged.
902 */
903 newfs->fs_csp = fs->fs_csp;
904 newfs->fs_maxcluster = fs->fs_maxcluster;
905 newfs->fs_contigdirs = fs->fs_contigdirs;
906 newfs->fs_ronly = fs->fs_ronly;
907 newfs->fs_active = fs->fs_active;
908 memcpy(fs, newfs, (u_int)fs_sbsize);
909 kmem_free(newfs, fs_sbsize);
910
911 /*
912 * Recheck for Apple UFS filesystem.
913 */
914 ump->um_flags &= ~UFS_ISAPPLEUFS;
915 if (ffs_is_appleufs(devvp, fs)) {
916 #ifdef APPLE_UFS
917 ump->um_flags |= UFS_ISAPPLEUFS;
918 #else
919 DPRINTF("AppleUFS not supported");
920 return (EIO); /* XXX: really? */
921 #endif
922 }
923
924 if (UFS_MPISAPPLEUFS(ump)) {
925 /* see comment about NeXT below */
926 ump->um_maxsymlinklen = APPLEUFS_MAXSYMLINKLEN;
927 ump->um_dirblksiz = APPLEUFS_DIRBLKSIZ;
928 mp->mnt_iflag |= IMNT_DTYPE;
929 } else {
930 ump->um_maxsymlinklen = fs->fs_maxsymlinklen;
931 ump->um_dirblksiz = UFS_DIRBLKSIZ;
932 if (ump->um_maxsymlinklen > 0)
933 mp->mnt_iflag |= IMNT_DTYPE;
934 else
935 mp->mnt_iflag &= ~IMNT_DTYPE;
936 }
937 ffs_oldfscompat_read(fs, ump, sblockloc);
938
939 mutex_enter(&ump->um_lock);
940 ump->um_maxfilesize = fs->fs_maxfilesize;
941 if (fs->fs_flags & ~(FS_KNOWN_FLAGS | FS_INTERNAL)) {
942 uprintf("%s: unknown ufs flags: 0x%08"PRIx32"%s\n",
943 mp->mnt_stat.f_mntonname, fs->fs_flags,
944 (mp->mnt_flag & MNT_FORCE) ? "" : ", not mounting");
945 if ((mp->mnt_flag & MNT_FORCE) == 0) {
946 mutex_exit(&ump->um_lock);
947 return (EINVAL);
948 }
949 }
950
951 if (fs->fs_pendingblocks != 0 || fs->fs_pendinginodes != 0) {
952 fs->fs_pendingblocks = 0;
953 fs->fs_pendinginodes = 0;
954 }
955 mutex_exit(&ump->um_lock);
956
957 ffs_statvfs(mp, &mp->mnt_stat);
958 /*
959 * Step 3: re-read summary information from disk.
960 */
961 blks = howmany(fs->fs_cssize, fs->fs_fsize);
962 space = fs->fs_csp;
963 for (i = 0; i < blks; i += fs->fs_frag) {
964 bsize = fs->fs_bsize;
965 if (i + fs->fs_frag > blks)
966 bsize = (blks - i) * fs->fs_fsize;
967 error = bread(devvp, FFS_FSBTODB(fs, fs->fs_csaddr + i), bsize,
968 0, &bp);
969 if (error) {
970 return (error);
971 }
972 #ifdef FFS_EI
973 if (UFS_FSNEEDSWAP(fs))
974 ffs_csum_swap((struct csum *)bp->b_data,
975 (struct csum *)space, bsize);
976 else
977 #endif
978 memcpy(space, bp->b_data, (size_t)bsize);
979 space = (char *)space + bsize;
980 brelse(bp, 0);
981 }
982 /*
983 * We no longer know anything about clusters per cylinder group.
984 */
985 if (fs->fs_contigsumsize > 0) {
986 lp = fs->fs_maxcluster;
987 for (i = 0; i < fs->fs_ncg; i++)
988 *lp++ = fs->fs_contigsumsize;
989 }
990
991 vfs_vnode_iterator_init(mp, &marker);
992 while ((vp = vfs_vnode_iterator_next(marker, NULL, NULL))) {
993 /*
994 * Step 4: invalidate all inactive vnodes.
995 */
996 if (vrecycle(vp))
997 continue;
998 /*
999 * Step 5: invalidate all cached file data.
1000 */
1001 if (vn_lock(vp, LK_EXCLUSIVE)) {
1002 vrele(vp);
1003 continue;
1004 }
1005 if (vinvalbuf(vp, 0, cred, l, 0, 0))
1006 panic("%s: dirty2", __func__);
1007 /*
1008 * Step 6: re-read inode data for all active vnodes.
1009 */
1010 ip = VTOI(vp);
1011 error = bread(devvp, FFS_FSBTODB(fs, ino_to_fsba(fs, ip->i_number)),
1012 (int)fs->fs_bsize, 0, &bp);
1013 if (error) {
1014 vput(vp);
1015 break;
1016 }
1017 ffs_load_inode(bp, ip, fs, ip->i_number);
1018 brelse(bp, 0);
1019 vput(vp);
1020 }
1021 vfs_vnode_iterator_destroy(marker);
1022 return (error);
1023 }
1024
1025 /*
1026 * Possible superblock locations ordered from most to least likely.
1027 */
1028 static const int sblock_try[] = SBLOCKSEARCH;
1029
1030
1031 static int
ffs_superblock_validate(struct fs * fs)1032 ffs_superblock_validate(struct fs *fs)
1033 {
1034 int32_t i, fs_bshift = 0, fs_fshift = 0, fs_fragshift = 0, fs_frag;
1035 int32_t fs_inopb;
1036
1037 /* Check the superblock size */
1038 if (fs->fs_sbsize > SBLOCKSIZE || fs->fs_sbsize < sizeof(struct fs))
1039 return 0;
1040
1041 /* Check the file system blocksize */
1042 if (fs->fs_bsize > MAXBSIZE || fs->fs_bsize < MINBSIZE)
1043 return 0;
1044 if (!powerof2(fs->fs_bsize))
1045 return 0;
1046
1047 /* Check the size of frag blocks */
1048 if (!powerof2(fs->fs_fsize))
1049 return 0;
1050 if (fs->fs_fsize == 0)
1051 return 0;
1052
1053 /*
1054 * XXX: these values are just zero-checked to prevent obvious
1055 * bugs. We need more strict checks.
1056 */
1057 if (fs->fs_size == 0 && fs->fs_old_size == 0)
1058 return 0;
1059 if (fs->fs_cssize == 0)
1060 return 0;
1061 if (fs->fs_ipg == 0)
1062 return 0;
1063 if (fs->fs_fpg == 0)
1064 return 0;
1065 if (fs->fs_ncg == 0)
1066 return 0;
1067 if (fs->fs_maxbpg == 0)
1068 return 0;
1069
1070 /* Check the number of inodes per block */
1071 if (fs->fs_magic == FS_UFS1_MAGIC)
1072 fs_inopb = fs->fs_bsize / sizeof(struct ufs1_dinode);
1073 else /* fs->fs_magic == FS_UFS2_MAGIC */
1074 fs_inopb = fs->fs_bsize / sizeof(struct ufs2_dinode);
1075 if (fs->fs_inopb != fs_inopb)
1076 return 0;
1077
1078 /* Block size cannot be smaller than fragment size */
1079 if (fs->fs_bsize < fs->fs_fsize)
1080 return 0;
1081
1082 /* Compute fs_bshift and ensure it is consistent */
1083 for (i = fs->fs_bsize; i > 1; i >>= 1)
1084 fs_bshift++;
1085 if (fs->fs_bshift != fs_bshift)
1086 return 0;
1087
1088 /* Compute fs_fshift and ensure it is consistent */
1089 for (i = fs->fs_fsize; i > 1; i >>= 1)
1090 fs_fshift++;
1091 if (fs->fs_fshift != fs_fshift)
1092 return 0;
1093
1094 /* Compute fs_fragshift and ensure it is consistent */
1095 for (i = fs->fs_frag; i > 1; i >>= 1)
1096 fs_fragshift++;
1097 if (fs->fs_fragshift != fs_fragshift)
1098 return 0;
1099
1100 /* Check the masks */
1101 if (fs->fs_bmask != ~(fs->fs_bsize - 1))
1102 return 0;
1103 if (fs->fs_fmask != ~(fs->fs_fsize - 1))
1104 return 0;
1105
1106 /*
1107 * Now that the shifts and masks are sanitized, we can use the ffs_ API.
1108 */
1109
1110 /* Check the number of frag blocks */
1111 if ((fs_frag = ffs_numfrags(fs, fs->fs_bsize)) > MAXFRAG)
1112 return 0;
1113 if (fs->fs_frag != fs_frag)
1114 return 0;
1115
1116 /* Check the size of cylinder groups */
1117 if ((fs->fs_cgsize < sizeof(struct cg)) ||
1118 (fs->fs_cgsize > fs->fs_bsize))
1119 return 0;
1120
1121 return 1;
1122 }
1123
1124 static int
ffs_is_appleufs(struct vnode * devvp,struct fs * fs)1125 ffs_is_appleufs(struct vnode *devvp, struct fs *fs)
1126 {
1127 struct dkwedge_info dkw;
1128 int ret = 0;
1129
1130 /*
1131 * First check to see if this is tagged as an Apple UFS filesystem
1132 * in the disklabel.
1133 */
1134 if (getdiskinfo(devvp, &dkw) == 0 &&
1135 strcmp(dkw.dkw_ptype, DKW_PTYPE_APPLEUFS) == 0)
1136 ret = 1;
1137 #ifdef APPLE_UFS
1138 else {
1139 struct appleufslabel *applefs;
1140 struct buf *bp;
1141 daddr_t blkno = APPLEUFS_LABEL_OFFSET / DEV_BSIZE;
1142 int error;
1143
1144 /*
1145 * Manually look for an Apple UFS label, and if a valid one
1146 * is found, then treat it like an Apple UFS filesystem anyway.
1147 */
1148 error = bread(devvp, blkno, APPLEUFS_LABEL_SIZE, 0, &bp);
1149 if (error) {
1150 DPRINTF("bread@0x%jx returned %d", (intmax_t)blkno, error);
1151 return 0;
1152 }
1153 applefs = (struct appleufslabel *)bp->b_data;
1154 error = ffs_appleufs_validate(fs->fs_fsmnt, applefs, NULL);
1155 if (error == 0)
1156 ret = 1;
1157 brelse(bp, 0);
1158 }
1159 #endif
1160
1161 return ret;
1162 }
1163
1164 /*
1165 * Common code for mount and mountroot
1166 */
1167 int
ffs_mountfs(struct vnode * devvp,struct mount * mp,struct lwp * l)1168 ffs_mountfs(struct vnode *devvp, struct mount *mp, struct lwp *l)
1169 {
1170 struct ufsmount *ump = NULL;
1171 struct buf *bp = NULL;
1172 struct fs *fs = NULL;
1173 dev_t dev;
1174 void *space;
1175 daddr_t sblockloc = 0;
1176 int blks, fstype = 0;
1177 int error, i, bsize, ronly, bset = 0;
1178 #ifdef FFS_EI
1179 int needswap = 0; /* keep gcc happy */
1180 #endif
1181 int32_t *lp;
1182 kauth_cred_t cred;
1183 u_int32_t allocsbsize, fs_sbsize = 0;
1184
1185 dev = devvp->v_rdev;
1186 cred = l ? l->l_cred : NOCRED;
1187
1188 /* Flush out any old buffers remaining from a previous use. */
1189 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
1190 error = vinvalbuf(devvp, V_SAVE, cred, l, 0, 0);
1191 VOP_UNLOCK(devvp);
1192 if (error) {
1193 DPRINTF("vinvalbuf returned %d", error);
1194 return error;
1195 }
1196
1197 ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
1198
1199 ump = kmem_zalloc(sizeof(*ump), KM_SLEEP);
1200 mutex_init(&ump->um_lock, MUTEX_DEFAULT, IPL_NONE);
1201 error = ffs_snapshot_init(ump);
1202 if (error) {
1203 DPRINTF("ffs_snapshot_init returned %d", error);
1204 goto out;
1205 }
1206 ump->um_ops = &ffs_ufsops;
1207
1208 #ifdef WAPBL
1209 sbagain:
1210 #endif
1211 /*
1212 * Try reading the superblock in each of its possible locations.
1213 */
1214 for (i = 0; ; i++) {
1215 daddr_t fs_sblockloc;
1216
1217 if (bp != NULL) {
1218 brelse(bp, BC_NOCACHE);
1219 bp = NULL;
1220 }
1221 if (sblock_try[i] == -1) {
1222 DPRINTF("no superblock found");
1223 error = EINVAL;
1224 fs = NULL;
1225 goto out;
1226 }
1227
1228 error = bread(devvp, sblock_try[i] / DEV_BSIZE, SBLOCKSIZE,
1229 0, &bp);
1230 if (error) {
1231 DPRINTF("bread@0x%x returned %d",
1232 sblock_try[i] / DEV_BSIZE, error);
1233 fs = NULL;
1234 goto out;
1235 }
1236 fs = (struct fs *)bp->b_data;
1237
1238 sblockloc = sblock_try[i];
1239 DPRINTF("fs_magic 0x%x", fs->fs_magic);
1240
1241 /*
1242 * Swap: here, we swap fs->fs_sbsize in order to get the correct
1243 * size to read the superblock. Once read, we swap the whole
1244 * superblock structure.
1245 */
1246 if (fs->fs_magic == FS_UFS2EA_MAGIC) {
1247 ump->um_flags |= UFS_EA;
1248 fs->fs_magic = FS_UFS2_MAGIC;
1249 } else if (fs->fs_magic == FS_UFS2EA_MAGIC_SWAPPED) {
1250 ump->um_flags |= UFS_EA;
1251 fs->fs_magic = FS_UFS2_MAGIC_SWAPPED;
1252 }
1253 if (fs->fs_magic == FS_UFS1_MAGIC) {
1254 fs_sbsize = fs->fs_sbsize;
1255 fstype = UFS1;
1256 #ifdef FFS_EI
1257 needswap = 0;
1258 } else if (fs->fs_magic == FS_UFS1_MAGIC_SWAPPED) {
1259 fs_sbsize = bswap32(fs->fs_sbsize);
1260 fstype = UFS1;
1261 needswap = 1;
1262 #endif
1263 } else if (fs->fs_magic == FS_UFS2_MAGIC) {
1264 fs_sbsize = fs->fs_sbsize;
1265 fstype = UFS2;
1266 #ifdef FFS_EI
1267 needswap = 0;
1268 } else if (fs->fs_magic == FS_UFS2_MAGIC_SWAPPED) {
1269 fs_sbsize = bswap32(fs->fs_sbsize);
1270 fstype = UFS2;
1271 needswap = 1;
1272 #endif
1273 } else
1274 continue;
1275
1276 /* fs->fs_sblockloc isn't defined for old filesystems */
1277 if (fstype == UFS1 && !(fs->fs_old_flags & FS_FLAGS_UPDATED)) {
1278 if (sblockloc == SBLOCK_UFS2)
1279 /*
1280 * This is likely to be the first alternate
1281 * in a filesystem with 64k blocks.
1282 * Don't use it.
1283 */
1284 continue;
1285 fs_sblockloc = sblockloc;
1286 } else {
1287 fs_sblockloc = fs->fs_sblockloc;
1288 #ifdef FFS_EI
1289 if (needswap)
1290 fs_sblockloc = bswap64(fs_sblockloc);
1291 #endif
1292 }
1293
1294 /* Check we haven't found an alternate superblock */
1295 if (fs_sblockloc != sblockloc)
1296 continue;
1297
1298 /* Check the superblock size */
1299 if (fs_sbsize > SBLOCKSIZE || fs_sbsize < sizeof(struct fs))
1300 continue;
1301 fs = kmem_alloc((u_long)fs_sbsize, KM_SLEEP);
1302 memcpy(fs, bp->b_data, fs_sbsize);
1303
1304 /* Swap the whole superblock structure, if necessary. */
1305 #ifdef FFS_EI
1306 if (needswap) {
1307 ffs_sb_swap((struct fs*)bp->b_data, fs);
1308 fs->fs_flags |= FS_SWAPPED;
1309 } else
1310 #endif
1311 fs->fs_flags &= ~FS_SWAPPED;
1312
1313 /*
1314 * Now that everything is swapped, the superblock is ready to
1315 * be sanitized.
1316 */
1317 if (!ffs_superblock_validate(fs)) {
1318 kmem_free(fs, fs_sbsize);
1319 continue;
1320 }
1321
1322 /* Ok seems to be a good superblock */
1323 break;
1324 }
1325
1326 ump->um_fs = fs;
1327
1328 #ifdef WAPBL
1329 if ((mp->mnt_wapbl_replay == 0) && (fs->fs_flags & FS_DOWAPBL)) {
1330 error = ffs_wapbl_replay_start(mp, fs, devvp);
1331 if (error && (mp->mnt_flag & MNT_FORCE) == 0) {
1332 DPRINTF("ffs_wapbl_replay_start returned %d", error);
1333 goto out;
1334 }
1335 if (!error) {
1336 if (!ronly) {
1337 /* XXX fsmnt may be stale. */
1338 printf("%s: replaying log to disk\n",
1339 fs->fs_fsmnt);
1340 error = wapbl_replay_write(mp->mnt_wapbl_replay,
1341 devvp);
1342 if (error) {
1343 DPRINTF("wapbl_replay_write returned %d",
1344 error);
1345 goto out;
1346 }
1347 wapbl_replay_stop(mp->mnt_wapbl_replay);
1348 fs->fs_clean = FS_WASCLEAN;
1349 } else {
1350 /* XXX fsmnt may be stale */
1351 printf("%s: replaying log to memory\n",
1352 fs->fs_fsmnt);
1353 }
1354
1355 /* Force a re-read of the superblock */
1356 brelse(bp, BC_INVAL);
1357 bp = NULL;
1358 kmem_free(fs, fs_sbsize);
1359 fs = NULL;
1360 goto sbagain;
1361 }
1362 }
1363 #else /* !WAPBL */
1364 if ((fs->fs_flags & FS_DOWAPBL) && (mp->mnt_flag & MNT_FORCE) == 0) {
1365 error = EPERM;
1366 DPRINTF("no force %d", error);
1367 goto out;
1368 }
1369 #endif /* !WAPBL */
1370
1371 ffs_oldfscompat_read(fs, ump, sblockloc);
1372 ump->um_maxfilesize = fs->fs_maxfilesize;
1373
1374 if (fs->fs_flags & ~(FS_KNOWN_FLAGS | FS_INTERNAL)) {
1375 uprintf("%s: unknown ufs flags: 0x%08"PRIx32"%s\n",
1376 mp->mnt_stat.f_mntonname, fs->fs_flags,
1377 (mp->mnt_flag & MNT_FORCE) ? "" : ", not mounting");
1378 if ((mp->mnt_flag & MNT_FORCE) == 0) {
1379 error = EINVAL;
1380 DPRINTF("no force %d", error);
1381 goto out;
1382 }
1383 }
1384
1385 fs->fs_fmod = 0;
1386 if (fs->fs_pendingblocks != 0 || fs->fs_pendinginodes != 0) {
1387 fs->fs_pendingblocks = 0;
1388 fs->fs_pendinginodes = 0;
1389 }
1390
1391 ump->um_fstype = fstype;
1392 if (fs->fs_sbsize < SBLOCKSIZE)
1393 brelse(bp, BC_INVAL);
1394 else
1395 brelse(bp, 0);
1396 bp = NULL;
1397
1398 if (ffs_is_appleufs(devvp, fs)) {
1399 #ifdef APPLE_UFS
1400 ump->um_flags |= UFS_ISAPPLEUFS;
1401 #else
1402 DPRINTF("AppleUFS not supported");
1403 error = EINVAL;
1404 goto out;
1405 #endif
1406 }
1407
1408 #if 0
1409 /*
1410 * XXX This code changes the behaviour of mounting dirty filesystems, to
1411 * XXX require "mount -f ..." to mount them. This doesn't match what
1412 * XXX mount(8) describes and is disabled for now.
1413 */
1414 /*
1415 * If the file system is not clean, don't allow it to be mounted
1416 * unless MNT_FORCE is specified. (Note: MNT_FORCE is always set
1417 * for the root file system.)
1418 */
1419 if (fs->fs_flags & FS_DOWAPBL) {
1420 /*
1421 * wapbl normally expects to be FS_WASCLEAN when the FS_DOWAPBL
1422 * bit is set, although there's a window in unmount where it
1423 * could be FS_ISCLEAN
1424 */
1425 if ((mp->mnt_flag & MNT_FORCE) == 0 &&
1426 (fs->fs_clean & (FS_WASCLEAN | FS_ISCLEAN)) == 0) {
1427 error = EPERM;
1428 goto out;
1429 }
1430 } else
1431 if ((fs->fs_clean & FS_ISCLEAN) == 0 &&
1432 (mp->mnt_flag & MNT_FORCE) == 0) {
1433 error = EPERM;
1434 goto out;
1435 }
1436 #endif
1437
1438 /*
1439 * Verify that we can access the last block in the fs
1440 * if we're mounting read/write.
1441 */
1442 if (!ronly) {
1443 error = bread(devvp, FFS_FSBTODB(fs, fs->fs_size - 1),
1444 fs->fs_fsize, 0, &bp);
1445 if (error) {
1446 DPRINTF("bread@0x%jx returned %d",
1447 (intmax_t)FFS_FSBTODB(fs, fs->fs_size - 1),
1448 error);
1449 bset = BC_INVAL;
1450 goto out;
1451 }
1452 if (bp->b_bcount != fs->fs_fsize) {
1453 DPRINTF("bcount %x != fsize %x", bp->b_bcount,
1454 fs->fs_fsize);
1455 error = EINVAL;
1456 bset = BC_INVAL;
1457 goto out;
1458 }
1459 brelse(bp, BC_INVAL);
1460 bp = NULL;
1461 }
1462
1463 fs->fs_ronly = ronly;
1464 /* Don't bump fs_clean if we're replaying journal */
1465 if (!((fs->fs_flags & FS_DOWAPBL) && (fs->fs_clean & FS_WASCLEAN))) {
1466 if (ronly == 0) {
1467 fs->fs_clean =
1468 fs->fs_clean == FS_ISCLEAN ? FS_WASCLEAN : 0;
1469 fs->fs_fmod = 1;
1470 }
1471 }
1472
1473 bsize = fs->fs_cssize;
1474 blks = howmany(bsize, fs->fs_fsize);
1475 if (fs->fs_contigsumsize > 0)
1476 bsize += fs->fs_ncg * sizeof(int32_t);
1477 bsize += fs->fs_ncg * sizeof(*fs->fs_contigdirs);
1478 allocsbsize = bsize;
1479 space = kmem_alloc((u_long)allocsbsize, KM_SLEEP);
1480 fs->fs_csp = space;
1481
1482 for (i = 0; i < blks; i += fs->fs_frag) {
1483 bsize = fs->fs_bsize;
1484 if (i + fs->fs_frag > blks)
1485 bsize = (blks - i) * fs->fs_fsize;
1486 error = bread(devvp, FFS_FSBTODB(fs, fs->fs_csaddr + i), bsize,
1487 0, &bp);
1488 if (error) {
1489 DPRINTF("bread@0x%jx %d",
1490 (intmax_t)FFS_FSBTODB(fs, fs->fs_csaddr + i),
1491 error);
1492 goto out1;
1493 }
1494 #ifdef FFS_EI
1495 if (needswap)
1496 ffs_csum_swap((struct csum *)bp->b_data,
1497 (struct csum *)space, bsize);
1498 else
1499 #endif
1500 memcpy(space, bp->b_data, (u_int)bsize);
1501
1502 space = (char *)space + bsize;
1503 brelse(bp, 0);
1504 bp = NULL;
1505 }
1506 if (fs->fs_contigsumsize > 0) {
1507 fs->fs_maxcluster = lp = space;
1508 for (i = 0; i < fs->fs_ncg; i++)
1509 *lp++ = fs->fs_contigsumsize;
1510 space = lp;
1511 }
1512 bsize = fs->fs_ncg * sizeof(*fs->fs_contigdirs);
1513 fs->fs_contigdirs = space;
1514 space = (char *)space + bsize;
1515 memset(fs->fs_contigdirs, 0, bsize);
1516
1517 /* Compatibility for old filesystems - XXX */
1518 if (fs->fs_avgfilesize <= 0)
1519 fs->fs_avgfilesize = AVFILESIZ;
1520 if (fs->fs_avgfpdir <= 0)
1521 fs->fs_avgfpdir = AFPDIR;
1522 fs->fs_active = NULL;
1523
1524 mp->mnt_data = ump;
1525 mp->mnt_stat.f_fsidx.__fsid_val[0] = (long)dev;
1526 mp->mnt_stat.f_fsidx.__fsid_val[1] = makefstype(MOUNT_FFS);
1527 mp->mnt_stat.f_fsid = mp->mnt_stat.f_fsidx.__fsid_val[0];
1528 mp->mnt_stat.f_namemax = FFS_MAXNAMLEN;
1529 if (UFS_MPISAPPLEUFS(ump)) {
1530 /* NeXT used to keep short symlinks in the inode even
1531 * when using FS_42INODEFMT. In that case fs->fs_maxsymlinklen
1532 * is probably -1, but we still need to be able to identify
1533 * short symlinks.
1534 */
1535 ump->um_maxsymlinklen = APPLEUFS_MAXSYMLINKLEN;
1536 ump->um_dirblksiz = APPLEUFS_DIRBLKSIZ;
1537 mp->mnt_iflag |= IMNT_DTYPE;
1538 } else {
1539 ump->um_maxsymlinklen = fs->fs_maxsymlinklen;
1540 ump->um_dirblksiz = UFS_DIRBLKSIZ;
1541 if (ump->um_maxsymlinklen > 0)
1542 mp->mnt_iflag |= IMNT_DTYPE;
1543 else
1544 mp->mnt_iflag &= ~IMNT_DTYPE;
1545 }
1546 mp->mnt_fs_bshift = fs->fs_bshift;
1547 mp->mnt_dev_bshift = DEV_BSHIFT; /* XXX */
1548 mp->mnt_flag |= MNT_LOCAL;
1549 mp->mnt_iflag |= IMNT_MPSAFE | IMNT_CAN_RWTORO | IMNT_SHRLOOKUP |
1550 IMNT_NCLOOKUP;
1551 #ifdef FFS_EI
1552 if (needswap)
1553 ump->um_flags |= UFS_NEEDSWAP;
1554 #endif
1555 error = ffs_acls(mp, fs->fs_flags);
1556 if (error)
1557 goto out1;
1558 ump->um_mountp = mp;
1559 ump->um_dev = dev;
1560 ump->um_devvp = devvp;
1561 ump->um_nindir = fs->fs_nindir;
1562 ump->um_lognindir = ffs(fs->fs_nindir) - 1;
1563 ump->um_bptrtodb = fs->fs_fshift - DEV_BSHIFT;
1564 ump->um_seqinc = fs->fs_frag;
1565 for (i = 0; i < MAXQUOTAS; i++)
1566 ump->um_quotas[i] = NULLVP;
1567 spec_node_setmountedfs(devvp, mp);
1568 if (ronly == 0 && fs->fs_snapinum[0] != 0)
1569 ffs_snapshot_mount(mp);
1570 #ifdef WAPBL
1571 if (!ronly) {
1572 KDASSERT(fs->fs_ronly == 0);
1573 /*
1574 * ffs_wapbl_start() needs mp->mnt_stat initialised if it
1575 * needs to create a new log file in-filesystem.
1576 */
1577 error = ffs_statvfs(mp, &mp->mnt_stat);
1578 if (error) {
1579 DPRINTF("ffs_statvfs returned %d", error);
1580 goto out1;
1581 }
1582
1583 error = ffs_wapbl_start(mp);
1584 if (error) {
1585 DPRINTF("ffs_wapbl_start returned %d", error);
1586 goto out1;
1587 }
1588 }
1589 #endif /* WAPBL */
1590 if (ronly == 0) {
1591 #ifdef QUOTA2
1592 error = ffs_quota2_mount(mp);
1593 if (error) {
1594 DPRINTF("ffs_quota2_mount returned %d", error);
1595 goto out1;
1596 }
1597 #else
1598 if (fs->fs_flags & FS_DOQUOTA2) {
1599 ump->um_flags |= UFS_QUOTA2;
1600 uprintf("%s: options QUOTA2 not enabled%s\n",
1601 mp->mnt_stat.f_mntonname,
1602 (mp->mnt_flag & MNT_FORCE) ? "" : ", not mounting");
1603 if ((mp->mnt_flag & MNT_FORCE) == 0) {
1604 error = EINVAL;
1605 DPRINTF("quota disabled %d", error);
1606 goto out1;
1607 }
1608 }
1609 #endif
1610 }
1611
1612 if (mp->mnt_flag & MNT_DISCARD)
1613 ump->um_discarddata = ffs_discard_init(devvp, fs);
1614
1615 return (0);
1616 out1:
1617 kmem_free(fs->fs_csp, allocsbsize);
1618 out:
1619 #ifdef WAPBL
1620 if (mp->mnt_wapbl_replay) {
1621 wapbl_replay_stop(mp->mnt_wapbl_replay);
1622 wapbl_replay_free(mp->mnt_wapbl_replay);
1623 mp->mnt_wapbl_replay = 0;
1624 }
1625 #endif
1626
1627 if (fs)
1628 kmem_free(fs, fs->fs_sbsize);
1629 spec_node_setmountedfs(devvp, NULL);
1630 if (bp)
1631 brelse(bp, bset);
1632 if (ump) {
1633 if (ump->um_oldfscompat)
1634 kmem_free(ump->um_oldfscompat, 512 + 3*sizeof(int32_t));
1635 mutex_destroy(&ump->um_lock);
1636 kmem_free(ump, sizeof(*ump));
1637 mp->mnt_data = NULL;
1638 }
1639 return (error);
1640 }
1641
1642 /*
1643 * Sanity checks for loading old filesystem superblocks.
1644 * See ffs_oldfscompat_write below for unwound actions.
1645 *
1646 * XXX - Parts get retired eventually.
1647 * Unfortunately new bits get added.
1648 */
1649 static void
ffs_oldfscompat_read(struct fs * fs,struct ufsmount * ump,daddr_t sblockloc)1650 ffs_oldfscompat_read(struct fs *fs, struct ufsmount *ump, daddr_t sblockloc)
1651 {
1652 off_t maxfilesize;
1653 int32_t *extrasave;
1654
1655 if ((fs->fs_magic != FS_UFS1_MAGIC) ||
1656 (fs->fs_old_flags & FS_FLAGS_UPDATED))
1657 return;
1658
1659 if (!ump->um_oldfscompat)
1660 ump->um_oldfscompat = kmem_alloc(512 + 3*sizeof(int32_t),
1661 KM_SLEEP);
1662
1663 memcpy(ump->um_oldfscompat, &fs->fs_old_postbl_start, 512);
1664 extrasave = ump->um_oldfscompat;
1665 extrasave += 512/sizeof(int32_t);
1666 extrasave[0] = fs->fs_old_npsect;
1667 extrasave[1] = fs->fs_old_interleave;
1668 extrasave[2] = fs->fs_old_trackskew;
1669
1670 /* These fields will be overwritten by their
1671 * original values in fs_oldfscompat_write, so it is harmless
1672 * to modify them here.
1673 */
1674 fs->fs_cstotal.cs_ndir = fs->fs_old_cstotal.cs_ndir;
1675 fs->fs_cstotal.cs_nbfree = fs->fs_old_cstotal.cs_nbfree;
1676 fs->fs_cstotal.cs_nifree = fs->fs_old_cstotal.cs_nifree;
1677 fs->fs_cstotal.cs_nffree = fs->fs_old_cstotal.cs_nffree;
1678
1679 fs->fs_maxbsize = fs->fs_bsize;
1680 fs->fs_time = fs->fs_old_time;
1681 fs->fs_size = fs->fs_old_size;
1682 fs->fs_dsize = fs->fs_old_dsize;
1683 fs->fs_csaddr = fs->fs_old_csaddr;
1684 fs->fs_sblockloc = sblockloc;
1685
1686 fs->fs_flags = fs->fs_old_flags | (fs->fs_flags & FS_INTERNAL);
1687
1688 if (fs->fs_old_postblformat == FS_42POSTBLFMT) {
1689 fs->fs_old_nrpos = 8;
1690 fs->fs_old_npsect = fs->fs_old_nsect;
1691 fs->fs_old_interleave = 1;
1692 fs->fs_old_trackskew = 0;
1693 }
1694
1695 if (fs->fs_magic == FS_UFS1_MAGIC &&
1696 fs->fs_old_inodefmt < FS_44INODEFMT) {
1697 fs->fs_maxfilesize = (u_quad_t) 1LL << 39;
1698 fs->fs_qbmask = ~fs->fs_bmask;
1699 fs->fs_qfmask = ~fs->fs_fmask;
1700 }
1701
1702 maxfilesize = (u_int64_t)0x80000000 * fs->fs_bsize - 1;
1703 if (fs->fs_maxfilesize > maxfilesize)
1704 fs->fs_maxfilesize = maxfilesize;
1705
1706 /* Compatibility for old filesystems */
1707 if (fs->fs_avgfilesize <= 0)
1708 fs->fs_avgfilesize = AVFILESIZ;
1709 if (fs->fs_avgfpdir <= 0)
1710 fs->fs_avgfpdir = AFPDIR;
1711
1712 #if 0
1713 if (bigcgs) {
1714 fs->fs_save_cgsize = fs->fs_cgsize;
1715 fs->fs_cgsize = fs->fs_bsize;
1716 }
1717 #endif
1718 }
1719
1720 /*
1721 * Unwinding superblock updates for old filesystems.
1722 * See ffs_oldfscompat_read above for details.
1723 *
1724 * XXX - Parts get retired eventually.
1725 * Unfortunately new bits get added.
1726 */
1727 static void
ffs_oldfscompat_write(struct fs * fs,struct ufsmount * ump)1728 ffs_oldfscompat_write(struct fs *fs, struct ufsmount *ump)
1729 {
1730 int32_t *extrasave;
1731
1732 if ((fs->fs_magic != FS_UFS1_MAGIC) ||
1733 (fs->fs_old_flags & FS_FLAGS_UPDATED))
1734 return;
1735
1736 fs->fs_old_time = fs->fs_time;
1737 fs->fs_old_cstotal.cs_ndir = fs->fs_cstotal.cs_ndir;
1738 fs->fs_old_cstotal.cs_nbfree = fs->fs_cstotal.cs_nbfree;
1739 fs->fs_old_cstotal.cs_nifree = fs->fs_cstotal.cs_nifree;
1740 fs->fs_old_cstotal.cs_nffree = fs->fs_cstotal.cs_nffree;
1741 fs->fs_old_flags = fs->fs_flags;
1742
1743 #if 0
1744 if (bigcgs) {
1745 fs->fs_cgsize = fs->fs_save_cgsize;
1746 }
1747 #endif
1748
1749 memcpy(&fs->fs_old_postbl_start, ump->um_oldfscompat, 512);
1750 extrasave = ump->um_oldfscompat;
1751 extrasave += 512/sizeof(int32_t);
1752 fs->fs_old_npsect = extrasave[0];
1753 fs->fs_old_interleave = extrasave[1];
1754 fs->fs_old_trackskew = extrasave[2];
1755
1756 }
1757
1758 /*
1759 * unmount vfs operation
1760 */
1761 int
ffs_unmount(struct mount * mp,int mntflags)1762 ffs_unmount(struct mount *mp, int mntflags)
1763 {
1764 struct lwp *l = curlwp;
1765 struct ufsmount *ump = VFSTOUFS(mp);
1766 struct fs *fs = ump->um_fs;
1767 int error, flags;
1768 u_int32_t bsize;
1769 #ifdef WAPBL
1770 extern int doforce;
1771 #endif
1772
1773 /* The file system is suspended. */
1774 KASSERT(fstrans_is_owner(mp));
1775
1776 if (ump->um_discarddata) {
1777 ffs_discard_finish(ump->um_discarddata, mntflags);
1778 ump->um_discarddata = NULL;
1779 }
1780
1781 flags = 0;
1782 if (mntflags & MNT_FORCE)
1783 flags |= FORCECLOSE;
1784 if ((error = ffs_flushfiles(mp, flags, l)) != 0)
1785 return (error);
1786 if (fs->fs_ronly == 0 && UFS_WAPBL_BEGIN(mp) == 0) {
1787 if (ffs_cgupdate(ump, MNT_WAIT) == 0 &&
1788 fs->fs_clean & FS_WASCLEAN) {
1789 mutex_enter(&ump->um_lock);
1790 fs->fs_clean = FS_ISCLEAN;
1791 fs->fs_fmod = 0;
1792 mutex_exit(&ump->um_lock);
1793 (void) ffs_sbupdate(ump, MNT_WAIT);
1794 }
1795 UFS_WAPBL_END(mp);
1796 }
1797 #ifdef WAPBL
1798 KASSERT(!(mp->mnt_wapbl_replay && mp->mnt_wapbl));
1799 if (mp->mnt_wapbl_replay) {
1800 KDASSERT(fs->fs_ronly);
1801 wapbl_replay_stop(mp->mnt_wapbl_replay);
1802 wapbl_replay_free(mp->mnt_wapbl_replay);
1803 mp->mnt_wapbl_replay = 0;
1804 }
1805 error = ffs_wapbl_stop(mp, doforce && (mntflags & MNT_FORCE));
1806 if (error) {
1807 return error;
1808 }
1809 #endif /* WAPBL */
1810
1811 if (ump->um_devvp->v_type != VBAD)
1812 spec_node_setmountedfs(ump->um_devvp, NULL);
1813 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
1814 (void)VOP_CLOSE(ump->um_devvp, fs->fs_ronly ? FREAD : FREAD | FWRITE,
1815 NOCRED);
1816 vput(ump->um_devvp);
1817
1818 bsize = fs->fs_cssize;
1819 if (fs->fs_contigsumsize > 0)
1820 bsize += fs->fs_ncg * sizeof(int32_t);
1821 bsize += fs->fs_ncg * sizeof(*fs->fs_contigdirs);
1822 kmem_free(fs->fs_csp, bsize);
1823
1824 kmem_free(fs, fs->fs_sbsize);
1825 if (ump->um_oldfscompat != NULL)
1826 kmem_free(ump->um_oldfscompat, 512 + 3*sizeof(int32_t));
1827 mutex_destroy(&ump->um_lock);
1828 ffs_snapshot_fini(ump);
1829 kmem_free(ump, sizeof(*ump));
1830 mp->mnt_data = NULL;
1831 mp->mnt_flag &= ~MNT_LOCAL;
1832 return (0);
1833 }
1834
1835 /*
1836 * Flush out all the files in a filesystem.
1837 */
1838 int
ffs_flushfiles(struct mount * mp,int flags,struct lwp * l)1839 ffs_flushfiles(struct mount *mp, int flags, struct lwp *l)
1840 {
1841 extern int doforce;
1842 struct ufsmount *ump;
1843 int error;
1844
1845 if (!doforce)
1846 flags &= ~FORCECLOSE;
1847 ump = VFSTOUFS(mp);
1848 #ifdef QUOTA
1849 if ((error = quota1_umount(mp, flags)) != 0)
1850 return (error);
1851 #endif
1852 #ifdef QUOTA2
1853 if ((error = quota2_umount(mp, flags)) != 0)
1854 return (error);
1855 #endif
1856 #ifdef UFS_EXTATTR
1857 if (ump->um_fstype == UFS1) {
1858 if (ump->um_extattr.uepm_flags & UFS_EXTATTR_UEPM_STARTED)
1859 ufs_extattr_stop(mp, l);
1860 if (ump->um_extattr.uepm_flags & UFS_EXTATTR_UEPM_INITIALIZED)
1861 ufs_extattr_uepm_destroy(&ump->um_extattr);
1862 mp->mnt_flag &= ~MNT_EXTATTR;
1863 }
1864 #endif
1865 if ((error = vflush(mp, 0, SKIPSYSTEM | flags)) != 0)
1866 return (error);
1867 ffs_snapshot_unmount(mp);
1868 /*
1869 * Flush all the files.
1870 */
1871 error = vflush(mp, NULLVP, flags);
1872 if (error)
1873 return (error);
1874 /*
1875 * Flush filesystem metadata.
1876 */
1877 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
1878 error = VOP_FSYNC(ump->um_devvp, l->l_cred, FSYNC_WAIT, 0, 0);
1879 VOP_UNLOCK(ump->um_devvp);
1880 if (flags & FORCECLOSE) /* XXXDBJ */
1881 error = 0;
1882
1883 #ifdef WAPBL
1884 if (error)
1885 return error;
1886 if (mp->mnt_wapbl) {
1887 error = wapbl_flush(mp->mnt_wapbl, 1);
1888 if (flags & FORCECLOSE)
1889 error = 0;
1890 }
1891 #endif
1892
1893 return (error);
1894 }
1895
1896 /*
1897 * Get file system statistics.
1898 */
1899 int
ffs_statvfs(struct mount * mp,struct statvfs * sbp)1900 ffs_statvfs(struct mount *mp, struct statvfs *sbp)
1901 {
1902 struct ufsmount *ump;
1903 struct fs *fs;
1904
1905 ump = VFSTOUFS(mp);
1906 fs = ump->um_fs;
1907 mutex_enter(&ump->um_lock);
1908 sbp->f_bsize = fs->fs_bsize;
1909 sbp->f_frsize = fs->fs_fsize;
1910 sbp->f_iosize = fs->fs_bsize;
1911 sbp->f_blocks = fs->fs_dsize;
1912 sbp->f_bfree = ffs_blkstofrags(fs, fs->fs_cstotal.cs_nbfree) +
1913 fs->fs_cstotal.cs_nffree + FFS_DBTOFSB(fs, fs->fs_pendingblocks);
1914 sbp->f_bresvd = ((u_int64_t) fs->fs_dsize * (u_int64_t)
1915 fs->fs_minfree) / (u_int64_t) 100;
1916 if (sbp->f_bfree > sbp->f_bresvd)
1917 sbp->f_bavail = sbp->f_bfree - sbp->f_bresvd;
1918 else
1919 sbp->f_bavail = 0;
1920 sbp->f_files = fs->fs_ncg * fs->fs_ipg - UFS_ROOTINO;
1921 sbp->f_ffree = fs->fs_cstotal.cs_nifree + fs->fs_pendinginodes;
1922 sbp->f_favail = sbp->f_ffree;
1923 sbp->f_fresvd = 0;
1924 mutex_exit(&ump->um_lock);
1925 copy_statvfs_info(sbp, mp);
1926
1927 return (0);
1928 }
1929
1930 struct ffs_sync_ctx {
1931 int waitfor;
1932 };
1933
1934 static bool
ffs_sync_selector(void * cl,struct vnode * vp)1935 ffs_sync_selector(void *cl, struct vnode *vp)
1936 {
1937 struct ffs_sync_ctx *c = cl;
1938 struct inode *ip;
1939
1940 KASSERT(mutex_owned(vp->v_interlock));
1941
1942 ip = VTOI(vp);
1943 /*
1944 * Skip the vnode/inode if inaccessible.
1945 */
1946 if (ip == NULL || vp->v_type == VNON)
1947 return false;
1948
1949 /*
1950 * We deliberately update inode times here. This will
1951 * prevent a massive queue of updates accumulating, only
1952 * to be handled by a call to unmount.
1953 *
1954 * XXX It would be better to have the syncer trickle these
1955 * out. Adjustment needed to allow registering vnodes for
1956 * sync when the vnode is clean, but the inode dirty. Or
1957 * have ufs itself trickle out inode updates.
1958 *
1959 * If doing a lazy sync, we don't care about metadata or
1960 * data updates, because they are handled by each vnode's
1961 * synclist entry. In this case we are only interested in
1962 * writing back modified inodes.
1963 */
1964 if ((ip->i_flag & (IN_ACCESS | IN_CHANGE | IN_UPDATE |
1965 IN_MODIFY | IN_MODIFIED | IN_ACCESSED)) == 0 &&
1966 (c->waitfor == MNT_LAZY || (LIST_EMPTY(&vp->v_dirtyblkhd) &&
1967 (vp->v_iflag & VI_ONWORKLST) == 0)))
1968 return false;
1969
1970 return true;
1971 }
1972
1973 /*
1974 * Go through the disk queues to initiate sandbagged IO;
1975 * go through the inodes to write those that have been modified;
1976 * initiate the writing of the super block if it has been modified.
1977 */
1978 int
ffs_sync(struct mount * mp,int waitfor,kauth_cred_t cred)1979 ffs_sync(struct mount *mp, int waitfor, kauth_cred_t cred)
1980 {
1981 struct vnode *vp;
1982 struct ufsmount *ump = VFSTOUFS(mp);
1983 struct fs *fs;
1984 struct vnode_iterator *marker;
1985 int error, allerror = 0;
1986 struct ffs_sync_ctx ctx;
1987
1988 fs = ump->um_fs;
1989 if (fs->fs_fmod != 0 && fs->fs_ronly != 0) { /* XXX */
1990 panic("%s: rofs mod, fs=%s", __func__, fs->fs_fsmnt);
1991 }
1992
1993 /*
1994 * Write back each (modified) inode.
1995 */
1996 vfs_vnode_iterator_init(mp, &marker);
1997
1998 ctx.waitfor = waitfor;
1999 while ((vp = vfs_vnode_iterator_next(marker, ffs_sync_selector, &ctx)))
2000 {
2001 error = vn_lock(vp,
2002 LK_EXCLUSIVE | (waitfor == MNT_LAZY ? LK_NOWAIT : 0));
2003 if (error) {
2004 vrele(vp);
2005 continue;
2006 }
2007 if (waitfor == MNT_LAZY) {
2008 error = UFS_WAPBL_BEGIN(vp->v_mount);
2009 if (!error) {
2010 error = ffs_update(vp, NULL, NULL,
2011 UPDATE_CLOSE);
2012 UFS_WAPBL_END(vp->v_mount);
2013 }
2014 } else {
2015 error = VOP_FSYNC(vp, cred, FSYNC_NOLOG |
2016 (waitfor == MNT_WAIT ? FSYNC_WAIT : 0), 0, 0);
2017 }
2018 if (error)
2019 allerror = error;
2020 vput(vp);
2021 }
2022 vfs_vnode_iterator_destroy(marker);
2023
2024 /*
2025 * Force stale file system control information to be flushed.
2026 */
2027 if (waitfor != MNT_LAZY) {
2028 bool need_devvp_fsync;
2029
2030 mutex_enter(ump->um_devvp->v_interlock);
2031 need_devvp_fsync = (ump->um_devvp->v_numoutput > 0 ||
2032 !LIST_EMPTY(&ump->um_devvp->v_dirtyblkhd));
2033 mutex_exit(ump->um_devvp->v_interlock);
2034 if (need_devvp_fsync) {
2035 int flags = FSYNC_NOLOG;
2036
2037 if (waitfor == MNT_WAIT)
2038 flags |= FSYNC_WAIT;
2039
2040 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
2041 if ((error = VOP_FSYNC(ump->um_devvp, cred, flags, 0,
2042 0)) != 0)
2043 allerror = error;
2044 VOP_UNLOCK(ump->um_devvp);
2045 }
2046 }
2047 #if defined(QUOTA) || defined(QUOTA2)
2048 qsync(mp);
2049 #endif
2050 /*
2051 * Write back modified superblock.
2052 */
2053 error = UFS_WAPBL_BEGIN(mp);
2054 if (error) {
2055 allerror = error;
2056 } else {
2057 mutex_enter(&ump->um_lock);
2058 if (fs->fs_fmod != 0) {
2059 fs->fs_fmod = 0;
2060 fs->fs_time = time_second;
2061 mutex_exit(&ump->um_lock);
2062 if ((error = ffs_cgupdate(ump, waitfor)))
2063 allerror = error;
2064 } else {
2065 mutex_exit(&ump->um_lock);
2066 }
2067 UFS_WAPBL_END(mp);
2068 }
2069
2070 #ifdef WAPBL
2071 if (mp->mnt_wapbl) {
2072 error = wapbl_flush(mp->mnt_wapbl, (waitfor == MNT_WAIT));
2073 if (error)
2074 allerror = error;
2075 }
2076 #endif
2077
2078 return (allerror);
2079 }
2080
2081 /*
2082 * Load inode from disk and initialize vnode.
2083 */
2084 static int
ffs_init_vnode(struct ufsmount * ump,struct vnode * vp,ino_t ino)2085 ffs_init_vnode(struct ufsmount *ump, struct vnode *vp, ino_t ino)
2086 {
2087 struct fs *fs;
2088 struct inode *ip;
2089 struct buf *bp;
2090 int error;
2091
2092 fs = ump->um_fs;
2093
2094 /* Read in the disk contents for the inode. */
2095 error = bread(ump->um_devvp, FFS_FSBTODB(fs, ino_to_fsba(fs, ino)),
2096 (int)fs->fs_bsize, 0, &bp);
2097 if (error)
2098 return error;
2099
2100 /* Allocate and initialize inode. */
2101 ip = pool_cache_get(ffs_inode_cache, PR_WAITOK);
2102 memset(ip, 0, sizeof(struct inode));
2103 ip->i_ump = ump;
2104 ip->i_fs = fs;
2105 ip->i_dev = ump->um_dev;
2106 ip->i_number = ino;
2107 if (ump->um_fstype == UFS1)
2108 ip->i_din.ffs1_din = pool_cache_get(ffs_dinode1_cache,
2109 PR_WAITOK);
2110 else
2111 ip->i_din.ffs2_din = pool_cache_get(ffs_dinode2_cache,
2112 PR_WAITOK);
2113 ffs_load_inode(bp, ip, fs, ino);
2114 brelse(bp, 0);
2115 ip->i_vnode = vp;
2116 #if defined(QUOTA) || defined(QUOTA2)
2117 ufsquota_init(ip);
2118 #endif
2119
2120 /* Initialise vnode with this inode. */
2121 vp->v_tag = VT_UFS;
2122 vp->v_op = ffs_vnodeop_p;
2123 vp->v_data = ip;
2124
2125 /* Initialize genfs node. */
2126 genfs_node_init(vp, &ffs_genfsops);
2127
2128 return 0;
2129 }
2130
2131 /*
2132 * Undo ffs_init_vnode().
2133 */
2134 static void
ffs_deinit_vnode(struct ufsmount * ump,struct vnode * vp)2135 ffs_deinit_vnode(struct ufsmount *ump, struct vnode *vp)
2136 {
2137 struct inode *ip = VTOI(vp);
2138
2139 genfs_node_destroy(vp);
2140 vp->v_data = NULL;
2141
2142 if (ump->um_fstype == UFS1)
2143 pool_cache_put(ffs_dinode1_cache, ip->i_din.ffs1_din);
2144 else
2145 pool_cache_put(ffs_dinode2_cache, ip->i_din.ffs2_din);
2146 pool_cache_put(ffs_inode_cache, ip);
2147 }
2148
2149 /*
2150 * Read an inode from disk and initialize this vnode / inode pair.
2151 * Caller assures no other thread will try to load this inode.
2152 */
2153 int
ffs_loadvnode(struct mount * mp,struct vnode * vp,const void * key,size_t key_len,const void ** new_key)2154 ffs_loadvnode(struct mount *mp, struct vnode *vp,
2155 const void *key, size_t key_len, const void **new_key)
2156 {
2157 ino_t ino;
2158 struct fs *fs;
2159 struct inode *ip;
2160 struct ufsmount *ump;
2161 int error;
2162
2163 KASSERT(key_len == sizeof(ino));
2164 memcpy(&ino, key, key_len);
2165 ump = VFSTOUFS(mp);
2166 fs = ump->um_fs;
2167
2168 error = ffs_init_vnode(ump, vp, ino);
2169 if (error)
2170 return error;
2171
2172 ip = VTOI(vp);
2173 if (ip->i_mode == 0) {
2174 ffs_deinit_vnode(ump, vp);
2175
2176 return ENOENT;
2177 }
2178
2179 /* Initialize the vnode from the inode. */
2180 ufs_vinit(mp, ffs_specop_p, ffs_fifoop_p, &vp);
2181
2182 /* Finish inode initialization. */
2183 ip->i_devvp = ump->um_devvp;
2184 vref(ip->i_devvp);
2185
2186 /*
2187 * Ensure that uid and gid are correct. This is a temporary
2188 * fix until fsck has been changed to do the update.
2189 */
2190
2191 if (fs->fs_magic == FS_UFS1_MAGIC && /* XXX */
2192 fs->fs_old_inodefmt < FS_44INODEFMT) { /* XXX */
2193 ip->i_uid = ip->i_ffs1_ouid; /* XXX */
2194 ip->i_gid = ip->i_ffs1_ogid; /* XXX */
2195 } /* XXX */
2196 uvm_vnp_setsize(vp, ip->i_size);
2197 cache_enter_id(vp, ip->i_mode, ip->i_uid, ip->i_gid, !HAS_ACLS(ip));
2198 *new_key = &ip->i_number;
2199 return 0;
2200 }
2201
2202 /*
2203 * Create a new inode on disk and initialize this vnode / inode pair.
2204 */
2205 int
ffs_newvnode(struct mount * mp,struct vnode * dvp,struct vnode * vp,struct vattr * vap,kauth_cred_t cred,void * extra,size_t * key_len,const void ** new_key)2206 ffs_newvnode(struct mount *mp, struct vnode *dvp, struct vnode *vp,
2207 struct vattr *vap, kauth_cred_t cred, void *extra,
2208 size_t *key_len, const void **new_key)
2209 {
2210 ino_t ino;
2211 struct fs *fs;
2212 struct inode *ip;
2213 struct timespec ts;
2214 struct ufsmount *ump;
2215 int error, mode;
2216
2217 KASSERT(dvp->v_mount == mp);
2218 KASSERT(vap->va_type != VNON);
2219
2220 *key_len = sizeof(ino);
2221 ump = VFSTOUFS(mp);
2222 fs = ump->um_fs;
2223 mode = MAKEIMODE(vap->va_type, vap->va_mode);
2224
2225 /* Allocate fresh inode. */
2226 error = ffs_valloc(dvp, mode, cred, &ino);
2227 if (error)
2228 return error;
2229
2230 /* Attach inode to vnode. */
2231 error = ffs_init_vnode(ump, vp, ino);
2232 if (error) {
2233 if (UFS_WAPBL_BEGIN(mp) == 0) {
2234 ffs_vfree(dvp, ino, mode);
2235 UFS_WAPBL_END(mp);
2236 }
2237 return error;
2238 }
2239
2240 ip = VTOI(vp);
2241 if (ip->i_mode) {
2242 panic("%s: dup alloc ino=%" PRId64 " on %s: mode %o/%o "
2243 "gen %x/%x size %" PRIx64 " blocks %" PRIx64,
2244 __func__, ino, fs->fs_fsmnt, DIP(ip, mode), ip->i_mode,
2245 DIP(ip, gen), ip->i_gen, DIP(ip, size), DIP(ip, blocks));
2246 }
2247 if (DIP(ip, size) || DIP(ip, blocks)) {
2248 printf("%s: ino=%" PRId64 " on %s: "
2249 "gen %x/%x has non zero blocks %" PRIx64 " or size %"
2250 PRIx64 "\n",
2251 __func__, ino, fs->fs_fsmnt, DIP(ip, gen), ip->i_gen,
2252 DIP(ip, blocks), DIP(ip, size));
2253 if ((ip)->i_ump->um_fstype == UFS1)
2254 panic("%s: dirty filesystem?", __func__);
2255 DIP_ASSIGN(ip, blocks, 0);
2256 DIP_ASSIGN(ip, size, 0);
2257 }
2258
2259 /* Set uid / gid. */
2260 if (cred == NOCRED || cred == FSCRED) {
2261 ip->i_gid = 0;
2262 ip->i_uid = 0;
2263 } else {
2264 ip->i_gid = VTOI(dvp)->i_gid;
2265 ip->i_uid = kauth_cred_geteuid(cred);
2266 }
2267 DIP_ASSIGN(ip, gid, ip->i_gid);
2268 DIP_ASSIGN(ip, uid, ip->i_uid);
2269
2270 #if defined(QUOTA) || defined(QUOTA2)
2271 error = UFS_WAPBL_BEGIN(mp);
2272 if (error) {
2273 ffs_deinit_vnode(ump, vp);
2274
2275 return error;
2276 }
2277 error = chkiq(ip, 1, cred, 0);
2278 if (error) {
2279 ffs_vfree(dvp, ino, mode);
2280 UFS_WAPBL_END(mp);
2281 ffs_deinit_vnode(ump, vp);
2282
2283 return error;
2284 }
2285 UFS_WAPBL_END(mp);
2286 #endif
2287
2288 /* Set type and finalize. */
2289 ip->i_flags = 0;
2290 DIP_ASSIGN(ip, flags, 0);
2291 ip->i_mode = mode;
2292 DIP_ASSIGN(ip, mode, mode);
2293 if (vap->va_rdev != VNOVAL) {
2294 /*
2295 * Want to be able to use this to make badblock
2296 * inodes, so don't truncate the dev number.
2297 */
2298 if (ump->um_fstype == UFS1)
2299 ip->i_ffs1_rdev = ufs_rw32(vap->va_rdev,
2300 UFS_MPNEEDSWAP(ump));
2301 else
2302 ip->i_ffs2_rdev = ufs_rw64(vap->va_rdev,
2303 UFS_MPNEEDSWAP(ump));
2304 }
2305 ufs_vinit(mp, ffs_specop_p, ffs_fifoop_p, &vp);
2306 ip->i_devvp = ump->um_devvp;
2307 vref(ip->i_devvp);
2308
2309 /* Set up a new generation number for this inode. */
2310 ip->i_gen++;
2311 DIP_ASSIGN(ip, gen, ip->i_gen);
2312 if (fs->fs_magic == FS_UFS2_MAGIC) {
2313 vfs_timestamp(&ts);
2314 ip->i_ffs2_birthtime = ts.tv_sec;
2315 ip->i_ffs2_birthnsec = ts.tv_nsec;
2316 }
2317
2318 uvm_vnp_setsize(vp, ip->i_size);
2319 cache_enter_id(vp, ip->i_mode, ip->i_uid, ip->i_gid, !HAS_ACLS(ip));
2320 *new_key = &ip->i_number;
2321 return 0;
2322 }
2323
2324 /*
2325 * File handle to vnode
2326 *
2327 * Have to be really careful about stale file handles:
2328 * - check that the inode number is valid
2329 * - call ffs_vget() to get the locked inode
2330 * - check for an unallocated inode (i_mode == 0)
2331 * - check that the given client host has export rights and return
2332 * those rights via. exflagsp and credanonp
2333 */
2334 int
ffs_fhtovp(struct mount * mp,struct fid * fhp,int lktype,struct vnode ** vpp)2335 ffs_fhtovp(struct mount *mp, struct fid *fhp, int lktype, struct vnode **vpp)
2336 {
2337 struct ufid ufh;
2338 int error;
2339
2340 if (fhp->fid_len != sizeof(struct ufid))
2341 return EINVAL;
2342
2343 memcpy(&ufh, fhp, sizeof(ufh));
2344 if ((error = ffs_checkrange(mp, ufh.ufid_ino)) != 0)
2345 return error;
2346
2347 return (ufs_fhtovp(mp, &ufh, lktype, vpp));
2348 }
2349
2350 /*
2351 * Vnode pointer to File handle
2352 */
2353 /* ARGSUSED */
2354 int
ffs_vptofh(struct vnode * vp,struct fid * fhp,size_t * fh_size)2355 ffs_vptofh(struct vnode *vp, struct fid *fhp, size_t *fh_size)
2356 {
2357 struct inode *ip;
2358 struct ufid ufh;
2359
2360 if (*fh_size < sizeof(struct ufid)) {
2361 *fh_size = sizeof(struct ufid);
2362 return E2BIG;
2363 }
2364 ip = VTOI(vp);
2365 *fh_size = sizeof(struct ufid);
2366 memset(&ufh, 0, sizeof(ufh));
2367 ufh.ufid_len = sizeof(struct ufid);
2368 ufh.ufid_ino = ip->i_number;
2369 ufh.ufid_gen = ip->i_gen;
2370 memcpy(fhp, &ufh, sizeof(ufh));
2371 return (0);
2372 }
2373
2374 void
ffs_init(void)2375 ffs_init(void)
2376 {
2377 if (ffs_initcount++ > 0)
2378 return;
2379
2380 ffs_inode_cache = pool_cache_init(sizeof(struct inode), 0, 0, 0,
2381 "ffsino", NULL, IPL_NONE, NULL, NULL, NULL);
2382 ffs_dinode1_cache = pool_cache_init(sizeof(struct ufs1_dinode), 0, 0, 0,
2383 "ffsdino1", NULL, IPL_NONE, NULL, NULL, NULL);
2384 ffs_dinode2_cache = pool_cache_init(sizeof(struct ufs2_dinode), 0, 0, 0,
2385 "ffsdino2", NULL, IPL_NONE, NULL, NULL, NULL);
2386 ufs_init();
2387 }
2388
2389 void
ffs_reinit(void)2390 ffs_reinit(void)
2391 {
2392 ufs_reinit();
2393 }
2394
2395 void
ffs_done(void)2396 ffs_done(void)
2397 {
2398 if (--ffs_initcount > 0)
2399 return;
2400
2401 ufs_done();
2402 pool_cache_destroy(ffs_dinode2_cache);
2403 pool_cache_destroy(ffs_dinode1_cache);
2404 pool_cache_destroy(ffs_inode_cache);
2405 }
2406
2407 /*
2408 * Write a superblock and associated information back to disk.
2409 */
2410 int
ffs_sbupdate(struct ufsmount * mp,int waitfor)2411 ffs_sbupdate(struct ufsmount *mp, int waitfor)
2412 {
2413 struct fs *fs = mp->um_fs;
2414 struct fs *bfs;
2415 struct buf *bp;
2416 int error;
2417
2418 error = ffs_getblk(mp->um_devvp,
2419 fs->fs_sblockloc / DEV_BSIZE, FFS_NOBLK,
2420 fs->fs_sbsize, false, &bp);
2421 if (error)
2422 return error;
2423
2424 mutex_enter(&mp->um_lock);
2425 memcpy(bp->b_data, fs, fs->fs_sbsize);
2426 mutex_exit(&mp->um_lock);
2427
2428 bfs = (struct fs *)bp->b_data;
2429
2430 bfs->fs_flags &= ~FS_INTERNAL;
2431 ffs_oldfscompat_write((struct fs *)bp->b_data, mp);
2432 if (mp->um_flags & UFS_EA) {
2433 KASSERT(bfs->fs_magic == FS_UFS2_MAGIC);
2434 bfs->fs_magic = FS_UFS2EA_MAGIC;
2435 }
2436 #ifdef FFS_EI
2437 if (mp->um_flags & UFS_NEEDSWAP)
2438 ffs_sb_swap(bfs, bfs);
2439 #endif
2440
2441 if (waitfor == MNT_WAIT)
2442 error = bwrite(bp);
2443 else
2444 bawrite(bp);
2445 return (error);
2446 }
2447
2448 int
ffs_cgupdate(struct ufsmount * mp,int waitfor)2449 ffs_cgupdate(struct ufsmount *mp, int waitfor)
2450 {
2451 struct fs *fs = mp->um_fs;
2452 struct buf *bp;
2453 int blks;
2454 void *space;
2455 int i, size, error = 0, allerror = 0;
2456
2457 UFS_WAPBL_JLOCK_ASSERT(mp->um_mountp);
2458
2459 allerror = ffs_sbupdate(mp, waitfor);
2460 blks = howmany(fs->fs_cssize, fs->fs_fsize);
2461 space = fs->fs_csp;
2462 for (i = 0; i < blks; i += fs->fs_frag) {
2463 size = fs->fs_bsize;
2464 if (i + fs->fs_frag > blks)
2465 size = (blks - i) * fs->fs_fsize;
2466 error = ffs_getblk(mp->um_devvp, FFS_FSBTODB(fs, fs->fs_csaddr + i),
2467 FFS_NOBLK, size, false, &bp);
2468 if (error)
2469 break;
2470 #ifdef FFS_EI
2471 if (mp->um_flags & UFS_NEEDSWAP)
2472 ffs_csum_swap((struct csum*)space,
2473 (struct csum*)bp->b_data, size);
2474 else
2475 #endif
2476 memcpy(bp->b_data, space, (u_int)size);
2477 space = (char *)space + size;
2478 if (waitfor == MNT_WAIT)
2479 error = bwrite(bp);
2480 else
2481 bawrite(bp);
2482 }
2483 if (!allerror && error)
2484 allerror = error;
2485 return (allerror);
2486 }
2487
2488 int
ffs_extattrctl(struct mount * mp,int cmd,struct vnode * vp,int attrnamespace,const char * attrname)2489 ffs_extattrctl(struct mount *mp, int cmd, struct vnode *vp,
2490 int attrnamespace, const char *attrname)
2491 {
2492 #ifdef UFS_EXTATTR
2493 /*
2494 * File-backed extended attributes are only supported on UFS1.
2495 * UFS2 has native extended attributes.
2496 */
2497 if (VFSTOUFS(mp)->um_fstype == UFS1)
2498 return (ufs_extattrctl(mp, cmd, vp, attrnamespace, attrname));
2499 #endif
2500 return (vfs_stdextattrctl(mp, cmd, vp, attrnamespace, attrname));
2501 }
2502
2503 /*
2504 * Synch vnode for a mounted file system.
2505 */
2506 static int
ffs_vfs_fsync(vnode_t * vp,int flags)2507 ffs_vfs_fsync(vnode_t *vp, int flags)
2508 {
2509 int error, i, pflags;
2510 #ifdef WAPBL
2511 struct mount *mp;
2512 #endif
2513
2514 KASSERT(vp->v_type == VBLK);
2515 KASSERT(spec_node_getmountedfs(vp) != NULL);
2516
2517 /*
2518 * Flush all dirty data associated with the vnode.
2519 */
2520 pflags = PGO_ALLPAGES | PGO_CLEANIT;
2521 if ((flags & FSYNC_WAIT) != 0)
2522 pflags |= PGO_SYNCIO;
2523 rw_enter(vp->v_uobj.vmobjlock, RW_WRITER);
2524 error = VOP_PUTPAGES(vp, 0, 0, pflags);
2525 if (error)
2526 return error;
2527
2528 #ifdef WAPBL
2529 mp = spec_node_getmountedfs(vp);
2530 if (mp && mp->mnt_wapbl) {
2531 /*
2532 * Don't bother writing out metadata if the syncer is
2533 * making the request. We will let the sync vnode
2534 * write it out in a single burst through a call to
2535 * VFS_SYNC().
2536 */
2537 if ((flags & (FSYNC_DATAONLY | FSYNC_LAZY | FSYNC_NOLOG)) != 0)
2538 return 0;
2539
2540 /*
2541 * Don't flush the log if the vnode being flushed
2542 * contains no dirty buffers that could be in the log.
2543 */
2544 if (!LIST_EMPTY(&vp->v_dirtyblkhd)) {
2545 error = wapbl_flush(mp->mnt_wapbl, 0);
2546 if (error)
2547 return error;
2548 }
2549
2550 if ((flags & FSYNC_WAIT) != 0) {
2551 mutex_enter(vp->v_interlock);
2552 while (vp->v_numoutput)
2553 cv_wait(&vp->v_cv, vp->v_interlock);
2554 mutex_exit(vp->v_interlock);
2555 }
2556
2557 return 0;
2558 }
2559 #endif /* WAPBL */
2560
2561 error = vflushbuf(vp, flags);
2562 if (error == 0 && (flags & FSYNC_CACHE) != 0) {
2563 i = 1;
2564 (void)VOP_IOCTL(vp, DIOCCACHESYNC, &i, FWRITE,
2565 kauth_cred_get());
2566 }
2567
2568 return error;
2569 }
2570