1/* $NetBSD: ffs_vfsops.c,v 1.341 2016/10/20 19:31:32 jdolecek 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 1.341 2016/10/20 19:31:32 jdolecek 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
113MODULE(MODULE_CLASS_VFS, ffs, "wapbl");
114#else
115MODULE(MODULE_CLASS_VFS, ffs, NULL);
116#endif
117
118static int ffs_vfs_fsync(vnode_t *, int);
119static int ffs_superblock_validate(struct fs *);
120static int ffs_is_appleufs(struct vnode *, struct fs *);
121
122static int ffs_init_vnode(struct ufsmount *, struct vnode *, ino_t);
123static void ffs_deinit_vnode(struct ufsmount *, struct vnode *);
124
125static struct sysctllog *ffs_sysctl_log;
126
127static kauth_listener_t ffs_snapshot_listener;
128
129/* how many times ffs_init() was called */
130int 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
138extern const struct vnodeopv_desc ffs_vnodeop_opv_desc;
139extern const struct vnodeopv_desc ffs_specop_opv_desc;
140extern const struct vnodeopv_desc ffs_fifoop_opv_desc;
141
142const 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
149struct 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
177static 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
184static 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
194static int
195ffs_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
242static int
243ffs_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
258static int
259ffs_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
348pool_cache_t ffs_inode_cache;
349pool_cache_t ffs_dinode1_cache;
350pool_cache_t ffs_dinode2_cache;
351
352static void ffs_oldfscompat_read(struct fs *, struct ufsmount *, daddr_t);
353static 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
359int
360ffs_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 */
402int
403ffs_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
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 if (bdevsw_lookup(devvp->v_rdev) == NULL) {
455 DPRINTF("can't find block device 0x%jx",
456 devvp->v_rdev);
457 error = ENXIO;
458 }
459 } else {
460 /*
461 * Be sure we're still naming the same device
462 * used for our initial mount
463 */
464 ump = VFSTOUFS(mp);
465 if (devvp != ump->um_devvp) {
466 if (devvp->v_rdev != ump->um_devvp->v_rdev) {
467 DPRINTF("wrong device 0x%jx != 0x%jx",
468 (uintmax_t)devvp->v_rdev,
469 (uintmax_t)ump->um_devvp->v_rdev);
470 error = EINVAL;
471 } else {
472 vrele(devvp);
473 devvp = ump->um_devvp;
474 vref(devvp);
475 }
476 }
477 }
478 } else {
479 if (!update) {
480 /* New mounts must have a filename for the device */
481 DPRINTF("no filename for mount");
482 return EINVAL;
483 } else {
484 /* Use the extant mount */
485 ump = VFSTOUFS(mp);
486 devvp = ump->um_devvp;
487 vref(devvp);
488 }
489 }
490
491 /*
492 * If mount by non-root, then verify that user has necessary
493 * permissions on the device.
494 *
495 * Permission to update a mount is checked higher, so here we presume
496 * updating the mount is okay (for example, as far as securelevel goes)
497 * which leaves us with the normal check.
498 */
499 if (error == 0) {
500 accessmode = VREAD;
501 if (update ?
502 (mp->mnt_iflag & IMNT_WANTRDWR) != 0 :
503 (mp->mnt_flag & MNT_RDONLY) == 0)
504 accessmode |= VWRITE;
505 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
506 error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT,
507 KAUTH_REQ_SYSTEM_MOUNT_DEVICE, mp, devvp,
508 KAUTH_ARG(accessmode));
509 if (error) {
510 DPRINTF("kauth returned %d", error);
511 }
512 VOP_UNLOCK(devvp);
513 }
514
515 if (error) {
516 vrele(devvp);
517 return (error);
518 }
519
520#ifdef WAPBL
521 /* WAPBL can only be enabled on a r/w mount. */
522 if ((mp->mnt_flag & MNT_RDONLY) && !(mp->mnt_iflag & IMNT_WANTRDWR)) {
523 mp->mnt_flag &= ~MNT_LOG;
524 }
525#else /* !WAPBL */
526 mp->mnt_flag &= ~MNT_LOG;
527#endif /* !WAPBL */
528
529 if (!update) {
530 int xflags;
531
532 if (mp->mnt_flag & MNT_RDONLY)
533 xflags = FREAD;
534 else
535 xflags = FREAD | FWRITE;
536 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
537 error = VOP_OPEN(devvp, xflags, FSCRED);
538 VOP_UNLOCK(devvp);
539 if (error) {
540 DPRINTF("VOP_OPEN returned %d", error);
541 goto fail;
542 }
543 error = ffs_mountfs(devvp, mp, l);
544 if (error) {
545 DPRINTF("ffs_mountfs returned %d", error);
546 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
547 (void)VOP_CLOSE(devvp, xflags, NOCRED);
548 VOP_UNLOCK(devvp);
549 goto fail;
550 }
551
552 ump = VFSTOUFS(mp);
553 fs = ump->um_fs;
554 } else {
555 /*
556 * Update the mount.
557 */
558
559 /*
560 * The initial mount got a reference on this
561 * device, so drop the one obtained via
562 * namei(), above.
563 */
564 vrele(devvp);
565
566 ump = VFSTOUFS(mp);
567 fs = ump->um_fs;
568 if (fs->fs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) {
569 /*
570 * Changing from r/w to r/o
571 */
572 flags = WRITECLOSE;
573 if (mp->mnt_flag & MNT_FORCE)
574 flags |= FORCECLOSE;
575 error = ffs_flushfiles(mp, flags, l);
576 if (error == 0)
577 error = UFS_WAPBL_BEGIN(mp);
578 if (error == 0 &&
579 ffs_cgupdate(ump, MNT_WAIT) == 0 &&
580 fs->fs_clean & FS_WASCLEAN) {
581 if (mp->mnt_flag & MNT_SOFTDEP)
582 fs->fs_flags &= ~FS_DOSOFTDEP;
583 fs->fs_clean = FS_ISCLEAN;
584 (void) ffs_sbupdate(ump, MNT_WAIT);
585 }
586 if (error) {
587 DPRINTF("wapbl %d", error);
588 return error;
589 }
590 UFS_WAPBL_END(mp);
591 }
592
593#ifdef WAPBL
594 if ((mp->mnt_flag & MNT_LOG) == 0) {
595 error = ffs_wapbl_stop(mp, mp->mnt_flag & MNT_FORCE);
596 if (error) {
597 DPRINTF("ffs_wapbl_stop returned %d", error);
598 return error;
599 }
600 }
601#endif /* WAPBL */
602
603 if (fs->fs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) {
604 /*
605 * Finish change from r/w to r/o
606 */
607 fs->fs_ronly = 1;
608 fs->fs_fmod = 0;
609 }
610
611 if (mp->mnt_flag & MNT_RELOAD) {
612 error = ffs_reload(mp, l->l_cred, l);
613 if (error) {
614 DPRINTF("ffs_reload returned %d", error);
615 return error;
616 }
617 }
618
619 if (fs->fs_ronly && (mp->mnt_iflag & IMNT_WANTRDWR)) {
620 /*
621 * Changing from read-only to read/write
622 */
623#ifndef QUOTA2
624 if (fs->fs_flags & FS_DOQUOTA2) {
625 ump->um_flags |= UFS_QUOTA2;
626 uprintf("%s: options QUOTA2 not enabled%s\n",
627 mp->mnt_stat.f_mntonname,
628 (mp->mnt_flag & MNT_FORCE) ? "" :
629 ", not mounting");
630 DPRINTF("ffs_quota2 %d", EINVAL);
631 return EINVAL;
632 }
633#endif
634 fs->fs_ronly = 0;
635 fs->fs_clean <<= 1;
636 fs->fs_fmod = 1;
637#ifdef WAPBL
638 if (fs->fs_flags & FS_DOWAPBL) {
639 const char *nm = mp->mnt_stat.f_mntonname;
640 if (!mp->mnt_wapbl_replay) {
641 printf("%s: log corrupted;"
642 " replay cancelled\n", nm);
643 return EFTYPE;
644 }
645 printf("%s: replaying log to disk\n", nm);
646 error = wapbl_replay_write(mp->mnt_wapbl_replay,
647 devvp);
648 if (error) {
649 DPRINTF("%s: wapbl_replay_write %d",
650 nm, error);
651 return error;
652 }
653 wapbl_replay_stop(mp->mnt_wapbl_replay);
654 fs->fs_clean = FS_WASCLEAN;
655 }
656#endif /* WAPBL */
657 if (fs->fs_snapinum[0] != 0)
658 ffs_snapshot_mount(mp);
659 }
660
661#ifdef WAPBL
662 error = ffs_wapbl_start(mp);
663 if (error) {
664 DPRINTF("ffs_wapbl_start returned %d", error);
665 return error;
666 }
667#endif /* WAPBL */
668
669#ifdef QUOTA2
670 if (!fs->fs_ronly) {
671 error = ffs_quota2_mount(mp);
672 if (error) {
673 DPRINTF("ffs_quota2_mount returned %d", error);
674 return error;
675 }
676 }
677#endif
678
679 if ((mp->mnt_flag & MNT_DISCARD) && !(ump->um_discarddata))
680 ump->um_discarddata = ffs_discard_init(devvp, fs);
681
682 if (args->fspec == NULL)
683 return 0;
684 }
685
686 error = set_statvfs_info(path, UIO_USERSPACE, args->fspec,
687 UIO_USERSPACE, mp->mnt_op->vfs_name, mp, l);
688 if (error == 0)
689 (void)strncpy(fs->fs_fsmnt, mp->mnt_stat.f_mntonname,
690 sizeof(fs->fs_fsmnt));
691 else {
692 DPRINTF("set_statvfs_info returned %d", error);
693 }
694 fs->fs_flags &= ~FS_DOSOFTDEP;
695 if (fs->fs_fmod != 0) { /* XXX */
696 int err;
697
698 fs->fs_fmod = 0;
699 if (fs->fs_clean & FS_WASCLEAN)
700 fs->fs_time = time_second;
701 else {
702 printf("%s: file system not clean (fs_clean=%#x); "
703 "please fsck(8)\n", mp->mnt_stat.f_mntfromname,
704 fs->fs_clean);
705 printf("%s: lost blocks %" PRId64 " files %d\n",
706 mp->mnt_stat.f_mntfromname, fs->fs_pendingblocks,
707 fs->fs_pendinginodes);
708 }
709 err = UFS_WAPBL_BEGIN(mp);
710 if (err == 0) {
711 (void) ffs_cgupdate(ump, MNT_WAIT);
712 UFS_WAPBL_END(mp);
713 }
714 }
715 if ((mp->mnt_flag & MNT_SOFTDEP) != 0) {
716 printf("%s: `-o softdep' is no longer supported, "
717 "consider `-o log'\n", mp->mnt_stat.f_mntfromname);
718 mp->mnt_flag &= ~MNT_SOFTDEP;
719 }
720
721 return (error);
722
723fail:
724 vrele(devvp);
725 return (error);
726}
727
728/*
729 * Reload all incore data for a filesystem (used after running fsck on
730 * the root filesystem and finding things to fix). The filesystem must
731 * be mounted read-only.
732 *
733 * Things to do to update the mount:
734 * 1) invalidate all cached meta-data.
735 * 2) re-read superblock from disk.
736 * 3) re-read summary information from disk.
737 * 4) invalidate all inactive vnodes.
738 * 5) invalidate all cached file data.
739 * 6) re-read inode data for all active vnodes.
740 */
741int
742ffs_reload(struct mount *mp, kauth_cred_t cred, struct lwp *l)
743{
744 struct vnode *vp, *devvp;
745 struct inode *ip;
746 void *space;
747 struct buf *bp;
748 struct fs *fs, *newfs;
749 int i, bsize, blks, error;
750 int32_t *lp, fs_sbsize;
751 struct ufsmount *ump;
752 daddr_t sblockloc;
753 struct vnode_iterator *marker;
754
755 if ((mp->mnt_flag & MNT_RDONLY) == 0)
756 return (EINVAL);
757
758 ump = VFSTOUFS(mp);
759
760 /*
761 * Step 1: invalidate all cached meta-data.
762 */
763 devvp = ump->um_devvp;
764 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
765 error = vinvalbuf(devvp, 0, cred, l, 0, 0);
766 VOP_UNLOCK(devvp);
767 if (error)
768 panic("%s: dirty1", __func__);
769
770 /*
771 * Step 2: re-read superblock from disk. XXX: We don't handle
772 * possibility that superblock moved. Which implies that we don't
773 * want its size to change either.
774 */
775 fs = ump->um_fs;
776 fs_sbsize = fs->fs_sbsize;
777 error = bread(devvp, fs->fs_sblockloc / DEV_BSIZE, fs_sbsize,
778 0, &bp);
779 if (error)
780 return (error);
781 newfs = kmem_alloc(fs_sbsize, KM_SLEEP);
782 memcpy(newfs, bp->b_data, fs_sbsize);
783
784#ifdef FFS_EI
785 if (ump->um_flags & UFS_NEEDSWAP) {
786 ffs_sb_swap((struct fs *)bp->b_data, newfs);
787 newfs->fs_flags |= FS_SWAPPED;
788 } else
789#endif
790 newfs->fs_flags &= ~FS_SWAPPED;
791
792 brelse(bp, 0);
793
794 if ((newfs->fs_magic != FS_UFS1_MAGIC) &&
795 (newfs->fs_magic != FS_UFS2_MAGIC)) {
796 kmem_free(newfs, fs_sbsize);
797 return (EIO); /* XXX needs translation */
798 }
799 if (!ffs_superblock_validate(newfs)) {
800 kmem_free(newfs, fs_sbsize);
801 return (EINVAL);
802 }
803
804 /*
805 * The current implementation doesn't handle the possibility that
806 * these values may have changed.
807 */
808 if ((newfs->fs_sbsize != fs_sbsize) ||
809 (newfs->fs_cssize != fs->fs_cssize) ||
810 (newfs->fs_contigsumsize != fs->fs_contigsumsize) ||
811 (newfs->fs_ncg != fs->fs_ncg)) {
812 kmem_free(newfs, fs_sbsize);
813 return (EINVAL);
814 }
815
816 /* Store off old fs_sblockloc for fs_oldfscompat_read. */
817 sblockloc = fs->fs_sblockloc;
818 /*
819 * Copy pointer fields back into superblock before copying in XXX
820 * new superblock. These should really be in the ufsmount. XXX
821 * Note that important parameters (eg fs_ncg) are unchanged.
822 */
823 newfs->fs_csp = fs->fs_csp;
824 newfs->fs_maxcluster = fs->fs_maxcluster;
825 newfs->fs_contigdirs = fs->fs_contigdirs;
826 newfs->fs_ronly = fs->fs_ronly;
827 newfs->fs_active = fs->fs_active;
828 memcpy(fs, newfs, (u_int)fs_sbsize);
829 kmem_free(newfs, fs_sbsize);
830
831 /*
832 * Recheck for Apple UFS filesystem.
833 */
834 ump->um_flags &= ~UFS_ISAPPLEUFS;
835 if (ffs_is_appleufs(devvp, fs)) {
836#ifdef APPLE_UFS
837 ump->um_flags |= UFS_ISAPPLEUFS;
838#else
839 DPRINTF("AppleUFS not supported");
840 return (EIO); /* XXX: really? */
841#endif
842 }
843
844 if (UFS_MPISAPPLEUFS(ump)) {
845 /* see comment about NeXT below */
846 ump->um_maxsymlinklen = APPLEUFS_MAXSYMLINKLEN;
847 ump->um_dirblksiz = APPLEUFS_DIRBLKSIZ;
848 mp->mnt_iflag |= IMNT_DTYPE;
849 } else {
850 ump->um_maxsymlinklen = fs->fs_maxsymlinklen;
851 ump->um_dirblksiz = UFS_DIRBLKSIZ;
852 if (ump->um_maxsymlinklen > 0)
853 mp->mnt_iflag |= IMNT_DTYPE;
854 else
855 mp->mnt_iflag &= ~IMNT_DTYPE;
856 }
857 ffs_oldfscompat_read(fs, ump, sblockloc);
858
859 mutex_enter(&ump->um_lock);
860 ump->um_maxfilesize = fs->fs_maxfilesize;
861 if (fs->fs_flags & ~(FS_KNOWN_FLAGS | FS_INTERNAL)) {
862 uprintf("%s: unknown ufs flags: 0x%08"PRIx32"%s\n",
863 mp->mnt_stat.f_mntonname, fs->fs_flags,
864 (mp->mnt_flag & MNT_FORCE) ? "" : ", not mounting");
865 if ((mp->mnt_flag & MNT_FORCE) == 0) {
866 mutex_exit(&ump->um_lock);
867 return (EINVAL);
868 }
869 }
870 if (fs->fs_pendingblocks != 0 || fs->fs_pendinginodes != 0) {
871 fs->fs_pendingblocks = 0;
872 fs->fs_pendinginodes = 0;
873 }
874 mutex_exit(&ump->um_lock);
875
876 ffs_statvfs(mp, &mp->mnt_stat);
877 /*
878 * Step 3: re-read summary information from disk.
879 */
880 blks = howmany(fs->fs_cssize, fs->fs_fsize);
881 space = fs->fs_csp;
882 for (i = 0; i < blks; i += fs->fs_frag) {
883 bsize = fs->fs_bsize;
884 if (i + fs->fs_frag > blks)
885 bsize = (blks - i) * fs->fs_fsize;
886 error = bread(devvp, FFS_FSBTODB(fs, fs->fs_csaddr + i), bsize,
887 0, &bp);
888 if (error) {
889 return (error);
890 }
891#ifdef FFS_EI
892 if (UFS_FSNEEDSWAP(fs))
893 ffs_csum_swap((struct csum *)bp->b_data,
894 (struct csum *)space, bsize);
895 else
896#endif
897 memcpy(space, bp->b_data, (size_t)bsize);
898 space = (char *)space + bsize;
899 brelse(bp, 0);
900 }
901 /*
902 * We no longer know anything about clusters per cylinder group.
903 */
904 if (fs->fs_contigsumsize > 0) {
905 lp = fs->fs_maxcluster;
906 for (i = 0; i < fs->fs_ncg; i++)
907 *lp++ = fs->fs_contigsumsize;
908 }
909
910 vfs_vnode_iterator_init(mp, &marker);
911 while ((vp = vfs_vnode_iterator_next(marker, NULL, NULL))) {
912 /*
913 * Step 4: invalidate all inactive vnodes.
914 */
915 if (vrecycle(vp))
916 continue;
917 /*
918 * Step 5: invalidate all cached file data.
919 */
920 if (vn_lock(vp, LK_EXCLUSIVE)) {
921 vrele(vp);
922 continue;
923 }
924 if (vinvalbuf(vp, 0, cred, l, 0, 0))
925 panic("%s: dirty2", __func__);
926 /*
927 * Step 6: re-read inode data for all active vnodes.
928 */
929 ip = VTOI(vp);
930 error = bread(devvp, FFS_FSBTODB(fs, ino_to_fsba(fs, ip->i_number)),
931 (int)fs->fs_bsize, 0, &bp);
932 if (error) {
933 vput(vp);
934 break;
935 }
936 ffs_load_inode(bp, ip, fs, ip->i_number);
937 brelse(bp, 0);
938 vput(vp);
939 }
940 vfs_vnode_iterator_destroy(marker);
941 return (error);
942}
943
944/*
945 * Possible superblock locations ordered from most to least likely.
946 */
947static const int sblock_try[] = SBLOCKSEARCH;
948
949
950static int
951ffs_superblock_validate(struct fs *fs)
952{
953 int32_t i, fs_bshift = 0, fs_fshift = 0, fs_fragshift = 0, fs_frag;
954 int32_t fs_inopb;
955
956 /* Check the superblock size */
957 if (fs->fs_sbsize > SBLOCKSIZE || fs->fs_sbsize < sizeof(struct fs))
958 return 0;
959
960 /* Check the file system blocksize */
961 if (fs->fs_bsize > MAXBSIZE || fs->fs_bsize < MINBSIZE)
962 return 0;
963 if (!powerof2(fs->fs_bsize))
964 return 0;
965
966 /* Check the size of frag blocks */
967 if (!powerof2(fs->fs_fsize))
968 return 0;
969 if (fs->fs_fsize == 0)
970 return 0;
971
972 /*
973 * XXX: these values are just zero-checked to prevent obvious
974 * bugs. We need more strict checks.
975 */
976 if (fs->fs_size == 0)
977 return 0;
978 if (fs->fs_cssize == 0)
979 return 0;
980 if (fs->fs_ipg == 0)
981 return 0;
982 if (fs->fs_fpg == 0)
983 return 0;
984 if (fs->fs_ncg == 0)
985 return 0;
986 if (fs->fs_maxbpg == 0)
987 return 0;
988
989 /* Check the number of inodes per block */
990 if (fs->fs_magic == FS_UFS1_MAGIC)
991 fs_inopb = fs->fs_bsize / sizeof(struct ufs1_dinode);
992 else /* fs->fs_magic == FS_UFS2_MAGIC */
993 fs_inopb = fs->fs_bsize / sizeof(struct ufs2_dinode);
994 if (fs->fs_inopb != fs_inopb)
995 return 0;
996
997 /* Block size cannot be smaller than fragment size */
998 if (fs->fs_bsize < fs->fs_fsize)
999 return 0;
1000
1001 /* Compute fs_bshift and ensure it is consistent */
1002 for (i = fs->fs_bsize; i > 1; i >>= 1)
1003 fs_bshift++;
1004 if (fs->fs_bshift != fs_bshift)
1005 return 0;
1006
1007 /* Compute fs_fshift and ensure it is consistent */
1008 for (i = fs->fs_fsize; i > 1; i >>= 1)
1009 fs_fshift++;
1010 if (fs->fs_fshift != fs_fshift)
1011 return 0;
1012
1013 /* Compute fs_fragshift and ensure it is consistent */
1014 for (i = fs->fs_frag; i > 1; i >>= 1)
1015 fs_fragshift++;
1016 if (fs->fs_fragshift != fs_fragshift)
1017 return 0;
1018
1019 /* Check the masks */
1020 if (fs->fs_bmask != ~(fs->fs_bsize - 1))
1021 return 0;
1022 if (fs->fs_fmask != ~(fs->fs_fsize - 1))
1023 return 0;
1024
1025 /*
1026 * Now that the shifts and masks are sanitized, we can use the ffs_ API.
1027 */
1028
1029 /* Check the number of frag blocks */
1030 if ((fs_frag = ffs_numfrags(fs, fs->fs_bsize)) > MAXFRAG)
1031 return 0;
1032 if (fs->fs_frag != fs_frag)
1033 return 0;
1034
1035 /* Check the size of cylinder groups */
1036 if ((fs->fs_cgsize < sizeof(struct cg)) ||
1037 (fs->fs_cgsize > fs->fs_bsize))
1038 return 0;
1039
1040 return 1;
1041}
1042
1043static int
1044ffs_is_appleufs(struct vnode *devvp, struct fs *fs)
1045{
1046 struct dkwedge_info dkw;
1047 int ret = 0;
1048
1049 /*
1050 * First check to see if this is tagged as an Apple UFS filesystem
1051 * in the disklabel.
1052 */
1053 if (getdiskinfo(devvp, &dkw) == 0 &&
1054 strcmp(dkw.dkw_ptype, DKW_PTYPE_APPLEUFS) == 0)
1055 ret = 1;
1056#ifdef APPLE_UFS
1057 else {
1058 struct appleufslabel *applefs;
1059 struct buf *bp;
1060 daddr_t blkno = APPLEUFS_LABEL_OFFSET / DEV_BSIZE;
1061 int error;
1062
1063 /*
1064 * Manually look for an Apple UFS label, and if a valid one
1065 * is found, then treat it like an Apple UFS filesystem anyway.
1066 */
1067 error = bread(devvp, blkno, APPLEUFS_LABEL_SIZE, 0, &bp);
1068 if (error) {
1069 DPRINTF("bread@0x%jx returned %d", (intmax_t)blkno, error);
1070 return 0;
1071 }
1072 applefs = (struct appleufslabel *)bp->b_data;
1073 error = ffs_appleufs_validate(fs->fs_fsmnt, applefs, NULL);
1074 if (error == 0)
1075 ret = 1;
1076 brelse(bp, 0);
1077 }
1078#endif
1079
1080 return ret;
1081}
1082
1083/*
1084 * Common code for mount and mountroot
1085 */
1086int
1087ffs_mountfs(struct vnode *devvp, struct mount *mp, struct lwp *l)
1088{
1089 struct ufsmount *ump = NULL;
1090 struct buf *bp = NULL;
1091 struct fs *fs = NULL;
1092 dev_t dev;
1093 void *space;
1094 daddr_t sblockloc = 0;
1095 int blks, fstype = 0;
1096 int error, i, bsize, ronly, bset = 0;
1097#ifdef FFS_EI
1098 int needswap = 0; /* keep gcc happy */
1099#endif
1100 int32_t *lp;
1101 kauth_cred_t cred;
1102 u_int32_t allocsbsize, fs_sbsize = 0;
1103
1104 dev = devvp->v_rdev;
1105 cred = l ? l->l_cred : NOCRED;
1106
1107 /* Flush out any old buffers remaining from a previous use. */
1108 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
1109 error = vinvalbuf(devvp, V_SAVE, cred, l, 0, 0);
1110 VOP_UNLOCK(devvp);
1111 if (error) {
1112 DPRINTF("vinvalbuf returned %d", error);
1113 return error;
1114 }
1115
1116 ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
1117
1118 error = fstrans_mount(mp);
1119 if (error) {
1120 DPRINTF("fstrans_mount returned %d", error);
1121 return error;
1122 }
1123
1124 ump = kmem_zalloc(sizeof(*ump), KM_SLEEP);
1125 mutex_init(&ump->um_lock, MUTEX_DEFAULT, IPL_NONE);
1126 error = ffs_snapshot_init(ump);
1127 if (error) {
1128 DPRINTF("ffs_snapshot_init returned %d", error);
1129 goto out;
1130 }
1131 ump->um_ops = &ffs_ufsops;
1132
1133#ifdef WAPBL
1134 sbagain:
1135#endif
1136 /*
1137 * Try reading the superblock in each of its possible locations.
1138 */
1139 for (i = 0; ; i++) {
1140 daddr_t fs_sblockloc;
1141
1142 if (bp != NULL) {
1143 brelse(bp, BC_NOCACHE);
1144 bp = NULL;
1145 }
1146 if (sblock_try[i] == -1) {
1147 DPRINTF("no superblock found");
1148 error = EINVAL;
1149 fs = NULL;
1150 goto out;
1151 }
1152
1153 error = bread(devvp, sblock_try[i] / DEV_BSIZE, SBLOCKSIZE,
1154 0, &bp);
1155 if (error) {
1156 DPRINTF("bread@0x%x returned %d",
1157 sblock_try[i] / DEV_BSIZE, error);
1158 fs = NULL;
1159 goto out;
1160 }
1161 fs = (struct fs *)bp->b_data;
1162
1163 sblockloc = sblock_try[i];
1164 DPRINTF("fs_magic 0x%x", fs->fs_magic);
1165
1166 /*
1167 * Swap: here, we swap fs->fs_sbsize in order to get the correct
1168 * size to read the superblock. Once read, we swap the whole
1169 * superblock structure.
1170 */
1171 if (fs->fs_magic == FS_UFS1_MAGIC) {
1172 fs_sbsize = fs->fs_sbsize;
1173 fstype = UFS1;
1174#ifdef FFS_EI
1175 needswap = 0;
1176 } else if (fs->fs_magic == FS_UFS1_MAGIC_SWAPPED) {
1177 fs_sbsize = bswap32(fs->fs_sbsize);
1178 fstype = UFS1;
1179 needswap = 1;
1180#endif
1181 } else if (fs->fs_magic == FS_UFS2_MAGIC) {
1182 fs_sbsize = fs->fs_sbsize;
1183 fstype = UFS2;
1184#ifdef FFS_EI
1185 needswap = 0;
1186 } else if (fs->fs_magic == FS_UFS2_MAGIC_SWAPPED) {
1187 fs_sbsize = bswap32(fs->fs_sbsize);
1188 fstype = UFS2;
1189 needswap = 1;
1190#endif
1191 } else
1192 continue;
1193
1194 /* fs->fs_sblockloc isn't defined for old filesystems */
1195 if (fstype == UFS1 && !(fs->fs_old_flags & FS_FLAGS_UPDATED)) {
1196 if (sblockloc == SBLOCK_UFS2)
1197 /*
1198 * This is likely to be the first alternate
1199 * in a filesystem with 64k blocks.
1200 * Don't use it.
1201 */
1202 continue;
1203 fs_sblockloc = sblockloc;
1204 } else {
1205 fs_sblockloc = fs->fs_sblockloc;
1206#ifdef FFS_EI
1207 if (needswap)
1208 fs_sblockloc = bswap64(fs_sblockloc);
1209#endif
1210 }
1211
1212 /* Check we haven't found an alternate superblock */
1213 if (fs_sblockloc != sblockloc)
1214 continue;
1215
1216 /* Check the superblock size */
1217 if (fs_sbsize > SBLOCKSIZE || fs_sbsize < sizeof(struct fs))
1218 continue;
1219 fs = kmem_alloc((u_long)fs_sbsize, KM_SLEEP);
1220 memcpy(fs, bp->b_data, fs_sbsize);
1221
1222 /* Swap the whole superblock structure, if necessary. */
1223#ifdef FFS_EI
1224 if (needswap) {
1225 ffs_sb_swap((struct fs*)bp->b_data, fs);
1226 fs->fs_flags |= FS_SWAPPED;
1227 } else
1228#endif
1229 fs->fs_flags &= ~FS_SWAPPED;
1230
1231 /*
1232 * Now that everything is swapped, the superblock is ready to
1233 * be sanitized.
1234 */
1235 if (!ffs_superblock_validate(fs)) {
1236 kmem_free(fs, fs_sbsize);
1237 continue;
1238 }
1239
1240 /* Ok seems to be a good superblock */
1241 break;
1242 }
1243
1244 ump->um_fs = fs;
1245
1246#ifdef WAPBL
1247 if ((mp->mnt_wapbl_replay == 0) && (fs->fs_flags & FS_DOWAPBL)) {
1248 error = ffs_wapbl_replay_start(mp, fs, devvp);
1249 if (error && (mp->mnt_flag & MNT_FORCE) == 0) {
1250 DPRINTF("ffs_wapbl_replay_start returned %d", error);
1251 goto out;
1252 }
1253 if (!error) {
1254 if (!ronly) {
1255 /* XXX fsmnt may be stale. */
1256 printf("%s: replaying log to disk\n",
1257 fs->fs_fsmnt);
1258 error = wapbl_replay_write(mp->mnt_wapbl_replay,
1259 devvp);
1260 if (error) {
1261 DPRINTF("wapbl_replay_write returned %d",
1262 error);
1263 goto out;
1264 }
1265 wapbl_replay_stop(mp->mnt_wapbl_replay);
1266 fs->fs_clean = FS_WASCLEAN;
1267 } else {
1268 /* XXX fsmnt may be stale */
1269 printf("%s: replaying log to memory\n",
1270 fs->fs_fsmnt);
1271 }
1272
1273 /* Force a re-read of the superblock */
1274 brelse(bp, BC_INVAL);
1275 bp = NULL;
1276 kmem_free(fs, fs_sbsize);
1277 fs = NULL;
1278 goto sbagain;
1279 }
1280 }
1281#else /* !WAPBL */
1282 if ((fs->fs_flags & FS_DOWAPBL) && (mp->mnt_flag & MNT_FORCE) == 0) {
1283 error = EPERM;
1284 DPRINTF("no force %d", error);
1285 goto out;
1286 }
1287#endif /* !WAPBL */
1288
1289 ffs_oldfscompat_read(fs, ump, sblockloc);
1290 ump->um_maxfilesize = fs->fs_maxfilesize;
1291
1292 if (fs->fs_flags & ~(FS_KNOWN_FLAGS | FS_INTERNAL)) {
1293 uprintf("%s: unknown ufs flags: 0x%08"PRIx32"%s\n",
1294 mp->mnt_stat.f_mntonname, fs->fs_flags,
1295 (mp->mnt_flag & MNT_FORCE) ? "" : ", not mounting");
1296 if ((mp->mnt_flag & MNT_FORCE) == 0) {
1297 error = EINVAL;
1298 DPRINTF("no force %d", error);
1299 goto out;
1300 }
1301 }
1302
1303 if (fs->fs_pendingblocks != 0 || fs->fs_pendinginodes != 0) {
1304 fs->fs_pendingblocks = 0;
1305 fs->fs_pendinginodes = 0;
1306 }
1307
1308 ump->um_fstype = fstype;
1309 if (fs->fs_sbsize < SBLOCKSIZE)
1310 brelse(bp, BC_INVAL);
1311 else
1312 brelse(bp, 0);
1313 bp = NULL;
1314
1315 if (ffs_is_appleufs(devvp, fs)) {
1316#ifdef APPLE_UFS
1317 ump->um_flags |= UFS_ISAPPLEUFS;
1318#else
1319 DPRINTF("AppleUFS not supported");
1320 error = EINVAL;
1321 goto out;
1322#endif
1323 }
1324
1325#if 0
1326/*
1327 * XXX This code changes the behaviour of mounting dirty filesystems, to
1328 * XXX require "mount -f ..." to mount them. This doesn't match what
1329 * XXX mount(8) describes and is disabled for now.
1330 */
1331 /*
1332 * If the file system is not clean, don't allow it to be mounted
1333 * unless MNT_FORCE is specified. (Note: MNT_FORCE is always set
1334 * for the root file system.)
1335 */
1336 if (fs->fs_flags & FS_DOWAPBL) {
1337 /*
1338 * wapbl normally expects to be FS_WASCLEAN when the FS_DOWAPBL
1339 * bit is set, although there's a window in unmount where it
1340 * could be FS_ISCLEAN
1341 */
1342 if ((mp->mnt_flag & MNT_FORCE) == 0 &&
1343 (fs->fs_clean & (FS_WASCLEAN | FS_ISCLEAN)) == 0) {
1344 error = EPERM;
1345 goto out;
1346 }
1347 } else
1348 if ((fs->fs_clean & FS_ISCLEAN) == 0 &&
1349 (mp->mnt_flag & MNT_FORCE) == 0) {
1350 error = EPERM;
1351 goto out;
1352 }
1353#endif
1354
1355 /*
1356 * Verify that we can access the last block in the fs
1357 * if we're mounting read/write.
1358 */
1359 if (!ronly) {
1360 error = bread(devvp, FFS_FSBTODB(fs, fs->fs_size - 1),
1361 fs->fs_fsize, 0, &bp);
1362 if (error) {
1363 DPRINTF("bread@0x%jx returned %d",
1364 (intmax_t)FFS_FSBTODB(fs, fs->fs_size - 1),
1365 error);
1366 bset = BC_INVAL;
1367 goto out;
1368 }
1369 if (bp->b_bcount != fs->fs_fsize) {
1370 DPRINTF("bcount %x != fsize %x", bp->b_bcount,
1371 fs->fs_fsize);
1372 error = EINVAL;
1373 bset = BC_INVAL;
1374 goto out;
1375 }
1376 brelse(bp, BC_INVAL);
1377 bp = NULL;
1378 }
1379
1380 fs->fs_ronly = ronly;
1381 /* Don't bump fs_clean if we're replaying journal */
1382 if (!((fs->fs_flags & FS_DOWAPBL) && (fs->fs_clean & FS_WASCLEAN))) {
1383 if (ronly == 0) {
1384 fs->fs_clean <<= 1;
1385 fs->fs_fmod = 1;
1386 }
1387 }
1388
1389 bsize = fs->fs_cssize;
1390 blks = howmany(bsize, fs->fs_fsize);
1391 if (fs->fs_contigsumsize > 0)
1392 bsize += fs->fs_ncg * sizeof(int32_t);
1393 bsize += fs->fs_ncg * sizeof(*fs->fs_contigdirs);
1394 allocsbsize = bsize;
1395 space = kmem_alloc((u_long)allocsbsize, KM_SLEEP);
1396 fs->fs_csp = space;
1397
1398 for (i = 0; i < blks; i += fs->fs_frag) {
1399 bsize = fs->fs_bsize;
1400 if (i + fs->fs_frag > blks)
1401 bsize = (blks - i) * fs->fs_fsize;
1402 error = bread(devvp, FFS_FSBTODB(fs, fs->fs_csaddr + i), bsize,
1403 0, &bp);
1404 if (error) {
1405 DPRINTF("bread@0x%jx %d",
1406 (intmax_t)FFS_FSBTODB(fs, fs->fs_csaddr + i),
1407 error);
1408 goto out1;
1409 }
1410#ifdef FFS_EI
1411 if (needswap)
1412 ffs_csum_swap((struct csum *)bp->b_data,
1413 (struct csum *)space, bsize);
1414 else
1415#endif
1416 memcpy(space, bp->b_data, (u_int)bsize);
1417
1418 space = (char *)space + bsize;
1419 brelse(bp, 0);
1420 bp = NULL;
1421 }
1422 if (fs->fs_contigsumsize > 0) {
1423 fs->fs_maxcluster = lp = space;
1424 for (i = 0; i < fs->fs_ncg; i++)
1425 *lp++ = fs->fs_contigsumsize;
1426 space = lp;
1427 }
1428 bsize = fs->fs_ncg * sizeof(*fs->fs_contigdirs);
1429 fs->fs_contigdirs = space;
1430 space = (char *)space + bsize;
1431 memset(fs->fs_contigdirs, 0, bsize);
1432
1433 /* Compatibility for old filesystems - XXX */
1434 if (fs->fs_avgfilesize <= 0)
1435 fs->fs_avgfilesize = AVFILESIZ;
1436 if (fs->fs_avgfpdir <= 0)
1437 fs->fs_avgfpdir = AFPDIR;
1438 fs->fs_active = NULL;
1439
1440 mp->mnt_data = ump;
1441 mp->mnt_stat.f_fsidx.__fsid_val[0] = (long)dev;
1442 mp->mnt_stat.f_fsidx.__fsid_val[1] = makefstype(MOUNT_FFS);
1443 mp->mnt_stat.f_fsid = mp->mnt_stat.f_fsidx.__fsid_val[0];
1444 mp->mnt_stat.f_namemax = FFS_MAXNAMLEN;
1445 if (UFS_MPISAPPLEUFS(ump)) {
1446 /* NeXT used to keep short symlinks in the inode even
1447 * when using FS_42INODEFMT. In that case fs->fs_maxsymlinklen
1448 * is probably -1, but we still need to be able to identify
1449 * short symlinks.
1450 */
1451 ump->um_maxsymlinklen = APPLEUFS_MAXSYMLINKLEN;
1452 ump->um_dirblksiz = APPLEUFS_DIRBLKSIZ;
1453 mp->mnt_iflag |= IMNT_DTYPE;
1454 } else {
1455 ump->um_maxsymlinklen = fs->fs_maxsymlinklen;
1456 ump->um_dirblksiz = UFS_DIRBLKSIZ;
1457 if (ump->um_maxsymlinklen > 0)
1458 mp->mnt_iflag |= IMNT_DTYPE;
1459 else
1460 mp->mnt_iflag &= ~IMNT_DTYPE;
1461 }
1462 mp->mnt_fs_bshift = fs->fs_bshift;
1463 mp->mnt_dev_bshift = DEV_BSHIFT; /* XXX */
1464 mp->mnt_flag |= MNT_LOCAL;
1465 mp->mnt_iflag |= IMNT_MPSAFE;
1466#ifdef FFS_EI
1467 if (needswap)
1468 ump->um_flags |= UFS_NEEDSWAP;
1469#endif
1470 ump->um_mountp = mp;
1471 ump->um_dev = dev;
1472 ump->um_devvp = devvp;
1473 ump->um_nindir = fs->fs_nindir;
1474 ump->um_lognindir = ffs(fs->fs_nindir) - 1;
1475 ump->um_bptrtodb = fs->fs_fshift - DEV_BSHIFT;
1476 ump->um_seqinc = fs->fs_frag;
1477 for (i = 0; i < MAXQUOTAS; i++)
1478 ump->um_quotas[i] = NULLVP;
1479 spec_node_setmountedfs(devvp, mp);
1480 if (ronly == 0 && fs->fs_snapinum[0] != 0)
1481 ffs_snapshot_mount(mp);
1482#ifdef WAPBL
1483 if (!ronly) {
1484 KDASSERT(fs->fs_ronly == 0);
1485 /*
1486 * ffs_wapbl_start() needs mp->mnt_stat initialised if it
1487 * needs to create a new log file in-filesystem.
1488 */
1489 error = ffs_statvfs(mp, &mp->mnt_stat);
1490 if (error) {
1491 DPRINTF("ffs_statvfs returned %d", error);
1492 goto out1;
1493 }
1494
1495 error = ffs_wapbl_start(mp);
1496 if (error) {
1497 DPRINTF("ffs_wapbl_start returned %d", error);
1498 goto out1;
1499 }
1500 }
1501#endif /* WAPBL */
1502 if (ronly == 0) {
1503#ifdef QUOTA2
1504 error = ffs_quota2_mount(mp);
1505 if (error) {
1506 DPRINTF("ffs_quota2_mount returned %d", error);
1507 goto out1;
1508 }
1509#else
1510 if (fs->fs_flags & FS_DOQUOTA2) {
1511 ump->um_flags |= UFS_QUOTA2;
1512 uprintf("%s: options QUOTA2 not enabled%s\n",
1513 mp->mnt_stat.f_mntonname,
1514 (mp->mnt_flag & MNT_FORCE) ? "" : ", not mounting");
1515 if ((mp->mnt_flag & MNT_FORCE) == 0) {
1516 error = EINVAL;
1517 DPRINTF("quota disabled %d", error);
1518 goto out1;
1519 }
1520 }
1521#endif
1522 }
1523
1524 if (mp->mnt_flag & MNT_DISCARD)
1525 ump->um_discarddata = ffs_discard_init(devvp, fs);
1526
1527 return (0);
1528out1:
1529 kmem_free(fs->fs_csp, allocsbsize);
1530out:
1531#ifdef WAPBL
1532 if (mp->mnt_wapbl_replay) {
1533 wapbl_replay_stop(mp->mnt_wapbl_replay);
1534 wapbl_replay_free(mp->mnt_wapbl_replay);
1535 mp->mnt_wapbl_replay = 0;
1536 }
1537#endif
1538
1539 fstrans_unmount(mp);
1540 if (fs)
1541 kmem_free(fs, fs->fs_sbsize);
1542 spec_node_setmountedfs(devvp, NULL);
1543 if (bp)
1544 brelse(bp, bset);
1545 if (ump) {
1546 if (ump->um_oldfscompat)
1547 kmem_free(ump->um_oldfscompat, 512 + 3*sizeof(int32_t));
1548 mutex_destroy(&ump->um_lock);
1549 kmem_free(ump, sizeof(*ump));
1550 mp->mnt_data = NULL;
1551 }
1552 return (error);
1553}
1554
1555/*
1556 * Sanity checks for loading old filesystem superblocks.
1557 * See ffs_oldfscompat_write below for unwound actions.
1558 *
1559 * XXX - Parts get retired eventually.
1560 * Unfortunately new bits get added.
1561 */
1562static void
1563ffs_oldfscompat_read(struct fs *fs, struct ufsmount *ump, daddr_t sblockloc)
1564{
1565 off_t maxfilesize;
1566 int32_t *extrasave;
1567
1568 if ((fs->fs_magic != FS_UFS1_MAGIC) ||
1569 (fs->fs_old_flags & FS_FLAGS_UPDATED))
1570 return;
1571
1572 if (!ump->um_oldfscompat)
1573 ump->um_oldfscompat = kmem_alloc(512 + 3*sizeof(int32_t),
1574 KM_SLEEP);
1575
1576 memcpy(ump->um_oldfscompat, &fs->fs_old_postbl_start, 512);
1577 extrasave = ump->um_oldfscompat;
1578 extrasave += 512/sizeof(int32_t);
1579 extrasave[0] = fs->fs_old_npsect;
1580 extrasave[1] = fs->fs_old_interleave;
1581 extrasave[2] = fs->fs_old_trackskew;
1582
1583 /* These fields will be overwritten by their
1584 * original values in fs_oldfscompat_write, so it is harmless
1585 * to modify them here.
1586 */
1587 fs->fs_cstotal.cs_ndir = fs->fs_old_cstotal.cs_ndir;
1588 fs->fs_cstotal.cs_nbfree = fs->fs_old_cstotal.cs_nbfree;
1589 fs->fs_cstotal.cs_nifree = fs->fs_old_cstotal.cs_nifree;
1590 fs->fs_cstotal.cs_nffree = fs->fs_old_cstotal.cs_nffree;
1591
1592 fs->fs_maxbsize = fs->fs_bsize;
1593 fs->fs_time = fs->fs_old_time;
1594 fs->fs_size = fs->fs_old_size;
1595 fs->fs_dsize = fs->fs_old_dsize;
1596 fs->fs_csaddr = fs->fs_old_csaddr;
1597 fs->fs_sblockloc = sblockloc;
1598
1599 fs->fs_flags = fs->fs_old_flags | (fs->fs_flags & FS_INTERNAL);
1600
1601 if (fs->fs_old_postblformat == FS_42POSTBLFMT) {
1602 fs->fs_old_nrpos = 8;
1603 fs->fs_old_npsect = fs->fs_old_nsect;
1604 fs->fs_old_interleave = 1;
1605 fs->fs_old_trackskew = 0;
1606 }
1607
1608 if (fs->fs_magic == FS_UFS1_MAGIC &&
1609 fs->fs_old_inodefmt < FS_44INODEFMT) {
1610 fs->fs_maxfilesize = (u_quad_t) 1LL << 39;
1611 fs->fs_qbmask = ~fs->fs_bmask;
1612 fs->fs_qfmask = ~fs->fs_fmask;
1613 }
1614
1615 maxfilesize = (u_int64_t)0x80000000 * fs->fs_bsize - 1;
1616 if (fs->fs_maxfilesize > maxfilesize)
1617 fs->fs_maxfilesize = maxfilesize;
1618
1619 /* Compatibility for old filesystems */
1620 if (fs->fs_avgfilesize <= 0)
1621 fs->fs_avgfilesize = AVFILESIZ;
1622 if (fs->fs_avgfpdir <= 0)
1623 fs->fs_avgfpdir = AFPDIR;
1624
1625#if 0
1626 if (bigcgs) {
1627 fs->fs_save_cgsize = fs->fs_cgsize;
1628 fs->fs_cgsize = fs->fs_bsize;
1629 }
1630#endif
1631}
1632
1633/*
1634 * Unwinding superblock updates for old filesystems.
1635 * See ffs_oldfscompat_read above for details.
1636 *
1637 * XXX - Parts get retired eventually.
1638 * Unfortunately new bits get added.
1639 */
1640static void
1641ffs_oldfscompat_write(struct fs *fs, struct ufsmount *ump)
1642{
1643 int32_t *extrasave;
1644
1645 if ((fs->fs_magic != FS_UFS1_MAGIC) ||
1646 (fs->fs_old_flags & FS_FLAGS_UPDATED))
1647 return;
1648
1649 fs->fs_old_time = fs->fs_time;
1650 fs->fs_old_cstotal.cs_ndir = fs->fs_cstotal.cs_ndir;
1651 fs->fs_old_cstotal.cs_nbfree = fs->fs_cstotal.cs_nbfree;
1652 fs->fs_old_cstotal.cs_nifree = fs->fs_cstotal.cs_nifree;
1653 fs->fs_old_cstotal.cs_nffree = fs->fs_cstotal.cs_nffree;
1654 fs->fs_old_flags = fs->fs_flags;
1655
1656#if 0
1657 if (bigcgs) {
1658 fs->fs_cgsize = fs->fs_save_cgsize;
1659 }
1660#endif
1661
1662 memcpy(&fs->fs_old_postbl_start, ump->um_oldfscompat, 512);
1663 extrasave = ump->um_oldfscompat;
1664 extrasave += 512/sizeof(int32_t);
1665 fs->fs_old_npsect = extrasave[0];
1666 fs->fs_old_interleave = extrasave[1];
1667 fs->fs_old_trackskew = extrasave[2];
1668
1669}
1670
1671/*
1672 * unmount vfs operation
1673 */
1674int
1675ffs_unmount(struct mount *mp, int mntflags)
1676{
1677 struct lwp *l = curlwp;
1678 struct ufsmount *ump = VFSTOUFS(mp);
1679 struct fs *fs = ump->um_fs;
1680 int error, flags;
1681 u_int32_t bsize;
1682#ifdef WAPBL
1683 extern int doforce;
1684#endif
1685
1686 if (ump->um_discarddata) {
1687 ffs_discard_finish(ump->um_discarddata, mntflags);
1688 ump->um_discarddata = NULL;
1689 }
1690
1691 flags = 0;
1692 if (mntflags & MNT_FORCE)
1693 flags |= FORCECLOSE;
1694 if ((error = ffs_flushfiles(mp, flags, l)) != 0)
1695 return (error);
1696 error = UFS_WAPBL_BEGIN(mp);
1697 if (error == 0)
1698 if (fs->fs_ronly == 0 &&
1699 ffs_cgupdate(ump, MNT_WAIT) == 0 &&
1700 fs->fs_clean & FS_WASCLEAN) {
1701 fs->fs_clean = FS_ISCLEAN;
1702 fs->fs_fmod = 0;
1703 (void) ffs_sbupdate(ump, MNT_WAIT);
1704 }
1705 if (error == 0)
1706 UFS_WAPBL_END(mp);
1707#ifdef WAPBL
1708 KASSERT(!(mp->mnt_wapbl_replay && mp->mnt_wapbl));
1709 if (mp->mnt_wapbl_replay) {
1710 KDASSERT(fs->fs_ronly);
1711 wapbl_replay_stop(mp->mnt_wapbl_replay);
1712 wapbl_replay_free(mp->mnt_wapbl_replay);
1713 mp->mnt_wapbl_replay = 0;
1714 }
1715 error = ffs_wapbl_stop(mp, doforce && (mntflags & MNT_FORCE));
1716 if (error) {
1717 return error;
1718 }
1719#endif /* WAPBL */
1720
1721 if (ump->um_devvp->v_type != VBAD)
1722 spec_node_setmountedfs(ump->um_devvp, NULL);
1723 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
1724 (void)VOP_CLOSE(ump->um_devvp, fs->fs_ronly ? FREAD : FREAD | FWRITE,
1725 NOCRED);
1726 vput(ump->um_devvp);
1727
1728 bsize = fs->fs_cssize;
1729 if (fs->fs_contigsumsize > 0)
1730 bsize += fs->fs_ncg * sizeof(int32_t);
1731 bsize += fs->fs_ncg * sizeof(*fs->fs_contigdirs);
1732 kmem_free(fs->fs_csp, bsize);
1733
1734 kmem_free(fs, fs->fs_sbsize);
1735 if (ump->um_oldfscompat != NULL)
1736 kmem_free(ump->um_oldfscompat, 512 + 3*sizeof(int32_t));
1737 mutex_destroy(&ump->um_lock);
1738 ffs_snapshot_fini(ump);
1739 kmem_free(ump, sizeof(*ump));
1740 mp->mnt_data = NULL;
1741 mp->mnt_flag &= ~MNT_LOCAL;
1742 fstrans_unmount(mp);
1743 return (0);
1744}
1745
1746/*
1747 * Flush out all the files in a filesystem.
1748 */
1749int
1750ffs_flushfiles(struct mount *mp, int flags, struct lwp *l)
1751{
1752 extern int doforce;
1753 struct ufsmount *ump;
1754 int error;
1755
1756 if (!doforce)
1757 flags &= ~FORCECLOSE;
1758 ump = VFSTOUFS(mp);
1759#ifdef QUOTA
1760 if ((error = quota1_umount(mp, flags)) != 0)
1761 return (error);
1762#endif
1763#ifdef QUOTA2
1764 if ((error = quota2_umount(mp, flags)) != 0)
1765 return (error);
1766#endif
1767#ifdef UFS_EXTATTR
1768 if (ump->um_fstype == UFS1) {
1769 if (ump->um_extattr.uepm_flags & UFS_EXTATTR_UEPM_STARTED)
1770 ufs_extattr_stop(mp, l);
1771 if (ump->um_extattr.uepm_flags & UFS_EXTATTR_UEPM_INITIALIZED)
1772 ufs_extattr_uepm_destroy(&ump->um_extattr);
1773 mp->mnt_flag &= ~MNT_EXTATTR;
1774 }
1775#endif
1776 if ((error = vflush(mp, 0, SKIPSYSTEM | flags)) != 0)
1777 return (error);
1778 ffs_snapshot_unmount(mp);
1779 /*
1780 * Flush all the files.
1781 */
1782 error = vflush(mp, NULLVP, flags);
1783 if (error)
1784 return (error);
1785 /*
1786 * Flush filesystem metadata.
1787 */
1788 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
1789 error = VOP_FSYNC(ump->um_devvp, l->l_cred, FSYNC_WAIT, 0, 0);
1790 VOP_UNLOCK(ump->um_devvp);
1791 if (flags & FORCECLOSE) /* XXXDBJ */
1792 error = 0;
1793
1794#ifdef WAPBL
1795 if (error)
1796 return error;
1797 if (mp->mnt_wapbl) {
1798 error = wapbl_flush(mp->mnt_wapbl, 1);
1799 if (flags & FORCECLOSE)
1800 error = 0;
1801 }
1802#endif
1803
1804 return (error);
1805}
1806
1807/*
1808 * Get file system statistics.
1809 */
1810int
1811ffs_statvfs(struct mount *mp, struct statvfs *sbp)
1812{
1813 struct ufsmount *ump;
1814 struct fs *fs;
1815
1816 ump = VFSTOUFS(mp);
1817 fs = ump->um_fs;
1818 mutex_enter(&ump->um_lock);
1819 sbp->f_bsize = fs->fs_bsize;
1820 sbp->f_frsize = fs->fs_fsize;
1821 sbp->f_iosize = fs->fs_bsize;
1822 sbp->f_blocks = fs->fs_dsize;
1823 sbp->f_bfree = ffs_blkstofrags(fs, fs->fs_cstotal.cs_nbfree) +
1824 fs->fs_cstotal.cs_nffree + FFS_DBTOFSB(fs, fs->fs_pendingblocks);
1825 sbp->f_bresvd = ((u_int64_t) fs->fs_dsize * (u_int64_t)
1826 fs->fs_minfree) / (u_int64_t) 100;
1827 if (sbp->f_bfree > sbp->f_bresvd)
1828 sbp->f_bavail = sbp->f_bfree - sbp->f_bresvd;
1829 else
1830 sbp->f_bavail = 0;
1831 sbp->f_files = fs->fs_ncg * fs->fs_ipg - UFS_ROOTINO;
1832 sbp->f_ffree = fs->fs_cstotal.cs_nifree + fs->fs_pendinginodes;
1833 sbp->f_favail = sbp->f_ffree;
1834 sbp->f_fresvd = 0;
1835 mutex_exit(&ump->um_lock);
1836 copy_statvfs_info(sbp, mp);
1837
1838 return (0);
1839}
1840
1841struct ffs_sync_ctx {
1842 int waitfor;
1843 bool is_suspending;
1844};
1845
1846static bool
1847ffs_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 if (vp->v_type == VBLK && c->is_suspending)
1881 return false;
1882
1883 return true;
1884}
1885
1886/*
1887 * Go through the disk queues to initiate sandbagged IO;
1888 * go through the inodes to write those that have been modified;
1889 * initiate the writing of the super block if it has been modified.
1890 *
1891 * Note: we are always called with the filesystem marked `MPBUSY'.
1892 */
1893int
1894ffs_sync(struct mount *mp, int waitfor, kauth_cred_t cred)
1895{
1896 struct vnode *vp;
1897 struct ufsmount *ump = VFSTOUFS(mp);
1898 struct fs *fs;
1899 struct vnode_iterator *marker;
1900 int error, allerror = 0;
1901 bool is_suspending;
1902 struct ffs_sync_ctx ctx;
1903
1904 fs = ump->um_fs;
1905 if (fs->fs_fmod != 0 && fs->fs_ronly != 0) { /* XXX */
1906 panic("%s: rofs mod, fs=%s", __func__, fs->fs_fsmnt);
1907 }
1908
1909 fstrans_start(mp, FSTRANS_SHARED);
1910 is_suspending = (fstrans_getstate(mp) == FSTRANS_SUSPENDING);
1911 /*
1912 * Write back each (modified) inode.
1913 */
1914 vfs_vnode_iterator_init(mp, &marker);
1915
1916 ctx.waitfor = waitfor;
1917 ctx.is_suspending = is_suspending;
1918 while ((vp = vfs_vnode_iterator_next(marker, ffs_sync_selector, &ctx)))
1919 {
1920 error = vn_lock(vp, LK_EXCLUSIVE);
1921 if (error) {
1922 vrele(vp);
1923 continue;
1924 }
1925 if (waitfor == MNT_LAZY) {
1926 error = UFS_WAPBL_BEGIN(vp->v_mount);
1927 if (!error) {
1928 error = ffs_update(vp, NULL, NULL,
1929 UPDATE_CLOSE);
1930 UFS_WAPBL_END(vp->v_mount);
1931 }
1932 } else {
1933 error = VOP_FSYNC(vp, cred, FSYNC_NOLOG |
1934 (waitfor == MNT_WAIT ? FSYNC_WAIT : 0), 0, 0);
1935 }
1936 if (error)
1937 allerror = error;
1938 vput(vp);
1939 }
1940 vfs_vnode_iterator_destroy(marker);
1941
1942 /*
1943 * Force stale file system control information to be flushed.
1944 */
1945 if (waitfor != MNT_LAZY && (ump->um_devvp->v_numoutput > 0 ||
1946 !LIST_EMPTY(&ump->um_devvp->v_dirtyblkhd))) {
1947 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
1948 if ((error = VOP_FSYNC(ump->um_devvp, cred,
1949 (waitfor == MNT_WAIT ? FSYNC_WAIT : 0) | FSYNC_NOLOG,
1950 0, 0)) != 0)
1951 allerror = error;
1952 VOP_UNLOCK(ump->um_devvp);
1953 }
1954#if defined(QUOTA) || defined(QUOTA2)
1955 qsync(mp);
1956#endif
1957 /*
1958 * Write back modified superblock.
1959 */
1960 if (fs->fs_fmod != 0) {
1961 fs->fs_fmod = 0;
1962 fs->fs_time = time_second;
1963 error = UFS_WAPBL_BEGIN(mp);
1964 if (error)
1965 allerror = error;
1966 else {
1967 if ((error = ffs_cgupdate(ump, waitfor)))
1968 allerror = error;
1969 UFS_WAPBL_END(mp);
1970 }
1971 }
1972
1973#ifdef WAPBL
1974 if (mp->mnt_wapbl) {
1975 error = wapbl_flush(mp->mnt_wapbl, 0);
1976 if (error)
1977 allerror = error;
1978 }
1979#endif
1980
1981 fstrans_done(mp);
1982 return (allerror);
1983}
1984
1985/*
1986 * Load inode from disk and initialize vnode.
1987 */
1988static int
1989ffs_init_vnode(struct ufsmount *ump, struct vnode *vp, ino_t ino)
1990{
1991 struct fs *fs;
1992 struct inode *ip;
1993 struct buf *bp;
1994 int error;
1995
1996 fs = ump->um_fs;
1997
1998 /* Read in the disk contents for the inode. */
1999 error = bread(ump->um_devvp, FFS_FSBTODB(fs, ino_to_fsba(fs, ino)),
2000 (int)fs->fs_bsize, 0, &bp);
2001 if (error)
2002 return error;
2003
2004 /* Allocate and initialize inode. */
2005 ip = pool_cache_get(ffs_inode_cache, PR_WAITOK);
2006 memset(ip, 0, sizeof(struct inode));
2007 ip->i_ump = ump;
2008 ip->i_fs = fs;
2009 ip->i_dev = ump->um_dev;
2010 ip->i_number = ino;
2011 if (ump->um_fstype == UFS1)
2012 ip->i_din.ffs1_din = pool_cache_get(ffs_dinode1_cache,
2013 PR_WAITOK);
2014 else
2015 ip->i_din.ffs2_din = pool_cache_get(ffs_dinode2_cache,
2016 PR_WAITOK);
2017 ffs_load_inode(bp, ip, fs, ino);
2018 brelse(bp, 0);
2019 ip->i_vnode = vp;
2020#if defined(QUOTA) || defined(QUOTA2)
2021 ufsquota_init(ip);
2022#endif
2023
2024 /* Initialise vnode with this inode. */
2025 vp->v_tag = VT_UFS;
2026 vp->v_op = ffs_vnodeop_p;
2027 vp->v_vflag |= VV_LOCKSWORK;
2028 vp->v_data = ip;
2029
2030 /* Initialize genfs node. */
2031 genfs_node_init(vp, &ffs_genfsops);
2032
2033 return 0;
2034}
2035
2036/*
2037 * Undo ffs_init_vnode().
2038 */
2039static void
2040ffs_deinit_vnode(struct ufsmount *ump, struct vnode *vp)
2041{
2042 struct inode *ip = VTOI(vp);
2043
2044 if (ump->um_fstype == UFS1)
2045 pool_cache_put(ffs_dinode1_cache, ip->i_din.ffs1_din);
2046 else
2047 pool_cache_put(ffs_dinode2_cache, ip->i_din.ffs2_din);
2048 pool_cache_put(ffs_inode_cache, ip);
2049
2050 genfs_node_destroy(vp);
2051 vp->v_data = NULL;
2052}
2053
2054/*
2055 * Read an inode from disk and initialize this vnode / inode pair.
2056 * Caller assures no other thread will try to load this inode.
2057 */
2058int
2059ffs_loadvnode(struct mount *mp, struct vnode *vp,
2060 const void *key, size_t key_len, const void **new_key)
2061{
2062 ino_t ino;
2063 struct fs *fs;
2064 struct inode *ip;
2065 struct ufsmount *ump;
2066 int error;
2067
2068 KASSERT(key_len == sizeof(ino));
2069 memcpy(&ino, key, key_len);
2070 ump = VFSTOUFS(mp);
2071 fs = ump->um_fs;
2072
2073 error = ffs_init_vnode(ump, vp, ino);
2074 if (error)
2075 return error;
2076
2077 ip = VTOI(vp);
2078 if (ip->i_mode == 0) {
2079 ffs_deinit_vnode(ump, vp);
2080
2081 return ENOENT;
2082 }
2083
2084 /* Initialize the vnode from the inode. */
2085 ufs_vinit(mp, ffs_specop_p, ffs_fifoop_p, &vp);
2086
2087 /* Finish inode initialization. */
2088 ip->i_devvp = ump->um_devvp;
2089 vref(ip->i_devvp);
2090
2091 /*
2092 * Ensure that uid and gid are correct. This is a temporary
2093 * fix until fsck has been changed to do the update.
2094 */
2095
2096 if (fs->fs_magic == FS_UFS1_MAGIC && /* XXX */
2097 fs->fs_old_inodefmt < FS_44INODEFMT) { /* XXX */
2098 ip->i_uid = ip->i_ffs1_ouid; /* XXX */
2099 ip->i_gid = ip->i_ffs1_ogid; /* XXX */
2100 } /* XXX */
2101 uvm_vnp_setsize(vp, ip->i_size);
2102 *new_key = &ip->i_number;
2103 return 0;
2104}
2105
2106/*
2107 * Create a new inode on disk and initialize this vnode / inode pair.
2108 */
2109int
2110ffs_newvnode(struct mount *mp, struct vnode *dvp, struct vnode *vp,
2111 struct vattr *vap, kauth_cred_t cred,
2112 size_t *key_len, const void **new_key)
2113{
2114 ino_t ino;
2115 struct fs *fs;
2116 struct inode *ip;
2117 struct timespec ts;
2118 struct ufsmount *ump;
2119 int error, mode;
2120
2121 KASSERT(dvp->v_mount == mp);
2122 KASSERT(vap->va_type != VNON);
2123
2124 *key_len = sizeof(ino);
2125 ump = VFSTOUFS(mp);
2126 fs = ump->um_fs;
2127 mode = MAKEIMODE(vap->va_type, vap->va_mode);
2128
2129 /* Allocate fresh inode. */
2130 error = ffs_valloc(dvp, mode, cred, &ino);
2131 if (error)
2132 return error;
2133
2134 /* Attach inode to vnode. */
2135 error = ffs_init_vnode(ump, vp, ino);
2136 if (error) {
2137 if (UFS_WAPBL_BEGIN(mp) == 0) {
2138 ffs_vfree(dvp, ino, mode);
2139 UFS_WAPBL_END(mp);
2140 }
2141 return error;
2142 }
2143
2144 ip = VTOI(vp);
2145 if (ip->i_mode) {
2146 panic("%s: dup alloc ino=%" PRId64 " on %s: mode %x/%x "
2147 "gen %x/%x size %" PRIx64 " blocks %" PRIx64,
2148 __func__, ino, fs->fs_fsmnt, DIP(ip, mode), ip->i_mode,
2149 DIP(ip, gen), ip->i_gen, DIP(ip, size), DIP(ip, blocks));
2150 }
2151 if (DIP(ip, size) || DIP(ip, blocks)) {
2152 printf("%s: ino=%" PRId64 " on %s: "
2153 "gen %x/%x has non zero blocks %" PRIx64 " or size %"
2154 PRIx64 "\n",
2155 __func__, ino, fs->fs_fsmnt, DIP(ip, gen), ip->i_gen,
2156 DIP(ip, blocks), DIP(ip, size));
2157 if ((ip)->i_ump->um_fstype == UFS1)
2158 panic("%s: dirty filesystem?", __func__);
2159 DIP_ASSIGN(ip, blocks, 0);
2160 DIP_ASSIGN(ip, size, 0);
2161 }
2162
2163 /* Set uid / gid. */
2164 if (cred == NOCRED || cred == FSCRED) {
2165 ip->i_gid = 0;
2166 ip->i_uid = 0;
2167 } else {
2168 ip->i_gid = VTOI(dvp)->i_gid;
2169 ip->i_uid = kauth_cred_geteuid(cred);
2170 }
2171 DIP_ASSIGN(ip, gid, ip->i_gid);
2172 DIP_ASSIGN(ip, uid, ip->i_uid);
2173
2174#if defined(QUOTA) || defined(QUOTA2)
2175 error = UFS_WAPBL_BEGIN(mp);
2176 if (error) {
2177 ffs_deinit_vnode(ump, vp);
2178
2179 return error;
2180 }
2181 error = chkiq(ip, 1, cred, 0);
2182 if (error) {
2183 ffs_vfree(dvp, ino, mode);
2184 UFS_WAPBL_END(mp);
2185 ffs_deinit_vnode(ump, vp);
2186
2187 return error;
2188 }
2189 UFS_WAPBL_END(mp);
2190#endif
2191
2192 /* Set type and finalize. */
2193 ip->i_flags = 0;
2194 DIP_ASSIGN(ip, flags, 0);
2195 ip->i_mode = mode;
2196 DIP_ASSIGN(ip, mode, mode);
2197 if (vap->va_rdev != VNOVAL) {
2198 /*
2199 * Want to be able to use this to make badblock
2200 * inodes, so don't truncate the dev number.
2201 */
2202 if (ump->um_fstype == UFS1)
2203 ip->i_ffs1_rdev = ufs_rw32(vap->va_rdev,
2204 UFS_MPNEEDSWAP(ump));
2205 else
2206 ip->i_ffs2_rdev = ufs_rw64(vap->va_rdev,
2207 UFS_MPNEEDSWAP(ump));
2208 }
2209 ufs_vinit(mp, ffs_specop_p, ffs_fifoop_p, &vp);
2210 ip->i_devvp = ump->um_devvp;
2211 vref(ip->i_devvp);
2212
2213 /* Set up a new generation number for this inode. */
2214 ip->i_gen++;
2215 DIP_ASSIGN(ip, gen, ip->i_gen);
2216 if (fs->fs_magic == FS_UFS2_MAGIC) {
2217 vfs_timestamp(&ts);
2218 ip->i_ffs2_birthtime = ts.tv_sec;
2219 ip->i_ffs2_birthnsec = ts.tv_nsec;
2220 }
2221
2222 uvm_vnp_setsize(vp, ip->i_size);
2223 *new_key = &ip->i_number;
2224 return 0;
2225}
2226
2227/*
2228 * File handle to vnode
2229 *
2230 * Have to be really careful about stale file handles:
2231 * - check that the inode number is valid
2232 * - call ffs_vget() to get the locked inode
2233 * - check for an unallocated inode (i_mode == 0)
2234 * - check that the given client host has export rights and return
2235 * those rights via. exflagsp and credanonp
2236 */
2237int
2238ffs_fhtovp(struct mount *mp, struct fid *fhp, struct vnode **vpp)
2239{
2240 struct ufid ufh;
2241 int error;
2242
2243 if (fhp->fid_len != sizeof(struct ufid))
2244 return EINVAL;
2245
2246 memcpy(&ufh, fhp, sizeof(ufh));
2247 if ((error = ffs_checkrange(mp, ufh.ufid_ino)) != 0)
2248 return error;
2249
2250 return (ufs_fhtovp(mp, &ufh, vpp));
2251}
2252
2253/*
2254 * Vnode pointer to File handle
2255 */
2256/* ARGSUSED */
2257int
2258ffs_vptofh(struct vnode *vp, struct fid *fhp, size_t *fh_size)
2259{
2260 struct inode *ip;
2261 struct ufid ufh;
2262
2263 if (*fh_size < sizeof(struct ufid)) {
2264 *fh_size = sizeof(struct ufid);
2265 return E2BIG;
2266 }
2267 ip = VTOI(vp);
2268 *fh_size = sizeof(struct ufid);
2269 memset(&ufh, 0, sizeof(ufh));
2270 ufh.ufid_len = sizeof(struct ufid);
2271 ufh.ufid_ino = ip->i_number;
2272 ufh.ufid_gen = ip->i_gen;
2273 memcpy(fhp, &ufh, sizeof(ufh));
2274 return (0);
2275}
2276
2277void
2278ffs_init(void)
2279{
2280 if (ffs_initcount++ > 0)
2281 return;
2282
2283 ffs_inode_cache = pool_cache_init(sizeof(struct inode), 0, 0, 0,
2284 "ffsino", NULL, IPL_NONE, NULL, NULL, NULL);
2285 ffs_dinode1_cache = pool_cache_init(sizeof(struct ufs1_dinode), 0, 0, 0,
2286 "ffsdino1", NULL, IPL_NONE, NULL, NULL, NULL);
2287 ffs_dinode2_cache = pool_cache_init(sizeof(struct ufs2_dinode), 0, 0, 0,
2288 "ffsdino2", NULL, IPL_NONE, NULL, NULL, NULL);
2289 ufs_init();
2290}
2291
2292void
2293ffs_reinit(void)
2294{
2295 ufs_reinit();
2296}
2297
2298void
2299ffs_done(void)
2300{
2301 if (--ffs_initcount > 0)
2302 return;
2303
2304 ufs_done();
2305 pool_cache_destroy(ffs_dinode2_cache);
2306 pool_cache_destroy(ffs_dinode1_cache);
2307 pool_cache_destroy(ffs_inode_cache);
2308}
2309
2310/*
2311 * Write a superblock and associated information back to disk.
2312 */
2313int
2314ffs_sbupdate(struct ufsmount *mp, int waitfor)
2315{
2316 struct fs *fs = mp->um_fs;
2317 struct buf *bp;
2318 int error;
2319 u_int32_t saveflag;
2320
2321 error = ffs_getblk(mp->um_devvp,
2322 fs->fs_sblockloc / DEV_BSIZE, FFS_NOBLK,
2323 fs->fs_sbsize, false, &bp);
2324 if (error)
2325 return error;
2326 saveflag = fs->fs_flags & FS_INTERNAL;
2327 fs->fs_flags &= ~FS_INTERNAL;
2328
2329 memcpy(bp->b_data, fs, fs->fs_sbsize);
2330
2331 ffs_oldfscompat_write((struct fs *)bp->b_data, mp);
2332#ifdef FFS_EI
2333 if (mp->um_flags & UFS_NEEDSWAP)
2334 ffs_sb_swap((struct fs *)bp->b_data, (struct fs *)bp->b_data);
2335#endif
2336 fs->fs_flags |= saveflag;
2337
2338 if (waitfor == MNT_WAIT)
2339 error = bwrite(bp);
2340 else
2341 bawrite(bp);
2342 return (error);
2343}
2344
2345int
2346ffs_cgupdate(struct ufsmount *mp, int waitfor)
2347{
2348 struct fs *fs = mp->um_fs;
2349 struct buf *bp;
2350 int blks;
2351 void *space;
2352 int i, size, error = 0, allerror = 0;
2353
2354 UFS_WAPBL_JLOCK_ASSERT(mp);
2355
2356 allerror = ffs_sbupdate(mp, waitfor);
2357 blks = howmany(fs->fs_cssize, fs->fs_fsize);
2358 space = fs->fs_csp;
2359 for (i = 0; i < blks; i += fs->fs_frag) {
2360 size = fs->fs_bsize;
2361 if (i + fs->fs_frag > blks)
2362 size = (blks - i) * fs->fs_fsize;
2363 error = ffs_getblk(mp->um_devvp, FFS_FSBTODB(fs, fs->fs_csaddr + i),
2364 FFS_NOBLK, size, false, &bp);
2365 if (error)
2366 break;
2367#ifdef FFS_EI
2368 if (mp->um_flags & UFS_NEEDSWAP)
2369 ffs_csum_swap((struct csum*)space,
2370 (struct csum*)bp->b_data, size);
2371 else
2372#endif
2373 memcpy(bp->b_data, space, (u_int)size);
2374 space = (char *)space + size;
2375 if (waitfor == MNT_WAIT)
2376 error = bwrite(bp);
2377 else
2378 bawrite(bp);
2379 }
2380 if (!allerror && error)
2381 allerror = error;
2382 return (allerror);
2383}
2384
2385int
2386ffs_extattrctl(struct mount *mp, int cmd, struct vnode *vp,
2387 int attrnamespace, const char *attrname)
2388{
2389#ifdef UFS_EXTATTR
2390 /*
2391 * File-backed extended attributes are only supported on UFS1.
2392 * UFS2 has native extended attributes.
2393 */
2394 if (VFSTOUFS(mp)->um_fstype == UFS1)
2395 return (ufs_extattrctl(mp, cmd, vp, attrnamespace, attrname));
2396#endif
2397 return (vfs_stdextattrctl(mp, cmd, vp, attrnamespace, attrname));
2398}
2399
2400int
2401ffs_suspendctl(struct mount *mp, int cmd)
2402{
2403 int error;
2404 struct lwp *l = curlwp;
2405
2406 switch (cmd) {
2407 case SUSPEND_SUSPEND:
2408 if ((error = fstrans_setstate(mp, FSTRANS_SUSPENDING)) != 0)
2409 return error;
2410 error = ffs_sync(mp, MNT_WAIT, l->l_proc->p_cred);
2411 if (error == 0)
2412 error = fstrans_setstate(mp, FSTRANS_SUSPENDED);
2413#ifdef WAPBL
2414 if (error == 0 && mp->mnt_wapbl)
2415 error = wapbl_flush(mp->mnt_wapbl, 1);
2416#endif
2417 if (error != 0) {
2418 (void) fstrans_setstate(mp, FSTRANS_NORMAL);
2419 return error;
2420 }
2421 return 0;
2422
2423 case SUSPEND_RESUME:
2424 return fstrans_setstate(mp, FSTRANS_NORMAL);
2425
2426 default:
2427 return EINVAL;
2428 }
2429}
2430
2431/*
2432 * Synch vnode for a mounted file system.
2433 */
2434static int
2435ffs_vfs_fsync(vnode_t *vp, int flags)
2436{
2437 int error, i, pflags;
2438#ifdef WAPBL
2439 struct mount *mp;
2440#endif
2441
2442 KASSERT(vp->v_type == VBLK);
2443 KASSERT(spec_node_getmountedfs(vp) != NULL);
2444
2445 /*
2446 * Flush all dirty data associated with the vnode.
2447 */
2448 pflags = PGO_ALLPAGES | PGO_CLEANIT;
2449 if ((flags & FSYNC_WAIT) != 0)
2450 pflags |= PGO_SYNCIO;
2451 mutex_enter(vp->v_interlock);
2452 error = VOP_PUTPAGES(vp, 0, 0, pflags);
2453 if (error)
2454 return error;
2455
2456#ifdef WAPBL
2457 mp = spec_node_getmountedfs(vp);
2458 if (mp && mp->mnt_wapbl) {
2459 /*
2460 * Don't bother writing out metadata if the syncer is
2461 * making the request. We will let the sync vnode
2462 * write it out in a single burst through a call to
2463 * VFS_SYNC().
2464 */
2465 if ((flags & (FSYNC_DATAONLY | FSYNC_LAZY | FSYNC_NOLOG)) != 0)
2466 return 0;
2467
2468 /*
2469 * Don't flush the log if the vnode being flushed
2470 * contains no dirty buffers that could be in the log.
2471 */
2472 if (!LIST_EMPTY(&vp->v_dirtyblkhd)) {
2473 error = wapbl_flush(mp->mnt_wapbl, 0);
2474 if (error)
2475 return error;
2476 }
2477
2478 if ((flags & FSYNC_WAIT) != 0) {
2479 mutex_enter(vp->v_interlock);
2480 while (vp->v_numoutput)
2481 cv_wait(&vp->v_cv, vp->v_interlock);
2482 mutex_exit(vp->v_interlock);
2483 }
2484
2485 return 0;
2486 }
2487#endif /* WAPBL */
2488
2489 error = vflushbuf(vp, flags);
2490 if (error == 0 && (flags & FSYNC_CACHE) != 0) {
2491 i = 1;
2492 (void)VOP_IOCTL(vp, DIOCCACHESYNC, &i, FWRITE,
2493 kauth_cred_get());
2494 }
2495
2496 return error;
2497}
2498