xref: /netbsd-src/sys/ufs/ext2fs/ext2fs_vfsops.c (revision 796c32c94f6e154afc9de0f63da35c91bb739b45)
1 /*	$NetBSD: ext2fs_vfsops.c,v 1.210 2017/07/30 14:23:54 riastradh 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  *
47  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
48  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
49  * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
50  * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
51  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
52  * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
53  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
54  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
55  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
56  * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
57  *
58  *	@(#)ffs_vfsops.c	8.14 (Berkeley) 11/28/94
59  * Modified for ext2fs by Manuel Bouyer.
60  */
61 
62 #include <sys/cdefs.h>
63 __KERNEL_RCSID(0, "$NetBSD: ext2fs_vfsops.c,v 1.210 2017/07/30 14:23:54 riastradh Exp $");
64 
65 #if defined(_KERNEL_OPT)
66 #include "opt_compat_netbsd.h"
67 #endif
68 
69 #include <sys/param.h>
70 #include <sys/systm.h>
71 #include <sys/sysctl.h>
72 #include <sys/namei.h>
73 #include <sys/proc.h>
74 #include <sys/kernel.h>
75 #include <sys/vnode.h>
76 #include <sys/socket.h>
77 #include <sys/mount.h>
78 #include <sys/buf.h>
79 #include <sys/device.h>
80 #include <sys/mbuf.h>
81 #include <sys/file.h>
82 #include <sys/disklabel.h>
83 #include <sys/ioctl.h>
84 #include <sys/errno.h>
85 #include <sys/pool.h>
86 #include <sys/lock.h>
87 #include <sys/conf.h>
88 #include <sys/kauth.h>
89 #include <sys/module.h>
90 
91 #include <miscfs/genfs/genfs.h>
92 #include <miscfs/specfs/specdev.h>
93 
94 #include <ufs/ufs/quota.h>
95 #include <ufs/ufs/ufsmount.h>
96 #include <ufs/ufs/inode.h>
97 #include <ufs/ufs/dir.h>
98 #include <ufs/ufs/ufs_extern.h>
99 
100 #include <ufs/ext2fs/ext2fs.h>
101 #include <ufs/ext2fs/ext2fs_dir.h>
102 #include <ufs/ext2fs/ext2fs_extern.h>
103 
104 MODULE(MODULE_CLASS_VFS, ext2fs, "ffs");
105 
106 int ext2fs_sbupdate(struct ufsmount *, int);
107 static int ext2fs_sbfill(struct m_ext2fs *, int);
108 
109 static struct sysctllog *ext2fs_sysctl_log;
110 
111 extern const struct vnodeopv_desc ext2fs_vnodeop_opv_desc;
112 extern const struct vnodeopv_desc ext2fs_specop_opv_desc;
113 extern const struct vnodeopv_desc ext2fs_fifoop_opv_desc;
114 
115 const struct vnodeopv_desc * const ext2fs_vnodeopv_descs[] = {
116 	&ext2fs_vnodeop_opv_desc,
117 	&ext2fs_specop_opv_desc,
118 	&ext2fs_fifoop_opv_desc,
119 	NULL,
120 };
121 
122 struct vfsops ext2fs_vfsops = {
123 	.vfs_name = MOUNT_EXT2FS,
124 	.vfs_min_mount_data = sizeof (struct ufs_args),
125 	.vfs_mount = ext2fs_mount,
126 	.vfs_start = ufs_start,
127 	.vfs_unmount = ext2fs_unmount,
128 	.vfs_root = ufs_root,
129 	.vfs_quotactl = ufs_quotactl,
130 	.vfs_statvfs = ext2fs_statvfs,
131 	.vfs_sync = ext2fs_sync,
132 	.vfs_vget = ufs_vget,
133 	.vfs_loadvnode = ext2fs_loadvnode,
134 	.vfs_newvnode = ext2fs_newvnode,
135 	.vfs_fhtovp = ext2fs_fhtovp,
136 	.vfs_vptofh = ext2fs_vptofh,
137 	.vfs_init = ext2fs_init,
138 	.vfs_reinit = ext2fs_reinit,
139 	.vfs_done = ext2fs_done,
140 	.vfs_mountroot = ext2fs_mountroot,
141 	.vfs_snapshot = (void *)eopnotsupp,
142 	.vfs_extattrctl = vfs_stdextattrctl,
143 	.vfs_suspendctl = genfs_suspendctl,
144 	.vfs_renamelock_enter = genfs_renamelock_enter,
145 	.vfs_renamelock_exit = genfs_renamelock_exit,
146 	.vfs_fsync = (void *)eopnotsupp,
147 	.vfs_opv_descs = ext2fs_vnodeopv_descs
148 };
149 
150 static const struct genfs_ops ext2fs_genfsops = {
151 	.gop_size = genfs_size,
152 	.gop_alloc = ext2fs_gop_alloc,
153 	.gop_write = genfs_gop_write,
154 	.gop_markupdate = ufs_gop_markupdate,
155 };
156 
157 static const struct ufs_ops ext2fs_ufsops = {
158 	.uo_itimes = ext2fs_itimes,
159 	.uo_update = ext2fs_update,
160 	.uo_bufrd = ext2fs_bufrd,
161 	.uo_bufwr = ext2fs_bufwr,
162 };
163 
164 /* Fill in the inode uid/gid from ext2 halves.  */
165 void
166 ext2fs_set_inode_guid(struct inode *ip)
167 {
168 
169 	ip->i_gid = ip->i_e2fs_gid;
170 	ip->i_uid = ip->i_e2fs_uid;
171 	if (ip->i_e2fs->e2fs.e2fs_rev > E2FS_REV0) {
172 		ip->i_gid |= ip->i_e2fs_gid_high << 16;
173 		ip->i_uid |= ip->i_e2fs_uid_high << 16;
174 	}
175 }
176 
177 static int
178 ext2fs_modcmd(modcmd_t cmd, void *arg)
179 {
180 	int error;
181 
182 	switch (cmd) {
183 	case MODULE_CMD_INIT:
184 		error = vfs_attach(&ext2fs_vfsops);
185 		if (error != 0)
186 			break;
187 		sysctl_createv(&ext2fs_sysctl_log, 0, NULL, NULL,
188 			       CTLFLAG_PERMANENT,
189 			       CTLTYPE_NODE, "ext2fs",
190 			       SYSCTL_DESCR("Linux EXT2FS file system"),
191 			       NULL, 0, NULL, 0,
192 			       CTL_VFS, 17, CTL_EOL);
193 		/*
194 		 * XXX the "17" above could be dynamic, thereby eliminating
195 		 * one more instance of the "number to vfs" mapping problem,
196 		 * but "17" is the order as taken from sys/mount.h
197 		 */
198 		break;
199 	case MODULE_CMD_FINI:
200 		error = vfs_detach(&ext2fs_vfsops);
201 		if (error != 0)
202 			break;
203 		sysctl_teardown(&ext2fs_sysctl_log);
204 		break;
205 	default:
206 		error = ENOTTY;
207 		break;
208 	}
209 
210 	return error;
211 }
212 
213 /*
214  * XXX Same structure as FFS inodes?  Should we share a common pool?
215  */
216 struct pool ext2fs_inode_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 	ufs_init();
227 }
228 
229 void
230 ext2fs_reinit(void)
231 {
232 	ufs_reinit();
233 }
234 
235 void
236 ext2fs_done(void)
237 {
238 
239 	ufs_done();
240 	pool_destroy(&ext2fs_inode_pool);
241 }
242 
243 static void
244 ext2fs_sb_setmountinfo(struct m_ext2fs *fs, struct mount *mp)
245 {
246 	(void)strlcpy(fs->e2fs_fsmnt, mp->mnt_stat.f_mntonname,
247             sizeof(fs->e2fs_fsmnt));
248 	if (fs->e2fs_ronly == 0 && fs->e2fs.e2fs_rev > E2FS_REV0) {
249 		(void)strlcpy(fs->e2fs.e2fs_fsmnt, mp->mnt_stat.f_mntonname,
250 		    sizeof(fs->e2fs.e2fs_fsmnt));
251 
252 		fs->e2fs.e2fs_mtime = time_second;
253 		fs->e2fs.e2fs_mnt_count++;
254 
255 		fs->e2fs_fmod = 1;
256 	}
257 }
258 
259 /*
260  * Called by main() when ext2fs is going to be mounted as root.
261  *
262  * Name is updated by mount(8) after booting.
263  */
264 
265 int
266 ext2fs_mountroot(void)
267 {
268 	extern struct vnode *rootvp;
269 	struct m_ext2fs *fs;
270 	struct mount *mp;
271 	struct ufsmount *ump;
272 	int error;
273 
274 	if (device_class(root_device) != DV_DISK)
275 		return ENODEV;
276 
277 	if ((error = vfs_rootmountalloc(MOUNT_EXT2FS, "root_device", &mp))) {
278 		vrele(rootvp);
279 		return error;
280 	}
281 
282 	if ((error = ext2fs_mountfs(rootvp, mp)) != 0) {
283 		vfs_unbusy(mp);
284 		vfs_rele(mp);
285 		return error;
286 	}
287 	mountlist_append(mp);
288 	ump = VFSTOUFS(mp);
289 	fs = ump->um_e2fs;
290 	ext2fs_sb_setmountinfo(fs, mp);
291 	(void)ext2fs_statvfs(mp, &mp->mnt_stat);
292 	vfs_unbusy(mp);
293 	setrootfstime((time_t)fs->e2fs.e2fs_wtime);
294 	return 0;
295 }
296 
297 /*
298  * VFS Operations.
299  *
300  * mount system call
301  */
302 int
303 ext2fs_mount(struct mount *mp, const char *path, void *data, size_t *data_len)
304 {
305 	struct lwp *l = curlwp;
306 	struct vnode *devvp;
307 	struct ufs_args *args = data;
308 	struct ufsmount *ump = NULL;
309 	struct m_ext2fs *fs;
310 	int error = 0, flags, update;
311 	mode_t accessmode;
312 
313 	if (args == NULL)
314 		return EINVAL;
315 	if (*data_len < sizeof *args)
316 		return EINVAL;
317 
318 	if (mp->mnt_flag & MNT_GETARGS) {
319 		ump = VFSTOUFS(mp);
320 		if (ump == NULL)
321 			return EIO;
322 		memset(args, 0, sizeof *args);
323 		args->fspec = NULL;
324 		*data_len = sizeof *args;
325 		return 0;
326 	}
327 
328 	update = mp->mnt_flag & MNT_UPDATE;
329 
330 	/* Check arguments */
331 	if (args->fspec != NULL) {
332 		/*
333 		 * Look up the name and verify that it's sane.
334 		 */
335 		error = namei_simple_user(args->fspec,
336 					NSM_FOLLOW_NOEMULROOT, &devvp);
337 		if (error != 0)
338 			return error;
339 
340 		if (!update) {
341 			/*
342 			 * Be sure this is a valid block device
343 			 */
344 			if (devvp->v_type != VBLK)
345 				error = ENOTBLK;
346 			else if (bdevsw_lookup(devvp->v_rdev) == NULL)
347 				error = ENXIO;
348 		} else {
349 		        /*
350 			 * Be sure we're still naming the same device
351 			 * used for our initial mount
352 			 */
353 			ump = VFSTOUFS(mp);
354 			if (devvp != ump->um_devvp) {
355 				if (devvp->v_rdev != ump->um_devvp->v_rdev)
356 					error = EINVAL;
357 				else {
358 					vrele(devvp);
359 					devvp = ump->um_devvp;
360 					vref(devvp);
361 				}
362 			}
363 		}
364 	} else {
365 		if (!update) {
366 			/* New mounts must have a filename for the device */
367 			return EINVAL;
368 		} else {
369 			ump = VFSTOUFS(mp);
370 			devvp = ump->um_devvp;
371 			vref(devvp);
372 		}
373 	}
374 
375 	/*
376 	 * If mount by non-root, then verify that user has necessary
377 	 * permissions on the device.
378 	 *
379 	 * Permission to update a mount is checked higher, so here we presume
380 	 * updating the mount is okay (for example, as far as securelevel goes)
381 	 * which leaves us with the normal check.
382 	 */
383 	if (error == 0) {
384 		accessmode = VREAD;
385 		if (update ?
386 		    (mp->mnt_iflag & IMNT_WANTRDWR) != 0 :
387 		    (mp->mnt_flag & MNT_RDONLY) == 0)
388 			accessmode |= VWRITE;
389 		vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
390 		error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT,
391 		    KAUTH_REQ_SYSTEM_MOUNT_DEVICE, mp, devvp,
392 		    KAUTH_ARG(accessmode));
393 		VOP_UNLOCK(devvp);
394 	}
395 
396 	if (error) {
397 		vrele(devvp);
398 		return error;
399 	}
400 
401 	if (!update) {
402 		int xflags;
403 
404 		if (mp->mnt_flag & MNT_RDONLY)
405 			xflags = FREAD;
406 		else
407 			xflags = FREAD|FWRITE;
408 		vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
409 		error = VOP_OPEN(devvp, xflags, FSCRED);
410 		VOP_UNLOCK(devvp);
411 		if (error)
412 			goto fail;
413 		error = ext2fs_mountfs(devvp, mp);
414 		if (error) {
415 			vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
416 			(void)VOP_CLOSE(devvp, xflags, NOCRED);
417 			VOP_UNLOCK(devvp);
418 			goto fail;
419 		}
420 
421 		ump = VFSTOUFS(mp);
422 		fs = ump->um_e2fs;
423 	} else {
424 		/*
425 		 * Update the mount.
426 		 */
427 
428 		/*
429 		 * The initial mount got a reference on this
430 		 * device, so drop the one obtained via
431 		 * namei(), above.
432 		 */
433 		vrele(devvp);
434 
435 		ump = VFSTOUFS(mp);
436 		fs = ump->um_e2fs;
437 		if (fs->e2fs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) {
438 			/*
439 			 * Changing from r/w to r/o
440 			 */
441 			flags = WRITECLOSE;
442 			if (mp->mnt_flag & MNT_FORCE)
443 				flags |= FORCECLOSE;
444 			error = ext2fs_flushfiles(mp, flags);
445 			if (error == 0 &&
446 			    ext2fs_cgupdate(ump, MNT_WAIT) == 0 &&
447 			    (fs->e2fs.e2fs_state & E2FS_ERRORS) == 0) {
448 				fs->e2fs.e2fs_state = E2FS_ISCLEAN;
449 				(void) ext2fs_sbupdate(ump, MNT_WAIT);
450 			}
451 			if (error)
452 				return error;
453 			fs->e2fs_ronly = 1;
454 		}
455 
456 		if (mp->mnt_flag & MNT_RELOAD) {
457 			error = ext2fs_reload(mp, l->l_cred, l);
458 			if (error)
459 				return error;
460 		}
461 
462 		if (fs->e2fs_ronly && (mp->mnt_iflag & IMNT_WANTRDWR)) {
463 			/*
464 			 * Changing from read-only to read/write
465 			 */
466 			fs->e2fs_ronly = 0;
467 			if (fs->e2fs.e2fs_state == E2FS_ISCLEAN)
468 				fs->e2fs.e2fs_state = 0;
469 			else
470 				fs->e2fs.e2fs_state = E2FS_ERRORS;
471 			fs->e2fs_fmod = 1;
472 		}
473 		if (args->fspec == NULL)
474 			return 0;
475 	}
476 
477 	error = set_statvfs_info(path, UIO_USERSPACE, args->fspec,
478 	    UIO_USERSPACE, mp->mnt_op->vfs_name, mp, l);
479 	if (error == 0)
480 		ext2fs_sb_setmountinfo(fs, mp);
481 
482 	if (fs->e2fs_fmod != 0) {	/* XXX */
483 		fs->e2fs_fmod = 0;
484 		if (fs->e2fs.e2fs_state == 0)
485 			fs->e2fs.e2fs_wtime = time_second;
486 		else
487 			printf("%s: file system not clean; please fsck(8)\n",
488 				mp->mnt_stat.f_mntfromname);
489 		(void) ext2fs_cgupdate(ump, MNT_WAIT);
490 	}
491 	return error;
492 
493 fail:
494 	vrele(devvp);
495 	return error;
496 }
497 
498 /*
499  * Sanity check the disk vnode content, and copy it over to inode structure.
500  */
501 static int
502 ext2fs_loadvnode_content(struct m_ext2fs *fs, ino_t ino, struct buf *bp, struct inode *ip)
503 {
504 	struct ext2fs_dinode *din;
505 	int error = 0;
506 
507 	din = (struct ext2fs_dinode *)((char *)bp->b_data + (ino_to_fsbo(fs, ino) * EXT2_DINODE_SIZE(fs)));
508 
509 	/* sanity checks - inode data NOT byteswapped at this point */
510 	if (EXT2_DINODE_FITS(din, e2di_extra_isize, EXT2_DINODE_SIZE(fs))
511 	    && (EXT2_DINODE_SIZE(fs) - EXT2_REV0_DINODE_SIZE) < fs2h16(din->e2di_extra_isize))
512 	{
513 		printf("ext2fs: inode %"PRIu64" bad extra_isize %u",
514 			ino, din->e2di_extra_isize);
515 		error = EINVAL;
516 		goto bad;
517 	}
518 
519 	/* everything allright, proceed with copy */
520 	if (ip->i_din.e2fs_din == NULL)
521 		ip->i_din.e2fs_din = kmem_alloc(EXT2_DINODE_SIZE(fs), KM_SLEEP);
522 
523 	e2fs_iload(din, ip->i_din.e2fs_din, EXT2_DINODE_SIZE(fs));
524 
525 	ext2fs_set_inode_guid(ip);
526 
527     bad:
528 	return error;
529 }
530 
531 /*
532  * Reload all incore data for a filesystem (used after running fsck on
533  * the root filesystem and finding things to fix). The filesystem must
534  * be mounted read-only.
535  *
536  * Things to do to update the mount:
537  *	1) invalidate all cached meta-data.
538  *	2) re-read superblock from disk.
539  *	3) re-read summary information from disk.
540  *	4) invalidate all inactive vnodes.
541  *	5) invalidate all cached file data.
542  *	6) re-read inode data for all active vnodes.
543  */
544 int
545 ext2fs_reload(struct mount *mp, kauth_cred_t cred, struct lwp *l)
546 {
547 	struct vnode *vp, *devvp;
548 	struct inode *ip;
549 	struct buf *bp;
550 	struct m_ext2fs *fs;
551 	struct ext2fs *newfs;
552 	int i, error;
553 	struct ufsmount *ump;
554 	struct vnode_iterator *marker;
555 
556 	if ((mp->mnt_flag & MNT_RDONLY) == 0)
557 		return EINVAL;
558 
559 	ump = VFSTOUFS(mp);
560 	/*
561 	 * Step 1: invalidate all cached meta-data.
562 	 */
563 	devvp = ump->um_devvp;
564 	vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
565 	error = vinvalbuf(devvp, 0, cred, l, 0, 0);
566 	VOP_UNLOCK(devvp);
567 	if (error)
568 		panic("ext2fs_reload: dirty1");
569 
570 	fs = ump->um_e2fs;
571 	/*
572 	 * Step 2: re-read superblock from disk. Copy in new superblock, and compute
573 	 * in-memory values.
574 	 */
575 	error = bread(devvp, SBLOCK, SBSIZE, 0, &bp);
576 	if (error)
577 		return error;
578 	newfs = (struct ext2fs *)bp->b_data;
579 	e2fs_sbload(newfs, &fs->e2fs);
580 
581 	brelse(bp, 0);
582 
583 	error = ext2fs_sbfill(fs, (mp->mnt_flag & MNT_RDONLY) != 0);
584 	if (error)
585 		return error;
586 
587 	/*
588 	 * Step 3: re-read summary information from disk.
589 	 */
590 	for (i = 0; i < fs->e2fs_ngdb; i++) {
591 		error = bread(devvp ,
592 		    EXT2_FSBTODB(fs, fs->e2fs.e2fs_first_dblock +
593 		    1 /* superblock */ + i),
594 		    fs->e2fs_bsize, 0, &bp);
595 		if (error) {
596 			return error;
597 		}
598 		e2fs_cgload((struct ext2_gd *)bp->b_data,
599 		    &fs->e2fs_gd[i * fs->e2fs_bsize / sizeof(struct ext2_gd)],
600 		    fs->e2fs_bsize);
601 		brelse(bp, 0);
602 	}
603 
604 	vfs_vnode_iterator_init(mp, &marker);
605 	while ((vp = vfs_vnode_iterator_next(marker, NULL, NULL))) {
606 		/*
607 		 * Step 4: invalidate all inactive vnodes.
608 		 */
609 		if (vrecycle(vp))
610 			continue;
611 		/*
612 		 * Step 5: invalidate all cached file data.
613 		 */
614 		if (vn_lock(vp, LK_EXCLUSIVE)) {
615 			vrele(vp);
616 			continue;
617 		}
618 		if (vinvalbuf(vp, 0, cred, l, 0, 0))
619 			panic("ext2fs_reload: dirty2");
620 		/*
621 		 * Step 6: re-read inode data for all active vnodes.
622 		 */
623 		ip = VTOI(vp);
624 		error = bread(devvp, EXT2_FSBTODB(fs, ino_to_fsba(fs, ip->i_number)),
625 		    (int)fs->e2fs_bsize, 0, &bp);
626 		if (error) {
627 			vput(vp);
628 			break;
629 		}
630 		error = ext2fs_loadvnode_content(fs, ip->i_number, bp, ip);
631 		brelse(bp, 0);
632 		if (error) {
633 			vput(vp);
634 			break;
635 		}
636 
637 		vput(vp);
638 	}
639 	vfs_vnode_iterator_destroy(marker);
640 	return error;
641 }
642 
643 /*
644  * Common code for mount and mountroot
645  */
646 int
647 ext2fs_mountfs(struct vnode *devvp, struct mount *mp)
648 {
649 	struct lwp *l = curlwp;
650 	struct ufsmount *ump;
651 	struct buf *bp;
652 	struct ext2fs *fs;
653 	struct m_ext2fs *m_fs;
654 	dev_t dev;
655 	int error, i, ronly;
656 	kauth_cred_t cred;
657 
658 	dev = devvp->v_rdev;
659 	cred = l->l_cred;
660 
661 	/* Flush out any old buffers remaining from a previous use. */
662 	vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
663 	error = vinvalbuf(devvp, V_SAVE, cred, l, 0, 0);
664 	VOP_UNLOCK(devvp);
665 	if (error)
666 		return error;
667 
668 	ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
669 
670 	bp = NULL;
671 	ump = NULL;
672 
673 	/* Read the superblock from disk, and swap it directly. */
674 	error = bread(devvp, SBLOCK, SBSIZE, 0, &bp);
675 	if (error)
676 		goto out;
677 	fs = (struct ext2fs *)bp->b_data;
678 	m_fs = kmem_zalloc(sizeof(*m_fs), KM_SLEEP);
679 	e2fs_sbload(fs, &m_fs->e2fs);
680 
681 	brelse(bp, 0);
682 	bp = NULL;
683 
684 	/* Once swapped, validate and fill in the superblock. */
685 	error = ext2fs_sbfill(m_fs, ronly);
686 	if (error) {
687 		kmem_free(m_fs, sizeof(*m_fs));
688 		goto out;
689 	}
690 	m_fs->e2fs_ronly = ronly;
691 
692 	ump = kmem_zalloc(sizeof(*ump), KM_SLEEP);
693 	ump->um_fstype = UFS1;
694 	ump->um_ops = &ext2fs_ufsops;
695 	ump->um_e2fs = m_fs;
696 
697 	if (ronly == 0) {
698 		if (m_fs->e2fs.e2fs_state == E2FS_ISCLEAN)
699 			m_fs->e2fs.e2fs_state = 0;
700 		else
701 			m_fs->e2fs.e2fs_state = E2FS_ERRORS;
702 		m_fs->e2fs_fmod = 1;
703 	}
704 
705 	/* XXX: should be added in ext2fs_sbfill()? */
706 	m_fs->e2fs_gd = kmem_alloc(m_fs->e2fs_ngdb * m_fs->e2fs_bsize, KM_SLEEP);
707 	for (i = 0; i < m_fs->e2fs_ngdb; i++) {
708 		error = bread(devvp,
709 		    EXT2_FSBTODB(m_fs, m_fs->e2fs.e2fs_first_dblock +
710 		    1 /* superblock */ + i),
711 		    m_fs->e2fs_bsize, 0, &bp);
712 		if (error) {
713 			kmem_free(m_fs->e2fs_gd,
714 			    m_fs->e2fs_ngdb * m_fs->e2fs_bsize);
715 			goto out;
716 		}
717 		e2fs_cgload((struct ext2_gd *)bp->b_data,
718 		    &m_fs->e2fs_gd[
719 			i * m_fs->e2fs_bsize / sizeof(struct ext2_gd)],
720 		    m_fs->e2fs_bsize);
721 		brelse(bp, 0);
722 		bp = NULL;
723 	}
724 
725 	error = ext2fs_cg_verify_and_initialize(devvp, m_fs, ronly);
726 	if (error) {
727 		kmem_free(m_fs->e2fs_gd, m_fs->e2fs_ngdb * m_fs->e2fs_bsize);
728 		goto out;
729 	}
730 
731 	mp->mnt_data = ump;
732 	mp->mnt_stat.f_fsidx.__fsid_val[0] = (long)dev;
733 	mp->mnt_stat.f_fsidx.__fsid_val[1] = makefstype(MOUNT_EXT2FS);
734 	mp->mnt_stat.f_fsid = mp->mnt_stat.f_fsidx.__fsid_val[0];
735 	mp->mnt_stat.f_namemax = EXT2FS_MAXNAMLEN;
736 	mp->mnt_flag |= MNT_LOCAL;
737 	mp->mnt_dev_bshift = DEV_BSHIFT;	/* XXX */
738 	mp->mnt_fs_bshift = m_fs->e2fs_bshift;
739 	mp->mnt_iflag |= IMNT_DTYPE;
740 	ump->um_flags = 0;
741 	ump->um_mountp = mp;
742 	ump->um_dev = dev;
743 	ump->um_devvp = devvp;
744 	ump->um_nindir = EXT2_NINDIR(m_fs);
745 	ump->um_lognindir = ffs(EXT2_NINDIR(m_fs)) - 1;
746 	ump->um_bptrtodb = m_fs->e2fs_fsbtodb;
747 	ump->um_seqinc = 1; /* no frags */
748 	ump->um_maxsymlinklen = EXT2_MAXSYMLINKLEN;
749 	ump->um_dirblksiz = m_fs->e2fs_bsize;
750 	ump->um_maxfilesize = ((uint64_t)0x80000000 * m_fs->e2fs_bsize - 1);
751 	spec_node_setmountedfs(devvp, mp);
752 	return 0;
753 
754 out:
755 	if (bp != NULL)
756 		brelse(bp, 0);
757 	if (ump) {
758 		kmem_free(ump->um_e2fs, sizeof(*m_fs));
759 		kmem_free(ump, sizeof(*ump));
760 		mp->mnt_data = NULL;
761 	}
762 	return error;
763 }
764 
765 /*
766  * unmount system call
767  */
768 int
769 ext2fs_unmount(struct mount *mp, int mntflags)
770 {
771 	struct ufsmount *ump;
772 	struct m_ext2fs *fs;
773 	int error, flags;
774 
775 	flags = 0;
776 	if (mntflags & MNT_FORCE)
777 		flags |= FORCECLOSE;
778 	if ((error = ext2fs_flushfiles(mp, flags)) != 0)
779 		return error;
780 	ump = VFSTOUFS(mp);
781 	fs = ump->um_e2fs;
782 	if (fs->e2fs_ronly == 0 &&
783 		ext2fs_cgupdate(ump, MNT_WAIT) == 0 &&
784 		(fs->e2fs.e2fs_state & E2FS_ERRORS) == 0) {
785 		fs->e2fs.e2fs_state = E2FS_ISCLEAN;
786 		(void) ext2fs_sbupdate(ump, MNT_WAIT);
787 	}
788 	if (ump->um_devvp->v_type != VBAD)
789 		spec_node_setmountedfs(ump->um_devvp, NULL);
790 	vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
791 	error = VOP_CLOSE(ump->um_devvp, fs->e2fs_ronly ? FREAD : FREAD|FWRITE,
792 	    NOCRED);
793 	vput(ump->um_devvp);
794 	kmem_free(fs->e2fs_gd, fs->e2fs_ngdb * fs->e2fs_bsize);
795 	kmem_free(fs, sizeof(*fs));
796 	kmem_free(ump, sizeof(*ump));
797 	mp->mnt_data = NULL;
798 	mp->mnt_flag &= ~MNT_LOCAL;
799 	return error;
800 }
801 
802 /*
803  * Flush out all the files in a filesystem.
804  */
805 int
806 ext2fs_flushfiles(struct mount *mp, int flags)
807 {
808 	extern int doforce;
809 	int error;
810 
811 	if (!doforce)
812 		flags &= ~FORCECLOSE;
813 	error = vflush(mp, NULLVP, flags);
814 	return error;
815 }
816 
817 /*
818  * Get file system statistics.
819  */
820 int
821 ext2fs_statvfs(struct mount *mp, struct statvfs *sbp)
822 {
823 	struct ufsmount *ump;
824 	struct m_ext2fs *fs;
825 	uint32_t overhead, overhead_per_group, ngdb;
826 	int i, ngroups;
827 
828 	ump = VFSTOUFS(mp);
829 	fs = ump->um_e2fs;
830 	if (fs->e2fs.e2fs_magic != E2FS_MAGIC)
831 		panic("ext2fs_statvfs");
832 
833 	/*
834 	 * Compute the overhead (FS structures)
835 	 */
836 	overhead_per_group =
837 	    1 /* block bitmap */ +
838 	    1 /* inode bitmap */ +
839 	    fs->e2fs_itpg;
840 	overhead = fs->e2fs.e2fs_first_dblock +
841 	    fs->e2fs_ncg * overhead_per_group;
842 	if (EXT2F_HAS_COMPAT_FEATURE(fs, EXT2F_COMPAT_SPARSESUPER2)) {
843 		/*
844 		 * Superblock and group descriptions is in group zero,
845 		 * then optionally 0, 1 or 2 extra copies.
846 		 */
847 		ngroups = 1
848 			+ (fs->e2fs.e4fs_backup_bgs[0] ? 1 : 0)
849 			+ (fs->e2fs.e4fs_backup_bgs[1] ? 1 : 0);
850 	} else if (EXT2F_HAS_ROCOMPAT_FEATURE(fs, EXT2F_ROCOMPAT_SPARSESUPER)) {
851 		for (i = 0, ngroups = 0; i < fs->e2fs_ncg; i++) {
852 			if (cg_has_sb(i))
853 				ngroups++;
854 		}
855 	} else {
856 		ngroups = fs->e2fs_ncg;
857 	}
858 	ngdb = fs->e2fs_ngdb;
859 	if (EXT2F_HAS_COMPAT_FEATURE(fs, EXT2F_COMPAT_RESIZE))
860 		ngdb += fs->e2fs.e2fs_reserved_ngdb;
861 	overhead += ngroups * (1 /* superblock */ + ngdb);
862 
863 	sbp->f_bsize = fs->e2fs_bsize;
864 	sbp->f_frsize = MINBSIZE << fs->e2fs.e2fs_fsize;
865 	sbp->f_iosize = fs->e2fs_bsize;
866 	sbp->f_blocks = fs->e2fs.e2fs_bcount - overhead;
867 	sbp->f_bfree = fs->e2fs.e2fs_fbcount;
868 	sbp->f_bresvd = fs->e2fs.e2fs_rbcount;
869 	if (sbp->f_bfree > sbp->f_bresvd)
870 		sbp->f_bavail = sbp->f_bfree - sbp->f_bresvd;
871 	else
872 		sbp->f_bavail = 0;
873 	sbp->f_files =  fs->e2fs.e2fs_icount;
874 	sbp->f_ffree = fs->e2fs.e2fs_ficount;
875 	sbp->f_favail = fs->e2fs.e2fs_ficount;
876 	sbp->f_fresvd = 0;
877 	copy_statvfs_info(sbp, mp);
878 	return 0;
879 }
880 
881 static bool
882 ext2fs_sync_selector(void *cl, struct vnode *vp)
883 {
884 	struct inode *ip;
885 
886 	KASSERT(mutex_owned(vp->v_interlock));
887 
888 	ip = VTOI(vp);
889 	/*
890 	 * Skip the vnode/inode if inaccessible.
891 	 */
892 	if (ip == NULL || vp->v_type == VNON)
893 		return false;
894 
895 	if (((ip->i_flag &
896 	      (IN_CHANGE | IN_UPDATE | IN_MODIFIED)) == 0 &&
897 	     LIST_EMPTY(&vp->v_dirtyblkhd) &&
898 	     UVM_OBJ_IS_CLEAN(&vp->v_uobj)))
899 		return false;
900 	return true;
901 }
902 
903 /*
904  * Go through the disk queues to initiate sandbagged IO;
905  * go through the inodes to write those that have been modified;
906  * initiate the writing of the super block if it has been modified.
907  *
908  * Note: we are always called with the filesystem marked `MPBUSY'.
909  */
910 int
911 ext2fs_sync(struct mount *mp, int waitfor, kauth_cred_t cred)
912 {
913 	struct vnode *vp;
914 	struct ufsmount *ump = VFSTOUFS(mp);
915 	struct m_ext2fs *fs;
916 	struct vnode_iterator *marker;
917 	int error, allerror = 0;
918 
919 	fs = ump->um_e2fs;
920 	if (fs->e2fs_fmod != 0 && fs->e2fs_ronly != 0) {	/* XXX */
921 		printf("fs = %s\n", fs->e2fs_fsmnt);
922 		panic("update: rofs mod");
923 	}
924 
925 	/*
926 	 * Write back each (modified) inode.
927 	 */
928 	vfs_vnode_iterator_init(mp, &marker);
929 	while ((vp = vfs_vnode_iterator_next(marker, ext2fs_sync_selector,
930 	    NULL)))
931 	{
932 		error = vn_lock(vp, LK_EXCLUSIVE);
933 		if (error) {
934 			vrele(vp);
935 			continue;
936 		}
937 		if (vp->v_type == VREG && waitfor == MNT_LAZY)
938 			error = ext2fs_update(vp, NULL, NULL, 0);
939 		else
940 			error = VOP_FSYNC(vp, cred,
941 			    waitfor == MNT_WAIT ? FSYNC_WAIT : 0, 0, 0);
942 		if (error)
943 			allerror = error;
944 		vput(vp);
945 	}
946 	vfs_vnode_iterator_destroy(marker);
947 	/*
948 	 * Force stale file system control information to be flushed.
949 	 */
950 	if (waitfor != MNT_LAZY) {
951 		vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
952 		if ((error = VOP_FSYNC(ump->um_devvp, cred,
953 		    waitfor == MNT_WAIT ? FSYNC_WAIT : 0, 0, 0)) != 0)
954 			allerror = error;
955 		VOP_UNLOCK(ump->um_devvp);
956 	}
957 	/*
958 	 * Write back modified superblock.
959 	 */
960 	if (fs->e2fs_fmod != 0) {
961 		fs->e2fs_fmod = 0;
962 		fs->e2fs.e2fs_wtime = time_second;
963 		if ((error = ext2fs_cgupdate(ump, waitfor)))
964 			allerror = error;
965 	}
966 	return allerror;
967 }
968 
969 /*
970  * Load inode from disk and initialize vnode.
971  */
972 static int
973 ext2fs_init_vnode(struct ufsmount *ump, struct vnode *vp, ino_t ino)
974 {
975 	struct m_ext2fs *fs;
976 	struct inode *ip;
977 	struct buf *bp;
978 	int error;
979 
980 	fs = ump->um_e2fs;
981 
982 	/* Read in the disk contents for the inode, copy into the inode. */
983 	error = bread(ump->um_devvp, EXT2_FSBTODB(fs, ino_to_fsba(fs, ino)),
984 	    (int)fs->e2fs_bsize, 0, &bp);
985 	if (error)
986 		return error;
987 
988 	/* Allocate and initialize inode. */
989 	ip = pool_get(&ext2fs_inode_pool, PR_WAITOK);
990 	memset(ip, 0, sizeof(struct inode));
991 	ip->i_vnode = vp;
992 	ip->i_ump = ump;
993 	ip->i_e2fs = fs;
994 	ip->i_dev = ump->um_dev;
995 	ip->i_number = ino;
996 	ip->i_e2fs_last_lblk = 0;
997 	ip->i_e2fs_last_blk = 0;
998 
999 	error = ext2fs_loadvnode_content(fs, ino, bp, ip);
1000 	brelse(bp, 0);
1001 	if (error) {
1002 		pool_put(&ext2fs_inode_pool, ip);
1003 		return error;
1004 	}
1005 
1006 	/* If the inode was deleted, reset all fields */
1007 	if (ip->i_e2fs_dtime != 0) {
1008 		ip->i_e2fs_mode = 0;
1009 		(void)ext2fs_setsize(ip, 0);
1010 		(void)ext2fs_setnblock(ip, 0);
1011 		memset(ip->i_e2fs_blocks, 0, sizeof(ip->i_e2fs_blocks));
1012 	}
1013 
1014 	/* Initialise vnode with this inode. */
1015 	vp->v_tag = VT_EXT2FS;
1016 	vp->v_op = ext2fs_vnodeop_p;
1017 	vp->v_vflag |= VV_LOCKSWORK;
1018 	vp->v_data = ip;
1019 
1020 	/* Initialize genfs node. */
1021 	genfs_node_init(vp, &ext2fs_genfsops);
1022 
1023 	return 0;
1024 }
1025 
1026 /*
1027  * Read an inode from disk and initialize this vnode / inode pair.
1028  * Caller assures no other thread will try to load this inode.
1029  */
1030 int
1031 ext2fs_loadvnode(struct mount *mp, struct vnode *vp,
1032     const void *key, size_t key_len, const void **new_key)
1033 {
1034 	ino_t ino;
1035 	struct inode *ip;
1036 	struct ufsmount *ump;
1037 	int error;
1038 
1039 	KASSERT(key_len == sizeof(ino));
1040 	memcpy(&ino, key, key_len);
1041 	ump = VFSTOUFS(mp);
1042 
1043 	error = ext2fs_init_vnode(ump, vp, ino);
1044 	if (error)
1045 		return error;
1046 
1047 	ip = VTOI(vp);
1048 
1049 	/* Initialize the vnode from the inode. */
1050 	ext2fs_vinit(mp, ext2fs_specop_p, ext2fs_fifoop_p, &vp);
1051 
1052 	/* Finish inode initialization. */
1053 	ip->i_devvp = ump->um_devvp;
1054 	vref(ip->i_devvp);
1055 
1056 	/*
1057 	 * Set up a generation number for this inode if it does not
1058 	 * already have one. This should only happen on old filesystems.
1059 	 */
1060 
1061 	if (ip->i_e2fs_gen == 0) {
1062 		if (++ext2gennumber < (u_long)time_second)
1063 			ext2gennumber = time_second;
1064 		ip->i_e2fs_gen = ext2gennumber;
1065 		if ((mp->mnt_flag & MNT_RDONLY) == 0)
1066 			ip->i_flag |= IN_MODIFIED;
1067 	}
1068 	uvm_vnp_setsize(vp, ext2fs_size(ip));
1069 	*new_key = &ip->i_number;
1070 	return 0;
1071 }
1072 
1073 /*
1074  * Create a new inode on disk and initialize this vnode / inode pair.
1075  */
1076 int
1077 ext2fs_newvnode(struct mount *mp, struct vnode *dvp, struct vnode *vp,
1078     struct vattr *vap, kauth_cred_t cred,
1079     size_t *key_len, const void **new_key)
1080 {
1081 	ino_t ino;
1082 	struct inode *ip, *pdir;
1083 	struct m_ext2fs *fs;
1084 	struct ufsmount *ump;
1085 	int error, mode;
1086 
1087 	KASSERT(dvp->v_mount == mp);
1088 	KASSERT(vap->va_type != VNON);
1089 
1090 	*key_len = sizeof(ino);
1091 
1092 	pdir = VTOI(dvp);
1093 	fs = pdir->i_e2fs;
1094 	ump = VFSTOUFS(mp);
1095 	mode = MAKEIMODE(vap->va_type, vap->va_mode);
1096 
1097 	/* Allocate fresh inode. */
1098 	error = ext2fs_valloc(dvp, mode, cred, &ino);
1099 	if (error)
1100 		return error;
1101 
1102 	/* Attach inode to vnode. */
1103 	error = ext2fs_init_vnode(ump, vp, ino);
1104 	if (error) {
1105 		ext2fs_vfree(dvp, ino, mode);
1106 		return error;
1107 	}
1108 
1109 	ip = VTOI(vp);
1110 
1111 	KASSERT(!E2FS_HAS_GD_CSUM(fs) || (fs->e2fs_gd[ino_to_cg(fs, ino)].ext2bgd_flags & h2fs16(E2FS_BG_INODE_ZEROED)) != 0);
1112 
1113 	/* check for already used inode; makes sense only for ZEROED itable */
1114 	if (__predict_false(ip->i_e2fs_mode && ip->i_e2fs_nlink != 0)) {
1115 		printf("mode = 0%o, nlinks %d, inum = %llu, fs = %s\n",
1116 		    ip->i_e2fs_mode, ip->i_e2fs_nlink,
1117 		    (unsigned long long)ip->i_number, fs->e2fs_fsmnt);
1118 		panic("ext2fs_valloc: dup alloc");
1119 	}
1120 
1121 	memset(ip->i_din.e2fs_din, 0, EXT2_DINODE_SIZE(fs));
1122 
1123 	/*
1124 	 * Set up a new generation number for this inode.
1125 	 */
1126 	if (++ext2gennumber < time_second)
1127 		ext2gennumber = time_second;
1128 	ip->i_e2fs_gen = ext2gennumber;
1129 
1130 	ip->i_uid = kauth_cred_geteuid(cred);
1131 	ip->i_e2fs_uid = ip->i_uid & 0xffff;
1132 	ip->i_e2fs_gid = pdir->i_e2fs_gid;
1133 	if (ip->i_e2fs->e2fs.e2fs_rev > E2FS_REV0) {
1134 		ip->i_e2fs_uid_high = (ip->i_uid >> 16) & 0xffff;
1135 		ip->i_e2fs_gid_high = pdir->i_e2fs_gid_high;
1136 	} else {
1137 		ip->i_e2fs_uid_high = 0;
1138 		ip->i_e2fs_gid_high = 0;
1139 	}
1140 	ip->i_gid = ip->i_e2fs_gid | (ip->i_e2fs_gid_high << 16);
1141 	ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
1142 	ip->i_e2fs_mode = mode;
1143 	vp->v_type = IFTOVT(mode);
1144 	ip->i_e2fs_nlink = 1;
1145 
1146 	/* Authorize setting SGID if needed. */
1147 	if (ip->i_e2fs_mode & ISGID) {
1148 		error = kauth_authorize_vnode(cred, KAUTH_VNODE_WRITE_SECURITY,
1149 		    vp, NULL, genfs_can_chmod(vp->v_type, cred, ip->i_uid,
1150 		    ip->i_gid, mode));
1151 		if (error)
1152 			ip->i_e2fs_mode &= ~ISGID;
1153 	}
1154 
1155 	/* Initialize extra_isize according to what is set in superblock */
1156 	if (EXT2F_HAS_ROCOMPAT_FEATURE(ip->i_e2fs, EXT2F_ROCOMPAT_EXTRA_ISIZE)
1157 	    && EXT2_DINODE_SIZE(ip->i_e2fs) > EXT2_REV0_DINODE_SIZE) {
1158 		ip->i_din.e2fs_din->e2di_extra_isize = ip->i_e2fs->e2fs.e4fs_want_extra_isize;
1159 	}
1160 
1161 	/* Set create time if possible */
1162 	if (EXT2_DINODE_FITS(ip->i_din.e2fs_din, e2di_crtime, EXT2_DINODE_SIZE(ip->i_e2fs))) {
1163 		struct timespec now;
1164 		vfs_timestamp(&now);
1165 		EXT2_DINODE_TIME_SET(&now, ip->i_din.e2fs_din, e2di_crtime, EXT2_DINODE_SIZE(ip->i_e2fs));
1166 	}
1167 
1168 	/* Initialize the vnode from the inode. */
1169 	ext2fs_vinit(mp, ext2fs_specop_p, ext2fs_fifoop_p, &vp);
1170 
1171 	/* Finish inode initialization. */
1172 	ip->i_devvp = ump->um_devvp;
1173 	vref(ip->i_devvp);
1174 
1175 	uvm_vnp_setsize(vp, ext2fs_size(ip));
1176 	*new_key = &ip->i_number;
1177 	return 0;
1178 }
1179 
1180 /*
1181  * File handle to vnode
1182  *
1183  * Have to be really careful about stale file handles:
1184  * - check that the inode number is valid
1185  * - call ext2fs_vget() to get the locked inode
1186  * - check for an unallocated inode (i_mode == 0)
1187  */
1188 int
1189 ext2fs_fhtovp(struct mount *mp, struct fid *fhp, struct vnode **vpp)
1190 {
1191 	struct inode *ip;
1192 	struct vnode *nvp;
1193 	int error;
1194 	struct ufid ufh;
1195 	struct m_ext2fs *fs;
1196 
1197 	if (fhp->fid_len != sizeof(struct ufid))
1198 		return EINVAL;
1199 
1200 	memcpy(&ufh, fhp, sizeof(struct ufid));
1201 	fs = VFSTOUFS(mp)->um_e2fs;
1202 	if ((ufh.ufid_ino < EXT2_FIRSTINO && ufh.ufid_ino != EXT2_ROOTINO) ||
1203 		ufh.ufid_ino >= fs->e2fs_ncg * fs->e2fs.e2fs_ipg)
1204 		return ESTALE;
1205 
1206 	if ((error = VFS_VGET(mp, ufh.ufid_ino, &nvp)) != 0) {
1207 		*vpp = NULLVP;
1208 		return error;
1209 	}
1210 	ip = VTOI(nvp);
1211 	if (ip->i_e2fs_mode == 0 || ip->i_e2fs_dtime != 0 ||
1212 		ip->i_e2fs_gen != ufh.ufid_gen) {
1213 		vput(nvp);
1214 		*vpp = NULLVP;
1215 		return ESTALE;
1216 	}
1217 	*vpp = nvp;
1218 	return 0;
1219 }
1220 
1221 /*
1222  * Vnode pointer to File handle
1223  */
1224 /* ARGSUSED */
1225 int
1226 ext2fs_vptofh(struct vnode *vp, struct fid *fhp, size_t *fh_size)
1227 {
1228 	struct inode *ip;
1229 	struct ufid ufh;
1230 
1231 	if (*fh_size < sizeof(struct ufid)) {
1232 		*fh_size = sizeof(struct ufid);
1233 		return E2BIG;
1234 	}
1235 	*fh_size = sizeof(struct ufid);
1236 
1237 	ip = VTOI(vp);
1238 	memset(&ufh, 0, sizeof(ufh));
1239 	ufh.ufid_len = sizeof(struct ufid);
1240 	ufh.ufid_ino = ip->i_number;
1241 	ufh.ufid_gen = ip->i_e2fs_gen;
1242 	memcpy(fhp, &ufh, sizeof(ufh));
1243 	return 0;
1244 }
1245 
1246 /*
1247  * Write a superblock and associated information back to disk.
1248  */
1249 int
1250 ext2fs_sbupdate(struct ufsmount *mp, int waitfor)
1251 {
1252 	struct m_ext2fs *fs = mp->um_e2fs;
1253 	struct buf *bp;
1254 	int error = 0;
1255 
1256 	bp = getblk(mp->um_devvp, SBLOCK, SBSIZE, 0, 0);
1257 	e2fs_sbsave(&fs->e2fs, (struct ext2fs*)bp->b_data);
1258 	if (waitfor == MNT_WAIT)
1259 		error = bwrite(bp);
1260 	else
1261 		bawrite(bp);
1262 	return error;
1263 }
1264 
1265 int
1266 ext2fs_cgupdate(struct ufsmount *mp, int waitfor)
1267 {
1268 	struct m_ext2fs *fs = mp->um_e2fs;
1269 	struct buf *bp;
1270 	int i, error = 0, allerror = 0;
1271 
1272 	allerror = ext2fs_sbupdate(mp, waitfor);
1273 	for (i = 0; i < fs->e2fs_ngdb; i++) {
1274 		bp = getblk(mp->um_devvp, EXT2_FSBTODB(fs,
1275 		    fs->e2fs.e2fs_first_dblock +
1276 		    1 /* superblock */ + i), fs->e2fs_bsize, 0, 0);
1277 		e2fs_cgsave(&fs->e2fs_gd[
1278 		    i * fs->e2fs_bsize / sizeof(struct ext2_gd)],
1279 		    (struct ext2_gd *)bp->b_data, fs->e2fs_bsize);
1280 		if (waitfor == MNT_WAIT)
1281 			error = bwrite(bp);
1282 		else
1283 			bawrite(bp);
1284 	}
1285 
1286 	if (!allerror && error)
1287 		allerror = error;
1288 	return allerror;
1289 }
1290 
1291 /*
1292  * Fill in the m_fs structure, and validate the fields of the superblock.
1293  * NOTE: here, the superblock is already swapped.
1294  */
1295 static int
1296 ext2fs_sbfill(struct m_ext2fs *m_fs, int ronly)
1297 {
1298 	uint32_t u32;
1299 	struct ext2fs *fs = &m_fs->e2fs;
1300 
1301 	/*
1302 	 * General sanity checks
1303 	 */
1304 	if (fs->e2fs_magic != E2FS_MAGIC)
1305 		return EINVAL;
1306 	if (fs->e2fs_rev > E2FS_REV1) {
1307 		printf("ext2fs: unsupported revision number: %x\n", fs->e2fs_rev);
1308 		return EINVAL;
1309 	}
1310 	if (fs->e2fs_log_bsize > 2) {
1311 		/* block size = 1024|2048|4096 */
1312 		printf("ext2fs: bad block size: %d\n", fs->e2fs_log_bsize);
1313 		return EINVAL;
1314 	}
1315 	if (fs->e2fs_bpg == 0) {
1316 		printf("ext2fs: zero blocks per group\n");
1317 		return EINVAL;
1318 	}
1319 	if (fs->e2fs_ipg == 0) {
1320 		printf("ext2fs: zero inodes per group\n");
1321 		return EINVAL;
1322 	}
1323 
1324 	if (fs->e2fs_first_dblock >= fs->e2fs_bcount) {
1325 		printf("ext2fs: invalid first data block\n");
1326 		return EINVAL;
1327 	}
1328 	if (fs->e2fs_rbcount > fs->e2fs_bcount ||
1329 	    fs->e2fs_fbcount > fs->e2fs_bcount) {
1330 		printf("ext2fs: invalid block count\n");
1331 		return EINVAL;
1332 	}
1333 
1334 	/*
1335 	 * Compute the fields of the superblock
1336 	 */
1337 	u32 = fs->e2fs_bcount - fs->e2fs_first_dblock; /* > 0 */
1338 	m_fs->e2fs_ncg = howmany(u32, fs->e2fs_bpg);
1339 	if (m_fs->e2fs_ncg == 0) {
1340 		printf("ext2fs: invalid number of cylinder groups\n");
1341 		return EINVAL;
1342 	}
1343 
1344 	m_fs->e2fs_fsbtodb = fs->e2fs_log_bsize + LOG_MINBSIZE - DEV_BSHIFT;
1345 	m_fs->e2fs_bsize = MINBSIZE << fs->e2fs_log_bsize;
1346 	m_fs->e2fs_bshift = LOG_MINBSIZE + fs->e2fs_log_bsize;
1347 	m_fs->e2fs_qbmask = m_fs->e2fs_bsize - 1;
1348 	m_fs->e2fs_bmask = ~m_fs->e2fs_qbmask;
1349 
1350 	if ((u32 = m_fs->e2fs_bsize / sizeof(struct ext2_gd)) == 0) {
1351 		/* Unlikely to happen */
1352 		printf("ext2fs: invalid block size\n");
1353 		return EINVAL;
1354 	}
1355 	m_fs->e2fs_ngdb = howmany(m_fs->e2fs_ncg, u32);
1356 	if (m_fs->e2fs_ngdb == 0) {
1357 		printf("ext2fs: invalid number of group descriptor blocks\n");
1358 		return EINVAL;
1359 	}
1360 
1361 	if (m_fs->e2fs_bsize < EXT2_DINODE_SIZE(m_fs)) {
1362 		printf("ext2fs: invalid inode size\n");
1363 		return EINVAL;
1364 	}
1365 	m_fs->e2fs_ipb = m_fs->e2fs_bsize / EXT2_DINODE_SIZE(m_fs);
1366 
1367 	m_fs->e2fs_itpg = fs->e2fs_ipg / m_fs->e2fs_ipb;
1368 
1369 	/*
1370 	 * Revision-specific checks
1371 	 */
1372 	if (fs->e2fs_rev > E2FS_REV0) {
1373 		char buf[256];
1374 		if (fs->e2fs_first_ino != EXT2_FIRSTINO) {
1375 			printf("ext2fs: unsupported first inode position\n");
1376 			return EINVAL;
1377 		}
1378 		u32 = fs->e2fs_features_incompat & ~EXT2F_INCOMPAT_SUPP;
1379 		if (u32) {
1380 			snprintb(buf, sizeof(buf), EXT2F_INCOMPAT_BITS, u32);
1381 			printf("ext2fs: unsupported incompat features: %s\n", buf);
1382 #ifndef EXT2_IGNORE_INCOMPAT_FEATURES
1383 			return EINVAL;
1384 #endif
1385 		}
1386 		u32 = fs->e2fs_features_rocompat & ~EXT2F_ROCOMPAT_SUPP;
1387 		if (!ronly && u32) {
1388 			snprintb(buf, sizeof(buf), EXT2F_ROCOMPAT_BITS, u32);
1389 			printf("ext2fs: unsupported ro-incompat features: %s\n",
1390 			    buf);
1391 #ifndef EXT2_IGNORE_ROCOMPAT_FEATURES
1392 			return EROFS;
1393 #endif
1394 		}
1395 		if (fs->e2fs_inode_size == 0 || !powerof2(fs->e2fs_inode_size) || fs->e2fs_inode_size > m_fs->e2fs_bsize) {
1396 			printf("ext2fs: bad inode size\n");
1397 			return EINVAL;
1398 		}
1399 	}
1400 
1401 	return 0;
1402 }
1403