ffs_vfsops.c revision 1.339.2.1 1 /* $NetBSD: ffs_vfsops.c,v 1.339.2.1 2016/07/20 23:47:57 pgoyette 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 1.339.2.1 2016/07/20 23:47:57 pgoyette Exp $");
65
66 #if defined(_KERNEL_OPT)
67 #include "opt_ffs.h"
68 #include "opt_quota.h"
69 #include "opt_wapbl.h"
70 #endif
71
72 #include <sys/param.h>
73 #include <sys/systm.h>
74 #include <sys/namei.h>
75 #include <sys/proc.h>
76 #include <sys/kernel.h>
77 #include <sys/vnode.h>
78 #include <sys/socket.h>
79 #include <sys/mount.h>
80 #include <sys/buf.h>
81 #include <sys/device.h>
82 #include <sys/disk.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/kmem.h>
89 #include <sys/pool.h>
90 #include <sys/lock.h>
91 #include <sys/sysctl.h>
92 #include <sys/conf.h>
93 #include <sys/kauth.h>
94 #include <sys/wapbl.h>
95 #include <sys/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 #ifdef WAPBL
113 MODULE(MODULE_CLASS_VFS, ffs, "wapbl");
114 #else
115 MODULE(MODULE_CLASS_VFS, ffs, NULL);
116 #endif
117
118 static int ffs_vfs_fsync(vnode_t *, int);
119 static int ffs_superblock_validate(struct fs *);
120 static int ffs_is_appleufs(struct vnode *, struct fs *);
121
122 static int ffs_init_vnode(struct ufsmount *, struct vnode *, ino_t);
123 static void ffs_deinit_vnode(struct ufsmount *, struct vnode *);
124
125 static struct sysctllog *ffs_sysctl_log;
126
127 static kauth_listener_t ffs_snapshot_listener;
128
129 /* how many times ffs_init() was called */
130 int ffs_initcount = 0;
131
132 #ifdef DEBUG_FFS_MOUNT
133 #define DPRINTF(_fmt, args...) printf("%s: " _fmt "\n", __func__, ##args)
134 #else
135 #define DPRINTF(_fmt, args...) do {} while (/*CONSTCOND*/0)
136 #endif
137
138 extern const struct vnodeopv_desc ffs_vnodeop_opv_desc;
139 extern const struct vnodeopv_desc ffs_specop_opv_desc;
140 extern const struct vnodeopv_desc ffs_fifoop_opv_desc;
141
142 const struct vnodeopv_desc * const ffs_vnodeopv_descs[] = {
143 &ffs_vnodeop_opv_desc,
144 &ffs_specop_opv_desc,
145 &ffs_fifoop_opv_desc,
146 NULL,
147 };
148
149 struct vfsops ffs_vfsops = {
150 .vfs_name = MOUNT_FFS,
151 .vfs_min_mount_data = sizeof (struct ufs_args),
152 .vfs_mount = ffs_mount,
153 .vfs_start = ufs_start,
154 .vfs_unmount = ffs_unmount,
155 .vfs_root = ufs_root,
156 .vfs_quotactl = ufs_quotactl,
157 .vfs_statvfs = ffs_statvfs,
158 .vfs_sync = ffs_sync,
159 .vfs_vget = ufs_vget,
160 .vfs_loadvnode = ffs_loadvnode,
161 .vfs_newvnode = ffs_newvnode,
162 .vfs_fhtovp = ffs_fhtovp,
163 .vfs_vptofh = ffs_vptofh,
164 .vfs_init = ffs_init,
165 .vfs_reinit = ffs_reinit,
166 .vfs_done = ffs_done,
167 .vfs_mountroot = ffs_mountroot,
168 .vfs_snapshot = ffs_snapshot,
169 .vfs_extattrctl = ffs_extattrctl,
170 .vfs_suspendctl = ffs_suspendctl,
171 .vfs_renamelock_enter = genfs_renamelock_enter,
172 .vfs_renamelock_exit = genfs_renamelock_exit,
173 .vfs_fsync = ffs_vfs_fsync,
174 .vfs_opv_descs = ffs_vnodeopv_descs
175 };
176
177 static const struct genfs_ops ffs_genfsops = {
178 .gop_size = ffs_gop_size,
179 .gop_alloc = ufs_gop_alloc,
180 .gop_write = genfs_gop_write,
181 .gop_markupdate = ufs_gop_markupdate,
182 };
183
184 static const struct ufs_ops ffs_ufsops = {
185 .uo_itimes = ffs_itimes,
186 .uo_update = ffs_update,
187 .uo_truncate = ffs_truncate,
188 .uo_balloc = ffs_balloc,
189 .uo_snapgone = ffs_snapgone,
190 .uo_bufrd = ffs_bufrd,
191 .uo_bufwr = ffs_bufwr,
192 };
193
194 static int
195 ffs_checkrange(struct mount *mp, uint32_t ino)
196 {
197 struct fs *fs = VFSTOUFS(mp)->um_fs;
198
199 if (ino < UFS_ROOTINO || ino >= fs->fs_ncg * fs->fs_ipg) {
200 DPRINTF("out of range %u\n", ino);
201 return ESTALE;
202 }
203
204 /*
205 * Need to check if inode is initialized because ffsv2 does
206 * lazy initialization and we can get here from nfs_fhtovp
207 */
208 if (fs->fs_magic != FS_UFS2_MAGIC)
209 return 0;
210
211 struct buf *bp;
212 int cg = ino_to_cg(fs, ino);
213 struct ufsmount *ump = VFSTOUFS(mp);
214
215 int error = bread(ump->um_devvp, FFS_FSBTODB(fs, cgtod(fs, cg)),
216 (int)fs->fs_cgsize, B_MODIFY, &bp);
217 if (error) {
218 DPRINTF("error %d reading cg %d ino %u\n", error, cg, ino);
219 return error;
220 }
221
222 const int needswap = UFS_FSNEEDSWAP(fs);
223
224 struct cg *cgp = (struct cg *)bp->b_data;
225 if (!cg_chkmagic(cgp, needswap)) {
226 brelse(bp, 0);
227 DPRINTF("bad cylinder group magic cg %d ino %u\n", cg, ino);
228 return ESTALE;
229 }
230
231 int32_t initediblk = ufs_rw32(cgp->cg_initediblk, needswap);
232 brelse(bp, 0);
233
234 if (cg * fs->fs_ipg + initediblk < ino) {
235 DPRINTF("cg=%d fs->fs_ipg=%d initediblk=%d ino=%u\n",
236 cg, fs->fs_ipg, initediblk, ino);
237 return ESTALE;
238 }
239 return 0;
240 }
241
242 static int
243 ffs_snapshot_cb(kauth_cred_t cred, kauth_action_t action, void *cookie,
244 void *arg0, void *arg1, void *arg2, void *arg3)
245 {
246 vnode_t *vp = arg2;
247 int result = KAUTH_RESULT_DEFER;
248
249 if (action != KAUTH_SYSTEM_FS_SNAPSHOT)
250 return result;
251
252 if (VTOI(vp)->i_uid == kauth_cred_geteuid(cred))
253 result = KAUTH_RESULT_ALLOW;
254
255 return result;
256 }
257
258 static int
259 ffs_modcmd(modcmd_t cmd, void *arg)
260 {
261 int error;
262
263 #if 0
264 extern int doasyncfree;
265 #endif
266 #ifdef UFS_EXTATTR
267 extern int ufs_extattr_autocreate;
268 #endif
269 extern int ffs_log_changeopt;
270
271 switch (cmd) {
272 case MODULE_CMD_INIT:
273 error = vfs_attach(&ffs_vfsops);
274 if (error != 0)
275 break;
276
277 sysctl_createv(&ffs_sysctl_log, 0, NULL, NULL,
278 CTLFLAG_PERMANENT,
279 CTLTYPE_NODE, "ffs",
280 SYSCTL_DESCR("Berkeley Fast File System"),
281 NULL, 0, NULL, 0,
282 CTL_VFS, 1, CTL_EOL);
283 /*
284 * @@@ should we even bother with these first three?
285 */
286 sysctl_createv(&ffs_sysctl_log, 0, NULL, NULL,
287 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
288 CTLTYPE_INT, "doclusterread", NULL,
289 sysctl_notavail, 0, NULL, 0,
290 CTL_VFS, 1, FFS_CLUSTERREAD, CTL_EOL);
291 sysctl_createv(&ffs_sysctl_log, 0, NULL, NULL,
292 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
293 CTLTYPE_INT, "doclusterwrite", NULL,
294 sysctl_notavail, 0, NULL, 0,
295 CTL_VFS, 1, FFS_CLUSTERWRITE, CTL_EOL);
296 sysctl_createv(&ffs_sysctl_log, 0, NULL, NULL,
297 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
298 CTLTYPE_INT, "doreallocblks", NULL,
299 sysctl_notavail, 0, NULL, 0,
300 CTL_VFS, 1, FFS_REALLOCBLKS, CTL_EOL);
301 #if 0
302 sysctl_createv(&ffs_sysctl_log, 0, NULL, NULL,
303 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
304 CTLTYPE_INT, "doasyncfree",
305 SYSCTL_DESCR("Release dirty blocks asynchronously"),
306 NULL, 0, &doasyncfree, 0,
307 CTL_VFS, 1, FFS_ASYNCFREE, CTL_EOL);
308 #endif
309 sysctl_createv(&ffs_sysctl_log, 0, NULL, NULL,
310 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
311 CTLTYPE_INT, "log_changeopt",
312 SYSCTL_DESCR("Log changes in optimization strategy"),
313 NULL, 0, &ffs_log_changeopt, 0,
314 CTL_VFS, 1, FFS_LOG_CHANGEOPT, CTL_EOL);
315 #ifdef UFS_EXTATTR
316 sysctl_createv(&ffs_sysctl_log, 0, NULL, NULL,
317 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
318 CTLTYPE_INT, "extattr_autocreate",
319 SYSCTL_DESCR("Size of attribute for "
320 "backing file autocreation"),
321 NULL, 0, &ufs_extattr_autocreate, 0,
322 CTL_VFS, 1, FFS_EXTATTR_AUTOCREATE, CTL_EOL);
323
324 #endif /* UFS_EXTATTR */
325
326 ffs_snapshot_listener = kauth_listen_scope(KAUTH_SCOPE_SYSTEM,
327 ffs_snapshot_cb, NULL);
328 if (ffs_snapshot_listener == NULL)
329 printf("ffs_modcmd: can't listen on system scope.\n");
330
331 break;
332 case MODULE_CMD_FINI:
333 error = vfs_detach(&ffs_vfsops);
334 if (error != 0)
335 break;
336 sysctl_teardown(&ffs_sysctl_log);
337 if (ffs_snapshot_listener != NULL)
338 kauth_unlisten_scope(ffs_snapshot_listener);
339 break;
340 default:
341 error = ENOTTY;
342 break;
343 }
344
345 return (error);
346 }
347
348 pool_cache_t ffs_inode_cache;
349 pool_cache_t ffs_dinode1_cache;
350 pool_cache_t ffs_dinode2_cache;
351
352 static void ffs_oldfscompat_read(struct fs *, struct ufsmount *, daddr_t);
353 static void ffs_oldfscompat_write(struct fs *, struct ufsmount *);
354
355 /*
356 * Called by main() when ffs is going to be mounted as root.
357 */
358
359 int
360 ffs_mountroot(void)
361 {
362 struct fs *fs;
363 struct mount *mp;
364 struct lwp *l = curlwp; /* XXX */
365 struct ufsmount *ump;
366 int error;
367
368 if (device_class(root_device) != DV_DISK)
369 return (ENODEV);
370
371 if ((error = vfs_rootmountalloc(MOUNT_FFS, "root_device", &mp))) {
372 vrele(rootvp);
373 return (error);
374 }
375
376 /*
377 * We always need to be able to mount the root file system.
378 */
379 mp->mnt_flag |= MNT_FORCE;
380 if ((error = ffs_mountfs(rootvp, mp, l)) != 0) {
381 vfs_unbusy(mp, false, NULL);
382 vfs_destroy(mp);
383 return (error);
384 }
385 mp->mnt_flag &= ~MNT_FORCE;
386 mountlist_append(mp);
387 ump = VFSTOUFS(mp);
388 fs = ump->um_fs;
389 memset(fs->fs_fsmnt, 0, sizeof(fs->fs_fsmnt));
390 (void)copystr(mp->mnt_stat.f_mntonname, fs->fs_fsmnt, MNAMELEN - 1, 0);
391 (void)ffs_statvfs(mp, &mp->mnt_stat);
392 vfs_unbusy(mp, false, NULL);
393 setrootfstime((time_t)fs->fs_time);
394 return (0);
395 }
396
397 /*
398 * VFS Operations.
399 *
400 * mount system call
401 */
402 int
403 ffs_mount(struct mount *mp, const char *path, void *data, size_t *data_len)
404 {
405 struct lwp *l = curlwp;
406 struct vnode *devvp = NULL;
407 struct ufs_args *args = data;
408 struct ufsmount *ump = NULL;
409 struct fs *fs;
410 int error = 0, flags, update;
411 mode_t accessmode;
412 const struct bdevsw *bdev = NULL;
413
414 if (args == NULL) {
415 DPRINTF("NULL args");
416 return EINVAL;
417 }
418 if (*data_len < sizeof(*args)) {
419 DPRINTF("bad size args %zu != %zu", *data_len, sizeof(*args));
420 return EINVAL;
421 }
422
423 if (mp->mnt_flag & MNT_GETARGS) {
424 ump = VFSTOUFS(mp);
425 if (ump == NULL) {
426 DPRINTF("no ump");
427 return EIO;
428 }
429 args->fspec = NULL;
430 *data_len = sizeof *args;
431 return 0;
432 }
433
434 update = mp->mnt_flag & MNT_UPDATE;
435
436 /* Check arguments */
437 if (args->fspec != NULL) {
438 /*
439 * Look up the name and verify that it's sane.
440 */
441 error = namei_simple_user(args->fspec,
442 NSM_FOLLOW_NOEMULROOT, &devvp);
443 if (error != 0) {
444 DPRINTF("namei_simple_user returned %d", error);
445 return error;
446 }
447
448 if (!update) {
449 /*
450 * Be sure this is a valid block device
451 */
452 if (devvp->v_type != VBLK) {
453 DPRINTF("non block device %d", devvp->v_type);
454 error = ENOTBLK;
455 } else if (bdevsw_lookup_acquire(devvp->v_rdev) == NULL) {
456 DPRINTF("can't find block device 0x%jx",
457 devvp->v_rdev);
458 error = ENXIO;
459 }
460 } else {
461 /*
462 * Be sure we're still naming the same device
463 * used for our initial mount
464 */
465 ump = VFSTOUFS(mp);
466 if (devvp != ump->um_devvp) {
467 if (devvp->v_rdev != ump->um_devvp->v_rdev) {
468 DPRINTF("wrong device 0x%jx != 0x%jx",
469 (uintmax_t)devvp->v_rdev,
470 (uintmax_t)ump->um_devvp->v_rdev);
471 error = EINVAL;
472 } else {
473 vrele(devvp);
474 devvp = ump->um_devvp;
475 vref(devvp);
476 }
477 }
478 }
479 } else {
480 if (!update) {
481 /* New mounts must have a filename for the device */
482 DPRINTF("no filename for mount");
483 return EINVAL;
484 } else {
485 /* Use the extant mount */
486 ump = VFSTOUFS(mp);
487 devvp = ump->um_devvp;
488 vref(devvp);
489 }
490 }
491
492 /*
493 * If mount by non-root, then verify that user has necessary
494 * permissions on the device.
495 *
496 * Permission to update a mount is checked higher, so here we presume
497 * updating the mount is okay (for example, as far as securelevel goes)
498 * which leaves us with the normal check.
499 */
500 if (error == 0) {
501 accessmode = VREAD;
502 if (update ?
503 (mp->mnt_iflag & IMNT_WANTRDWR) != 0 :
504 (mp->mnt_flag & MNT_RDONLY) == 0)
505 accessmode |= VWRITE;
506 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
507 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT,
508 KAUTH_REQ_SYSTEM_MOUNT_DEVICE, mp, devvp,
509 KAUTH_ARG(accessmode));
510 if (error) {
511 DPRINTF("kauth returned %d", error);
512 }
513 VOP_UNLOCK(devvp);
514 }
515
516 if (error)
517 goto fail;
518
519 #ifdef WAPBL
520 /* WAPBL can only be enabled on a r/w mount. */
521 if ((mp->mnt_flag & MNT_RDONLY) && !(mp->mnt_iflag & IMNT_WANTRDWR)) {
522 mp->mnt_flag &= ~MNT_LOG;
523 }
524 #else /* !WAPBL */
525 mp->mnt_flag &= ~MNT_LOG;
526 #endif /* !WAPBL */
527
528 if (!update) {
529 int xflags;
530
531 if (mp->mnt_flag & MNT_RDONLY)
532 xflags = FREAD;
533 else
534 xflags = FREAD | FWRITE;
535 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
536 error = VOP_OPEN(devvp, xflags, FSCRED);
537 VOP_UNLOCK(devvp);
538 if (error) {
539 DPRINTF("VOP_OPEN returned %d", error);
540 goto fail;
541 }
542 error = ffs_mountfs(devvp, mp, l);
543 if (error) {
544 DPRINTF("ffs_mountfs returned %d", error);
545 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
546 (void)VOP_CLOSE(devvp, xflags, NOCRED);
547 VOP_UNLOCK(devvp);
548 goto fail;
549 }
550
551 ump = VFSTOUFS(mp);
552 fs = ump->um_fs;
553 } else {
554 /*
555 * Update the mount.
556 */
557
558 /*
559 * The initial mount got a reference on this
560 * device, so drop the one obtained via
561 * namei(), above.
562 */
563 vrele(devvp);
564
565 ump = VFSTOUFS(mp);
566 fs = ump->um_fs;
567 if (fs->fs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) {
568 /*
569 * Changing from r/w to r/o
570 */
571 flags = WRITECLOSE;
572 if (mp->mnt_flag & MNT_FORCE)
573 flags |= FORCECLOSE;
574 error = ffs_flushfiles(mp, flags, l);
575 if (error == 0)
576 error = UFS_WAPBL_BEGIN(mp);
577 if (error == 0 &&
578 ffs_cgupdate(ump, MNT_WAIT) == 0 &&
579 fs->fs_clean & FS_WASCLEAN) {
580 if (mp->mnt_flag & MNT_SOFTDEP)
581 fs->fs_flags &= ~FS_DOSOFTDEP;
582 fs->fs_clean = FS_ISCLEAN;
583 (void) ffs_sbupdate(ump, MNT_WAIT);
584 }
585 if (error) {
586 DPRINTF("wapbl %d", error);
587 return error;
588 }
589 UFS_WAPBL_END(mp);
590 }
591
592 #ifdef WAPBL
593 if ((mp->mnt_flag & MNT_LOG) == 0) {
594 error = ffs_wapbl_stop(mp, mp->mnt_flag & MNT_FORCE);
595 if (error) {
596 DPRINTF("ffs_wapbl_stop returned %d", error);
597 return error;
598 }
599 }
600 #endif /* WAPBL */
601
602 if (fs->fs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) {
603 /*
604 * Finish change from r/w to r/o
605 */
606 fs->fs_ronly = 1;
607 fs->fs_fmod = 0;
608 }
609
610 if (mp->mnt_flag & MNT_RELOAD) {
611 error = ffs_reload(mp, l->l_cred, l);
612 if (error) {
613 DPRINTF("ffs_reload returned %d", error);
614 return error;
615 }
616 }
617
618 if (fs->fs_ronly && (mp->mnt_iflag & IMNT_WANTRDWR)) {
619 /*
620 * Changing from read-only to read/write
621 */
622 #ifndef QUOTA2
623 if (fs->fs_flags & FS_DOQUOTA2) {
624 ump->um_flags |= UFS_QUOTA2;
625 uprintf("%s: options QUOTA2 not enabled%s\n",
626 mp->mnt_stat.f_mntonname,
627 (mp->mnt_flag & MNT_FORCE) ? "" :
628 ", not mounting");
629 DPRINTF("ffs_quota2 %d", EINVAL);
630 return EINVAL;
631 }
632 #endif
633 fs->fs_ronly = 0;
634 fs->fs_clean <<= 1;
635 fs->fs_fmod = 1;
636 #ifdef WAPBL
637 if (fs->fs_flags & FS_DOWAPBL) {
638 const char *nm = mp->mnt_stat.f_mntonname;
639 if (!mp->mnt_wapbl_replay) {
640 printf("%s: log corrupted;"
641 " replay cancelled\n", nm);
642 return EFTYPE;
643 }
644 printf("%s: replaying log to disk\n", nm);
645 error = wapbl_replay_write(mp->mnt_wapbl_replay,
646 devvp);
647 if (error) {
648 DPRINTF("%s: wapbl_replay_write %d",
649 nm, error);
650 return error;
651 }
652 wapbl_replay_stop(mp->mnt_wapbl_replay);
653 fs->fs_clean = FS_WASCLEAN;
654 }
655 #endif /* WAPBL */
656 if (fs->fs_snapinum[0] != 0)
657 ffs_snapshot_mount(mp);
658 }
659
660 #ifdef WAPBL
661 error = ffs_wapbl_start(mp);
662 if (error) {
663 DPRINTF("ffs_wapbl_start returned %d", error);
664 return error;
665 }
666 #endif /* WAPBL */
667
668 #ifdef QUOTA2
669 if (!fs->fs_ronly) {
670 error = ffs_quota2_mount(mp);
671 if (error) {
672 DPRINTF("ffs_quota2_mount returned %d", error);
673 return error;
674 }
675 }
676 #endif
677
678 if ((mp->mnt_flag & MNT_DISCARD) && !(ump->um_discarddata))
679 ump->um_discarddata = ffs_discard_init(devvp, fs);
680
681 if (args->fspec == NULL)
682 return 0;
683 }
684
685 error = set_statvfs_info(path, UIO_USERSPACE, args->fspec,
686 UIO_USERSPACE, mp->mnt_op->vfs_name, mp, l);
687 if (error == 0)
688 (void)strncpy(fs->fs_fsmnt, mp->mnt_stat.f_mntonname,
689 sizeof(fs->fs_fsmnt));
690 else {
691 DPRINTF("set_statvfs_info returned %d", error);
692 }
693 fs->fs_flags &= ~FS_DOSOFTDEP;
694 if (fs->fs_fmod != 0) { /* XXX */
695 int err;
696
697 fs->fs_fmod = 0;
698 if (fs->fs_clean & FS_WASCLEAN)
699 fs->fs_time = time_second;
700 else {
701 printf("%s: file system not clean (fs_clean=%#x); "
702 "please fsck(8)\n", mp->mnt_stat.f_mntfromname,
703 fs->fs_clean);
704 printf("%s: lost blocks %" PRId64 " files %d\n",
705 mp->mnt_stat.f_mntfromname, fs->fs_pendingblocks,
706 fs->fs_pendinginodes);
707 }
708 err = UFS_WAPBL_BEGIN(mp);
709 if (err == 0) {
710 (void) ffs_cgupdate(ump, MNT_WAIT);
711 UFS_WAPBL_END(mp);
712 }
713 }
714 if ((mp->mnt_flag & MNT_SOFTDEP) != 0) {
715 printf("%s: `-o softdep' is no longer supported, "
716 "consider `-o log'\n", mp->mnt_stat.f_mntfromname);
717 mp->mnt_flag &= ~MNT_SOFTDEP;
718 }
719
720 if (bdev != NULL)
721 bdevsw_release(bdev);
722 return (error);
723
724 fail:
725 if (bdev != NULL)
726 bdevsw_release(bdev);
727 vrele(devvp);
728 return (error);
729 }
730
731 /*
732 * Reload all incore data for a filesystem (used after running fsck on
733 * the root filesystem and finding things to fix). The filesystem must
734 * be mounted read-only.
735 *
736 * Things to do to update the mount:
737 * 1) invalidate all cached meta-data.
738 * 2) re-read superblock from disk.
739 * 3) re-read summary information from disk.
740 * 4) invalidate all inactive vnodes.
741 * 5) invalidate all cached file data.
742 * 6) re-read inode data for all active vnodes.
743 */
744 int
745 ffs_reload(struct mount *mp, kauth_cred_t cred, struct lwp *l)
746 {
747 struct vnode *vp, *devvp;
748 struct inode *ip;
749 void *space;
750 struct buf *bp;
751 struct fs *fs, *newfs;
752 int i, bsize, blks, error;
753 int32_t *lp, fs_sbsize;
754 struct ufsmount *ump;
755 daddr_t sblockloc;
756 struct vnode_iterator *marker;
757
758 if ((mp->mnt_flag & MNT_RDONLY) == 0)
759 return (EINVAL);
760
761 ump = VFSTOUFS(mp);
762
763 /*
764 * Step 1: invalidate all cached meta-data.
765 */
766 devvp = ump->um_devvp;
767 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
768 error = vinvalbuf(devvp, 0, cred, l, 0, 0);
769 VOP_UNLOCK(devvp);
770 if (error)
771 panic("%s: dirty1", __func__);
772
773 /*
774 * Step 2: re-read superblock from disk. XXX: We don't handle
775 * possibility that superblock moved. Which implies that we don't
776 * want its size to change either.
777 */
778 fs = ump->um_fs;
779 fs_sbsize = fs->fs_sbsize;
780 error = bread(devvp, fs->fs_sblockloc / DEV_BSIZE, fs_sbsize,
781 0, &bp);
782 if (error)
783 return (error);
784 newfs = kmem_alloc(fs_sbsize, KM_SLEEP);
785 memcpy(newfs, bp->b_data, fs_sbsize);
786
787 #ifdef FFS_EI
788 if (ump->um_flags & UFS_NEEDSWAP) {
789 ffs_sb_swap((struct fs *)bp->b_data, newfs);
790 newfs->fs_flags |= FS_SWAPPED;
791 } else
792 #endif
793 newfs->fs_flags &= ~FS_SWAPPED;
794
795 brelse(bp, 0);
796
797 if ((newfs->fs_magic != FS_UFS1_MAGIC) &&
798 (newfs->fs_magic != FS_UFS2_MAGIC)) {
799 kmem_free(newfs, fs_sbsize);
800 return (EIO); /* XXX needs translation */
801 }
802 if (!ffs_superblock_validate(newfs)) {
803 kmem_free(newfs, fs_sbsize);
804 return (EINVAL);
805 }
806
807 /*
808 * The current implementation doesn't handle the possibility that
809 * these values may have changed.
810 */
811 if ((newfs->fs_sbsize != fs_sbsize) ||
812 (newfs->fs_cssize != fs->fs_cssize) ||
813 (newfs->fs_contigsumsize != fs->fs_contigsumsize) ||
814 (newfs->fs_ncg != fs->fs_ncg)) {
815 kmem_free(newfs, fs_sbsize);
816 return (EINVAL);
817 }
818
819 /* Store off old fs_sblockloc for fs_oldfscompat_read. */
820 sblockloc = fs->fs_sblockloc;
821 /*
822 * Copy pointer fields back into superblock before copying in XXX
823 * new superblock. These should really be in the ufsmount. XXX
824 * Note that important parameters (eg fs_ncg) are unchanged.
825 */
826 newfs->fs_csp = fs->fs_csp;
827 newfs->fs_maxcluster = fs->fs_maxcluster;
828 newfs->fs_contigdirs = fs->fs_contigdirs;
829 newfs->fs_ronly = fs->fs_ronly;
830 newfs->fs_active = fs->fs_active;
831 memcpy(fs, newfs, (u_int)fs_sbsize);
832 kmem_free(newfs, fs_sbsize);
833
834 /*
835 * Recheck for Apple UFS filesystem.
836 */
837 ump->um_flags &= ~UFS_ISAPPLEUFS;
838 if (ffs_is_appleufs(devvp, fs)) {
839 #ifdef APPLE_UFS
840 ump->um_flags |= UFS_ISAPPLEUFS;
841 #else
842 DPRINTF("AppleUFS not supported");
843 return (EIO); /* XXX: really? */
844 #endif
845 }
846
847 if (UFS_MPISAPPLEUFS(ump)) {
848 /* see comment about NeXT below */
849 ump->um_maxsymlinklen = APPLEUFS_MAXSYMLINKLEN;
850 ump->um_dirblksiz = APPLEUFS_DIRBLKSIZ;
851 mp->mnt_iflag |= IMNT_DTYPE;
852 } else {
853 ump->um_maxsymlinklen = fs->fs_maxsymlinklen;
854 ump->um_dirblksiz = UFS_DIRBLKSIZ;
855 if (ump->um_maxsymlinklen > 0)
856 mp->mnt_iflag |= IMNT_DTYPE;
857 else
858 mp->mnt_iflag &= ~IMNT_DTYPE;
859 }
860 ffs_oldfscompat_read(fs, ump, sblockloc);
861
862 mutex_enter(&ump->um_lock);
863 ump->um_maxfilesize = fs->fs_maxfilesize;
864 if (fs->fs_flags & ~(FS_KNOWN_FLAGS | FS_INTERNAL)) {
865 uprintf("%s: unknown ufs flags: 0x%08"PRIx32"%s\n",
866 mp->mnt_stat.f_mntonname, fs->fs_flags,
867 (mp->mnt_flag & MNT_FORCE) ? "" : ", not mounting");
868 if ((mp->mnt_flag & MNT_FORCE) == 0) {
869 mutex_exit(&ump->um_lock);
870 return (EINVAL);
871 }
872 }
873 if (fs->fs_pendingblocks != 0 || fs->fs_pendinginodes != 0) {
874 fs->fs_pendingblocks = 0;
875 fs->fs_pendinginodes = 0;
876 }
877 mutex_exit(&ump->um_lock);
878
879 ffs_statvfs(mp, &mp->mnt_stat);
880 /*
881 * Step 3: re-read summary information from disk.
882 */
883 blks = howmany(fs->fs_cssize, fs->fs_fsize);
884 space = fs->fs_csp;
885 for (i = 0; i < blks; i += fs->fs_frag) {
886 bsize = fs->fs_bsize;
887 if (i + fs->fs_frag > blks)
888 bsize = (blks - i) * fs->fs_fsize;
889 error = bread(devvp, FFS_FSBTODB(fs, fs->fs_csaddr + i), bsize,
890 0, &bp);
891 if (error) {
892 return (error);
893 }
894 #ifdef FFS_EI
895 if (UFS_FSNEEDSWAP(fs))
896 ffs_csum_swap((struct csum *)bp->b_data,
897 (struct csum *)space, bsize);
898 else
899 #endif
900 memcpy(space, bp->b_data, (size_t)bsize);
901 space = (char *)space + bsize;
902 brelse(bp, 0);
903 }
904 /*
905 * We no longer know anything about clusters per cylinder group.
906 */
907 if (fs->fs_contigsumsize > 0) {
908 lp = fs->fs_maxcluster;
909 for (i = 0; i < fs->fs_ncg; i++)
910 *lp++ = fs->fs_contigsumsize;
911 }
912
913 vfs_vnode_iterator_init(mp, &marker);
914 while ((vp = vfs_vnode_iterator_next(marker, NULL, NULL))) {
915 /*
916 * Step 4: invalidate all inactive vnodes.
917 */
918 if (vrecycle(vp))
919 continue;
920 /*
921 * Step 5: invalidate all cached file data.
922 */
923 if (vn_lock(vp, LK_EXCLUSIVE)) {
924 vrele(vp);
925 continue;
926 }
927 if (vinvalbuf(vp, 0, cred, l, 0, 0))
928 panic("%s: dirty2", __func__);
929 /*
930 * Step 6: re-read inode data for all active vnodes.
931 */
932 ip = VTOI(vp);
933 error = bread(devvp, FFS_FSBTODB(fs, ino_to_fsba(fs, ip->i_number)),
934 (int)fs->fs_bsize, 0, &bp);
935 if (error) {
936 vput(vp);
937 break;
938 }
939 ffs_load_inode(bp, ip, fs, ip->i_number);
940 brelse(bp, 0);
941 vput(vp);
942 }
943 vfs_vnode_iterator_destroy(marker);
944 return (error);
945 }
946
947 /*
948 * Possible superblock locations ordered from most to least likely.
949 */
950 static const int sblock_try[] = SBLOCKSEARCH;
951
952
953 static int
954 ffs_superblock_validate(struct fs *fs)
955 {
956 int32_t i, fs_bshift = 0, fs_fshift = 0, fs_fragshift = 0, fs_frag;
957 int32_t fs_inopb;
958
959 /* Check the superblock size */
960 if (fs->fs_sbsize > SBLOCKSIZE || fs->fs_sbsize < sizeof(struct fs))
961 return 0;
962
963 /* Check the file system blocksize */
964 if (fs->fs_bsize > MAXBSIZE || fs->fs_bsize < MINBSIZE)
965 return 0;
966 if (!powerof2(fs->fs_bsize))
967 return 0;
968
969 /* Check the size of frag blocks */
970 if (!powerof2(fs->fs_fsize))
971 return 0;
972 if (fs->fs_fsize == 0)
973 return 0;
974
975 /*
976 * XXX: these values are just zero-checked to prevent obvious
977 * bugs. We need more strict checks.
978 */
979 if (fs->fs_size == 0)
980 return 0;
981 if (fs->fs_cssize == 0)
982 return 0;
983 if (fs->fs_ipg == 0)
984 return 0;
985 if (fs->fs_fpg == 0)
986 return 0;
987 if (fs->fs_ncg == 0)
988 return 0;
989 if (fs->fs_maxbpg == 0)
990 return 0;
991
992 /* Check the number of inodes per block */
993 if (fs->fs_magic == FS_UFS1_MAGIC)
994 fs_inopb = fs->fs_bsize / sizeof(struct ufs1_dinode);
995 else /* fs->fs_magic == FS_UFS2_MAGIC */
996 fs_inopb = fs->fs_bsize / sizeof(struct ufs2_dinode);
997 if (fs->fs_inopb != fs_inopb)
998 return 0;
999
1000 /* Block size cannot be smaller than fragment size */
1001 if (fs->fs_bsize < fs->fs_fsize)
1002 return 0;
1003
1004 /* Compute fs_bshift and ensure it is consistent */
1005 for (i = fs->fs_bsize; i > 1; i >>= 1)
1006 fs_bshift++;
1007 if (fs->fs_bshift != fs_bshift)
1008 return 0;
1009
1010 /* Compute fs_fshift and ensure it is consistent */
1011 for (i = fs->fs_fsize; i > 1; i >>= 1)
1012 fs_fshift++;
1013 if (fs->fs_fshift != fs_fshift)
1014 return 0;
1015
1016 /* Compute fs_fragshift and ensure it is consistent */
1017 for (i = fs->fs_frag; i > 1; i >>= 1)
1018 fs_fragshift++;
1019 if (fs->fs_fragshift != fs_fragshift)
1020 return 0;
1021
1022 /* Check the masks */
1023 if (fs->fs_bmask != ~(fs->fs_bsize - 1))
1024 return 0;
1025 if (fs->fs_fmask != ~(fs->fs_fsize - 1))
1026 return 0;
1027
1028 /*
1029 * Now that the shifts and masks are sanitized, we can use the ffs_ API.
1030 */
1031
1032 /* Check the number of frag blocks */
1033 if ((fs_frag = ffs_numfrags(fs, fs->fs_bsize)) > MAXFRAG)
1034 return 0;
1035 if (fs->fs_frag != fs_frag)
1036 return 0;
1037
1038 /* Check the size of cylinder groups */
1039 if ((fs->fs_cgsize < sizeof(struct cg)) ||
1040 (fs->fs_cgsize > fs->fs_bsize))
1041 return 0;
1042
1043 return 1;
1044 }
1045
1046 static int
1047 ffs_is_appleufs(struct vnode *devvp, struct fs *fs)
1048 {
1049 struct dkwedge_info dkw;
1050 int ret = 0;
1051
1052 /*
1053 * First check to see if this is tagged as an Apple UFS filesystem
1054 * in the disklabel.
1055 */
1056 if (getdiskinfo(devvp, &dkw) == 0 &&
1057 strcmp(dkw.dkw_ptype, DKW_PTYPE_APPLEUFS) == 0)
1058 ret = 1;
1059 #ifdef APPLE_UFS
1060 else {
1061 struct appleufslabel *applefs;
1062 struct buf *bp;
1063 daddr_t blkno = APPLEUFS_LABEL_OFFSET / DEV_BSIZE;
1064 int error;
1065
1066 /*
1067 * Manually look for an Apple UFS label, and if a valid one
1068 * is found, then treat it like an Apple UFS filesystem anyway.
1069 */
1070 error = bread(devvp, blkno, APPLEUFS_LABEL_SIZE, 0, &bp);
1071 if (error) {
1072 DPRINTF("bread@0x%jx returned %d", (intmax_t)blkno, error);
1073 return 0;
1074 }
1075 applefs = (struct appleufslabel *)bp->b_data;
1076 error = ffs_appleufs_validate(fs->fs_fsmnt, applefs, NULL);
1077 if (error == 0)
1078 ret = 1;
1079 brelse(bp, 0);
1080 }
1081 #endif
1082
1083 return ret;
1084 }
1085
1086 /*
1087 * Common code for mount and mountroot
1088 */
1089 int
1090 ffs_mountfs(struct vnode *devvp, struct mount *mp, struct lwp *l)
1091 {
1092 struct ufsmount *ump = NULL;
1093 struct buf *bp = NULL;
1094 struct fs *fs = NULL;
1095 dev_t dev;
1096 void *space;
1097 daddr_t sblockloc = 0;
1098 int blks, fstype = 0;
1099 int error, i, bsize, ronly, bset = 0;
1100 #ifdef FFS_EI
1101 int needswap = 0; /* keep gcc happy */
1102 #endif
1103 int32_t *lp;
1104 kauth_cred_t cred;
1105 u_int32_t allocsbsize, fs_sbsize = 0;
1106
1107 dev = devvp->v_rdev;
1108 cred = l ? l->l_cred : NOCRED;
1109
1110 /* Flush out any old buffers remaining from a previous use. */
1111 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
1112 error = vinvalbuf(devvp, V_SAVE, cred, l, 0, 0);
1113 VOP_UNLOCK(devvp);
1114 if (error) {
1115 DPRINTF("vinvalbuf returned %d", error);
1116 return error;
1117 }
1118
1119 ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
1120
1121 error = fstrans_mount(mp);
1122 if (error) {
1123 DPRINTF("fstrans_mount returned %d", error);
1124 return error;
1125 }
1126
1127 ump = kmem_zalloc(sizeof(*ump), KM_SLEEP);
1128 mutex_init(&ump->um_lock, MUTEX_DEFAULT, IPL_NONE);
1129 error = ffs_snapshot_init(ump);
1130 if (error) {
1131 DPRINTF("ffs_snapshot_init returned %d", error);
1132 goto out;
1133 }
1134 ump->um_ops = &ffs_ufsops;
1135
1136 #ifdef WAPBL
1137 sbagain:
1138 #endif
1139 /*
1140 * Try reading the superblock in each of its possible locations.
1141 */
1142 for (i = 0; ; i++) {
1143 daddr_t fs_sblockloc;
1144
1145 if (bp != NULL) {
1146 brelse(bp, BC_NOCACHE);
1147 bp = NULL;
1148 }
1149 if (sblock_try[i] == -1) {
1150 DPRINTF("no superblock found");
1151 error = EINVAL;
1152 fs = NULL;
1153 goto out;
1154 }
1155
1156 error = bread(devvp, sblock_try[i] / DEV_BSIZE, SBLOCKSIZE,
1157 0, &bp);
1158 if (error) {
1159 DPRINTF("bread@0x%x returned %d",
1160 sblock_try[i] / DEV_BSIZE, error);
1161 fs = NULL;
1162 goto out;
1163 }
1164 fs = (struct fs *)bp->b_data;
1165
1166 sblockloc = sblock_try[i];
1167 DPRINTF("fs_magic 0x%x", fs->fs_magic);
1168
1169 /*
1170 * Swap: here, we swap fs->fs_sbsize in order to get the correct
1171 * size to read the superblock. Once read, we swap the whole
1172 * superblock structure.
1173 */
1174 if (fs->fs_magic == FS_UFS1_MAGIC) {
1175 fs_sbsize = fs->fs_sbsize;
1176 fstype = UFS1;
1177 #ifdef FFS_EI
1178 needswap = 0;
1179 } else if (fs->fs_magic == FS_UFS1_MAGIC_SWAPPED) {
1180 fs_sbsize = bswap32(fs->fs_sbsize);
1181 fstype = UFS1;
1182 needswap = 1;
1183 #endif
1184 } else if (fs->fs_magic == FS_UFS2_MAGIC) {
1185 fs_sbsize = fs->fs_sbsize;
1186 fstype = UFS2;
1187 #ifdef FFS_EI
1188 needswap = 0;
1189 } else if (fs->fs_magic == FS_UFS2_MAGIC_SWAPPED) {
1190 fs_sbsize = bswap32(fs->fs_sbsize);
1191 fstype = UFS2;
1192 needswap = 1;
1193 #endif
1194 } else
1195 continue;
1196
1197 /* fs->fs_sblockloc isn't defined for old filesystems */
1198 if (fstype == UFS1 && !(fs->fs_old_flags & FS_FLAGS_UPDATED)) {
1199 if (sblockloc == SBLOCK_UFS2)
1200 /*
1201 * This is likely to be the first alternate
1202 * in a filesystem with 64k blocks.
1203 * Don't use it.
1204 */
1205 continue;
1206 fs_sblockloc = sblockloc;
1207 } else {
1208 fs_sblockloc = fs->fs_sblockloc;
1209 #ifdef FFS_EI
1210 if (needswap)
1211 fs_sblockloc = bswap64(fs_sblockloc);
1212 #endif
1213 }
1214
1215 /* Check we haven't found an alternate superblock */
1216 if (fs_sblockloc != sblockloc)
1217 continue;
1218
1219 /* Check the superblock size */
1220 if (fs_sbsize > SBLOCKSIZE || fs_sbsize < sizeof(struct fs))
1221 continue;
1222 fs = kmem_alloc((u_long)fs_sbsize, KM_SLEEP);
1223 memcpy(fs, bp->b_data, fs_sbsize);
1224
1225 /* Swap the whole superblock structure, if necessary. */
1226 #ifdef FFS_EI
1227 if (needswap) {
1228 ffs_sb_swap((struct fs*)bp->b_data, fs);
1229 fs->fs_flags |= FS_SWAPPED;
1230 } else
1231 #endif
1232 fs->fs_flags &= ~FS_SWAPPED;
1233
1234 /*
1235 * Now that everything is swapped, the superblock is ready to
1236 * be sanitized.
1237 */
1238 if (!ffs_superblock_validate(fs)) {
1239 kmem_free(fs, fs_sbsize);
1240 continue;
1241 }
1242
1243 /* Ok seems to be a good superblock */
1244 break;
1245 }
1246
1247 ump->um_fs = fs;
1248
1249 #ifdef WAPBL
1250 if ((mp->mnt_wapbl_replay == 0) && (fs->fs_flags & FS_DOWAPBL)) {
1251 error = ffs_wapbl_replay_start(mp, fs, devvp);
1252 if (error && (mp->mnt_flag & MNT_FORCE) == 0) {
1253 DPRINTF("ffs_wapbl_replay_start returned %d", error);
1254 goto out;
1255 }
1256 if (!error) {
1257 if (!ronly) {
1258 /* XXX fsmnt may be stale. */
1259 printf("%s: replaying log to disk\n",
1260 fs->fs_fsmnt);
1261 error = wapbl_replay_write(mp->mnt_wapbl_replay,
1262 devvp);
1263 if (error) {
1264 DPRINTF("wapbl_replay_write returned %d",
1265 error);
1266 goto out;
1267 }
1268 wapbl_replay_stop(mp->mnt_wapbl_replay);
1269 fs->fs_clean = FS_WASCLEAN;
1270 } else {
1271 /* XXX fsmnt may be stale */
1272 printf("%s: replaying log to memory\n",
1273 fs->fs_fsmnt);
1274 }
1275
1276 /* Force a re-read of the superblock */
1277 brelse(bp, BC_INVAL);
1278 bp = NULL;
1279 kmem_free(fs, fs_sbsize);
1280 fs = NULL;
1281 goto sbagain;
1282 }
1283 }
1284 #else /* !WAPBL */
1285 if ((fs->fs_flags & FS_DOWAPBL) && (mp->mnt_flag & MNT_FORCE) == 0) {
1286 error = EPERM;
1287 DPRINTF("no force %d", error);
1288 goto out;
1289 }
1290 #endif /* !WAPBL */
1291
1292 ffs_oldfscompat_read(fs, ump, sblockloc);
1293 ump->um_maxfilesize = fs->fs_maxfilesize;
1294
1295 if (fs->fs_flags & ~(FS_KNOWN_FLAGS | FS_INTERNAL)) {
1296 uprintf("%s: unknown ufs flags: 0x%08"PRIx32"%s\n",
1297 mp->mnt_stat.f_mntonname, fs->fs_flags,
1298 (mp->mnt_flag & MNT_FORCE) ? "" : ", not mounting");
1299 if ((mp->mnt_flag & MNT_FORCE) == 0) {
1300 error = EINVAL;
1301 DPRINTF("no force %d", error);
1302 goto out;
1303 }
1304 }
1305
1306 if (fs->fs_pendingblocks != 0 || fs->fs_pendinginodes != 0) {
1307 fs->fs_pendingblocks = 0;
1308 fs->fs_pendinginodes = 0;
1309 }
1310
1311 ump->um_fstype = fstype;
1312 if (fs->fs_sbsize < SBLOCKSIZE)
1313 brelse(bp, BC_INVAL);
1314 else
1315 brelse(bp, 0);
1316 bp = NULL;
1317
1318 if (ffs_is_appleufs(devvp, fs)) {
1319 #ifdef APPLE_UFS
1320 ump->um_flags |= UFS_ISAPPLEUFS;
1321 #else
1322 DPRINTF("AppleUFS not supported");
1323 error = EINVAL;
1324 goto out;
1325 #endif
1326 }
1327
1328 #if 0
1329 /*
1330 * XXX This code changes the behaviour of mounting dirty filesystems, to
1331 * XXX require "mount -f ..." to mount them. This doesn't match what
1332 * XXX mount(8) describes and is disabled for now.
1333 */
1334 /*
1335 * If the file system is not clean, don't allow it to be mounted
1336 * unless MNT_FORCE is specified. (Note: MNT_FORCE is always set
1337 * for the root file system.)
1338 */
1339 if (fs->fs_flags & FS_DOWAPBL) {
1340 /*
1341 * wapbl normally expects to be FS_WASCLEAN when the FS_DOWAPBL
1342 * bit is set, although there's a window in unmount where it
1343 * could be FS_ISCLEAN
1344 */
1345 if ((mp->mnt_flag & MNT_FORCE) == 0 &&
1346 (fs->fs_clean & (FS_WASCLEAN | FS_ISCLEAN)) == 0) {
1347 error = EPERM;
1348 goto out;
1349 }
1350 } else
1351 if ((fs->fs_clean & FS_ISCLEAN) == 0 &&
1352 (mp->mnt_flag & MNT_FORCE) == 0) {
1353 error = EPERM;
1354 goto out;
1355 }
1356 #endif
1357
1358 /*
1359 * Verify that we can access the last block in the fs
1360 * if we're mounting read/write.
1361 */
1362 if (!ronly) {
1363 error = bread(devvp, FFS_FSBTODB(fs, fs->fs_size - 1),
1364 fs->fs_fsize, 0, &bp);
1365 if (error) {
1366 DPRINTF("bread@0x%jx returned %d",
1367 (intmax_t)FFS_FSBTODB(fs, fs->fs_size - 1),
1368 error);
1369 bset = BC_INVAL;
1370 goto out;
1371 }
1372 if (bp->b_bcount != fs->fs_fsize) {
1373 DPRINTF("bcount %x != fsize %x", bp->b_bcount,
1374 fs->fs_fsize);
1375 error = EINVAL;
1376 bset = BC_INVAL;
1377 goto out;
1378 }
1379 brelse(bp, BC_INVAL);
1380 bp = NULL;
1381 }
1382
1383 fs->fs_ronly = ronly;
1384 /* Don't bump fs_clean if we're replaying journal */
1385 if (!((fs->fs_flags & FS_DOWAPBL) && (fs->fs_clean & FS_WASCLEAN))) {
1386 if (ronly == 0) {
1387 fs->fs_clean <<= 1;
1388 fs->fs_fmod = 1;
1389 }
1390 }
1391
1392 bsize = fs->fs_cssize;
1393 blks = howmany(bsize, fs->fs_fsize);
1394 if (fs->fs_contigsumsize > 0)
1395 bsize += fs->fs_ncg * sizeof(int32_t);
1396 bsize += fs->fs_ncg * sizeof(*fs->fs_contigdirs);
1397 allocsbsize = bsize;
1398 space = kmem_alloc((u_long)allocsbsize, KM_SLEEP);
1399 fs->fs_csp = space;
1400
1401 for (i = 0; i < blks; i += fs->fs_frag) {
1402 bsize = fs->fs_bsize;
1403 if (i + fs->fs_frag > blks)
1404 bsize = (blks - i) * fs->fs_fsize;
1405 error = bread(devvp, FFS_FSBTODB(fs, fs->fs_csaddr + i), bsize,
1406 0, &bp);
1407 if (error) {
1408 DPRINTF("bread@0x%jx %d",
1409 (intmax_t)FFS_FSBTODB(fs, fs->fs_csaddr + i),
1410 error);
1411 goto out1;
1412 }
1413 #ifdef FFS_EI
1414 if (needswap)
1415 ffs_csum_swap((struct csum *)bp->b_data,
1416 (struct csum *)space, bsize);
1417 else
1418 #endif
1419 memcpy(space, bp->b_data, (u_int)bsize);
1420
1421 space = (char *)space + bsize;
1422 brelse(bp, 0);
1423 bp = NULL;
1424 }
1425 if (fs->fs_contigsumsize > 0) {
1426 fs->fs_maxcluster = lp = space;
1427 for (i = 0; i < fs->fs_ncg; i++)
1428 *lp++ = fs->fs_contigsumsize;
1429 space = lp;
1430 }
1431 bsize = fs->fs_ncg * sizeof(*fs->fs_contigdirs);
1432 fs->fs_contigdirs = space;
1433 space = (char *)space + bsize;
1434 memset(fs->fs_contigdirs, 0, bsize);
1435
1436 /* Compatibility for old filesystems - XXX */
1437 if (fs->fs_avgfilesize <= 0)
1438 fs->fs_avgfilesize = AVFILESIZ;
1439 if (fs->fs_avgfpdir <= 0)
1440 fs->fs_avgfpdir = AFPDIR;
1441 fs->fs_active = NULL;
1442
1443 mp->mnt_data = ump;
1444 mp->mnt_stat.f_fsidx.__fsid_val[0] = (long)dev;
1445 mp->mnt_stat.f_fsidx.__fsid_val[1] = makefstype(MOUNT_FFS);
1446 mp->mnt_stat.f_fsid = mp->mnt_stat.f_fsidx.__fsid_val[0];
1447 mp->mnt_stat.f_namemax = FFS_MAXNAMLEN;
1448 if (UFS_MPISAPPLEUFS(ump)) {
1449 /* NeXT used to keep short symlinks in the inode even
1450 * when using FS_42INODEFMT. In that case fs->fs_maxsymlinklen
1451 * is probably -1, but we still need to be able to identify
1452 * short symlinks.
1453 */
1454 ump->um_maxsymlinklen = APPLEUFS_MAXSYMLINKLEN;
1455 ump->um_dirblksiz = APPLEUFS_DIRBLKSIZ;
1456 mp->mnt_iflag |= IMNT_DTYPE;
1457 } else {
1458 ump->um_maxsymlinklen = fs->fs_maxsymlinklen;
1459 ump->um_dirblksiz = UFS_DIRBLKSIZ;
1460 if (ump->um_maxsymlinklen > 0)
1461 mp->mnt_iflag |= IMNT_DTYPE;
1462 else
1463 mp->mnt_iflag &= ~IMNT_DTYPE;
1464 }
1465 mp->mnt_fs_bshift = fs->fs_bshift;
1466 mp->mnt_dev_bshift = DEV_BSHIFT; /* XXX */
1467 mp->mnt_flag |= MNT_LOCAL;
1468 mp->mnt_iflag |= IMNT_MPSAFE;
1469 #ifdef FFS_EI
1470 if (needswap)
1471 ump->um_flags |= UFS_NEEDSWAP;
1472 #endif
1473 ump->um_mountp = mp;
1474 ump->um_dev = dev;
1475 ump->um_devvp = devvp;
1476 ump->um_nindir = fs->fs_nindir;
1477 ump->um_lognindir = ffs(fs->fs_nindir) - 1;
1478 ump->um_bptrtodb = fs->fs_fshift - DEV_BSHIFT;
1479 ump->um_seqinc = fs->fs_frag;
1480 for (i = 0; i < MAXQUOTAS; i++)
1481 ump->um_quotas[i] = NULLVP;
1482 spec_node_setmountedfs(devvp, mp);
1483 if (ronly == 0 && fs->fs_snapinum[0] != 0)
1484 ffs_snapshot_mount(mp);
1485 #ifdef WAPBL
1486 if (!ronly) {
1487 KDASSERT(fs->fs_ronly == 0);
1488 /*
1489 * ffs_wapbl_start() needs mp->mnt_stat initialised if it
1490 * needs to create a new log file in-filesystem.
1491 */
1492 error = ffs_statvfs(mp, &mp->mnt_stat);
1493 if (error) {
1494 DPRINTF("ffs_statvfs returned %d", error);
1495 goto out1;
1496 }
1497
1498 error = ffs_wapbl_start(mp);
1499 if (error) {
1500 DPRINTF("ffs_wapbl_start returned %d", error);
1501 goto out1;
1502 }
1503 }
1504 #endif /* WAPBL */
1505 if (ronly == 0) {
1506 #ifdef QUOTA2
1507 error = ffs_quota2_mount(mp);
1508 if (error) {
1509 DPRINTF("ffs_quota2_mount returned %d", error);
1510 goto out1;
1511 }
1512 #else
1513 if (fs->fs_flags & FS_DOQUOTA2) {
1514 ump->um_flags |= UFS_QUOTA2;
1515 uprintf("%s: options QUOTA2 not enabled%s\n",
1516 mp->mnt_stat.f_mntonname,
1517 (mp->mnt_flag & MNT_FORCE) ? "" : ", not mounting");
1518 if ((mp->mnt_flag & MNT_FORCE) == 0) {
1519 error = EINVAL;
1520 DPRINTF("quota disabled %d", error);
1521 goto out1;
1522 }
1523 }
1524 #endif
1525 }
1526
1527 if (mp->mnt_flag & MNT_DISCARD)
1528 ump->um_discarddata = ffs_discard_init(devvp, fs);
1529
1530 return (0);
1531 out1:
1532 kmem_free(fs->fs_csp, allocsbsize);
1533 out:
1534 #ifdef WAPBL
1535 if (mp->mnt_wapbl_replay) {
1536 wapbl_replay_stop(mp->mnt_wapbl_replay);
1537 wapbl_replay_free(mp->mnt_wapbl_replay);
1538 mp->mnt_wapbl_replay = 0;
1539 }
1540 #endif
1541
1542 fstrans_unmount(mp);
1543 if (fs)
1544 kmem_free(fs, fs->fs_sbsize);
1545 spec_node_setmountedfs(devvp, NULL);
1546 if (bp)
1547 brelse(bp, bset);
1548 if (ump) {
1549 if (ump->um_oldfscompat)
1550 kmem_free(ump->um_oldfscompat, 512 + 3*sizeof(int32_t));
1551 mutex_destroy(&ump->um_lock);
1552 kmem_free(ump, sizeof(*ump));
1553 mp->mnt_data = NULL;
1554 }
1555 return (error);
1556 }
1557
1558 /*
1559 * Sanity checks for loading old filesystem superblocks.
1560 * See ffs_oldfscompat_write below for unwound actions.
1561 *
1562 * XXX - Parts get retired eventually.
1563 * Unfortunately new bits get added.
1564 */
1565 static void
1566 ffs_oldfscompat_read(struct fs *fs, struct ufsmount *ump, daddr_t sblockloc)
1567 {
1568 off_t maxfilesize;
1569 int32_t *extrasave;
1570
1571 if ((fs->fs_magic != FS_UFS1_MAGIC) ||
1572 (fs->fs_old_flags & FS_FLAGS_UPDATED))
1573 return;
1574
1575 if (!ump->um_oldfscompat)
1576 ump->um_oldfscompat = kmem_alloc(512 + 3*sizeof(int32_t),
1577 KM_SLEEP);
1578
1579 memcpy(ump->um_oldfscompat, &fs->fs_old_postbl_start, 512);
1580 extrasave = ump->um_oldfscompat;
1581 extrasave += 512/sizeof(int32_t);
1582 extrasave[0] = fs->fs_old_npsect;
1583 extrasave[1] = fs->fs_old_interleave;
1584 extrasave[2] = fs->fs_old_trackskew;
1585
1586 /* These fields will be overwritten by their
1587 * original values in fs_oldfscompat_write, so it is harmless
1588 * to modify them here.
1589 */
1590 fs->fs_cstotal.cs_ndir = fs->fs_old_cstotal.cs_ndir;
1591 fs->fs_cstotal.cs_nbfree = fs->fs_old_cstotal.cs_nbfree;
1592 fs->fs_cstotal.cs_nifree = fs->fs_old_cstotal.cs_nifree;
1593 fs->fs_cstotal.cs_nffree = fs->fs_old_cstotal.cs_nffree;
1594
1595 fs->fs_maxbsize = fs->fs_bsize;
1596 fs->fs_time = fs->fs_old_time;
1597 fs->fs_size = fs->fs_old_size;
1598 fs->fs_dsize = fs->fs_old_dsize;
1599 fs->fs_csaddr = fs->fs_old_csaddr;
1600 fs->fs_sblockloc = sblockloc;
1601
1602 fs->fs_flags = fs->fs_old_flags | (fs->fs_flags & FS_INTERNAL);
1603
1604 if (fs->fs_old_postblformat == FS_42POSTBLFMT) {
1605 fs->fs_old_nrpos = 8;
1606 fs->fs_old_npsect = fs->fs_old_nsect;
1607 fs->fs_old_interleave = 1;
1608 fs->fs_old_trackskew = 0;
1609 }
1610
1611 if (fs->fs_old_inodefmt < FS_44INODEFMT) {
1612 fs->fs_maxfilesize = (u_quad_t) 1LL << 39;
1613 fs->fs_qbmask = ~fs->fs_bmask;
1614 fs->fs_qfmask = ~fs->fs_fmask;
1615 }
1616
1617 maxfilesize = (u_int64_t)0x80000000 * fs->fs_bsize - 1;
1618 if (fs->fs_maxfilesize > maxfilesize)
1619 fs->fs_maxfilesize = maxfilesize;
1620
1621 /* Compatibility for old filesystems */
1622 if (fs->fs_avgfilesize <= 0)
1623 fs->fs_avgfilesize = AVFILESIZ;
1624 if (fs->fs_avgfpdir <= 0)
1625 fs->fs_avgfpdir = AFPDIR;
1626
1627 #if 0
1628 if (bigcgs) {
1629 fs->fs_save_cgsize = fs->fs_cgsize;
1630 fs->fs_cgsize = fs->fs_bsize;
1631 }
1632 #endif
1633 }
1634
1635 /*
1636 * Unwinding superblock updates for old filesystems.
1637 * See ffs_oldfscompat_read above for details.
1638 *
1639 * XXX - Parts get retired eventually.
1640 * Unfortunately new bits get added.
1641 */
1642 static void
1643 ffs_oldfscompat_write(struct fs *fs, struct ufsmount *ump)
1644 {
1645 int32_t *extrasave;
1646
1647 if ((fs->fs_magic != FS_UFS1_MAGIC) ||
1648 (fs->fs_old_flags & FS_FLAGS_UPDATED))
1649 return;
1650
1651 fs->fs_old_time = fs->fs_time;
1652 fs->fs_old_cstotal.cs_ndir = fs->fs_cstotal.cs_ndir;
1653 fs->fs_old_cstotal.cs_nbfree = fs->fs_cstotal.cs_nbfree;
1654 fs->fs_old_cstotal.cs_nifree = fs->fs_cstotal.cs_nifree;
1655 fs->fs_old_cstotal.cs_nffree = fs->fs_cstotal.cs_nffree;
1656 fs->fs_old_flags = fs->fs_flags;
1657
1658 #if 0
1659 if (bigcgs) {
1660 fs->fs_cgsize = fs->fs_save_cgsize;
1661 }
1662 #endif
1663
1664 memcpy(&fs->fs_old_postbl_start, ump->um_oldfscompat, 512);
1665 extrasave = ump->um_oldfscompat;
1666 extrasave += 512/sizeof(int32_t);
1667 fs->fs_old_npsect = extrasave[0];
1668 fs->fs_old_interleave = extrasave[1];
1669 fs->fs_old_trackskew = extrasave[2];
1670
1671 }
1672
1673 /*
1674 * unmount vfs operation
1675 */
1676 int
1677 ffs_unmount(struct mount *mp, int mntflags)
1678 {
1679 struct lwp *l = curlwp;
1680 struct ufsmount *ump = VFSTOUFS(mp);
1681 struct fs *fs = ump->um_fs;
1682 int error, flags;
1683 u_int32_t bsize;
1684 #ifdef WAPBL
1685 extern int doforce;
1686 #endif
1687
1688 if (ump->um_discarddata) {
1689 ffs_discard_finish(ump->um_discarddata, mntflags);
1690 ump->um_discarddata = NULL;
1691 }
1692
1693 flags = 0;
1694 if (mntflags & MNT_FORCE)
1695 flags |= FORCECLOSE;
1696 if ((error = ffs_flushfiles(mp, flags, l)) != 0)
1697 return (error);
1698 error = UFS_WAPBL_BEGIN(mp);
1699 if (error == 0)
1700 if (fs->fs_ronly == 0 &&
1701 ffs_cgupdate(ump, MNT_WAIT) == 0 &&
1702 fs->fs_clean & FS_WASCLEAN) {
1703 fs->fs_clean = FS_ISCLEAN;
1704 fs->fs_fmod = 0;
1705 (void) ffs_sbupdate(ump, MNT_WAIT);
1706 }
1707 if (error == 0)
1708 UFS_WAPBL_END(mp);
1709 #ifdef WAPBL
1710 KASSERT(!(mp->mnt_wapbl_replay && mp->mnt_wapbl));
1711 if (mp->mnt_wapbl_replay) {
1712 KDASSERT(fs->fs_ronly);
1713 wapbl_replay_stop(mp->mnt_wapbl_replay);
1714 wapbl_replay_free(mp->mnt_wapbl_replay);
1715 mp->mnt_wapbl_replay = 0;
1716 }
1717 error = ffs_wapbl_stop(mp, doforce && (mntflags & MNT_FORCE));
1718 if (error) {
1719 return error;
1720 }
1721 #endif /* WAPBL */
1722
1723 if (ump->um_devvp->v_type != VBAD)
1724 spec_node_setmountedfs(ump->um_devvp, NULL);
1725 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
1726 (void)VOP_CLOSE(ump->um_devvp, fs->fs_ronly ? FREAD : FREAD | FWRITE,
1727 NOCRED);
1728 vput(ump->um_devvp);
1729
1730 bsize = fs->fs_cssize;
1731 if (fs->fs_contigsumsize > 0)
1732 bsize += fs->fs_ncg * sizeof(int32_t);
1733 bsize += fs->fs_ncg * sizeof(*fs->fs_contigdirs);
1734 kmem_free(fs->fs_csp, bsize);
1735
1736 kmem_free(fs, fs->fs_sbsize);
1737 if (ump->um_oldfscompat != NULL)
1738 kmem_free(ump->um_oldfscompat, 512 + 3*sizeof(int32_t));
1739 mutex_destroy(&ump->um_lock);
1740 ffs_snapshot_fini(ump);
1741 kmem_free(ump, sizeof(*ump));
1742 mp->mnt_data = NULL;
1743 mp->mnt_flag &= ~MNT_LOCAL;
1744 fstrans_unmount(mp);
1745 return (0);
1746 }
1747
1748 /*
1749 * Flush out all the files in a filesystem.
1750 */
1751 int
1752 ffs_flushfiles(struct mount *mp, int flags, struct lwp *l)
1753 {
1754 extern int doforce;
1755 struct ufsmount *ump;
1756 int error;
1757
1758 if (!doforce)
1759 flags &= ~FORCECLOSE;
1760 ump = VFSTOUFS(mp);
1761 #ifdef QUOTA
1762 if ((error = quota1_umount(mp, flags)) != 0)
1763 return (error);
1764 #endif
1765 #ifdef QUOTA2
1766 if ((error = quota2_umount(mp, flags)) != 0)
1767 return (error);
1768 #endif
1769 #ifdef UFS_EXTATTR
1770 if (ump->um_fstype == UFS1) {
1771 if (ump->um_extattr.uepm_flags & UFS_EXTATTR_UEPM_STARTED)
1772 ufs_extattr_stop(mp, l);
1773 if (ump->um_extattr.uepm_flags & UFS_EXTATTR_UEPM_INITIALIZED)
1774 ufs_extattr_uepm_destroy(&ump->um_extattr);
1775 mp->mnt_flag &= ~MNT_EXTATTR;
1776 }
1777 #endif
1778 if ((error = vflush(mp, 0, SKIPSYSTEM | flags)) != 0)
1779 return (error);
1780 ffs_snapshot_unmount(mp);
1781 /*
1782 * Flush all the files.
1783 */
1784 error = vflush(mp, NULLVP, flags);
1785 if (error)
1786 return (error);
1787 /*
1788 * Flush filesystem metadata.
1789 */
1790 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
1791 error = VOP_FSYNC(ump->um_devvp, l->l_cred, FSYNC_WAIT, 0, 0);
1792 VOP_UNLOCK(ump->um_devvp);
1793 if (flags & FORCECLOSE) /* XXXDBJ */
1794 error = 0;
1795
1796 #ifdef WAPBL
1797 if (error)
1798 return error;
1799 if (mp->mnt_wapbl) {
1800 error = wapbl_flush(mp->mnt_wapbl, 1);
1801 if (flags & FORCECLOSE)
1802 error = 0;
1803 }
1804 #endif
1805
1806 return (error);
1807 }
1808
1809 /*
1810 * Get file system statistics.
1811 */
1812 int
1813 ffs_statvfs(struct mount *mp, struct statvfs *sbp)
1814 {
1815 struct ufsmount *ump;
1816 struct fs *fs;
1817
1818 ump = VFSTOUFS(mp);
1819 fs = ump->um_fs;
1820 mutex_enter(&ump->um_lock);
1821 sbp->f_bsize = fs->fs_bsize;
1822 sbp->f_frsize = fs->fs_fsize;
1823 sbp->f_iosize = fs->fs_bsize;
1824 sbp->f_blocks = fs->fs_dsize;
1825 sbp->f_bfree = ffs_blkstofrags(fs, fs->fs_cstotal.cs_nbfree) +
1826 fs->fs_cstotal.cs_nffree + FFS_DBTOFSB(fs, fs->fs_pendingblocks);
1827 sbp->f_bresvd = ((u_int64_t) fs->fs_dsize * (u_int64_t)
1828 fs->fs_minfree) / (u_int64_t) 100;
1829 if (sbp->f_bfree > sbp->f_bresvd)
1830 sbp->f_bavail = sbp->f_bfree - sbp->f_bresvd;
1831 else
1832 sbp->f_bavail = 0;
1833 sbp->f_files = fs->fs_ncg * fs->fs_ipg - UFS_ROOTINO;
1834 sbp->f_ffree = fs->fs_cstotal.cs_nifree + fs->fs_pendinginodes;
1835 sbp->f_favail = sbp->f_ffree;
1836 sbp->f_fresvd = 0;
1837 mutex_exit(&ump->um_lock);
1838 copy_statvfs_info(sbp, mp);
1839
1840 return (0);
1841 }
1842
1843 struct ffs_sync_ctx {
1844 int waitfor;
1845 bool is_suspending;
1846 };
1847
1848 static bool
1849 ffs_sync_selector(void *cl, struct vnode *vp)
1850 {
1851 struct ffs_sync_ctx *c = cl;
1852 struct inode *ip;
1853
1854 ip = VTOI(vp);
1855 /*
1856 * Skip the vnode/inode if inaccessible.
1857 */
1858 if (ip == NULL || vp->v_type == VNON)
1859 return false;
1860
1861 /*
1862 * We deliberately update inode times here. This will
1863 * prevent a massive queue of updates accumulating, only
1864 * to be handled by a call to unmount.
1865 *
1866 * XXX It would be better to have the syncer trickle these
1867 * out. Adjustment needed to allow registering vnodes for
1868 * sync when the vnode is clean, but the inode dirty. Or
1869 * have ufs itself trickle out inode updates.
1870 *
1871 * If doing a lazy sync, we don't care about metadata or
1872 * data updates, because they are handled by each vnode's
1873 * synclist entry. In this case we are only interested in
1874 * writing back modified inodes.
1875 */
1876 if ((ip->i_flag & (IN_ACCESS | IN_CHANGE | IN_UPDATE |
1877 IN_MODIFY | IN_MODIFIED | IN_ACCESSED)) == 0 &&
1878 (c->waitfor == MNT_LAZY || (LIST_EMPTY(&vp->v_dirtyblkhd) &&
1879 UVM_OBJ_IS_CLEAN(&vp->v_uobj))))
1880 return false;
1881
1882 if (vp->v_type == VBLK && c->is_suspending)
1883 return false;
1884
1885 return true;
1886 }
1887
1888 /*
1889 * Go through the disk queues to initiate sandbagged IO;
1890 * go through the inodes to write those that have been modified;
1891 * initiate the writing of the super block if it has been modified.
1892 *
1893 * Note: we are always called with the filesystem marked `MPBUSY'.
1894 */
1895 int
1896 ffs_sync(struct mount *mp, int waitfor, kauth_cred_t cred)
1897 {
1898 struct vnode *vp;
1899 struct ufsmount *ump = VFSTOUFS(mp);
1900 struct fs *fs;
1901 struct vnode_iterator *marker;
1902 int error, allerror = 0;
1903 bool is_suspending;
1904 struct ffs_sync_ctx ctx;
1905
1906 fs = ump->um_fs;
1907 if (fs->fs_fmod != 0 && fs->fs_ronly != 0) { /* XXX */
1908 panic("%s: rofs mod, fs=%s", __func__, fs->fs_fsmnt);
1909 }
1910
1911 fstrans_start(mp, FSTRANS_SHARED);
1912 is_suspending = (fstrans_getstate(mp) == FSTRANS_SUSPENDING);
1913 /*
1914 * Write back each (modified) inode.
1915 */
1916 vfs_vnode_iterator_init(mp, &marker);
1917
1918 ctx.waitfor = waitfor;
1919 ctx.is_suspending = is_suspending;
1920 while ((vp = vfs_vnode_iterator_next(marker, ffs_sync_selector, &ctx)))
1921 {
1922 error = vn_lock(vp, LK_EXCLUSIVE);
1923 if (error) {
1924 vrele(vp);
1925 continue;
1926 }
1927 if (waitfor == MNT_LAZY) {
1928 error = UFS_WAPBL_BEGIN(vp->v_mount);
1929 if (!error) {
1930 error = ffs_update(vp, NULL, NULL,
1931 UPDATE_CLOSE);
1932 UFS_WAPBL_END(vp->v_mount);
1933 }
1934 } else {
1935 error = VOP_FSYNC(vp, cred, FSYNC_NOLOG |
1936 (waitfor == MNT_WAIT ? FSYNC_WAIT : 0), 0, 0);
1937 }
1938 if (error)
1939 allerror = error;
1940 vput(vp);
1941 }
1942 vfs_vnode_iterator_destroy(marker);
1943
1944 /*
1945 * Force stale file system control information to be flushed.
1946 */
1947 if (waitfor != MNT_LAZY && (ump->um_devvp->v_numoutput > 0 ||
1948 !LIST_EMPTY(&ump->um_devvp->v_dirtyblkhd))) {
1949 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
1950 if ((error = VOP_FSYNC(ump->um_devvp, cred,
1951 (waitfor == MNT_WAIT ? FSYNC_WAIT : 0) | FSYNC_NOLOG,
1952 0, 0)) != 0)
1953 allerror = error;
1954 VOP_UNLOCK(ump->um_devvp);
1955 }
1956 #if defined(QUOTA) || defined(QUOTA2)
1957 qsync(mp);
1958 #endif
1959 /*
1960 * Write back modified superblock.
1961 */
1962 if (fs->fs_fmod != 0) {
1963 fs->fs_fmod = 0;
1964 fs->fs_time = time_second;
1965 error = UFS_WAPBL_BEGIN(mp);
1966 if (error)
1967 allerror = error;
1968 else {
1969 if ((error = ffs_cgupdate(ump, waitfor)))
1970 allerror = error;
1971 UFS_WAPBL_END(mp);
1972 }
1973 }
1974
1975 #ifdef WAPBL
1976 if (mp->mnt_wapbl) {
1977 error = wapbl_flush(mp->mnt_wapbl, 0);
1978 if (error)
1979 allerror = error;
1980 }
1981 #endif
1982
1983 fstrans_done(mp);
1984 return (allerror);
1985 }
1986
1987 /*
1988 * Load inode from disk and initialize vnode.
1989 */
1990 static int
1991 ffs_init_vnode(struct ufsmount *ump, struct vnode *vp, ino_t ino)
1992 {
1993 struct fs *fs;
1994 struct inode *ip;
1995 struct buf *bp;
1996 int error;
1997
1998 fs = ump->um_fs;
1999
2000 /* Read in the disk contents for the inode. */
2001 error = bread(ump->um_devvp, FFS_FSBTODB(fs, ino_to_fsba(fs, ino)),
2002 (int)fs->fs_bsize, 0, &bp);
2003 if (error)
2004 return error;
2005
2006 /* Allocate and initialize inode. */
2007 ip = pool_cache_get(ffs_inode_cache, PR_WAITOK);
2008 memset(ip, 0, sizeof(struct inode));
2009 ip->i_ump = ump;
2010 ip->i_fs = fs;
2011 ip->i_dev = ump->um_dev;
2012 ip->i_number = ino;
2013 if (ump->um_fstype == UFS1)
2014 ip->i_din.ffs1_din = pool_cache_get(ffs_dinode1_cache,
2015 PR_WAITOK);
2016 else
2017 ip->i_din.ffs2_din = pool_cache_get(ffs_dinode2_cache,
2018 PR_WAITOK);
2019 ffs_load_inode(bp, ip, fs, ino);
2020 brelse(bp, 0);
2021 ip->i_vnode = vp;
2022 #if defined(QUOTA) || defined(QUOTA2)
2023 ufsquota_init(ip);
2024 #endif
2025
2026 /* Initialise vnode with this inode. */
2027 vp->v_tag = VT_UFS;
2028 vp->v_op = ffs_vnodeop_p;
2029 vp->v_vflag |= VV_LOCKSWORK;
2030 vp->v_data = ip;
2031
2032 /* Initialize genfs node. */
2033 genfs_node_init(vp, &ffs_genfsops);
2034
2035 return 0;
2036 }
2037
2038 /*
2039 * Undo ffs_init_vnode().
2040 */
2041 static void
2042 ffs_deinit_vnode(struct ufsmount *ump, struct vnode *vp)
2043 {
2044 struct inode *ip = VTOI(vp);
2045
2046 if (ump->um_fstype == UFS1)
2047 pool_cache_put(ffs_dinode1_cache, ip->i_din.ffs1_din);
2048 else
2049 pool_cache_put(ffs_dinode2_cache, ip->i_din.ffs2_din);
2050 pool_cache_put(ffs_inode_cache, ip);
2051
2052 genfs_node_destroy(vp);
2053 vp->v_data = NULL;
2054 }
2055
2056 /*
2057 * Read an inode from disk and initialize this vnode / inode pair.
2058 * Caller assures no other thread will try to load this inode.
2059 */
2060 int
2061 ffs_loadvnode(struct mount *mp, struct vnode *vp,
2062 const void *key, size_t key_len, const void **new_key)
2063 {
2064 ino_t ino;
2065 struct fs *fs;
2066 struct inode *ip;
2067 struct ufsmount *ump;
2068 int error;
2069
2070 KASSERT(key_len == sizeof(ino));
2071 memcpy(&ino, key, key_len);
2072 ump = VFSTOUFS(mp);
2073 fs = ump->um_fs;
2074
2075 error = ffs_init_vnode(ump, vp, ino);
2076 if (error)
2077 return error;
2078
2079 ip = VTOI(vp);
2080 if (ip->i_mode == 0) {
2081 ffs_deinit_vnode(ump, vp);
2082
2083 return ENOENT;
2084 }
2085
2086 /* Initialize the vnode from the inode. */
2087 ufs_vinit(mp, ffs_specop_p, ffs_fifoop_p, &vp);
2088
2089 /* Finish inode initialization. */
2090 ip->i_devvp = ump->um_devvp;
2091 vref(ip->i_devvp);
2092
2093 /*
2094 * Ensure that uid and gid are correct. This is a temporary
2095 * fix until fsck has been changed to do the update.
2096 */
2097
2098 if (fs->fs_old_inodefmt < FS_44INODEFMT) { /* XXX */
2099 ip->i_uid = ip->i_ffs1_ouid; /* XXX */
2100 ip->i_gid = ip->i_ffs1_ogid; /* XXX */
2101 } /* XXX */
2102 uvm_vnp_setsize(vp, ip->i_size);
2103 *new_key = &ip->i_number;
2104 return 0;
2105 }
2106
2107 /*
2108 * Create a new inode on disk and initialize this vnode / inode pair.
2109 */
2110 int
2111 ffs_newvnode(struct mount *mp, struct vnode *dvp, struct vnode *vp,
2112 struct vattr *vap, kauth_cred_t cred,
2113 size_t *key_len, const void **new_key)
2114 {
2115 ino_t ino;
2116 struct fs *fs;
2117 struct inode *ip;
2118 struct timespec ts;
2119 struct ufsmount *ump;
2120 int error, mode;
2121
2122 KASSERT(dvp->v_mount == mp);
2123 KASSERT(vap->va_type != VNON);
2124
2125 *key_len = sizeof(ino);
2126 ump = VFSTOUFS(mp);
2127 fs = ump->um_fs;
2128 mode = MAKEIMODE(vap->va_type, vap->va_mode);
2129
2130 /* Allocate fresh inode. */
2131 error = ffs_valloc(dvp, mode, cred, &ino);
2132 if (error)
2133 return error;
2134
2135 /* Attach inode to vnode. */
2136 error = ffs_init_vnode(ump, vp, ino);
2137 if (error) {
2138 if (UFS_WAPBL_BEGIN(mp) == 0) {
2139 ffs_vfree(dvp, ino, mode);
2140 UFS_WAPBL_END(mp);
2141 }
2142 return error;
2143 }
2144
2145 ip = VTOI(vp);
2146 if (ip->i_mode) {
2147 panic("%s: dup alloc ino=%" PRId64 " on %s: mode %x/%x "
2148 "gen %x/%x size %" PRIx64 " blocks %" PRIx64,
2149 __func__, ino, fs->fs_fsmnt, DIP(ip, mode), ip->i_mode,
2150 DIP(ip, gen), ip->i_gen, DIP(ip, size), DIP(ip, blocks));
2151 }
2152 if (DIP(ip, size) || DIP(ip, blocks)) {
2153 printf("%s: ino=%" PRId64 " on %s: "
2154 "gen %x/%x has non zero blocks %" PRIx64 " or size %"
2155 PRIx64 "\n",
2156 __func__, ino, fs->fs_fsmnt, DIP(ip, gen), ip->i_gen,
2157 DIP(ip, blocks), DIP(ip, size));
2158 if ((ip)->i_ump->um_fstype == UFS1)
2159 panic("%s: dirty filesystem?", __func__);
2160 DIP_ASSIGN(ip, blocks, 0);
2161 DIP_ASSIGN(ip, size, 0);
2162 }
2163
2164 /* Set uid / gid. */
2165 if (cred == NOCRED || cred == FSCRED) {
2166 ip->i_gid = 0;
2167 ip->i_uid = 0;
2168 } else {
2169 ip->i_gid = VTOI(dvp)->i_gid;
2170 ip->i_uid = kauth_cred_geteuid(cred);
2171 }
2172 DIP_ASSIGN(ip, gid, ip->i_gid);
2173 DIP_ASSIGN(ip, uid, ip->i_uid);
2174
2175 #if defined(QUOTA) || defined(QUOTA2)
2176 error = UFS_WAPBL_BEGIN(mp);
2177 if (error) {
2178 ffs_deinit_vnode(ump, vp);
2179
2180 return error;
2181 }
2182 error = chkiq(ip, 1, cred, 0);
2183 if (error) {
2184 ffs_vfree(dvp, ino, mode);
2185 UFS_WAPBL_END(mp);
2186 ffs_deinit_vnode(ump, vp);
2187
2188 return error;
2189 }
2190 UFS_WAPBL_END(mp);
2191 #endif
2192
2193 /* Set type and finalize. */
2194 ip->i_flags = 0;
2195 DIP_ASSIGN(ip, flags, 0);
2196 ip->i_mode = mode;
2197 DIP_ASSIGN(ip, mode, mode);
2198 if (vap->va_rdev != VNOVAL) {
2199 /*
2200 * Want to be able to use this to make badblock
2201 * inodes, so don't truncate the dev number.
2202 */
2203 if (ump->um_fstype == UFS1)
2204 ip->i_ffs1_rdev = ufs_rw32(vap->va_rdev,
2205 UFS_MPNEEDSWAP(ump));
2206 else
2207 ip->i_ffs2_rdev = ufs_rw64(vap->va_rdev,
2208 UFS_MPNEEDSWAP(ump));
2209 }
2210 ufs_vinit(mp, ffs_specop_p, ffs_fifoop_p, &vp);
2211 ip->i_devvp = ump->um_devvp;
2212 vref(ip->i_devvp);
2213
2214 /* Set up a new generation number for this inode. */
2215 ip->i_gen++;
2216 DIP_ASSIGN(ip, gen, ip->i_gen);
2217 if (fs->fs_magic == FS_UFS2_MAGIC) {
2218 vfs_timestamp(&ts);
2219 ip->i_ffs2_birthtime = ts.tv_sec;
2220 ip->i_ffs2_birthnsec = ts.tv_nsec;
2221 }
2222
2223 uvm_vnp_setsize(vp, ip->i_size);
2224 *new_key = &ip->i_number;
2225 return 0;
2226 }
2227
2228 /*
2229 * File handle to vnode
2230 *
2231 * Have to be really careful about stale file handles:
2232 * - check that the inode number is valid
2233 * - call ffs_vget() to get the locked inode
2234 * - check for an unallocated inode (i_mode == 0)
2235 * - check that the given client host has export rights and return
2236 * those rights via. exflagsp and credanonp
2237 */
2238 int
2239 ffs_fhtovp(struct mount *mp, struct fid *fhp, struct vnode **vpp)
2240 {
2241 struct ufid ufh;
2242 int error;
2243
2244 if (fhp->fid_len != sizeof(struct ufid))
2245 return EINVAL;
2246
2247 memcpy(&ufh, fhp, sizeof(ufh));
2248 if ((error = ffs_checkrange(mp, ufh.ufid_ino)) != 0)
2249 return error;
2250
2251 return (ufs_fhtovp(mp, &ufh, vpp));
2252 }
2253
2254 /*
2255 * Vnode pointer to File handle
2256 */
2257 /* ARGSUSED */
2258 int
2259 ffs_vptofh(struct vnode *vp, struct fid *fhp, size_t *fh_size)
2260 {
2261 struct inode *ip;
2262 struct ufid ufh;
2263
2264 if (*fh_size < sizeof(struct ufid)) {
2265 *fh_size = sizeof(struct ufid);
2266 return E2BIG;
2267 }
2268 ip = VTOI(vp);
2269 *fh_size = sizeof(struct ufid);
2270 memset(&ufh, 0, sizeof(ufh));
2271 ufh.ufid_len = sizeof(struct ufid);
2272 ufh.ufid_ino = ip->i_number;
2273 ufh.ufid_gen = ip->i_gen;
2274 memcpy(fhp, &ufh, sizeof(ufh));
2275 return (0);
2276 }
2277
2278 void
2279 ffs_init(void)
2280 {
2281 if (ffs_initcount++ > 0)
2282 return;
2283
2284 ffs_inode_cache = pool_cache_init(sizeof(struct inode), 0, 0, 0,
2285 "ffsino", NULL, IPL_NONE, NULL, NULL, NULL);
2286 ffs_dinode1_cache = pool_cache_init(sizeof(struct ufs1_dinode), 0, 0, 0,
2287 "ffsdino1", NULL, IPL_NONE, NULL, NULL, NULL);
2288 ffs_dinode2_cache = pool_cache_init(sizeof(struct ufs2_dinode), 0, 0, 0,
2289 "ffsdino2", NULL, IPL_NONE, NULL, NULL, NULL);
2290 ufs_init();
2291 }
2292
2293 void
2294 ffs_reinit(void)
2295 {
2296 ufs_reinit();
2297 }
2298
2299 void
2300 ffs_done(void)
2301 {
2302 if (--ffs_initcount > 0)
2303 return;
2304
2305 ufs_done();
2306 pool_cache_destroy(ffs_dinode2_cache);
2307 pool_cache_destroy(ffs_dinode1_cache);
2308 pool_cache_destroy(ffs_inode_cache);
2309 }
2310
2311 /*
2312 * Write a superblock and associated information back to disk.
2313 */
2314 int
2315 ffs_sbupdate(struct ufsmount *mp, int waitfor)
2316 {
2317 struct fs *fs = mp->um_fs;
2318 struct buf *bp;
2319 int error;
2320 u_int32_t saveflag;
2321
2322 error = ffs_getblk(mp->um_devvp,
2323 fs->fs_sblockloc / DEV_BSIZE, FFS_NOBLK,
2324 fs->fs_sbsize, false, &bp);
2325 if (error)
2326 return error;
2327 saveflag = fs->fs_flags & FS_INTERNAL;
2328 fs->fs_flags &= ~FS_INTERNAL;
2329
2330 memcpy(bp->b_data, fs, fs->fs_sbsize);
2331
2332 ffs_oldfscompat_write((struct fs *)bp->b_data, mp);
2333 #ifdef FFS_EI
2334 if (mp->um_flags & UFS_NEEDSWAP)
2335 ffs_sb_swap((struct fs *)bp->b_data, (struct fs *)bp->b_data);
2336 #endif
2337 fs->fs_flags |= saveflag;
2338
2339 if (waitfor == MNT_WAIT)
2340 error = bwrite(bp);
2341 else
2342 bawrite(bp);
2343 return (error);
2344 }
2345
2346 int
2347 ffs_cgupdate(struct ufsmount *mp, int waitfor)
2348 {
2349 struct fs *fs = mp->um_fs;
2350 struct buf *bp;
2351 int blks;
2352 void *space;
2353 int i, size, error = 0, allerror = 0;
2354
2355 allerror = ffs_sbupdate(mp, waitfor);
2356 blks = howmany(fs->fs_cssize, fs->fs_fsize);
2357 space = fs->fs_csp;
2358 for (i = 0; i < blks; i += fs->fs_frag) {
2359 size = fs->fs_bsize;
2360 if (i + fs->fs_frag > blks)
2361 size = (blks - i) * fs->fs_fsize;
2362 error = ffs_getblk(mp->um_devvp, FFS_FSBTODB(fs, fs->fs_csaddr + i),
2363 FFS_NOBLK, size, false, &bp);
2364 if (error)
2365 break;
2366 #ifdef FFS_EI
2367 if (mp->um_flags & UFS_NEEDSWAP)
2368 ffs_csum_swap((struct csum*)space,
2369 (struct csum*)bp->b_data, size);
2370 else
2371 #endif
2372 memcpy(bp->b_data, space, (u_int)size);
2373 space = (char *)space + size;
2374 if (waitfor == MNT_WAIT)
2375 error = bwrite(bp);
2376 else
2377 bawrite(bp);
2378 }
2379 if (!allerror && error)
2380 allerror = error;
2381 return (allerror);
2382 }
2383
2384 int
2385 ffs_extattrctl(struct mount *mp, int cmd, struct vnode *vp,
2386 int attrnamespace, const char *attrname)
2387 {
2388 #ifdef UFS_EXTATTR
2389 /*
2390 * File-backed extended attributes are only supported on UFS1.
2391 * UFS2 has native extended attributes.
2392 */
2393 if (VFSTOUFS(mp)->um_fstype == UFS1)
2394 return (ufs_extattrctl(mp, cmd, vp, attrnamespace, attrname));
2395 #endif
2396 return (vfs_stdextattrctl(mp, cmd, vp, attrnamespace, attrname));
2397 }
2398
2399 int
2400 ffs_suspendctl(struct mount *mp, int cmd)
2401 {
2402 int error;
2403 struct lwp *l = curlwp;
2404
2405 switch (cmd) {
2406 case SUSPEND_SUSPEND:
2407 if ((error = fstrans_setstate(mp, FSTRANS_SUSPENDING)) != 0)
2408 return error;
2409 error = ffs_sync(mp, MNT_WAIT, l->l_proc->p_cred);
2410 if (error == 0)
2411 error = fstrans_setstate(mp, FSTRANS_SUSPENDED);
2412 #ifdef WAPBL
2413 if (error == 0 && mp->mnt_wapbl)
2414 error = wapbl_flush(mp->mnt_wapbl, 1);
2415 #endif
2416 if (error != 0) {
2417 (void) fstrans_setstate(mp, FSTRANS_NORMAL);
2418 return error;
2419 }
2420 return 0;
2421
2422 case SUSPEND_RESUME:
2423 return fstrans_setstate(mp, FSTRANS_NORMAL);
2424
2425 default:
2426 return EINVAL;
2427 }
2428 }
2429
2430 /*
2431 * Synch vnode for a mounted file system.
2432 */
2433 static int
2434 ffs_vfs_fsync(vnode_t *vp, int flags)
2435 {
2436 int error, i, pflags;
2437 #ifdef WAPBL
2438 struct mount *mp;
2439 #endif
2440
2441 KASSERT(vp->v_type == VBLK);
2442 KASSERT(spec_node_getmountedfs(vp) != NULL);
2443
2444 /*
2445 * Flush all dirty data associated with the vnode.
2446 */
2447 pflags = PGO_ALLPAGES | PGO_CLEANIT;
2448 if ((flags & FSYNC_WAIT) != 0)
2449 pflags |= PGO_SYNCIO;
2450 mutex_enter(vp->v_interlock);
2451 error = VOP_PUTPAGES(vp, 0, 0, pflags);
2452 if (error)
2453 return error;
2454
2455 #ifdef WAPBL
2456 mp = spec_node_getmountedfs(vp);
2457 if (mp && mp->mnt_wapbl) {
2458 /*
2459 * Don't bother writing out metadata if the syncer is
2460 * making the request. We will let the sync vnode
2461 * write it out in a single burst through a call to
2462 * VFS_SYNC().
2463 */
2464 if ((flags & (FSYNC_DATAONLY | FSYNC_LAZY | FSYNC_NOLOG)) != 0)
2465 return 0;
2466
2467 /*
2468 * Don't flush the log if the vnode being flushed
2469 * contains no dirty buffers that could be in the log.
2470 */
2471 if (!LIST_EMPTY(&vp->v_dirtyblkhd)) {
2472 error = wapbl_flush(mp->mnt_wapbl, 0);
2473 if (error)
2474 return error;
2475 }
2476
2477 if ((flags & FSYNC_WAIT) != 0) {
2478 mutex_enter(vp->v_interlock);
2479 while (vp->v_numoutput)
2480 cv_wait(&vp->v_cv, vp->v_interlock);
2481 mutex_exit(vp->v_interlock);
2482 }
2483
2484 return 0;
2485 }
2486 #endif /* WAPBL */
2487
2488 error = vflushbuf(vp, flags);
2489 if (error == 0 && (flags & FSYNC_CACHE) != 0) {
2490 i = 1;
2491 (void)VOP_IOCTL(vp, DIOCCACHESYNC, &i, FWRITE,
2492 kauth_cred_get());
2493 }
2494
2495 return error;
2496 }
2497