xref: /netbsd-src/sys/ufs/ext2fs/ext2fs_vfsops.c (revision 82d56013d7b633d116a93943de88e08335357a7c)
1 /*	$NetBSD: ext2fs_vfsops.c,v 1.219 2020/05/16 18:31:53 christos 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 1.219 2020/05/16 18:31:53 christos 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/file.h>
81 #include <sys/disklabel.h>
82 #include <sys/ioctl.h>
83 #include <sys/errno.h>
84 #include <sys/pool.h>
85 #include <sys/lock.h>
86 #include <sys/conf.h>
87 #include <sys/kauth.h>
88 #include <sys/module.h>
89 
90 #include <miscfs/genfs/genfs.h>
91 #include <miscfs/specfs/specdev.h>
92 
93 #include <ufs/ufs/quota.h>
94 #include <ufs/ufs/ufsmount.h>
95 #include <ufs/ufs/inode.h>
96 #include <ufs/ufs/dir.h>
97 #include <ufs/ufs/ufs_extern.h>
98 
99 #include <ufs/ext2fs/ext2fs.h>
100 #include <ufs/ext2fs/ext2fs_dir.h>
101 #include <ufs/ext2fs/ext2fs_extern.h>
102 
103 MODULE(MODULE_CLASS_VFS, ext2fs, "ufs");
104 
105 int ext2fs_sbupdate(struct ufsmount *, int);
106 static int ext2fs_sbfill(struct m_ext2fs *, int);
107 
108 extern const struct vnodeopv_desc ext2fs_vnodeop_opv_desc;
109 extern const struct vnodeopv_desc ext2fs_specop_opv_desc;
110 extern const struct vnodeopv_desc ext2fs_fifoop_opv_desc;
111 
112 const struct vnodeopv_desc * const ext2fs_vnodeopv_descs[] = {
113 	&ext2fs_vnodeop_opv_desc,
114 	&ext2fs_specop_opv_desc,
115 	&ext2fs_fifoop_opv_desc,
116 	NULL,
117 };
118 
119 struct vfsops ext2fs_vfsops = {
120 	.vfs_name = MOUNT_EXT2FS,
121 	.vfs_min_mount_data = sizeof (struct ufs_args),
122 	.vfs_mount = ext2fs_mount,
123 	.vfs_start = ufs_start,
124 	.vfs_unmount = ext2fs_unmount,
125 	.vfs_root = ufs_root,
126 	.vfs_quotactl = ufs_quotactl,
127 	.vfs_statvfs = ext2fs_statvfs,
128 	.vfs_sync = ext2fs_sync,
129 	.vfs_vget = ufs_vget,
130 	.vfs_loadvnode = ext2fs_loadvnode,
131 	.vfs_newvnode = ext2fs_newvnode,
132 	.vfs_fhtovp = ext2fs_fhtovp,
133 	.vfs_vptofh = ext2fs_vptofh,
134 	.vfs_init = ext2fs_init,
135 	.vfs_reinit = ext2fs_reinit,
136 	.vfs_done = ext2fs_done,
137 	.vfs_mountroot = ext2fs_mountroot,
138 	.vfs_snapshot = (void *)eopnotsupp,
139 	.vfs_extattrctl = vfs_stdextattrctl,
140 	.vfs_suspendctl = genfs_suspendctl,
141 	.vfs_renamelock_enter = genfs_renamelock_enter,
142 	.vfs_renamelock_exit = genfs_renamelock_exit,
143 	.vfs_fsync = (void *)eopnotsupp,
144 	.vfs_opv_descs = ext2fs_vnodeopv_descs
145 };
146 
147 static const struct genfs_ops ext2fs_genfsops = {
148 	.gop_size = genfs_size,
149 	.gop_alloc = ext2fs_gop_alloc,
150 	.gop_write = genfs_gop_write,
151 	.gop_markupdate = ufs_gop_markupdate,
152 	.gop_putrange = genfs_gop_putrange,
153 };
154 
155 static const struct ufs_ops ext2fs_ufsops = {
156 	.uo_itimes = ext2fs_itimes,
157 	.uo_update = ext2fs_update,
158 	.uo_bufrd = ext2fs_bufrd,
159 	.uo_bufwr = ext2fs_bufwr,
160 };
161 
162 /* Fill in the inode uid/gid from ext2 halves.  */
163 void
164 ext2fs_set_inode_guid(struct inode *ip)
165 {
166 
167 	ip->i_gid = ip->i_e2fs_gid;
168 	ip->i_uid = ip->i_e2fs_uid;
169 	if (ip->i_e2fs->e2fs.e2fs_rev > E2FS_REV0) {
170 		ip->i_gid |= ip->i_e2fs_gid_high << 16;
171 		ip->i_uid |= ip->i_e2fs_uid_high << 16;
172 	}
173 }
174 
175 SYSCTL_SETUP(ext2fs_sysctl_setup, "ext2fs sysctl")
176 {
177 
178 		sysctl_createv(clog, 0, NULL, NULL,
179 			       CTLFLAG_PERMANENT,
180 			       CTLTYPE_NODE, "ext2fs",
181 			       SYSCTL_DESCR("Linux EXT2FS file system"),
182 			       NULL, 0, NULL, 0,
183 			       CTL_VFS, 17, CTL_EOL);
184 		/*
185 		 * XXX the "17" above could be dynamic, thereby eliminating
186 		 * one more instance of the "number to vfs" mapping problem,
187 		 * but "17" is the order as taken from sys/mount.h
188 		 */
189 }
190 
191 static int
192 ext2fs_modcmd(modcmd_t cmd, void *arg)
193 {
194 	int error;
195 
196 	switch (cmd) {
197 	case MODULE_CMD_INIT:
198 		error = vfs_attach(&ext2fs_vfsops);
199 		if (error != 0)
200 			break;
201 		break;
202 	case MODULE_CMD_FINI:
203 		error = vfs_detach(&ext2fs_vfsops);
204 		if (error != 0)
205 			break;
206 		break;
207 	default:
208 		error = ENOTTY;
209 		break;
210 	}
211 
212 	return error;
213 }
214 
215 /*
216  * XXX Same structure as FFS inodes?  Should we share a common pool?
217  */
218 struct pool ext2fs_inode_pool;
219 
220 extern u_long ext2gennumber;
221 
222 void
223 ext2fs_init(void)
224 {
225 
226 	pool_init(&ext2fs_inode_pool, sizeof(struct inode), 0, 0, 0,
227 	    "ext2fsinopl", &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 }
244 
245 static void
246 ext2fs_sb_setmountinfo(struct m_ext2fs *fs, struct mount *mp)
247 {
248 	(void)strlcpy(fs->e2fs_fsmnt, mp->mnt_stat.f_mntonname,
249             sizeof(fs->e2fs_fsmnt));
250 	if (fs->e2fs_ronly == 0 && fs->e2fs.e2fs_rev > E2FS_REV0) {
251 		(void)strlcpy(fs->e2fs.e2fs_fsmnt, mp->mnt_stat.f_mntonname,
252 		    sizeof(fs->e2fs.e2fs_fsmnt));
253 
254 		fs->e2fs.e2fs_mtime = time_second;
255 		fs->e2fs.e2fs_mnt_count++;
256 
257 		fs->e2fs_fmod = 1;
258 	}
259 }
260 
261 /*
262  * Called by main() when ext2fs is going to be mounted as root.
263  *
264  * Name is updated by mount(8) after booting.
265  */
266 
267 int
268 ext2fs_mountroot(void)
269 {
270 	extern struct vnode *rootvp;
271 	struct m_ext2fs *fs;
272 	struct mount *mp;
273 	struct ufsmount *ump;
274 	int error;
275 
276 	if (device_class(root_device) != DV_DISK)
277 		return ENODEV;
278 
279 	if ((error = vfs_rootmountalloc(MOUNT_EXT2FS, "root_device", &mp))) {
280 		vrele(rootvp);
281 		return error;
282 	}
283 
284 	if ((error = ext2fs_mountfs(rootvp, mp)) != 0) {
285 		vfs_unbusy(mp);
286 		vfs_rele(mp);
287 		return error;
288 	}
289 	mountlist_append(mp);
290 	ump = VFSTOUFS(mp);
291 	fs = ump->um_e2fs;
292 	ext2fs_sb_setmountinfo(fs, mp);
293 	(void)ext2fs_statvfs(mp, &mp->mnt_stat);
294 	vfs_unbusy(mp);
295 	setrootfstime((time_t)fs->e2fs.e2fs_wtime);
296 	return 0;
297 }
298 
299 /*
300  * VFS Operations.
301  *
302  * mount system call
303  */
304 int
305 ext2fs_mount(struct mount *mp, const char *path, void *data, size_t *data_len)
306 {
307 	struct lwp *l = curlwp;
308 	struct vnode *devvp;
309 	struct ufs_args *args = data;
310 	struct ufsmount *ump = NULL;
311 	struct m_ext2fs *fs;
312 	int error = 0, flags, update;
313 	mode_t accessmode;
314 
315 	if (args == NULL)
316 		return EINVAL;
317 	if (*data_len < sizeof *args)
318 		return EINVAL;
319 
320 	if (mp->mnt_flag & MNT_GETARGS) {
321 		ump = VFSTOUFS(mp);
322 		if (ump == NULL)
323 			return EIO;
324 		memset(args, 0, sizeof *args);
325 		args->fspec = NULL;
326 		*data_len = sizeof *args;
327 		return 0;
328 	}
329 
330 	update = mp->mnt_flag & MNT_UPDATE;
331 
332 	/* Check arguments */
333 	if (args->fspec != NULL) {
334 		/*
335 		 * Look up the name and verify that it's sane.
336 		 */
337 		error = namei_simple_user(args->fspec,
338 					NSM_FOLLOW_NOEMULROOT, &devvp);
339 		if (error != 0)
340 			return error;
341 
342 		if (!update) {
343 			/*
344 			 * Be sure this is a valid block device
345 			 */
346 			if (devvp->v_type != VBLK)
347 				error = ENOTBLK;
348 			else if (bdevsw_lookup(devvp->v_rdev) == NULL)
349 				error = ENXIO;
350 		} else {
351 		        /*
352 			 * Be sure we're still naming the same device
353 			 * used for our initial mount
354 			 */
355 			ump = VFSTOUFS(mp);
356 			if (devvp != ump->um_devvp) {
357 				if (devvp->v_rdev != ump->um_devvp->v_rdev)
358 					error = EINVAL;
359 				else {
360 					vrele(devvp);
361 					devvp = ump->um_devvp;
362 					vref(devvp);
363 				}
364 			}
365 		}
366 	} else {
367 		if (!update) {
368 			/* New mounts must have a filename for the device */
369 			return EINVAL;
370 		} else {
371 			ump = VFSTOUFS(mp);
372 			devvp = ump->um_devvp;
373 			vref(devvp);
374 		}
375 	}
376 
377 	/*
378 	 * If mount by non-root, then verify that user has necessary
379 	 * permissions on the device.
380 	 *
381 	 * Permission to update a mount is checked higher, so here we presume
382 	 * updating the mount is okay (for example, as far as securelevel goes)
383 	 * which leaves us with the normal check.
384 	 */
385 	if (error == 0) {
386 		accessmode = VREAD;
387 		if (update ?
388 		    (mp->mnt_iflag & IMNT_WANTRDWR) != 0 :
389 		    (mp->mnt_flag & MNT_RDONLY) == 0)
390 			accessmode |= VWRITE;
391 		vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
392 		error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT,
393 		    KAUTH_REQ_SYSTEM_MOUNT_DEVICE, mp, devvp,
394 		    KAUTH_ARG(accessmode));
395 		VOP_UNLOCK(devvp);
396 	}
397 
398 	if (error) {
399 		vrele(devvp);
400 		return error;
401 	}
402 
403 	if (!update) {
404 		int xflags;
405 
406 		if (mp->mnt_flag & MNT_RDONLY)
407 			xflags = FREAD;
408 		else
409 			xflags = FREAD|FWRITE;
410 		vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
411 		error = VOP_OPEN(devvp, xflags, FSCRED);
412 		VOP_UNLOCK(devvp);
413 		if (error)
414 			goto fail;
415 		error = ext2fs_mountfs(devvp, mp);
416 		if (error) {
417 			vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
418 			(void)VOP_CLOSE(devvp, xflags, NOCRED);
419 			VOP_UNLOCK(devvp);
420 			goto fail;
421 		}
422 
423 		ump = VFSTOUFS(mp);
424 		fs = ump->um_e2fs;
425 	} else {
426 		/*
427 		 * Update the mount.
428 		 */
429 
430 		/*
431 		 * The initial mount got a reference on this
432 		 * device, so drop the one obtained via
433 		 * namei(), above.
434 		 */
435 		vrele(devvp);
436 
437 		ump = VFSTOUFS(mp);
438 		fs = ump->um_e2fs;
439 		if (fs->e2fs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) {
440 			/*
441 			 * Changing from r/w to r/o
442 			 */
443 			flags = WRITECLOSE;
444 			if (mp->mnt_flag & MNT_FORCE)
445 				flags |= FORCECLOSE;
446 			error = ext2fs_flushfiles(mp, flags);
447 			if (error == 0 &&
448 			    ext2fs_cgupdate(ump, MNT_WAIT) == 0 &&
449 			    (fs->e2fs.e2fs_state & E2FS_ERRORS) == 0) {
450 				fs->e2fs.e2fs_state = E2FS_ISCLEAN;
451 				(void) ext2fs_sbupdate(ump, MNT_WAIT);
452 			}
453 			if (error)
454 				return error;
455 			fs->e2fs_ronly = 1;
456 		}
457 
458 		if (mp->mnt_flag & MNT_RELOAD) {
459 			error = ext2fs_reload(mp, l->l_cred, l);
460 			if (error)
461 				return error;
462 		}
463 
464 		if (fs->e2fs_ronly && (mp->mnt_iflag & IMNT_WANTRDWR)) {
465 			/*
466 			 * Changing from read-only to read/write
467 			 */
468 			fs->e2fs_ronly = 0;
469 			if (fs->e2fs.e2fs_state == E2FS_ISCLEAN)
470 				fs->e2fs.e2fs_state = 0;
471 			else
472 				fs->e2fs.e2fs_state = E2FS_ERRORS;
473 			fs->e2fs_fmod = 1;
474 		}
475 		if (args->fspec == NULL)
476 			return 0;
477 	}
478 
479 	error = set_statvfs_info(path, UIO_USERSPACE, args->fspec,
480 	    UIO_USERSPACE, mp->mnt_op->vfs_name, mp, l);
481 	if (error == 0)
482 		ext2fs_sb_setmountinfo(fs, mp);
483 
484 	if (fs->e2fs_fmod != 0) {	/* XXX */
485 		fs->e2fs_fmod = 0;
486 		if (fs->e2fs.e2fs_state == 0)
487 			fs->e2fs.e2fs_wtime = time_second;
488 		else
489 			printf("%s: file system not clean; please fsck(8)\n",
490 				mp->mnt_stat.f_mntfromname);
491 		(void) ext2fs_cgupdate(ump, MNT_WAIT);
492 	}
493 	return error;
494 
495 fail:
496 	vrele(devvp);
497 	return error;
498 }
499 
500 /*
501  * Sanity check the disk vnode content, and copy it over to inode structure.
502  */
503 static int
504 ext2fs_loadvnode_content(struct m_ext2fs *fs, ino_t ino, struct buf *bp, struct inode *ip)
505 {
506 	struct ext2fs_dinode *din;
507 	int error = 0;
508 
509 	din = (struct ext2fs_dinode *)((char *)bp->b_data + (ino_to_fsbo(fs, ino) * EXT2_DINODE_SIZE(fs)));
510 
511 	/* sanity checks - inode data NOT byteswapped at this point */
512 	if (EXT2_DINODE_FITS(din, e2di_extra_isize, EXT2_DINODE_SIZE(fs))
513 	    && (EXT2_DINODE_SIZE(fs) - EXT2_REV0_DINODE_SIZE) < fs2h16(din->e2di_extra_isize))
514 	{
515 		printf("ext2fs: inode %"PRIu64" bad extra_isize %u",
516 			ino, din->e2di_extra_isize);
517 		error = EINVAL;
518 		goto bad;
519 	}
520 
521 	/* everything allright, proceed with copy */
522 	if (ip->i_din.e2fs_din == NULL)
523 		ip->i_din.e2fs_din = kmem_alloc(EXT2_DINODE_SIZE(fs), KM_SLEEP);
524 
525 	e2fs_iload(din, ip->i_din.e2fs_din, EXT2_DINODE_SIZE(fs));
526 
527 	ext2fs_set_inode_guid(ip);
528 
529     bad:
530 	return error;
531 }
532 
533 /*
534  * Reload all incore data for a filesystem (used after running fsck on
535  * the root filesystem and finding things to fix). The filesystem must
536  * be mounted read-only.
537  *
538  * Things to do to update the mount:
539  *	1) invalidate all cached meta-data.
540  *	2) re-read superblock from disk.
541  *	3) re-read summary information from disk.
542  *	4) invalidate all inactive vnodes.
543  *	5) invalidate all cached file data.
544  *	6) re-read inode data for all active vnodes.
545  */
546 int
547 ext2fs_reload(struct mount *mp, kauth_cred_t cred, struct lwp *l)
548 {
549 	struct vnode *vp, *devvp;
550 	struct inode *ip;
551 	struct buf *bp;
552 	struct m_ext2fs *fs;
553 	struct ext2fs *newfs;
554 	int i, error;
555 	struct ufsmount *ump;
556 	struct vnode_iterator *marker;
557 
558 	if ((mp->mnt_flag & MNT_RDONLY) == 0)
559 		return EINVAL;
560 
561 	ump = VFSTOUFS(mp);
562 	/*
563 	 * Step 1: invalidate all cached meta-data.
564 	 */
565 	devvp = ump->um_devvp;
566 	vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
567 	error = vinvalbuf(devvp, 0, cred, l, 0, 0);
568 	VOP_UNLOCK(devvp);
569 	if (error)
570 		panic("ext2fs_reload: dirty1");
571 
572 	fs = ump->um_e2fs;
573 	/*
574 	 * Step 2: re-read superblock from disk. Copy in new superblock, and compute
575 	 * in-memory values.
576 	 */
577 	error = bread(devvp, SBLOCK, SBSIZE, 0, &bp);
578 	if (error)
579 		return error;
580 	newfs = (struct ext2fs *)bp->b_data;
581 	e2fs_sbload(newfs, &fs->e2fs);
582 
583 	brelse(bp, 0);
584 
585 	error = ext2fs_sbfill(fs, (mp->mnt_flag & MNT_RDONLY) != 0);
586 	if (error)
587 		return error;
588 
589 	/*
590 	 * Step 3: re-read summary information from disk.
591 	 */
592 	for (i = 0; i < fs->e2fs_ngdb; i++) {
593 		error = bread(devvp ,
594 		    EXT2_FSBTODB(fs, fs->e2fs.e2fs_first_dblock +
595 		    1 /* superblock */ + i),
596 		    fs->e2fs_bsize, 0, &bp);
597 		if (error) {
598 			return error;
599 		}
600 		e2fs_cgload((struct ext2_gd *)bp->b_data,
601 		    &fs->e2fs_gd[i * fs->e2fs_bsize / sizeof(struct ext2_gd)],
602 		    fs->e2fs_bsize);
603 		brelse(bp, 0);
604 	}
605 
606 	vfs_vnode_iterator_init(mp, &marker);
607 	while ((vp = vfs_vnode_iterator_next(marker, NULL, NULL))) {
608 		/*
609 		 * Step 4: invalidate all inactive vnodes.
610 		 */
611 		if (vrecycle(vp))
612 			continue;
613 		/*
614 		 * Step 5: invalidate all cached file data.
615 		 */
616 		if (vn_lock(vp, LK_EXCLUSIVE)) {
617 			vrele(vp);
618 			continue;
619 		}
620 		if (vinvalbuf(vp, 0, cred, l, 0, 0))
621 			panic("ext2fs_reload: dirty2");
622 		/*
623 		 * Step 6: re-read inode data for all active vnodes.
624 		 */
625 		ip = VTOI(vp);
626 		error = bread(devvp, EXT2_FSBTODB(fs, ino_to_fsba(fs, ip->i_number)),
627 		    (int)fs->e2fs_bsize, 0, &bp);
628 		if (error) {
629 			vput(vp);
630 			break;
631 		}
632 		error = ext2fs_loadvnode_content(fs, ip->i_number, bp, ip);
633 		brelse(bp, 0);
634 		if (error) {
635 			vput(vp);
636 			break;
637 		}
638 
639 		vput(vp);
640 	}
641 	vfs_vnode_iterator_destroy(marker);
642 	return error;
643 }
644 
645 /*
646  * Common code for mount and mountroot
647  */
648 int
649 ext2fs_mountfs(struct vnode *devvp, struct mount *mp)
650 {
651 	struct lwp *l = curlwp;
652 	struct ufsmount *ump;
653 	struct buf *bp;
654 	struct ext2fs *fs;
655 	struct m_ext2fs *m_fs;
656 	dev_t dev;
657 	int error, i, ronly;
658 	kauth_cred_t cred;
659 
660 	dev = devvp->v_rdev;
661 	cred = l->l_cred;
662 
663 	/* Flush out any old buffers remaining from a previous use. */
664 	vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
665 	error = vinvalbuf(devvp, V_SAVE, cred, l, 0, 0);
666 	VOP_UNLOCK(devvp);
667 	if (error)
668 		return error;
669 
670 	ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
671 
672 	bp = NULL;
673 	ump = NULL;
674 
675 	/* Read the superblock from disk, and swap it directly. */
676 	error = bread(devvp, SBLOCK, SBSIZE, 0, &bp);
677 	if (error)
678 		goto out;
679 	fs = (struct ext2fs *)bp->b_data;
680 	m_fs = kmem_zalloc(sizeof(*m_fs), KM_SLEEP);
681 	e2fs_sbload(fs, &m_fs->e2fs);
682 
683 	brelse(bp, 0);
684 	bp = NULL;
685 
686 	/* Once swapped, validate and fill in the superblock. */
687 	error = ext2fs_sbfill(m_fs, ronly);
688 	if (error) {
689 		kmem_free(m_fs, sizeof(*m_fs));
690 		goto out;
691 	}
692 	m_fs->e2fs_ronly = ronly;
693 
694 	ump = kmem_zalloc(sizeof(*ump), KM_SLEEP);
695 	ump->um_fstype = UFS1;
696 	ump->um_ops = &ext2fs_ufsops;
697 	ump->um_e2fs = m_fs;
698 
699 	if (ronly == 0) {
700 		if (m_fs->e2fs.e2fs_state == E2FS_ISCLEAN)
701 			m_fs->e2fs.e2fs_state = 0;
702 		else
703 			m_fs->e2fs.e2fs_state = E2FS_ERRORS;
704 		m_fs->e2fs_fmod = 1;
705 	}
706 
707 	/* XXX: should be added in ext2fs_sbfill()? */
708 	m_fs->e2fs_gd = kmem_alloc(m_fs->e2fs_ngdb * m_fs->e2fs_bsize, KM_SLEEP);
709 	for (i = 0; i < m_fs->e2fs_ngdb; i++) {
710 		error = bread(devvp,
711 		    EXT2_FSBTODB(m_fs, m_fs->e2fs.e2fs_first_dblock +
712 		    1 /* superblock */ + i),
713 		    m_fs->e2fs_bsize, 0, &bp);
714 		if (error) {
715 			kmem_free(m_fs->e2fs_gd,
716 			    m_fs->e2fs_ngdb * m_fs->e2fs_bsize);
717 			goto out;
718 		}
719 		e2fs_cgload((struct ext2_gd *)bp->b_data,
720 		    &m_fs->e2fs_gd[
721 			i * m_fs->e2fs_bsize / sizeof(struct ext2_gd)],
722 		    m_fs->e2fs_bsize);
723 		brelse(bp, 0);
724 		bp = NULL;
725 	}
726 
727 	error = ext2fs_cg_verify_and_initialize(devvp, m_fs, ronly);
728 	if (error) {
729 		kmem_free(m_fs->e2fs_gd, m_fs->e2fs_ngdb * m_fs->e2fs_bsize);
730 		goto out;
731 	}
732 
733 	mp->mnt_data = ump;
734 	mp->mnt_stat.f_fsidx.__fsid_val[0] = (long)dev;
735 	mp->mnt_stat.f_fsidx.__fsid_val[1] = makefstype(MOUNT_EXT2FS);
736 	mp->mnt_stat.f_fsid = mp->mnt_stat.f_fsidx.__fsid_val[0];
737 	mp->mnt_stat.f_namemax = EXT2FS_MAXNAMLEN;
738 	mp->mnt_flag |= MNT_LOCAL;
739 	mp->mnt_dev_bshift = DEV_BSHIFT;	/* XXX */
740 	mp->mnt_fs_bshift = m_fs->e2fs_bshift;
741 	mp->mnt_iflag |= IMNT_DTYPE | IMNT_SHRLOOKUP;
742 	ump->um_flags = 0;
743 	ump->um_mountp = mp;
744 	ump->um_dev = dev;
745 	ump->um_devvp = devvp;
746 	ump->um_nindir = EXT2_NINDIR(m_fs);
747 	ump->um_lognindir = ffs(EXT2_NINDIR(m_fs)) - 1;
748 	ump->um_bptrtodb = m_fs->e2fs_fsbtodb;
749 	ump->um_seqinc = 1; /* no frags */
750 	ump->um_maxsymlinklen = EXT2_MAXSYMLINKLEN;
751 	ump->um_dirblksiz = m_fs->e2fs_bsize;
752 	ump->um_maxfilesize = ((uint64_t)0x80000000 * m_fs->e2fs_bsize - 1);
753 	spec_node_setmountedfs(devvp, mp);
754 	return 0;
755 
756 out:
757 	if (bp != NULL)
758 		brelse(bp, 0);
759 	if (ump) {
760 		kmem_free(ump->um_e2fs, sizeof(*m_fs));
761 		kmem_free(ump, sizeof(*ump));
762 		mp->mnt_data = NULL;
763 	}
764 	return error;
765 }
766 
767 /*
768  * unmount system call
769  */
770 int
771 ext2fs_unmount(struct mount *mp, int mntflags)
772 {
773 	struct ufsmount *ump;
774 	struct m_ext2fs *fs;
775 	int error, flags;
776 
777 	flags = 0;
778 	if (mntflags & MNT_FORCE)
779 		flags |= FORCECLOSE;
780 	if ((error = ext2fs_flushfiles(mp, flags)) != 0)
781 		return error;
782 	ump = VFSTOUFS(mp);
783 	fs = ump->um_e2fs;
784 	if (fs->e2fs_ronly == 0 &&
785 		ext2fs_cgupdate(ump, MNT_WAIT) == 0 &&
786 		(fs->e2fs.e2fs_state & E2FS_ERRORS) == 0) {
787 		fs->e2fs.e2fs_state = E2FS_ISCLEAN;
788 		(void) ext2fs_sbupdate(ump, MNT_WAIT);
789 	}
790 	if (ump->um_devvp->v_type != VBAD)
791 		spec_node_setmountedfs(ump->um_devvp, NULL);
792 	vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
793 	error = VOP_CLOSE(ump->um_devvp, fs->e2fs_ronly ? FREAD : FREAD|FWRITE,
794 	    NOCRED);
795 	vput(ump->um_devvp);
796 	kmem_free(fs->e2fs_gd, fs->e2fs_ngdb * fs->e2fs_bsize);
797 	kmem_free(fs, sizeof(*fs));
798 	kmem_free(ump, sizeof(*ump));
799 	mp->mnt_data = NULL;
800 	mp->mnt_flag &= ~MNT_LOCAL;
801 	return error;
802 }
803 
804 /*
805  * Flush out all the files in a filesystem.
806  */
807 int
808 ext2fs_flushfiles(struct mount *mp, int flags)
809 {
810 	extern int doforce;
811 	int error;
812 
813 	if (!doforce)
814 		flags &= ~FORCECLOSE;
815 	error = vflush(mp, NULLVP, flags);
816 	return error;
817 }
818 
819 /*
820  * Get file system statistics.
821  */
822 int
823 ext2fs_statvfs(struct mount *mp, struct statvfs *sbp)
824 {
825 	struct ufsmount *ump;
826 	struct m_ext2fs *fs;
827 	uint32_t overhead, overhead_per_group, ngdb;
828 	int i, ngroups;
829 
830 	ump = VFSTOUFS(mp);
831 	fs = ump->um_e2fs;
832 	if (fs->e2fs.e2fs_magic != E2FS_MAGIC)
833 		panic("ext2fs_statvfs");
834 
835 	/*
836 	 * Compute the overhead (FS structures)
837 	 */
838 	overhead_per_group =
839 	    1 /* block bitmap */ +
840 	    1 /* inode bitmap */ +
841 	    fs->e2fs_itpg;
842 	overhead = fs->e2fs.e2fs_first_dblock +
843 	    fs->e2fs_ncg * overhead_per_group;
844 	if (EXT2F_HAS_COMPAT_FEATURE(fs, EXT2F_COMPAT_SPARSESUPER2)) {
845 		/*
846 		 * Superblock and group descriptions is in group zero,
847 		 * then optionally 0, 1 or 2 extra copies.
848 		 */
849 		ngroups = 1
850 			+ (fs->e2fs.e4fs_backup_bgs[0] ? 1 : 0)
851 			+ (fs->e2fs.e4fs_backup_bgs[1] ? 1 : 0);
852 	} else if (EXT2F_HAS_ROCOMPAT_FEATURE(fs, 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 (EXT2F_HAS_COMPAT_FEATURE(fs, EXT2F_COMPAT_RESIZE))
862 		ngdb += fs->e2fs.e2fs_reserved_ngdb;
863 	overhead += ngroups * (1 /* superblock */ + ngdb);
864 
865 	sbp->f_bsize = fs->e2fs_bsize;
866 	sbp->f_frsize = MINBSIZE << fs->e2fs.e2fs_fsize;
867 	sbp->f_iosize = fs->e2fs_bsize;
868 	sbp->f_blocks = fs->e2fs.e2fs_bcount - overhead;
869 	sbp->f_bfree = fs->e2fs.e2fs_fbcount;
870 	sbp->f_bresvd = fs->e2fs.e2fs_rbcount;
871 	if (sbp->f_bfree > sbp->f_bresvd)
872 		sbp->f_bavail = sbp->f_bfree - sbp->f_bresvd;
873 	else
874 		sbp->f_bavail = 0;
875 	sbp->f_files =  fs->e2fs.e2fs_icount;
876 	sbp->f_ffree = fs->e2fs.e2fs_ficount;
877 	sbp->f_favail = fs->e2fs.e2fs_ficount;
878 	sbp->f_fresvd = 0;
879 	copy_statvfs_info(sbp, mp);
880 	return 0;
881 }
882 
883 static bool
884 ext2fs_sync_selector(void *cl, struct vnode *vp)
885 {
886 	struct inode *ip;
887 
888 	KASSERT(mutex_owned(vp->v_interlock));
889 
890 	ip = VTOI(vp);
891 	/*
892 	 * Skip the vnode/inode if inaccessible.
893 	 */
894 	if (ip == NULL || vp->v_type == VNON)
895 		return false;
896 
897 	if (((ip->i_flag &
898 	      (IN_CHANGE | IN_UPDATE | IN_MODIFIED)) == 0 &&
899 	     LIST_EMPTY(&vp->v_dirtyblkhd) &&
900 	     (vp->v_iflag & VI_ONWORKLST) == 0))
901 		return false;
902 	return true;
903 }
904 
905 /*
906  * Go through the disk queues to initiate sandbagged IO;
907  * go through the inodes to write those that have been modified;
908  * initiate the writing of the super block if it has been modified.
909  *
910  * Note: we are always called with the filesystem marked `MPBUSY'.
911  */
912 int
913 ext2fs_sync(struct mount *mp, int waitfor, kauth_cred_t cred)
914 {
915 	struct vnode *vp;
916 	struct ufsmount *ump = VFSTOUFS(mp);
917 	struct m_ext2fs *fs;
918 	struct vnode_iterator *marker;
919 	int error, allerror = 0;
920 
921 	fs = ump->um_e2fs;
922 	if (fs->e2fs_fmod != 0 && fs->e2fs_ronly != 0) {	/* XXX */
923 		printf("fs = %s\n", fs->e2fs_fsmnt);
924 		panic("update: rofs mod");
925 	}
926 
927 	/*
928 	 * Write back each (modified) inode.
929 	 */
930 	vfs_vnode_iterator_init(mp, &marker);
931 	while ((vp = vfs_vnode_iterator_next(marker, ext2fs_sync_selector,
932 	    NULL)))
933 	{
934 		error = vn_lock(vp, LK_EXCLUSIVE);
935 		if (error) {
936 			vrele(vp);
937 			continue;
938 		}
939 		if (vp->v_type == VREG && waitfor == MNT_LAZY)
940 			error = ext2fs_update(vp, NULL, NULL, 0);
941 		else
942 			error = VOP_FSYNC(vp, cred,
943 			    waitfor == MNT_WAIT ? FSYNC_WAIT : 0, 0, 0);
944 		if (error)
945 			allerror = error;
946 		vput(vp);
947 	}
948 	vfs_vnode_iterator_destroy(marker);
949 	/*
950 	 * Force stale file system control information to be flushed.
951 	 */
952 	if (waitfor != MNT_LAZY) {
953 		vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
954 		if ((error = VOP_FSYNC(ump->um_devvp, cred,
955 		    waitfor == MNT_WAIT ? FSYNC_WAIT : 0, 0, 0)) != 0)
956 			allerror = error;
957 		VOP_UNLOCK(ump->um_devvp);
958 	}
959 	/*
960 	 * Write back modified superblock.
961 	 */
962 	if (fs->e2fs_fmod != 0) {
963 		fs->e2fs_fmod = 0;
964 		fs->e2fs.e2fs_wtime = time_second;
965 		if ((error = ext2fs_cgupdate(ump, waitfor)))
966 			allerror = error;
967 	}
968 	return allerror;
969 }
970 
971 /*
972  * Load inode from disk and initialize vnode.
973  */
974 static int
975 ext2fs_init_vnode(struct ufsmount *ump, struct vnode *vp, ino_t ino)
976 {
977 	struct m_ext2fs *fs;
978 	struct inode *ip;
979 	struct buf *bp;
980 	int error;
981 
982 	fs = ump->um_e2fs;
983 
984 	/* Read in the disk contents for the inode, copy into the inode. */
985 	error = bread(ump->um_devvp, EXT2_FSBTODB(fs, ino_to_fsba(fs, ino)),
986 	    (int)fs->e2fs_bsize, 0, &bp);
987 	if (error)
988 		return error;
989 
990 	/* Allocate and initialize inode. */
991 	ip = pool_get(&ext2fs_inode_pool, PR_WAITOK);
992 	memset(ip, 0, sizeof(struct inode));
993 	ip->i_vnode = vp;
994 	ip->i_ump = ump;
995 	ip->i_e2fs = fs;
996 	ip->i_dev = ump->um_dev;
997 	ip->i_number = ino;
998 	ip->i_e2fs_last_lblk = 0;
999 	ip->i_e2fs_last_blk = 0;
1000 
1001 	error = ext2fs_loadvnode_content(fs, ino, bp, ip);
1002 	brelse(bp, 0);
1003 	if (error) {
1004 		pool_put(&ext2fs_inode_pool, ip);
1005 		return error;
1006 	}
1007 
1008 	/* If the inode was deleted, reset all fields */
1009 	if (ip->i_e2fs_dtime != 0) {
1010 		ip->i_e2fs_mode = 0;
1011 		(void)ext2fs_setsize(ip, 0);
1012 		(void)ext2fs_setnblock(ip, 0);
1013 		memset(ip->i_e2fs_blocks, 0, sizeof(ip->i_e2fs_blocks));
1014 	}
1015 
1016 	/* Initialise vnode with this inode. */
1017 	vp->v_tag = VT_EXT2FS;
1018 	vp->v_op = ext2fs_vnodeop_p;
1019 	vp->v_vflag |= VV_LOCKSWORK;
1020 	vp->v_data = ip;
1021 
1022 	/* Initialize genfs node. */
1023 	genfs_node_init(vp, &ext2fs_genfsops);
1024 
1025 	return 0;
1026 }
1027 
1028 /*
1029  * Read an inode from disk and initialize this vnode / inode pair.
1030  * Caller assures no other thread will try to load this inode.
1031  */
1032 int
1033 ext2fs_loadvnode(struct mount *mp, struct vnode *vp,
1034     const void *key, size_t key_len, const void **new_key)
1035 {
1036 	ino_t ino;
1037 	struct inode *ip;
1038 	struct ufsmount *ump;
1039 	int error;
1040 
1041 	KASSERT(key_len == sizeof(ino));
1042 	memcpy(&ino, key, key_len);
1043 	ump = VFSTOUFS(mp);
1044 
1045 	error = ext2fs_init_vnode(ump, vp, ino);
1046 	if (error)
1047 		return error;
1048 
1049 	ip = VTOI(vp);
1050 
1051 	/* Initialize the vnode from the inode. */
1052 	ext2fs_vinit(mp, ext2fs_specop_p, ext2fs_fifoop_p, &vp);
1053 
1054 	/* Finish inode initialization. */
1055 	ip->i_devvp = ump->um_devvp;
1056 	vref(ip->i_devvp);
1057 
1058 	/*
1059 	 * Set up a generation number for this inode if it does not
1060 	 * already have one. This should only happen on old filesystems.
1061 	 */
1062 
1063 	if (ip->i_e2fs_gen == 0) {
1064 		if (++ext2gennumber < (u_long)time_second)
1065 			ext2gennumber = time_second;
1066 		ip->i_e2fs_gen = ext2gennumber;
1067 		if ((mp->mnt_flag & MNT_RDONLY) == 0)
1068 			ip->i_flag |= IN_MODIFIED;
1069 	}
1070 	uvm_vnp_setsize(vp, ext2fs_size(ip));
1071 	*new_key = &ip->i_number;
1072 	return 0;
1073 }
1074 
1075 /*
1076  * Create a new inode on disk and initialize this vnode / inode pair.
1077  */
1078 int
1079 ext2fs_newvnode(struct mount *mp, struct vnode *dvp, struct vnode *vp,
1080     struct vattr *vap, kauth_cred_t cred, void *extra,
1081     size_t *key_len, const void **new_key)
1082 {
1083 	ino_t ino;
1084 	struct inode *ip, *pdir;
1085 	struct m_ext2fs *fs;
1086 	struct ufsmount *ump;
1087 	int error, mode;
1088 
1089 	KASSERT(dvp->v_mount == mp);
1090 	KASSERT(vap->va_type != VNON);
1091 
1092 	*key_len = sizeof(ino);
1093 
1094 	pdir = VTOI(dvp);
1095 	fs = pdir->i_e2fs;
1096 	ump = VFSTOUFS(mp);
1097 	mode = MAKEIMODE(vap->va_type, vap->va_mode);
1098 
1099 	/* Allocate fresh inode. */
1100 	error = ext2fs_valloc(dvp, mode, cred, &ino);
1101 	if (error)
1102 		return error;
1103 
1104 	/* Attach inode to vnode. */
1105 	error = ext2fs_init_vnode(ump, vp, ino);
1106 	if (error) {
1107 		ext2fs_vfree(dvp, ino, mode);
1108 		return error;
1109 	}
1110 
1111 	ip = VTOI(vp);
1112 
1113 	KASSERT(!E2FS_HAS_GD_CSUM(fs) || (fs->e2fs_gd[ino_to_cg(fs, ino)].ext2bgd_flags & h2fs16(E2FS_BG_INODE_ZEROED)) != 0);
1114 
1115 	/* check for already used inode; makes sense only for ZEROED itable */
1116 	if (__predict_false(ip->i_e2fs_mode && ip->i_e2fs_nlink != 0)) {
1117 		printf("mode = 0%o, nlinks %d, inum = %llu, fs = %s\n",
1118 		    ip->i_e2fs_mode, ip->i_e2fs_nlink,
1119 		    (unsigned long long)ip->i_number, fs->e2fs_fsmnt);
1120 		panic("ext2fs_valloc: dup alloc");
1121 	}
1122 
1123 	memset(ip->i_din.e2fs_din, 0, EXT2_DINODE_SIZE(fs));
1124 
1125 	/*
1126 	 * Set up a new generation number for this inode.
1127 	 */
1128 	if (++ext2gennumber < time_second)
1129 		ext2gennumber = time_second;
1130 	ip->i_e2fs_gen = ext2gennumber;
1131 
1132 	ip->i_uid = kauth_cred_geteuid(cred);
1133 	ip->i_e2fs_uid = ip->i_uid & 0xffff;
1134 	ip->i_e2fs_gid = pdir->i_e2fs_gid;
1135 	if (ip->i_e2fs->e2fs.e2fs_rev > E2FS_REV0) {
1136 		ip->i_e2fs_uid_high = (ip->i_uid >> 16) & 0xffff;
1137 		ip->i_e2fs_gid_high = pdir->i_e2fs_gid_high;
1138 	} else {
1139 		ip->i_e2fs_uid_high = 0;
1140 		ip->i_e2fs_gid_high = 0;
1141 	}
1142 	ip->i_gid = ip->i_e2fs_gid | (ip->i_e2fs_gid_high << 16);
1143 	ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
1144 	ip->i_e2fs_mode = mode;
1145 	vp->v_type = IFTOVT(mode);
1146 	ip->i_e2fs_nlink = 1;
1147 
1148 	/* Authorize setting SGID if needed. */
1149 	if (ip->i_e2fs_mode & ISGID) {
1150 		error = kauth_authorize_vnode(cred, KAUTH_VNODE_WRITE_SECURITY,
1151 		    vp, NULL, genfs_can_chmod(vp, cred, ip->i_uid, ip->i_gid,
1152 		    mode));
1153 		if (error)
1154 			ip->i_e2fs_mode &= ~ISGID;
1155 	}
1156 
1157 	/* Initialize extra_isize according to what is set in superblock */
1158 	if (EXT2F_HAS_ROCOMPAT_FEATURE(ip->i_e2fs, EXT2F_ROCOMPAT_EXTRA_ISIZE)
1159 	    && EXT2_DINODE_SIZE(ip->i_e2fs) > EXT2_REV0_DINODE_SIZE) {
1160 		ip->i_din.e2fs_din->e2di_extra_isize = ip->i_e2fs->e2fs.e4fs_want_extra_isize;
1161 	}
1162 
1163 	/* Set create time if possible */
1164 	if (EXT2_DINODE_FITS(ip->i_din.e2fs_din, e2di_crtime, EXT2_DINODE_SIZE(ip->i_e2fs))) {
1165 		struct timespec now;
1166 		vfs_timestamp(&now);
1167 		EXT2_DINODE_TIME_SET(&now, ip->i_din.e2fs_din, e2di_crtime, EXT2_DINODE_SIZE(ip->i_e2fs));
1168 	}
1169 
1170 	/* Initialize the vnode from the inode. */
1171 	ext2fs_vinit(mp, ext2fs_specop_p, ext2fs_fifoop_p, &vp);
1172 
1173 	/* Finish inode initialization. */
1174 	ip->i_devvp = ump->um_devvp;
1175 	vref(ip->i_devvp);
1176 
1177 	uvm_vnp_setsize(vp, ext2fs_size(ip));
1178 	*new_key = &ip->i_number;
1179 	return 0;
1180 }
1181 
1182 /*
1183  * File handle to vnode
1184  *
1185  * Have to be really careful about stale file handles:
1186  * - check that the inode number is valid
1187  * - call ext2fs_vget() to get the locked inode
1188  * - check for an unallocated inode (i_mode == 0)
1189  */
1190 int
1191 ext2fs_fhtovp(struct mount *mp, struct fid *fhp, int lktype, struct vnode **vpp)
1192 {
1193 	struct inode *ip;
1194 	struct vnode *nvp;
1195 	int error;
1196 	struct ufid ufh;
1197 	struct m_ext2fs *fs;
1198 
1199 	if (fhp->fid_len != sizeof(struct ufid))
1200 		return EINVAL;
1201 
1202 	memcpy(&ufh, fhp, sizeof(struct ufid));
1203 	fs = VFSTOUFS(mp)->um_e2fs;
1204 	if ((ufh.ufid_ino < EXT2_FIRSTINO && ufh.ufid_ino != EXT2_ROOTINO) ||
1205 		ufh.ufid_ino >= fs->e2fs_ncg * fs->e2fs.e2fs_ipg)
1206 		return ESTALE;
1207 
1208 	if ((error = VFS_VGET(mp, ufh.ufid_ino, lktype, &nvp)) != 0) {
1209 		*vpp = NULLVP;
1210 		return error;
1211 	}
1212 	ip = VTOI(nvp);
1213 	if (ip->i_e2fs_mode == 0 || ip->i_e2fs_dtime != 0 ||
1214 		ip->i_e2fs_gen != ufh.ufid_gen) {
1215 		vput(nvp);
1216 		*vpp = NULLVP;
1217 		return ESTALE;
1218 	}
1219 	*vpp = nvp;
1220 	return 0;
1221 }
1222 
1223 /*
1224  * Vnode pointer to File handle
1225  */
1226 /* ARGSUSED */
1227 int
1228 ext2fs_vptofh(struct vnode *vp, struct fid *fhp, size_t *fh_size)
1229 {
1230 	struct inode *ip;
1231 	struct ufid ufh;
1232 
1233 	if (*fh_size < sizeof(struct ufid)) {
1234 		*fh_size = sizeof(struct ufid);
1235 		return E2BIG;
1236 	}
1237 	*fh_size = sizeof(struct ufid);
1238 
1239 	ip = VTOI(vp);
1240 	memset(&ufh, 0, sizeof(ufh));
1241 	ufh.ufid_len = sizeof(struct ufid);
1242 	ufh.ufid_ino = ip->i_number;
1243 	ufh.ufid_gen = ip->i_e2fs_gen;
1244 	memcpy(fhp, &ufh, sizeof(ufh));
1245 	return 0;
1246 }
1247 
1248 /*
1249  * Write a superblock and associated information back to disk.
1250  */
1251 int
1252 ext2fs_sbupdate(struct ufsmount *mp, int waitfor)
1253 {
1254 	struct m_ext2fs *fs = mp->um_e2fs;
1255 	struct buf *bp;
1256 	int error = 0;
1257 
1258 	bp = getblk(mp->um_devvp, SBLOCK, SBSIZE, 0, 0);
1259 	e2fs_sbsave(&fs->e2fs, (struct ext2fs*)bp->b_data);
1260 	if (waitfor == MNT_WAIT)
1261 		error = bwrite(bp);
1262 	else
1263 		bawrite(bp);
1264 	return error;
1265 }
1266 
1267 int
1268 ext2fs_cgupdate(struct ufsmount *mp, int waitfor)
1269 {
1270 	struct m_ext2fs *fs = mp->um_e2fs;
1271 	struct buf *bp;
1272 	int i, error = 0, allerror = 0;
1273 
1274 	allerror = ext2fs_sbupdate(mp, waitfor);
1275 	for (i = 0; i < fs->e2fs_ngdb; i++) {
1276 		bp = getblk(mp->um_devvp, EXT2_FSBTODB(fs,
1277 		    fs->e2fs.e2fs_first_dblock +
1278 		    1 /* superblock */ + i), fs->e2fs_bsize, 0, 0);
1279 		e2fs_cgsave(&fs->e2fs_gd[
1280 		    i * fs->e2fs_bsize / sizeof(struct ext2_gd)],
1281 		    (struct ext2_gd *)bp->b_data, fs->e2fs_bsize);
1282 		if (waitfor == MNT_WAIT)
1283 			error = bwrite(bp);
1284 		else
1285 			bawrite(bp);
1286 	}
1287 
1288 	if (!allerror && error)
1289 		allerror = error;
1290 	return allerror;
1291 }
1292 
1293 /*
1294  * Fill in the m_fs structure, and validate the fields of the superblock.
1295  * NOTE: here, the superblock is already swapped.
1296  */
1297 static int
1298 ext2fs_sbfill(struct m_ext2fs *m_fs, int ronly)
1299 {
1300 	uint32_t u32;
1301 	struct ext2fs *fs = &m_fs->e2fs;
1302 
1303 	/*
1304 	 * General sanity checks
1305 	 */
1306 	if (fs->e2fs_magic != E2FS_MAGIC)
1307 		return EINVAL;
1308 	if (fs->e2fs_rev > E2FS_REV1) {
1309 		printf("ext2fs: unsupported revision number: %x\n", fs->e2fs_rev);
1310 		return EINVAL;
1311 	}
1312 	if (fs->e2fs_log_bsize > 2) {
1313 		/* block size = 1024|2048|4096 */
1314 		printf("ext2fs: bad block size: %d\n", fs->e2fs_log_bsize);
1315 		return EINVAL;
1316 	}
1317 	if (fs->e2fs_bpg == 0) {
1318 		printf("ext2fs: zero blocks per group\n");
1319 		return EINVAL;
1320 	}
1321 	if (fs->e2fs_ipg == 0) {
1322 		printf("ext2fs: zero inodes per group\n");
1323 		return EINVAL;
1324 	}
1325 
1326 	if (fs->e2fs_first_dblock >= fs->e2fs_bcount) {
1327 		printf("ext2fs: invalid first data block\n");
1328 		return EINVAL;
1329 	}
1330 	if (fs->e2fs_rbcount > fs->e2fs_bcount ||
1331 	    fs->e2fs_fbcount > fs->e2fs_bcount) {
1332 		printf("ext2fs: invalid block count\n");
1333 		return EINVAL;
1334 	}
1335 
1336 	/*
1337 	 * Compute the fields of the superblock
1338 	 */
1339 	u32 = fs->e2fs_bcount - fs->e2fs_first_dblock; /* > 0 */
1340 	m_fs->e2fs_ncg = howmany(u32, fs->e2fs_bpg);
1341 	if (m_fs->e2fs_ncg == 0) {
1342 		printf("ext2fs: invalid number of cylinder groups\n");
1343 		return EINVAL;
1344 	}
1345 
1346 	m_fs->e2fs_fsbtodb = fs->e2fs_log_bsize + LOG_MINBSIZE - DEV_BSHIFT;
1347 	m_fs->e2fs_bsize = MINBSIZE << fs->e2fs_log_bsize;
1348 	m_fs->e2fs_bshift = LOG_MINBSIZE + fs->e2fs_log_bsize;
1349 	m_fs->e2fs_qbmask = m_fs->e2fs_bsize - 1;
1350 	m_fs->e2fs_bmask = ~m_fs->e2fs_qbmask;
1351 
1352 	if ((u32 = m_fs->e2fs_bsize / sizeof(struct ext2_gd)) == 0) {
1353 		/* Unlikely to happen */
1354 		printf("ext2fs: invalid block size\n");
1355 		return EINVAL;
1356 	}
1357 	m_fs->e2fs_ngdb = howmany(m_fs->e2fs_ncg, u32);
1358 	if (m_fs->e2fs_ngdb == 0) {
1359 		printf("ext2fs: invalid number of group descriptor blocks\n");
1360 		return EINVAL;
1361 	}
1362 
1363 	if (m_fs->e2fs_bsize < EXT2_DINODE_SIZE(m_fs)) {
1364 		printf("ext2fs: invalid inode size\n");
1365 		return EINVAL;
1366 	}
1367 	m_fs->e2fs_ipb = m_fs->e2fs_bsize / EXT2_DINODE_SIZE(m_fs);
1368 
1369 	m_fs->e2fs_itpg = fs->e2fs_ipg / m_fs->e2fs_ipb;
1370 
1371 	/*
1372 	 * Revision-specific checks
1373 	 */
1374 	if (fs->e2fs_rev > E2FS_REV0) {
1375 		char buf[256];
1376 		if (fs->e2fs_first_ino != EXT2_FIRSTINO) {
1377 			printf("ext2fs: unsupported first inode position\n");
1378 			return EINVAL;
1379 		}
1380 		u32 = fs->e2fs_features_incompat & ~EXT2F_INCOMPAT_SUPP;
1381 		if (u32) {
1382 			snprintb(buf, sizeof(buf), EXT2F_INCOMPAT_BITS, u32);
1383 			printf("ext2fs: unsupported incompat features: %s\n", buf);
1384 #ifndef EXT2_IGNORE_INCOMPAT_FEATURES
1385 			return EINVAL;
1386 #endif
1387 		}
1388 		u32 = fs->e2fs_features_rocompat & ~EXT2F_ROCOMPAT_SUPP;
1389 		if (!ronly && u32) {
1390 			snprintb(buf, sizeof(buf), EXT2F_ROCOMPAT_BITS, u32);
1391 			printf("ext2fs: unsupported ro-incompat features: %s\n",
1392 			    buf);
1393 #ifndef EXT2_IGNORE_ROCOMPAT_FEATURES
1394 			return EROFS;
1395 #endif
1396 		}
1397 		if (fs->e2fs_inode_size == 0 || !powerof2(fs->e2fs_inode_size) || fs->e2fs_inode_size > m_fs->e2fs_bsize) {
1398 			printf("ext2fs: bad inode size\n");
1399 			return EINVAL;
1400 		}
1401 	}
1402 
1403 	return 0;
1404 }
1405