xref: /netbsd-src/sys/ufs/ext2fs/ext2fs_vfsops.c (revision bdc22b2e01993381dcefeff2bc9b56ca75a4235c)
1 /*	$NetBSD: ext2fs_vfsops.c,v 1.211 2018/05/28 21:04:38 chs 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 1.211 2018/05/28 21:04:38 chs 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 	.gop_putrange = genfs_gop_putrange,
156 };
157 
158 static const struct ufs_ops ext2fs_ufsops = {
159 	.uo_itimes = ext2fs_itimes,
160 	.uo_update = ext2fs_update,
161 	.uo_bufrd = ext2fs_bufrd,
162 	.uo_bufwr = ext2fs_bufwr,
163 };
164 
165 /* Fill in the inode uid/gid from ext2 halves.  */
166 void
167 ext2fs_set_inode_guid(struct inode *ip)
168 {
169 
170 	ip->i_gid = ip->i_e2fs_gid;
171 	ip->i_uid = ip->i_e2fs_uid;
172 	if (ip->i_e2fs->e2fs.e2fs_rev > E2FS_REV0) {
173 		ip->i_gid |= ip->i_e2fs_gid_high << 16;
174 		ip->i_uid |= ip->i_e2fs_uid_high << 16;
175 	}
176 }
177 
178 static int
179 ext2fs_modcmd(modcmd_t cmd, void *arg)
180 {
181 	int error;
182 
183 	switch (cmd) {
184 	case MODULE_CMD_INIT:
185 		error = vfs_attach(&ext2fs_vfsops);
186 		if (error != 0)
187 			break;
188 		sysctl_createv(&ext2fs_sysctl_log, 0, NULL, NULL,
189 			       CTLFLAG_PERMANENT,
190 			       CTLTYPE_NODE, "ext2fs",
191 			       SYSCTL_DESCR("Linux EXT2FS file system"),
192 			       NULL, 0, NULL, 0,
193 			       CTL_VFS, 17, CTL_EOL);
194 		/*
195 		 * XXX the "17" above could be dynamic, thereby eliminating
196 		 * one more instance of the "number to vfs" mapping problem,
197 		 * but "17" is the order as taken from sys/mount.h
198 		 */
199 		break;
200 	case MODULE_CMD_FINI:
201 		error = vfs_detach(&ext2fs_vfsops);
202 		if (error != 0)
203 			break;
204 		sysctl_teardown(&ext2fs_sysctl_log);
205 		break;
206 	default:
207 		error = ENOTTY;
208 		break;
209 	}
210 
211 	return error;
212 }
213 
214 /*
215  * XXX Same structure as FFS inodes?  Should we share a common pool?
216  */
217 struct pool ext2fs_inode_pool;
218 
219 extern u_long ext2gennumber;
220 
221 void
222 ext2fs_init(void)
223 {
224 
225 	pool_init(&ext2fs_inode_pool, sizeof(struct inode), 0, 0, 0,
226 	    "ext2fsinopl", &pool_allocator_nointr, IPL_NONE);
227 	ufs_init();
228 }
229 
230 void
231 ext2fs_reinit(void)
232 {
233 	ufs_reinit();
234 }
235 
236 void
237 ext2fs_done(void)
238 {
239 
240 	ufs_done();
241 	pool_destroy(&ext2fs_inode_pool);
242 }
243 
244 static void
245 ext2fs_sb_setmountinfo(struct m_ext2fs *fs, struct mount *mp)
246 {
247 	(void)strlcpy(fs->e2fs_fsmnt, mp->mnt_stat.f_mntonname,
248             sizeof(fs->e2fs_fsmnt));
249 	if (fs->e2fs_ronly == 0 && fs->e2fs.e2fs_rev > E2FS_REV0) {
250 		(void)strlcpy(fs->e2fs.e2fs_fsmnt, mp->mnt_stat.f_mntonname,
251 		    sizeof(fs->e2fs.e2fs_fsmnt));
252 
253 		fs->e2fs.e2fs_mtime = time_second;
254 		fs->e2fs.e2fs_mnt_count++;
255 
256 		fs->e2fs_fmod = 1;
257 	}
258 }
259 
260 /*
261  * Called by main() when ext2fs is going to be mounted as root.
262  *
263  * Name is updated by mount(8) after booting.
264  */
265 
266 int
267 ext2fs_mountroot(void)
268 {
269 	extern struct vnode *rootvp;
270 	struct m_ext2fs *fs;
271 	struct mount *mp;
272 	struct ufsmount *ump;
273 	int error;
274 
275 	if (device_class(root_device) != DV_DISK)
276 		return ENODEV;
277 
278 	if ((error = vfs_rootmountalloc(MOUNT_EXT2FS, "root_device", &mp))) {
279 		vrele(rootvp);
280 		return error;
281 	}
282 
283 	if ((error = ext2fs_mountfs(rootvp, mp)) != 0) {
284 		vfs_unbusy(mp);
285 		vfs_rele(mp);
286 		return error;
287 	}
288 	mountlist_append(mp);
289 	ump = VFSTOUFS(mp);
290 	fs = ump->um_e2fs;
291 	ext2fs_sb_setmountinfo(fs, mp);
292 	(void)ext2fs_statvfs(mp, &mp->mnt_stat);
293 	vfs_unbusy(mp);
294 	setrootfstime((time_t)fs->e2fs.e2fs_wtime);
295 	return 0;
296 }
297 
298 /*
299  * VFS Operations.
300  *
301  * mount system call
302  */
303 int
304 ext2fs_mount(struct mount *mp, const char *path, void *data, size_t *data_len)
305 {
306 	struct lwp *l = curlwp;
307 	struct vnode *devvp;
308 	struct ufs_args *args = data;
309 	struct ufsmount *ump = NULL;
310 	struct m_ext2fs *fs;
311 	int error = 0, flags, update;
312 	mode_t accessmode;
313 
314 	if (args == NULL)
315 		return EINVAL;
316 	if (*data_len < sizeof *args)
317 		return EINVAL;
318 
319 	if (mp->mnt_flag & MNT_GETARGS) {
320 		ump = VFSTOUFS(mp);
321 		if (ump == NULL)
322 			return EIO;
323 		memset(args, 0, sizeof *args);
324 		args->fspec = NULL;
325 		*data_len = sizeof *args;
326 		return 0;
327 	}
328 
329 	update = mp->mnt_flag & MNT_UPDATE;
330 
331 	/* Check arguments */
332 	if (args->fspec != NULL) {
333 		/*
334 		 * Look up the name and verify that it's sane.
335 		 */
336 		error = namei_simple_user(args->fspec,
337 					NSM_FOLLOW_NOEMULROOT, &devvp);
338 		if (error != 0)
339 			return error;
340 
341 		if (!update) {
342 			/*
343 			 * Be sure this is a valid block device
344 			 */
345 			if (devvp->v_type != VBLK)
346 				error = ENOTBLK;
347 			else if (bdevsw_lookup(devvp->v_rdev) == NULL)
348 				error = ENXIO;
349 		} else {
350 		        /*
351 			 * Be sure we're still naming the same device
352 			 * used for our initial mount
353 			 */
354 			ump = VFSTOUFS(mp);
355 			if (devvp != ump->um_devvp) {
356 				if (devvp->v_rdev != ump->um_devvp->v_rdev)
357 					error = EINVAL;
358 				else {
359 					vrele(devvp);
360 					devvp = ump->um_devvp;
361 					vref(devvp);
362 				}
363 			}
364 		}
365 	} else {
366 		if (!update) {
367 			/* New mounts must have a filename for the device */
368 			return EINVAL;
369 		} else {
370 			ump = VFSTOUFS(mp);
371 			devvp = ump->um_devvp;
372 			vref(devvp);
373 		}
374 	}
375 
376 	/*
377 	 * If mount by non-root, then verify that user has necessary
378 	 * permissions on the device.
379 	 *
380 	 * Permission to update a mount is checked higher, so here we presume
381 	 * updating the mount is okay (for example, as far as securelevel goes)
382 	 * which leaves us with the normal check.
383 	 */
384 	if (error == 0) {
385 		accessmode = VREAD;
386 		if (update ?
387 		    (mp->mnt_iflag & IMNT_WANTRDWR) != 0 :
388 		    (mp->mnt_flag & MNT_RDONLY) == 0)
389 			accessmode |= VWRITE;
390 		vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
391 		error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT,
392 		    KAUTH_REQ_SYSTEM_MOUNT_DEVICE, mp, devvp,
393 		    KAUTH_ARG(accessmode));
394 		VOP_UNLOCK(devvp);
395 	}
396 
397 	if (error) {
398 		vrele(devvp);
399 		return error;
400 	}
401 
402 	if (!update) {
403 		int xflags;
404 
405 		if (mp->mnt_flag & MNT_RDONLY)
406 			xflags = FREAD;
407 		else
408 			xflags = FREAD|FWRITE;
409 		vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
410 		error = VOP_OPEN(devvp, xflags, FSCRED);
411 		VOP_UNLOCK(devvp);
412 		if (error)
413 			goto fail;
414 		error = ext2fs_mountfs(devvp, mp);
415 		if (error) {
416 			vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
417 			(void)VOP_CLOSE(devvp, xflags, NOCRED);
418 			VOP_UNLOCK(devvp);
419 			goto fail;
420 		}
421 
422 		ump = VFSTOUFS(mp);
423 		fs = ump->um_e2fs;
424 	} else {
425 		/*
426 		 * Update the mount.
427 		 */
428 
429 		/*
430 		 * The initial mount got a reference on this
431 		 * device, so drop the one obtained via
432 		 * namei(), above.
433 		 */
434 		vrele(devvp);
435 
436 		ump = VFSTOUFS(mp);
437 		fs = ump->um_e2fs;
438 		if (fs->e2fs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) {
439 			/*
440 			 * Changing from r/w to r/o
441 			 */
442 			flags = WRITECLOSE;
443 			if (mp->mnt_flag & MNT_FORCE)
444 				flags |= FORCECLOSE;
445 			error = ext2fs_flushfiles(mp, flags);
446 			if (error == 0 &&
447 			    ext2fs_cgupdate(ump, MNT_WAIT) == 0 &&
448 			    (fs->e2fs.e2fs_state & E2FS_ERRORS) == 0) {
449 				fs->e2fs.e2fs_state = E2FS_ISCLEAN;
450 				(void) ext2fs_sbupdate(ump, MNT_WAIT);
451 			}
452 			if (error)
453 				return error;
454 			fs->e2fs_ronly = 1;
455 		}
456 
457 		if (mp->mnt_flag & MNT_RELOAD) {
458 			error = ext2fs_reload(mp, l->l_cred, l);
459 			if (error)
460 				return error;
461 		}
462 
463 		if (fs->e2fs_ronly && (mp->mnt_iflag & IMNT_WANTRDWR)) {
464 			/*
465 			 * Changing from read-only to read/write
466 			 */
467 			fs->e2fs_ronly = 0;
468 			if (fs->e2fs.e2fs_state == E2FS_ISCLEAN)
469 				fs->e2fs.e2fs_state = 0;
470 			else
471 				fs->e2fs.e2fs_state = E2FS_ERRORS;
472 			fs->e2fs_fmod = 1;
473 		}
474 		if (args->fspec == NULL)
475 			return 0;
476 	}
477 
478 	error = set_statvfs_info(path, UIO_USERSPACE, args->fspec,
479 	    UIO_USERSPACE, mp->mnt_op->vfs_name, mp, l);
480 	if (error == 0)
481 		ext2fs_sb_setmountinfo(fs, mp);
482 
483 	if (fs->e2fs_fmod != 0) {	/* XXX */
484 		fs->e2fs_fmod = 0;
485 		if (fs->e2fs.e2fs_state == 0)
486 			fs->e2fs.e2fs_wtime = time_second;
487 		else
488 			printf("%s: file system not clean; please fsck(8)\n",
489 				mp->mnt_stat.f_mntfromname);
490 		(void) ext2fs_cgupdate(ump, MNT_WAIT);
491 	}
492 	return error;
493 
494 fail:
495 	vrele(devvp);
496 	return error;
497 }
498 
499 /*
500  * Sanity check the disk vnode content, and copy it over to inode structure.
501  */
502 static int
503 ext2fs_loadvnode_content(struct m_ext2fs *fs, ino_t ino, struct buf *bp, struct inode *ip)
504 {
505 	struct ext2fs_dinode *din;
506 	int error = 0;
507 
508 	din = (struct ext2fs_dinode *)((char *)bp->b_data + (ino_to_fsbo(fs, ino) * EXT2_DINODE_SIZE(fs)));
509 
510 	/* sanity checks - inode data NOT byteswapped at this point */
511 	if (EXT2_DINODE_FITS(din, e2di_extra_isize, EXT2_DINODE_SIZE(fs))
512 	    && (EXT2_DINODE_SIZE(fs) - EXT2_REV0_DINODE_SIZE) < fs2h16(din->e2di_extra_isize))
513 	{
514 		printf("ext2fs: inode %"PRIu64" bad extra_isize %u",
515 			ino, din->e2di_extra_isize);
516 		error = EINVAL;
517 		goto bad;
518 	}
519 
520 	/* everything allright, proceed with copy */
521 	if (ip->i_din.e2fs_din == NULL)
522 		ip->i_din.e2fs_din = kmem_alloc(EXT2_DINODE_SIZE(fs), KM_SLEEP);
523 
524 	e2fs_iload(din, ip->i_din.e2fs_din, EXT2_DINODE_SIZE(fs));
525 
526 	ext2fs_set_inode_guid(ip);
527 
528     bad:
529 	return error;
530 }
531 
532 /*
533  * Reload all incore data for a filesystem (used after running fsck on
534  * the root filesystem and finding things to fix). The filesystem must
535  * be mounted read-only.
536  *
537  * Things to do to update the mount:
538  *	1) invalidate all cached meta-data.
539  *	2) re-read superblock from disk.
540  *	3) re-read summary information from disk.
541  *	4) invalidate all inactive vnodes.
542  *	5) invalidate all cached file data.
543  *	6) re-read inode data for all active vnodes.
544  */
545 int
546 ext2fs_reload(struct mount *mp, kauth_cred_t cred, struct lwp *l)
547 {
548 	struct vnode *vp, *devvp;
549 	struct inode *ip;
550 	struct buf *bp;
551 	struct m_ext2fs *fs;
552 	struct ext2fs *newfs;
553 	int i, error;
554 	struct ufsmount *ump;
555 	struct vnode_iterator *marker;
556 
557 	if ((mp->mnt_flag & MNT_RDONLY) == 0)
558 		return EINVAL;
559 
560 	ump = VFSTOUFS(mp);
561 	/*
562 	 * Step 1: invalidate all cached meta-data.
563 	 */
564 	devvp = ump->um_devvp;
565 	vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
566 	error = vinvalbuf(devvp, 0, cred, l, 0, 0);
567 	VOP_UNLOCK(devvp);
568 	if (error)
569 		panic("ext2fs_reload: dirty1");
570 
571 	fs = ump->um_e2fs;
572 	/*
573 	 * Step 2: re-read superblock from disk. Copy in new superblock, and compute
574 	 * in-memory values.
575 	 */
576 	error = bread(devvp, SBLOCK, SBSIZE, 0, &bp);
577 	if (error)
578 		return error;
579 	newfs = (struct ext2fs *)bp->b_data;
580 	e2fs_sbload(newfs, &fs->e2fs);
581 
582 	brelse(bp, 0);
583 
584 	error = ext2fs_sbfill(fs, (mp->mnt_flag & MNT_RDONLY) != 0);
585 	if (error)
586 		return error;
587 
588 	/*
589 	 * Step 3: re-read summary information from disk.
590 	 */
591 	for (i = 0; i < fs->e2fs_ngdb; i++) {
592 		error = bread(devvp ,
593 		    EXT2_FSBTODB(fs, fs->e2fs.e2fs_first_dblock +
594 		    1 /* superblock */ + i),
595 		    fs->e2fs_bsize, 0, &bp);
596 		if (error) {
597 			return error;
598 		}
599 		e2fs_cgload((struct ext2_gd *)bp->b_data,
600 		    &fs->e2fs_gd[i * fs->e2fs_bsize / sizeof(struct ext2_gd)],
601 		    fs->e2fs_bsize);
602 		brelse(bp, 0);
603 	}
604 
605 	vfs_vnode_iterator_init(mp, &marker);
606 	while ((vp = vfs_vnode_iterator_next(marker, NULL, NULL))) {
607 		/*
608 		 * Step 4: invalidate all inactive vnodes.
609 		 */
610 		if (vrecycle(vp))
611 			continue;
612 		/*
613 		 * Step 5: invalidate all cached file data.
614 		 */
615 		if (vn_lock(vp, LK_EXCLUSIVE)) {
616 			vrele(vp);
617 			continue;
618 		}
619 		if (vinvalbuf(vp, 0, cred, l, 0, 0))
620 			panic("ext2fs_reload: dirty2");
621 		/*
622 		 * Step 6: re-read inode data for all active vnodes.
623 		 */
624 		ip = VTOI(vp);
625 		error = bread(devvp, EXT2_FSBTODB(fs, ino_to_fsba(fs, ip->i_number)),
626 		    (int)fs->e2fs_bsize, 0, &bp);
627 		if (error) {
628 			vput(vp);
629 			break;
630 		}
631 		error = ext2fs_loadvnode_content(fs, ip->i_number, bp, ip);
632 		brelse(bp, 0);
633 		if (error) {
634 			vput(vp);
635 			break;
636 		}
637 
638 		vput(vp);
639 	}
640 	vfs_vnode_iterator_destroy(marker);
641 	return error;
642 }
643 
644 /*
645  * Common code for mount and mountroot
646  */
647 int
648 ext2fs_mountfs(struct vnode *devvp, struct mount *mp)
649 {
650 	struct lwp *l = curlwp;
651 	struct ufsmount *ump;
652 	struct buf *bp;
653 	struct ext2fs *fs;
654 	struct m_ext2fs *m_fs;
655 	dev_t dev;
656 	int error, i, ronly;
657 	kauth_cred_t cred;
658 
659 	dev = devvp->v_rdev;
660 	cred = l->l_cred;
661 
662 	/* Flush out any old buffers remaining from a previous use. */
663 	vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
664 	error = vinvalbuf(devvp, V_SAVE, cred, l, 0, 0);
665 	VOP_UNLOCK(devvp);
666 	if (error)
667 		return error;
668 
669 	ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
670 
671 	bp = NULL;
672 	ump = NULL;
673 
674 	/* Read the superblock from disk, and swap it directly. */
675 	error = bread(devvp, SBLOCK, SBSIZE, 0, &bp);
676 	if (error)
677 		goto out;
678 	fs = (struct ext2fs *)bp->b_data;
679 	m_fs = kmem_zalloc(sizeof(*m_fs), KM_SLEEP);
680 	e2fs_sbload(fs, &m_fs->e2fs);
681 
682 	brelse(bp, 0);
683 	bp = NULL;
684 
685 	/* Once swapped, validate and fill in the superblock. */
686 	error = ext2fs_sbfill(m_fs, ronly);
687 	if (error) {
688 		kmem_free(m_fs, sizeof(*m_fs));
689 		goto out;
690 	}
691 	m_fs->e2fs_ronly = ronly;
692 
693 	ump = kmem_zalloc(sizeof(*ump), KM_SLEEP);
694 	ump->um_fstype = UFS1;
695 	ump->um_ops = &ext2fs_ufsops;
696 	ump->um_e2fs = m_fs;
697 
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 	/* XXX: should be added in ext2fs_sbfill()? */
707 	m_fs->e2fs_gd = kmem_alloc(m_fs->e2fs_ngdb * m_fs->e2fs_bsize, KM_SLEEP);
708 	for (i = 0; i < m_fs->e2fs_ngdb; i++) {
709 		error = bread(devvp,
710 		    EXT2_FSBTODB(m_fs, m_fs->e2fs.e2fs_first_dblock +
711 		    1 /* superblock */ + i),
712 		    m_fs->e2fs_bsize, 0, &bp);
713 		if (error) {
714 			kmem_free(m_fs->e2fs_gd,
715 			    m_fs->e2fs_ngdb * m_fs->e2fs_bsize);
716 			goto out;
717 		}
718 		e2fs_cgload((struct ext2_gd *)bp->b_data,
719 		    &m_fs->e2fs_gd[
720 			i * m_fs->e2fs_bsize / sizeof(struct ext2_gd)],
721 		    m_fs->e2fs_bsize);
722 		brelse(bp, 0);
723 		bp = NULL;
724 	}
725 
726 	error = ext2fs_cg_verify_and_initialize(devvp, m_fs, ronly);
727 	if (error) {
728 		kmem_free(m_fs->e2fs_gd, m_fs->e2fs_ngdb * m_fs->e2fs_bsize);
729 		goto out;
730 	}
731 
732 	mp->mnt_data = ump;
733 	mp->mnt_stat.f_fsidx.__fsid_val[0] = (long)dev;
734 	mp->mnt_stat.f_fsidx.__fsid_val[1] = makefstype(MOUNT_EXT2FS);
735 	mp->mnt_stat.f_fsid = mp->mnt_stat.f_fsidx.__fsid_val[0];
736 	mp->mnt_stat.f_namemax = EXT2FS_MAXNAMLEN;
737 	mp->mnt_flag |= MNT_LOCAL;
738 	mp->mnt_dev_bshift = DEV_BSHIFT;	/* XXX */
739 	mp->mnt_fs_bshift = m_fs->e2fs_bshift;
740 	mp->mnt_iflag |= IMNT_DTYPE;
741 	ump->um_flags = 0;
742 	ump->um_mountp = mp;
743 	ump->um_dev = dev;
744 	ump->um_devvp = devvp;
745 	ump->um_nindir = EXT2_NINDIR(m_fs);
746 	ump->um_lognindir = ffs(EXT2_NINDIR(m_fs)) - 1;
747 	ump->um_bptrtodb = m_fs->e2fs_fsbtodb;
748 	ump->um_seqinc = 1; /* no frags */
749 	ump->um_maxsymlinklen = EXT2_MAXSYMLINKLEN;
750 	ump->um_dirblksiz = m_fs->e2fs_bsize;
751 	ump->um_maxfilesize = ((uint64_t)0x80000000 * m_fs->e2fs_bsize - 1);
752 	spec_node_setmountedfs(devvp, mp);
753 	return 0;
754 
755 out:
756 	if (bp != NULL)
757 		brelse(bp, 0);
758 	if (ump) {
759 		kmem_free(ump->um_e2fs, sizeof(*m_fs));
760 		kmem_free(ump, sizeof(*ump));
761 		mp->mnt_data = NULL;
762 	}
763 	return error;
764 }
765 
766 /*
767  * unmount system call
768  */
769 int
770 ext2fs_unmount(struct mount *mp, int mntflags)
771 {
772 	struct ufsmount *ump;
773 	struct m_ext2fs *fs;
774 	int error, flags;
775 
776 	flags = 0;
777 	if (mntflags & MNT_FORCE)
778 		flags |= FORCECLOSE;
779 	if ((error = ext2fs_flushfiles(mp, flags)) != 0)
780 		return error;
781 	ump = VFSTOUFS(mp);
782 	fs = ump->um_e2fs;
783 	if (fs->e2fs_ronly == 0 &&
784 		ext2fs_cgupdate(ump, MNT_WAIT) == 0 &&
785 		(fs->e2fs.e2fs_state & E2FS_ERRORS) == 0) {
786 		fs->e2fs.e2fs_state = E2FS_ISCLEAN;
787 		(void) ext2fs_sbupdate(ump, MNT_WAIT);
788 	}
789 	if (ump->um_devvp->v_type != VBAD)
790 		spec_node_setmountedfs(ump->um_devvp, NULL);
791 	vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
792 	error = VOP_CLOSE(ump->um_devvp, fs->e2fs_ronly ? FREAD : FREAD|FWRITE,
793 	    NOCRED);
794 	vput(ump->um_devvp);
795 	kmem_free(fs->e2fs_gd, fs->e2fs_ngdb * fs->e2fs_bsize);
796 	kmem_free(fs, sizeof(*fs));
797 	kmem_free(ump, sizeof(*ump));
798 	mp->mnt_data = NULL;
799 	mp->mnt_flag &= ~MNT_LOCAL;
800 	return error;
801 }
802 
803 /*
804  * Flush out all the files in a filesystem.
805  */
806 int
807 ext2fs_flushfiles(struct mount *mp, int flags)
808 {
809 	extern int doforce;
810 	int error;
811 
812 	if (!doforce)
813 		flags &= ~FORCECLOSE;
814 	error = vflush(mp, NULLVP, flags);
815 	return error;
816 }
817 
818 /*
819  * Get file system statistics.
820  */
821 int
822 ext2fs_statvfs(struct mount *mp, struct statvfs *sbp)
823 {
824 	struct ufsmount *ump;
825 	struct m_ext2fs *fs;
826 	uint32_t overhead, overhead_per_group, ngdb;
827 	int i, ngroups;
828 
829 	ump = VFSTOUFS(mp);
830 	fs = ump->um_e2fs;
831 	if (fs->e2fs.e2fs_magic != E2FS_MAGIC)
832 		panic("ext2fs_statvfs");
833 
834 	/*
835 	 * Compute the overhead (FS structures)
836 	 */
837 	overhead_per_group =
838 	    1 /* block bitmap */ +
839 	    1 /* inode bitmap */ +
840 	    fs->e2fs_itpg;
841 	overhead = fs->e2fs.e2fs_first_dblock +
842 	    fs->e2fs_ncg * overhead_per_group;
843 	if (EXT2F_HAS_COMPAT_FEATURE(fs, EXT2F_COMPAT_SPARSESUPER2)) {
844 		/*
845 		 * Superblock and group descriptions is in group zero,
846 		 * then optionally 0, 1 or 2 extra copies.
847 		 */
848 		ngroups = 1
849 			+ (fs->e2fs.e4fs_backup_bgs[0] ? 1 : 0)
850 			+ (fs->e2fs.e4fs_backup_bgs[1] ? 1 : 0);
851 	} else if (EXT2F_HAS_ROCOMPAT_FEATURE(fs, EXT2F_ROCOMPAT_SPARSESUPER)) {
852 		for (i = 0, ngroups = 0; i < fs->e2fs_ncg; i++) {
853 			if (cg_has_sb(i))
854 				ngroups++;
855 		}
856 	} else {
857 		ngroups = fs->e2fs_ncg;
858 	}
859 	ngdb = fs->e2fs_ngdb;
860 	if (EXT2F_HAS_COMPAT_FEATURE(fs, EXT2F_COMPAT_RESIZE))
861 		ngdb += fs->e2fs.e2fs_reserved_ngdb;
862 	overhead += ngroups * (1 /* superblock */ + ngdb);
863 
864 	sbp->f_bsize = fs->e2fs_bsize;
865 	sbp->f_frsize = MINBSIZE << fs->e2fs.e2fs_fsize;
866 	sbp->f_iosize = fs->e2fs_bsize;
867 	sbp->f_blocks = fs->e2fs.e2fs_bcount - overhead;
868 	sbp->f_bfree = fs->e2fs.e2fs_fbcount;
869 	sbp->f_bresvd = fs->e2fs.e2fs_rbcount;
870 	if (sbp->f_bfree > sbp->f_bresvd)
871 		sbp->f_bavail = sbp->f_bfree - sbp->f_bresvd;
872 	else
873 		sbp->f_bavail = 0;
874 	sbp->f_files =  fs->e2fs.e2fs_icount;
875 	sbp->f_ffree = fs->e2fs.e2fs_ficount;
876 	sbp->f_favail = fs->e2fs.e2fs_ficount;
877 	sbp->f_fresvd = 0;
878 	copy_statvfs_info(sbp, mp);
879 	return 0;
880 }
881 
882 static bool
883 ext2fs_sync_selector(void *cl, struct vnode *vp)
884 {
885 	struct inode *ip;
886 
887 	KASSERT(mutex_owned(vp->v_interlock));
888 
889 	ip = VTOI(vp);
890 	/*
891 	 * Skip the vnode/inode if inaccessible.
892 	 */
893 	if (ip == NULL || vp->v_type == VNON)
894 		return false;
895 
896 	if (((ip->i_flag &
897 	      (IN_CHANGE | IN_UPDATE | IN_MODIFIED)) == 0 &&
898 	     LIST_EMPTY(&vp->v_dirtyblkhd) &&
899 	     UVM_OBJ_IS_CLEAN(&vp->v_uobj)))
900 		return false;
901 	return true;
902 }
903 
904 /*
905  * Go through the disk queues to initiate sandbagged IO;
906  * go through the inodes to write those that have been modified;
907  * initiate the writing of the super block if it has been modified.
908  *
909  * Note: we are always called with the filesystem marked `MPBUSY'.
910  */
911 int
912 ext2fs_sync(struct mount *mp, int waitfor, kauth_cred_t cred)
913 {
914 	struct vnode *vp;
915 	struct ufsmount *ump = VFSTOUFS(mp);
916 	struct m_ext2fs *fs;
917 	struct vnode_iterator *marker;
918 	int error, allerror = 0;
919 
920 	fs = ump->um_e2fs;
921 	if (fs->e2fs_fmod != 0 && fs->e2fs_ronly != 0) {	/* XXX */
922 		printf("fs = %s\n", fs->e2fs_fsmnt);
923 		panic("update: rofs mod");
924 	}
925 
926 	/*
927 	 * Write back each (modified) inode.
928 	 */
929 	vfs_vnode_iterator_init(mp, &marker);
930 	while ((vp = vfs_vnode_iterator_next(marker, ext2fs_sync_selector,
931 	    NULL)))
932 	{
933 		error = vn_lock(vp, LK_EXCLUSIVE);
934 		if (error) {
935 			vrele(vp);
936 			continue;
937 		}
938 		if (vp->v_type == VREG && waitfor == MNT_LAZY)
939 			error = ext2fs_update(vp, NULL, NULL, 0);
940 		else
941 			error = VOP_FSYNC(vp, cred,
942 			    waitfor == MNT_WAIT ? FSYNC_WAIT : 0, 0, 0);
943 		if (error)
944 			allerror = error;
945 		vput(vp);
946 	}
947 	vfs_vnode_iterator_destroy(marker);
948 	/*
949 	 * Force stale file system control information to be flushed.
950 	 */
951 	if (waitfor != MNT_LAZY) {
952 		vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
953 		if ((error = VOP_FSYNC(ump->um_devvp, cred,
954 		    waitfor == MNT_WAIT ? FSYNC_WAIT : 0, 0, 0)) != 0)
955 			allerror = error;
956 		VOP_UNLOCK(ump->um_devvp);
957 	}
958 	/*
959 	 * Write back modified superblock.
960 	 */
961 	if (fs->e2fs_fmod != 0) {
962 		fs->e2fs_fmod = 0;
963 		fs->e2fs.e2fs_wtime = time_second;
964 		if ((error = ext2fs_cgupdate(ump, waitfor)))
965 			allerror = error;
966 	}
967 	return allerror;
968 }
969 
970 /*
971  * Load inode from disk and initialize vnode.
972  */
973 static int
974 ext2fs_init_vnode(struct ufsmount *ump, struct vnode *vp, ino_t ino)
975 {
976 	struct m_ext2fs *fs;
977 	struct inode *ip;
978 	struct buf *bp;
979 	int error;
980 
981 	fs = ump->um_e2fs;
982 
983 	/* Read in the disk contents for the inode, copy into the inode. */
984 	error = bread(ump->um_devvp, EXT2_FSBTODB(fs, ino_to_fsba(fs, ino)),
985 	    (int)fs->e2fs_bsize, 0, &bp);
986 	if (error)
987 		return error;
988 
989 	/* Allocate and initialize inode. */
990 	ip = pool_get(&ext2fs_inode_pool, PR_WAITOK);
991 	memset(ip, 0, sizeof(struct inode));
992 	ip->i_vnode = vp;
993 	ip->i_ump = ump;
994 	ip->i_e2fs = fs;
995 	ip->i_dev = ump->um_dev;
996 	ip->i_number = ino;
997 	ip->i_e2fs_last_lblk = 0;
998 	ip->i_e2fs_last_blk = 0;
999 
1000 	error = ext2fs_loadvnode_content(fs, ino, bp, ip);
1001 	brelse(bp, 0);
1002 	if (error) {
1003 		pool_put(&ext2fs_inode_pool, ip);
1004 		return error;
1005 	}
1006 
1007 	/* If the inode was deleted, reset all fields */
1008 	if (ip->i_e2fs_dtime != 0) {
1009 		ip->i_e2fs_mode = 0;
1010 		(void)ext2fs_setsize(ip, 0);
1011 		(void)ext2fs_setnblock(ip, 0);
1012 		memset(ip->i_e2fs_blocks, 0, sizeof(ip->i_e2fs_blocks));
1013 	}
1014 
1015 	/* Initialise vnode with this inode. */
1016 	vp->v_tag = VT_EXT2FS;
1017 	vp->v_op = ext2fs_vnodeop_p;
1018 	vp->v_vflag |= VV_LOCKSWORK;
1019 	vp->v_data = ip;
1020 
1021 	/* Initialize genfs node. */
1022 	genfs_node_init(vp, &ext2fs_genfsops);
1023 
1024 	return 0;
1025 }
1026 
1027 /*
1028  * Read an inode from disk and initialize this vnode / inode pair.
1029  * Caller assures no other thread will try to load this inode.
1030  */
1031 int
1032 ext2fs_loadvnode(struct mount *mp, struct vnode *vp,
1033     const void *key, size_t key_len, const void **new_key)
1034 {
1035 	ino_t ino;
1036 	struct inode *ip;
1037 	struct ufsmount *ump;
1038 	int error;
1039 
1040 	KASSERT(key_len == sizeof(ino));
1041 	memcpy(&ino, key, key_len);
1042 	ump = VFSTOUFS(mp);
1043 
1044 	error = ext2fs_init_vnode(ump, vp, ino);
1045 	if (error)
1046 		return error;
1047 
1048 	ip = VTOI(vp);
1049 
1050 	/* Initialize the vnode from the inode. */
1051 	ext2fs_vinit(mp, ext2fs_specop_p, ext2fs_fifoop_p, &vp);
1052 
1053 	/* Finish inode initialization. */
1054 	ip->i_devvp = ump->um_devvp;
1055 	vref(ip->i_devvp);
1056 
1057 	/*
1058 	 * Set up a generation number for this inode if it does not
1059 	 * already have one. This should only happen on old filesystems.
1060 	 */
1061 
1062 	if (ip->i_e2fs_gen == 0) {
1063 		if (++ext2gennumber < (u_long)time_second)
1064 			ext2gennumber = time_second;
1065 		ip->i_e2fs_gen = ext2gennumber;
1066 		if ((mp->mnt_flag & MNT_RDONLY) == 0)
1067 			ip->i_flag |= IN_MODIFIED;
1068 	}
1069 	uvm_vnp_setsize(vp, ext2fs_size(ip));
1070 	*new_key = &ip->i_number;
1071 	return 0;
1072 }
1073 
1074 /*
1075  * Create a new inode on disk and initialize this vnode / inode pair.
1076  */
1077 int
1078 ext2fs_newvnode(struct mount *mp, struct vnode *dvp, struct vnode *vp,
1079     struct vattr *vap, kauth_cred_t cred,
1080     size_t *key_len, const void **new_key)
1081 {
1082 	ino_t ino;
1083 	struct inode *ip, *pdir;
1084 	struct m_ext2fs *fs;
1085 	struct ufsmount *ump;
1086 	int error, mode;
1087 
1088 	KASSERT(dvp->v_mount == mp);
1089 	KASSERT(vap->va_type != VNON);
1090 
1091 	*key_len = sizeof(ino);
1092 
1093 	pdir = VTOI(dvp);
1094 	fs = pdir->i_e2fs;
1095 	ump = VFSTOUFS(mp);
1096 	mode = MAKEIMODE(vap->va_type, vap->va_mode);
1097 
1098 	/* Allocate fresh inode. */
1099 	error = ext2fs_valloc(dvp, mode, cred, &ino);
1100 	if (error)
1101 		return error;
1102 
1103 	/* Attach inode to vnode. */
1104 	error = ext2fs_init_vnode(ump, vp, ino);
1105 	if (error) {
1106 		ext2fs_vfree(dvp, ino, mode);
1107 		return error;
1108 	}
1109 
1110 	ip = VTOI(vp);
1111 
1112 	KASSERT(!E2FS_HAS_GD_CSUM(fs) || (fs->e2fs_gd[ino_to_cg(fs, ino)].ext2bgd_flags & h2fs16(E2FS_BG_INODE_ZEROED)) != 0);
1113 
1114 	/* check for already used inode; makes sense only for ZEROED itable */
1115 	if (__predict_false(ip->i_e2fs_mode && ip->i_e2fs_nlink != 0)) {
1116 		printf("mode = 0%o, nlinks %d, inum = %llu, fs = %s\n",
1117 		    ip->i_e2fs_mode, ip->i_e2fs_nlink,
1118 		    (unsigned long long)ip->i_number, fs->e2fs_fsmnt);
1119 		panic("ext2fs_valloc: dup alloc");
1120 	}
1121 
1122 	memset(ip->i_din.e2fs_din, 0, EXT2_DINODE_SIZE(fs));
1123 
1124 	/*
1125 	 * Set up a new generation number for this inode.
1126 	 */
1127 	if (++ext2gennumber < time_second)
1128 		ext2gennumber = time_second;
1129 	ip->i_e2fs_gen = ext2gennumber;
1130 
1131 	ip->i_uid = kauth_cred_geteuid(cred);
1132 	ip->i_e2fs_uid = ip->i_uid & 0xffff;
1133 	ip->i_e2fs_gid = pdir->i_e2fs_gid;
1134 	if (ip->i_e2fs->e2fs.e2fs_rev > E2FS_REV0) {
1135 		ip->i_e2fs_uid_high = (ip->i_uid >> 16) & 0xffff;
1136 		ip->i_e2fs_gid_high = pdir->i_e2fs_gid_high;
1137 	} else {
1138 		ip->i_e2fs_uid_high = 0;
1139 		ip->i_e2fs_gid_high = 0;
1140 	}
1141 	ip->i_gid = ip->i_e2fs_gid | (ip->i_e2fs_gid_high << 16);
1142 	ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
1143 	ip->i_e2fs_mode = mode;
1144 	vp->v_type = IFTOVT(mode);
1145 	ip->i_e2fs_nlink = 1;
1146 
1147 	/* Authorize setting SGID if needed. */
1148 	if (ip->i_e2fs_mode & ISGID) {
1149 		error = kauth_authorize_vnode(cred, KAUTH_VNODE_WRITE_SECURITY,
1150 		    vp, NULL, genfs_can_chmod(vp->v_type, cred, ip->i_uid,
1151 		    ip->i_gid, mode));
1152 		if (error)
1153 			ip->i_e2fs_mode &= ~ISGID;
1154 	}
1155 
1156 	/* Initialize extra_isize according to what is set in superblock */
1157 	if (EXT2F_HAS_ROCOMPAT_FEATURE(ip->i_e2fs, EXT2F_ROCOMPAT_EXTRA_ISIZE)
1158 	    && EXT2_DINODE_SIZE(ip->i_e2fs) > EXT2_REV0_DINODE_SIZE) {
1159 		ip->i_din.e2fs_din->e2di_extra_isize = ip->i_e2fs->e2fs.e4fs_want_extra_isize;
1160 	}
1161 
1162 	/* Set create time if possible */
1163 	if (EXT2_DINODE_FITS(ip->i_din.e2fs_din, e2di_crtime, EXT2_DINODE_SIZE(ip->i_e2fs))) {
1164 		struct timespec now;
1165 		vfs_timestamp(&now);
1166 		EXT2_DINODE_TIME_SET(&now, ip->i_din.e2fs_din, e2di_crtime, EXT2_DINODE_SIZE(ip->i_e2fs));
1167 	}
1168 
1169 	/* Initialize the vnode from the inode. */
1170 	ext2fs_vinit(mp, ext2fs_specop_p, ext2fs_fifoop_p, &vp);
1171 
1172 	/* Finish inode initialization. */
1173 	ip->i_devvp = ump->um_devvp;
1174 	vref(ip->i_devvp);
1175 
1176 	uvm_vnp_setsize(vp, ext2fs_size(ip));
1177 	*new_key = &ip->i_number;
1178 	return 0;
1179 }
1180 
1181 /*
1182  * File handle to vnode
1183  *
1184  * Have to be really careful about stale file handles:
1185  * - check that the inode number is valid
1186  * - call ext2fs_vget() to get the locked inode
1187  * - check for an unallocated inode (i_mode == 0)
1188  */
1189 int
1190 ext2fs_fhtovp(struct mount *mp, struct fid *fhp, struct vnode **vpp)
1191 {
1192 	struct inode *ip;
1193 	struct vnode *nvp;
1194 	int error;
1195 	struct ufid ufh;
1196 	struct m_ext2fs *fs;
1197 
1198 	if (fhp->fid_len != sizeof(struct ufid))
1199 		return EINVAL;
1200 
1201 	memcpy(&ufh, fhp, sizeof(struct ufid));
1202 	fs = VFSTOUFS(mp)->um_e2fs;
1203 	if ((ufh.ufid_ino < EXT2_FIRSTINO && ufh.ufid_ino != EXT2_ROOTINO) ||
1204 		ufh.ufid_ino >= fs->e2fs_ncg * fs->e2fs.e2fs_ipg)
1205 		return ESTALE;
1206 
1207 	if ((error = VFS_VGET(mp, ufh.ufid_ino, &nvp)) != 0) {
1208 		*vpp = NULLVP;
1209 		return error;
1210 	}
1211 	ip = VTOI(nvp);
1212 	if (ip->i_e2fs_mode == 0 || ip->i_e2fs_dtime != 0 ||
1213 		ip->i_e2fs_gen != ufh.ufid_gen) {
1214 		vput(nvp);
1215 		*vpp = NULLVP;
1216 		return ESTALE;
1217 	}
1218 	*vpp = nvp;
1219 	return 0;
1220 }
1221 
1222 /*
1223  * Vnode pointer to File handle
1224  */
1225 /* ARGSUSED */
1226 int
1227 ext2fs_vptofh(struct vnode *vp, struct fid *fhp, size_t *fh_size)
1228 {
1229 	struct inode *ip;
1230 	struct ufid ufh;
1231 
1232 	if (*fh_size < sizeof(struct ufid)) {
1233 		*fh_size = sizeof(struct ufid);
1234 		return E2BIG;
1235 	}
1236 	*fh_size = sizeof(struct ufid);
1237 
1238 	ip = VTOI(vp);
1239 	memset(&ufh, 0, sizeof(ufh));
1240 	ufh.ufid_len = sizeof(struct ufid);
1241 	ufh.ufid_ino = ip->i_number;
1242 	ufh.ufid_gen = ip->i_e2fs_gen;
1243 	memcpy(fhp, &ufh, sizeof(ufh));
1244 	return 0;
1245 }
1246 
1247 /*
1248  * Write a superblock and associated information back to disk.
1249  */
1250 int
1251 ext2fs_sbupdate(struct ufsmount *mp, int waitfor)
1252 {
1253 	struct m_ext2fs *fs = mp->um_e2fs;
1254 	struct buf *bp;
1255 	int error = 0;
1256 
1257 	bp = getblk(mp->um_devvp, SBLOCK, SBSIZE, 0, 0);
1258 	e2fs_sbsave(&fs->e2fs, (struct ext2fs*)bp->b_data);
1259 	if (waitfor == MNT_WAIT)
1260 		error = bwrite(bp);
1261 	else
1262 		bawrite(bp);
1263 	return error;
1264 }
1265 
1266 int
1267 ext2fs_cgupdate(struct ufsmount *mp, int waitfor)
1268 {
1269 	struct m_ext2fs *fs = mp->um_e2fs;
1270 	struct buf *bp;
1271 	int i, error = 0, allerror = 0;
1272 
1273 	allerror = ext2fs_sbupdate(mp, waitfor);
1274 	for (i = 0; i < fs->e2fs_ngdb; i++) {
1275 		bp = getblk(mp->um_devvp, EXT2_FSBTODB(fs,
1276 		    fs->e2fs.e2fs_first_dblock +
1277 		    1 /* superblock */ + i), fs->e2fs_bsize, 0, 0);
1278 		e2fs_cgsave(&fs->e2fs_gd[
1279 		    i * fs->e2fs_bsize / sizeof(struct ext2_gd)],
1280 		    (struct ext2_gd *)bp->b_data, fs->e2fs_bsize);
1281 		if (waitfor == MNT_WAIT)
1282 			error = bwrite(bp);
1283 		else
1284 			bawrite(bp);
1285 	}
1286 
1287 	if (!allerror && error)
1288 		allerror = error;
1289 	return allerror;
1290 }
1291 
1292 /*
1293  * Fill in the m_fs structure, and validate the fields of the superblock.
1294  * NOTE: here, the superblock is already swapped.
1295  */
1296 static int
1297 ext2fs_sbfill(struct m_ext2fs *m_fs, int ronly)
1298 {
1299 	uint32_t u32;
1300 	struct ext2fs *fs = &m_fs->e2fs;
1301 
1302 	/*
1303 	 * General sanity checks
1304 	 */
1305 	if (fs->e2fs_magic != E2FS_MAGIC)
1306 		return EINVAL;
1307 	if (fs->e2fs_rev > E2FS_REV1) {
1308 		printf("ext2fs: unsupported revision number: %x\n", fs->e2fs_rev);
1309 		return EINVAL;
1310 	}
1311 	if (fs->e2fs_log_bsize > 2) {
1312 		/* block size = 1024|2048|4096 */
1313 		printf("ext2fs: bad block size: %d\n", fs->e2fs_log_bsize);
1314 		return EINVAL;
1315 	}
1316 	if (fs->e2fs_bpg == 0) {
1317 		printf("ext2fs: zero blocks per group\n");
1318 		return EINVAL;
1319 	}
1320 	if (fs->e2fs_ipg == 0) {
1321 		printf("ext2fs: zero inodes per group\n");
1322 		return EINVAL;
1323 	}
1324 
1325 	if (fs->e2fs_first_dblock >= fs->e2fs_bcount) {
1326 		printf("ext2fs: invalid first data block\n");
1327 		return EINVAL;
1328 	}
1329 	if (fs->e2fs_rbcount > fs->e2fs_bcount ||
1330 	    fs->e2fs_fbcount > fs->e2fs_bcount) {
1331 		printf("ext2fs: invalid block count\n");
1332 		return EINVAL;
1333 	}
1334 
1335 	/*
1336 	 * Compute the fields of the superblock
1337 	 */
1338 	u32 = fs->e2fs_bcount - fs->e2fs_first_dblock; /* > 0 */
1339 	m_fs->e2fs_ncg = howmany(u32, fs->e2fs_bpg);
1340 	if (m_fs->e2fs_ncg == 0) {
1341 		printf("ext2fs: invalid number of cylinder groups\n");
1342 		return EINVAL;
1343 	}
1344 
1345 	m_fs->e2fs_fsbtodb = fs->e2fs_log_bsize + LOG_MINBSIZE - DEV_BSHIFT;
1346 	m_fs->e2fs_bsize = MINBSIZE << fs->e2fs_log_bsize;
1347 	m_fs->e2fs_bshift = LOG_MINBSIZE + fs->e2fs_log_bsize;
1348 	m_fs->e2fs_qbmask = m_fs->e2fs_bsize - 1;
1349 	m_fs->e2fs_bmask = ~m_fs->e2fs_qbmask;
1350 
1351 	if ((u32 = m_fs->e2fs_bsize / sizeof(struct ext2_gd)) == 0) {
1352 		/* Unlikely to happen */
1353 		printf("ext2fs: invalid block size\n");
1354 		return EINVAL;
1355 	}
1356 	m_fs->e2fs_ngdb = howmany(m_fs->e2fs_ncg, u32);
1357 	if (m_fs->e2fs_ngdb == 0) {
1358 		printf("ext2fs: invalid number of group descriptor blocks\n");
1359 		return EINVAL;
1360 	}
1361 
1362 	if (m_fs->e2fs_bsize < EXT2_DINODE_SIZE(m_fs)) {
1363 		printf("ext2fs: invalid inode size\n");
1364 		return EINVAL;
1365 	}
1366 	m_fs->e2fs_ipb = m_fs->e2fs_bsize / EXT2_DINODE_SIZE(m_fs);
1367 
1368 	m_fs->e2fs_itpg = fs->e2fs_ipg / m_fs->e2fs_ipb;
1369 
1370 	/*
1371 	 * Revision-specific checks
1372 	 */
1373 	if (fs->e2fs_rev > E2FS_REV0) {
1374 		char buf[256];
1375 		if (fs->e2fs_first_ino != EXT2_FIRSTINO) {
1376 			printf("ext2fs: unsupported first inode position\n");
1377 			return EINVAL;
1378 		}
1379 		u32 = fs->e2fs_features_incompat & ~EXT2F_INCOMPAT_SUPP;
1380 		if (u32) {
1381 			snprintb(buf, sizeof(buf), EXT2F_INCOMPAT_BITS, u32);
1382 			printf("ext2fs: unsupported incompat features: %s\n", buf);
1383 #ifndef EXT2_IGNORE_INCOMPAT_FEATURES
1384 			return EINVAL;
1385 #endif
1386 		}
1387 		u32 = fs->e2fs_features_rocompat & ~EXT2F_ROCOMPAT_SUPP;
1388 		if (!ronly && u32) {
1389 			snprintb(buf, sizeof(buf), EXT2F_ROCOMPAT_BITS, u32);
1390 			printf("ext2fs: unsupported ro-incompat features: %s\n",
1391 			    buf);
1392 #ifndef EXT2_IGNORE_ROCOMPAT_FEATURES
1393 			return EROFS;
1394 #endif
1395 		}
1396 		if (fs->e2fs_inode_size == 0 || !powerof2(fs->e2fs_inode_size) || fs->e2fs_inode_size > m_fs->e2fs_bsize) {
1397 			printf("ext2fs: bad inode size\n");
1398 			return EINVAL;
1399 		}
1400 	}
1401 
1402 	return 0;
1403 }
1404