xref: /netbsd-src/sys/ufs/ext2fs/ext2fs_vfsops.c (revision 2980e352a13e8f0b545a366830c411e7a542ada8)
1 /*	$NetBSD: ext2fs_vfsops.c,v 1.137 2008/06/28 01:34:05 rumble Exp $	*/
2 
3 /*
4  * Copyright (c) 1989, 1991, 1993, 1994
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the University nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  *
31  *	@(#)ffs_vfsops.c	8.14 (Berkeley) 11/28/94
32  * Modified for ext2fs by Manuel Bouyer.
33  */
34 
35 /*
36  * Copyright (c) 1997 Manuel Bouyer.
37  *
38  * Redistribution and use in source and binary forms, with or without
39  * modification, are permitted provided that the following conditions
40  * are met:
41  * 1. Redistributions of source code must retain the above copyright
42  *    notice, this list of conditions and the following disclaimer.
43  * 2. Redistributions in binary form must reproduce the above copyright
44  *    notice, this list of conditions and the following disclaimer in the
45  *    documentation and/or other materials provided with the distribution.
46  * 3. All advertising materials mentioning features or use of this software
47  *    must display the following acknowledgement:
48  *	This product includes software developed by Manuel Bouyer.
49  * 4. The name of the author may not be used to endorse or promote products
50  *    derived from this software without specific prior written permission.
51  *
52  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
53  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
54  * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
55  * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
56  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
57  * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
58  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
59  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
60  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
61  * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
62  *
63  *	@(#)ffs_vfsops.c	8.14 (Berkeley) 11/28/94
64  * Modified for ext2fs by Manuel Bouyer.
65  */
66 
67 #include <sys/cdefs.h>
68 __KERNEL_RCSID(0, "$NetBSD: ext2fs_vfsops.c,v 1.137 2008/06/28 01:34:05 rumble Exp $");
69 
70 #if defined(_KERNEL_OPT)
71 #include "opt_compat_netbsd.h"
72 #endif
73 
74 #include <sys/param.h>
75 #include <sys/systm.h>
76 #include <sys/sysctl.h>
77 #include <sys/namei.h>
78 #include <sys/proc.h>
79 #include <sys/kernel.h>
80 #include <sys/vnode.h>
81 #include <sys/socket.h>
82 #include <sys/mount.h>
83 #include <sys/buf.h>
84 #include <sys/device.h>
85 #include <sys/mbuf.h>
86 #include <sys/file.h>
87 #include <sys/disklabel.h>
88 #include <sys/ioctl.h>
89 #include <sys/errno.h>
90 #include <sys/malloc.h>
91 #include <sys/pool.h>
92 #include <sys/lock.h>
93 #include <sys/conf.h>
94 #include <sys/kauth.h>
95 #include <sys/module.h>
96 
97 #include <miscfs/genfs/genfs.h>
98 #include <miscfs/specfs/specdev.h>
99 
100 #include <ufs/ufs/quota.h>
101 #include <ufs/ufs/ufsmount.h>
102 #include <ufs/ufs/inode.h>
103 #include <ufs/ufs/dir.h>
104 #include <ufs/ufs/ufs_extern.h>
105 
106 #include <ufs/ext2fs/ext2fs.h>
107 #include <ufs/ext2fs/ext2fs_dir.h>
108 #include <ufs/ext2fs/ext2fs_extern.h>
109 
110 MODULE(MODULE_CLASS_VFS, ext2fs, NULL);
111 
112 extern kmutex_t ufs_hashlock;
113 
114 int ext2fs_sbupdate(struct ufsmount *, int);
115 static int ext2fs_checksb(struct ext2fs *, int);
116 
117 static struct sysctllog *ext2fs_sysctl_log;
118 
119 extern const struct vnodeopv_desc ext2fs_vnodeop_opv_desc;
120 extern const struct vnodeopv_desc ext2fs_specop_opv_desc;
121 extern const struct vnodeopv_desc ext2fs_fifoop_opv_desc;
122 
123 const struct vnodeopv_desc * const ext2fs_vnodeopv_descs[] = {
124 	&ext2fs_vnodeop_opv_desc,
125 	&ext2fs_specop_opv_desc,
126 	&ext2fs_fifoop_opv_desc,
127 	NULL,
128 };
129 
130 struct vfsops ext2fs_vfsops = {
131 	MOUNT_EXT2FS,
132 	sizeof (struct ufs_args),
133 	ext2fs_mount,
134 	ufs_start,
135 	ext2fs_unmount,
136 	ufs_root,
137 	ufs_quotactl,
138 	ext2fs_statvfs,
139 	ext2fs_sync,
140 	ext2fs_vget,
141 	ext2fs_fhtovp,
142 	ext2fs_vptofh,
143 	ext2fs_init,
144 	ext2fs_reinit,
145 	ext2fs_done,
146 	ext2fs_mountroot,
147 	(int (*)(struct mount *, struct vnode *, struct timespec *)) eopnotsupp,
148 	vfs_stdextattrctl,
149 	(void *)eopnotsupp,	/* vfs_suspendctl */
150 	genfs_renamelock_enter,
151 	genfs_renamelock_exit,
152 	(void *)eopnotsupp,
153 	ext2fs_vnodeopv_descs,
154 	0,
155 	{ NULL, NULL },
156 };
157 
158 static const struct genfs_ops ext2fs_genfsops = {
159 	.gop_size = genfs_size,
160 	.gop_alloc = ext2fs_gop_alloc,
161 	.gop_write = genfs_gop_write,
162 	.gop_markupdate = ufs_gop_markupdate,
163 };
164 
165 static const struct ufs_ops ext2fs_ufsops = {
166 	.uo_itimes = ext2fs_itimes,
167 	.uo_update = ext2fs_update,
168 	.uo_vfree = ext2fs_vfree,
169 };
170 
171 static int
172 ext2fs_modcmd(modcmd_t cmd, void *arg)
173 {
174 	int error;
175 
176 	switch (cmd) {
177 	case MODULE_CMD_INIT:
178 		error = vfs_attach(&ext2fs_vfsops);
179 		if (error != 0)
180 			break;
181 		sysctl_createv(&ext2fs_sysctl_log, 0, NULL, NULL,
182 			       CTLFLAG_PERMANENT,
183 			       CTLTYPE_NODE, "vfs", NULL,
184 			       NULL, 0, NULL, 0,
185 			       CTL_VFS, CTL_EOL);
186 		sysctl_createv(&ext2fs_sysctl_log, 0, NULL, NULL,
187 			       CTLFLAG_PERMANENT,
188 			       CTLTYPE_NODE, "ext2fs",
189 			       SYSCTL_DESCR("Linux EXT2FS file system"),
190 			       NULL, 0, NULL, 0,
191 			       CTL_VFS, 17, CTL_EOL);
192 		/*
193 		 * XXX the "17" above could be dynamic, thereby eliminating
194 		 * one more instance of the "number to vfs" mapping problem,
195 		 * but "17" is the order as taken from sys/mount.h
196 		 */
197 		break;
198 	case MODULE_CMD_FINI:
199 		error = vfs_detach(&ext2fs_vfsops);
200 		if (error != 0)
201 			break;
202 		sysctl_teardown(&ext2fs_sysctl_log);
203 		break;
204 	default:
205 		error = ENOTTY;
206 		break;
207 	}
208 
209 	return (error);
210 }
211 
212 /*
213  * XXX Same structure as FFS inodes?  Should we share a common pool?
214  */
215 struct pool ext2fs_inode_pool;
216 struct pool ext2fs_dinode_pool;
217 
218 extern u_long ext2gennumber;
219 
220 void
221 ext2fs_init(void)
222 {
223 
224 	pool_init(&ext2fs_inode_pool, sizeof(struct inode), 0, 0, 0,
225 	    "ext2fsinopl", &pool_allocator_nointr, IPL_NONE);
226 	pool_init(&ext2fs_dinode_pool, sizeof(struct ext2fs_dinode), 0, 0, 0,
227 	    "ext2dinopl", &pool_allocator_nointr, IPL_NONE);
228 	ufs_init();
229 }
230 
231 void
232 ext2fs_reinit(void)
233 {
234 	ufs_reinit();
235 }
236 
237 void
238 ext2fs_done(void)
239 {
240 
241 	ufs_done();
242 	pool_destroy(&ext2fs_inode_pool);
243 	pool_destroy(&ext2fs_dinode_pool);
244 }
245 
246 /*
247  * Called by main() when ext2fs is going to be mounted as root.
248  *
249  * Name is updated by mount(8) after booting.
250  */
251 #define ROOTNAME	"root_device"
252 
253 int
254 ext2fs_mountroot(void)
255 {
256 	extern struct vnode *rootvp;
257 	struct m_ext2fs *fs;
258 	struct mount *mp;
259 	struct ufsmount *ump;
260 	int error;
261 
262 	if (device_class(root_device) != DV_DISK)
263 		return (ENODEV);
264 
265 	if ((error = vfs_rootmountalloc(MOUNT_EXT2FS, "root_device", &mp))) {
266 		vrele(rootvp);
267 		return (error);
268 	}
269 
270 	if ((error = ext2fs_mountfs(rootvp, mp)) != 0) {
271 		vfs_unbusy(mp, false, NULL);
272 		vfs_destroy(mp);
273 		return (error);
274 	}
275 	mutex_enter(&mountlist_lock);
276 	CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list);
277 	mutex_exit(&mountlist_lock);
278 	ump = VFSTOUFS(mp);
279 	fs = ump->um_e2fs;
280 	memset(fs->e2fs_fsmnt, 0, sizeof(fs->e2fs_fsmnt));
281 	(void) copystr(mp->mnt_stat.f_mntonname, fs->e2fs_fsmnt,
282 	    sizeof(fs->e2fs_fsmnt) - 1, 0);
283 	if (fs->e2fs.e2fs_rev > E2FS_REV0) {
284 		memset(fs->e2fs.e2fs_fsmnt, 0, sizeof(fs->e2fs.e2fs_fsmnt));
285 		(void) copystr(mp->mnt_stat.f_mntonname, fs->e2fs.e2fs_fsmnt,
286 		    sizeof(fs->e2fs.e2fs_fsmnt) - 1, 0);
287 	}
288 	(void)ext2fs_statvfs(mp, &mp->mnt_stat);
289 	vfs_unbusy(mp, false, NULL);
290 	setrootfstime((time_t)fs->e2fs.e2fs_wtime);
291 	return (0);
292 }
293 
294 /*
295  * VFS Operations.
296  *
297  * mount system call
298  */
299 int
300 ext2fs_mount(struct mount *mp, const char *path, void *data, size_t *data_len)
301 {
302 	struct lwp *l = curlwp;
303 	struct nameidata nd;
304 	struct vnode *devvp;
305 	struct ufs_args *args = data;
306 	struct ufsmount *ump = NULL;
307 	struct m_ext2fs *fs;
308 	size_t size;
309 	int error = 0, flags, update;
310 	mode_t accessmode;
311 
312 	if (*data_len < sizeof *args)
313 		return EINVAL;
314 
315 	if (mp->mnt_flag & MNT_GETARGS) {
316 		ump = VFSTOUFS(mp);
317 		if (ump == NULL)
318 			return EIO;
319 		memset(args, 0, sizeof *args);
320 		args->fspec = NULL;
321 		*data_len = sizeof *args;
322 		return 0;
323 	}
324 
325 	update = mp->mnt_flag & MNT_UPDATE;
326 
327 	/* Check arguments */
328 	if (args->fspec != NULL) {
329 		/*
330 		 * Look up the name and verify that it's sane.
331 		 */
332 		NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, args->fspec);
333 		if ((error = namei(&nd)) != 0)
334 			return (error);
335 		devvp = nd.ni_vp;
336 
337 		if (!update) {
338 			/*
339 			 * Be sure this is a valid block device
340 			 */
341 			if (devvp->v_type != VBLK)
342 				error = ENOTBLK;
343 			else if (bdevsw_lookup(devvp->v_rdev) == NULL)
344 				error = ENXIO;
345 		} else {
346 		        /*
347 			 * Be sure we're still naming the same device
348 			 * used for our initial mount
349 			 */
350 			ump = VFSTOUFS(mp);
351 			if (devvp != ump->um_devvp)
352 				error = EINVAL;
353 		}
354 	} else {
355 		if (!update) {
356 			/* New mounts must have a filename for the device */
357 			return (EINVAL);
358 		} else {
359 			ump = VFSTOUFS(mp);
360 			devvp = ump->um_devvp;
361 			vref(devvp);
362 		}
363 	}
364 
365 	/*
366 	 * If mount by non-root, then verify that user has necessary
367 	 * permissions on the device.
368 	 */
369 	if (error == 0 && kauth_authorize_generic(l->l_cred,
370 	    KAUTH_GENERIC_ISSUSER, NULL) != 0) {
371 		accessmode = VREAD;
372 		if (update ?
373 		    (mp->mnt_iflag & IMNT_WANTRDWR) != 0 :
374 		    (mp->mnt_flag & MNT_RDONLY) == 0)
375 			accessmode |= VWRITE;
376 		vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
377 		error = VOP_ACCESS(devvp, accessmode, l->l_cred);
378 		VOP_UNLOCK(devvp, 0);
379 	}
380 
381 	if (error) {
382 		vrele(devvp);
383 		return (error);
384 	}
385 
386 	if (!update) {
387 		int xflags;
388 
389 		if (mp->mnt_flag & MNT_RDONLY)
390 			xflags = FREAD;
391 		else
392 			xflags = FREAD|FWRITE;
393 		error = VOP_OPEN(devvp, xflags, FSCRED);
394 		if (error)
395 			goto fail;
396 		error = ext2fs_mountfs(devvp, mp);
397 		if (error) {
398 			vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
399 			(void)VOP_CLOSE(devvp, xflags, NOCRED);
400 			VOP_UNLOCK(devvp, 0);
401 			goto fail;
402 		}
403 
404 		ump = VFSTOUFS(mp);
405 		fs = ump->um_e2fs;
406 	} else {
407 		/*
408 		 * Update the mount.
409 		 */
410 
411 		/*
412 		 * The initial mount got a reference on this
413 		 * device, so drop the one obtained via
414 		 * namei(), above.
415 		 */
416 		vrele(devvp);
417 
418 		ump = VFSTOUFS(mp);
419 		fs = ump->um_e2fs;
420 		if (fs->e2fs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) {
421 			/*
422 			 * Changing from r/w to r/o
423 			 */
424 			flags = WRITECLOSE;
425 			if (mp->mnt_flag & MNT_FORCE)
426 				flags |= FORCECLOSE;
427 			error = ext2fs_flushfiles(mp, flags);
428 			if (error == 0 &&
429 			    ext2fs_cgupdate(ump, MNT_WAIT) == 0 &&
430 			    (fs->e2fs.e2fs_state & E2FS_ERRORS) == 0) {
431 				fs->e2fs.e2fs_state = E2FS_ISCLEAN;
432 				(void) ext2fs_sbupdate(ump, MNT_WAIT);
433 			}
434 			if (error)
435 				return (error);
436 			fs->e2fs_ronly = 1;
437 		}
438 
439 		if (mp->mnt_flag & MNT_RELOAD) {
440 			error = ext2fs_reload(mp, l->l_cred);
441 			if (error)
442 				return (error);
443 		}
444 
445 		if (fs->e2fs_ronly && (mp->mnt_iflag & IMNT_WANTRDWR)) {
446 			/*
447 			 * Changing from read-only to read/write
448 			 */
449 			fs->e2fs_ronly = 0;
450 			if (fs->e2fs.e2fs_state == E2FS_ISCLEAN)
451 				fs->e2fs.e2fs_state = 0;
452 			else
453 				fs->e2fs.e2fs_state = E2FS_ERRORS;
454 			fs->e2fs_fmod = 1;
455 		}
456 		if (args->fspec == NULL)
457 			return EINVAL;
458 	}
459 
460 	error = set_statvfs_info(path, UIO_USERSPACE, args->fspec,
461 	    UIO_USERSPACE, mp->mnt_op->vfs_name, mp, l);
462 	(void) copystr(mp->mnt_stat.f_mntonname, fs->e2fs_fsmnt,
463 	    sizeof(fs->e2fs_fsmnt) - 1, &size);
464 	memset(fs->e2fs_fsmnt + size, 0, sizeof(fs->e2fs_fsmnt) - size);
465 	if (fs->e2fs.e2fs_rev > E2FS_REV0) {
466 		(void) copystr(mp->mnt_stat.f_mntonname, fs->e2fs.e2fs_fsmnt,
467 		    sizeof(fs->e2fs.e2fs_fsmnt) - 1, &size);
468 		memset(fs->e2fs.e2fs_fsmnt, 0,
469 		    sizeof(fs->e2fs.e2fs_fsmnt) - size);
470 	}
471 	if (fs->e2fs_fmod != 0) {	/* XXX */
472 		fs->e2fs_fmod = 0;
473 		if (fs->e2fs.e2fs_state == 0)
474 			fs->e2fs.e2fs_wtime = time_second;
475 		else
476 			printf("%s: file system not clean; please fsck(8)\n",
477 				mp->mnt_stat.f_mntfromname);
478 		(void) ext2fs_cgupdate(ump, MNT_WAIT);
479 	}
480 	return (error);
481 
482 fail:
483 	vrele(devvp);
484 	return (error);
485 }
486 
487 /*
488  * Reload all incore data for a filesystem (used after running fsck on
489  * the root filesystem and finding things to fix). The filesystem must
490  * be mounted read-only.
491  *
492  * Things to do to update the mount:
493  *	1) invalidate all cached meta-data.
494  *	2) re-read superblock from disk.
495  *	3) re-read summary information from disk.
496  *	4) invalidate all inactive vnodes.
497  *	5) invalidate all cached file data.
498  *	6) re-read inode data for all active vnodes.
499  */
500 int
501 ext2fs_reload(struct mount *mountp, kauth_cred_t cred)
502 {
503 	struct lwp *l = curlwp;
504 	struct vnode *vp, *mvp, *devvp;
505 	struct inode *ip;
506 	struct buf *bp;
507 	struct m_ext2fs *fs;
508 	struct ext2fs *newfs;
509 	struct partinfo dpart;
510 	int i, size, error;
511 	void *cp;
512 
513 	if ((mountp->mnt_flag & MNT_RDONLY) == 0)
514 		return (EINVAL);
515 
516 	/*
517 	 * Step 1: invalidate all cached meta-data.
518 	 */
519 	devvp = VFSTOUFS(mountp)->um_devvp;
520 	vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
521 	error = vinvalbuf(devvp, 0, cred, l, 0, 0);
522 	VOP_UNLOCK(devvp, 0);
523 	if (error)
524 		panic("ext2fs_reload: dirty1");
525 	/*
526 	 * Step 2: re-read superblock from disk.
527 	 */
528 	if (VOP_IOCTL(devvp, DIOCGPART, &dpart, FREAD, NOCRED) != 0)
529 		size = DEV_BSIZE;
530 	else
531 		size = dpart.disklab->d_secsize;
532 	error = bread(devvp, (daddr_t)(SBOFF / size), SBSIZE, NOCRED, 0, &bp);
533 	if (error) {
534 		brelse(bp, 0);
535 		return (error);
536 	}
537 	newfs = (struct ext2fs *)bp->b_data;
538 	error = ext2fs_checksb(newfs, (mountp->mnt_flag & MNT_RDONLY) != 0);
539 	if (error) {
540 		brelse(bp, 0);
541 		return (error);
542 	}
543 
544 	fs = VFSTOUFS(mountp)->um_e2fs;
545 	/*
546 	 * copy in new superblock, and compute in-memory values
547 	 */
548 	e2fs_sbload(newfs, &fs->e2fs);
549 	fs->e2fs_ncg =
550 	    howmany(fs->e2fs.e2fs_bcount - fs->e2fs.e2fs_first_dblock,
551 	    fs->e2fs.e2fs_bpg);
552 	/* XXX assume hw bsize = 512 */
553 	fs->e2fs_fsbtodb = fs->e2fs.e2fs_log_bsize + 1;
554 	fs->e2fs_bsize = MINBSIZE << fs->e2fs.e2fs_log_bsize;
555 	fs->e2fs_bshift = LOG_MINBSIZE + fs->e2fs.e2fs_log_bsize;
556 	fs->e2fs_qbmask = fs->e2fs_bsize - 1;
557 	fs->e2fs_bmask = ~fs->e2fs_qbmask;
558 	fs->e2fs_ngdb =
559 	    howmany(fs->e2fs_ncg, fs->e2fs_bsize / sizeof(struct ext2_gd));
560 	fs->e2fs_ipb = fs->e2fs_bsize / EXT2_DINODE_SIZE;
561 	fs->e2fs_itpg = fs->e2fs.e2fs_ipg / fs->e2fs_ipb;
562 
563 	/*
564 	 * Step 3: re-read summary information from disk.
565 	 */
566 
567 	for (i = 0; i < fs->e2fs_ngdb; i++) {
568 		error = bread(devvp ,
569 		    fsbtodb(fs, fs->e2fs.e2fs_first_dblock +
570 		    1 /* superblock */ + i),
571 		    fs->e2fs_bsize, NOCRED, 0, &bp);
572 		if (error) {
573 			brelse(bp, 0);
574 			return (error);
575 		}
576 		e2fs_cgload((struct ext2_gd *)bp->b_data,
577 		    &fs->e2fs_gd[i * fs->e2fs_bsize / sizeof(struct ext2_gd)],
578 		    fs->e2fs_bsize);
579 		brelse(bp, 0);
580 	}
581 
582 	/* Allocate a marker vnode. */
583 	if ((mvp = vnalloc(mountp)) == NULL)
584 		return (ENOMEM);
585 	/*
586 	 * NOTE: not using the TAILQ_FOREACH here since in this loop vgone()
587 	 * and vclean() can be called indirectly
588 	 */
589 	mutex_enter(&mntvnode_lock);
590 loop:
591 	for (vp = TAILQ_FIRST(&mountp->mnt_vnodelist); vp; vp = vunmark(mvp)) {
592 		vmark(mvp, vp);
593 		if (vp->v_mount != mountp || vismarker(vp))
594 			continue;
595 		/*
596 		 * Step 4: invalidate all inactive vnodes.
597 		 */
598 		if (vrecycle(vp, &mntvnode_lock, l)) {
599 			mutex_enter(&mntvnode_lock);
600 			(void)vunmark(mvp);
601 			goto loop;
602 		}
603 		/*
604 		 * Step 5: invalidate all cached file data.
605 		 */
606 		mutex_enter(&vp->v_interlock);
607 		mutex_exit(&mntvnode_lock);
608 		if (vget(vp, LK_EXCLUSIVE | LK_INTERLOCK)) {
609 			mutex_enter(&mntvnode_lock);
610 			(void)vunmark(mvp);
611 			goto loop;
612 		}
613 		if (vinvalbuf(vp, 0, cred, l, 0, 0))
614 			panic("ext2fs_reload: dirty2");
615 		/*
616 		 * Step 6: re-read inode data for all active vnodes.
617 		 */
618 		ip = VTOI(vp);
619 		error = bread(devvp, fsbtodb(fs, ino_to_fsba(fs, ip->i_number)),
620 		    (int)fs->e2fs_bsize, NOCRED, 0, &bp);
621 		if (error) {
622 			vput(vp);
623 			mutex_enter(&mntvnode_lock);
624 			(void)vunmark(mvp);
625 			break;
626 		}
627 		cp = (char *)bp->b_data +
628 		    (ino_to_fsbo(fs, ip->i_number) * EXT2_DINODE_SIZE);
629 		e2fs_iload((struct ext2fs_dinode *)cp, ip->i_din.e2fs_din);
630 		brelse(bp, 0);
631 		vput(vp);
632 		mutex_enter(&mntvnode_lock);
633 	}
634 	mutex_exit(&mntvnode_lock);
635 	vnfree(mvp);
636 	return (error);
637 }
638 
639 /*
640  * Common code for mount and mountroot
641  */
642 int
643 ext2fs_mountfs(struct vnode *devvp, struct mount *mp)
644 {
645 	struct lwp *l = curlwp;
646 	struct ufsmount *ump;
647 	struct buf *bp;
648 	struct ext2fs *fs;
649 	struct m_ext2fs *m_fs;
650 	dev_t dev;
651 	struct partinfo dpart;
652 	int error, i, size, ronly;
653 	kauth_cred_t cred;
654 	struct proc *p;
655 
656 	dev = devvp->v_rdev;
657 	p = l ? l->l_proc : NULL;
658 	cred = l ? l->l_cred : NOCRED;
659 
660 	/* Flush out any old buffers remaining from a previous use. */
661 	vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
662 	error = vinvalbuf(devvp, V_SAVE, cred, l, 0, 0);
663 	VOP_UNLOCK(devvp, 0);
664 	if (error)
665 		return (error);
666 
667 	ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
668 	if (VOP_IOCTL(devvp, DIOCGPART, &dpart, FREAD, cred) != 0)
669 		size = DEV_BSIZE;
670 	else
671 		size = dpart.disklab->d_secsize;
672 
673 	bp = NULL;
674 	ump = NULL;
675 
676 #ifdef DEBUG_EXT2
677 	printf("sb size: %d ino size %d\n", sizeof(struct ext2fs),
678 	    EXT2_DINODE_SIZE);
679 #endif
680 	error = bread(devvp, (SBOFF / size), SBSIZE, cred, 0, &bp);
681 	if (error)
682 		goto out;
683 	fs = (struct ext2fs *)bp->b_data;
684 	error = ext2fs_checksb(fs, ronly);
685 	if (error)
686 		goto out;
687 	ump = malloc(sizeof(*ump), M_UFSMNT, M_WAITOK);
688 	memset(ump, 0, sizeof(*ump));
689 	ump->um_fstype = UFS1;
690 	ump->um_ops = &ext2fs_ufsops;
691 	ump->um_e2fs = malloc(sizeof(struct m_ext2fs), M_UFSMNT, M_WAITOK);
692 	memset(ump->um_e2fs, 0, sizeof(struct m_ext2fs));
693 	e2fs_sbload((struct ext2fs *)bp->b_data, &ump->um_e2fs->e2fs);
694 	brelse(bp, 0);
695 	bp = NULL;
696 	m_fs = ump->um_e2fs;
697 	m_fs->e2fs_ronly = ronly;
698 	if (ronly == 0) {
699 		if (m_fs->e2fs.e2fs_state == E2FS_ISCLEAN)
700 			m_fs->e2fs.e2fs_state = 0;
701 		else
702 			m_fs->e2fs.e2fs_state = E2FS_ERRORS;
703 		m_fs->e2fs_fmod = 1;
704 	}
705 
706 	/* compute dynamic sb infos */
707 	m_fs->e2fs_ncg =
708 	    howmany(m_fs->e2fs.e2fs_bcount - m_fs->e2fs.e2fs_first_dblock,
709 	    m_fs->e2fs.e2fs_bpg);
710 	/* XXX assume hw bsize = 512 */
711 	m_fs->e2fs_fsbtodb = m_fs->e2fs.e2fs_log_bsize + 1;
712 	m_fs->e2fs_bsize = MINBSIZE << m_fs->e2fs.e2fs_log_bsize;
713 	m_fs->e2fs_bshift = LOG_MINBSIZE + m_fs->e2fs.e2fs_log_bsize;
714 	m_fs->e2fs_qbmask = m_fs->e2fs_bsize - 1;
715 	m_fs->e2fs_bmask = ~m_fs->e2fs_qbmask;
716 	m_fs->e2fs_ngdb =
717 	    howmany(m_fs->e2fs_ncg, m_fs->e2fs_bsize / sizeof(struct ext2_gd));
718 	m_fs->e2fs_ipb = m_fs->e2fs_bsize / EXT2_DINODE_SIZE;
719 	m_fs->e2fs_itpg = m_fs->e2fs.e2fs_ipg / m_fs->e2fs_ipb;
720 
721 	m_fs->e2fs_gd = malloc(m_fs->e2fs_ngdb * m_fs->e2fs_bsize,
722 	    M_UFSMNT, M_WAITOK);
723 	for (i = 0; i < m_fs->e2fs_ngdb; i++) {
724 		error = bread(devvp ,
725 		    fsbtodb(m_fs, m_fs->e2fs.e2fs_first_dblock +
726 		    1 /* superblock */ + i),
727 		    m_fs->e2fs_bsize, NOCRED, 0, &bp);
728 		if (error) {
729 			free(m_fs->e2fs_gd, M_UFSMNT);
730 			goto out;
731 		}
732 		e2fs_cgload((struct ext2_gd *)bp->b_data,
733 		    &m_fs->e2fs_gd[
734 			i * m_fs->e2fs_bsize / sizeof(struct ext2_gd)],
735 		    m_fs->e2fs_bsize);
736 		brelse(bp, 0);
737 		bp = NULL;
738 	}
739 
740 	mp->mnt_data = ump;
741 	mp->mnt_stat.f_fsidx.__fsid_val[0] = (long)dev;
742 	mp->mnt_stat.f_fsidx.__fsid_val[1] = makefstype(MOUNT_EXT2FS);
743 	mp->mnt_stat.f_fsid = mp->mnt_stat.f_fsidx.__fsid_val[0];
744 	mp->mnt_stat.f_namemax = EXT2FS_MAXNAMLEN;
745 	mp->mnt_flag |= MNT_LOCAL;
746 	mp->mnt_dev_bshift = DEV_BSHIFT;	/* XXX */
747 	mp->mnt_fs_bshift = m_fs->e2fs_bshift;
748 	mp->mnt_iflag |= IMNT_DTYPE;
749 	ump->um_flags = 0;
750 	ump->um_mountp = mp;
751 	ump->um_dev = dev;
752 	ump->um_devvp = devvp;
753 	ump->um_nindir = NINDIR(m_fs);
754 	ump->um_lognindir = ffs(NINDIR(m_fs)) - 1;
755 	ump->um_bptrtodb = m_fs->e2fs_fsbtodb;
756 	ump->um_seqinc = 1; /* no frags */
757 	ump->um_maxsymlinklen = EXT2_MAXSYMLINKLEN;
758 	ump->um_dirblksiz = m_fs->e2fs_bsize;
759 	ump->um_maxfilesize = ((uint64_t)0x80000000 * m_fs->e2fs_bsize - 1);
760 	devvp->v_specmountpoint = mp;
761 	return (0);
762 
763 out:
764 	KASSERT(bp != NULL);
765 	brelse(bp, 0);
766 	if (ump) {
767 		free(ump->um_e2fs, M_UFSMNT);
768 		free(ump, M_UFSMNT);
769 		mp->mnt_data = NULL;
770 	}
771 	return (error);
772 }
773 
774 /*
775  * unmount system call
776  */
777 int
778 ext2fs_unmount(struct mount *mp, int mntflags)
779 {
780 	struct ufsmount *ump;
781 	struct m_ext2fs *fs;
782 	int error, flags;
783 
784 	flags = 0;
785 	if (mntflags & MNT_FORCE)
786 		flags |= FORCECLOSE;
787 	if ((error = ext2fs_flushfiles(mp, flags)) != 0)
788 		return (error);
789 	ump = VFSTOUFS(mp);
790 	fs = ump->um_e2fs;
791 	if (fs->e2fs_ronly == 0 &&
792 		ext2fs_cgupdate(ump, MNT_WAIT) == 0 &&
793 		(fs->e2fs.e2fs_state & E2FS_ERRORS) == 0) {
794 		fs->e2fs.e2fs_state = E2FS_ISCLEAN;
795 		(void) ext2fs_sbupdate(ump, MNT_WAIT);
796 	}
797 	if (ump->um_devvp->v_type != VBAD)
798 		ump->um_devvp->v_specmountpoint = NULL;
799 	vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
800 	error = VOP_CLOSE(ump->um_devvp, fs->e2fs_ronly ? FREAD : FREAD|FWRITE,
801 	    NOCRED);
802 	vput(ump->um_devvp);
803 	free(fs->e2fs_gd, M_UFSMNT);
804 	free(fs, M_UFSMNT);
805 	free(ump, M_UFSMNT);
806 	mp->mnt_data = NULL;
807 	mp->mnt_flag &= ~MNT_LOCAL;
808 	return (error);
809 }
810 
811 /*
812  * Flush out all the files in a filesystem.
813  */
814 int
815 ext2fs_flushfiles(struct mount *mp, int flags)
816 {
817 	extern int doforce;
818 	int error;
819 
820 	if (!doforce)
821 		flags &= ~FORCECLOSE;
822 	error = vflush(mp, NULLVP, flags);
823 	return (error);
824 }
825 
826 /*
827  * Get file system statistics.
828  */
829 int
830 ext2fs_statvfs(struct mount *mp, struct statvfs *sbp)
831 {
832 	struct ufsmount *ump;
833 	struct m_ext2fs *fs;
834 	uint32_t overhead, overhead_per_group, ngdb;
835 	int i, ngroups;
836 
837 	ump = VFSTOUFS(mp);
838 	fs = ump->um_e2fs;
839 	if (fs->e2fs.e2fs_magic != E2FS_MAGIC)
840 		panic("ext2fs_statvfs");
841 
842 	/*
843 	 * Compute the overhead (FS structures)
844 	 */
845 	overhead_per_group =
846 	    1 /* block bitmap */ +
847 	    1 /* inode bitmap */ +
848 	    fs->e2fs_itpg;
849 	overhead = fs->e2fs.e2fs_first_dblock +
850 	    fs->e2fs_ncg * overhead_per_group;
851 	if (fs->e2fs.e2fs_rev > E2FS_REV0 &&
852 	    fs->e2fs.e2fs_features_rocompat & EXT2F_ROCOMPAT_SPARSESUPER) {
853 		for (i = 0, ngroups = 0; i < fs->e2fs_ncg; i++) {
854 			if (cg_has_sb(i))
855 				ngroups++;
856 		}
857 	} else {
858 		ngroups = fs->e2fs_ncg;
859 	}
860 	ngdb = fs->e2fs_ngdb;
861 	if (fs->e2fs.e2fs_rev > E2FS_REV0 &&
862 	    fs->e2fs.e2fs_features_compat & EXT2F_COMPAT_RESIZE)
863 		ngdb += fs->e2fs.e2fs_reserved_ngdb;
864 	overhead += ngroups * (1 /* superblock */ + ngdb);
865 
866 	sbp->f_bsize = fs->e2fs_bsize;
867 	sbp->f_frsize = MINBSIZE << fs->e2fs.e2fs_fsize;
868 	sbp->f_iosize = fs->e2fs_bsize;
869 	sbp->f_blocks = fs->e2fs.e2fs_bcount - overhead;
870 	sbp->f_bfree = fs->e2fs.e2fs_fbcount;
871 	sbp->f_bresvd = fs->e2fs.e2fs_rbcount;
872 	if (sbp->f_bfree > sbp->f_bresvd)
873 		sbp->f_bavail = sbp->f_bfree - sbp->f_bresvd;
874 	else
875 		sbp->f_bavail = 0;
876 	sbp->f_files =  fs->e2fs.e2fs_icount;
877 	sbp->f_ffree = fs->e2fs.e2fs_ficount;
878 	sbp->f_favail = fs->e2fs.e2fs_ficount;
879 	sbp->f_fresvd = 0;
880 	copy_statvfs_info(sbp, mp);
881 	return (0);
882 }
883 
884 /*
885  * Go through the disk queues to initiate sandbagged IO;
886  * go through the inodes to write those that have been modified;
887  * initiate the writing of the super block if it has been modified.
888  *
889  * Note: we are always called with the filesystem marked `MPBUSY'.
890  */
891 int
892 ext2fs_sync(struct mount *mp, int waitfor, kauth_cred_t cred)
893 {
894 	struct vnode *vp, *mvp;
895 	struct inode *ip;
896 	struct ufsmount *ump = VFSTOUFS(mp);
897 	struct m_ext2fs *fs;
898 	int error, allerror = 0;
899 
900 	fs = ump->um_e2fs;
901 	if (fs->e2fs_fmod != 0 && fs->e2fs_ronly != 0) {	/* XXX */
902 		printf("fs = %s\n", fs->e2fs_fsmnt);
903 		panic("update: rofs mod");
904 	}
905 
906 	/* Allocate a marker vnode. */
907 	if ((mvp = vnalloc(mp)) == NULL)
908 		return (ENOMEM);
909 
910 	/*
911 	 * Write back each (modified) inode.
912 	 */
913 	mutex_enter(&mntvnode_lock);
914 loop:
915 	/*
916 	 * NOTE: not using the TAILQ_FOREACH here since in this loop vgone()
917 	 * and vclean() can be called indirectly
918 	 */
919 	for (vp = TAILQ_FIRST(&mp->mnt_vnodelist); vp; vp = vunmark(mvp)) {
920 		vmark(mvp, vp);
921 		if (vp->v_mount != mp || vismarker(vp))
922 			continue;
923 		mutex_enter(&vp->v_interlock);
924 		ip = VTOI(vp);
925 		if (ip == NULL || (vp->v_iflag & (VI_XLOCK|VI_CLEAN)) != 0 ||
926 		    vp->v_type == VNON ||
927 		    ((ip->i_flag &
928 		      (IN_CHANGE | IN_UPDATE | IN_MODIFIED)) == 0 &&
929 		     LIST_EMPTY(&vp->v_dirtyblkhd) &&
930 		     UVM_OBJ_IS_CLEAN(&vp->v_uobj)))
931 		{
932 			mutex_exit(&vp->v_interlock);
933 			continue;
934 		}
935 		mutex_exit(&mntvnode_lock);
936 		error = vget(vp, LK_EXCLUSIVE | LK_NOWAIT | LK_INTERLOCK);
937 		if (error) {
938 			mutex_enter(&mntvnode_lock);
939 			if (error == ENOENT) {
940 				mutex_enter(&mntvnode_lock);
941 				(void)vunmark(mvp);
942 				goto loop;
943 			}
944 			continue;
945 		}
946 		if (vp->v_type == VREG && waitfor == MNT_LAZY)
947 			error = ext2fs_update(vp, NULL, NULL, 0);
948 		else
949 			error = VOP_FSYNC(vp, cred,
950 			    waitfor == MNT_WAIT ? FSYNC_WAIT : 0, 0, 0);
951 		if (error)
952 			allerror = error;
953 		vput(vp);
954 		mutex_enter(&mntvnode_lock);
955 	}
956 	mutex_exit(&mntvnode_lock);
957 	vnfree(mvp);
958 	/*
959 	 * Force stale file system control information to be flushed.
960 	 */
961 	if (waitfor != MNT_LAZY) {
962 		vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
963 		if ((error = VOP_FSYNC(ump->um_devvp, cred,
964 		    waitfor == MNT_WAIT ? FSYNC_WAIT : 0, 0, 0)) != 0)
965 			allerror = error;
966 		VOP_UNLOCK(ump->um_devvp, 0);
967 	}
968 	/*
969 	 * Write back modified superblock.
970 	 */
971 	if (fs->e2fs_fmod != 0) {
972 		fs->e2fs_fmod = 0;
973 		fs->e2fs.e2fs_wtime = time_second;
974 		if ((error = ext2fs_cgupdate(ump, waitfor)))
975 			allerror = error;
976 	}
977 	return (allerror);
978 }
979 
980 /*
981  * Look up a EXT2FS dinode number to find its incore vnode, otherwise read it
982  * in from disk.  If it is in core, wait for the lock bit to clear, then
983  * return the inode locked.  Detection and handling of mount points must be
984  * done by the calling routine.
985  */
986 int
987 ext2fs_vget(struct mount *mp, ino_t ino, struct vnode **vpp)
988 {
989 	struct m_ext2fs *fs;
990 	struct inode *ip;
991 	struct ufsmount *ump;
992 	struct buf *bp;
993 	struct vnode *vp;
994 	dev_t dev;
995 	int error;
996 	void *cp;
997 
998 	ump = VFSTOUFS(mp);
999 	dev = ump->um_dev;
1000 retry:
1001 	if ((*vpp = ufs_ihashget(dev, ino, LK_EXCLUSIVE)) != NULL)
1002 		return (0);
1003 
1004 	/* Allocate a new vnode/inode. */
1005 	if ((error = getnewvnode(VT_EXT2FS, mp, ext2fs_vnodeop_p, &vp)) != 0) {
1006 		*vpp = NULL;
1007 		return (error);
1008 	}
1009 	ip = pool_get(&ext2fs_inode_pool, PR_WAITOK);
1010 
1011 	mutex_enter(&ufs_hashlock);
1012 	if ((*vpp = ufs_ihashget(dev, ino, 0)) != NULL) {
1013 		mutex_exit(&ufs_hashlock);
1014 		ungetnewvnode(vp);
1015 		pool_put(&ext2fs_inode_pool, ip);
1016 		goto retry;
1017 	}
1018 
1019 	vp->v_vflag |= VV_LOCKSWORK;
1020 
1021 	memset(ip, 0, sizeof(struct inode));
1022 	vp->v_data = ip;
1023 	ip->i_vnode = vp;
1024 	ip->i_ump = ump;
1025 	ip->i_e2fs = fs = ump->um_e2fs;
1026 	ip->i_dev = dev;
1027 	ip->i_number = ino;
1028 	ip->i_e2fs_last_lblk = 0;
1029 	ip->i_e2fs_last_blk = 0;
1030 	genfs_node_init(vp, &ext2fs_genfsops);
1031 
1032 	/*
1033 	 * Put it onto its hash chain and lock it so that other requests for
1034 	 * this inode will block if they arrive while we are sleeping waiting
1035 	 * for old data structures to be purged or for the contents of the
1036 	 * disk portion of this inode to be read.
1037 	 */
1038 
1039 	ufs_ihashins(ip);
1040 	mutex_exit(&ufs_hashlock);
1041 
1042 	/* Read in the disk contents for the inode, copy into the inode. */
1043 	error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)),
1044 	    (int)fs->e2fs_bsize, NOCRED, 0, &bp);
1045 	if (error) {
1046 
1047 		/*
1048 		 * The inode does not contain anything useful, so it would
1049 		 * be misleading to leave it on its hash chain. With mode
1050 		 * still zero, it will be unlinked and returned to the free
1051 		 * list by vput().
1052 		 */
1053 
1054 		vput(vp);
1055 		brelse(bp, 0);
1056 		*vpp = NULL;
1057 		return (error);
1058 	}
1059 	cp = (char *)bp->b_data + (ino_to_fsbo(fs, ino) * EXT2_DINODE_SIZE);
1060 	ip->i_din.e2fs_din = pool_get(&ext2fs_dinode_pool, PR_WAITOK);
1061 	e2fs_iload((struct ext2fs_dinode *)cp, ip->i_din.e2fs_din);
1062 	brelse(bp, 0);
1063 
1064 	/* If the inode was deleted, reset all fields */
1065 	if (ip->i_e2fs_dtime != 0) {
1066 		ip->i_e2fs_mode = ip->i_e2fs_nblock = 0;
1067 		(void)ext2fs_setsize(ip, 0);
1068 		memset(ip->i_e2fs_blocks, 0, sizeof(ip->i_e2fs_blocks));
1069 	}
1070 
1071 	/*
1072 	 * Initialize the vnode from the inode, check for aliases.
1073 	 * Note that the underlying vnode may have changed.
1074 	 */
1075 
1076 	error = ext2fs_vinit(mp, ext2fs_specop_p, ext2fs_fifoop_p, &vp);
1077 	if (error) {
1078 		vput(vp);
1079 		*vpp = NULL;
1080 		return (error);
1081 	}
1082 	/*
1083 	 * Finish inode initialization now that aliasing has been resolved.
1084 	 */
1085 
1086 	ip->i_devvp = ump->um_devvp;
1087 	VREF(ip->i_devvp);
1088 
1089 	/*
1090 	 * Set up a generation number for this inode if it does not
1091 	 * already have one. This should only happen on old filesystems.
1092 	 */
1093 
1094 	if (ip->i_e2fs_gen == 0) {
1095 		if (++ext2gennumber < (u_long)time_second)
1096 			ext2gennumber = time_second;
1097 		ip->i_e2fs_gen = ext2gennumber;
1098 		if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0)
1099 			ip->i_flag |= IN_MODIFIED;
1100 	}
1101 	uvm_vnp_setsize(vp, ext2fs_size(ip));
1102 	*vpp = vp;
1103 	return (0);
1104 }
1105 
1106 /*
1107  * File handle to vnode
1108  *
1109  * Have to be really careful about stale file handles:
1110  * - check that the inode number is valid
1111  * - call ext2fs_vget() to get the locked inode
1112  * - check for an unallocated inode (i_mode == 0)
1113  */
1114 int
1115 ext2fs_fhtovp(struct mount *mp, struct fid *fhp, struct vnode **vpp)
1116 {
1117 	struct inode *ip;
1118 	struct vnode *nvp;
1119 	int error;
1120 	struct ufid ufh;
1121 	struct m_ext2fs *fs;
1122 
1123 	if (fhp->fid_len != sizeof(struct ufid))
1124 		return EINVAL;
1125 
1126 	memcpy(&ufh, fhp, sizeof(struct ufid));
1127 	fs = VFSTOUFS(mp)->um_e2fs;
1128 	if ((ufh.ufid_ino < EXT2_FIRSTINO && ufh.ufid_ino != EXT2_ROOTINO) ||
1129 		ufh.ufid_ino >= fs->e2fs_ncg * fs->e2fs.e2fs_ipg)
1130 		return (ESTALE);
1131 
1132 	if ((error = VFS_VGET(mp, ufh.ufid_ino, &nvp)) != 0) {
1133 		*vpp = NULLVP;
1134 		return (error);
1135 	}
1136 	ip = VTOI(nvp);
1137 	if (ip->i_e2fs_mode == 0 || ip->i_e2fs_dtime != 0 ||
1138 		ip->i_e2fs_gen != ufh.ufid_gen) {
1139 		vput(nvp);
1140 		*vpp = NULLVP;
1141 		return (ESTALE);
1142 	}
1143 	*vpp = nvp;
1144 	return (0);
1145 }
1146 
1147 /*
1148  * Vnode pointer to File handle
1149  */
1150 /* ARGSUSED */
1151 int
1152 ext2fs_vptofh(struct vnode *vp, struct fid *fhp, size_t *fh_size)
1153 {
1154 	struct inode *ip;
1155 	struct ufid ufh;
1156 
1157 	if (*fh_size < sizeof(struct ufid)) {
1158 		*fh_size = sizeof(struct ufid);
1159 		return E2BIG;
1160 	}
1161 	*fh_size = sizeof(struct ufid);
1162 
1163 	ip = VTOI(vp);
1164 	memset(&ufh, 0, sizeof(ufh));
1165 	ufh.ufid_len = sizeof(struct ufid);
1166 	ufh.ufid_ino = ip->i_number;
1167 	ufh.ufid_gen = ip->i_e2fs_gen;
1168 	memcpy(fhp, &ufh, sizeof(ufh));
1169 	return (0);
1170 }
1171 
1172 /*
1173  * Write a superblock and associated information back to disk.
1174  */
1175 int
1176 ext2fs_sbupdate(struct ufsmount *mp, int waitfor)
1177 {
1178 	struct m_ext2fs *fs = mp->um_e2fs;
1179 	struct buf *bp;
1180 	int error = 0;
1181 
1182 	bp = getblk(mp->um_devvp, SBLOCK, SBSIZE, 0, 0);
1183 	e2fs_sbsave(&fs->e2fs, (struct ext2fs*)bp->b_data);
1184 	if (waitfor == MNT_WAIT)
1185 		error = bwrite(bp);
1186 	else
1187 		bawrite(bp);
1188 	return (error);
1189 }
1190 
1191 int
1192 ext2fs_cgupdate(struct ufsmount *mp, int waitfor)
1193 {
1194 	struct m_ext2fs *fs = mp->um_e2fs;
1195 	struct buf *bp;
1196 	int i, error = 0, allerror = 0;
1197 
1198 	allerror = ext2fs_sbupdate(mp, waitfor);
1199 	for (i = 0; i < fs->e2fs_ngdb; i++) {
1200 		bp = getblk(mp->um_devvp, fsbtodb(fs,
1201 		    fs->e2fs.e2fs_first_dblock +
1202 		    1 /* superblock */ + i), fs->e2fs_bsize, 0, 0);
1203 		e2fs_cgsave(&fs->e2fs_gd[
1204 		    i * fs->e2fs_bsize / sizeof(struct ext2_gd)],
1205 		    (struct ext2_gd *)bp->b_data, fs->e2fs_bsize);
1206 		if (waitfor == MNT_WAIT)
1207 			error = bwrite(bp);
1208 		else
1209 			bawrite(bp);
1210 	}
1211 
1212 	if (!allerror && error)
1213 		allerror = error;
1214 	return (allerror);
1215 }
1216 
1217 static int
1218 ext2fs_checksb(struct ext2fs *fs, int ronly)
1219 {
1220 
1221 	if (fs2h16(fs->e2fs_magic) != E2FS_MAGIC) {
1222 		return (EINVAL);		/* XXX needs translation */
1223 	}
1224 	if (fs2h32(fs->e2fs_rev) > E2FS_REV1) {
1225 #ifdef DIAGNOSTIC
1226 		printf("Ext2 fs: unsupported revision number: %x\n",
1227 		    fs2h32(fs->e2fs_rev));
1228 #endif
1229 		return (EINVAL);		/* XXX needs translation */
1230 	}
1231 	if (fs2h32(fs->e2fs_log_bsize) > 2) { /* block size = 1024|2048|4096 */
1232 #ifdef DIAGNOSTIC
1233 		printf("Ext2 fs: bad block size: %d "
1234 		    "(expected <= 2 for ext2 fs)\n",
1235 		    fs2h32(fs->e2fs_log_bsize));
1236 #endif
1237 		return (EINVAL);	   /* XXX needs translation */
1238 	}
1239 	if (fs2h32(fs->e2fs_rev) > E2FS_REV0) {
1240 		if (fs2h32(fs->e2fs_first_ino) != EXT2_FIRSTINO ||
1241 		    fs2h16(fs->e2fs_inode_size) != EXT2_DINODE_SIZE) {
1242 			printf("Ext2 fs: unsupported inode size\n");
1243 			return (EINVAL);      /* XXX needs translation */
1244 		}
1245 		if (fs2h32(fs->e2fs_features_incompat) &
1246 		    ~EXT2F_INCOMPAT_SUPP) {
1247 			printf("Ext2 fs: unsupported optional feature\n");
1248 			return (EINVAL);      /* XXX needs translation */
1249 		}
1250 		if (!ronly && fs2h32(fs->e2fs_features_rocompat) &
1251 		    ~EXT2F_ROCOMPAT_SUPP) {
1252 			return (EROFS);      /* XXX needs translation */
1253 		}
1254 	}
1255 	return (0);
1256 }
1257