ffs_vfsops.c revision 1.339.2.6 1 /* $NetBSD: ffs_vfsops.c,v 1.339.2.6 2017/03/20 06:57:54 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 1.339.2.6 2017/03/20 06:57:54 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/module.h>
96
97 #include <miscfs/genfs/genfs.h>
98 #include <miscfs/specfs/specdev.h>
99
100 #include <ufs/ufs/quota.h>
101 #include <ufs/ufs/ufsmount.h>
102 #include <ufs/ufs/inode.h>
103 #include <ufs/ufs/dir.h>
104 #include <ufs/ufs/ufs_extern.h>
105 #include <ufs/ufs/ufs_bswap.h>
106 #include <ufs/ufs/ufs_wapbl.h>
107
108 #include <ufs/ffs/fs.h>
109 #include <ufs/ffs/ffs_extern.h>
110
111 #ifdef WAPBL
112 MODULE(MODULE_CLASS_VFS, ffs, "wapbl");
113 #else
114 MODULE(MODULE_CLASS_VFS, ffs, NULL);
115 #endif
116
117 static int ffs_vfs_fsync(vnode_t *, int);
118 static int ffs_superblock_validate(struct fs *);
119 static int ffs_is_appleufs(struct vnode *, struct fs *);
120
121 static int ffs_init_vnode(struct ufsmount *, struct vnode *, ino_t);
122 static void ffs_deinit_vnode(struct ufsmount *, struct vnode *);
123
124 static struct sysctllog *ffs_sysctl_log;
125
126 static kauth_listener_t ffs_snapshot_listener;
127
128 /* how many times ffs_init() was called */
129 int ffs_initcount = 0;
130
131 #ifdef DEBUG_FFS_MOUNT
132 #define DPRINTF(_fmt, args...) printf("%s: " _fmt "\n", __func__, ##args)
133 #else
134 #define DPRINTF(_fmt, args...) do {} while (/*CONSTCOND*/0)
135 #endif
136
137 extern const struct vnodeopv_desc ffs_vnodeop_opv_desc;
138 extern const struct vnodeopv_desc ffs_specop_opv_desc;
139 extern const struct vnodeopv_desc ffs_fifoop_opv_desc;
140
141 const struct vnodeopv_desc * const ffs_vnodeopv_descs[] = {
142 &ffs_vnodeop_opv_desc,
143 &ffs_specop_opv_desc,
144 &ffs_fifoop_opv_desc,
145 NULL,
146 };
147
148 struct vfsops ffs_vfsops = {
149 .vfs_name = MOUNT_FFS,
150 .vfs_min_mount_data = sizeof (struct ufs_args),
151 .vfs_mount = ffs_mount,
152 .vfs_start = ufs_start,
153 .vfs_unmount = ffs_unmount,
154 .vfs_root = ufs_root,
155 .vfs_quotactl = ufs_quotactl,
156 .vfs_statvfs = ffs_statvfs,
157 .vfs_sync = ffs_sync,
158 .vfs_vget = ufs_vget,
159 .vfs_loadvnode = ffs_loadvnode,
160 .vfs_newvnode = ffs_newvnode,
161 .vfs_fhtovp = ffs_fhtovp,
162 .vfs_vptofh = ffs_vptofh,
163 .vfs_init = ffs_init,
164 .vfs_reinit = ffs_reinit,
165 .vfs_done = ffs_done,
166 .vfs_mountroot = ffs_mountroot,
167 .vfs_snapshot = ffs_snapshot,
168 .vfs_extattrctl = ffs_extattrctl,
169 .vfs_suspendctl = genfs_suspendctl,
170 .vfs_renamelock_enter = genfs_renamelock_enter,
171 .vfs_renamelock_exit = genfs_renamelock_exit,
172 .vfs_fsync = ffs_vfs_fsync,
173 .vfs_opv_descs = ffs_vnodeopv_descs
174 };
175
176 static const struct genfs_ops ffs_genfsops = {
177 .gop_size = ffs_gop_size,
178 .gop_alloc = ufs_gop_alloc,
179 .gop_write = genfs_gop_write,
180 .gop_markupdate = ufs_gop_markupdate,
181 };
182
183 static const struct ufs_ops ffs_ufsops = {
184 .uo_itimes = ffs_itimes,
185 .uo_update = ffs_update,
186 .uo_truncate = ffs_truncate,
187 .uo_balloc = ffs_balloc,
188 .uo_snapgone = ffs_snapgone,
189 .uo_bufrd = ffs_bufrd,
190 .uo_bufwr = ffs_bufwr,
191 };
192
193 static int
194 ffs_checkrange(struct mount *mp, uint32_t ino)
195 {
196 struct fs *fs = VFSTOUFS(mp)->um_fs;
197
198 if (ino < UFS_ROOTINO || ino >= fs->fs_ncg * fs->fs_ipg) {
199 DPRINTF("out of range %u\n", ino);
200 return ESTALE;
201 }
202
203 /*
204 * Need to check if inode is initialized because ffsv2 does
205 * lazy initialization and we can get here from nfs_fhtovp
206 */
207 if (fs->fs_magic != FS_UFS2_MAGIC)
208 return 0;
209
210 struct buf *bp;
211 int cg = ino_to_cg(fs, ino);
212 struct ufsmount *ump = VFSTOUFS(mp);
213
214 int error = bread(ump->um_devvp, FFS_FSBTODB(fs, cgtod(fs, cg)),
215 (int)fs->fs_cgsize, B_MODIFY, &bp);
216 if (error) {
217 DPRINTF("error %d reading cg %d ino %u\n", error, cg, ino);
218 return error;
219 }
220
221 const int needswap = UFS_FSNEEDSWAP(fs);
222
223 struct cg *cgp = (struct cg *)bp->b_data;
224 if (!cg_chkmagic(cgp, needswap)) {
225 brelse(bp, 0);
226 DPRINTF("bad cylinder group magic cg %d ino %u\n", cg, ino);
227 return ESTALE;
228 }
229
230 int32_t initediblk = ufs_rw32(cgp->cg_initediblk, needswap);
231 brelse(bp, 0);
232
233 if (cg * fs->fs_ipg + initediblk < ino) {
234 DPRINTF("cg=%d fs->fs_ipg=%d initediblk=%d ino=%u\n",
235 cg, fs->fs_ipg, initediblk, ino);
236 return ESTALE;
237 }
238 return 0;
239 }
240
241 static int
242 ffs_snapshot_cb(kauth_cred_t cred, kauth_action_t action, void *cookie,
243 void *arg0, void *arg1, void *arg2, void *arg3)
244 {
245 vnode_t *vp = arg2;
246 int result = KAUTH_RESULT_DEFER;
247
248 if (action != KAUTH_SYSTEM_FS_SNAPSHOT)
249 return result;
250
251 if (VTOI(vp)->i_uid == kauth_cred_geteuid(cred))
252 result = KAUTH_RESULT_ALLOW;
253
254 return result;
255 }
256
257 static int
258 ffs_modcmd(modcmd_t cmd, void *arg)
259 {
260 int error;
261
262 #if 0
263 extern int doasyncfree;
264 #endif
265 #ifdef UFS_EXTATTR
266 extern int ufs_extattr_autocreate;
267 #endif
268 extern int ffs_log_changeopt;
269
270 switch (cmd) {
271 case MODULE_CMD_INIT:
272 error = vfs_attach(&ffs_vfsops);
273 if (error != 0)
274 break;
275
276 sysctl_createv(&ffs_sysctl_log, 0, NULL, NULL,
277 CTLFLAG_PERMANENT,
278 CTLTYPE_NODE, "ffs",
279 SYSCTL_DESCR("Berkeley Fast File System"),
280 NULL, 0, NULL, 0,
281 CTL_VFS, 1, CTL_EOL);
282 /*
283 * @@@ should we even bother with these first three?
284 */
285 sysctl_createv(&ffs_sysctl_log, 0, NULL, NULL,
286 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
287 CTLTYPE_INT, "doclusterread", NULL,
288 sysctl_notavail, 0, NULL, 0,
289 CTL_VFS, 1, FFS_CLUSTERREAD, CTL_EOL);
290 sysctl_createv(&ffs_sysctl_log, 0, NULL, NULL,
291 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
292 CTLTYPE_INT, "doclusterwrite", NULL,
293 sysctl_notavail, 0, NULL, 0,
294 CTL_VFS, 1, FFS_CLUSTERWRITE, CTL_EOL);
295 sysctl_createv(&ffs_sysctl_log, 0, NULL, NULL,
296 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
297 CTLTYPE_INT, "doreallocblks", NULL,
298 sysctl_notavail, 0, NULL, 0,
299 CTL_VFS, 1, FFS_REALLOCBLKS, CTL_EOL);
300 #if 0
301 sysctl_createv(&ffs_sysctl_log, 0, NULL, NULL,
302 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
303 CTLTYPE_INT, "doasyncfree",
304 SYSCTL_DESCR("Release dirty blocks asynchronously"),
305 NULL, 0, &doasyncfree, 0,
306 CTL_VFS, 1, FFS_ASYNCFREE, CTL_EOL);
307 #endif
308 sysctl_createv(&ffs_sysctl_log, 0, NULL, NULL,
309 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
310 CTLTYPE_INT, "log_changeopt",
311 SYSCTL_DESCR("Log changes in optimization strategy"),
312 NULL, 0, &ffs_log_changeopt, 0,
313 CTL_VFS, 1, FFS_LOG_CHANGEOPT, CTL_EOL);
314 #ifdef UFS_EXTATTR
315 sysctl_createv(&ffs_sysctl_log, 0, NULL, NULL,
316 CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
317 CTLTYPE_INT, "extattr_autocreate",
318 SYSCTL_DESCR("Size of attribute for "
319 "backing file autocreation"),
320 NULL, 0, &ufs_extattr_autocreate, 0,
321 CTL_VFS, 1, FFS_EXTATTR_AUTOCREATE, CTL_EOL);
322
323 #endif /* UFS_EXTATTR */
324
325 ffs_snapshot_listener = kauth_listen_scope(KAUTH_SCOPE_SYSTEM,
326 ffs_snapshot_cb, NULL);
327 if (ffs_snapshot_listener == NULL)
328 printf("ffs_modcmd: can't listen on system scope.\n");
329
330 break;
331 case MODULE_CMD_FINI:
332 error = vfs_detach(&ffs_vfsops);
333 if (error != 0)
334 break;
335 sysctl_teardown(&ffs_sysctl_log);
336 if (ffs_snapshot_listener != NULL)
337 kauth_unlisten_scope(ffs_snapshot_listener);
338 break;
339 default:
340 error = ENOTTY;
341 break;
342 }
343
344 return (error);
345 }
346
347 pool_cache_t ffs_inode_cache;
348 pool_cache_t ffs_dinode1_cache;
349 pool_cache_t ffs_dinode2_cache;
350
351 static void ffs_oldfscompat_read(struct fs *, struct ufsmount *, daddr_t);
352 static void ffs_oldfscompat_write(struct fs *, struct ufsmount *);
353
354 /*
355 * Called by main() when ffs is going to be mounted as root.
356 */
357
358 int
359 ffs_mountroot(void)
360 {
361 struct fs *fs;
362 struct mount *mp;
363 struct lwp *l = curlwp; /* XXX */
364 struct ufsmount *ump;
365 int error;
366
367 if (device_class(root_device) != DV_DISK)
368 return (ENODEV);
369
370 if ((error = vfs_rootmountalloc(MOUNT_FFS, "root_device", &mp))) {
371 vrele(rootvp);
372 return (error);
373 }
374
375 /*
376 * We always need to be able to mount the root file system.
377 */
378 mp->mnt_flag |= MNT_FORCE;
379 if ((error = ffs_mountfs(rootvp, mp, l)) != 0) {
380 vfs_unbusy(mp, false, NULL);
381 vfs_destroy(mp);
382 return (error);
383 }
384 mp->mnt_flag &= ~MNT_FORCE;
385 mountlist_append(mp);
386 ump = VFSTOUFS(mp);
387 fs = ump->um_fs;
388 memset(fs->fs_fsmnt, 0, sizeof(fs->fs_fsmnt));
389 (void)copystr(mp->mnt_stat.f_mntonname, fs->fs_fsmnt, MNAMELEN - 1, 0);
390 (void)ffs_statvfs(mp, &mp->mnt_stat);
391 vfs_unbusy(mp, false, NULL);
392 setrootfstime((time_t)fs->fs_time);
393 return (0);
394 }
395
396 /*
397 * VFS Operations.
398 *
399 * mount system call
400 */
401 int
402 ffs_mount(struct mount *mp, const char *path, void *data, size_t *data_len)
403 {
404 struct lwp *l = curlwp;
405 struct vnode *devvp = NULL;
406 struct ufs_args *args = data;
407 struct ufsmount *ump = NULL;
408 struct fs *fs;
409 int error = 0, flags, update;
410 mode_t accessmode;
411 const struct bdevsw *bdev = NULL;
412
413 if (args == NULL) {
414 DPRINTF("NULL args");
415 return EINVAL;
416 }
417 if (*data_len < sizeof(*args)) {
418 DPRINTF("bad size args %zu != %zu", *data_len, sizeof(*args));
419 return EINVAL;
420 }
421
422 if (mp->mnt_flag & MNT_GETARGS) {
423 ump = VFSTOUFS(mp);
424 if (ump == NULL) {
425 DPRINTF("no ump");
426 return EIO;
427 }
428 args->fspec = NULL;
429 *data_len = sizeof *args;
430 return 0;
431 }
432
433 update = mp->mnt_flag & MNT_UPDATE;
434
435 /* Check arguments */
436 if (args->fspec != NULL) {
437 /*
438 * Look up the name and verify that it's sane.
439 */
440 error = namei_simple_user(args->fspec,
441 NSM_FOLLOW_NOEMULROOT, &devvp);
442 if (error != 0) {
443 DPRINTF("namei_simple_user returned %d", error);
444 return error;
445 }
446
447 if (!update) {
448 /*
449 * Be sure this is a valid block device
450 */
451 if (devvp->v_type != VBLK) {
452 DPRINTF("non block device %d", devvp->v_type);
453 error = ENOTBLK;
454 } else {
455 bdev = bdevsw_lookup_acquire(devvp->v_rdev);
456 if (bdev == NULL) {
457 DPRINTF("can't find block device 0x%jx",
458 devvp->v_rdev);
459 error = ENXIO;
460 }
461 }
462 } else {
463 /*
464 * Be sure we're still naming the same device
465 * used for our initial mount
466 */
467 ump = VFSTOUFS(mp);
468 if (devvp != ump->um_devvp) {
469 if (devvp->v_rdev != ump->um_devvp->v_rdev) {
470 DPRINTF("wrong device 0x%jx != 0x%jx",
471 (uintmax_t)devvp->v_rdev,
472 (uintmax_t)ump->um_devvp->v_rdev);
473 error = EINVAL;
474 } else {
475 vrele(devvp);
476 devvp = ump->um_devvp;
477 vref(devvp);
478 }
479 }
480 }
481 } else {
482 if (!update) {
483 /* New mounts must have a filename for the device */
484 DPRINTF("no filename for mount");
485 return EINVAL;
486 } else {
487 /* Use the extant mount */
488 ump = VFSTOUFS(mp);
489 devvp = ump->um_devvp;
490 vref(devvp);
491 }
492 }
493
494 /*
495 * If mount by non-root, then verify that user has necessary
496 * permissions on the device.
497 *
498 * Permission to update a mount is checked higher, so here we presume
499 * updating the mount is okay (for example, as far as securelevel goes)
500 * which leaves us with the normal check.
501 */
502 if (error == 0) {
503 accessmode = VREAD;
504 if (update ?
505 (mp->mnt_iflag & IMNT_WANTRDWR) != 0 :
506 (mp->mnt_flag & MNT_RDONLY) == 0)
507 accessmode |= VWRITE;
508 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
509 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT,
510 KAUTH_REQ_SYSTEM_MOUNT_DEVICE, mp, devvp,
511 KAUTH_ARG(accessmode));
512 if (error) {
513 DPRINTF("kauth returned %d", error);
514 }
515 VOP_UNLOCK(devvp);
516 }
517
518 if (error)
519 goto fail;
520
521 #ifdef WAPBL
522 /* WAPBL can only be enabled on a r/w mount. */
523 if (((mp->mnt_flag & MNT_RDONLY) && !(mp->mnt_iflag & IMNT_WANTRDWR)) ||
524 (mp->mnt_iflag & IMNT_WANTRDONLY)) {
525 mp->mnt_flag &= ~MNT_LOG;
526 }
527 #else /* !WAPBL */
528 mp->mnt_flag &= ~MNT_LOG;
529 #endif /* !WAPBL */
530
531 if (!update) {
532 int xflags;
533
534 if (mp->mnt_flag & MNT_RDONLY)
535 xflags = FREAD;
536 else
537 xflags = FREAD | FWRITE;
538 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
539 error = VOP_OPEN(devvp, xflags, FSCRED);
540 VOP_UNLOCK(devvp);
541 if (error) {
542 DPRINTF("VOP_OPEN returned %d", error);
543 goto fail;
544 }
545 error = ffs_mountfs(devvp, mp, l);
546 if (error) {
547 DPRINTF("ffs_mountfs returned %d", error);
548 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
549 (void)VOP_CLOSE(devvp, xflags, NOCRED);
550 VOP_UNLOCK(devvp);
551 goto fail;
552 }
553
554 ump = VFSTOUFS(mp);
555 fs = ump->um_fs;
556 } else {
557 /*
558 * Update the mount.
559 */
560
561 /*
562 * The initial mount got a reference on this
563 * device, so drop the one obtained via
564 * namei(), above.
565 */
566 vrele(devvp);
567
568 ump = VFSTOUFS(mp);
569 fs = ump->um_fs;
570 if (fs->fs_ronly == 0 && (mp->mnt_iflag & IMNT_WANTRDONLY)) {
571 /*
572 * Changing from r/w to r/o
573 */
574 flags = WRITECLOSE;
575 if (mp->mnt_flag & MNT_FORCE)
576 flags |= FORCECLOSE;
577 error = ffs_flushfiles(mp, flags, l);
578 if (error)
579 return error;
580
581 error = UFS_WAPBL_BEGIN(mp);
582 if (error) {
583 DPRINTF("wapbl %d", error);
584 return error;
585 }
586
587 if (ffs_cgupdate(ump, MNT_WAIT) == 0 &&
588 fs->fs_clean & FS_WASCLEAN) {
589 if (mp->mnt_flag & MNT_SOFTDEP)
590 fs->fs_flags &= ~FS_DOSOFTDEP;
591 fs->fs_clean = FS_ISCLEAN;
592 (void) ffs_sbupdate(ump, MNT_WAIT);
593 }
594
595 UFS_WAPBL_END(mp);
596 }
597
598 #ifdef WAPBL
599 if ((mp->mnt_flag & MNT_LOG) == 0) {
600 error = ffs_wapbl_stop(mp, mp->mnt_flag & MNT_FORCE);
601 if (error) {
602 DPRINTF("ffs_wapbl_stop returned %d", error);
603 return error;
604 }
605 }
606 #endif /* WAPBL */
607
608 if (fs->fs_ronly == 0 && (mp->mnt_iflag & IMNT_WANTRDONLY)) {
609 /*
610 * Finish change from r/w to r/o
611 */
612 fs->fs_ronly = 1;
613 fs->fs_fmod = 0;
614 }
615
616 if (mp->mnt_flag & MNT_RELOAD) {
617 error = ffs_reload(mp, l->l_cred, l);
618 if (error) {
619 DPRINTF("ffs_reload returned %d", error);
620 return error;
621 }
622 }
623
624 if (fs->fs_ronly && (mp->mnt_iflag & IMNT_WANTRDWR)) {
625 /*
626 * Changing from read-only to read/write
627 */
628 #ifndef QUOTA2
629 if (fs->fs_flags & FS_DOQUOTA2) {
630 ump->um_flags |= UFS_QUOTA2;
631 uprintf("%s: options QUOTA2 not enabled%s\n",
632 mp->mnt_stat.f_mntonname,
633 (mp->mnt_flag & MNT_FORCE) ? "" :
634 ", not mounting");
635 DPRINTF("ffs_quota2 %d", EINVAL);
636 return EINVAL;
637 }
638 #endif
639 fs->fs_ronly = 0;
640 fs->fs_clean <<= 1;
641 fs->fs_fmod = 1;
642 #ifdef WAPBL
643 if (fs->fs_flags & FS_DOWAPBL) {
644 const char *nm = mp->mnt_stat.f_mntonname;
645 if (!mp->mnt_wapbl_replay) {
646 printf("%s: log corrupted;"
647 " replay cancelled\n", nm);
648 return EFTYPE;
649 }
650 printf("%s: replaying log to disk\n", nm);
651 error = wapbl_replay_write(mp->mnt_wapbl_replay,
652 devvp);
653 if (error) {
654 DPRINTF("%s: wapbl_replay_write %d",
655 nm, error);
656 return error;
657 }
658 wapbl_replay_stop(mp->mnt_wapbl_replay);
659 fs->fs_clean = FS_WASCLEAN;
660 }
661 #endif /* WAPBL */
662 if (fs->fs_snapinum[0] != 0)
663 ffs_snapshot_mount(mp);
664 }
665
666 #ifdef WAPBL
667 error = ffs_wapbl_start(mp);
668 if (error) {
669 DPRINTF("ffs_wapbl_start returned %d", error);
670 return error;
671 }
672 #endif /* WAPBL */
673
674 #ifdef QUOTA2
675 if (!fs->fs_ronly) {
676 error = ffs_quota2_mount(mp);
677 if (error) {
678 DPRINTF("ffs_quota2_mount returned %d", error);
679 return error;
680 }
681 }
682 #endif
683
684 if ((mp->mnt_flag & MNT_DISCARD) && !(ump->um_discarddata))
685 ump->um_discarddata = ffs_discard_init(devvp, fs);
686
687 if (args->fspec == NULL)
688 return 0;
689 }
690
691 error = set_statvfs_info(path, UIO_USERSPACE, args->fspec,
692 UIO_USERSPACE, mp->mnt_op->vfs_name, mp, l);
693 if (error == 0)
694 (void)strncpy(fs->fs_fsmnt, mp->mnt_stat.f_mntonname,
695 sizeof(fs->fs_fsmnt));
696 else {
697 DPRINTF("set_statvfs_info returned %d", error);
698 }
699 fs->fs_flags &= ~FS_DOSOFTDEP;
700 if (fs->fs_fmod != 0) { /* XXX */
701 int err;
702
703 fs->fs_fmod = 0;
704 if (fs->fs_clean & FS_WASCLEAN)
705 fs->fs_time = time_second;
706 else {
707 printf("%s: file system not clean (fs_clean=%#x); "
708 "please fsck(8)\n", mp->mnt_stat.f_mntfromname,
709 fs->fs_clean);
710 printf("%s: lost blocks %" PRId64 " files %d\n",
711 mp->mnt_stat.f_mntfromname, fs->fs_pendingblocks,
712 fs->fs_pendinginodes);
713 }
714 err = UFS_WAPBL_BEGIN(mp);
715 if (err == 0) {
716 (void) ffs_cgupdate(ump, MNT_WAIT);
717 UFS_WAPBL_END(mp);
718 }
719 }
720 if ((mp->mnt_flag & MNT_SOFTDEP) != 0) {
721 printf("%s: `-o softdep' is no longer supported, "
722 "consider `-o log'\n", mp->mnt_stat.f_mntfromname);
723 mp->mnt_flag &= ~MNT_SOFTDEP;
724 }
725
726 if (bdev != NULL)
727 bdevsw_release(bdev);
728 return (error);
729
730 fail:
731 if (bdev != NULL)
732 bdevsw_release(bdev);
733 vrele(devvp);
734 return (error);
735 }
736
737 /*
738 * Reload all incore data for a filesystem (used after running fsck on
739 * the root filesystem and finding things to fix). The filesystem must
740 * be mounted read-only.
741 *
742 * Things to do to update the mount:
743 * 1) invalidate all cached meta-data.
744 * 2) re-read superblock from disk.
745 * 3) re-read summary information from disk.
746 * 4) invalidate all inactive vnodes.
747 * 5) invalidate all cached file data.
748 * 6) re-read inode data for all active vnodes.
749 */
750 int
751 ffs_reload(struct mount *mp, kauth_cred_t cred, struct lwp *l)
752 {
753 struct vnode *vp, *devvp;
754 struct inode *ip;
755 void *space;
756 struct buf *bp;
757 struct fs *fs, *newfs;
758 int i, bsize, blks, error;
759 int32_t *lp, fs_sbsize;
760 struct ufsmount *ump;
761 daddr_t sblockloc;
762 struct vnode_iterator *marker;
763
764 if ((mp->mnt_flag & MNT_RDONLY) == 0)
765 return (EINVAL);
766
767 ump = VFSTOUFS(mp);
768
769 /*
770 * Step 1: invalidate all cached meta-data.
771 */
772 devvp = ump->um_devvp;
773 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
774 error = vinvalbuf(devvp, 0, cred, l, 0, 0);
775 VOP_UNLOCK(devvp);
776 if (error)
777 panic("%s: dirty1", __func__);
778
779 /*
780 * Step 2: re-read superblock from disk. XXX: We don't handle
781 * possibility that superblock moved. Which implies that we don't
782 * want its size to change either.
783 */
784 fs = ump->um_fs;
785 fs_sbsize = fs->fs_sbsize;
786 error = bread(devvp, fs->fs_sblockloc / DEV_BSIZE, fs_sbsize,
787 0, &bp);
788 if (error)
789 return (error);
790 newfs = kmem_alloc(fs_sbsize, KM_SLEEP);
791 memcpy(newfs, bp->b_data, fs_sbsize);
792
793 #ifdef FFS_EI
794 if (ump->um_flags & UFS_NEEDSWAP) {
795 ffs_sb_swap((struct fs *)bp->b_data, newfs);
796 newfs->fs_flags |= FS_SWAPPED;
797 } else
798 #endif
799 newfs->fs_flags &= ~FS_SWAPPED;
800
801 brelse(bp, 0);
802
803 if ((newfs->fs_magic != FS_UFS1_MAGIC) &&
804 (newfs->fs_magic != FS_UFS2_MAGIC)) {
805 kmem_free(newfs, fs_sbsize);
806 return (EIO); /* XXX needs translation */
807 }
808 if (!ffs_superblock_validate(newfs)) {
809 kmem_free(newfs, fs_sbsize);
810 return (EINVAL);
811 }
812
813 /*
814 * The current implementation doesn't handle the possibility that
815 * these values may have changed.
816 */
817 if ((newfs->fs_sbsize != fs_sbsize) ||
818 (newfs->fs_cssize != fs->fs_cssize) ||
819 (newfs->fs_contigsumsize != fs->fs_contigsumsize) ||
820 (newfs->fs_ncg != fs->fs_ncg)) {
821 kmem_free(newfs, fs_sbsize);
822 return (EINVAL);
823 }
824
825 /* Store off old fs_sblockloc for fs_oldfscompat_read. */
826 sblockloc = fs->fs_sblockloc;
827 /*
828 * Copy pointer fields back into superblock before copying in XXX
829 * new superblock. These should really be in the ufsmount. XXX
830 * Note that important parameters (eg fs_ncg) are unchanged.
831 */
832 newfs->fs_csp = fs->fs_csp;
833 newfs->fs_maxcluster = fs->fs_maxcluster;
834 newfs->fs_contigdirs = fs->fs_contigdirs;
835 newfs->fs_ronly = fs->fs_ronly;
836 newfs->fs_active = fs->fs_active;
837 memcpy(fs, newfs, (u_int)fs_sbsize);
838 kmem_free(newfs, fs_sbsize);
839
840 /*
841 * Recheck for Apple UFS filesystem.
842 */
843 ump->um_flags &= ~UFS_ISAPPLEUFS;
844 if (ffs_is_appleufs(devvp, fs)) {
845 #ifdef APPLE_UFS
846 ump->um_flags |= UFS_ISAPPLEUFS;
847 #else
848 DPRINTF("AppleUFS not supported");
849 return (EIO); /* XXX: really? */
850 #endif
851 }
852
853 if (UFS_MPISAPPLEUFS(ump)) {
854 /* see comment about NeXT below */
855 ump->um_maxsymlinklen = APPLEUFS_MAXSYMLINKLEN;
856 ump->um_dirblksiz = APPLEUFS_DIRBLKSIZ;
857 mp->mnt_iflag |= IMNT_DTYPE;
858 } else {
859 ump->um_maxsymlinklen = fs->fs_maxsymlinklen;
860 ump->um_dirblksiz = UFS_DIRBLKSIZ;
861 if (ump->um_maxsymlinklen > 0)
862 mp->mnt_iflag |= IMNT_DTYPE;
863 else
864 mp->mnt_iflag &= ~IMNT_DTYPE;
865 }
866 ffs_oldfscompat_read(fs, ump, sblockloc);
867
868 mutex_enter(&ump->um_lock);
869 ump->um_maxfilesize = fs->fs_maxfilesize;
870 if (fs->fs_flags & ~(FS_KNOWN_FLAGS | FS_INTERNAL)) {
871 uprintf("%s: unknown ufs flags: 0x%08"PRIx32"%s\n",
872 mp->mnt_stat.f_mntonname, fs->fs_flags,
873 (mp->mnt_flag & MNT_FORCE) ? "" : ", not mounting");
874 if ((mp->mnt_flag & MNT_FORCE) == 0) {
875 mutex_exit(&ump->um_lock);
876 return (EINVAL);
877 }
878 }
879 if (fs->fs_pendingblocks != 0 || fs->fs_pendinginodes != 0) {
880 fs->fs_pendingblocks = 0;
881 fs->fs_pendinginodes = 0;
882 }
883 mutex_exit(&ump->um_lock);
884
885 ffs_statvfs(mp, &mp->mnt_stat);
886 /*
887 * Step 3: re-read summary information from disk.
888 */
889 blks = howmany(fs->fs_cssize, fs->fs_fsize);
890 space = fs->fs_csp;
891 for (i = 0; i < blks; i += fs->fs_frag) {
892 bsize = fs->fs_bsize;
893 if (i + fs->fs_frag > blks)
894 bsize = (blks - i) * fs->fs_fsize;
895 error = bread(devvp, FFS_FSBTODB(fs, fs->fs_csaddr + i), bsize,
896 0, &bp);
897 if (error) {
898 return (error);
899 }
900 #ifdef FFS_EI
901 if (UFS_FSNEEDSWAP(fs))
902 ffs_csum_swap((struct csum *)bp->b_data,
903 (struct csum *)space, bsize);
904 else
905 #endif
906 memcpy(space, bp->b_data, (size_t)bsize);
907 space = (char *)space + bsize;
908 brelse(bp, 0);
909 }
910 /*
911 * We no longer know anything about clusters per cylinder group.
912 */
913 if (fs->fs_contigsumsize > 0) {
914 lp = fs->fs_maxcluster;
915 for (i = 0; i < fs->fs_ncg; i++)
916 *lp++ = fs->fs_contigsumsize;
917 }
918
919 vfs_vnode_iterator_init(mp, &marker);
920 while ((vp = vfs_vnode_iterator_next(marker, NULL, NULL))) {
921 /*
922 * Step 4: invalidate all inactive vnodes.
923 */
924 if (vrecycle(vp))
925 continue;
926 /*
927 * Step 5: invalidate all cached file data.
928 */
929 if (vn_lock(vp, LK_EXCLUSIVE)) {
930 vrele(vp);
931 continue;
932 }
933 if (vinvalbuf(vp, 0, cred, l, 0, 0))
934 panic("%s: dirty2", __func__);
935 /*
936 * Step 6: re-read inode data for all active vnodes.
937 */
938 ip = VTOI(vp);
939 error = bread(devvp, FFS_FSBTODB(fs, ino_to_fsba(fs, ip->i_number)),
940 (int)fs->fs_bsize, 0, &bp);
941 if (error) {
942 vput(vp);
943 break;
944 }
945 ffs_load_inode(bp, ip, fs, ip->i_number);
946 brelse(bp, 0);
947 vput(vp);
948 }
949 vfs_vnode_iterator_destroy(marker);
950 return (error);
951 }
952
953 /*
954 * Possible superblock locations ordered from most to least likely.
955 */
956 static const int sblock_try[] = SBLOCKSEARCH;
957
958
959 static int
960 ffs_superblock_validate(struct fs *fs)
961 {
962 int32_t i, fs_bshift = 0, fs_fshift = 0, fs_fragshift = 0, fs_frag;
963 int32_t fs_inopb;
964
965 /* Check the superblock size */
966 if (fs->fs_sbsize > SBLOCKSIZE || fs->fs_sbsize < sizeof(struct fs))
967 return 0;
968
969 /* Check the file system blocksize */
970 if (fs->fs_bsize > MAXBSIZE || fs->fs_bsize < MINBSIZE)
971 return 0;
972 if (!powerof2(fs->fs_bsize))
973 return 0;
974
975 /* Check the size of frag blocks */
976 if (!powerof2(fs->fs_fsize))
977 return 0;
978 if (fs->fs_fsize == 0)
979 return 0;
980
981 /*
982 * XXX: these values are just zero-checked to prevent obvious
983 * bugs. We need more strict checks.
984 */
985 if (fs->fs_size == 0)
986 return 0;
987 if (fs->fs_cssize == 0)
988 return 0;
989 if (fs->fs_ipg == 0)
990 return 0;
991 if (fs->fs_fpg == 0)
992 return 0;
993 if (fs->fs_ncg == 0)
994 return 0;
995 if (fs->fs_maxbpg == 0)
996 return 0;
997
998 /* Check the number of inodes per block */
999 if (fs->fs_magic == FS_UFS1_MAGIC)
1000 fs_inopb = fs->fs_bsize / sizeof(struct ufs1_dinode);
1001 else /* fs->fs_magic == FS_UFS2_MAGIC */
1002 fs_inopb = fs->fs_bsize / sizeof(struct ufs2_dinode);
1003 if (fs->fs_inopb != fs_inopb)
1004 return 0;
1005
1006 /* Block size cannot be smaller than fragment size */
1007 if (fs->fs_bsize < fs->fs_fsize)
1008 return 0;
1009
1010 /* Compute fs_bshift and ensure it is consistent */
1011 for (i = fs->fs_bsize; i > 1; i >>= 1)
1012 fs_bshift++;
1013 if (fs->fs_bshift != fs_bshift)
1014 return 0;
1015
1016 /* Compute fs_fshift and ensure it is consistent */
1017 for (i = fs->fs_fsize; i > 1; i >>= 1)
1018 fs_fshift++;
1019 if (fs->fs_fshift != fs_fshift)
1020 return 0;
1021
1022 /* Compute fs_fragshift and ensure it is consistent */
1023 for (i = fs->fs_frag; i > 1; i >>= 1)
1024 fs_fragshift++;
1025 if (fs->fs_fragshift != fs_fragshift)
1026 return 0;
1027
1028 /* Check the masks */
1029 if (fs->fs_bmask != ~(fs->fs_bsize - 1))
1030 return 0;
1031 if (fs->fs_fmask != ~(fs->fs_fsize - 1))
1032 return 0;
1033
1034 /*
1035 * Now that the shifts and masks are sanitized, we can use the ffs_ API.
1036 */
1037
1038 /* Check the number of frag blocks */
1039 if ((fs_frag = ffs_numfrags(fs, fs->fs_bsize)) > MAXFRAG)
1040 return 0;
1041 if (fs->fs_frag != fs_frag)
1042 return 0;
1043
1044 /* Check the size of cylinder groups */
1045 if ((fs->fs_cgsize < sizeof(struct cg)) ||
1046 (fs->fs_cgsize > fs->fs_bsize))
1047 return 0;
1048
1049 return 1;
1050 }
1051
1052 static int
1053 ffs_is_appleufs(struct vnode *devvp, struct fs *fs)
1054 {
1055 struct dkwedge_info dkw;
1056 int ret = 0;
1057
1058 /*
1059 * First check to see if this is tagged as an Apple UFS filesystem
1060 * in the disklabel.
1061 */
1062 if (getdiskinfo(devvp, &dkw) == 0 &&
1063 strcmp(dkw.dkw_ptype, DKW_PTYPE_APPLEUFS) == 0)
1064 ret = 1;
1065 #ifdef APPLE_UFS
1066 else {
1067 struct appleufslabel *applefs;
1068 struct buf *bp;
1069 daddr_t blkno = APPLEUFS_LABEL_OFFSET / DEV_BSIZE;
1070 int error;
1071
1072 /*
1073 * Manually look for an Apple UFS label, and if a valid one
1074 * is found, then treat it like an Apple UFS filesystem anyway.
1075 */
1076 error = bread(devvp, blkno, APPLEUFS_LABEL_SIZE, 0, &bp);
1077 if (error) {
1078 DPRINTF("bread@0x%jx returned %d", (intmax_t)blkno, error);
1079 return 0;
1080 }
1081 applefs = (struct appleufslabel *)bp->b_data;
1082 error = ffs_appleufs_validate(fs->fs_fsmnt, applefs, NULL);
1083 if (error == 0)
1084 ret = 1;
1085 brelse(bp, 0);
1086 }
1087 #endif
1088
1089 return ret;
1090 }
1091
1092 /*
1093 * Common code for mount and mountroot
1094 */
1095 int
1096 ffs_mountfs(struct vnode *devvp, struct mount *mp, struct lwp *l)
1097 {
1098 struct ufsmount *ump = NULL;
1099 struct buf *bp = NULL;
1100 struct fs *fs = NULL;
1101 dev_t dev;
1102 void *space;
1103 daddr_t sblockloc = 0;
1104 int blks, fstype = 0;
1105 int error, i, bsize, ronly, bset = 0;
1106 #ifdef FFS_EI
1107 int needswap = 0; /* keep gcc happy */
1108 #endif
1109 int32_t *lp;
1110 kauth_cred_t cred;
1111 u_int32_t allocsbsize, fs_sbsize = 0;
1112
1113 dev = devvp->v_rdev;
1114 cred = l ? l->l_cred : NOCRED;
1115
1116 /* Flush out any old buffers remaining from a previous use. */
1117 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
1118 error = vinvalbuf(devvp, V_SAVE, cred, l, 0, 0);
1119 VOP_UNLOCK(devvp);
1120 if (error) {
1121 DPRINTF("vinvalbuf returned %d", error);
1122 return error;
1123 }
1124
1125 ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
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 | IMNT_CAN_RWTORO;
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 if (fs)
1543 kmem_free(fs, fs->fs_sbsize);
1544 spec_node_setmountedfs(devvp, NULL);
1545 if (bp)
1546 brelse(bp, bset);
1547 if (ump) {
1548 if (ump->um_oldfscompat)
1549 kmem_free(ump->um_oldfscompat, 512 + 3*sizeof(int32_t));
1550 mutex_destroy(&ump->um_lock);
1551 kmem_free(ump, sizeof(*ump));
1552 mp->mnt_data = NULL;
1553 }
1554 return (error);
1555 }
1556
1557 /*
1558 * Sanity checks for loading old filesystem superblocks.
1559 * See ffs_oldfscompat_write below for unwound actions.
1560 *
1561 * XXX - Parts get retired eventually.
1562 * Unfortunately new bits get added.
1563 */
1564 static void
1565 ffs_oldfscompat_read(struct fs *fs, struct ufsmount *ump, daddr_t sblockloc)
1566 {
1567 off_t maxfilesize;
1568 int32_t *extrasave;
1569
1570 if ((fs->fs_magic != FS_UFS1_MAGIC) ||
1571 (fs->fs_old_flags & FS_FLAGS_UPDATED))
1572 return;
1573
1574 if (!ump->um_oldfscompat)
1575 ump->um_oldfscompat = kmem_alloc(512 + 3*sizeof(int32_t),
1576 KM_SLEEP);
1577
1578 memcpy(ump->um_oldfscompat, &fs->fs_old_postbl_start, 512);
1579 extrasave = ump->um_oldfscompat;
1580 extrasave += 512/sizeof(int32_t);
1581 extrasave[0] = fs->fs_old_npsect;
1582 extrasave[1] = fs->fs_old_interleave;
1583 extrasave[2] = fs->fs_old_trackskew;
1584
1585 /* These fields will be overwritten by their
1586 * original values in fs_oldfscompat_write, so it is harmless
1587 * to modify them here.
1588 */
1589 fs->fs_cstotal.cs_ndir = fs->fs_old_cstotal.cs_ndir;
1590 fs->fs_cstotal.cs_nbfree = fs->fs_old_cstotal.cs_nbfree;
1591 fs->fs_cstotal.cs_nifree = fs->fs_old_cstotal.cs_nifree;
1592 fs->fs_cstotal.cs_nffree = fs->fs_old_cstotal.cs_nffree;
1593
1594 fs->fs_maxbsize = fs->fs_bsize;
1595 fs->fs_time = fs->fs_old_time;
1596 fs->fs_size = fs->fs_old_size;
1597 fs->fs_dsize = fs->fs_old_dsize;
1598 fs->fs_csaddr = fs->fs_old_csaddr;
1599 fs->fs_sblockloc = sblockloc;
1600
1601 fs->fs_flags = fs->fs_old_flags | (fs->fs_flags & FS_INTERNAL);
1602
1603 if (fs->fs_old_postblformat == FS_42POSTBLFMT) {
1604 fs->fs_old_nrpos = 8;
1605 fs->fs_old_npsect = fs->fs_old_nsect;
1606 fs->fs_old_interleave = 1;
1607 fs->fs_old_trackskew = 0;
1608 }
1609
1610 if (fs->fs_magic == FS_UFS1_MAGIC &&
1611 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 return (0);
1745 }
1746
1747 /*
1748 * Flush out all the files in a filesystem.
1749 */
1750 int
1751 ffs_flushfiles(struct mount *mp, int flags, struct lwp *l)
1752 {
1753 extern int doforce;
1754 struct ufsmount *ump;
1755 int error;
1756
1757 if (!doforce)
1758 flags &= ~FORCECLOSE;
1759 ump = VFSTOUFS(mp);
1760 #ifdef QUOTA
1761 if ((error = quota1_umount(mp, flags)) != 0)
1762 return (error);
1763 #endif
1764 #ifdef QUOTA2
1765 if ((error = quota2_umount(mp, flags)) != 0)
1766 return (error);
1767 #endif
1768 #ifdef UFS_EXTATTR
1769 if (ump->um_fstype == UFS1) {
1770 if (ump->um_extattr.uepm_flags & UFS_EXTATTR_UEPM_STARTED)
1771 ufs_extattr_stop(mp, l);
1772 if (ump->um_extattr.uepm_flags & UFS_EXTATTR_UEPM_INITIALIZED)
1773 ufs_extattr_uepm_destroy(&ump->um_extattr);
1774 mp->mnt_flag &= ~MNT_EXTATTR;
1775 }
1776 #endif
1777 if ((error = vflush(mp, 0, SKIPSYSTEM | flags)) != 0)
1778 return (error);
1779 ffs_snapshot_unmount(mp);
1780 /*
1781 * Flush all the files.
1782 */
1783 error = vflush(mp, NULLVP, flags);
1784 if (error)
1785 return (error);
1786 /*
1787 * Flush filesystem metadata.
1788 */
1789 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
1790 error = VOP_FSYNC(ump->um_devvp, l->l_cred, FSYNC_WAIT, 0, 0);
1791 VOP_UNLOCK(ump->um_devvp);
1792 if (flags & FORCECLOSE) /* XXXDBJ */
1793 error = 0;
1794
1795 #ifdef WAPBL
1796 if (error)
1797 return error;
1798 if (mp->mnt_wapbl) {
1799 error = wapbl_flush(mp->mnt_wapbl, 1);
1800 if (flags & FORCECLOSE)
1801 error = 0;
1802 }
1803 #endif
1804
1805 return (error);
1806 }
1807
1808 /*
1809 * Get file system statistics.
1810 */
1811 int
1812 ffs_statvfs(struct mount *mp, struct statvfs *sbp)
1813 {
1814 struct ufsmount *ump;
1815 struct fs *fs;
1816
1817 ump = VFSTOUFS(mp);
1818 fs = ump->um_fs;
1819 mutex_enter(&ump->um_lock);
1820 sbp->f_bsize = fs->fs_bsize;
1821 sbp->f_frsize = fs->fs_fsize;
1822 sbp->f_iosize = fs->fs_bsize;
1823 sbp->f_blocks = fs->fs_dsize;
1824 sbp->f_bfree = ffs_blkstofrags(fs, fs->fs_cstotal.cs_nbfree) +
1825 fs->fs_cstotal.cs_nffree + FFS_DBTOFSB(fs, fs->fs_pendingblocks);
1826 sbp->f_bresvd = ((u_int64_t) fs->fs_dsize * (u_int64_t)
1827 fs->fs_minfree) / (u_int64_t) 100;
1828 if (sbp->f_bfree > sbp->f_bresvd)
1829 sbp->f_bavail = sbp->f_bfree - sbp->f_bresvd;
1830 else
1831 sbp->f_bavail = 0;
1832 sbp->f_files = fs->fs_ncg * fs->fs_ipg - UFS_ROOTINO;
1833 sbp->f_ffree = fs->fs_cstotal.cs_nifree + fs->fs_pendinginodes;
1834 sbp->f_favail = sbp->f_ffree;
1835 sbp->f_fresvd = 0;
1836 mutex_exit(&ump->um_lock);
1837 copy_statvfs_info(sbp, mp);
1838
1839 return (0);
1840 }
1841
1842 struct ffs_sync_ctx {
1843 int waitfor;
1844 };
1845
1846 static bool
1847 ffs_sync_selector(void *cl, struct vnode *vp)
1848 {
1849 struct ffs_sync_ctx *c = cl;
1850 struct inode *ip;
1851
1852 ip = VTOI(vp);
1853 /*
1854 * Skip the vnode/inode if inaccessible.
1855 */
1856 if (ip == NULL || vp->v_type == VNON)
1857 return false;
1858
1859 /*
1860 * We deliberately update inode times here. This will
1861 * prevent a massive queue of updates accumulating, only
1862 * to be handled by a call to unmount.
1863 *
1864 * XXX It would be better to have the syncer trickle these
1865 * out. Adjustment needed to allow registering vnodes for
1866 * sync when the vnode is clean, but the inode dirty. Or
1867 * have ufs itself trickle out inode updates.
1868 *
1869 * If doing a lazy sync, we don't care about metadata or
1870 * data updates, because they are handled by each vnode's
1871 * synclist entry. In this case we are only interested in
1872 * writing back modified inodes.
1873 */
1874 if ((ip->i_flag & (IN_ACCESS | IN_CHANGE | IN_UPDATE |
1875 IN_MODIFY | IN_MODIFIED | IN_ACCESSED)) == 0 &&
1876 (c->waitfor == MNT_LAZY || (LIST_EMPTY(&vp->v_dirtyblkhd) &&
1877 UVM_OBJ_IS_CLEAN(&vp->v_uobj))))
1878 return false;
1879
1880 return true;
1881 }
1882
1883 /*
1884 * Go through the disk queues to initiate sandbagged IO;
1885 * go through the inodes to write those that have been modified;
1886 * initiate the writing of the super block if it has been modified.
1887 *
1888 * Note: we are always called with the filesystem marked `MPBUSY'.
1889 */
1890 int
1891 ffs_sync(struct mount *mp, int waitfor, kauth_cred_t cred)
1892 {
1893 struct vnode *vp;
1894 struct ufsmount *ump = VFSTOUFS(mp);
1895 struct fs *fs;
1896 struct vnode_iterator *marker;
1897 int error, allerror = 0;
1898 struct ffs_sync_ctx ctx;
1899
1900 fs = ump->um_fs;
1901 if (fs->fs_fmod != 0 && fs->fs_ronly != 0) { /* XXX */
1902 panic("%s: rofs mod, fs=%s", __func__, fs->fs_fsmnt);
1903 }
1904
1905 /*
1906 * Write back each (modified) inode.
1907 */
1908 vfs_vnode_iterator_init(mp, &marker);
1909
1910 ctx.waitfor = waitfor;
1911 while ((vp = vfs_vnode_iterator_next(marker, ffs_sync_selector, &ctx)))
1912 {
1913 error = vn_lock(vp,
1914 LK_EXCLUSIVE | (waitfor == MNT_LAZY ? LK_NOWAIT : 0));
1915 if (error) {
1916 vrele(vp);
1917 continue;
1918 }
1919 if (waitfor == MNT_LAZY) {
1920 error = UFS_WAPBL_BEGIN(vp->v_mount);
1921 if (!error) {
1922 error = ffs_update(vp, NULL, NULL,
1923 UPDATE_CLOSE);
1924 UFS_WAPBL_END(vp->v_mount);
1925 }
1926 } else {
1927 error = VOP_FSYNC(vp, cred, FSYNC_NOLOG |
1928 (waitfor == MNT_WAIT ? FSYNC_WAIT : 0), 0, 0);
1929 }
1930 if (error)
1931 allerror = error;
1932 vput(vp);
1933 }
1934 vfs_vnode_iterator_destroy(marker);
1935
1936 /*
1937 * Force stale file system control information to be flushed.
1938 */
1939 if (waitfor != MNT_LAZY && (ump->um_devvp->v_numoutput > 0 ||
1940 !LIST_EMPTY(&ump->um_devvp->v_dirtyblkhd))) {
1941 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
1942 if ((error = VOP_FSYNC(ump->um_devvp, cred,
1943 (waitfor == MNT_WAIT ? FSYNC_WAIT : 0) | FSYNC_NOLOG,
1944 0, 0)) != 0)
1945 allerror = error;
1946 VOP_UNLOCK(ump->um_devvp);
1947 }
1948 #if defined(QUOTA) || defined(QUOTA2)
1949 qsync(mp);
1950 #endif
1951 /*
1952 * Write back modified superblock.
1953 */
1954 if (fs->fs_fmod != 0) {
1955 fs->fs_fmod = 0;
1956 fs->fs_time = time_second;
1957 error = UFS_WAPBL_BEGIN(mp);
1958 if (error)
1959 allerror = error;
1960 else {
1961 if ((error = ffs_cgupdate(ump, waitfor)))
1962 allerror = error;
1963 UFS_WAPBL_END(mp);
1964 }
1965 }
1966
1967 #ifdef WAPBL
1968 if (mp->mnt_wapbl) {
1969 error = wapbl_flush(mp->mnt_wapbl, (waitfor == MNT_WAIT));
1970 if (error)
1971 allerror = error;
1972 }
1973 #endif
1974
1975 return (allerror);
1976 }
1977
1978 /*
1979 * Load inode from disk and initialize vnode.
1980 */
1981 static int
1982 ffs_init_vnode(struct ufsmount *ump, struct vnode *vp, ino_t ino)
1983 {
1984 struct fs *fs;
1985 struct inode *ip;
1986 struct buf *bp;
1987 int error;
1988
1989 fs = ump->um_fs;
1990
1991 /* Read in the disk contents for the inode. */
1992 error = bread(ump->um_devvp, FFS_FSBTODB(fs, ino_to_fsba(fs, ino)),
1993 (int)fs->fs_bsize, 0, &bp);
1994 if (error)
1995 return error;
1996
1997 /* Allocate and initialize inode. */
1998 ip = pool_cache_get(ffs_inode_cache, PR_WAITOK);
1999 memset(ip, 0, sizeof(struct inode));
2000 ip->i_ump = ump;
2001 ip->i_fs = fs;
2002 ip->i_dev = ump->um_dev;
2003 ip->i_number = ino;
2004 if (ump->um_fstype == UFS1)
2005 ip->i_din.ffs1_din = pool_cache_get(ffs_dinode1_cache,
2006 PR_WAITOK);
2007 else
2008 ip->i_din.ffs2_din = pool_cache_get(ffs_dinode2_cache,
2009 PR_WAITOK);
2010 ffs_load_inode(bp, ip, fs, ino);
2011 brelse(bp, 0);
2012 ip->i_vnode = vp;
2013 #if defined(QUOTA) || defined(QUOTA2)
2014 ufsquota_init(ip);
2015 #endif
2016
2017 /* Initialise vnode with this inode. */
2018 vp->v_tag = VT_UFS;
2019 vp->v_op = ffs_vnodeop_p;
2020 vp->v_vflag |= VV_LOCKSWORK;
2021 vp->v_data = ip;
2022
2023 /* Initialize genfs node. */
2024 genfs_node_init(vp, &ffs_genfsops);
2025
2026 return 0;
2027 }
2028
2029 /*
2030 * Undo ffs_init_vnode().
2031 */
2032 static void
2033 ffs_deinit_vnode(struct ufsmount *ump, struct vnode *vp)
2034 {
2035 struct inode *ip = VTOI(vp);
2036
2037 if (ump->um_fstype == UFS1)
2038 pool_cache_put(ffs_dinode1_cache, ip->i_din.ffs1_din);
2039 else
2040 pool_cache_put(ffs_dinode2_cache, ip->i_din.ffs2_din);
2041 pool_cache_put(ffs_inode_cache, ip);
2042
2043 genfs_node_destroy(vp);
2044 vp->v_data = NULL;
2045 }
2046
2047 /*
2048 * Read an inode from disk and initialize this vnode / inode pair.
2049 * Caller assures no other thread will try to load this inode.
2050 */
2051 int
2052 ffs_loadvnode(struct mount *mp, struct vnode *vp,
2053 const void *key, size_t key_len, const void **new_key)
2054 {
2055 ino_t ino;
2056 struct fs *fs;
2057 struct inode *ip;
2058 struct ufsmount *ump;
2059 int error;
2060
2061 KASSERT(key_len == sizeof(ino));
2062 memcpy(&ino, key, key_len);
2063 ump = VFSTOUFS(mp);
2064 fs = ump->um_fs;
2065
2066 error = ffs_init_vnode(ump, vp, ino);
2067 if (error)
2068 return error;
2069
2070 ip = VTOI(vp);
2071 if (ip->i_mode == 0) {
2072 ffs_deinit_vnode(ump, vp);
2073
2074 return ENOENT;
2075 }
2076
2077 /* Initialize the vnode from the inode. */
2078 ufs_vinit(mp, ffs_specop_p, ffs_fifoop_p, &vp);
2079
2080 /* Finish inode initialization. */
2081 ip->i_devvp = ump->um_devvp;
2082 vref(ip->i_devvp);
2083
2084 /*
2085 * Ensure that uid and gid are correct. This is a temporary
2086 * fix until fsck has been changed to do the update.
2087 */
2088
2089 if (fs->fs_magic == FS_UFS1_MAGIC && /* XXX */
2090 fs->fs_old_inodefmt < FS_44INODEFMT) { /* XXX */
2091 ip->i_uid = ip->i_ffs1_ouid; /* XXX */
2092 ip->i_gid = ip->i_ffs1_ogid; /* XXX */
2093 } /* XXX */
2094 uvm_vnp_setsize(vp, ip->i_size);
2095 *new_key = &ip->i_number;
2096 return 0;
2097 }
2098
2099 /*
2100 * Create a new inode on disk and initialize this vnode / inode pair.
2101 */
2102 int
2103 ffs_newvnode(struct mount *mp, struct vnode *dvp, struct vnode *vp,
2104 struct vattr *vap, kauth_cred_t cred,
2105 size_t *key_len, const void **new_key)
2106 {
2107 ino_t ino;
2108 struct fs *fs;
2109 struct inode *ip;
2110 struct timespec ts;
2111 struct ufsmount *ump;
2112 int error, mode;
2113
2114 KASSERT(dvp->v_mount == mp);
2115 KASSERT(vap->va_type != VNON);
2116
2117 *key_len = sizeof(ino);
2118 ump = VFSTOUFS(mp);
2119 fs = ump->um_fs;
2120 mode = MAKEIMODE(vap->va_type, vap->va_mode);
2121
2122 /* Allocate fresh inode. */
2123 error = ffs_valloc(dvp, mode, cred, &ino);
2124 if (error)
2125 return error;
2126
2127 /* Attach inode to vnode. */
2128 error = ffs_init_vnode(ump, vp, ino);
2129 if (error) {
2130 if (UFS_WAPBL_BEGIN(mp) == 0) {
2131 ffs_vfree(dvp, ino, mode);
2132 UFS_WAPBL_END(mp);
2133 }
2134 return error;
2135 }
2136
2137 ip = VTOI(vp);
2138 if (ip->i_mode) {
2139 panic("%s: dup alloc ino=%" PRId64 " on %s: mode %x/%x "
2140 "gen %x/%x size %" PRIx64 " blocks %" PRIx64,
2141 __func__, ino, fs->fs_fsmnt, DIP(ip, mode), ip->i_mode,
2142 DIP(ip, gen), ip->i_gen, DIP(ip, size), DIP(ip, blocks));
2143 }
2144 if (DIP(ip, size) || DIP(ip, blocks)) {
2145 printf("%s: ino=%" PRId64 " on %s: "
2146 "gen %x/%x has non zero blocks %" PRIx64 " or size %"
2147 PRIx64 "\n",
2148 __func__, ino, fs->fs_fsmnt, DIP(ip, gen), ip->i_gen,
2149 DIP(ip, blocks), DIP(ip, size));
2150 if ((ip)->i_ump->um_fstype == UFS1)
2151 panic("%s: dirty filesystem?", __func__);
2152 DIP_ASSIGN(ip, blocks, 0);
2153 DIP_ASSIGN(ip, size, 0);
2154 }
2155
2156 /* Set uid / gid. */
2157 if (cred == NOCRED || cred == FSCRED) {
2158 ip->i_gid = 0;
2159 ip->i_uid = 0;
2160 } else {
2161 ip->i_gid = VTOI(dvp)->i_gid;
2162 ip->i_uid = kauth_cred_geteuid(cred);
2163 }
2164 DIP_ASSIGN(ip, gid, ip->i_gid);
2165 DIP_ASSIGN(ip, uid, ip->i_uid);
2166
2167 #if defined(QUOTA) || defined(QUOTA2)
2168 error = UFS_WAPBL_BEGIN(mp);
2169 if (error) {
2170 ffs_deinit_vnode(ump, vp);
2171
2172 return error;
2173 }
2174 error = chkiq(ip, 1, cred, 0);
2175 if (error) {
2176 ffs_vfree(dvp, ino, mode);
2177 UFS_WAPBL_END(mp);
2178 ffs_deinit_vnode(ump, vp);
2179
2180 return error;
2181 }
2182 UFS_WAPBL_END(mp);
2183 #endif
2184
2185 /* Set type and finalize. */
2186 ip->i_flags = 0;
2187 DIP_ASSIGN(ip, flags, 0);
2188 ip->i_mode = mode;
2189 DIP_ASSIGN(ip, mode, mode);
2190 if (vap->va_rdev != VNOVAL) {
2191 /*
2192 * Want to be able to use this to make badblock
2193 * inodes, so don't truncate the dev number.
2194 */
2195 if (ump->um_fstype == UFS1)
2196 ip->i_ffs1_rdev = ufs_rw32(vap->va_rdev,
2197 UFS_MPNEEDSWAP(ump));
2198 else
2199 ip->i_ffs2_rdev = ufs_rw64(vap->va_rdev,
2200 UFS_MPNEEDSWAP(ump));
2201 }
2202 ufs_vinit(mp, ffs_specop_p, ffs_fifoop_p, &vp);
2203 ip->i_devvp = ump->um_devvp;
2204 vref(ip->i_devvp);
2205
2206 /* Set up a new generation number for this inode. */
2207 ip->i_gen++;
2208 DIP_ASSIGN(ip, gen, ip->i_gen);
2209 if (fs->fs_magic == FS_UFS2_MAGIC) {
2210 vfs_timestamp(&ts);
2211 ip->i_ffs2_birthtime = ts.tv_sec;
2212 ip->i_ffs2_birthnsec = ts.tv_nsec;
2213 }
2214
2215 uvm_vnp_setsize(vp, ip->i_size);
2216 *new_key = &ip->i_number;
2217 return 0;
2218 }
2219
2220 /*
2221 * File handle to vnode
2222 *
2223 * Have to be really careful about stale file handles:
2224 * - check that the inode number is valid
2225 * - call ffs_vget() to get the locked inode
2226 * - check for an unallocated inode (i_mode == 0)
2227 * - check that the given client host has export rights and return
2228 * those rights via. exflagsp and credanonp
2229 */
2230 int
2231 ffs_fhtovp(struct mount *mp, struct fid *fhp, struct vnode **vpp)
2232 {
2233 struct ufid ufh;
2234 int error;
2235
2236 if (fhp->fid_len != sizeof(struct ufid))
2237 return EINVAL;
2238
2239 memcpy(&ufh, fhp, sizeof(ufh));
2240 if ((error = ffs_checkrange(mp, ufh.ufid_ino)) != 0)
2241 return error;
2242
2243 return (ufs_fhtovp(mp, &ufh, vpp));
2244 }
2245
2246 /*
2247 * Vnode pointer to File handle
2248 */
2249 /* ARGSUSED */
2250 int
2251 ffs_vptofh(struct vnode *vp, struct fid *fhp, size_t *fh_size)
2252 {
2253 struct inode *ip;
2254 struct ufid ufh;
2255
2256 if (*fh_size < sizeof(struct ufid)) {
2257 *fh_size = sizeof(struct ufid);
2258 return E2BIG;
2259 }
2260 ip = VTOI(vp);
2261 *fh_size = sizeof(struct ufid);
2262 memset(&ufh, 0, sizeof(ufh));
2263 ufh.ufid_len = sizeof(struct ufid);
2264 ufh.ufid_ino = ip->i_number;
2265 ufh.ufid_gen = ip->i_gen;
2266 memcpy(fhp, &ufh, sizeof(ufh));
2267 return (0);
2268 }
2269
2270 void
2271 ffs_init(void)
2272 {
2273 if (ffs_initcount++ > 0)
2274 return;
2275
2276 ffs_inode_cache = pool_cache_init(sizeof(struct inode), 0, 0, 0,
2277 "ffsino", NULL, IPL_NONE, NULL, NULL, NULL);
2278 ffs_dinode1_cache = pool_cache_init(sizeof(struct ufs1_dinode), 0, 0, 0,
2279 "ffsdino1", NULL, IPL_NONE, NULL, NULL, NULL);
2280 ffs_dinode2_cache = pool_cache_init(sizeof(struct ufs2_dinode), 0, 0, 0,
2281 "ffsdino2", NULL, IPL_NONE, NULL, NULL, NULL);
2282 ufs_init();
2283 }
2284
2285 void
2286 ffs_reinit(void)
2287 {
2288 ufs_reinit();
2289 }
2290
2291 void
2292 ffs_done(void)
2293 {
2294 if (--ffs_initcount > 0)
2295 return;
2296
2297 ufs_done();
2298 pool_cache_destroy(ffs_dinode2_cache);
2299 pool_cache_destroy(ffs_dinode1_cache);
2300 pool_cache_destroy(ffs_inode_cache);
2301 }
2302
2303 /*
2304 * Write a superblock and associated information back to disk.
2305 */
2306 int
2307 ffs_sbupdate(struct ufsmount *mp, int waitfor)
2308 {
2309 struct fs *fs = mp->um_fs;
2310 struct buf *bp;
2311 int error;
2312 u_int32_t saveflag;
2313
2314 error = ffs_getblk(mp->um_devvp,
2315 fs->fs_sblockloc / DEV_BSIZE, FFS_NOBLK,
2316 fs->fs_sbsize, false, &bp);
2317 if (error)
2318 return error;
2319 saveflag = fs->fs_flags & FS_INTERNAL;
2320 fs->fs_flags &= ~FS_INTERNAL;
2321
2322 memcpy(bp->b_data, fs, fs->fs_sbsize);
2323
2324 ffs_oldfscompat_write((struct fs *)bp->b_data, mp);
2325 #ifdef FFS_EI
2326 if (mp->um_flags & UFS_NEEDSWAP)
2327 ffs_sb_swap((struct fs *)bp->b_data, (struct fs *)bp->b_data);
2328 #endif
2329 fs->fs_flags |= saveflag;
2330
2331 if (waitfor == MNT_WAIT)
2332 error = bwrite(bp);
2333 else
2334 bawrite(bp);
2335 return (error);
2336 }
2337
2338 int
2339 ffs_cgupdate(struct ufsmount *mp, int waitfor)
2340 {
2341 struct fs *fs = mp->um_fs;
2342 struct buf *bp;
2343 int blks;
2344 void *space;
2345 int i, size, error = 0, allerror = 0;
2346
2347 UFS_WAPBL_JLOCK_ASSERT(mp);
2348
2349 allerror = ffs_sbupdate(mp, waitfor);
2350 blks = howmany(fs->fs_cssize, fs->fs_fsize);
2351 space = fs->fs_csp;
2352 for (i = 0; i < blks; i += fs->fs_frag) {
2353 size = fs->fs_bsize;
2354 if (i + fs->fs_frag > blks)
2355 size = (blks - i) * fs->fs_fsize;
2356 error = ffs_getblk(mp->um_devvp, FFS_FSBTODB(fs, fs->fs_csaddr + i),
2357 FFS_NOBLK, size, false, &bp);
2358 if (error)
2359 break;
2360 #ifdef FFS_EI
2361 if (mp->um_flags & UFS_NEEDSWAP)
2362 ffs_csum_swap((struct csum*)space,
2363 (struct csum*)bp->b_data, size);
2364 else
2365 #endif
2366 memcpy(bp->b_data, space, (u_int)size);
2367 space = (char *)space + size;
2368 if (waitfor == MNT_WAIT)
2369 error = bwrite(bp);
2370 else
2371 bawrite(bp);
2372 }
2373 if (!allerror && error)
2374 allerror = error;
2375 return (allerror);
2376 }
2377
2378 int
2379 ffs_extattrctl(struct mount *mp, int cmd, struct vnode *vp,
2380 int attrnamespace, const char *attrname)
2381 {
2382 #ifdef UFS_EXTATTR
2383 /*
2384 * File-backed extended attributes are only supported on UFS1.
2385 * UFS2 has native extended attributes.
2386 */
2387 if (VFSTOUFS(mp)->um_fstype == UFS1)
2388 return (ufs_extattrctl(mp, cmd, vp, attrnamespace, attrname));
2389 #endif
2390 return (vfs_stdextattrctl(mp, cmd, vp, attrnamespace, attrname));
2391 }
2392
2393 /*
2394 * Synch vnode for a mounted file system.
2395 */
2396 static int
2397 ffs_vfs_fsync(vnode_t *vp, int flags)
2398 {
2399 int error, i, pflags;
2400 #ifdef WAPBL
2401 struct mount *mp;
2402 #endif
2403
2404 KASSERT(vp->v_type == VBLK);
2405 KASSERT(spec_node_getmountedfs(vp) != NULL);
2406
2407 /*
2408 * Flush all dirty data associated with the vnode.
2409 */
2410 pflags = PGO_ALLPAGES | PGO_CLEANIT;
2411 if ((flags & FSYNC_WAIT) != 0)
2412 pflags |= PGO_SYNCIO;
2413 mutex_enter(vp->v_interlock);
2414 error = VOP_PUTPAGES(vp, 0, 0, pflags);
2415 if (error)
2416 return error;
2417
2418 #ifdef WAPBL
2419 mp = spec_node_getmountedfs(vp);
2420 if (mp && mp->mnt_wapbl) {
2421 /*
2422 * Don't bother writing out metadata if the syncer is
2423 * making the request. We will let the sync vnode
2424 * write it out in a single burst through a call to
2425 * VFS_SYNC().
2426 */
2427 if ((flags & (FSYNC_DATAONLY | FSYNC_LAZY | FSYNC_NOLOG)) != 0)
2428 return 0;
2429
2430 /*
2431 * Don't flush the log if the vnode being flushed
2432 * contains no dirty buffers that could be in the log.
2433 */
2434 if (!LIST_EMPTY(&vp->v_dirtyblkhd)) {
2435 error = wapbl_flush(mp->mnt_wapbl, 0);
2436 if (error)
2437 return error;
2438 }
2439
2440 if ((flags & FSYNC_WAIT) != 0) {
2441 mutex_enter(vp->v_interlock);
2442 while (vp->v_numoutput)
2443 cv_wait(&vp->v_cv, vp->v_interlock);
2444 mutex_exit(vp->v_interlock);
2445 }
2446
2447 return 0;
2448 }
2449 #endif /* WAPBL */
2450
2451 error = vflushbuf(vp, flags);
2452 if (error == 0 && (flags & FSYNC_CACHE) != 0) {
2453 i = 1;
2454 (void)VOP_IOCTL(vp, DIOCCACHESYNC, &i, FWRITE,
2455 kauth_cred_get());
2456 }
2457
2458 return error;
2459 }
2460