xref: /netbsd-src/sys/ufs/ffs/ffs_vfsops.c (revision af56d1fe9956bd7c616e18c1b7f025f464618471)
1 /*	$NetBSD: ffs_vfsops.c,v 1.282 2013/01/22 09:39:16 dholland 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 1.282 2013/01/22 09:39:16 dholland 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 MODULE(MODULE_CLASS_VFS, ffs, NULL);
113 
114 static int	ffs_vfs_fsync(vnode_t *, int);
115 
116 static struct sysctllog *ffs_sysctl_log;
117 
118 static kauth_listener_t ffs_snapshot_listener;
119 
120 /* how many times ffs_init() was called */
121 int ffs_initcount = 0;
122 
123 extern const struct vnodeopv_desc ffs_vnodeop_opv_desc;
124 extern const struct vnodeopv_desc ffs_specop_opv_desc;
125 extern const struct vnodeopv_desc ffs_fifoop_opv_desc;
126 
127 const struct vnodeopv_desc * const ffs_vnodeopv_descs[] = {
128 	&ffs_vnodeop_opv_desc,
129 	&ffs_specop_opv_desc,
130 	&ffs_fifoop_opv_desc,
131 	NULL,
132 };
133 
134 struct vfsops ffs_vfsops = {
135 	MOUNT_FFS,
136 	sizeof (struct ufs_args),
137 	ffs_mount,
138 	ufs_start,
139 	ffs_unmount,
140 	ufs_root,
141 	ufs_quotactl,
142 	ffs_statvfs,
143 	ffs_sync,
144 	ffs_vget,
145 	ffs_fhtovp,
146 	ffs_vptofh,
147 	ffs_init,
148 	ffs_reinit,
149 	ffs_done,
150 	ffs_mountroot,
151 	ffs_snapshot,
152 	ffs_extattrctl,
153 	ffs_suspendctl,
154 	genfs_renamelock_enter,
155 	genfs_renamelock_exit,
156 	ffs_vfs_fsync,
157 	ffs_vnodeopv_descs,
158 	0,
159 	{ NULL, NULL },
160 };
161 
162 static const struct genfs_ops ffs_genfsops = {
163 	.gop_size = ffs_gop_size,
164 	.gop_alloc = ufs_gop_alloc,
165 	.gop_write = genfs_gop_write,
166 	.gop_markupdate = ufs_gop_markupdate,
167 };
168 
169 static const struct ufs_ops ffs_ufsops = {
170 	.uo_itimes = ffs_itimes,
171 	.uo_update = ffs_update,
172 	.uo_truncate = ffs_truncate,
173 	.uo_valloc = ffs_valloc,
174 	.uo_vfree = ffs_vfree,
175 	.uo_balloc = ffs_balloc,
176 	.uo_unmark_vnode = (void (*)(vnode_t *))nullop,
177 };
178 
179 static int
180 ffs_snapshot_cb(kauth_cred_t cred, kauth_action_t action, void *cookie,
181     void *arg0, void *arg1, void *arg2, void *arg3)
182 {
183 	vnode_t *vp = arg2;
184 	int result = KAUTH_RESULT_DEFER;;
185 
186 	if (action != KAUTH_SYSTEM_FS_SNAPSHOT)
187 		return result;
188 
189 	if (VTOI(vp)->i_uid == kauth_cred_geteuid(cred))
190 		result = KAUTH_RESULT_ALLOW;
191 
192 	return result;
193 }
194 
195 static int
196 ffs_modcmd(modcmd_t cmd, void *arg)
197 {
198 	int error;
199 
200 #if 0
201 	extern int doasyncfree;
202 #endif
203 #ifdef UFS_EXTATTR
204 	extern int ufs_extattr_autocreate;
205 #endif
206 	extern int ffs_log_changeopt;
207 
208 	switch (cmd) {
209 	case MODULE_CMD_INIT:
210 		error = vfs_attach(&ffs_vfsops);
211 		if (error != 0)
212 			break;
213 
214 		sysctl_createv(&ffs_sysctl_log, 0, NULL, NULL,
215 			       CTLFLAG_PERMANENT,
216 			       CTLTYPE_NODE, "vfs", NULL,
217 			       NULL, 0, NULL, 0,
218 			       CTL_VFS, CTL_EOL);
219 		sysctl_createv(&ffs_sysctl_log, 0, NULL, NULL,
220 			       CTLFLAG_PERMANENT,
221 			       CTLTYPE_NODE, "ffs",
222 			       SYSCTL_DESCR("Berkeley Fast File System"),
223 			       NULL, 0, NULL, 0,
224 			       CTL_VFS, 1, CTL_EOL);
225 		/*
226 		 * @@@ should we even bother with these first three?
227 		 */
228 		sysctl_createv(&ffs_sysctl_log, 0, NULL, NULL,
229 			       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
230 			       CTLTYPE_INT, "doclusterread", NULL,
231 			       sysctl_notavail, 0, NULL, 0,
232 			       CTL_VFS, 1, FFS_CLUSTERREAD, CTL_EOL);
233 		sysctl_createv(&ffs_sysctl_log, 0, NULL, NULL,
234 			       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
235 			       CTLTYPE_INT, "doclusterwrite", NULL,
236 			       sysctl_notavail, 0, NULL, 0,
237 			       CTL_VFS, 1, FFS_CLUSTERWRITE, CTL_EOL);
238 		sysctl_createv(&ffs_sysctl_log, 0, NULL, NULL,
239 			       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
240 			       CTLTYPE_INT, "doreallocblks", NULL,
241 			       sysctl_notavail, 0, NULL, 0,
242 			       CTL_VFS, 1, FFS_REALLOCBLKS, CTL_EOL);
243 #if 0
244 		sysctl_createv(&ffs_sysctl_log, 0, NULL, NULL,
245 			       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
246 			       CTLTYPE_INT, "doasyncfree",
247 			       SYSCTL_DESCR("Release dirty blocks asynchronously"),
248 			       NULL, 0, &doasyncfree, 0,
249 			       CTL_VFS, 1, FFS_ASYNCFREE, CTL_EOL);
250 #endif
251 		sysctl_createv(&ffs_sysctl_log, 0, NULL, NULL,
252 			       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
253 			       CTLTYPE_INT, "log_changeopt",
254 			       SYSCTL_DESCR("Log changes in optimization strategy"),
255 			       NULL, 0, &ffs_log_changeopt, 0,
256 			       CTL_VFS, 1, FFS_LOG_CHANGEOPT, CTL_EOL);
257 #ifdef UFS_EXTATTR
258 		sysctl_createv(&ffs_sysctl_log, 0, NULL, NULL,
259 			       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
260 			       CTLTYPE_INT, "extattr_autocreate",
261 			       SYSCTL_DESCR("Size of attribute for "
262 					    "backing file autocreation"),
263 			       NULL, 0, &ufs_extattr_autocreate, 0,
264 			       CTL_VFS, 1, FFS_EXTATTR_AUTOCREATE, CTL_EOL);
265 
266 #endif /* UFS_EXTATTR */
267 
268 		ffs_snapshot_listener = kauth_listen_scope(KAUTH_SCOPE_SYSTEM,
269 		    ffs_snapshot_cb, NULL);
270 		if (ffs_snapshot_listener == NULL)
271 			printf("ffs_modcmd: can't listen on system scope.\n");
272 
273 		break;
274 	case MODULE_CMD_FINI:
275 		error = vfs_detach(&ffs_vfsops);
276 		if (error != 0)
277 			break;
278 		sysctl_teardown(&ffs_sysctl_log);
279 		if (ffs_snapshot_listener != NULL)
280 			kauth_unlisten_scope(ffs_snapshot_listener);
281 		break;
282 	default:
283 		error = ENOTTY;
284 		break;
285 	}
286 
287 	return (error);
288 }
289 
290 pool_cache_t ffs_inode_cache;
291 pool_cache_t ffs_dinode1_cache;
292 pool_cache_t ffs_dinode2_cache;
293 
294 static void ffs_oldfscompat_read(struct fs *, struct ufsmount *, daddr_t);
295 static void ffs_oldfscompat_write(struct fs *, struct ufsmount *);
296 
297 /*
298  * Called by main() when ffs is going to be mounted as root.
299  */
300 
301 int
302 ffs_mountroot(void)
303 {
304 	struct fs *fs;
305 	struct mount *mp;
306 	struct lwp *l = curlwp;			/* XXX */
307 	struct ufsmount *ump;
308 	int error;
309 
310 	if (device_class(root_device) != DV_DISK)
311 		return (ENODEV);
312 
313 	if ((error = vfs_rootmountalloc(MOUNT_FFS, "root_device", &mp))) {
314 		vrele(rootvp);
315 		return (error);
316 	}
317 
318 	/*
319 	 * We always need to be able to mount the root file system.
320 	 */
321 	mp->mnt_flag |= MNT_FORCE;
322 	if ((error = ffs_mountfs(rootvp, mp, l)) != 0) {
323 		vfs_unbusy(mp, false, NULL);
324 		vfs_destroy(mp);
325 		return (error);
326 	}
327 	mp->mnt_flag &= ~MNT_FORCE;
328 	mutex_enter(&mountlist_lock);
329 	CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list);
330 	mutex_exit(&mountlist_lock);
331 	ump = VFSTOUFS(mp);
332 	fs = ump->um_fs;
333 	memset(fs->fs_fsmnt, 0, sizeof(fs->fs_fsmnt));
334 	(void)copystr(mp->mnt_stat.f_mntonname, fs->fs_fsmnt, MNAMELEN - 1, 0);
335 	(void)ffs_statvfs(mp, &mp->mnt_stat);
336 	vfs_unbusy(mp, false, NULL);
337 	setrootfstime((time_t)fs->fs_time);
338 	return (0);
339 }
340 
341 /*
342  * VFS Operations.
343  *
344  * mount system call
345  */
346 int
347 ffs_mount(struct mount *mp, const char *path, void *data, size_t *data_len)
348 {
349 	struct lwp *l = curlwp;
350 	struct vnode *devvp = NULL;
351 	struct ufs_args *args = data;
352 	struct ufsmount *ump = NULL;
353 	struct fs *fs;
354 	int error = 0, flags, update;
355 	mode_t accessmode;
356 
357 	if (*data_len < sizeof *args)
358 		return EINVAL;
359 
360 	if (mp->mnt_flag & MNT_GETARGS) {
361 		ump = VFSTOUFS(mp);
362 		if (ump == NULL)
363 			return EIO;
364 		args->fspec = NULL;
365 		*data_len = sizeof *args;
366 		return 0;
367 	}
368 
369 	update = mp->mnt_flag & MNT_UPDATE;
370 
371 	/* Check arguments */
372 	if (args->fspec != NULL) {
373 		/*
374 		 * Look up the name and verify that it's sane.
375 		 */
376 		error = namei_simple_user(args->fspec,
377 					NSM_FOLLOW_NOEMULROOT, &devvp);
378 		if (error != 0)
379 			return (error);
380 
381 		if (!update) {
382 			/*
383 			 * Be sure this is a valid block device
384 			 */
385 			if (devvp->v_type != VBLK)
386 				error = ENOTBLK;
387 			else if (bdevsw_lookup(devvp->v_rdev) == NULL)
388 				error = ENXIO;
389 		} else {
390 			/*
391 			 * Be sure we're still naming the same device
392 			 * used for our initial mount
393 			 */
394 			ump = VFSTOUFS(mp);
395 			if (devvp != ump->um_devvp) {
396 				if (devvp->v_rdev != ump->um_devvp->v_rdev)
397 					error = EINVAL;
398 				else {
399 					vrele(devvp);
400 					devvp = ump->um_devvp;
401 					vref(devvp);
402 				}
403 			}
404 		}
405 	} else {
406 		if (!update) {
407 			/* New mounts must have a filename for the device */
408 			return (EINVAL);
409 		} else {
410 			/* Use the extant mount */
411 			ump = VFSTOUFS(mp);
412 			devvp = ump->um_devvp;
413 			vref(devvp);
414 		}
415 	}
416 
417 	/*
418 	 * If mount by non-root, then verify that user has necessary
419 	 * permissions on the device.
420 	 *
421 	 * Permission to update a mount is checked higher, so here we presume
422 	 * updating the mount is okay (for example, as far as securelevel goes)
423 	 * which leaves us with the normal check.
424 	 */
425 	if (error == 0) {
426 		accessmode = VREAD;
427 		if (update ?
428 		    (mp->mnt_iflag & IMNT_WANTRDWR) != 0 :
429 		    (mp->mnt_flag & MNT_RDONLY) == 0)
430 			accessmode |= VWRITE;
431 		vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
432 		error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT,
433 		    KAUTH_REQ_SYSTEM_MOUNT_DEVICE, mp, devvp,
434 		    KAUTH_ARG(accessmode));
435 		VOP_UNLOCK(devvp);
436 	}
437 
438 	if (error) {
439 		vrele(devvp);
440 		return (error);
441 	}
442 
443 #ifdef WAPBL
444 	/* WAPBL can only be enabled on a r/w mount. */
445 	if ((mp->mnt_flag & MNT_RDONLY) && !(mp->mnt_iflag & IMNT_WANTRDWR)) {
446 		mp->mnt_flag &= ~MNT_LOG;
447 	}
448 #else /* !WAPBL */
449 	mp->mnt_flag &= ~MNT_LOG;
450 #endif /* !WAPBL */
451 
452 	if (!update) {
453 		int xflags;
454 
455 		if (mp->mnt_flag & MNT_RDONLY)
456 			xflags = FREAD;
457 		else
458 			xflags = FREAD | FWRITE;
459 		vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
460 		error = VOP_OPEN(devvp, xflags, FSCRED);
461 		VOP_UNLOCK(devvp);
462 		if (error)
463 			goto fail;
464 		error = ffs_mountfs(devvp, mp, l);
465 		if (error) {
466 			vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
467 			(void)VOP_CLOSE(devvp, xflags, NOCRED);
468 			VOP_UNLOCK(devvp);
469 			goto fail;
470 		}
471 
472 		ump = VFSTOUFS(mp);
473 		fs = ump->um_fs;
474 	} else {
475 		/*
476 		 * Update the mount.
477 		 */
478 
479 		/*
480 		 * The initial mount got a reference on this
481 		 * device, so drop the one obtained via
482 		 * namei(), above.
483 		 */
484 		vrele(devvp);
485 
486 		ump = VFSTOUFS(mp);
487 		fs = ump->um_fs;
488 		if (fs->fs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) {
489 			/*
490 			 * Changing from r/w to r/o
491 			 */
492 			flags = WRITECLOSE;
493 			if (mp->mnt_flag & MNT_FORCE)
494 				flags |= FORCECLOSE;
495 			error = ffs_flushfiles(mp, flags, l);
496 			if (error == 0)
497 				error = UFS_WAPBL_BEGIN(mp);
498 			if (error == 0 &&
499 			    ffs_cgupdate(ump, MNT_WAIT) == 0 &&
500 			    fs->fs_clean & FS_WASCLEAN) {
501 				if (mp->mnt_flag & MNT_SOFTDEP)
502 					fs->fs_flags &= ~FS_DOSOFTDEP;
503 				fs->fs_clean = FS_ISCLEAN;
504 				(void) ffs_sbupdate(ump, MNT_WAIT);
505 			}
506 			if (error == 0)
507 				UFS_WAPBL_END(mp);
508 			if (error)
509 				return (error);
510 		}
511 
512 #ifdef WAPBL
513 		if ((mp->mnt_flag & MNT_LOG) == 0) {
514 			error = ffs_wapbl_stop(mp, mp->mnt_flag & MNT_FORCE);
515 			if (error)
516 				return error;
517 		}
518 #endif /* WAPBL */
519 
520 		if (fs->fs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) {
521 			/*
522 			 * Finish change from r/w to r/o
523 			 */
524 			fs->fs_ronly = 1;
525 			fs->fs_fmod = 0;
526 		}
527 
528 		if (mp->mnt_flag & MNT_RELOAD) {
529 			error = ffs_reload(mp, l->l_cred, l);
530 			if (error)
531 				return (error);
532 		}
533 
534 		if (fs->fs_ronly && (mp->mnt_iflag & IMNT_WANTRDWR)) {
535 			/*
536 			 * Changing from read-only to read/write
537 			 */
538 #ifndef QUOTA2
539 			if (fs->fs_flags & FS_DOQUOTA2) {
540 				ump->um_flags |= UFS_QUOTA2;
541 				uprintf("%s: options QUOTA2 not enabled%s\n",
542 				    mp->mnt_stat.f_mntonname,
543 				    (mp->mnt_flag & MNT_FORCE) ? "" :
544 				    ", not mounting");
545 				return EINVAL;
546 			}
547 #endif
548 			fs->fs_ronly = 0;
549 			fs->fs_clean <<= 1;
550 			fs->fs_fmod = 1;
551 #ifdef WAPBL
552 			if (fs->fs_flags & FS_DOWAPBL) {
553 				printf("%s: replaying log to disk\n",
554 				    mp->mnt_stat.f_mntonname);
555 				KDASSERT(mp->mnt_wapbl_replay);
556 				error = wapbl_replay_write(mp->mnt_wapbl_replay,
557 							   devvp);
558 				if (error) {
559 					return error;
560 				}
561 				wapbl_replay_stop(mp->mnt_wapbl_replay);
562 				fs->fs_clean = FS_WASCLEAN;
563 			}
564 #endif /* WAPBL */
565 			if (fs->fs_snapinum[0] != 0)
566 				ffs_snapshot_mount(mp);
567 		}
568 
569 #ifdef WAPBL
570 		error = ffs_wapbl_start(mp);
571 		if (error)
572 			return error;
573 #endif /* WAPBL */
574 
575 #ifdef QUOTA2
576 		if (!fs->fs_ronly) {
577 			error = ffs_quota2_mount(mp);
578 			if (error) {
579 				return error;
580 			}
581 		}
582 #endif
583 
584 		if ((mp->mnt_flag & MNT_DISCARD) && !(ump->um_discarddata))
585 			ump->um_discarddata = ffs_discard_init(devvp, fs);
586 
587 		if (args->fspec == NULL)
588 			return 0;
589 	}
590 
591 	error = set_statvfs_info(path, UIO_USERSPACE, args->fspec,
592 	    UIO_USERSPACE, mp->mnt_op->vfs_name, mp, l);
593 	if (error == 0)
594 		(void)strncpy(fs->fs_fsmnt, mp->mnt_stat.f_mntonname,
595 		    sizeof(fs->fs_fsmnt));
596 	fs->fs_flags &= ~FS_DOSOFTDEP;
597 	if (fs->fs_fmod != 0) {	/* XXX */
598 		int err;
599 
600 		fs->fs_fmod = 0;
601 		if (fs->fs_clean & FS_WASCLEAN)
602 			fs->fs_time = time_second;
603 		else {
604 			printf("%s: file system not clean (fs_clean=%#x); "
605 			    "please fsck(8)\n", mp->mnt_stat.f_mntfromname,
606 			    fs->fs_clean);
607 			printf("%s: lost blocks %" PRId64 " files %d\n",
608 			    mp->mnt_stat.f_mntfromname, fs->fs_pendingblocks,
609 			    fs->fs_pendinginodes);
610 		}
611 		err = UFS_WAPBL_BEGIN(mp);
612 		if (err == 0) {
613 			(void) ffs_cgupdate(ump, MNT_WAIT);
614 			UFS_WAPBL_END(mp);
615 		}
616 	}
617 	if ((mp->mnt_flag & MNT_SOFTDEP) != 0) {
618 		printf("%s: `-o softdep' is no longer supported, "
619 		    "consider `-o log'\n", mp->mnt_stat.f_mntfromname);
620 		mp->mnt_flag &= ~MNT_SOFTDEP;
621 	}
622 
623 	return (error);
624 
625 fail:
626 	vrele(devvp);
627 	return (error);
628 }
629 
630 /*
631  * Reload all incore data for a filesystem (used after running fsck on
632  * the root filesystem and finding things to fix). The filesystem must
633  * be mounted read-only.
634  *
635  * Things to do to update the mount:
636  *	1) invalidate all cached meta-data.
637  *	2) re-read superblock from disk.
638  *	3) re-read summary information from disk.
639  *	4) invalidate all inactive vnodes.
640  *	5) invalidate all cached file data.
641  *	6) re-read inode data for all active vnodes.
642  */
643 int
644 ffs_reload(struct mount *mp, kauth_cred_t cred, struct lwp *l)
645 {
646 	struct vnode *vp, *mvp, *devvp;
647 	struct inode *ip;
648 	void *space;
649 	struct buf *bp;
650 	struct fs *fs, *newfs;
651 	struct dkwedge_info dkw;
652 	int i, bsize, blks, error;
653 	int32_t *lp;
654 	struct ufsmount *ump;
655 	daddr_t sblockloc;
656 
657 	if ((mp->mnt_flag & MNT_RDONLY) == 0)
658 		return (EINVAL);
659 
660 	ump = VFSTOUFS(mp);
661 	/*
662 	 * Step 1: invalidate all cached meta-data.
663 	 */
664 	devvp = ump->um_devvp;
665 	vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
666 	error = vinvalbuf(devvp, 0, cred, l, 0, 0);
667 	VOP_UNLOCK(devvp);
668 	if (error)
669 		panic("ffs_reload: dirty1");
670 	/*
671 	 * Step 2: re-read superblock from disk.
672 	 */
673 	fs = ump->um_fs;
674 
675 	/* XXX we don't handle possibility that superblock moved. */
676 	error = bread(devvp, fs->fs_sblockloc / DEV_BSIZE, fs->fs_sbsize,
677 		      NOCRED, 0, &bp);
678 	if (error) {
679 		return (error);
680 	}
681 	newfs = kmem_alloc(fs->fs_sbsize, KM_SLEEP);
682 	memcpy(newfs, bp->b_data, fs->fs_sbsize);
683 #ifdef FFS_EI
684 	if (ump->um_flags & UFS_NEEDSWAP) {
685 		ffs_sb_swap((struct fs*)bp->b_data, newfs);
686 		fs->fs_flags |= FS_SWAPPED;
687 	} else
688 #endif
689 		fs->fs_flags &= ~FS_SWAPPED;
690 	if ((newfs->fs_magic != FS_UFS1_MAGIC &&
691 	     newfs->fs_magic != FS_UFS2_MAGIC)||
692 	     newfs->fs_bsize > MAXBSIZE ||
693 	     newfs->fs_bsize < sizeof(struct fs)) {
694 		brelse(bp, 0);
695 		kmem_free(newfs, fs->fs_sbsize);
696 		return (EIO);		/* XXX needs translation */
697 	}
698 	/* Store off old fs_sblockloc for fs_oldfscompat_read. */
699 	sblockloc = fs->fs_sblockloc;
700 	/*
701 	 * Copy pointer fields back into superblock before copying in	XXX
702 	 * new superblock. These should really be in the ufsmount.	XXX
703 	 * Note that important parameters (eg fs_ncg) are unchanged.
704 	 */
705 	newfs->fs_csp = fs->fs_csp;
706 	newfs->fs_maxcluster = fs->fs_maxcluster;
707 	newfs->fs_contigdirs = fs->fs_contigdirs;
708 	newfs->fs_ronly = fs->fs_ronly;
709 	newfs->fs_active = fs->fs_active;
710 	memcpy(fs, newfs, (u_int)fs->fs_sbsize);
711 	brelse(bp, 0);
712 	kmem_free(newfs, fs->fs_sbsize);
713 
714 	/* Recheck for apple UFS filesystem */
715 	ump->um_flags &= ~UFS_ISAPPLEUFS;
716 	/* First check to see if this is tagged as an Apple UFS filesystem
717 	 * in the disklabel
718 	 */
719 	if (getdiskinfo(devvp, &dkw) == 0 &&
720 	    strcmp(dkw.dkw_ptype, DKW_PTYPE_APPLEUFS) == 0)
721 		ump->um_flags |= UFS_ISAPPLEUFS;
722 #ifdef APPLE_UFS
723 	else {
724 		/* Manually look for an apple ufs label, and if a valid one
725 		 * is found, then treat it like an Apple UFS filesystem anyway
726 		 *
727 		 * EINVAL is most probably a blocksize or alignment problem,
728 		 * it is unlikely that this is an Apple UFS filesystem then.
729 		 */
730 		error = bread(devvp, (daddr_t)(APPLEUFS_LABEL_OFFSET / DEV_BSIZE),
731 			APPLEUFS_LABEL_SIZE, cred, 0, &bp);
732 		if (error && error != EINVAL) {
733 			return (error);
734 		}
735 		if (error == 0) {
736 			error = ffs_appleufs_validate(fs->fs_fsmnt,
737 				(struct appleufslabel *)bp->b_data, NULL);
738 			if (error == 0)
739 				ump->um_flags |= UFS_ISAPPLEUFS;
740 			brelse(bp, 0);
741 		}
742 		bp = NULL;
743 	}
744 #else
745 	if (ump->um_flags & UFS_ISAPPLEUFS)
746 		return (EIO);
747 #endif
748 
749 	if (UFS_MPISAPPLEUFS(ump)) {
750 		/* see comment about NeXT below */
751 		ump->um_maxsymlinklen = APPLEUFS_MAXSYMLINKLEN;
752 		ump->um_dirblksiz = APPLEUFS_DIRBLKSIZ;
753 		mp->mnt_iflag |= IMNT_DTYPE;
754 	} else {
755 		ump->um_maxsymlinklen = fs->fs_maxsymlinklen;
756 		ump->um_dirblksiz = DIRBLKSIZ;
757 		if (ump->um_maxsymlinklen > 0)
758 			mp->mnt_iflag |= IMNT_DTYPE;
759 		else
760 			mp->mnt_iflag &= ~IMNT_DTYPE;
761 	}
762 	ffs_oldfscompat_read(fs, ump, sblockloc);
763 
764 	mutex_enter(&ump->um_lock);
765 	ump->um_maxfilesize = fs->fs_maxfilesize;
766 	if (fs->fs_flags & ~(FS_KNOWN_FLAGS | FS_INTERNAL)) {
767 		uprintf("%s: unknown ufs flags: 0x%08"PRIx32"%s\n",
768 		    mp->mnt_stat.f_mntonname, fs->fs_flags,
769 		    (mp->mnt_flag & MNT_FORCE) ? "" : ", not mounting");
770 		if ((mp->mnt_flag & MNT_FORCE) == 0) {
771 			mutex_exit(&ump->um_lock);
772 			return (EINVAL);
773 		}
774 	}
775 	if (fs->fs_pendingblocks != 0 || fs->fs_pendinginodes != 0) {
776 		fs->fs_pendingblocks = 0;
777 		fs->fs_pendinginodes = 0;
778 	}
779 	mutex_exit(&ump->um_lock);
780 
781 	ffs_statvfs(mp, &mp->mnt_stat);
782 	/*
783 	 * Step 3: re-read summary information from disk.
784 	 */
785 	blks = howmany(fs->fs_cssize, fs->fs_fsize);
786 	space = fs->fs_csp;
787 	for (i = 0; i < blks; i += fs->fs_frag) {
788 		bsize = fs->fs_bsize;
789 		if (i + fs->fs_frag > blks)
790 			bsize = (blks - i) * fs->fs_fsize;
791 		error = bread(devvp, fsbtodb(fs, fs->fs_csaddr + i), bsize,
792 			      NOCRED, 0, &bp);
793 		if (error) {
794 			return (error);
795 		}
796 #ifdef FFS_EI
797 		if (UFS_FSNEEDSWAP(fs))
798 			ffs_csum_swap((struct csum *)bp->b_data,
799 			    (struct csum *)space, bsize);
800 		else
801 #endif
802 			memcpy(space, bp->b_data, (size_t)bsize);
803 		space = (char *)space + bsize;
804 		brelse(bp, 0);
805 	}
806 	if (fs->fs_snapinum[0] != 0)
807 		ffs_snapshot_mount(mp);
808 	/*
809 	 * We no longer know anything about clusters per cylinder group.
810 	 */
811 	if (fs->fs_contigsumsize > 0) {
812 		lp = fs->fs_maxcluster;
813 		for (i = 0; i < fs->fs_ncg; i++)
814 			*lp++ = fs->fs_contigsumsize;
815 	}
816 
817 	/* Allocate a marker vnode. */
818 	mvp = vnalloc(mp);
819 	/*
820 	 * NOTE: not using the TAILQ_FOREACH here since in this loop vgone()
821 	 * and vclean() can be called indirectly
822 	 */
823 	mutex_enter(&mntvnode_lock);
824  loop:
825 	for (vp = TAILQ_FIRST(&mp->mnt_vnodelist); vp; vp = vunmark(mvp)) {
826 		vmark(mvp, vp);
827 		if (vp->v_mount != mp || vismarker(vp))
828 			continue;
829 		/*
830 		 * Step 4: invalidate all inactive vnodes.
831 		 */
832 		if (vrecycle(vp, &mntvnode_lock, l)) {
833 			mutex_enter(&mntvnode_lock);
834 			(void)vunmark(mvp);
835 			goto loop;
836 		}
837 		/*
838 		 * Step 5: invalidate all cached file data.
839 		 */
840 		mutex_enter(vp->v_interlock);
841 		mutex_exit(&mntvnode_lock);
842 		if (vget(vp, LK_EXCLUSIVE)) {
843 			(void)vunmark(mvp);
844 			goto loop;
845 		}
846 		if (vinvalbuf(vp, 0, cred, l, 0, 0))
847 			panic("ffs_reload: dirty2");
848 		/*
849 		 * Step 6: re-read inode data for all active vnodes.
850 		 */
851 		ip = VTOI(vp);
852 		error = bread(devvp, fsbtodb(fs, ino_to_fsba(fs, ip->i_number)),
853 			      (int)fs->fs_bsize, NOCRED, 0, &bp);
854 		if (error) {
855 			vput(vp);
856 			(void)vunmark(mvp);
857 			break;
858 		}
859 		ffs_load_inode(bp, ip, fs, ip->i_number);
860 		brelse(bp, 0);
861 		vput(vp);
862 		mutex_enter(&mntvnode_lock);
863 	}
864 	mutex_exit(&mntvnode_lock);
865 	vnfree(mvp);
866 	return (error);
867 }
868 
869 /*
870  * Possible superblock locations ordered from most to least likely.
871  */
872 static const int sblock_try[] = SBLOCKSEARCH;
873 
874 /*
875  * Common code for mount and mountroot
876  */
877 int
878 ffs_mountfs(struct vnode *devvp, struct mount *mp, struct lwp *l)
879 {
880 	struct ufsmount *ump;
881 	struct buf *bp;
882 	struct fs *fs;
883 	dev_t dev;
884 	struct dkwedge_info dkw;
885 	void *space;
886 	daddr_t sblockloc, fsblockloc;
887 	int blks, fstype;
888 	int error, i, bsize, ronly, bset = 0;
889 #ifdef FFS_EI
890 	int needswap = 0;		/* keep gcc happy */
891 #endif
892 	int32_t *lp;
893 	kauth_cred_t cred;
894 	u_int32_t sbsize = 8192;	/* keep gcc happy*/
895 	u_int32_t allocsbsize;
896 	int32_t fsbsize;
897 
898 	dev = devvp->v_rdev;
899 	cred = l ? l->l_cred : NOCRED;
900 
901 	/* Flush out any old buffers remaining from a previous use. */
902 	vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
903 	error = vinvalbuf(devvp, V_SAVE, cred, l, 0, 0);
904 	VOP_UNLOCK(devvp);
905 	if (error)
906 		return (error);
907 
908 	ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
909 
910 	bp = NULL;
911 	ump = NULL;
912 	fs = NULL;
913 	sblockloc = 0;
914 	fstype = 0;
915 
916 	error = fstrans_mount(mp);
917 	if (error)
918 		return error;
919 
920 	ump = kmem_zalloc(sizeof(*ump), KM_SLEEP);
921 	mutex_init(&ump->um_lock, MUTEX_DEFAULT, IPL_NONE);
922 	error = ffs_snapshot_init(ump);
923 	if (error)
924 		goto out;
925 	ump->um_ops = &ffs_ufsops;
926 
927 #ifdef WAPBL
928  sbagain:
929 #endif
930 	/*
931 	 * Try reading the superblock in each of its possible locations.
932 	 */
933 	for (i = 0; ; i++) {
934 		if (bp != NULL) {
935 			brelse(bp, BC_NOCACHE);
936 			bp = NULL;
937 		}
938 		if (sblock_try[i] == -1) {
939 			error = EINVAL;
940 			fs = NULL;
941 			goto out;
942 		}
943 		error = bread(devvp, sblock_try[i] / DEV_BSIZE, SBLOCKSIZE, cred,
944 			      0, &bp);
945 		if (error) {
946 			fs = NULL;
947 			goto out;
948 		}
949 		fs = (struct fs*)bp->b_data;
950 		fsblockloc = sblockloc = sblock_try[i];
951 		if (fs->fs_magic == FS_UFS1_MAGIC) {
952 			sbsize = fs->fs_sbsize;
953 			fstype = UFS1;
954 			fsbsize = fs->fs_bsize;
955 #ifdef FFS_EI
956 			needswap = 0;
957 		} else if (fs->fs_magic == FS_UFS1_MAGIC_SWAPPED) {
958 			sbsize = bswap32(fs->fs_sbsize);
959 			fstype = UFS1;
960 			fsbsize = bswap32(fs->fs_bsize);
961 			needswap = 1;
962 #endif
963 		} else if (fs->fs_magic == FS_UFS2_MAGIC) {
964 			sbsize = fs->fs_sbsize;
965 			fstype = UFS2;
966 			fsbsize = fs->fs_bsize;
967 #ifdef FFS_EI
968 			needswap = 0;
969 		} else if (fs->fs_magic == FS_UFS2_MAGIC_SWAPPED) {
970 			sbsize = bswap32(fs->fs_sbsize);
971 			fstype = UFS2;
972 			fsbsize = bswap32(fs->fs_bsize);
973 			needswap = 1;
974 #endif
975 		} else
976 			continue;
977 
978 
979 		/* fs->fs_sblockloc isn't defined for old filesystems */
980 		if (fstype == UFS1 && !(fs->fs_old_flags & FS_FLAGS_UPDATED)) {
981 			if (sblockloc == SBLOCK_UFS2)
982 				/*
983 				 * This is likely to be the first alternate
984 				 * in a filesystem with 64k blocks.
985 				 * Don't use it.
986 				 */
987 				continue;
988 			fsblockloc = sblockloc;
989 		} else {
990 			fsblockloc = fs->fs_sblockloc;
991 #ifdef FFS_EI
992 			if (needswap)
993 				fsblockloc = bswap64(fsblockloc);
994 #endif
995 		}
996 
997 		/* Check we haven't found an alternate superblock */
998 		if (fsblockloc != sblockloc)
999 			continue;
1000 
1001 		/* Validate size of superblock */
1002 		if (sbsize > MAXBSIZE || sbsize < sizeof(struct fs))
1003 			continue;
1004 
1005 		/* Check that we can handle the file system blocksize */
1006 		if (fsbsize > MAXBSIZE) {
1007 			printf("ffs_mountfs: block size (%d) > MAXBSIZE (%d)\n",
1008 			    fsbsize, MAXBSIZE);
1009 			continue;
1010 		}
1011 
1012 		/* Ok seems to be a good superblock */
1013 		break;
1014 	}
1015 
1016 	fs = kmem_alloc((u_long)sbsize, KM_SLEEP);
1017 	memcpy(fs, bp->b_data, sbsize);
1018 	ump->um_fs = fs;
1019 
1020 #ifdef FFS_EI
1021 	if (needswap) {
1022 		ffs_sb_swap((struct fs*)bp->b_data, fs);
1023 		fs->fs_flags |= FS_SWAPPED;
1024 	} else
1025 #endif
1026 		fs->fs_flags &= ~FS_SWAPPED;
1027 
1028 #ifdef WAPBL
1029 	if ((mp->mnt_wapbl_replay == 0) && (fs->fs_flags & FS_DOWAPBL)) {
1030 		error = ffs_wapbl_replay_start(mp, fs, devvp);
1031 		if (error && (mp->mnt_flag & MNT_FORCE) == 0)
1032 			goto out;
1033 		if (!error) {
1034 			if (!ronly) {
1035 				/* XXX fsmnt may be stale. */
1036 				printf("%s: replaying log to disk\n",
1037 				    fs->fs_fsmnt);
1038 				error = wapbl_replay_write(mp->mnt_wapbl_replay,
1039 				    devvp);
1040 				if (error)
1041 					goto out;
1042 				wapbl_replay_stop(mp->mnt_wapbl_replay);
1043 				fs->fs_clean = FS_WASCLEAN;
1044 			} else {
1045 				/* XXX fsmnt may be stale */
1046 				printf("%s: replaying log to memory\n",
1047 				    fs->fs_fsmnt);
1048 			}
1049 
1050 			/* Force a re-read of the superblock */
1051 			brelse(bp, BC_INVAL);
1052 			bp = NULL;
1053 			kmem_free(fs, sbsize);
1054 			fs = NULL;
1055 			goto sbagain;
1056 		}
1057 	}
1058 #else /* !WAPBL */
1059 	if ((fs->fs_flags & FS_DOWAPBL) && (mp->mnt_flag & MNT_FORCE) == 0) {
1060 		error = EPERM;
1061 		goto out;
1062 	}
1063 #endif /* !WAPBL */
1064 
1065 	ffs_oldfscompat_read(fs, ump, sblockloc);
1066 	ump->um_maxfilesize = fs->fs_maxfilesize;
1067 
1068 	if (fs->fs_flags & ~(FS_KNOWN_FLAGS | FS_INTERNAL)) {
1069 		uprintf("%s: unknown ufs flags: 0x%08"PRIx32"%s\n",
1070 		    mp->mnt_stat.f_mntonname, fs->fs_flags,
1071 		    (mp->mnt_flag & MNT_FORCE) ? "" : ", not mounting");
1072 		if ((mp->mnt_flag & MNT_FORCE) == 0) {
1073 			error = EINVAL;
1074 			goto out;
1075 		}
1076 	}
1077 
1078 	if (fs->fs_pendingblocks != 0 || fs->fs_pendinginodes != 0) {
1079 		fs->fs_pendingblocks = 0;
1080 		fs->fs_pendinginodes = 0;
1081 	}
1082 
1083 	ump->um_fstype = fstype;
1084 	if (fs->fs_sbsize < SBLOCKSIZE)
1085 		brelse(bp, BC_INVAL);
1086 	else
1087 		brelse(bp, 0);
1088 	bp = NULL;
1089 
1090 	/* First check to see if this is tagged as an Apple UFS filesystem
1091 	 * in the disklabel
1092 	 */
1093 	if (getdiskinfo(devvp, &dkw) == 0 &&
1094 	    strcmp(dkw.dkw_ptype, DKW_PTYPE_APPLEUFS) == 0)
1095 		ump->um_flags |= UFS_ISAPPLEUFS;
1096 #ifdef APPLE_UFS
1097 	else {
1098 		/* Manually look for an apple ufs label, and if a valid one
1099 		 * is found, then treat it like an Apple UFS filesystem anyway
1100 		 */
1101 		error = bread(devvp, (daddr_t)(APPLEUFS_LABEL_OFFSET / DEV_BSIZE),
1102 			APPLEUFS_LABEL_SIZE, cred, 0, &bp);
1103 		if (error)
1104 			goto out;
1105 		error = ffs_appleufs_validate(fs->fs_fsmnt,
1106 			(struct appleufslabel *)bp->b_data, NULL);
1107 		if (error == 0) {
1108 			ump->um_flags |= UFS_ISAPPLEUFS;
1109 		}
1110 		brelse(bp, 0);
1111 		bp = NULL;
1112 	}
1113 #else
1114 	if (ump->um_flags & UFS_ISAPPLEUFS) {
1115 		error = EINVAL;
1116 		goto out;
1117 	}
1118 #endif
1119 
1120 #if 0
1121 /*
1122  * XXX This code changes the behaviour of mounting dirty filesystems, to
1123  * XXX require "mount -f ..." to mount them.  This doesn't match what
1124  * XXX mount(8) describes and is disabled for now.
1125  */
1126 	/*
1127 	 * If the file system is not clean, don't allow it to be mounted
1128 	 * unless MNT_FORCE is specified.  (Note: MNT_FORCE is always set
1129 	 * for the root file system.)
1130 	 */
1131 	if (fs->fs_flags & FS_DOWAPBL) {
1132 		/*
1133 		 * wapbl normally expects to be FS_WASCLEAN when the FS_DOWAPBL
1134 		 * bit is set, although there's a window in unmount where it
1135 		 * could be FS_ISCLEAN
1136 		 */
1137 		if ((mp->mnt_flag & MNT_FORCE) == 0 &&
1138 		    (fs->fs_clean & (FS_WASCLEAN | FS_ISCLEAN)) == 0) {
1139 			error = EPERM;
1140 			goto out;
1141 		}
1142 	} else
1143 		if ((fs->fs_clean & FS_ISCLEAN) == 0 &&
1144 		    (mp->mnt_flag & MNT_FORCE) == 0) {
1145 			error = EPERM;
1146 			goto out;
1147 		}
1148 #endif
1149 
1150 	/*
1151 	 * verify that we can access the last block in the fs
1152 	 * if we're mounting read/write.
1153 	 */
1154 
1155 	if (!ronly) {
1156 		error = bread(devvp, fsbtodb(fs, fs->fs_size - 1), fs->fs_fsize,
1157 		    cred, 0, &bp);
1158 		if (bp->b_bcount != fs->fs_fsize)
1159 			error = EINVAL;
1160 		if (error) {
1161 			bset = BC_INVAL;
1162 			goto out;
1163 		}
1164 		brelse(bp, BC_INVAL);
1165 		bp = NULL;
1166 	}
1167 
1168 	fs->fs_ronly = ronly;
1169 	/* Don't bump fs_clean if we're replaying journal */
1170 	if (!((fs->fs_flags & FS_DOWAPBL) && (fs->fs_clean & FS_WASCLEAN)))
1171 		if (ronly == 0) {
1172 			fs->fs_clean <<= 1;
1173 			fs->fs_fmod = 1;
1174 		}
1175 	bsize = fs->fs_cssize;
1176 	blks = howmany(bsize, fs->fs_fsize);
1177 	if (fs->fs_contigsumsize > 0)
1178 		bsize += fs->fs_ncg * sizeof(int32_t);
1179 	bsize += fs->fs_ncg * sizeof(*fs->fs_contigdirs);
1180 	allocsbsize = bsize;
1181 	space = kmem_alloc((u_long)allocsbsize, KM_SLEEP);
1182 	fs->fs_csp = space;
1183 	for (i = 0; i < blks; i += fs->fs_frag) {
1184 		bsize = fs->fs_bsize;
1185 		if (i + fs->fs_frag > blks)
1186 			bsize = (blks - i) * fs->fs_fsize;
1187 		error = bread(devvp, fsbtodb(fs, fs->fs_csaddr + i), bsize,
1188 			      cred, 0, &bp);
1189 		if (error) {
1190 			kmem_free(fs->fs_csp, allocsbsize);
1191 			goto out;
1192 		}
1193 #ifdef FFS_EI
1194 		if (needswap)
1195 			ffs_csum_swap((struct csum *)bp->b_data,
1196 				(struct csum *)space, bsize);
1197 		else
1198 #endif
1199 			memcpy(space, bp->b_data, (u_int)bsize);
1200 
1201 		space = (char *)space + bsize;
1202 		brelse(bp, 0);
1203 		bp = NULL;
1204 	}
1205 	if (fs->fs_contigsumsize > 0) {
1206 		fs->fs_maxcluster = lp = space;
1207 		for (i = 0; i < fs->fs_ncg; i++)
1208 			*lp++ = fs->fs_contigsumsize;
1209 		space = lp;
1210 	}
1211 	bsize = fs->fs_ncg * sizeof(*fs->fs_contigdirs);
1212 	fs->fs_contigdirs = space;
1213 	space = (char *)space + bsize;
1214 	memset(fs->fs_contigdirs, 0, bsize);
1215 		/* Compatibility for old filesystems - XXX */
1216 	if (fs->fs_avgfilesize <= 0)
1217 		fs->fs_avgfilesize = AVFILESIZ;
1218 	if (fs->fs_avgfpdir <= 0)
1219 		fs->fs_avgfpdir = AFPDIR;
1220 	fs->fs_active = NULL;
1221 	mp->mnt_data = ump;
1222 	mp->mnt_stat.f_fsidx.__fsid_val[0] = (long)dev;
1223 	mp->mnt_stat.f_fsidx.__fsid_val[1] = makefstype(MOUNT_FFS);
1224 	mp->mnt_stat.f_fsid = mp->mnt_stat.f_fsidx.__fsid_val[0];
1225 	mp->mnt_stat.f_namemax = FFS_MAXNAMLEN;
1226 	if (UFS_MPISAPPLEUFS(ump)) {
1227 		/* NeXT used to keep short symlinks in the inode even
1228 		 * when using FS_42INODEFMT.  In that case fs->fs_maxsymlinklen
1229 		 * is probably -1, but we still need to be able to identify
1230 		 * short symlinks.
1231 		 */
1232 		ump->um_maxsymlinklen = APPLEUFS_MAXSYMLINKLEN;
1233 		ump->um_dirblksiz = APPLEUFS_DIRBLKSIZ;
1234 		mp->mnt_iflag |= IMNT_DTYPE;
1235 	} else {
1236 		ump->um_maxsymlinklen = fs->fs_maxsymlinklen;
1237 		ump->um_dirblksiz = DIRBLKSIZ;
1238 		if (ump->um_maxsymlinklen > 0)
1239 			mp->mnt_iflag |= IMNT_DTYPE;
1240 		else
1241 			mp->mnt_iflag &= ~IMNT_DTYPE;
1242 	}
1243 	mp->mnt_fs_bshift = fs->fs_bshift;
1244 	mp->mnt_dev_bshift = DEV_BSHIFT;	/* XXX */
1245 	mp->mnt_flag |= MNT_LOCAL;
1246 	mp->mnt_iflag |= IMNT_MPSAFE;
1247 #ifdef FFS_EI
1248 	if (needswap)
1249 		ump->um_flags |= UFS_NEEDSWAP;
1250 #endif
1251 	ump->um_mountp = mp;
1252 	ump->um_dev = dev;
1253 	ump->um_devvp = devvp;
1254 	ump->um_nindir = fs->fs_nindir;
1255 	ump->um_lognindir = ffs(fs->fs_nindir) - 1;
1256 	ump->um_bptrtodb = fs->fs_fshift - DEV_BSHIFT;
1257 	ump->um_seqinc = fs->fs_frag;
1258 	for (i = 0; i < MAXQUOTAS; i++)
1259 		ump->um_quotas[i] = NULLVP;
1260 	devvp->v_specmountpoint = mp;
1261 	if (ronly == 0 && fs->fs_snapinum[0] != 0)
1262 		ffs_snapshot_mount(mp);
1263 #ifdef WAPBL
1264 	if (!ronly) {
1265 		KDASSERT(fs->fs_ronly == 0);
1266 		/*
1267 		 * ffs_wapbl_start() needs mp->mnt_stat initialised if it
1268 		 * needs to create a new log file in-filesystem.
1269 		 */
1270 		ffs_statvfs(mp, &mp->mnt_stat);
1271 
1272 		error = ffs_wapbl_start(mp);
1273 		if (error) {
1274 			kmem_free(fs->fs_csp, allocsbsize);
1275 			goto out;
1276 		}
1277 	}
1278 #endif /* WAPBL */
1279 	if (ronly == 0) {
1280 #ifdef QUOTA2
1281 		error = ffs_quota2_mount(mp);
1282 		if (error) {
1283 			kmem_free(fs->fs_csp, allocsbsize);
1284 			goto out;
1285 		}
1286 #else
1287 		if (fs->fs_flags & FS_DOQUOTA2) {
1288 			ump->um_flags |= UFS_QUOTA2;
1289 			uprintf("%s: options QUOTA2 not enabled%s\n",
1290 			    mp->mnt_stat.f_mntonname,
1291 			    (mp->mnt_flag & MNT_FORCE) ? "" : ", not mounting");
1292 			if ((mp->mnt_flag & MNT_FORCE) == 0) {
1293 				error = EINVAL;
1294 				kmem_free(fs->fs_csp, allocsbsize);
1295 				goto out;
1296 			}
1297 		}
1298 #endif
1299 	 }
1300 #ifdef UFS_EXTATTR
1301 	/*
1302 	 * Initialize file-backed extended attributes on UFS1 file
1303 	 * systems.
1304 	 */
1305 	if (ump->um_fstype == UFS1)
1306 		ufs_extattr_uepm_init(&ump->um_extattr);
1307 #endif /* UFS_EXTATTR */
1308 
1309 	if (mp->mnt_flag & MNT_DISCARD)
1310 		ump->um_discarddata = ffs_discard_init(devvp, fs);
1311 
1312 	return (0);
1313 out:
1314 #ifdef WAPBL
1315 	if (mp->mnt_wapbl_replay) {
1316 		wapbl_replay_stop(mp->mnt_wapbl_replay);
1317 		wapbl_replay_free(mp->mnt_wapbl_replay);
1318 		mp->mnt_wapbl_replay = 0;
1319 	}
1320 #endif
1321 
1322 	fstrans_unmount(mp);
1323 	if (fs)
1324 		kmem_free(fs, fs->fs_sbsize);
1325 	devvp->v_specmountpoint = NULL;
1326 	if (bp)
1327 		brelse(bp, bset);
1328 	if (ump) {
1329 		if (ump->um_oldfscompat)
1330 			kmem_free(ump->um_oldfscompat, 512 + 3*sizeof(int32_t));
1331 		mutex_destroy(&ump->um_lock);
1332 		kmem_free(ump, sizeof(*ump));
1333 		mp->mnt_data = NULL;
1334 	}
1335 	return (error);
1336 }
1337 
1338 /*
1339  * Sanity checks for loading old filesystem superblocks.
1340  * See ffs_oldfscompat_write below for unwound actions.
1341  *
1342  * XXX - Parts get retired eventually.
1343  * Unfortunately new bits get added.
1344  */
1345 static void
1346 ffs_oldfscompat_read(struct fs *fs, struct ufsmount *ump, daddr_t sblockloc)
1347 {
1348 	off_t maxfilesize;
1349 	int32_t *extrasave;
1350 
1351 	if ((fs->fs_magic != FS_UFS1_MAGIC) ||
1352 	    (fs->fs_old_flags & FS_FLAGS_UPDATED))
1353 		return;
1354 
1355 	if (!ump->um_oldfscompat)
1356 		ump->um_oldfscompat = kmem_alloc(512 + 3*sizeof(int32_t),
1357 		    KM_SLEEP);
1358 
1359 	memcpy(ump->um_oldfscompat, &fs->fs_old_postbl_start, 512);
1360 	extrasave = ump->um_oldfscompat;
1361 	extrasave += 512/sizeof(int32_t);
1362 	extrasave[0] = fs->fs_old_npsect;
1363 	extrasave[1] = fs->fs_old_interleave;
1364 	extrasave[2] = fs->fs_old_trackskew;
1365 
1366 	/* These fields will be overwritten by their
1367 	 * original values in fs_oldfscompat_write, so it is harmless
1368 	 * to modify them here.
1369 	 */
1370 	fs->fs_cstotal.cs_ndir = fs->fs_old_cstotal.cs_ndir;
1371 	fs->fs_cstotal.cs_nbfree = fs->fs_old_cstotal.cs_nbfree;
1372 	fs->fs_cstotal.cs_nifree = fs->fs_old_cstotal.cs_nifree;
1373 	fs->fs_cstotal.cs_nffree = fs->fs_old_cstotal.cs_nffree;
1374 
1375 	fs->fs_maxbsize = fs->fs_bsize;
1376 	fs->fs_time = fs->fs_old_time;
1377 	fs->fs_size = fs->fs_old_size;
1378 	fs->fs_dsize = fs->fs_old_dsize;
1379 	fs->fs_csaddr = fs->fs_old_csaddr;
1380 	fs->fs_sblockloc = sblockloc;
1381 
1382 	fs->fs_flags = fs->fs_old_flags | (fs->fs_flags & FS_INTERNAL);
1383 
1384 	if (fs->fs_old_postblformat == FS_42POSTBLFMT) {
1385 		fs->fs_old_nrpos = 8;
1386 		fs->fs_old_npsect = fs->fs_old_nsect;
1387 		fs->fs_old_interleave = 1;
1388 		fs->fs_old_trackskew = 0;
1389 	}
1390 
1391 	if (fs->fs_old_inodefmt < FS_44INODEFMT) {
1392 		fs->fs_maxfilesize = (u_quad_t) 1LL << 39;
1393 		fs->fs_qbmask = ~fs->fs_bmask;
1394 		fs->fs_qfmask = ~fs->fs_fmask;
1395 	}
1396 
1397 	maxfilesize = (u_int64_t)0x80000000 * fs->fs_bsize - 1;
1398 	if (fs->fs_maxfilesize > maxfilesize)
1399 		fs->fs_maxfilesize = maxfilesize;
1400 
1401 	/* Compatibility for old filesystems */
1402 	if (fs->fs_avgfilesize <= 0)
1403 		fs->fs_avgfilesize = AVFILESIZ;
1404 	if (fs->fs_avgfpdir <= 0)
1405 		fs->fs_avgfpdir = AFPDIR;
1406 
1407 #if 0
1408 	if (bigcgs) {
1409 		fs->fs_save_cgsize = fs->fs_cgsize;
1410 		fs->fs_cgsize = fs->fs_bsize;
1411 	}
1412 #endif
1413 }
1414 
1415 /*
1416  * Unwinding superblock updates for old filesystems.
1417  * See ffs_oldfscompat_read above for details.
1418  *
1419  * XXX - Parts get retired eventually.
1420  * Unfortunately new bits get added.
1421  */
1422 static void
1423 ffs_oldfscompat_write(struct fs *fs, struct ufsmount *ump)
1424 {
1425 	int32_t *extrasave;
1426 
1427 	if ((fs->fs_magic != FS_UFS1_MAGIC) ||
1428 	    (fs->fs_old_flags & FS_FLAGS_UPDATED))
1429 		return;
1430 
1431 	fs->fs_old_time = fs->fs_time;
1432 	fs->fs_old_cstotal.cs_ndir = fs->fs_cstotal.cs_ndir;
1433 	fs->fs_old_cstotal.cs_nbfree = fs->fs_cstotal.cs_nbfree;
1434 	fs->fs_old_cstotal.cs_nifree = fs->fs_cstotal.cs_nifree;
1435 	fs->fs_old_cstotal.cs_nffree = fs->fs_cstotal.cs_nffree;
1436 	fs->fs_old_flags = fs->fs_flags;
1437 
1438 #if 0
1439 	if (bigcgs) {
1440 		fs->fs_cgsize = fs->fs_save_cgsize;
1441 	}
1442 #endif
1443 
1444 	memcpy(&fs->fs_old_postbl_start, ump->um_oldfscompat, 512);
1445 	extrasave = ump->um_oldfscompat;
1446 	extrasave += 512/sizeof(int32_t);
1447 	fs->fs_old_npsect = extrasave[0];
1448 	fs->fs_old_interleave = extrasave[1];
1449 	fs->fs_old_trackskew = extrasave[2];
1450 
1451 }
1452 
1453 /*
1454  * unmount vfs operation
1455  */
1456 int
1457 ffs_unmount(struct mount *mp, int mntflags)
1458 {
1459 	struct lwp *l = curlwp;
1460 	struct ufsmount *ump = VFSTOUFS(mp);
1461 	struct fs *fs = ump->um_fs;
1462 	int error, flags;
1463 	u_int32_t bsize;
1464 #ifdef WAPBL
1465 	extern int doforce;
1466 #endif
1467 
1468 	if (ump->um_discarddata) {
1469 		ffs_discard_finish(ump->um_discarddata, mntflags);
1470 		ump->um_discarddata = NULL;
1471 	}
1472 
1473 	flags = 0;
1474 	if (mntflags & MNT_FORCE)
1475 		flags |= FORCECLOSE;
1476 	if ((error = ffs_flushfiles(mp, flags, l)) != 0)
1477 		return (error);
1478 	error = UFS_WAPBL_BEGIN(mp);
1479 	if (error == 0)
1480 		if (fs->fs_ronly == 0 &&
1481 		    ffs_cgupdate(ump, MNT_WAIT) == 0 &&
1482 		    fs->fs_clean & FS_WASCLEAN) {
1483 			fs->fs_clean = FS_ISCLEAN;
1484 			fs->fs_fmod = 0;
1485 			(void) ffs_sbupdate(ump, MNT_WAIT);
1486 		}
1487 	if (error == 0)
1488 		UFS_WAPBL_END(mp);
1489 #ifdef WAPBL
1490 	KASSERT(!(mp->mnt_wapbl_replay && mp->mnt_wapbl));
1491 	if (mp->mnt_wapbl_replay) {
1492 		KDASSERT(fs->fs_ronly);
1493 		wapbl_replay_stop(mp->mnt_wapbl_replay);
1494 		wapbl_replay_free(mp->mnt_wapbl_replay);
1495 		mp->mnt_wapbl_replay = 0;
1496 	}
1497 	error = ffs_wapbl_stop(mp, doforce && (mntflags & MNT_FORCE));
1498 	if (error) {
1499 		return error;
1500 	}
1501 #endif /* WAPBL */
1502 
1503 	if (ump->um_devvp->v_type != VBAD)
1504 		ump->um_devvp->v_specmountpoint = NULL;
1505 	vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
1506 	(void)VOP_CLOSE(ump->um_devvp, fs->fs_ronly ? FREAD : FREAD | FWRITE,
1507 		NOCRED);
1508 	vput(ump->um_devvp);
1509 
1510 	bsize = fs->fs_cssize;
1511 	if (fs->fs_contigsumsize > 0)
1512 		bsize += fs->fs_ncg * sizeof(int32_t);
1513 	bsize += fs->fs_ncg * sizeof(*fs->fs_contigdirs);
1514 	kmem_free(fs->fs_csp, bsize);
1515 
1516 	kmem_free(fs, fs->fs_sbsize);
1517 	if (ump->um_oldfscompat != NULL)
1518 		kmem_free(ump->um_oldfscompat, 512 + 3*sizeof(int32_t));
1519 	mutex_destroy(&ump->um_lock);
1520 	ffs_snapshot_fini(ump);
1521 	kmem_free(ump, sizeof(*ump));
1522 	mp->mnt_data = NULL;
1523 	mp->mnt_flag &= ~MNT_LOCAL;
1524 	fstrans_unmount(mp);
1525 	return (0);
1526 }
1527 
1528 /*
1529  * Flush out all the files in a filesystem.
1530  */
1531 int
1532 ffs_flushfiles(struct mount *mp, int flags, struct lwp *l)
1533 {
1534 	extern int doforce;
1535 	struct ufsmount *ump;
1536 	int error;
1537 
1538 	if (!doforce)
1539 		flags &= ~FORCECLOSE;
1540 	ump = VFSTOUFS(mp);
1541 #ifdef QUOTA
1542 	if ((error = quota1_umount(mp, flags)) != 0)
1543 		return (error);
1544 #endif
1545 #ifdef QUOTA2
1546 	if ((error = quota2_umount(mp, flags)) != 0)
1547 		return (error);
1548 #endif
1549 #ifdef UFS_EXTATTR
1550 	if (ump->um_fstype == UFS1) {
1551 		if (ump->um_extattr.uepm_flags & UFS_EXTATTR_UEPM_STARTED)
1552 			ufs_extattr_stop(mp, l);
1553 		if (ump->um_extattr.uepm_flags & UFS_EXTATTR_UEPM_INITIALIZED)
1554 			ufs_extattr_uepm_destroy(&ump->um_extattr);
1555 	}
1556 #endif
1557 	if ((error = vflush(mp, 0, SKIPSYSTEM | flags)) != 0)
1558 		return (error);
1559 	ffs_snapshot_unmount(mp);
1560 	/*
1561 	 * Flush all the files.
1562 	 */
1563 	error = vflush(mp, NULLVP, flags);
1564 	if (error)
1565 		return (error);
1566 	/*
1567 	 * Flush filesystem metadata.
1568 	 */
1569 	vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
1570 	error = VOP_FSYNC(ump->um_devvp, l->l_cred, FSYNC_WAIT, 0, 0);
1571 	VOP_UNLOCK(ump->um_devvp);
1572 	if (flags & FORCECLOSE) /* XXXDBJ */
1573 		error = 0;
1574 
1575 #ifdef WAPBL
1576 	if (error)
1577 		return error;
1578 	if (mp->mnt_wapbl) {
1579 		error = wapbl_flush(mp->mnt_wapbl, 1);
1580 		if (flags & FORCECLOSE)
1581 			error = 0;
1582 	}
1583 #endif
1584 
1585 	return (error);
1586 }
1587 
1588 /*
1589  * Get file system statistics.
1590  */
1591 int
1592 ffs_statvfs(struct mount *mp, struct statvfs *sbp)
1593 {
1594 	struct ufsmount *ump;
1595 	struct fs *fs;
1596 
1597 	ump = VFSTOUFS(mp);
1598 	fs = ump->um_fs;
1599 	mutex_enter(&ump->um_lock);
1600 	sbp->f_bsize = fs->fs_bsize;
1601 	sbp->f_frsize = fs->fs_fsize;
1602 	sbp->f_iosize = fs->fs_bsize;
1603 	sbp->f_blocks = fs->fs_dsize;
1604 	sbp->f_bfree = blkstofrags(fs, fs->fs_cstotal.cs_nbfree) +
1605 	    fs->fs_cstotal.cs_nffree + dbtofsb(fs, fs->fs_pendingblocks);
1606 	sbp->f_bresvd = ((u_int64_t) fs->fs_dsize * (u_int64_t)
1607 	    fs->fs_minfree) / (u_int64_t) 100;
1608 	if (sbp->f_bfree > sbp->f_bresvd)
1609 		sbp->f_bavail = sbp->f_bfree - sbp->f_bresvd;
1610 	else
1611 		sbp->f_bavail = 0;
1612 	sbp->f_files =  fs->fs_ncg * fs->fs_ipg - UFS_ROOTINO;
1613 	sbp->f_ffree = fs->fs_cstotal.cs_nifree + fs->fs_pendinginodes;
1614 	sbp->f_favail = sbp->f_ffree;
1615 	sbp->f_fresvd = 0;
1616 	mutex_exit(&ump->um_lock);
1617 	copy_statvfs_info(sbp, mp);
1618 
1619 	return (0);
1620 }
1621 
1622 /*
1623  * Go through the disk queues to initiate sandbagged IO;
1624  * go through the inodes to write those that have been modified;
1625  * initiate the writing of the super block if it has been modified.
1626  *
1627  * Note: we are always called with the filesystem marked `MPBUSY'.
1628  */
1629 int
1630 ffs_sync(struct mount *mp, int waitfor, kauth_cred_t cred)
1631 {
1632 	struct vnode *vp, *mvp, *nvp;
1633 	struct inode *ip;
1634 	struct ufsmount *ump = VFSTOUFS(mp);
1635 	struct fs *fs;
1636 	int error, allerror = 0;
1637 	bool is_suspending;
1638 
1639 	fs = ump->um_fs;
1640 	if (fs->fs_fmod != 0 && fs->fs_ronly != 0) {		/* XXX */
1641 		printf("fs = %s\n", fs->fs_fsmnt);
1642 		panic("update: rofs mod");
1643 	}
1644 
1645 	/* Allocate a marker vnode. */
1646 	mvp = vnalloc(mp);
1647 
1648 	fstrans_start(mp, FSTRANS_SHARED);
1649 	is_suspending = (fstrans_getstate(mp) == FSTRANS_SUSPENDING);
1650 	/*
1651 	 * Write back each (modified) inode.
1652 	 */
1653 	mutex_enter(&mntvnode_lock);
1654 loop:
1655 	/*
1656 	 * NOTE: not using the TAILQ_FOREACH here since in this loop vgone()
1657 	 * and vclean() can be called indirectly
1658 	 */
1659 	for (vp = TAILQ_FIRST(&mp->mnt_vnodelist); vp; vp = nvp) {
1660 		nvp = TAILQ_NEXT(vp, v_mntvnodes);
1661 		/*
1662 		 * If the vnode that we are about to sync is no longer
1663 		 * associated with this mount point, start over.
1664 		 */
1665 		if (vp->v_mount != mp)
1666 			goto loop;
1667 		/*
1668 		 * Don't interfere with concurrent scans of this FS.
1669 		 */
1670 		if (vismarker(vp))
1671 			continue;
1672 		mutex_enter(vp->v_interlock);
1673 		ip = VTOI(vp);
1674 
1675 		/*
1676 		 * Skip the vnode/inode if inaccessible.
1677 		 */
1678 		if (ip == NULL || (vp->v_iflag & (VI_XLOCK | VI_CLEAN)) != 0 ||
1679 		    vp->v_type == VNON) {
1680 			mutex_exit(vp->v_interlock);
1681 			continue;
1682 		}
1683 
1684 		/*
1685 		 * We deliberately update inode times here.  This will
1686 		 * prevent a massive queue of updates accumulating, only
1687 		 * to be handled by a call to unmount.
1688 		 *
1689 		 * XXX It would be better to have the syncer trickle these
1690 		 * out.  Adjustment needed to allow registering vnodes for
1691 		 * sync when the vnode is clean, but the inode dirty.  Or
1692 		 * have ufs itself trickle out inode updates.
1693 		 *
1694 		 * If doing a lazy sync, we don't care about metadata or
1695 		 * data updates, because they are handled by each vnode's
1696 		 * synclist entry.  In this case we are only interested in
1697 		 * writing back modified inodes.
1698 		 */
1699 		if ((ip->i_flag & (IN_ACCESS | IN_CHANGE | IN_UPDATE |
1700 		    IN_MODIFY | IN_MODIFIED | IN_ACCESSED)) == 0 &&
1701 		    (waitfor == MNT_LAZY || (LIST_EMPTY(&vp->v_dirtyblkhd) &&
1702 		    UVM_OBJ_IS_CLEAN(&vp->v_uobj)))) {
1703 			mutex_exit(vp->v_interlock);
1704 			continue;
1705 		}
1706 		if (vp->v_type == VBLK && is_suspending) {
1707 			mutex_exit(vp->v_interlock);
1708 			continue;
1709 		}
1710 		vmark(mvp, vp);
1711 		mutex_exit(&mntvnode_lock);
1712 		error = vget(vp, LK_EXCLUSIVE | LK_NOWAIT);
1713 		if (error) {
1714 			mutex_enter(&mntvnode_lock);
1715 			nvp = vunmark(mvp);
1716 			if (error == ENOENT) {
1717 				goto loop;
1718 			}
1719 			continue;
1720 		}
1721 		if (waitfor == MNT_LAZY) {
1722 			error = UFS_WAPBL_BEGIN(vp->v_mount);
1723 			if (!error) {
1724 				error = ffs_update(vp, NULL, NULL,
1725 				    UPDATE_CLOSE);
1726 				UFS_WAPBL_END(vp->v_mount);
1727 			}
1728 		} else {
1729 			error = VOP_FSYNC(vp, cred, FSYNC_NOLOG |
1730 			    (waitfor == MNT_WAIT ? FSYNC_WAIT : 0), 0, 0);
1731 		}
1732 		if (error)
1733 			allerror = error;
1734 		vput(vp);
1735 		mutex_enter(&mntvnode_lock);
1736 		nvp = vunmark(mvp);
1737 	}
1738 	mutex_exit(&mntvnode_lock);
1739 	/*
1740 	 * Force stale file system control information to be flushed.
1741 	 */
1742 	if (waitfor != MNT_LAZY && (ump->um_devvp->v_numoutput > 0 ||
1743 	    !LIST_EMPTY(&ump->um_devvp->v_dirtyblkhd))) {
1744 		vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
1745 		if ((error = VOP_FSYNC(ump->um_devvp, cred,
1746 		    (waitfor == MNT_WAIT ? FSYNC_WAIT : 0) | FSYNC_NOLOG,
1747 		    0, 0)) != 0)
1748 			allerror = error;
1749 		VOP_UNLOCK(ump->um_devvp);
1750 		if (allerror == 0 && waitfor == MNT_WAIT && !mp->mnt_wapbl) {
1751 			mutex_enter(&mntvnode_lock);
1752 			goto loop;
1753 		}
1754 	}
1755 #if defined(QUOTA) || defined(QUOTA2)
1756 	qsync(mp);
1757 #endif
1758 	/*
1759 	 * Write back modified superblock.
1760 	 */
1761 	if (fs->fs_fmod != 0) {
1762 		fs->fs_fmod = 0;
1763 		fs->fs_time = time_second;
1764 		error = UFS_WAPBL_BEGIN(mp);
1765 		if (error)
1766 			allerror = error;
1767 		else {
1768 			if ((error = ffs_cgupdate(ump, waitfor)))
1769 				allerror = error;
1770 			UFS_WAPBL_END(mp);
1771 		}
1772 	}
1773 
1774 #ifdef WAPBL
1775 	if (mp->mnt_wapbl) {
1776 		error = wapbl_flush(mp->mnt_wapbl, 0);
1777 		if (error)
1778 			allerror = error;
1779 	}
1780 #endif
1781 
1782 	fstrans_done(mp);
1783 	vnfree(mvp);
1784 	return (allerror);
1785 }
1786 
1787 /*
1788  * Look up a FFS dinode number to find its incore vnode, otherwise read it
1789  * in from disk.  If it is in core, wait for the lock bit to clear, then
1790  * return the inode locked.  Detection and handling of mount points must be
1791  * done by the calling routine.
1792  */
1793 int
1794 ffs_vget(struct mount *mp, ino_t ino, struct vnode **vpp)
1795 {
1796 	struct fs *fs;
1797 	struct inode *ip;
1798 	struct ufsmount *ump;
1799 	struct buf *bp;
1800 	struct vnode *vp;
1801 	dev_t dev;
1802 	int error;
1803 
1804 	ump = VFSTOUFS(mp);
1805 	dev = ump->um_dev;
1806 
1807  retry:
1808 	if ((*vpp = ufs_ihashget(dev, ino, LK_EXCLUSIVE)) != NULL)
1809 		return (0);
1810 
1811 	/* Allocate a new vnode/inode. */
1812 	error = getnewvnode(VT_UFS, mp, ffs_vnodeop_p, NULL, &vp);
1813 	if (error) {
1814 		*vpp = NULL;
1815 		return (error);
1816 	}
1817 	ip = pool_cache_get(ffs_inode_cache, PR_WAITOK);
1818 
1819 	/*
1820 	 * If someone beat us to it, put back the freshly allocated
1821 	 * vnode/inode pair and retry.
1822 	 */
1823 	mutex_enter(&ufs_hashlock);
1824 	if (ufs_ihashget(dev, ino, 0) != NULL) {
1825 		mutex_exit(&ufs_hashlock);
1826 		ungetnewvnode(vp);
1827 		pool_cache_put(ffs_inode_cache, ip);
1828 		goto retry;
1829 	}
1830 
1831 	vp->v_vflag |= VV_LOCKSWORK;
1832 
1833 	/*
1834 	 * XXX MFS ends up here, too, to allocate an inode.  Should we
1835 	 * XXX create another pool for MFS inodes?
1836 	 */
1837 
1838 	memset(ip, 0, sizeof(struct inode));
1839 	vp->v_data = ip;
1840 	ip->i_vnode = vp;
1841 	ip->i_ump = ump;
1842 	ip->i_fs = fs = ump->um_fs;
1843 	ip->i_dev = dev;
1844 	ip->i_number = ino;
1845 #if defined(QUOTA) || defined(QUOTA2)
1846 	ufsquota_init(ip);
1847 #endif
1848 
1849 	/*
1850 	 * Initialize genfs node, we might proceed to destroy it in
1851 	 * error branches.
1852 	 */
1853 	genfs_node_init(vp, &ffs_genfsops);
1854 
1855 	/*
1856 	 * Put it onto its hash chain and lock it so that other requests for
1857 	 * this inode will block if they arrive while we are sleeping waiting
1858 	 * for old data structures to be purged or for the contents of the
1859 	 * disk portion of this inode to be read.
1860 	 */
1861 
1862 	ufs_ihashins(ip);
1863 	mutex_exit(&ufs_hashlock);
1864 
1865 	/* Read in the disk contents for the inode, copy into the inode. */
1866 	error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)),
1867 		      (int)fs->fs_bsize, NOCRED, 0, &bp);
1868 	if (error) {
1869 
1870 		/*
1871 		 * The inode does not contain anything useful, so it would
1872 		 * be misleading to leave it on its hash chain. With mode
1873 		 * still zero, it will be unlinked and returned to the free
1874 		 * list by vput().
1875 		 */
1876 
1877 		vput(vp);
1878 		*vpp = NULL;
1879 		return (error);
1880 	}
1881 	if (ip->i_ump->um_fstype == UFS1)
1882 		ip->i_din.ffs1_din = pool_cache_get(ffs_dinode1_cache,
1883 		    PR_WAITOK);
1884 	else
1885 		ip->i_din.ffs2_din = pool_cache_get(ffs_dinode2_cache,
1886 		    PR_WAITOK);
1887 	ffs_load_inode(bp, ip, fs, ino);
1888 	brelse(bp, 0);
1889 
1890 	/*
1891 	 * Initialize the vnode from the inode, check for aliases.
1892 	 * Note that the underlying vnode may have changed.
1893 	 */
1894 
1895 	ufs_vinit(mp, ffs_specop_p, ffs_fifoop_p, &vp);
1896 
1897 	/*
1898 	 * Finish inode initialization now that aliasing has been resolved.
1899 	 */
1900 
1901 	ip->i_devvp = ump->um_devvp;
1902 	vref(ip->i_devvp);
1903 
1904 	/*
1905 	 * Ensure that uid and gid are correct. This is a temporary
1906 	 * fix until fsck has been changed to do the update.
1907 	 */
1908 
1909 	if (fs->fs_old_inodefmt < FS_44INODEFMT) {		/* XXX */
1910 		ip->i_uid = ip->i_ffs1_ouid;			/* XXX */
1911 		ip->i_gid = ip->i_ffs1_ogid;			/* XXX */
1912 	}							/* XXX */
1913 	uvm_vnp_setsize(vp, ip->i_size);
1914 	*vpp = vp;
1915 	return (0);
1916 }
1917 
1918 /*
1919  * File handle to vnode
1920  *
1921  * Have to be really careful about stale file handles:
1922  * - check that the inode number is valid
1923  * - call ffs_vget() to get the locked inode
1924  * - check for an unallocated inode (i_mode == 0)
1925  * - check that the given client host has export rights and return
1926  *   those rights via. exflagsp and credanonp
1927  */
1928 int
1929 ffs_fhtovp(struct mount *mp, struct fid *fhp, struct vnode **vpp)
1930 {
1931 	struct ufid ufh;
1932 	struct fs *fs;
1933 
1934 	if (fhp->fid_len != sizeof(struct ufid))
1935 		return EINVAL;
1936 
1937 	memcpy(&ufh, fhp, sizeof(ufh));
1938 	fs = VFSTOUFS(mp)->um_fs;
1939 	if (ufh.ufid_ino < UFS_ROOTINO ||
1940 	    ufh.ufid_ino >= fs->fs_ncg * fs->fs_ipg)
1941 		return (ESTALE);
1942 	return (ufs_fhtovp(mp, &ufh, vpp));
1943 }
1944 
1945 /*
1946  * Vnode pointer to File handle
1947  */
1948 /* ARGSUSED */
1949 int
1950 ffs_vptofh(struct vnode *vp, struct fid *fhp, size_t *fh_size)
1951 {
1952 	struct inode *ip;
1953 	struct ufid ufh;
1954 
1955 	if (*fh_size < sizeof(struct ufid)) {
1956 		*fh_size = sizeof(struct ufid);
1957 		return E2BIG;
1958 	}
1959 	ip = VTOI(vp);
1960 	*fh_size = sizeof(struct ufid);
1961 	memset(&ufh, 0, sizeof(ufh));
1962 	ufh.ufid_len = sizeof(struct ufid);
1963 	ufh.ufid_ino = ip->i_number;
1964 	ufh.ufid_gen = ip->i_gen;
1965 	memcpy(fhp, &ufh, sizeof(ufh));
1966 	return (0);
1967 }
1968 
1969 void
1970 ffs_init(void)
1971 {
1972 	if (ffs_initcount++ > 0)
1973 		return;
1974 
1975 	ffs_inode_cache = pool_cache_init(sizeof(struct inode), 0, 0, 0,
1976 	    "ffsino", NULL, IPL_NONE, NULL, NULL, NULL);
1977 	ffs_dinode1_cache = pool_cache_init(sizeof(struct ufs1_dinode), 0, 0, 0,
1978 	    "ffsdino1", NULL, IPL_NONE, NULL, NULL, NULL);
1979 	ffs_dinode2_cache = pool_cache_init(sizeof(struct ufs2_dinode), 0, 0, 0,
1980 	    "ffsdino2", NULL, IPL_NONE, NULL, NULL, NULL);
1981 	ufs_init();
1982 }
1983 
1984 void
1985 ffs_reinit(void)
1986 {
1987 
1988 	ufs_reinit();
1989 }
1990 
1991 void
1992 ffs_done(void)
1993 {
1994 	if (--ffs_initcount > 0)
1995 		return;
1996 
1997 	ufs_done();
1998 	pool_cache_destroy(ffs_dinode2_cache);
1999 	pool_cache_destroy(ffs_dinode1_cache);
2000 	pool_cache_destroy(ffs_inode_cache);
2001 }
2002 
2003 /*
2004  * Write a superblock and associated information back to disk.
2005  */
2006 int
2007 ffs_sbupdate(struct ufsmount *mp, int waitfor)
2008 {
2009 	struct fs *fs = mp->um_fs;
2010 	struct buf *bp;
2011 	int error = 0;
2012 	u_int32_t saveflag;
2013 
2014 	error = ffs_getblk(mp->um_devvp,
2015 	    fs->fs_sblockloc / DEV_BSIZE, FFS_NOBLK,
2016 	    fs->fs_sbsize, false, &bp);
2017 	if (error)
2018 		return error;
2019 	saveflag = fs->fs_flags & FS_INTERNAL;
2020 	fs->fs_flags &= ~FS_INTERNAL;
2021 
2022 	memcpy(bp->b_data, fs, fs->fs_sbsize);
2023 
2024 	ffs_oldfscompat_write((struct fs *)bp->b_data, mp);
2025 #ifdef FFS_EI
2026 	if (mp->um_flags & UFS_NEEDSWAP)
2027 		ffs_sb_swap((struct fs *)bp->b_data, (struct fs *)bp->b_data);
2028 #endif
2029 	fs->fs_flags |= saveflag;
2030 
2031 	if (waitfor == MNT_WAIT)
2032 		error = bwrite(bp);
2033 	else
2034 		bawrite(bp);
2035 	return (error);
2036 }
2037 
2038 int
2039 ffs_cgupdate(struct ufsmount *mp, int waitfor)
2040 {
2041 	struct fs *fs = mp->um_fs;
2042 	struct buf *bp;
2043 	int blks;
2044 	void *space;
2045 	int i, size, error = 0, allerror = 0;
2046 
2047 	allerror = ffs_sbupdate(mp, waitfor);
2048 	blks = howmany(fs->fs_cssize, fs->fs_fsize);
2049 	space = fs->fs_csp;
2050 	for (i = 0; i < blks; i += fs->fs_frag) {
2051 		size = fs->fs_bsize;
2052 		if (i + fs->fs_frag > blks)
2053 			size = (blks - i) * fs->fs_fsize;
2054 		error = ffs_getblk(mp->um_devvp, fsbtodb(fs, fs->fs_csaddr + i),
2055 		    FFS_NOBLK, size, false, &bp);
2056 		if (error)
2057 			break;
2058 #ifdef FFS_EI
2059 		if (mp->um_flags & UFS_NEEDSWAP)
2060 			ffs_csum_swap((struct csum*)space,
2061 			    (struct csum*)bp->b_data, size);
2062 		else
2063 #endif
2064 			memcpy(bp->b_data, space, (u_int)size);
2065 		space = (char *)space + size;
2066 		if (waitfor == MNT_WAIT)
2067 			error = bwrite(bp);
2068 		else
2069 			bawrite(bp);
2070 	}
2071 	if (!allerror && error)
2072 		allerror = error;
2073 	return (allerror);
2074 }
2075 
2076 int
2077 ffs_extattrctl(struct mount *mp, int cmd, struct vnode *vp,
2078     int attrnamespace, const char *attrname)
2079 {
2080 #ifdef UFS_EXTATTR
2081 	/*
2082 	 * File-backed extended attributes are only supported on UFS1.
2083 	 * UFS2 has native extended attributes.
2084 	 */
2085 	if (VFSTOUFS(mp)->um_fstype == UFS1)
2086 		return (ufs_extattrctl(mp, cmd, vp, attrnamespace, attrname));
2087 #endif
2088 	return (vfs_stdextattrctl(mp, cmd, vp, attrnamespace, attrname));
2089 }
2090 
2091 int
2092 ffs_suspendctl(struct mount *mp, int cmd)
2093 {
2094 	int error;
2095 	struct lwp *l = curlwp;
2096 
2097 	switch (cmd) {
2098 	case SUSPEND_SUSPEND:
2099 		if ((error = fstrans_setstate(mp, FSTRANS_SUSPENDING)) != 0)
2100 			return error;
2101 		error = ffs_sync(mp, MNT_WAIT, l->l_proc->p_cred);
2102 		if (error == 0)
2103 			error = fstrans_setstate(mp, FSTRANS_SUSPENDED);
2104 #ifdef WAPBL
2105 		if (error == 0 && mp->mnt_wapbl)
2106 			error = wapbl_flush(mp->mnt_wapbl, 1);
2107 #endif
2108 		if (error != 0) {
2109 			(void) fstrans_setstate(mp, FSTRANS_NORMAL);
2110 			return error;
2111 		}
2112 		return 0;
2113 
2114 	case SUSPEND_RESUME:
2115 		return fstrans_setstate(mp, FSTRANS_NORMAL);
2116 
2117 	default:
2118 		return EINVAL;
2119 	}
2120 }
2121 
2122 /*
2123  * Synch vnode for a mounted file system.
2124  */
2125 static int
2126 ffs_vfs_fsync(vnode_t *vp, int flags)
2127 {
2128 	int error, i, pflags;
2129 #ifdef WAPBL
2130 	struct mount *mp;
2131 #endif
2132 
2133 	KASSERT(vp->v_type == VBLK);
2134 	KASSERT(vp->v_specmountpoint != NULL);
2135 
2136 	/*
2137 	 * Flush all dirty data associated with the vnode.
2138 	 */
2139 	pflags = PGO_ALLPAGES | PGO_CLEANIT;
2140 	if ((flags & FSYNC_WAIT) != 0)
2141 		pflags |= PGO_SYNCIO;
2142 	mutex_enter(vp->v_interlock);
2143 	error = VOP_PUTPAGES(vp, 0, 0, pflags);
2144 	if (error)
2145 		return error;
2146 
2147 #ifdef WAPBL
2148 	mp = vp->v_specmountpoint;
2149 	if (mp && mp->mnt_wapbl) {
2150 		/*
2151 		 * Don't bother writing out metadata if the syncer is
2152 		 * making the request.  We will let the sync vnode
2153 		 * write it out in a single burst through a call to
2154 		 * VFS_SYNC().
2155 		 */
2156 		if ((flags & (FSYNC_DATAONLY | FSYNC_LAZY | FSYNC_NOLOG)) != 0)
2157 			return 0;
2158 
2159 		/*
2160 		 * Don't flush the log if the vnode being flushed
2161 		 * contains no dirty buffers that could be in the log.
2162 		 */
2163 		if (!LIST_EMPTY(&vp->v_dirtyblkhd)) {
2164 			error = wapbl_flush(mp->mnt_wapbl, 0);
2165 			if (error)
2166 				return error;
2167 		}
2168 
2169 		if ((flags & FSYNC_WAIT) != 0) {
2170 			mutex_enter(vp->v_interlock);
2171 			while (vp->v_numoutput)
2172 				cv_wait(&vp->v_cv, vp->v_interlock);
2173 			mutex_exit(vp->v_interlock);
2174 		}
2175 
2176 		return 0;
2177 	}
2178 #endif /* WAPBL */
2179 
2180 	error = vflushbuf(vp, flags);
2181 	if (error == 0 && (flags & FSYNC_CACHE) != 0) {
2182 		i = 1;
2183 		(void)VOP_IOCTL(vp, DIOCCACHESYNC, &i, FWRITE,
2184 		    kauth_cred_get());
2185 	}
2186 
2187 	return error;
2188 }
2189