ffs_vfsops.c revision 1.257.2.3 1 /* $NetBSD: ffs_vfsops.c,v 1.257.2.3 2010/03/23 01:58:14 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 1.257.2.3 2010/03/23 01:58:14 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, 0);
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, 0);
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, 0);
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 | LK_INTERLOCK)) {
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 uint64_t numsecs;
827 unsigned secsize;
828 #ifdef FFS_EI
829 int needswap = 0; /* keep gcc happy */
830 #endif
831 int32_t *lp;
832 kauth_cred_t cred;
833 u_int32_t sbsize = 8192; /* keep gcc happy*/
834
835 dev = devvp->v_rdev;
836 cred = l ? l->l_cred : NOCRED;
837
838 /* Flush out any old buffers remaining from a previous use. */
839 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
840 error = vinvalbuf(devvp, V_SAVE, cred, l, 0, 0);
841 VOP_UNLOCK(devvp, 0);
842 if (error)
843 return (error);
844
845 ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
846 error = getdisksize(devvp, &numsecs, &secsize);
847 if (error)
848 return (error);
849
850 bp = NULL;
851 ump = NULL;
852 fs = NULL;
853 sblockloc = 0;
854 fstype = 0;
855
856 error = fstrans_mount(mp);
857 if (error)
858 return error;
859
860 ump = malloc(sizeof *ump, M_UFSMNT, M_WAITOK);
861 memset(ump, 0, sizeof *ump);
862 mutex_init(&ump->um_lock, MUTEX_DEFAULT, IPL_NONE);
863 error = ffs_snapshot_init(ump);
864 if (error)
865 goto out;
866 ump->um_ops = &ffs_ufsops;
867
868 #ifdef WAPBL
869 sbagain:
870 #endif
871 /*
872 * Try reading the superblock in each of its possible locations.
873 */
874 for (i = 0; ; i++) {
875 if (bp != NULL) {
876 brelse(bp, BC_NOCACHE);
877 bp = NULL;
878 }
879 if (sblock_try[i] == -1) {
880 error = EINVAL;
881 fs = NULL;
882 goto out;
883 }
884 error = bread(devvp, sblock_try[i] / DEV_BSIZE, SBLOCKSIZE, cred,
885 0, &bp);
886 if (error) {
887 fs = NULL;
888 goto out;
889 }
890 fs = (struct fs*)bp->b_data;
891 fsblockloc = sblockloc = sblock_try[i];
892 if (fs->fs_magic == FS_UFS1_MAGIC) {
893 sbsize = fs->fs_sbsize;
894 fstype = UFS1;
895 #ifdef FFS_EI
896 needswap = 0;
897 } else if (fs->fs_magic == bswap32(FS_UFS1_MAGIC)) {
898 sbsize = bswap32(fs->fs_sbsize);
899 fstype = UFS1;
900 needswap = 1;
901 #endif
902 } else if (fs->fs_magic == FS_UFS2_MAGIC) {
903 sbsize = fs->fs_sbsize;
904 fstype = UFS2;
905 #ifdef FFS_EI
906 needswap = 0;
907 } else if (fs->fs_magic == bswap32(FS_UFS2_MAGIC)) {
908 sbsize = bswap32(fs->fs_sbsize);
909 fstype = UFS2;
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 /* Ok seems to be a good superblock */
943 break;
944 }
945
946 fs = malloc((u_long)sbsize, M_UFSMNT, M_WAITOK);
947 memcpy(fs, bp->b_data, sbsize);
948 ump->um_fs = fs;
949
950 #ifdef FFS_EI
951 if (needswap) {
952 ffs_sb_swap((struct fs*)bp->b_data, fs);
953 fs->fs_flags |= FS_SWAPPED;
954 } else
955 #endif
956 fs->fs_flags &= ~FS_SWAPPED;
957
958 #ifdef WAPBL
959 if ((mp->mnt_wapbl_replay == 0) && (fs->fs_flags & FS_DOWAPBL)) {
960 error = ffs_wapbl_replay_start(mp, fs, devvp);
961 if (error && (mp->mnt_flag & MNT_FORCE) == 0)
962 goto out;
963 if (!error) {
964 if (!ronly) {
965 /* XXX fsmnt may be stale. */
966 printf("%s: replaying log to disk\n",
967 fs->fs_fsmnt);
968 error = wapbl_replay_write(mp->mnt_wapbl_replay,
969 devvp);
970 if (error)
971 goto out;
972 wapbl_replay_stop(mp->mnt_wapbl_replay);
973 fs->fs_clean = FS_WASCLEAN;
974 } else {
975 /* XXX fsmnt may be stale */
976 printf("%s: replaying log to memory\n",
977 fs->fs_fsmnt);
978 }
979
980 /* Force a re-read of the superblock */
981 brelse(bp, BC_INVAL);
982 bp = NULL;
983 free(fs, M_UFSMNT);
984 fs = NULL;
985 goto sbagain;
986 }
987 }
988 #else /* !WAPBL */
989 if ((fs->fs_flags & FS_DOWAPBL) && (mp->mnt_flag & MNT_FORCE) == 0) {
990 error = EPERM;
991 goto out;
992 }
993 #endif /* !WAPBL */
994
995 ffs_oldfscompat_read(fs, ump, sblockloc);
996 ump->um_maxfilesize = fs->fs_maxfilesize;
997
998 if (fs->fs_flags & ~(FS_KNOWN_FLAGS | FS_INTERNAL)) {
999 uprintf("%s: unknown ufs flags: 0x%08"PRIx32"%s\n",
1000 mp->mnt_stat.f_mntonname, fs->fs_flags,
1001 (mp->mnt_flag & MNT_FORCE) ? "" : ", not mounting");
1002 if ((mp->mnt_flag & MNT_FORCE) == 0) {
1003 error = EINVAL;
1004 goto out;
1005 }
1006 }
1007
1008 if (fs->fs_pendingblocks != 0 || fs->fs_pendinginodes != 0) {
1009 fs->fs_pendingblocks = 0;
1010 fs->fs_pendinginodes = 0;
1011 }
1012
1013 ump->um_fstype = fstype;
1014 if (fs->fs_sbsize < SBLOCKSIZE)
1015 brelse(bp, BC_INVAL);
1016 else
1017 brelse(bp, 0);
1018 bp = NULL;
1019
1020 /* First check to see if this is tagged as an Apple UFS filesystem
1021 * in the disklabel
1022 */
1023 if ((VOP_IOCTL(devvp, DIOCGPART, &dpart, FREAD, cred) == 0) &&
1024 (dpart.part->p_fstype == FS_APPLEUFS)) {
1025 ump->um_flags |= UFS_ISAPPLEUFS;
1026 }
1027 #ifdef APPLE_UFS
1028 else {
1029 /* Manually look for an apple ufs label, and if a valid one
1030 * is found, then treat it like an Apple UFS filesystem anyway
1031 */
1032 error = bread(devvp, (daddr_t)(APPLEUFS_LABEL_OFFSET / DEV_BSIZE),
1033 APPLEUFS_LABEL_SIZE, cred, 0, &bp);
1034 if (error)
1035 goto out;
1036 error = ffs_appleufs_validate(fs->fs_fsmnt,
1037 (struct appleufslabel *)bp->b_data, NULL);
1038 if (error == 0) {
1039 ump->um_flags |= UFS_ISAPPLEUFS;
1040 }
1041 brelse(bp, 0);
1042 bp = NULL;
1043 }
1044 #else
1045 if (ump->um_flags & UFS_ISAPPLEUFS) {
1046 error = EINVAL;
1047 goto out;
1048 }
1049 #endif
1050
1051 #if 0
1052 /*
1053 * XXX This code changes the behaviour of mounting dirty filesystems, to
1054 * XXX require "mount -f ..." to mount them. This doesn't match what
1055 * XXX mount(8) describes and is disabled for now.
1056 */
1057 /*
1058 * If the file system is not clean, don't allow it to be mounted
1059 * unless MNT_FORCE is specified. (Note: MNT_FORCE is always set
1060 * for the root file system.)
1061 */
1062 if (fs->fs_flags & FS_DOWAPBL) {
1063 /*
1064 * wapbl normally expects to be FS_WASCLEAN when the FS_DOWAPBL
1065 * bit is set, although there's a window in unmount where it
1066 * could be FS_ISCLEAN
1067 */
1068 if ((mp->mnt_flag & MNT_FORCE) == 0 &&
1069 (fs->fs_clean & (FS_WASCLEAN | FS_ISCLEAN)) == 0) {
1070 error = EPERM;
1071 goto out;
1072 }
1073 } else
1074 if ((fs->fs_clean & FS_ISCLEAN) == 0 &&
1075 (mp->mnt_flag & MNT_FORCE) == 0) {
1076 error = EPERM;
1077 goto out;
1078 }
1079 #endif
1080
1081 /*
1082 * verify that we can access the last block in the fs
1083 * if we're mounting read/write.
1084 */
1085
1086 if (!ronly) {
1087 error = bread(devvp, fsbtodb(fs, fs->fs_size - 1), fs->fs_fsize,
1088 cred, 0, &bp);
1089 if (bp->b_bcount != fs->fs_fsize)
1090 error = EINVAL;
1091 if (error) {
1092 bset = BC_INVAL;
1093 goto out;
1094 }
1095 brelse(bp, BC_INVAL);
1096 bp = NULL;
1097 }
1098
1099 fs->fs_ronly = ronly;
1100 /* Don't bump fs_clean if we're replaying journal */
1101 if (!((fs->fs_flags & FS_DOWAPBL) && (fs->fs_clean & FS_WASCLEAN)))
1102 if (ronly == 0) {
1103 fs->fs_clean <<= 1;
1104 fs->fs_fmod = 1;
1105 }
1106 bsize = fs->fs_cssize;
1107 blks = howmany(bsize, fs->fs_fsize);
1108 if (fs->fs_contigsumsize > 0)
1109 bsize += fs->fs_ncg * sizeof(int32_t);
1110 bsize += fs->fs_ncg * sizeof(*fs->fs_contigdirs);
1111 space = malloc((u_long)bsize, M_UFSMNT, M_WAITOK);
1112 fs->fs_csp = space;
1113 for (i = 0; i < blks; i += fs->fs_frag) {
1114 bsize = fs->fs_bsize;
1115 if (i + fs->fs_frag > blks)
1116 bsize = (blks - i) * fs->fs_fsize;
1117 error = bread(devvp, fsbtodb(fs, fs->fs_csaddr + i), bsize,
1118 cred, 0, &bp);
1119 if (error) {
1120 free(fs->fs_csp, M_UFSMNT);
1121 goto out;
1122 }
1123 #ifdef FFS_EI
1124 if (needswap)
1125 ffs_csum_swap((struct csum *)bp->b_data,
1126 (struct csum *)space, bsize);
1127 else
1128 #endif
1129 memcpy(space, bp->b_data, (u_int)bsize);
1130
1131 space = (char *)space + bsize;
1132 brelse(bp, 0);
1133 bp = NULL;
1134 }
1135 if (fs->fs_contigsumsize > 0) {
1136 fs->fs_maxcluster = lp = space;
1137 for (i = 0; i < fs->fs_ncg; i++)
1138 *lp++ = fs->fs_contigsumsize;
1139 space = lp;
1140 }
1141 bsize = fs->fs_ncg * sizeof(*fs->fs_contigdirs);
1142 fs->fs_contigdirs = space;
1143 space = (char *)space + bsize;
1144 memset(fs->fs_contigdirs, 0, bsize);
1145 /* Compatibility for old filesystems - XXX */
1146 if (fs->fs_avgfilesize <= 0)
1147 fs->fs_avgfilesize = AVFILESIZ;
1148 if (fs->fs_avgfpdir <= 0)
1149 fs->fs_avgfpdir = AFPDIR;
1150 fs->fs_active = NULL;
1151 mp->mnt_data = ump;
1152 mp->mnt_stat.f_fsidx.__fsid_val[0] = (long)dev;
1153 mp->mnt_stat.f_fsidx.__fsid_val[1] = makefstype(MOUNT_FFS);
1154 mp->mnt_stat.f_fsid = mp->mnt_stat.f_fsidx.__fsid_val[0];
1155 mp->mnt_stat.f_namemax = FFS_MAXNAMLEN;
1156 if (UFS_MPISAPPLEUFS(ump)) {
1157 /* NeXT used to keep short symlinks in the inode even
1158 * when using FS_42INODEFMT. In that case fs->fs_maxsymlinklen
1159 * is probably -1, but we still need to be able to identify
1160 * short symlinks.
1161 */
1162 ump->um_maxsymlinklen = APPLEUFS_MAXSYMLINKLEN;
1163 ump->um_dirblksiz = APPLEUFS_DIRBLKSIZ;
1164 mp->mnt_iflag |= IMNT_DTYPE;
1165 } else {
1166 ump->um_maxsymlinklen = fs->fs_maxsymlinklen;
1167 ump->um_dirblksiz = DIRBLKSIZ;
1168 if (ump->um_maxsymlinklen > 0)
1169 mp->mnt_iflag |= IMNT_DTYPE;
1170 else
1171 mp->mnt_iflag &= ~IMNT_DTYPE;
1172 }
1173 mp->mnt_fs_bshift = fs->fs_bshift;
1174 mp->mnt_dev_bshift = DEV_BSHIFT; /* XXX */
1175 mp->mnt_flag |= MNT_LOCAL;
1176 mp->mnt_iflag |= IMNT_MPSAFE;
1177 #ifdef XIP
1178 paddr_t phys_addr = 0;
1179 if (mp->mnt_flag & MNT_XIP) {
1180 if ((mp->mnt_flag & MNT_RDONLY) == 0)
1181 printf("XIP needs read-only mount\n");
1182 else {
1183 if ((VOP_IOCTL(devvp, DIOCGPHYSADDR, &phys_addr, FREAD,
1184 cred) == 0) &&
1185 phys_addr != 0) {
1186 mp->mnt_iflag |= IMNT_XIP;
1187 devvp->v_phys_addr = phys_addr; /* XXX */
1188 } else
1189 printf("device doesn't support DIOCGPHYSADDR\n");
1190 }
1191 }
1192 #endif
1193 #ifdef FFS_EI
1194 if (needswap)
1195 ump->um_flags |= UFS_NEEDSWAP;
1196 #endif
1197 ump->um_mountp = mp;
1198 ump->um_dev = dev;
1199 ump->um_devvp = devvp;
1200 ump->um_nindir = fs->fs_nindir;
1201 ump->um_lognindir = ffs(fs->fs_nindir) - 1;
1202 ump->um_bptrtodb = fs->fs_fshift - DEV_BSHIFT;
1203 ump->um_seqinc = fs->fs_frag;
1204 for (i = 0; i < MAXQUOTAS; i++)
1205 ump->um_quotas[i] = NULLVP;
1206 devvp->v_specmountpoint = mp;
1207 if (ronly == 0 && fs->fs_snapinum[0] != 0)
1208 ffs_snapshot_mount(mp);
1209
1210 #ifdef WAPBL
1211 if (!ronly) {
1212 KDASSERT(fs->fs_ronly == 0);
1213 /*
1214 * ffs_wapbl_start() needs mp->mnt_stat initialised if it
1215 * needs to create a new log file in-filesystem.
1216 */
1217 ffs_statvfs(mp, &mp->mnt_stat);
1218
1219 error = ffs_wapbl_start(mp);
1220 if (error) {
1221 free(fs->fs_csp, M_UFSMNT);
1222 goto out;
1223 }
1224 }
1225 #endif /* WAPBL */
1226 #ifdef UFS_EXTATTR
1227 /*
1228 * Initialize file-backed extended attributes on UFS1 file
1229 * systems.
1230 */
1231 if (ump->um_fstype == UFS1) {
1232 ufs_extattr_uepm_init(&ump->um_extattr);
1233 #ifdef UFS_EXTATTR_AUTOSTART
1234 /*
1235 * XXX Just ignore errors. Not clear that we should
1236 * XXX fail the mount in this case.
1237 */
1238 (void) ufs_extattr_autostart(mp, l);
1239 #endif
1240 }
1241 #endif /* UFS_EXTATTR */
1242 return (0);
1243 out:
1244 #ifdef WAPBL
1245 if (mp->mnt_wapbl_replay) {
1246 wapbl_replay_stop(mp->mnt_wapbl_replay);
1247 wapbl_replay_free(mp->mnt_wapbl_replay);
1248 mp->mnt_wapbl_replay = 0;
1249 }
1250 #endif
1251
1252 fstrans_unmount(mp);
1253 if (fs)
1254 free(fs, M_UFSMNT);
1255 devvp->v_specmountpoint = NULL;
1256 if (bp)
1257 brelse(bp, bset);
1258 if (ump) {
1259 if (ump->um_oldfscompat)
1260 free(ump->um_oldfscompat, M_UFSMNT);
1261 mutex_destroy(&ump->um_lock);
1262 free(ump, M_UFSMNT);
1263 mp->mnt_data = NULL;
1264 }
1265 return (error);
1266 }
1267
1268 /*
1269 * Sanity checks for loading old filesystem superblocks.
1270 * See ffs_oldfscompat_write below for unwound actions.
1271 *
1272 * XXX - Parts get retired eventually.
1273 * Unfortunately new bits get added.
1274 */
1275 static void
1276 ffs_oldfscompat_read(struct fs *fs, struct ufsmount *ump, daddr_t sblockloc)
1277 {
1278 off_t maxfilesize;
1279 int32_t *extrasave;
1280
1281 if ((fs->fs_magic != FS_UFS1_MAGIC) ||
1282 (fs->fs_old_flags & FS_FLAGS_UPDATED))
1283 return;
1284
1285 if (!ump->um_oldfscompat)
1286 ump->um_oldfscompat = malloc(512 + 3*sizeof(int32_t),
1287 M_UFSMNT, M_WAITOK);
1288
1289 memcpy(ump->um_oldfscompat, &fs->fs_old_postbl_start, 512);
1290 extrasave = ump->um_oldfscompat;
1291 extrasave += 512/sizeof(int32_t);
1292 extrasave[0] = fs->fs_old_npsect;
1293 extrasave[1] = fs->fs_old_interleave;
1294 extrasave[2] = fs->fs_old_trackskew;
1295
1296 /* These fields will be overwritten by their
1297 * original values in fs_oldfscompat_write, so it is harmless
1298 * to modify them here.
1299 */
1300 fs->fs_cstotal.cs_ndir = fs->fs_old_cstotal.cs_ndir;
1301 fs->fs_cstotal.cs_nbfree = fs->fs_old_cstotal.cs_nbfree;
1302 fs->fs_cstotal.cs_nifree = fs->fs_old_cstotal.cs_nifree;
1303 fs->fs_cstotal.cs_nffree = fs->fs_old_cstotal.cs_nffree;
1304
1305 fs->fs_maxbsize = fs->fs_bsize;
1306 fs->fs_time = fs->fs_old_time;
1307 fs->fs_size = fs->fs_old_size;
1308 fs->fs_dsize = fs->fs_old_dsize;
1309 fs->fs_csaddr = fs->fs_old_csaddr;
1310 fs->fs_sblockloc = sblockloc;
1311
1312 fs->fs_flags = fs->fs_old_flags | (fs->fs_flags & FS_INTERNAL);
1313
1314 if (fs->fs_old_postblformat == FS_42POSTBLFMT) {
1315 fs->fs_old_nrpos = 8;
1316 fs->fs_old_npsect = fs->fs_old_nsect;
1317 fs->fs_old_interleave = 1;
1318 fs->fs_old_trackskew = 0;
1319 }
1320
1321 if (fs->fs_old_inodefmt < FS_44INODEFMT) {
1322 fs->fs_maxfilesize = (u_quad_t) 1LL << 39;
1323 fs->fs_qbmask = ~fs->fs_bmask;
1324 fs->fs_qfmask = ~fs->fs_fmask;
1325 }
1326
1327 maxfilesize = (u_int64_t)0x80000000 * fs->fs_bsize - 1;
1328 if (fs->fs_maxfilesize > maxfilesize)
1329 fs->fs_maxfilesize = maxfilesize;
1330
1331 /* Compatibility for old filesystems */
1332 if (fs->fs_avgfilesize <= 0)
1333 fs->fs_avgfilesize = AVFILESIZ;
1334 if (fs->fs_avgfpdir <= 0)
1335 fs->fs_avgfpdir = AFPDIR;
1336
1337 #if 0
1338 if (bigcgs) {
1339 fs->fs_save_cgsize = fs->fs_cgsize;
1340 fs->fs_cgsize = fs->fs_bsize;
1341 }
1342 #endif
1343 }
1344
1345 /*
1346 * Unwinding superblock updates for old filesystems.
1347 * See ffs_oldfscompat_read above for details.
1348 *
1349 * XXX - Parts get retired eventually.
1350 * Unfortunately new bits get added.
1351 */
1352 static void
1353 ffs_oldfscompat_write(struct fs *fs, struct ufsmount *ump)
1354 {
1355 int32_t *extrasave;
1356
1357 if ((fs->fs_magic != FS_UFS1_MAGIC) ||
1358 (fs->fs_old_flags & FS_FLAGS_UPDATED))
1359 return;
1360
1361 fs->fs_old_time = fs->fs_time;
1362 fs->fs_old_cstotal.cs_ndir = fs->fs_cstotal.cs_ndir;
1363 fs->fs_old_cstotal.cs_nbfree = fs->fs_cstotal.cs_nbfree;
1364 fs->fs_old_cstotal.cs_nifree = fs->fs_cstotal.cs_nifree;
1365 fs->fs_old_cstotal.cs_nffree = fs->fs_cstotal.cs_nffree;
1366 fs->fs_old_flags = fs->fs_flags;
1367
1368 #if 0
1369 if (bigcgs) {
1370 fs->fs_cgsize = fs->fs_save_cgsize;
1371 }
1372 #endif
1373
1374 memcpy(&fs->fs_old_postbl_start, ump->um_oldfscompat, 512);
1375 extrasave = ump->um_oldfscompat;
1376 extrasave += 512/sizeof(int32_t);
1377 fs->fs_old_npsect = extrasave[0];
1378 fs->fs_old_interleave = extrasave[1];
1379 fs->fs_old_trackskew = extrasave[2];
1380
1381 }
1382
1383 /*
1384 * unmount vfs operation
1385 */
1386 int
1387 ffs_unmount(struct mount *mp, int mntflags)
1388 {
1389 struct lwp *l = curlwp;
1390 struct ufsmount *ump = VFSTOUFS(mp);
1391 struct fs *fs = ump->um_fs;
1392 int error, flags;
1393 #ifdef WAPBL
1394 extern int doforce;
1395 #endif
1396
1397 flags = 0;
1398 if (mntflags & MNT_FORCE)
1399 flags |= FORCECLOSE;
1400 if ((error = ffs_flushfiles(mp, flags, l)) != 0)
1401 return (error);
1402 error = UFS_WAPBL_BEGIN(mp);
1403 if (error == 0)
1404 if (fs->fs_ronly == 0 &&
1405 ffs_cgupdate(ump, MNT_WAIT) == 0 &&
1406 fs->fs_clean & FS_WASCLEAN) {
1407 fs->fs_clean = FS_ISCLEAN;
1408 fs->fs_fmod = 0;
1409 (void) ffs_sbupdate(ump, MNT_WAIT);
1410 }
1411 if (error == 0)
1412 UFS_WAPBL_END(mp);
1413 #ifdef WAPBL
1414 KASSERT(!(mp->mnt_wapbl_replay && mp->mnt_wapbl));
1415 if (mp->mnt_wapbl_replay) {
1416 KDASSERT(fs->fs_ronly);
1417 wapbl_replay_stop(mp->mnt_wapbl_replay);
1418 wapbl_replay_free(mp->mnt_wapbl_replay);
1419 mp->mnt_wapbl_replay = 0;
1420 }
1421 error = ffs_wapbl_stop(mp, doforce && (mntflags & MNT_FORCE));
1422 if (error) {
1423 return error;
1424 }
1425 #endif /* WAPBL */
1426 #ifdef UFS_EXTATTR
1427 if (ump->um_fstype == UFS1) {
1428 ufs_extattr_stop(mp, l);
1429 ufs_extattr_uepm_destroy(&ump->um_extattr);
1430 }
1431 #endif /* UFS_EXTATTR */
1432
1433 if (ump->um_devvp->v_type != VBAD)
1434 ump->um_devvp->v_specmountpoint = NULL;
1435 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
1436 (void)VOP_CLOSE(ump->um_devvp, fs->fs_ronly ? FREAD : FREAD | FWRITE,
1437 NOCRED);
1438 vput(ump->um_devvp);
1439 free(fs->fs_csp, M_UFSMNT);
1440 free(fs, M_UFSMNT);
1441 if (ump->um_oldfscompat != NULL)
1442 free(ump->um_oldfscompat, M_UFSMNT);
1443 mutex_destroy(&ump->um_lock);
1444 ffs_snapshot_fini(ump);
1445 free(ump, M_UFSMNT);
1446 mp->mnt_data = NULL;
1447 mp->mnt_flag &= ~MNT_LOCAL;
1448 fstrans_unmount(mp);
1449 return (0);
1450 }
1451
1452 /*
1453 * Flush out all the files in a filesystem.
1454 */
1455 int
1456 ffs_flushfiles(struct mount *mp, int flags, struct lwp *l)
1457 {
1458 extern int doforce;
1459 struct ufsmount *ump;
1460 int error;
1461
1462 if (!doforce)
1463 flags &= ~FORCECLOSE;
1464 ump = VFSTOUFS(mp);
1465 #ifdef QUOTA
1466 if (mp->mnt_flag & MNT_QUOTA) {
1467 int i;
1468 if ((error = vflush(mp, NULLVP, SKIPSYSTEM | flags)) != 0)
1469 return (error);
1470 for (i = 0; i < MAXQUOTAS; i++) {
1471 if (ump->um_quotas[i] == NULLVP)
1472 continue;
1473 quotaoff(l, mp, i);
1474 }
1475 /*
1476 * Here we fall through to vflush again to ensure
1477 * that we have gotten rid of all the system vnodes.
1478 */
1479 }
1480 #endif
1481 if ((error = vflush(mp, 0, SKIPSYSTEM | flags)) != 0)
1482 return (error);
1483 ffs_snapshot_unmount(mp);
1484 /*
1485 * Flush all the files.
1486 */
1487 error = vflush(mp, NULLVP, flags);
1488 if (error)
1489 return (error);
1490 /*
1491 * Flush filesystem metadata.
1492 */
1493 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
1494 error = VOP_FSYNC(ump->um_devvp, l->l_cred, FSYNC_WAIT, 0, 0);
1495 VOP_UNLOCK(ump->um_devvp, 0);
1496 if (flags & FORCECLOSE) /* XXXDBJ */
1497 error = 0;
1498
1499 #ifdef WAPBL
1500 if (error)
1501 return error;
1502 if (mp->mnt_wapbl) {
1503 error = wapbl_flush(mp->mnt_wapbl, 1);
1504 if (flags & FORCECLOSE)
1505 error = 0;
1506 }
1507 #endif
1508
1509 return (error);
1510 }
1511
1512 /*
1513 * Get file system statistics.
1514 */
1515 int
1516 ffs_statvfs(struct mount *mp, struct statvfs *sbp)
1517 {
1518 struct ufsmount *ump;
1519 struct fs *fs;
1520
1521 ump = VFSTOUFS(mp);
1522 fs = ump->um_fs;
1523 mutex_enter(&ump->um_lock);
1524 sbp->f_bsize = fs->fs_bsize;
1525 sbp->f_frsize = fs->fs_fsize;
1526 sbp->f_iosize = fs->fs_bsize;
1527 sbp->f_blocks = fs->fs_dsize;
1528 sbp->f_bfree = blkstofrags(fs, fs->fs_cstotal.cs_nbfree) +
1529 fs->fs_cstotal.cs_nffree + dbtofsb(fs, fs->fs_pendingblocks);
1530 sbp->f_bresvd = ((u_int64_t) fs->fs_dsize * (u_int64_t)
1531 fs->fs_minfree) / (u_int64_t) 100;
1532 if (sbp->f_bfree > sbp->f_bresvd)
1533 sbp->f_bavail = sbp->f_bfree - sbp->f_bresvd;
1534 else
1535 sbp->f_bavail = 0;
1536 sbp->f_files = fs->fs_ncg * fs->fs_ipg - ROOTINO;
1537 sbp->f_ffree = fs->fs_cstotal.cs_nifree + fs->fs_pendinginodes;
1538 sbp->f_favail = sbp->f_ffree;
1539 sbp->f_fresvd = 0;
1540 mutex_exit(&ump->um_lock);
1541 copy_statvfs_info(sbp, mp);
1542
1543 return (0);
1544 }
1545
1546 /*
1547 * Go through the disk queues to initiate sandbagged IO;
1548 * go through the inodes to write those that have been modified;
1549 * initiate the writing of the super block if it has been modified.
1550 *
1551 * Note: we are always called with the filesystem marked `MPBUSY'.
1552 */
1553 int
1554 ffs_sync(struct mount *mp, int waitfor, kauth_cred_t cred)
1555 {
1556 struct vnode *vp, *mvp, *nvp;
1557 struct inode *ip;
1558 struct ufsmount *ump = VFSTOUFS(mp);
1559 struct fs *fs;
1560 int lk_flags, error, allerror = 0;
1561 bool is_suspending;
1562
1563 fs = ump->um_fs;
1564 if (fs->fs_fmod != 0 && fs->fs_ronly != 0) { /* XXX */
1565 printf("fs = %s\n", fs->fs_fsmnt);
1566 panic("update: rofs mod");
1567 }
1568
1569 /* Allocate a marker vnode. */
1570 if ((mvp = vnalloc(mp)) == NULL)
1571 return (ENOMEM);
1572
1573 fstrans_start(mp, FSTRANS_SHARED);
1574 is_suspending = (fstrans_getstate(mp) == FSTRANS_SUSPENDING);
1575 /*
1576 * We can't lock vnodes while the file system is suspending because
1577 * threads waiting on fstrans may have locked vnodes.
1578 */
1579 if (is_suspending)
1580 lk_flags = LK_INTERLOCK;
1581 else
1582 lk_flags = LK_EXCLUSIVE | LK_NOWAIT | LK_INTERLOCK;
1583 /*
1584 * Write back each (modified) inode.
1585 */
1586 mutex_enter(&mntvnode_lock);
1587 loop:
1588 /*
1589 * NOTE: not using the TAILQ_FOREACH here since in this loop vgone()
1590 * and vclean() can be called indirectly
1591 */
1592 for (vp = TAILQ_FIRST(&mp->mnt_vnodelist); vp; vp = nvp) {
1593 nvp = TAILQ_NEXT(vp, v_mntvnodes);
1594 /*
1595 * If the vnode that we are about to sync is no longer
1596 * associated with this mount point, start over.
1597 */
1598 if (vp->v_mount != mp)
1599 goto loop;
1600 /*
1601 * Don't interfere with concurrent scans of this FS.
1602 */
1603 if (vismarker(vp))
1604 continue;
1605 mutex_enter(&vp->v_interlock);
1606 ip = VTOI(vp);
1607
1608 /*
1609 * Skip the vnode/inode if inaccessible.
1610 */
1611 if (ip == NULL || (vp->v_iflag & (VI_XLOCK | VI_CLEAN)) != 0 ||
1612 vp->v_type == VNON) {
1613 mutex_exit(&vp->v_interlock);
1614 continue;
1615 }
1616
1617 /*
1618 * We deliberately update inode times here. This will
1619 * prevent a massive queue of updates accumulating, only
1620 * to be handled by a call to unmount.
1621 *
1622 * XXX It would be better to have the syncer trickle these
1623 * out. Adjustment needed to allow registering vnodes for
1624 * sync when the vnode is clean, but the inode dirty. Or
1625 * have ufs itself trickle out inode updates.
1626 *
1627 * If doing a lazy sync, we don't care about metadata or
1628 * data updates, because they are handled by each vnode's
1629 * synclist entry. In this case we are only interested in
1630 * writing back modified inodes.
1631 */
1632 if ((ip->i_flag & (IN_ACCESS | IN_CHANGE | IN_UPDATE |
1633 IN_MODIFY | IN_MODIFIED | IN_ACCESSED)) == 0 &&
1634 (waitfor == MNT_LAZY || (LIST_EMPTY(&vp->v_dirtyblkhd) &&
1635 UVM_OBJ_IS_CLEAN(&vp->v_uobj)))) {
1636 mutex_exit(&vp->v_interlock);
1637 continue;
1638 }
1639 if (vp->v_type == VBLK && is_suspending) {
1640 mutex_exit(&vp->v_interlock);
1641 continue;
1642 }
1643 vmark(mvp, vp);
1644 mutex_exit(&mntvnode_lock);
1645 error = vget(vp, lk_flags);
1646 if (error) {
1647 mutex_enter(&mntvnode_lock);
1648 nvp = vunmark(mvp);
1649 if (error == ENOENT) {
1650 goto loop;
1651 }
1652 continue;
1653 }
1654 if (waitfor == MNT_LAZY) {
1655 error = UFS_WAPBL_BEGIN(vp->v_mount);
1656 if (!error) {
1657 error = ffs_update(vp, NULL, NULL,
1658 UPDATE_CLOSE);
1659 UFS_WAPBL_END(vp->v_mount);
1660 }
1661 } else {
1662 error = VOP_FSYNC(vp, cred, FSYNC_NOLOG |
1663 (waitfor == MNT_WAIT ? FSYNC_WAIT : 0), 0, 0);
1664 }
1665 if (error)
1666 allerror = error;
1667 if (is_suspending)
1668 vrele(vp);
1669 else
1670 vput(vp);
1671 mutex_enter(&mntvnode_lock);
1672 nvp = vunmark(mvp);
1673 }
1674 mutex_exit(&mntvnode_lock);
1675 /*
1676 * Force stale file system control information to be flushed.
1677 */
1678 if (waitfor != MNT_LAZY && (ump->um_devvp->v_numoutput > 0 ||
1679 !LIST_EMPTY(&ump->um_devvp->v_dirtyblkhd))) {
1680 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
1681 if ((error = VOP_FSYNC(ump->um_devvp, cred,
1682 (waitfor == MNT_WAIT ? FSYNC_WAIT : 0) | FSYNC_NOLOG,
1683 0, 0)) != 0)
1684 allerror = error;
1685 VOP_UNLOCK(ump->um_devvp, 0);
1686 if (allerror == 0 && waitfor == MNT_WAIT && !mp->mnt_wapbl) {
1687 mutex_enter(&mntvnode_lock);
1688 goto loop;
1689 }
1690 }
1691 #ifdef QUOTA
1692 qsync(mp);
1693 #endif
1694 /*
1695 * Write back modified superblock.
1696 */
1697 if (fs->fs_fmod != 0) {
1698 fs->fs_fmod = 0;
1699 fs->fs_time = time_second;
1700 error = UFS_WAPBL_BEGIN(mp);
1701 if (error)
1702 allerror = error;
1703 else {
1704 if ((error = ffs_cgupdate(ump, waitfor)))
1705 allerror = error;
1706 UFS_WAPBL_END(mp);
1707 }
1708 }
1709
1710 #ifdef WAPBL
1711 if (mp->mnt_wapbl) {
1712 error = wapbl_flush(mp->mnt_wapbl, 0);
1713 if (error)
1714 allerror = error;
1715 }
1716 #endif
1717
1718 fstrans_done(mp);
1719 vnfree(mvp);
1720 return (allerror);
1721 }
1722
1723 /*
1724 * Look up a FFS dinode number to find its incore vnode, otherwise read it
1725 * in from disk. If it is in core, wait for the lock bit to clear, then
1726 * return the inode locked. Detection and handling of mount points must be
1727 * done by the calling routine.
1728 */
1729 int
1730 ffs_vget(struct mount *mp, ino_t ino, struct vnode **vpp)
1731 {
1732 struct fs *fs;
1733 struct inode *ip;
1734 struct ufsmount *ump;
1735 struct buf *bp;
1736 struct vnode *vp;
1737 dev_t dev;
1738 int error;
1739
1740 ump = VFSTOUFS(mp);
1741 dev = ump->um_dev;
1742
1743 retry:
1744 if ((*vpp = ufs_ihashget(dev, ino, LK_EXCLUSIVE)) != NULL)
1745 return (0);
1746
1747 /* Allocate a new vnode/inode. */
1748 if ((error = getnewvnode(VT_UFS, mp, ffs_vnodeop_p, &vp)) != 0) {
1749 *vpp = NULL;
1750 return (error);
1751 }
1752 ip = pool_cache_get(ffs_inode_cache, PR_WAITOK);
1753
1754 /*
1755 * If someone beat us to it, put back the freshly allocated
1756 * vnode/inode pair and retry.
1757 */
1758 mutex_enter(&ufs_hashlock);
1759 if (ufs_ihashget(dev, ino, 0) != NULL) {
1760 mutex_exit(&ufs_hashlock);
1761 ungetnewvnode(vp);
1762 pool_cache_put(ffs_inode_cache, ip);
1763 goto retry;
1764 }
1765
1766 vp->v_vflag |= VV_LOCKSWORK;
1767
1768 #ifdef XIP
1769 if ((vp->v_mount->mnt_iflag & IMNT_XIP) != 0) {
1770 vp->v_vflag |= VV_XIP;
1771 }
1772 #endif
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 /*
1840 * Finish inode initialization now that aliasing has been resolved.
1841 */
1842
1843 ip->i_devvp = ump->um_devvp;
1844 vref(ip->i_devvp);
1845
1846 /*
1847 * Ensure that uid and gid are correct. This is a temporary
1848 * fix until fsck has been changed to do the update.
1849 */
1850
1851 if (fs->fs_old_inodefmt < FS_44INODEFMT) { /* XXX */
1852 ip->i_uid = ip->i_ffs1_ouid; /* XXX */
1853 ip->i_gid = ip->i_ffs1_ogid; /* XXX */
1854 } /* XXX */
1855 uvm_vnp_setsize(vp, ip->i_size);
1856 *vpp = vp;
1857 return (0);
1858 }
1859
1860 /*
1861 * File handle to vnode
1862 *
1863 * Have to be really careful about stale file handles:
1864 * - check that the inode number is valid
1865 * - call ffs_vget() to get the locked inode
1866 * - check for an unallocated inode (i_mode == 0)
1867 * - check that the given client host has export rights and return
1868 * those rights via. exflagsp and credanonp
1869 */
1870 int
1871 ffs_fhtovp(struct mount *mp, struct fid *fhp, struct vnode **vpp)
1872 {
1873 struct ufid ufh;
1874 struct fs *fs;
1875
1876 if (fhp->fid_len != sizeof(struct ufid))
1877 return EINVAL;
1878
1879 memcpy(&ufh, fhp, sizeof(ufh));
1880 fs = VFSTOUFS(mp)->um_fs;
1881 if (ufh.ufid_ino < ROOTINO ||
1882 ufh.ufid_ino >= fs->fs_ncg * fs->fs_ipg)
1883 return (ESTALE);
1884 return (ufs_fhtovp(mp, &ufh, vpp));
1885 }
1886
1887 /*
1888 * Vnode pointer to File handle
1889 */
1890 /* ARGSUSED */
1891 int
1892 ffs_vptofh(struct vnode *vp, struct fid *fhp, size_t *fh_size)
1893 {
1894 struct inode *ip;
1895 struct ufid ufh;
1896
1897 if (*fh_size < sizeof(struct ufid)) {
1898 *fh_size = sizeof(struct ufid);
1899 return E2BIG;
1900 }
1901 ip = VTOI(vp);
1902 *fh_size = sizeof(struct ufid);
1903 memset(&ufh, 0, sizeof(ufh));
1904 ufh.ufid_len = sizeof(struct ufid);
1905 ufh.ufid_ino = ip->i_number;
1906 ufh.ufid_gen = ip->i_gen;
1907 memcpy(fhp, &ufh, sizeof(ufh));
1908 return (0);
1909 }
1910
1911 void
1912 ffs_init(void)
1913 {
1914 if (ffs_initcount++ > 0)
1915 return;
1916
1917 ffs_inode_cache = pool_cache_init(sizeof(struct inode), 0, 0, 0,
1918 "ffsino", NULL, IPL_NONE, NULL, NULL, NULL);
1919 ffs_dinode1_cache = pool_cache_init(sizeof(struct ufs1_dinode), 0, 0, 0,
1920 "ffsdino1", NULL, IPL_NONE, NULL, NULL, NULL);
1921 ffs_dinode2_cache = pool_cache_init(sizeof(struct ufs2_dinode), 0, 0, 0,
1922 "ffsdino2", NULL, IPL_NONE, NULL, NULL, NULL);
1923 ufs_init();
1924 }
1925
1926 void
1927 ffs_reinit(void)
1928 {
1929
1930 ufs_reinit();
1931 }
1932
1933 void
1934 ffs_done(void)
1935 {
1936 if (--ffs_initcount > 0)
1937 return;
1938
1939 ufs_done();
1940 pool_cache_destroy(ffs_dinode2_cache);
1941 pool_cache_destroy(ffs_dinode1_cache);
1942 pool_cache_destroy(ffs_inode_cache);
1943 }
1944
1945 /*
1946 * Write a superblock and associated information back to disk.
1947 */
1948 int
1949 ffs_sbupdate(struct ufsmount *mp, int waitfor)
1950 {
1951 struct fs *fs = mp->um_fs;
1952 struct buf *bp;
1953 int error = 0;
1954 u_int32_t saveflag;
1955
1956 error = ffs_getblk(mp->um_devvp,
1957 fs->fs_sblockloc / DEV_BSIZE, FFS_NOBLK,
1958 fs->fs_sbsize, false, &bp);
1959 if (error)
1960 return error;
1961 saveflag = fs->fs_flags & FS_INTERNAL;
1962 fs->fs_flags &= ~FS_INTERNAL;
1963
1964 memcpy(bp->b_data, fs, fs->fs_sbsize);
1965
1966 ffs_oldfscompat_write((struct fs *)bp->b_data, mp);
1967 #ifdef FFS_EI
1968 if (mp->um_flags & UFS_NEEDSWAP)
1969 ffs_sb_swap((struct fs *)bp->b_data, (struct fs *)bp->b_data);
1970 #endif
1971 fs->fs_flags |= saveflag;
1972
1973 if (waitfor == MNT_WAIT)
1974 error = bwrite(bp);
1975 else
1976 bawrite(bp);
1977 return (error);
1978 }
1979
1980 int
1981 ffs_cgupdate(struct ufsmount *mp, int waitfor)
1982 {
1983 struct fs *fs = mp->um_fs;
1984 struct buf *bp;
1985 int blks;
1986 void *space;
1987 int i, size, error = 0, allerror = 0;
1988
1989 allerror = ffs_sbupdate(mp, waitfor);
1990 blks = howmany(fs->fs_cssize, fs->fs_fsize);
1991 space = fs->fs_csp;
1992 for (i = 0; i < blks; i += fs->fs_frag) {
1993 size = fs->fs_bsize;
1994 if (i + fs->fs_frag > blks)
1995 size = (blks - i) * fs->fs_fsize;
1996 error = ffs_getblk(mp->um_devvp, fsbtodb(fs, fs->fs_csaddr + i),
1997 FFS_NOBLK, size, false, &bp);
1998 if (error)
1999 break;
2000 #ifdef FFS_EI
2001 if (mp->um_flags & UFS_NEEDSWAP)
2002 ffs_csum_swap((struct csum*)space,
2003 (struct csum*)bp->b_data, size);
2004 else
2005 #endif
2006 memcpy(bp->b_data, space, (u_int)size);
2007 space = (char *)space + size;
2008 if (waitfor == MNT_WAIT)
2009 error = bwrite(bp);
2010 else
2011 bawrite(bp);
2012 }
2013 if (!allerror && error)
2014 allerror = error;
2015 return (allerror);
2016 }
2017
2018 int
2019 ffs_extattrctl(struct mount *mp, int cmd, struct vnode *vp,
2020 int attrnamespace, const char *attrname)
2021 {
2022 #ifdef UFS_EXTATTR
2023 /*
2024 * File-backed extended attributes are only supported on UFS1.
2025 * UFS2 has native extended attributes.
2026 */
2027 if (VFSTOUFS(mp)->um_fstype == UFS1)
2028 return (ufs_extattrctl(mp, cmd, vp, attrnamespace, attrname));
2029 #endif
2030 return (vfs_stdextattrctl(mp, cmd, vp, attrnamespace, attrname));
2031 }
2032
2033 int
2034 ffs_suspendctl(struct mount *mp, int cmd)
2035 {
2036 int error;
2037 struct lwp *l = curlwp;
2038
2039 switch (cmd) {
2040 case SUSPEND_SUSPEND:
2041 if ((error = fstrans_setstate(mp, FSTRANS_SUSPENDING)) != 0)
2042 return error;
2043 error = ffs_sync(mp, MNT_WAIT, l->l_proc->p_cred);
2044 if (error == 0)
2045 error = fstrans_setstate(mp, FSTRANS_SUSPENDED);
2046 #ifdef WAPBL
2047 if (error == 0 && mp->mnt_wapbl)
2048 error = wapbl_flush(mp->mnt_wapbl, 1);
2049 #endif
2050 if (error != 0) {
2051 (void) fstrans_setstate(mp, FSTRANS_NORMAL);
2052 return error;
2053 }
2054 return 0;
2055
2056 case SUSPEND_RESUME:
2057 return fstrans_setstate(mp, FSTRANS_NORMAL);
2058
2059 default:
2060 return EINVAL;
2061 }
2062 }
2063
2064 /*
2065 * Synch vnode for a mounted file system. This is called for foreign
2066 * vnodes, i.e. non-ffs.
2067 */
2068 static int
2069 ffs_vfs_fsync(vnode_t *vp, int flags)
2070 {
2071 int error, passes, skipmeta, i, pflags;
2072 buf_t *bp, *nbp;
2073 #ifdef WAPBL
2074 struct mount *mp;
2075 #endif
2076
2077 KASSERT(vp->v_type == VBLK);
2078 KASSERT(vp->v_specmountpoint != NULL);
2079
2080 /*
2081 * Flush all dirty data associated with the vnode.
2082 */
2083 pflags = PGO_ALLPAGES | PGO_CLEANIT;
2084 if ((flags & FSYNC_WAIT) != 0)
2085 pflags |= PGO_SYNCIO;
2086 mutex_enter(&vp->v_interlock);
2087 error = VOP_PUTPAGES(vp, 0, 0, pflags);
2088 if (error)
2089 return error;
2090
2091 #ifdef WAPBL
2092 mp = vp->v_specmountpoint;
2093 if (mp && mp->mnt_wapbl) {
2094 /*
2095 * Don't bother writing out metadata if the syncer is
2096 * making the request. We will let the sync vnode
2097 * write it out in a single burst through a call to
2098 * VFS_SYNC().
2099 */
2100 if ((flags & (FSYNC_DATAONLY | FSYNC_LAZY | FSYNC_NOLOG)) != 0)
2101 return 0;
2102
2103 /*
2104 * Don't flush the log if the vnode being flushed
2105 * contains no dirty buffers that could be in the log.
2106 */
2107 if (!LIST_EMPTY(&vp->v_dirtyblkhd)) {
2108 error = wapbl_flush(mp->mnt_wapbl, 0);
2109 if (error)
2110 return error;
2111 }
2112
2113 if ((flags & FSYNC_WAIT) != 0) {
2114 mutex_enter(&vp->v_interlock);
2115 while (vp->v_numoutput)
2116 cv_wait(&vp->v_cv, &vp->v_interlock);
2117 mutex_exit(&vp->v_interlock);
2118 }
2119
2120 return 0;
2121 }
2122 #endif /* WAPBL */
2123
2124 /*
2125 * Write out metadata for non-logging file systems. XXX This block
2126 * should be simplified now that softdep is gone.
2127 */
2128 passes = NIADDR + 1;
2129 skipmeta = 0;
2130 if (flags & FSYNC_WAIT)
2131 skipmeta = 1;
2132
2133 loop:
2134 mutex_enter(&bufcache_lock);
2135 LIST_FOREACH(bp, &vp->v_dirtyblkhd, b_vnbufs) {
2136 bp->b_cflags &= ~BC_SCANNED;
2137 }
2138 for (bp = LIST_FIRST(&vp->v_dirtyblkhd); bp; bp = nbp) {
2139 nbp = LIST_NEXT(bp, b_vnbufs);
2140 if (bp->b_cflags & (BC_BUSY | BC_SCANNED))
2141 continue;
2142 if ((bp->b_oflags & BO_DELWRI) == 0)
2143 panic("ffs_fsync: not dirty");
2144 if (skipmeta && bp->b_lblkno < 0)
2145 continue;
2146 bp->b_cflags |= BC_BUSY | BC_VFLUSH | BC_SCANNED;
2147 mutex_exit(&bufcache_lock);
2148 /*
2149 * On our final pass through, do all I/O synchronously
2150 * so that we can find out if our flush is failing
2151 * because of write errors.
2152 */
2153 if (passes > 0 || !(flags & FSYNC_WAIT))
2154 (void) bawrite(bp);
2155 else if ((error = bwrite(bp)) != 0)
2156 return (error);
2157 /*
2158 * Since we unlocked during the I/O, we need
2159 * to start from a known point.
2160 */
2161 mutex_enter(&bufcache_lock);
2162 nbp = LIST_FIRST(&vp->v_dirtyblkhd);
2163 }
2164 mutex_exit(&bufcache_lock);
2165 if (skipmeta) {
2166 skipmeta = 0;
2167 goto loop;
2168 }
2169
2170 if ((flags & FSYNC_WAIT) != 0) {
2171 mutex_enter(&vp->v_interlock);
2172 while (vp->v_numoutput) {
2173 cv_wait(&vp->v_cv, &vp->v_interlock);
2174 }
2175 mutex_exit(&vp->v_interlock);
2176
2177 if (!LIST_EMPTY(&vp->v_dirtyblkhd)) {
2178 /*
2179 * Block devices associated with filesystems may
2180 * have new I/O requests posted for them even if
2181 * the vnode is locked, so no amount of trying will
2182 * get them clean. Thus we give block devices a
2183 * good effort, then just give up. For all other file
2184 * types, go around and try again until it is clean.
2185 */
2186 if (passes > 0) {
2187 passes--;
2188 goto loop;
2189 }
2190 #ifdef DIAGNOSTIC
2191 if (vp->v_type != VBLK)
2192 vprint("ffs_fsync: dirty", vp);
2193 #endif
2194 }
2195 }
2196
2197 if (error == 0 && (flags & FSYNC_CACHE) != 0) {
2198 (void)VOP_IOCTL(vp, DIOCCACHESYNC, &i, FWRITE,
2199 kauth_cred_get());
2200 }
2201
2202 return error;
2203 }
2204