ffs_vfsops.c revision 1.257.2.9 1 /* $NetBSD: ffs_vfsops.c,v 1.257.2.9 2010/09/26 06:35:32 uebayasi 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 1.257.2.9 2010/09/26 06:35:32 uebayasi Exp $");
65
66 #if defined(_KERNEL_OPT)
67 #include "opt_ffs.h"
68 #include "opt_quota.h"
69 #include "opt_wapbl.h"
70 #include "opt_xip.h"
71 #endif
72
73 #include <sys/param.h>
74 #include <sys/systm.h>
75 #include <sys/namei.h>
76 #include <sys/proc.h>
77 #include <sys/kernel.h>
78 #include <sys/vnode.h>
79 #include <sys/socket.h>
80 #include <sys/mount.h>
81 #include <sys/buf.h>
82 #include <sys/device.h>
83 #include <sys/mbuf.h>
84 #include <sys/file.h>
85 #include <sys/disklabel.h>
86 #include <sys/ioctl.h>
87 #include <sys/errno.h>
88 #include <sys/malloc.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/fstrans.h>
96 #include <sys/module.h>
97
98 #include <miscfs/genfs/genfs.h>
99 #include <miscfs/specfs/specdev.h>
100
101 #include <ufs/ufs/quota.h>
102 #include <ufs/ufs/ufsmount.h>
103 #include <ufs/ufs/inode.h>
104 #include <ufs/ufs/dir.h>
105 #include <ufs/ufs/ufs_extern.h>
106 #include <ufs/ufs/ufs_bswap.h>
107 #include <ufs/ufs/ufs_wapbl.h>
108
109 #include <ufs/ffs/fs.h>
110 #include <ufs/ffs/ffs_extern.h>
111
112 MODULE(MODULE_CLASS_VFS, ffs, NULL);
113
114 static int ffs_vfs_fsync(vnode_t *, int);
115
116 static struct sysctllog *ffs_sysctl_log;
117
118 /* how many times ffs_init() was called */
119 int ffs_initcount = 0;
120
121 extern const struct vnodeopv_desc ffs_vnodeop_opv_desc;
122 extern const struct vnodeopv_desc ffs_specop_opv_desc;
123 extern const struct vnodeopv_desc ffs_fifoop_opv_desc;
124
125 const struct vnodeopv_desc * const ffs_vnodeopv_descs[] = {
126 &ffs_vnodeop_opv_desc,
127 &ffs_specop_opv_desc,
128 &ffs_fifoop_opv_desc,
129 NULL,
130 };
131
132 struct vfsops ffs_vfsops = {
133 MOUNT_FFS,
134 sizeof (struct ufs_args),
135 ffs_mount,
136 ufs_start,
137 ffs_unmount,
138 ufs_root,
139 ufs_quotactl,
140 ffs_statvfs,
141 ffs_sync,
142 ffs_vget,
143 ffs_fhtovp,
144 ffs_vptofh,
145 ffs_init,
146 ffs_reinit,
147 ffs_done,
148 ffs_mountroot,
149 ffs_snapshot,
150 ffs_extattrctl,
151 ffs_suspendctl,
152 genfs_renamelock_enter,
153 genfs_renamelock_exit,
154 ffs_vfs_fsync,
155 ffs_vnodeopv_descs,
156 0,
157 { NULL, NULL },
158 };
159
160 static const struct genfs_ops ffs_genfsops = {
161 .gop_size = ffs_gop_size,
162 .gop_alloc = ufs_gop_alloc,
163 .gop_write = genfs_gop_write,
164 .gop_markupdate = ufs_gop_markupdate,
165 };
166
167 static const struct ufs_ops ffs_ufsops = {
168 .uo_itimes = ffs_itimes,
169 .uo_update = ffs_update,
170 .uo_truncate = ffs_truncate,
171 .uo_valloc = ffs_valloc,
172 .uo_vfree = ffs_vfree,
173 .uo_balloc = ffs_balloc,
174 .uo_unmark_vnode = (void (*)(vnode_t *))nullop,
175 };
176
177 static int
178 ffs_modcmd(modcmd_t cmd, void *arg)
179 {
180 int error;
181
182 #if 0
183 extern int doasyncfree;
184 #endif
185 extern int ffs_log_changeopt;
186
187 switch (cmd) {
188 case MODULE_CMD_INIT:
189 error = vfs_attach(&ffs_vfsops);
190 if (error != 0)
191 break;
192
193 sysctl_createv(&ffs_sysctl_log, 0, NULL, NULL,
194 CTLFLAG_PERMANENT,
195 CTLTYPE_NODE, "vfs", NULL,
196 NULL, 0, NULL, 0,
197 CTL_VFS, CTL_EOL);
198 sysctl_createv(&ffs_sysctl_log, 0, NULL, NULL,
199 CTLFLAG_PERMANENT,
200 CTLTYPE_NODE, "ffs",
201 SYSCTL_DESCR("Berkeley Fast File System"),
202 NULL, 0, NULL, 0,
203 CTL_VFS, 1, CTL_EOL);
204
205 /*
206 * @@@ should we even bother with these first three?
207 */
208 sysctl_createv(&ffs_sysctl_log, 0, NULL, NULL,
209 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
210 CTLTYPE_INT, "doclusterread", NULL,
211 sysctl_notavail, 0, NULL, 0,
212 CTL_VFS, 1, FFS_CLUSTERREAD, CTL_EOL);
213 sysctl_createv(&ffs_sysctl_log, 0, NULL, NULL,
214 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
215 CTLTYPE_INT, "doclusterwrite", NULL,
216 sysctl_notavail, 0, NULL, 0,
217 CTL_VFS, 1, FFS_CLUSTERWRITE, CTL_EOL);
218 sysctl_createv(&ffs_sysctl_log, 0, NULL, NULL,
219 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
220 CTLTYPE_INT, "doreallocblks", NULL,
221 sysctl_notavail, 0, NULL, 0,
222 CTL_VFS, 1, FFS_REALLOCBLKS, CTL_EOL);
223 #if 0
224 sysctl_createv(&ffs_sysctl_log, 0, NULL, NULL,
225 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
226 CTLTYPE_INT, "doasyncfree",
227 SYSCTL_DESCR("Release dirty blocks asynchronously"),
228 NULL, 0, &doasyncfree, 0,
229 CTL_VFS, 1, FFS_ASYNCFREE, CTL_EOL);
230 #endif
231 sysctl_createv(&ffs_sysctl_log, 0, NULL, NULL,
232 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
233 CTLTYPE_INT, "log_changeopt",
234 SYSCTL_DESCR("Log changes in optimization strategy"),
235 NULL, 0, &ffs_log_changeopt, 0,
236 CTL_VFS, 1, FFS_LOG_CHANGEOPT, CTL_EOL);
237 break;
238 case MODULE_CMD_FINI:
239 error = vfs_detach(&ffs_vfsops);
240 if (error != 0)
241 break;
242 sysctl_teardown(&ffs_sysctl_log);
243 break;
244 default:
245 error = ENOTTY;
246 break;
247 }
248
249 return (error);
250 }
251
252 pool_cache_t ffs_inode_cache;
253 pool_cache_t ffs_dinode1_cache;
254 pool_cache_t ffs_dinode2_cache;
255
256 static void ffs_oldfscompat_read(struct fs *, struct ufsmount *, daddr_t);
257 static void ffs_oldfscompat_write(struct fs *, struct ufsmount *);
258
259 /*
260 * Called by main() when ffs is going to be mounted as root.
261 */
262
263 int
264 ffs_mountroot(void)
265 {
266 struct fs *fs;
267 struct mount *mp;
268 struct lwp *l = curlwp; /* XXX */
269 struct ufsmount *ump;
270 int error;
271
272 if (device_class(root_device) != DV_DISK)
273 return (ENODEV);
274
275 if ((error = vfs_rootmountalloc(MOUNT_FFS, "root_device", &mp))) {
276 vrele(rootvp);
277 return (error);
278 }
279
280 /*
281 * We always need to be able to mount the root file system.
282 */
283 mp->mnt_flag |= MNT_FORCE;
284 if ((error = ffs_mountfs(rootvp, mp, l)) != 0) {
285 vfs_unbusy(mp, false, NULL);
286 vfs_destroy(mp);
287 return (error);
288 }
289 mp->mnt_flag &= ~MNT_FORCE;
290 mutex_enter(&mountlist_lock);
291 CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list);
292 mutex_exit(&mountlist_lock);
293 ump = VFSTOUFS(mp);
294 fs = ump->um_fs;
295 memset(fs->fs_fsmnt, 0, sizeof(fs->fs_fsmnt));
296 (void)copystr(mp->mnt_stat.f_mntonname, fs->fs_fsmnt, MNAMELEN - 1, 0);
297 (void)ffs_statvfs(mp, &mp->mnt_stat);
298 vfs_unbusy(mp, false, NULL);
299 setrootfstime((time_t)fs->fs_time);
300 return (0);
301 }
302
303 /*
304 * VFS Operations.
305 *
306 * mount system call
307 */
308 int
309 ffs_mount(struct mount *mp, const char *path, void *data, size_t *data_len)
310 {
311 struct lwp *l = curlwp;
312 struct vnode *devvp = NULL;
313 struct ufs_args *args = data;
314 struct ufsmount *ump = NULL;
315 struct fs *fs;
316 int error = 0, flags, update;
317 mode_t accessmode;
318
319 if (*data_len < sizeof *args)
320 return EINVAL;
321
322 if (mp->mnt_flag & MNT_GETARGS) {
323 ump = VFSTOUFS(mp);
324 if (ump == NULL)
325 return EIO;
326 args->fspec = NULL;
327 *data_len = sizeof *args;
328 return 0;
329 }
330
331 update = mp->mnt_flag & MNT_UPDATE;
332
333 /* Check arguments */
334 if (args->fspec != NULL) {
335 /*
336 * Look up the name and verify that it's sane.
337 */
338 error = namei_simple_user(args->fspec,
339 NSM_FOLLOW_NOEMULROOT, &devvp);
340 if (error != 0)
341 return (error);
342
343 if (!update) {
344 /*
345 * Be sure this is a valid block device
346 */
347 if (devvp->v_type != VBLK)
348 error = ENOTBLK;
349 else if (bdevsw_lookup(devvp->v_rdev) == NULL)
350 error = ENXIO;
351 } else {
352 /*
353 * Be sure we're still naming the same device
354 * used for our initial mount
355 */
356 ump = VFSTOUFS(mp);
357 if (devvp != ump->um_devvp) {
358 if (devvp->v_rdev != ump->um_devvp->v_rdev)
359 error = EINVAL;
360 else {
361 vrele(devvp);
362 devvp = ump->um_devvp;
363 vref(devvp);
364 }
365 }
366 }
367 } else {
368 if (!update) {
369 /* New mounts must have a filename for the device */
370 return (EINVAL);
371 } else {
372 /* Use the extant mount */
373 ump = VFSTOUFS(mp);
374 devvp = ump->um_devvp;
375 vref(devvp);
376 }
377 }
378
379 /*
380 * If mount by non-root, then verify that user has necessary
381 * permissions on the device.
382 *
383 * Permission to update a mount is checked higher, so here we presume
384 * updating the mount is okay (for example, as far as securelevel goes)
385 * which leaves us with the normal check.
386 */
387 if (error == 0) {
388 accessmode = VREAD;
389 if (update ?
390 (mp->mnt_iflag & IMNT_WANTRDWR) != 0 :
391 (mp->mnt_flag & MNT_RDONLY) == 0)
392 accessmode |= VWRITE;
393 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
394 error = genfs_can_mount(devvp, accessmode, l->l_cred);
395 VOP_UNLOCK(devvp);
396 }
397
398 if (error) {
399 vrele(devvp);
400 return (error);
401 }
402
403 #ifdef WAPBL
404 /* WAPBL can only be enabled on a r/w mount. */
405 if ((mp->mnt_flag & MNT_RDONLY) && !(mp->mnt_iflag & IMNT_WANTRDWR)) {
406 mp->mnt_flag &= ~MNT_LOG;
407 }
408 #else /* !WAPBL */
409 mp->mnt_flag &= ~MNT_LOG;
410 #endif /* !WAPBL */
411
412 if (!update) {
413 int xflags;
414
415 if (mp->mnt_flag & MNT_RDONLY)
416 xflags = FREAD;
417 else
418 xflags = FREAD | FWRITE;
419 error = VOP_OPEN(devvp, xflags, FSCRED);
420 if (error)
421 goto fail;
422 error = ffs_mountfs(devvp, mp, l);
423 if (error) {
424 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
425 (void)VOP_CLOSE(devvp, xflags, NOCRED);
426 VOP_UNLOCK(devvp);
427 goto fail;
428 }
429
430 ump = VFSTOUFS(mp);
431 fs = ump->um_fs;
432 } else {
433 /*
434 * Update the mount.
435 */
436
437 /*
438 * The initial mount got a reference on this
439 * device, so drop the one obtained via
440 * namei(), above.
441 */
442 vrele(devvp);
443
444 ump = VFSTOUFS(mp);
445 fs = ump->um_fs;
446 if (fs->fs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) {
447 /*
448 * Changing from r/w to r/o
449 */
450 flags = WRITECLOSE;
451 if (mp->mnt_flag & MNT_FORCE)
452 flags |= FORCECLOSE;
453 error = ffs_flushfiles(mp, flags, l);
454 if (error == 0)
455 error = UFS_WAPBL_BEGIN(mp);
456 if (error == 0 &&
457 ffs_cgupdate(ump, MNT_WAIT) == 0 &&
458 fs->fs_clean & FS_WASCLEAN) {
459 if (mp->mnt_flag & MNT_SOFTDEP)
460 fs->fs_flags &= ~FS_DOSOFTDEP;
461 fs->fs_clean = FS_ISCLEAN;
462 (void) ffs_sbupdate(ump, MNT_WAIT);
463 }
464 if (error == 0)
465 UFS_WAPBL_END(mp);
466 if (error)
467 return (error);
468 }
469
470 #ifdef WAPBL
471 if ((mp->mnt_flag & MNT_LOG) == 0) {
472 error = ffs_wapbl_stop(mp, mp->mnt_flag & MNT_FORCE);
473 if (error)
474 return error;
475 }
476 #endif /* WAPBL */
477
478 if (fs->fs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) {
479 /*
480 * Finish change from r/w to r/o
481 */
482 fs->fs_ronly = 1;
483 fs->fs_fmod = 0;
484 }
485
486 if (mp->mnt_flag & MNT_RELOAD) {
487 error = ffs_reload(mp, l->l_cred, l);
488 if (error)
489 return (error);
490 }
491
492 if (fs->fs_ronly && (mp->mnt_iflag & IMNT_WANTRDWR)) {
493 /*
494 * Changing from read-only to read/write
495 */
496 fs->fs_ronly = 0;
497 fs->fs_clean <<= 1;
498 fs->fs_fmod = 1;
499 #ifdef WAPBL
500 if (fs->fs_flags & FS_DOWAPBL) {
501 printf("%s: replaying log to disk\n",
502 fs->fs_fsmnt);
503 KDASSERT(mp->mnt_wapbl_replay);
504 error = wapbl_replay_write(mp->mnt_wapbl_replay,
505 devvp);
506 if (error) {
507 return error;
508 }
509 wapbl_replay_stop(mp->mnt_wapbl_replay);
510 fs->fs_clean = FS_WASCLEAN;
511 }
512 #endif /* WAPBL */
513 if (fs->fs_snapinum[0] != 0)
514 ffs_snapshot_mount(mp);
515 }
516
517 #ifdef WAPBL
518 error = ffs_wapbl_start(mp);
519 if (error)
520 return error;
521 #endif /* WAPBL */
522
523 if (args->fspec == NULL)
524 return 0;
525 }
526
527 error = set_statvfs_info(path, UIO_USERSPACE, args->fspec,
528 UIO_USERSPACE, mp->mnt_op->vfs_name, mp, l);
529 if (error == 0)
530 (void)strncpy(fs->fs_fsmnt, mp->mnt_stat.f_mntonname,
531 sizeof(fs->fs_fsmnt));
532 fs->fs_flags &= ~FS_DOSOFTDEP;
533 if (fs->fs_fmod != 0) { /* XXX */
534 int err;
535
536 fs->fs_fmod = 0;
537 if (fs->fs_clean & FS_WASCLEAN)
538 fs->fs_time = time_second;
539 else {
540 printf("%s: file system not clean (fs_clean=%#x); "
541 "please fsck(8)\n", mp->mnt_stat.f_mntfromname,
542 fs->fs_clean);
543 printf("%s: lost blocks %" PRId64 " files %d\n",
544 mp->mnt_stat.f_mntfromname, fs->fs_pendingblocks,
545 fs->fs_pendinginodes);
546 }
547 err = UFS_WAPBL_BEGIN(mp);
548 if (err == 0) {
549 (void) ffs_cgupdate(ump, MNT_WAIT);
550 UFS_WAPBL_END(mp);
551 }
552 }
553 if ((mp->mnt_flag & MNT_SOFTDEP) != 0) {
554 printf("%s: `-o softdep' is no longer supported, "
555 "consider `-o log'\n", mp->mnt_stat.f_mntfromname);
556 mp->mnt_flag &= ~MNT_SOFTDEP;
557 }
558
559 return (error);
560
561 fail:
562 vrele(devvp);
563 return (error);
564 }
565
566 /*
567 * Reload all incore data for a filesystem (used after running fsck on
568 * the root filesystem and finding things to fix). The filesystem must
569 * be mounted read-only.
570 *
571 * Things to do to update the mount:
572 * 1) invalidate all cached meta-data.
573 * 2) re-read superblock from disk.
574 * 3) re-read summary information from disk.
575 * 4) invalidate all inactive vnodes.
576 * 5) invalidate all cached file data.
577 * 6) re-read inode data for all active vnodes.
578 */
579 int
580 ffs_reload(struct mount *mp, kauth_cred_t cred, struct lwp *l)
581 {
582 struct vnode *vp, *mvp, *devvp;
583 struct inode *ip;
584 void *space;
585 struct buf *bp;
586 struct fs *fs, *newfs;
587 struct partinfo dpart;
588 int i, bsize, blks, error;
589 int32_t *lp;
590 struct ufsmount *ump;
591 daddr_t sblockloc;
592
593 if ((mp->mnt_flag & MNT_RDONLY) == 0)
594 return (EINVAL);
595
596 ump = VFSTOUFS(mp);
597 /*
598 * Step 1: invalidate all cached meta-data.
599 */
600 devvp = ump->um_devvp;
601 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
602 error = vinvalbuf(devvp, 0, cred, l, 0, 0);
603 VOP_UNLOCK(devvp);
604 if (error)
605 panic("ffs_reload: dirty1");
606 /*
607 * Step 2: re-read superblock from disk.
608 */
609 fs = ump->um_fs;
610
611 /* XXX we don't handle possibility that superblock moved. */
612 error = bread(devvp, fs->fs_sblockloc / DEV_BSIZE, fs->fs_sbsize,
613 NOCRED, 0, &bp);
614 if (error) {
615 brelse(bp, 0);
616 return (error);
617 }
618 newfs = malloc(fs->fs_sbsize, M_UFSMNT, M_WAITOK);
619 memcpy(newfs, bp->b_data, fs->fs_sbsize);
620 #ifdef FFS_EI
621 if (ump->um_flags & UFS_NEEDSWAP) {
622 ffs_sb_swap((struct fs*)bp->b_data, newfs);
623 fs->fs_flags |= FS_SWAPPED;
624 } else
625 #endif
626 fs->fs_flags &= ~FS_SWAPPED;
627 if ((newfs->fs_magic != FS_UFS1_MAGIC &&
628 newfs->fs_magic != FS_UFS2_MAGIC)||
629 newfs->fs_bsize > MAXBSIZE ||
630 newfs->fs_bsize < sizeof(struct fs)) {
631 brelse(bp, 0);
632 free(newfs, M_UFSMNT);
633 return (EIO); /* XXX needs translation */
634 }
635 /* Store off old fs_sblockloc for fs_oldfscompat_read. */
636 sblockloc = fs->fs_sblockloc;
637 /*
638 * Copy pointer fields back into superblock before copying in XXX
639 * new superblock. These should really be in the ufsmount. XXX
640 * Note that important parameters (eg fs_ncg) are unchanged.
641 */
642 newfs->fs_csp = fs->fs_csp;
643 newfs->fs_maxcluster = fs->fs_maxcluster;
644 newfs->fs_contigdirs = fs->fs_contigdirs;
645 newfs->fs_ronly = fs->fs_ronly;
646 newfs->fs_active = fs->fs_active;
647 memcpy(fs, newfs, (u_int)fs->fs_sbsize);
648 brelse(bp, 0);
649 free(newfs, M_UFSMNT);
650
651 /* Recheck for apple UFS filesystem */
652 ump->um_flags &= ~UFS_ISAPPLEUFS;
653 /* First check to see if this is tagged as an Apple UFS filesystem
654 * in the disklabel
655 */
656 if ((VOP_IOCTL(devvp, DIOCGPART, &dpart, FREAD, cred) == 0) &&
657 (dpart.part->p_fstype == FS_APPLEUFS)) {
658 ump->um_flags |= UFS_ISAPPLEUFS;
659 }
660 #ifdef APPLE_UFS
661 else {
662 /* Manually look for an apple ufs label, and if a valid one
663 * is found, then treat it like an Apple UFS filesystem anyway
664 */
665 error = bread(devvp, (daddr_t)(APPLEUFS_LABEL_OFFSET / DEV_BSIZE),
666 APPLEUFS_LABEL_SIZE, cred, 0, &bp);
667 if (error) {
668 brelse(bp, 0);
669 return (error);
670 }
671 error = ffs_appleufs_validate(fs->fs_fsmnt,
672 (struct appleufslabel *)bp->b_data, NULL);
673 if (error == 0)
674 ump->um_flags |= UFS_ISAPPLEUFS;
675 brelse(bp, 0);
676 bp = NULL;
677 }
678 #else
679 if (ump->um_flags & UFS_ISAPPLEUFS)
680 return (EIO);
681 #endif
682
683 if (UFS_MPISAPPLEUFS(ump)) {
684 /* see comment about NeXT below */
685 ump->um_maxsymlinklen = APPLEUFS_MAXSYMLINKLEN;
686 ump->um_dirblksiz = APPLEUFS_DIRBLKSIZ;
687 mp->mnt_iflag |= IMNT_DTYPE;
688 } else {
689 ump->um_maxsymlinklen = fs->fs_maxsymlinklen;
690 ump->um_dirblksiz = DIRBLKSIZ;
691 if (ump->um_maxsymlinklen > 0)
692 mp->mnt_iflag |= IMNT_DTYPE;
693 else
694 mp->mnt_iflag &= ~IMNT_DTYPE;
695 }
696 ffs_oldfscompat_read(fs, ump, sblockloc);
697
698 mutex_enter(&ump->um_lock);
699 ump->um_maxfilesize = fs->fs_maxfilesize;
700 if (fs->fs_flags & ~(FS_KNOWN_FLAGS | FS_INTERNAL)) {
701 uprintf("%s: unknown ufs flags: 0x%08"PRIx32"%s\n",
702 mp->mnt_stat.f_mntonname, fs->fs_flags,
703 (mp->mnt_flag & MNT_FORCE) ? "" : ", not mounting");
704 if ((mp->mnt_flag & MNT_FORCE) == 0) {
705 mutex_exit(&ump->um_lock);
706 return (EINVAL);
707 }
708 }
709 if (fs->fs_pendingblocks != 0 || fs->fs_pendinginodes != 0) {
710 fs->fs_pendingblocks = 0;
711 fs->fs_pendinginodes = 0;
712 }
713 mutex_exit(&ump->um_lock);
714
715 ffs_statvfs(mp, &mp->mnt_stat);
716 /*
717 * Step 3: re-read summary information from disk.
718 */
719 blks = howmany(fs->fs_cssize, fs->fs_fsize);
720 space = fs->fs_csp;
721 for (i = 0; i < blks; i += fs->fs_frag) {
722 bsize = fs->fs_bsize;
723 if (i + fs->fs_frag > blks)
724 bsize = (blks - i) * fs->fs_fsize;
725 error = bread(devvp, fsbtodb(fs, fs->fs_csaddr + i), bsize,
726 NOCRED, 0, &bp);
727 if (error) {
728 brelse(bp, 0);
729 return (error);
730 }
731 #ifdef FFS_EI
732 if (UFS_FSNEEDSWAP(fs))
733 ffs_csum_swap((struct csum *)bp->b_data,
734 (struct csum *)space, bsize);
735 else
736 #endif
737 memcpy(space, bp->b_data, (size_t)bsize);
738 space = (char *)space + bsize;
739 brelse(bp, 0);
740 }
741 if (fs->fs_snapinum[0] != 0)
742 ffs_snapshot_mount(mp);
743 /*
744 * We no longer know anything about clusters per cylinder group.
745 */
746 if (fs->fs_contigsumsize > 0) {
747 lp = fs->fs_maxcluster;
748 for (i = 0; i < fs->fs_ncg; i++)
749 *lp++ = fs->fs_contigsumsize;
750 }
751
752 /* Allocate a marker vnode. */
753 if ((mvp = vnalloc(mp)) == NULL)
754 return ENOMEM;
755 /*
756 * NOTE: not using the TAILQ_FOREACH here since in this loop vgone()
757 * and vclean() can be called indirectly
758 */
759 mutex_enter(&mntvnode_lock);
760 loop:
761 for (vp = TAILQ_FIRST(&mp->mnt_vnodelist); vp; vp = vunmark(mvp)) {
762 vmark(mvp, vp);
763 if (vp->v_mount != mp || vismarker(vp))
764 continue;
765 /*
766 * Step 4: invalidate all inactive vnodes.
767 */
768 if (vrecycle(vp, &mntvnode_lock, l)) {
769 mutex_enter(&mntvnode_lock);
770 (void)vunmark(mvp);
771 goto loop;
772 }
773 /*
774 * Step 5: invalidate all cached file data.
775 */
776 mutex_enter(&vp->v_interlock);
777 mutex_exit(&mntvnode_lock);
778 if (vget(vp, LK_EXCLUSIVE)) {
779 (void)vunmark(mvp);
780 goto loop;
781 }
782 if (vinvalbuf(vp, 0, cred, l, 0, 0))
783 panic("ffs_reload: dirty2");
784 /*
785 * Step 6: re-read inode data for all active vnodes.
786 */
787 ip = VTOI(vp);
788 error = bread(devvp, fsbtodb(fs, ino_to_fsba(fs, ip->i_number)),
789 (int)fs->fs_bsize, NOCRED, 0, &bp);
790 if (error) {
791 brelse(bp, 0);
792 vput(vp);
793 (void)vunmark(mvp);
794 break;
795 }
796 ffs_load_inode(bp, ip, fs, ip->i_number);
797 brelse(bp, 0);
798 vput(vp);
799 mutex_enter(&mntvnode_lock);
800 }
801 mutex_exit(&mntvnode_lock);
802 vnfree(mvp);
803 return (error);
804 }
805
806 /*
807 * Possible superblock locations ordered from most to least likely.
808 */
809 static const int sblock_try[] = SBLOCKSEARCH;
810
811 /*
812 * Common code for mount and mountroot
813 */
814 int
815 ffs_mountfs(struct vnode *devvp, struct mount *mp, struct lwp *l)
816 {
817 struct ufsmount *ump;
818 struct buf *bp;
819 struct fs *fs;
820 dev_t dev;
821 struct partinfo dpart;
822 void *space;
823 daddr_t sblockloc, fsblockloc;
824 int blks, fstype;
825 int error, i, bsize, ronly, bset = 0;
826 #ifdef FFS_EI
827 int needswap = 0; /* keep gcc happy */
828 #endif
829 int32_t *lp;
830 kauth_cred_t cred;
831 u_int32_t sbsize = 8192; /* keep gcc happy*/
832 int32_t fsbsize;
833
834 dev = devvp->v_rdev;
835 cred = l ? l->l_cred : NOCRED;
836
837 /* Flush out any old buffers remaining from a previous use. */
838 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
839 error = vinvalbuf(devvp, V_SAVE, cred, l, 0, 0);
840 VOP_UNLOCK(devvp);
841 if (error)
842 return (error);
843
844 ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
845
846 bp = NULL;
847 ump = NULL;
848 fs = NULL;
849 sblockloc = 0;
850 fstype = 0;
851
852 error = fstrans_mount(mp);
853 if (error)
854 return error;
855
856 ump = malloc(sizeof *ump, M_UFSMNT, M_WAITOK);
857 memset(ump, 0, sizeof *ump);
858 mutex_init(&ump->um_lock, MUTEX_DEFAULT, IPL_NONE);
859 error = ffs_snapshot_init(ump);
860 if (error)
861 goto out;
862 ump->um_ops = &ffs_ufsops;
863
864 #ifdef WAPBL
865 sbagain:
866 #endif
867 /*
868 * Try reading the superblock in each of its possible locations.
869 */
870 for (i = 0; ; i++) {
871 if (bp != NULL) {
872 brelse(bp, BC_NOCACHE);
873 bp = NULL;
874 }
875 if (sblock_try[i] == -1) {
876 error = EINVAL;
877 fs = NULL;
878 goto out;
879 }
880 error = bread(devvp, sblock_try[i] / DEV_BSIZE, SBLOCKSIZE, cred,
881 0, &bp);
882 if (error) {
883 fs = NULL;
884 goto out;
885 }
886 fs = (struct fs*)bp->b_data;
887 fsblockloc = sblockloc = sblock_try[i];
888 if (fs->fs_magic == FS_UFS1_MAGIC) {
889 sbsize = fs->fs_sbsize;
890 fstype = UFS1;
891 fsbsize = fs->fs_bsize;
892 #ifdef FFS_EI
893 needswap = 0;
894 } else if (fs->fs_magic == bswap32(FS_UFS1_MAGIC)) {
895 sbsize = bswap32(fs->fs_sbsize);
896 fstype = UFS1;
897 fsbsize = bswap32(fs->fs_bsize);
898 needswap = 1;
899 #endif
900 } else if (fs->fs_magic == FS_UFS2_MAGIC) {
901 sbsize = fs->fs_sbsize;
902 fstype = UFS2;
903 fsbsize = fs->fs_bsize;
904 #ifdef FFS_EI
905 needswap = 0;
906 } else if (fs->fs_magic == bswap32(FS_UFS2_MAGIC)) {
907 sbsize = bswap32(fs->fs_sbsize);
908 fstype = UFS2;
909 fsbsize = bswap32(fs->fs_bsize);
910 needswap = 1;
911 #endif
912 } else
913 continue;
914
915
916 /* fs->fs_sblockloc isn't defined for old filesystems */
917 if (fstype == UFS1 && !(fs->fs_old_flags & FS_FLAGS_UPDATED)) {
918 if (sblockloc == SBLOCK_UFS2)
919 /*
920 * This is likely to be the first alternate
921 * in a filesystem with 64k blocks.
922 * Don't use it.
923 */
924 continue;
925 fsblockloc = sblockloc;
926 } else {
927 fsblockloc = fs->fs_sblockloc;
928 #ifdef FFS_EI
929 if (needswap)
930 fsblockloc = bswap64(fsblockloc);
931 #endif
932 }
933
934 /* Check we haven't found an alternate superblock */
935 if (fsblockloc != sblockloc)
936 continue;
937
938 /* Validate size of superblock */
939 if (sbsize > MAXBSIZE || sbsize < sizeof(struct fs))
940 continue;
941
942 /* Check that we can handle the file system blocksize */
943 if (fsbsize > MAXBSIZE) {
944 printf("ffs_mountfs: block size (%d) > MAXBSIZE (%d)\n",
945 fsbsize, MAXBSIZE);
946 continue;
947 }
948
949 /* Ok seems to be a good superblock */
950 break;
951 }
952
953 fs = malloc((u_long)sbsize, M_UFSMNT, M_WAITOK);
954 memcpy(fs, bp->b_data, sbsize);
955 ump->um_fs = fs;
956
957 #ifdef FFS_EI
958 if (needswap) {
959 ffs_sb_swap((struct fs*)bp->b_data, fs);
960 fs->fs_flags |= FS_SWAPPED;
961 } else
962 #endif
963 fs->fs_flags &= ~FS_SWAPPED;
964
965 #ifdef WAPBL
966 if ((mp->mnt_wapbl_replay == 0) && (fs->fs_flags & FS_DOWAPBL)) {
967 error = ffs_wapbl_replay_start(mp, fs, devvp);
968 if (error && (mp->mnt_flag & MNT_FORCE) == 0)
969 goto out;
970 if (!error) {
971 if (!ronly) {
972 /* XXX fsmnt may be stale. */
973 printf("%s: replaying log to disk\n",
974 fs->fs_fsmnt);
975 error = wapbl_replay_write(mp->mnt_wapbl_replay,
976 devvp);
977 if (error)
978 goto out;
979 wapbl_replay_stop(mp->mnt_wapbl_replay);
980 fs->fs_clean = FS_WASCLEAN;
981 } else {
982 /* XXX fsmnt may be stale */
983 printf("%s: replaying log to memory\n",
984 fs->fs_fsmnt);
985 }
986
987 /* Force a re-read of the superblock */
988 brelse(bp, BC_INVAL);
989 bp = NULL;
990 free(fs, M_UFSMNT);
991 fs = NULL;
992 goto sbagain;
993 }
994 }
995 #else /* !WAPBL */
996 if ((fs->fs_flags & FS_DOWAPBL) && (mp->mnt_flag & MNT_FORCE) == 0) {
997 error = EPERM;
998 goto out;
999 }
1000 #endif /* !WAPBL */
1001
1002 ffs_oldfscompat_read(fs, ump, sblockloc);
1003 ump->um_maxfilesize = fs->fs_maxfilesize;
1004
1005 if (fs->fs_flags & ~(FS_KNOWN_FLAGS | FS_INTERNAL)) {
1006 uprintf("%s: unknown ufs flags: 0x%08"PRIx32"%s\n",
1007 mp->mnt_stat.f_mntonname, fs->fs_flags,
1008 (mp->mnt_flag & MNT_FORCE) ? "" : ", not mounting");
1009 if ((mp->mnt_flag & MNT_FORCE) == 0) {
1010 error = EINVAL;
1011 goto out;
1012 }
1013 }
1014
1015 if (fs->fs_pendingblocks != 0 || fs->fs_pendinginodes != 0) {
1016 fs->fs_pendingblocks = 0;
1017 fs->fs_pendinginodes = 0;
1018 }
1019
1020 ump->um_fstype = fstype;
1021 if (fs->fs_sbsize < SBLOCKSIZE)
1022 brelse(bp, BC_INVAL);
1023 else
1024 brelse(bp, 0);
1025 bp = NULL;
1026
1027 /* First check to see if this is tagged as an Apple UFS filesystem
1028 * in the disklabel
1029 */
1030 if ((VOP_IOCTL(devvp, DIOCGPART, &dpart, FREAD, cred) == 0) &&
1031 (dpart.part->p_fstype == FS_APPLEUFS)) {
1032 ump->um_flags |= UFS_ISAPPLEUFS;
1033 }
1034 #ifdef APPLE_UFS
1035 else {
1036 /* Manually look for an apple ufs label, and if a valid one
1037 * is found, then treat it like an Apple UFS filesystem anyway
1038 */
1039 error = bread(devvp, (daddr_t)(APPLEUFS_LABEL_OFFSET / DEV_BSIZE),
1040 APPLEUFS_LABEL_SIZE, cred, 0, &bp);
1041 if (error)
1042 goto out;
1043 error = ffs_appleufs_validate(fs->fs_fsmnt,
1044 (struct appleufslabel *)bp->b_data, NULL);
1045 if (error == 0) {
1046 ump->um_flags |= UFS_ISAPPLEUFS;
1047 }
1048 brelse(bp, 0);
1049 bp = NULL;
1050 }
1051 #else
1052 if (ump->um_flags & UFS_ISAPPLEUFS) {
1053 error = EINVAL;
1054 goto out;
1055 }
1056 #endif
1057
1058 #if 0
1059 /*
1060 * XXX This code changes the behaviour of mounting dirty filesystems, to
1061 * XXX require "mount -f ..." to mount them. This doesn't match what
1062 * XXX mount(8) describes and is disabled for now.
1063 */
1064 /*
1065 * If the file system is not clean, don't allow it to be mounted
1066 * unless MNT_FORCE is specified. (Note: MNT_FORCE is always set
1067 * for the root file system.)
1068 */
1069 if (fs->fs_flags & FS_DOWAPBL) {
1070 /*
1071 * wapbl normally expects to be FS_WASCLEAN when the FS_DOWAPBL
1072 * bit is set, although there's a window in unmount where it
1073 * could be FS_ISCLEAN
1074 */
1075 if ((mp->mnt_flag & MNT_FORCE) == 0 &&
1076 (fs->fs_clean & (FS_WASCLEAN | FS_ISCLEAN)) == 0) {
1077 error = EPERM;
1078 goto out;
1079 }
1080 } else
1081 if ((fs->fs_clean & FS_ISCLEAN) == 0 &&
1082 (mp->mnt_flag & MNT_FORCE) == 0) {
1083 error = EPERM;
1084 goto out;
1085 }
1086 #endif
1087
1088 /*
1089 * verify that we can access the last block in the fs
1090 * if we're mounting read/write.
1091 */
1092
1093 if (!ronly) {
1094 error = bread(devvp, fsbtodb(fs, fs->fs_size - 1), fs->fs_fsize,
1095 cred, 0, &bp);
1096 if (bp->b_bcount != fs->fs_fsize)
1097 error = EINVAL;
1098 if (error) {
1099 bset = BC_INVAL;
1100 goto out;
1101 }
1102 brelse(bp, BC_INVAL);
1103 bp = NULL;
1104 }
1105
1106 fs->fs_ronly = ronly;
1107 /* Don't bump fs_clean if we're replaying journal */
1108 if (!((fs->fs_flags & FS_DOWAPBL) && (fs->fs_clean & FS_WASCLEAN)))
1109 if (ronly == 0) {
1110 fs->fs_clean <<= 1;
1111 fs->fs_fmod = 1;
1112 }
1113 bsize = fs->fs_cssize;
1114 blks = howmany(bsize, fs->fs_fsize);
1115 if (fs->fs_contigsumsize > 0)
1116 bsize += fs->fs_ncg * sizeof(int32_t);
1117 bsize += fs->fs_ncg * sizeof(*fs->fs_contigdirs);
1118 space = malloc((u_long)bsize, M_UFSMNT, M_WAITOK);
1119 fs->fs_csp = space;
1120 for (i = 0; i < blks; i += fs->fs_frag) {
1121 bsize = fs->fs_bsize;
1122 if (i + fs->fs_frag > blks)
1123 bsize = (blks - i) * fs->fs_fsize;
1124 error = bread(devvp, fsbtodb(fs, fs->fs_csaddr + i), bsize,
1125 cred, 0, &bp);
1126 if (error) {
1127 free(fs->fs_csp, M_UFSMNT);
1128 goto out;
1129 }
1130 #ifdef FFS_EI
1131 if (needswap)
1132 ffs_csum_swap((struct csum *)bp->b_data,
1133 (struct csum *)space, bsize);
1134 else
1135 #endif
1136 memcpy(space, bp->b_data, (u_int)bsize);
1137
1138 space = (char *)space + bsize;
1139 brelse(bp, 0);
1140 bp = NULL;
1141 }
1142 if (fs->fs_contigsumsize > 0) {
1143 fs->fs_maxcluster = lp = space;
1144 for (i = 0; i < fs->fs_ncg; i++)
1145 *lp++ = fs->fs_contigsumsize;
1146 space = lp;
1147 }
1148 bsize = fs->fs_ncg * sizeof(*fs->fs_contigdirs);
1149 fs->fs_contigdirs = space;
1150 space = (char *)space + bsize;
1151 memset(fs->fs_contigdirs, 0, bsize);
1152 /* Compatibility for old filesystems - XXX */
1153 if (fs->fs_avgfilesize <= 0)
1154 fs->fs_avgfilesize = AVFILESIZ;
1155 if (fs->fs_avgfpdir <= 0)
1156 fs->fs_avgfpdir = AFPDIR;
1157 fs->fs_active = NULL;
1158 mp->mnt_data = ump;
1159 mp->mnt_stat.f_fsidx.__fsid_val[0] = (long)dev;
1160 mp->mnt_stat.f_fsidx.__fsid_val[1] = makefstype(MOUNT_FFS);
1161 mp->mnt_stat.f_fsid = mp->mnt_stat.f_fsidx.__fsid_val[0];
1162 mp->mnt_stat.f_namemax = FFS_MAXNAMLEN;
1163 if (UFS_MPISAPPLEUFS(ump)) {
1164 /* NeXT used to keep short symlinks in the inode even
1165 * when using FS_42INODEFMT. In that case fs->fs_maxsymlinklen
1166 * is probably -1, but we still need to be able to identify
1167 * short symlinks.
1168 */
1169 ump->um_maxsymlinklen = APPLEUFS_MAXSYMLINKLEN;
1170 ump->um_dirblksiz = APPLEUFS_DIRBLKSIZ;
1171 mp->mnt_iflag |= IMNT_DTYPE;
1172 } else {
1173 ump->um_maxsymlinklen = fs->fs_maxsymlinklen;
1174 ump->um_dirblksiz = DIRBLKSIZ;
1175 if (ump->um_maxsymlinklen > 0)
1176 mp->mnt_iflag |= IMNT_DTYPE;
1177 else
1178 mp->mnt_iflag &= ~IMNT_DTYPE;
1179 }
1180 mp->mnt_fs_bshift = fs->fs_bshift;
1181 mp->mnt_dev_bshift = DEV_BSHIFT; /* XXX */
1182 mp->mnt_flag |= MNT_LOCAL;
1183 mp->mnt_iflag |= IMNT_MPSAFE;
1184 #ifdef XIP
1185 if ((mp->mnt_flag & MNT_RDONLY) == 0)
1186 printf("XIP needs read-only mount\n");
1187 else {
1188 void *physseg = NULL;
1189
1190 if ((VOP_IOCTL(devvp, DIOCGPHYSSEG, &physseg, FREAD,
1191 cred) == 0) &&
1192 physseg != NULL) {
1193 mp->mnt_iflag |= IMNT_XIP;
1194 devvp->v_physseg = physseg;
1195 } else
1196 printf("device doesn't support DIOCGPHYSSEG\n");
1197 }
1198 #endif
1199 #ifdef FFS_EI
1200 if (needswap)
1201 ump->um_flags |= UFS_NEEDSWAP;
1202 #endif
1203 ump->um_mountp = mp;
1204 ump->um_dev = dev;
1205 ump->um_devvp = devvp;
1206 ump->um_nindir = fs->fs_nindir;
1207 ump->um_lognindir = ffs(fs->fs_nindir) - 1;
1208 ump->um_bptrtodb = fs->fs_fshift - DEV_BSHIFT;
1209 ump->um_seqinc = fs->fs_frag;
1210 for (i = 0; i < MAXQUOTAS; i++)
1211 ump->um_quotas[i] = NULLVP;
1212 devvp->v_specmountpoint = mp;
1213 if (ronly == 0 && fs->fs_snapinum[0] != 0)
1214 ffs_snapshot_mount(mp);
1215
1216 #ifdef WAPBL
1217 if (!ronly) {
1218 KDASSERT(fs->fs_ronly == 0);
1219 /*
1220 * ffs_wapbl_start() needs mp->mnt_stat initialised if it
1221 * needs to create a new log file in-filesystem.
1222 */
1223 ffs_statvfs(mp, &mp->mnt_stat);
1224
1225 error = ffs_wapbl_start(mp);
1226 if (error) {
1227 free(fs->fs_csp, M_UFSMNT);
1228 goto out;
1229 }
1230 }
1231 #endif /* WAPBL */
1232 #ifdef UFS_EXTATTR
1233 /*
1234 * Initialize file-backed extended attributes on UFS1 file
1235 * systems.
1236 */
1237 if (ump->um_fstype == UFS1) {
1238 ufs_extattr_uepm_init(&ump->um_extattr);
1239 #ifdef UFS_EXTATTR_AUTOSTART
1240 /*
1241 * XXX Just ignore errors. Not clear that we should
1242 * XXX fail the mount in this case.
1243 */
1244 (void) ufs_extattr_autostart(mp, l);
1245 #endif
1246 }
1247 #endif /* UFS_EXTATTR */
1248 return (0);
1249 out:
1250 #ifdef WAPBL
1251 if (mp->mnt_wapbl_replay) {
1252 wapbl_replay_stop(mp->mnt_wapbl_replay);
1253 wapbl_replay_free(mp->mnt_wapbl_replay);
1254 mp->mnt_wapbl_replay = 0;
1255 }
1256 #endif
1257
1258 fstrans_unmount(mp);
1259 if (fs)
1260 free(fs, M_UFSMNT);
1261 devvp->v_specmountpoint = NULL;
1262 if (bp)
1263 brelse(bp, bset);
1264 if (ump) {
1265 if (ump->um_oldfscompat)
1266 free(ump->um_oldfscompat, M_UFSMNT);
1267 mutex_destroy(&ump->um_lock);
1268 free(ump, M_UFSMNT);
1269 mp->mnt_data = NULL;
1270 }
1271 return (error);
1272 }
1273
1274 /*
1275 * Sanity checks for loading old filesystem superblocks.
1276 * See ffs_oldfscompat_write below for unwound actions.
1277 *
1278 * XXX - Parts get retired eventually.
1279 * Unfortunately new bits get added.
1280 */
1281 static void
1282 ffs_oldfscompat_read(struct fs *fs, struct ufsmount *ump, daddr_t sblockloc)
1283 {
1284 off_t maxfilesize;
1285 int32_t *extrasave;
1286
1287 if ((fs->fs_magic != FS_UFS1_MAGIC) ||
1288 (fs->fs_old_flags & FS_FLAGS_UPDATED))
1289 return;
1290
1291 if (!ump->um_oldfscompat)
1292 ump->um_oldfscompat = malloc(512 + 3*sizeof(int32_t),
1293 M_UFSMNT, M_WAITOK);
1294
1295 memcpy(ump->um_oldfscompat, &fs->fs_old_postbl_start, 512);
1296 extrasave = ump->um_oldfscompat;
1297 extrasave += 512/sizeof(int32_t);
1298 extrasave[0] = fs->fs_old_npsect;
1299 extrasave[1] = fs->fs_old_interleave;
1300 extrasave[2] = fs->fs_old_trackskew;
1301
1302 /* These fields will be overwritten by their
1303 * original values in fs_oldfscompat_write, so it is harmless
1304 * to modify them here.
1305 */
1306 fs->fs_cstotal.cs_ndir = fs->fs_old_cstotal.cs_ndir;
1307 fs->fs_cstotal.cs_nbfree = fs->fs_old_cstotal.cs_nbfree;
1308 fs->fs_cstotal.cs_nifree = fs->fs_old_cstotal.cs_nifree;
1309 fs->fs_cstotal.cs_nffree = fs->fs_old_cstotal.cs_nffree;
1310
1311 fs->fs_maxbsize = fs->fs_bsize;
1312 fs->fs_time = fs->fs_old_time;
1313 fs->fs_size = fs->fs_old_size;
1314 fs->fs_dsize = fs->fs_old_dsize;
1315 fs->fs_csaddr = fs->fs_old_csaddr;
1316 fs->fs_sblockloc = sblockloc;
1317
1318 fs->fs_flags = fs->fs_old_flags | (fs->fs_flags & FS_INTERNAL);
1319
1320 if (fs->fs_old_postblformat == FS_42POSTBLFMT) {
1321 fs->fs_old_nrpos = 8;
1322 fs->fs_old_npsect = fs->fs_old_nsect;
1323 fs->fs_old_interleave = 1;
1324 fs->fs_old_trackskew = 0;
1325 }
1326
1327 if (fs->fs_old_inodefmt < FS_44INODEFMT) {
1328 fs->fs_maxfilesize = (u_quad_t) 1LL << 39;
1329 fs->fs_qbmask = ~fs->fs_bmask;
1330 fs->fs_qfmask = ~fs->fs_fmask;
1331 }
1332
1333 maxfilesize = (u_int64_t)0x80000000 * fs->fs_bsize - 1;
1334 if (fs->fs_maxfilesize > maxfilesize)
1335 fs->fs_maxfilesize = maxfilesize;
1336
1337 /* Compatibility for old filesystems */
1338 if (fs->fs_avgfilesize <= 0)
1339 fs->fs_avgfilesize = AVFILESIZ;
1340 if (fs->fs_avgfpdir <= 0)
1341 fs->fs_avgfpdir = AFPDIR;
1342
1343 #if 0
1344 if (bigcgs) {
1345 fs->fs_save_cgsize = fs->fs_cgsize;
1346 fs->fs_cgsize = fs->fs_bsize;
1347 }
1348 #endif
1349 }
1350
1351 /*
1352 * Unwinding superblock updates for old filesystems.
1353 * See ffs_oldfscompat_read above for details.
1354 *
1355 * XXX - Parts get retired eventually.
1356 * Unfortunately new bits get added.
1357 */
1358 static void
1359 ffs_oldfscompat_write(struct fs *fs, struct ufsmount *ump)
1360 {
1361 int32_t *extrasave;
1362
1363 if ((fs->fs_magic != FS_UFS1_MAGIC) ||
1364 (fs->fs_old_flags & FS_FLAGS_UPDATED))
1365 return;
1366
1367 fs->fs_old_time = fs->fs_time;
1368 fs->fs_old_cstotal.cs_ndir = fs->fs_cstotal.cs_ndir;
1369 fs->fs_old_cstotal.cs_nbfree = fs->fs_cstotal.cs_nbfree;
1370 fs->fs_old_cstotal.cs_nifree = fs->fs_cstotal.cs_nifree;
1371 fs->fs_old_cstotal.cs_nffree = fs->fs_cstotal.cs_nffree;
1372 fs->fs_old_flags = fs->fs_flags;
1373
1374 #if 0
1375 if (bigcgs) {
1376 fs->fs_cgsize = fs->fs_save_cgsize;
1377 }
1378 #endif
1379
1380 memcpy(&fs->fs_old_postbl_start, ump->um_oldfscompat, 512);
1381 extrasave = ump->um_oldfscompat;
1382 extrasave += 512/sizeof(int32_t);
1383 fs->fs_old_npsect = extrasave[0];
1384 fs->fs_old_interleave = extrasave[1];
1385 fs->fs_old_trackskew = extrasave[2];
1386
1387 }
1388
1389 /*
1390 * unmount vfs operation
1391 */
1392 int
1393 ffs_unmount(struct mount *mp, int mntflags)
1394 {
1395 struct lwp *l = curlwp;
1396 struct ufsmount *ump = VFSTOUFS(mp);
1397 struct fs *fs = ump->um_fs;
1398 int error, flags;
1399 #ifdef WAPBL
1400 extern int doforce;
1401 #endif
1402
1403 flags = 0;
1404 if (mntflags & MNT_FORCE)
1405 flags |= FORCECLOSE;
1406 if ((error = ffs_flushfiles(mp, flags, l)) != 0)
1407 return (error);
1408 error = UFS_WAPBL_BEGIN(mp);
1409 if (error == 0)
1410 if (fs->fs_ronly == 0 &&
1411 ffs_cgupdate(ump, MNT_WAIT) == 0 &&
1412 fs->fs_clean & FS_WASCLEAN) {
1413 fs->fs_clean = FS_ISCLEAN;
1414 fs->fs_fmod = 0;
1415 (void) ffs_sbupdate(ump, MNT_WAIT);
1416 }
1417 if (error == 0)
1418 UFS_WAPBL_END(mp);
1419 #ifdef WAPBL
1420 KASSERT(!(mp->mnt_wapbl_replay && mp->mnt_wapbl));
1421 if (mp->mnt_wapbl_replay) {
1422 KDASSERT(fs->fs_ronly);
1423 wapbl_replay_stop(mp->mnt_wapbl_replay);
1424 wapbl_replay_free(mp->mnt_wapbl_replay);
1425 mp->mnt_wapbl_replay = 0;
1426 }
1427 error = ffs_wapbl_stop(mp, doforce && (mntflags & MNT_FORCE));
1428 if (error) {
1429 return error;
1430 }
1431 #endif /* WAPBL */
1432 #ifdef UFS_EXTATTR
1433 if (ump->um_fstype == UFS1) {
1434 ufs_extattr_stop(mp, l);
1435 ufs_extattr_uepm_destroy(&ump->um_extattr);
1436 }
1437 #endif /* UFS_EXTATTR */
1438
1439 if (ump->um_devvp->v_type != VBAD)
1440 ump->um_devvp->v_specmountpoint = NULL;
1441 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
1442 (void)VOP_CLOSE(ump->um_devvp, fs->fs_ronly ? FREAD : FREAD | FWRITE,
1443 NOCRED);
1444 vput(ump->um_devvp);
1445 free(fs->fs_csp, M_UFSMNT);
1446 free(fs, M_UFSMNT);
1447 if (ump->um_oldfscompat != NULL)
1448 free(ump->um_oldfscompat, M_UFSMNT);
1449 mutex_destroy(&ump->um_lock);
1450 ffs_snapshot_fini(ump);
1451 free(ump, M_UFSMNT);
1452 mp->mnt_data = NULL;
1453 mp->mnt_flag &= ~MNT_LOCAL;
1454 fstrans_unmount(mp);
1455 return (0);
1456 }
1457
1458 /*
1459 * Flush out all the files in a filesystem.
1460 */
1461 int
1462 ffs_flushfiles(struct mount *mp, int flags, struct lwp *l)
1463 {
1464 extern int doforce;
1465 struct ufsmount *ump;
1466 int error;
1467
1468 if (!doforce)
1469 flags &= ~FORCECLOSE;
1470 ump = VFSTOUFS(mp);
1471 #ifdef QUOTA
1472 if (mp->mnt_flag & MNT_QUOTA) {
1473 int i;
1474 if ((error = vflush(mp, NULLVP, SKIPSYSTEM | flags)) != 0)
1475 return (error);
1476 for (i = 0; i < MAXQUOTAS; i++) {
1477 if (ump->um_quotas[i] == NULLVP)
1478 continue;
1479 quotaoff(l, mp, i);
1480 }
1481 /*
1482 * Here we fall through to vflush again to ensure
1483 * that we have gotten rid of all the system vnodes.
1484 */
1485 }
1486 #endif
1487 if ((error = vflush(mp, 0, SKIPSYSTEM | flags)) != 0)
1488 return (error);
1489 ffs_snapshot_unmount(mp);
1490 /*
1491 * Flush all the files.
1492 */
1493 error = vflush(mp, NULLVP, flags);
1494 if (error)
1495 return (error);
1496 /*
1497 * Flush filesystem metadata.
1498 */
1499 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
1500 error = VOP_FSYNC(ump->um_devvp, l->l_cred, FSYNC_WAIT, 0, 0);
1501 VOP_UNLOCK(ump->um_devvp);
1502 if (flags & FORCECLOSE) /* XXXDBJ */
1503 error = 0;
1504
1505 #ifdef WAPBL
1506 if (error)
1507 return error;
1508 if (mp->mnt_wapbl) {
1509 error = wapbl_flush(mp->mnt_wapbl, 1);
1510 if (flags & FORCECLOSE)
1511 error = 0;
1512 }
1513 #endif
1514
1515 return (error);
1516 }
1517
1518 /*
1519 * Get file system statistics.
1520 */
1521 int
1522 ffs_statvfs(struct mount *mp, struct statvfs *sbp)
1523 {
1524 struct ufsmount *ump;
1525 struct fs *fs;
1526
1527 ump = VFSTOUFS(mp);
1528 fs = ump->um_fs;
1529 mutex_enter(&ump->um_lock);
1530 sbp->f_bsize = fs->fs_bsize;
1531 sbp->f_frsize = fs->fs_fsize;
1532 sbp->f_iosize = fs->fs_bsize;
1533 sbp->f_blocks = fs->fs_dsize;
1534 sbp->f_bfree = blkstofrags(fs, fs->fs_cstotal.cs_nbfree) +
1535 fs->fs_cstotal.cs_nffree + dbtofsb(fs, fs->fs_pendingblocks);
1536 sbp->f_bresvd = ((u_int64_t) fs->fs_dsize * (u_int64_t)
1537 fs->fs_minfree) / (u_int64_t) 100;
1538 if (sbp->f_bfree > sbp->f_bresvd)
1539 sbp->f_bavail = sbp->f_bfree - sbp->f_bresvd;
1540 else
1541 sbp->f_bavail = 0;
1542 sbp->f_files = fs->fs_ncg * fs->fs_ipg - ROOTINO;
1543 sbp->f_ffree = fs->fs_cstotal.cs_nifree + fs->fs_pendinginodes;
1544 sbp->f_favail = sbp->f_ffree;
1545 sbp->f_fresvd = 0;
1546 mutex_exit(&ump->um_lock);
1547 copy_statvfs_info(sbp, mp);
1548
1549 return (0);
1550 }
1551
1552 /*
1553 * Go through the disk queues to initiate sandbagged IO;
1554 * go through the inodes to write those that have been modified;
1555 * initiate the writing of the super block if it has been modified.
1556 *
1557 * Note: we are always called with the filesystem marked `MPBUSY'.
1558 */
1559 int
1560 ffs_sync(struct mount *mp, int waitfor, kauth_cred_t cred)
1561 {
1562 struct vnode *vp, *mvp, *nvp;
1563 struct inode *ip;
1564 struct ufsmount *ump = VFSTOUFS(mp);
1565 struct fs *fs;
1566 int lk_flags, error, allerror = 0;
1567 bool is_suspending;
1568
1569 fs = ump->um_fs;
1570 if (fs->fs_fmod != 0 && fs->fs_ronly != 0) { /* XXX */
1571 printf("fs = %s\n", fs->fs_fsmnt);
1572 panic("update: rofs mod");
1573 }
1574
1575 /* Allocate a marker vnode. */
1576 if ((mvp = vnalloc(mp)) == NULL)
1577 return (ENOMEM);
1578
1579 fstrans_start(mp, FSTRANS_SHARED);
1580 is_suspending = (fstrans_getstate(mp) == FSTRANS_SUSPENDING);
1581 /*
1582 * We can't lock vnodes while the file system is suspending because
1583 * threads waiting on fstrans may have locked vnodes.
1584 */
1585 if (is_suspending)
1586 lk_flags = 0;
1587 else
1588 lk_flags = LK_EXCLUSIVE | LK_NOWAIT;
1589 /*
1590 * Write back each (modified) inode.
1591 */
1592 mutex_enter(&mntvnode_lock);
1593 loop:
1594 /*
1595 * NOTE: not using the TAILQ_FOREACH here since in this loop vgone()
1596 * and vclean() can be called indirectly
1597 */
1598 for (vp = TAILQ_FIRST(&mp->mnt_vnodelist); vp; vp = nvp) {
1599 nvp = TAILQ_NEXT(vp, v_mntvnodes);
1600 /*
1601 * If the vnode that we are about to sync is no longer
1602 * associated with this mount point, start over.
1603 */
1604 if (vp->v_mount != mp)
1605 goto loop;
1606 /*
1607 * Don't interfere with concurrent scans of this FS.
1608 */
1609 if (vismarker(vp))
1610 continue;
1611 mutex_enter(&vp->v_interlock);
1612 ip = VTOI(vp);
1613
1614 /*
1615 * Skip the vnode/inode if inaccessible.
1616 */
1617 if (ip == NULL || (vp->v_iflag & (VI_XLOCK | VI_CLEAN)) != 0 ||
1618 vp->v_type == VNON) {
1619 mutex_exit(&vp->v_interlock);
1620 continue;
1621 }
1622
1623 /*
1624 * We deliberately update inode times here. This will
1625 * prevent a massive queue of updates accumulating, only
1626 * to be handled by a call to unmount.
1627 *
1628 * XXX It would be better to have the syncer trickle these
1629 * out. Adjustment needed to allow registering vnodes for
1630 * sync when the vnode is clean, but the inode dirty. Or
1631 * have ufs itself trickle out inode updates.
1632 *
1633 * If doing a lazy sync, we don't care about metadata or
1634 * data updates, because they are handled by each vnode's
1635 * synclist entry. In this case we are only interested in
1636 * writing back modified inodes.
1637 */
1638 if ((ip->i_flag & (IN_ACCESS | IN_CHANGE | IN_UPDATE |
1639 IN_MODIFY | IN_MODIFIED | IN_ACCESSED)) == 0 &&
1640 (waitfor == MNT_LAZY || (LIST_EMPTY(&vp->v_dirtyblkhd) &&
1641 UVM_OBJ_IS_CLEAN(&vp->v_uobj)))) {
1642 mutex_exit(&vp->v_interlock);
1643 continue;
1644 }
1645 if (vp->v_type == VBLK && is_suspending) {
1646 mutex_exit(&vp->v_interlock);
1647 continue;
1648 }
1649 vmark(mvp, vp);
1650 mutex_exit(&mntvnode_lock);
1651 error = vget(vp, lk_flags);
1652 if (error) {
1653 mutex_enter(&mntvnode_lock);
1654 nvp = vunmark(mvp);
1655 if (error == ENOENT) {
1656 goto loop;
1657 }
1658 continue;
1659 }
1660 if (waitfor == MNT_LAZY) {
1661 error = UFS_WAPBL_BEGIN(vp->v_mount);
1662 if (!error) {
1663 error = ffs_update(vp, NULL, NULL,
1664 UPDATE_CLOSE);
1665 UFS_WAPBL_END(vp->v_mount);
1666 }
1667 } else {
1668 error = VOP_FSYNC(vp, cred, FSYNC_NOLOG |
1669 (waitfor == MNT_WAIT ? FSYNC_WAIT : 0), 0, 0);
1670 }
1671 if (error)
1672 allerror = error;
1673 if (is_suspending)
1674 vrele(vp);
1675 else
1676 vput(vp);
1677 mutex_enter(&mntvnode_lock);
1678 nvp = vunmark(mvp);
1679 }
1680 mutex_exit(&mntvnode_lock);
1681 /*
1682 * Force stale file system control information to be flushed.
1683 */
1684 if (waitfor != MNT_LAZY && (ump->um_devvp->v_numoutput > 0 ||
1685 !LIST_EMPTY(&ump->um_devvp->v_dirtyblkhd))) {
1686 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
1687 if ((error = VOP_FSYNC(ump->um_devvp, cred,
1688 (waitfor == MNT_WAIT ? FSYNC_WAIT : 0) | FSYNC_NOLOG,
1689 0, 0)) != 0)
1690 allerror = error;
1691 VOP_UNLOCK(ump->um_devvp);
1692 if (allerror == 0 && waitfor == MNT_WAIT && !mp->mnt_wapbl) {
1693 mutex_enter(&mntvnode_lock);
1694 goto loop;
1695 }
1696 }
1697 #ifdef QUOTA
1698 qsync(mp);
1699 #endif
1700 /*
1701 * Write back modified superblock.
1702 */
1703 if (fs->fs_fmod != 0) {
1704 fs->fs_fmod = 0;
1705 fs->fs_time = time_second;
1706 error = UFS_WAPBL_BEGIN(mp);
1707 if (error)
1708 allerror = error;
1709 else {
1710 if ((error = ffs_cgupdate(ump, waitfor)))
1711 allerror = error;
1712 UFS_WAPBL_END(mp);
1713 }
1714 }
1715
1716 #ifdef WAPBL
1717 if (mp->mnt_wapbl) {
1718 error = wapbl_flush(mp->mnt_wapbl, 0);
1719 if (error)
1720 allerror = error;
1721 }
1722 #endif
1723
1724 fstrans_done(mp);
1725 vnfree(mvp);
1726 return (allerror);
1727 }
1728
1729 /*
1730 * Look up a FFS dinode number to find its incore vnode, otherwise read it
1731 * in from disk. If it is in core, wait for the lock bit to clear, then
1732 * return the inode locked. Detection and handling of mount points must be
1733 * done by the calling routine.
1734 */
1735 int
1736 ffs_vget(struct mount *mp, ino_t ino, struct vnode **vpp)
1737 {
1738 struct fs *fs;
1739 struct inode *ip;
1740 struct ufsmount *ump;
1741 struct buf *bp;
1742 struct vnode *vp;
1743 dev_t dev;
1744 int error;
1745
1746 ump = VFSTOUFS(mp);
1747 dev = ump->um_dev;
1748
1749 retry:
1750 if ((*vpp = ufs_ihashget(dev, ino, LK_EXCLUSIVE)) != NULL)
1751 return (0);
1752
1753 /* Allocate a new vnode/inode. */
1754 if ((error = getnewvnode(VT_UFS, mp, ffs_vnodeop_p, &vp)) != 0) {
1755 *vpp = NULL;
1756 return (error);
1757 }
1758 ip = pool_cache_get(ffs_inode_cache, PR_WAITOK);
1759
1760 /*
1761 * If someone beat us to it, put back the freshly allocated
1762 * vnode/inode pair and retry.
1763 */
1764 mutex_enter(&ufs_hashlock);
1765 if (ufs_ihashget(dev, ino, 0) != NULL) {
1766 mutex_exit(&ufs_hashlock);
1767 ungetnewvnode(vp);
1768 pool_cache_put(ffs_inode_cache, ip);
1769 goto retry;
1770 }
1771
1772 vp->v_vflag |= VV_LOCKSWORK;
1773
1774 /*
1775 * XXX MFS ends up here, too, to allocate an inode. Should we
1776 * XXX create another pool for MFS inodes?
1777 */
1778
1779 memset(ip, 0, sizeof(struct inode));
1780 vp->v_data = ip;
1781 ip->i_vnode = vp;
1782 ip->i_ump = ump;
1783 ip->i_fs = fs = ump->um_fs;
1784 ip->i_dev = dev;
1785 ip->i_number = ino;
1786 #ifdef QUOTA
1787 ufsquota_init(ip);
1788 #endif
1789
1790 /*
1791 * Initialize genfs node, we might proceed to destroy it in
1792 * error branches.
1793 */
1794 genfs_node_init(vp, &ffs_genfsops);
1795
1796 /*
1797 * Put it onto its hash chain and lock it so that other requests for
1798 * this inode will block if they arrive while we are sleeping waiting
1799 * for old data structures to be purged or for the contents of the
1800 * disk portion of this inode to be read.
1801 */
1802
1803 ufs_ihashins(ip);
1804 mutex_exit(&ufs_hashlock);
1805
1806 /* Read in the disk contents for the inode, copy into the inode. */
1807 error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)),
1808 (int)fs->fs_bsize, NOCRED, 0, &bp);
1809 if (error) {
1810
1811 /*
1812 * The inode does not contain anything useful, so it would
1813 * be misleading to leave it on its hash chain. With mode
1814 * still zero, it will be unlinked and returned to the free
1815 * list by vput().
1816 */
1817
1818 vput(vp);
1819 brelse(bp, 0);
1820 *vpp = NULL;
1821 return (error);
1822 }
1823 if (ip->i_ump->um_fstype == UFS1)
1824 ip->i_din.ffs1_din = pool_cache_get(ffs_dinode1_cache,
1825 PR_WAITOK);
1826 else
1827 ip->i_din.ffs2_din = pool_cache_get(ffs_dinode2_cache,
1828 PR_WAITOK);
1829 ffs_load_inode(bp, ip, fs, ino);
1830 brelse(bp, 0);
1831
1832 /*
1833 * Initialize the vnode from the inode, check for aliases.
1834 * Note that the underlying vnode may have changed.
1835 */
1836
1837 ufs_vinit(mp, ffs_specop_p, ffs_fifoop_p, &vp);
1838
1839 #ifdef XIP
1840 if ((vp->v_mount->mnt_iflag & IMNT_XIP) != 0 &&
1841 vp->v_type == VREG) {
1842 vp->v_vflag |= VV_XIP;
1843 }
1844 #endif
1845
1846 /*
1847 * Finish inode initialization now that aliasing has been resolved.
1848 */
1849
1850 ip->i_devvp = ump->um_devvp;
1851 vref(ip->i_devvp);
1852
1853 /*
1854 * Ensure that uid and gid are correct. This is a temporary
1855 * fix until fsck has been changed to do the update.
1856 */
1857
1858 if (fs->fs_old_inodefmt < FS_44INODEFMT) { /* XXX */
1859 ip->i_uid = ip->i_ffs1_ouid; /* XXX */
1860 ip->i_gid = ip->i_ffs1_ogid; /* XXX */
1861 } /* XXX */
1862 uvm_vnp_setsize(vp, ip->i_size);
1863 *vpp = vp;
1864 return (0);
1865 }
1866
1867 /*
1868 * File handle to vnode
1869 *
1870 * Have to be really careful about stale file handles:
1871 * - check that the inode number is valid
1872 * - call ffs_vget() to get the locked inode
1873 * - check for an unallocated inode (i_mode == 0)
1874 * - check that the given client host has export rights and return
1875 * those rights via. exflagsp and credanonp
1876 */
1877 int
1878 ffs_fhtovp(struct mount *mp, struct fid *fhp, struct vnode **vpp)
1879 {
1880 struct ufid ufh;
1881 struct fs *fs;
1882
1883 if (fhp->fid_len != sizeof(struct ufid))
1884 return EINVAL;
1885
1886 memcpy(&ufh, fhp, sizeof(ufh));
1887 fs = VFSTOUFS(mp)->um_fs;
1888 if (ufh.ufid_ino < ROOTINO ||
1889 ufh.ufid_ino >= fs->fs_ncg * fs->fs_ipg)
1890 return (ESTALE);
1891 return (ufs_fhtovp(mp, &ufh, vpp));
1892 }
1893
1894 /*
1895 * Vnode pointer to File handle
1896 */
1897 /* ARGSUSED */
1898 int
1899 ffs_vptofh(struct vnode *vp, struct fid *fhp, size_t *fh_size)
1900 {
1901 struct inode *ip;
1902 struct ufid ufh;
1903
1904 if (*fh_size < sizeof(struct ufid)) {
1905 *fh_size = sizeof(struct ufid);
1906 return E2BIG;
1907 }
1908 ip = VTOI(vp);
1909 *fh_size = sizeof(struct ufid);
1910 memset(&ufh, 0, sizeof(ufh));
1911 ufh.ufid_len = sizeof(struct ufid);
1912 ufh.ufid_ino = ip->i_number;
1913 ufh.ufid_gen = ip->i_gen;
1914 memcpy(fhp, &ufh, sizeof(ufh));
1915 return (0);
1916 }
1917
1918 void
1919 ffs_init(void)
1920 {
1921 if (ffs_initcount++ > 0)
1922 return;
1923
1924 ffs_inode_cache = pool_cache_init(sizeof(struct inode), 0, 0, 0,
1925 "ffsino", NULL, IPL_NONE, NULL, NULL, NULL);
1926 ffs_dinode1_cache = pool_cache_init(sizeof(struct ufs1_dinode), 0, 0, 0,
1927 "ffsdino1", NULL, IPL_NONE, NULL, NULL, NULL);
1928 ffs_dinode2_cache = pool_cache_init(sizeof(struct ufs2_dinode), 0, 0, 0,
1929 "ffsdino2", NULL, IPL_NONE, NULL, NULL, NULL);
1930 ufs_init();
1931 }
1932
1933 void
1934 ffs_reinit(void)
1935 {
1936
1937 ufs_reinit();
1938 }
1939
1940 void
1941 ffs_done(void)
1942 {
1943 if (--ffs_initcount > 0)
1944 return;
1945
1946 ufs_done();
1947 pool_cache_destroy(ffs_dinode2_cache);
1948 pool_cache_destroy(ffs_dinode1_cache);
1949 pool_cache_destroy(ffs_inode_cache);
1950 }
1951
1952 /*
1953 * Write a superblock and associated information back to disk.
1954 */
1955 int
1956 ffs_sbupdate(struct ufsmount *mp, int waitfor)
1957 {
1958 struct fs *fs = mp->um_fs;
1959 struct buf *bp;
1960 int error = 0;
1961 u_int32_t saveflag;
1962
1963 error = ffs_getblk(mp->um_devvp,
1964 fs->fs_sblockloc / DEV_BSIZE, FFS_NOBLK,
1965 fs->fs_sbsize, false, &bp);
1966 if (error)
1967 return error;
1968 saveflag = fs->fs_flags & FS_INTERNAL;
1969 fs->fs_flags &= ~FS_INTERNAL;
1970
1971 memcpy(bp->b_data, fs, fs->fs_sbsize);
1972
1973 ffs_oldfscompat_write((struct fs *)bp->b_data, mp);
1974 #ifdef FFS_EI
1975 if (mp->um_flags & UFS_NEEDSWAP)
1976 ffs_sb_swap((struct fs *)bp->b_data, (struct fs *)bp->b_data);
1977 #endif
1978 fs->fs_flags |= saveflag;
1979
1980 if (waitfor == MNT_WAIT)
1981 error = bwrite(bp);
1982 else
1983 bawrite(bp);
1984 return (error);
1985 }
1986
1987 int
1988 ffs_cgupdate(struct ufsmount *mp, int waitfor)
1989 {
1990 struct fs *fs = mp->um_fs;
1991 struct buf *bp;
1992 int blks;
1993 void *space;
1994 int i, size, error = 0, allerror = 0;
1995
1996 allerror = ffs_sbupdate(mp, waitfor);
1997 blks = howmany(fs->fs_cssize, fs->fs_fsize);
1998 space = fs->fs_csp;
1999 for (i = 0; i < blks; i += fs->fs_frag) {
2000 size = fs->fs_bsize;
2001 if (i + fs->fs_frag > blks)
2002 size = (blks - i) * fs->fs_fsize;
2003 error = ffs_getblk(mp->um_devvp, fsbtodb(fs, fs->fs_csaddr + i),
2004 FFS_NOBLK, size, false, &bp);
2005 if (error)
2006 break;
2007 #ifdef FFS_EI
2008 if (mp->um_flags & UFS_NEEDSWAP)
2009 ffs_csum_swap((struct csum*)space,
2010 (struct csum*)bp->b_data, size);
2011 else
2012 #endif
2013 memcpy(bp->b_data, space, (u_int)size);
2014 space = (char *)space + size;
2015 if (waitfor == MNT_WAIT)
2016 error = bwrite(bp);
2017 else
2018 bawrite(bp);
2019 }
2020 if (!allerror && error)
2021 allerror = error;
2022 return (allerror);
2023 }
2024
2025 int
2026 ffs_extattrctl(struct mount *mp, int cmd, struct vnode *vp,
2027 int attrnamespace, const char *attrname)
2028 {
2029 #ifdef UFS_EXTATTR
2030 /*
2031 * File-backed extended attributes are only supported on UFS1.
2032 * UFS2 has native extended attributes.
2033 */
2034 if (VFSTOUFS(mp)->um_fstype == UFS1)
2035 return (ufs_extattrctl(mp, cmd, vp, attrnamespace, attrname));
2036 #endif
2037 return (vfs_stdextattrctl(mp, cmd, vp, attrnamespace, attrname));
2038 }
2039
2040 int
2041 ffs_suspendctl(struct mount *mp, int cmd)
2042 {
2043 int error;
2044 struct lwp *l = curlwp;
2045
2046 switch (cmd) {
2047 case SUSPEND_SUSPEND:
2048 if ((error = fstrans_setstate(mp, FSTRANS_SUSPENDING)) != 0)
2049 return error;
2050 error = ffs_sync(mp, MNT_WAIT, l->l_proc->p_cred);
2051 if (error == 0)
2052 error = fstrans_setstate(mp, FSTRANS_SUSPENDED);
2053 #ifdef WAPBL
2054 if (error == 0 && mp->mnt_wapbl)
2055 error = wapbl_flush(mp->mnt_wapbl, 1);
2056 #endif
2057 if (error != 0) {
2058 (void) fstrans_setstate(mp, FSTRANS_NORMAL);
2059 return error;
2060 }
2061 return 0;
2062
2063 case SUSPEND_RESUME:
2064 return fstrans_setstate(mp, FSTRANS_NORMAL);
2065
2066 default:
2067 return EINVAL;
2068 }
2069 }
2070
2071 /*
2072 * Synch vnode for a mounted file system. This is called for foreign
2073 * vnodes, i.e. non-ffs.
2074 */
2075 static int
2076 ffs_vfs_fsync(vnode_t *vp, int flags)
2077 {
2078 int error, passes, skipmeta, i, pflags;
2079 buf_t *bp, *nbp;
2080 #ifdef WAPBL
2081 struct mount *mp;
2082 #endif
2083
2084 KASSERT(vp->v_type == VBLK);
2085 KASSERT(vp->v_specmountpoint != NULL);
2086
2087 /*
2088 * Flush all dirty data associated with the vnode.
2089 */
2090 pflags = PGO_ALLPAGES | PGO_CLEANIT;
2091 if ((flags & FSYNC_WAIT) != 0)
2092 pflags |= PGO_SYNCIO;
2093 mutex_enter(&vp->v_interlock);
2094 error = VOP_PUTPAGES(vp, 0, 0, pflags);
2095 if (error)
2096 return error;
2097
2098 #ifdef WAPBL
2099 mp = vp->v_specmountpoint;
2100 if (mp && mp->mnt_wapbl) {
2101 /*
2102 * Don't bother writing out metadata if the syncer is
2103 * making the request. We will let the sync vnode
2104 * write it out in a single burst through a call to
2105 * VFS_SYNC().
2106 */
2107 if ((flags & (FSYNC_DATAONLY | FSYNC_LAZY | FSYNC_NOLOG)) != 0)
2108 return 0;
2109
2110 /*
2111 * Don't flush the log if the vnode being flushed
2112 * contains no dirty buffers that could be in the log.
2113 */
2114 if (!LIST_EMPTY(&vp->v_dirtyblkhd)) {
2115 error = wapbl_flush(mp->mnt_wapbl, 0);
2116 if (error)
2117 return error;
2118 }
2119
2120 if ((flags & FSYNC_WAIT) != 0) {
2121 mutex_enter(&vp->v_interlock);
2122 while (vp->v_numoutput)
2123 cv_wait(&vp->v_cv, &vp->v_interlock);
2124 mutex_exit(&vp->v_interlock);
2125 }
2126
2127 return 0;
2128 }
2129 #endif /* WAPBL */
2130
2131 /*
2132 * Write out metadata for non-logging file systems. XXX This block
2133 * should be simplified now that softdep is gone.
2134 */
2135 passes = NIADDR + 1;
2136 skipmeta = 0;
2137 if (flags & FSYNC_WAIT)
2138 skipmeta = 1;
2139
2140 loop:
2141 mutex_enter(&bufcache_lock);
2142 LIST_FOREACH(bp, &vp->v_dirtyblkhd, b_vnbufs) {
2143 bp->b_cflags &= ~BC_SCANNED;
2144 }
2145 for (bp = LIST_FIRST(&vp->v_dirtyblkhd); bp; bp = nbp) {
2146 nbp = LIST_NEXT(bp, b_vnbufs);
2147 if (bp->b_cflags & (BC_BUSY | BC_SCANNED))
2148 continue;
2149 if ((bp->b_oflags & BO_DELWRI) == 0)
2150 panic("ffs_fsync: not dirty");
2151 if (skipmeta && bp->b_lblkno < 0)
2152 continue;
2153 bp->b_cflags |= BC_BUSY | BC_VFLUSH | BC_SCANNED;
2154 mutex_exit(&bufcache_lock);
2155 /*
2156 * On our final pass through, do all I/O synchronously
2157 * so that we can find out if our flush is failing
2158 * because of write errors.
2159 */
2160 if (passes > 0 || !(flags & FSYNC_WAIT))
2161 (void) bawrite(bp);
2162 else if ((error = bwrite(bp)) != 0)
2163 return (error);
2164 /*
2165 * Since we unlocked during the I/O, we need
2166 * to start from a known point.
2167 */
2168 mutex_enter(&bufcache_lock);
2169 nbp = LIST_FIRST(&vp->v_dirtyblkhd);
2170 }
2171 mutex_exit(&bufcache_lock);
2172 if (skipmeta) {
2173 skipmeta = 0;
2174 goto loop;
2175 }
2176
2177 if ((flags & FSYNC_WAIT) != 0) {
2178 mutex_enter(&vp->v_interlock);
2179 while (vp->v_numoutput) {
2180 cv_wait(&vp->v_cv, &vp->v_interlock);
2181 }
2182 mutex_exit(&vp->v_interlock);
2183
2184 if (!LIST_EMPTY(&vp->v_dirtyblkhd)) {
2185 /*
2186 * Block devices associated with filesystems may
2187 * have new I/O requests posted for them even if
2188 * the vnode is locked, so no amount of trying will
2189 * get them clean. Thus we give block devices a
2190 * good effort, then just give up. For all other file
2191 * types, go around and try again until it is clean.
2192 */
2193 if (passes > 0) {
2194 passes--;
2195 goto loop;
2196 }
2197 #ifdef DIAGNOSTIC
2198 if (vp->v_type != VBLK)
2199 vprint("ffs_fsync: dirty", vp);
2200 #endif
2201 }
2202 }
2203
2204 if (error == 0 && (flags & FSYNC_CACHE) != 0) {
2205 (void)VOP_IOCTL(vp, DIOCCACHESYNC, &i, FWRITE,
2206 kauth_cred_get());
2207 }
2208
2209 return error;
2210 }
2211