xref: /netbsd-src/sys/ufs/ext2fs/ext2fs_vfsops.c (revision 8ac07aec990b9d2e483062509d0a9fa5b4f57cf2)
1 /*	$NetBSD: ext2fs_vfsops.c,v 1.131 2008/02/05 15:21:19 ad Exp $	*/
2 
3 /*
4  * Copyright (c) 1989, 1991, 1993, 1994
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the University nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  *
31  *	@(#)ffs_vfsops.c	8.14 (Berkeley) 11/28/94
32  * Modified for ext2fs by Manuel Bouyer.
33  */
34 
35 /*
36  * Copyright (c) 1997 Manuel Bouyer.
37  *
38  * Redistribution and use in source and binary forms, with or without
39  * modification, are permitted provided that the following conditions
40  * are met:
41  * 1. Redistributions of source code must retain the above copyright
42  *    notice, this list of conditions and the following disclaimer.
43  * 2. Redistributions in binary form must reproduce the above copyright
44  *    notice, this list of conditions and the following disclaimer in the
45  *    documentation and/or other materials provided with the distribution.
46  * 3. All advertising materials mentioning features or use of this software
47  *    must display the following acknowledgement:
48  *	This product includes software developed by Manuel Bouyer.
49  * 4. The name of the author may not be used to endorse or promote products
50  *    derived from this software without specific prior written permission.
51  *
52  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
53  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
54  * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
55  * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
56  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
57  * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
58  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
59  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
60  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
61  * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
62  *
63  *	@(#)ffs_vfsops.c	8.14 (Berkeley) 11/28/94
64  * Modified for ext2fs by Manuel Bouyer.
65  */
66 
67 #include <sys/cdefs.h>
68 __KERNEL_RCSID(0, "$NetBSD: ext2fs_vfsops.c,v 1.131 2008/02/05 15:21:19 ad Exp $");
69 
70 #if defined(_KERNEL_OPT)
71 #include "opt_compat_netbsd.h"
72 #endif
73 
74 #include <sys/param.h>
75 #include <sys/systm.h>
76 #include <sys/sysctl.h>
77 #include <sys/namei.h>
78 #include <sys/proc.h>
79 #include <sys/kernel.h>
80 #include <sys/vnode.h>
81 #include <sys/socket.h>
82 #include <sys/mount.h>
83 #include <sys/buf.h>
84 #include <sys/device.h>
85 #include <sys/mbuf.h>
86 #include <sys/file.h>
87 #include <sys/disklabel.h>
88 #include <sys/ioctl.h>
89 #include <sys/errno.h>
90 #include <sys/malloc.h>
91 #include <sys/pool.h>
92 #include <sys/lock.h>
93 #include <sys/conf.h>
94 #include <sys/kauth.h>
95 
96 #include <miscfs/genfs/genfs.h>
97 #include <miscfs/specfs/specdev.h>
98 
99 #include <ufs/ufs/quota.h>
100 #include <ufs/ufs/ufsmount.h>
101 #include <ufs/ufs/inode.h>
102 #include <ufs/ufs/dir.h>
103 #include <ufs/ufs/ufs_extern.h>
104 
105 #include <ufs/ext2fs/ext2fs.h>
106 #include <ufs/ext2fs/ext2fs_dir.h>
107 #include <ufs/ext2fs/ext2fs_extern.h>
108 
109 extern kmutex_t ufs_hashlock;
110 
111 int ext2fs_sbupdate(struct ufsmount *, int);
112 static int ext2fs_checksb(struct ext2fs *, int);
113 
114 extern const struct vnodeopv_desc ext2fs_vnodeop_opv_desc;
115 extern const struct vnodeopv_desc ext2fs_specop_opv_desc;
116 extern const struct vnodeopv_desc ext2fs_fifoop_opv_desc;
117 
118 const struct vnodeopv_desc * const ext2fs_vnodeopv_descs[] = {
119 	&ext2fs_vnodeop_opv_desc,
120 	&ext2fs_specop_opv_desc,
121 	&ext2fs_fifoop_opv_desc,
122 	NULL,
123 };
124 
125 struct vfsops ext2fs_vfsops = {
126 	MOUNT_EXT2FS,
127 	sizeof (struct ufs_args),
128 	ext2fs_mount,
129 	ufs_start,
130 	ext2fs_unmount,
131 	ufs_root,
132 	ufs_quotactl,
133 	ext2fs_statvfs,
134 	ext2fs_sync,
135 	ext2fs_vget,
136 	ext2fs_fhtovp,
137 	ext2fs_vptofh,
138 	ext2fs_init,
139 	ext2fs_reinit,
140 	ext2fs_done,
141 	ext2fs_mountroot,
142 	(int (*)(struct mount *, struct vnode *, struct timespec *)) eopnotsupp,
143 	vfs_stdextattrctl,
144 	(void *)eopnotsupp,	/* vfs_suspendctl */
145 	genfs_renamelock_enter,
146 	genfs_renamelock_exit,
147 	ext2fs_vnodeopv_descs,
148 	0,
149 	{ NULL, NULL },
150 };
151 VFS_ATTACH(ext2fs_vfsops);
152 
153 static const struct genfs_ops ext2fs_genfsops = {
154 	.gop_size = genfs_size,
155 	.gop_alloc = ext2fs_gop_alloc,
156 	.gop_write = genfs_gop_write,
157 	.gop_markupdate = ufs_gop_markupdate,
158 };
159 
160 static const struct ufs_ops ext2fs_ufsops = {
161 	.uo_itimes = ext2fs_itimes,
162 	.uo_update = ext2fs_update,
163 	.uo_vfree = ext2fs_vfree,
164 };
165 
166 /*
167  * XXX Same structure as FFS inodes?  Should we share a common pool?
168  */
169 struct pool ext2fs_inode_pool;
170 struct pool ext2fs_dinode_pool;
171 
172 extern u_long ext2gennumber;
173 
174 void
175 ext2fs_init(void)
176 {
177 
178 	pool_init(&ext2fs_inode_pool, sizeof(struct inode), 0, 0, 0,
179 	    "ext2fsinopl", &pool_allocator_nointr, IPL_NONE);
180 	pool_init(&ext2fs_dinode_pool, sizeof(struct ext2fs_dinode), 0, 0, 0,
181 	    "ext2dinopl", &pool_allocator_nointr, IPL_NONE);
182 	ufs_init();
183 }
184 
185 void
186 ext2fs_reinit(void)
187 {
188 	ufs_reinit();
189 }
190 
191 void
192 ext2fs_done(void)
193 {
194 
195 	ufs_done();
196 	pool_destroy(&ext2fs_inode_pool);
197 	pool_destroy(&ext2fs_dinode_pool);
198 }
199 
200 /*
201  * Called by main() when ext2fs is going to be mounted as root.
202  *
203  * Name is updated by mount(8) after booting.
204  */
205 #define ROOTNAME	"root_device"
206 
207 int
208 ext2fs_mountroot(void)
209 {
210 	extern struct vnode *rootvp;
211 	struct m_ext2fs *fs;
212 	struct mount *mp;
213 	struct ufsmount *ump;
214 	int error;
215 
216 	if (device_class(root_device) != DV_DISK)
217 		return (ENODEV);
218 
219 	if ((error = vfs_rootmountalloc(MOUNT_EXT2FS, "root_device", &mp))) {
220 		vrele(rootvp);
221 		return (error);
222 	}
223 
224 	if ((error = ext2fs_mountfs(rootvp, mp)) != 0) {
225 		vfs_unbusy(mp, false);
226 		vfs_destroy(mp);
227 		return (error);
228 	}
229 	mutex_enter(&mountlist_lock);
230 	CIRCLEQ_INSERT_TAIL(&mountlist, mp, mnt_list);
231 	mutex_exit(&mountlist_lock);
232 	ump = VFSTOUFS(mp);
233 	fs = ump->um_e2fs;
234 	memset(fs->e2fs_fsmnt, 0, sizeof(fs->e2fs_fsmnt));
235 	(void) copystr(mp->mnt_stat.f_mntonname, fs->e2fs_fsmnt,
236 	    sizeof(fs->e2fs_fsmnt) - 1, 0);
237 	if (fs->e2fs.e2fs_rev > E2FS_REV0) {
238 		memset(fs->e2fs.e2fs_fsmnt, 0, sizeof(fs->e2fs.e2fs_fsmnt));
239 		(void) copystr(mp->mnt_stat.f_mntonname, fs->e2fs.e2fs_fsmnt,
240 		    sizeof(fs->e2fs.e2fs_fsmnt) - 1, 0);
241 	}
242 	(void)ext2fs_statvfs(mp, &mp->mnt_stat);
243 	vfs_unbusy(mp, false);
244 	setrootfstime((time_t)fs->e2fs.e2fs_wtime);
245 	return (0);
246 }
247 
248 /*
249  * VFS Operations.
250  *
251  * mount system call
252  */
253 int
254 ext2fs_mount(struct mount *mp, const char *path, void *data, size_t *data_len)
255 {
256 	struct lwp *l = curlwp;
257 	struct nameidata nd;
258 	struct vnode *devvp;
259 	struct ufs_args *args = data;
260 	struct ufsmount *ump = NULL;
261 	struct m_ext2fs *fs;
262 	size_t size;
263 	int error = 0, flags, update;
264 	mode_t accessmode;
265 
266 	if (*data_len < sizeof *args)
267 		return EINVAL;
268 
269 	if (mp->mnt_flag & MNT_GETARGS) {
270 		ump = VFSTOUFS(mp);
271 		if (ump == NULL)
272 			return EIO;
273 		memset(args, 0, sizeof *args);
274 		args->fspec = NULL;
275 		*data_len = sizeof *args;
276 		return 0;
277 	}
278 
279 	update = mp->mnt_flag & MNT_UPDATE;
280 
281 	/* Check arguments */
282 	if (args->fspec != NULL) {
283 		/*
284 		 * Look up the name and verify that it's sane.
285 		 */
286 		NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, args->fspec);
287 		if ((error = namei(&nd)) != 0)
288 			return (error);
289 		devvp = nd.ni_vp;
290 
291 		if (!update) {
292 			/*
293 			 * Be sure this is a valid block device
294 			 */
295 			if (devvp->v_type != VBLK)
296 				error = ENOTBLK;
297 			else if (bdevsw_lookup(devvp->v_rdev) == NULL)
298 				error = ENXIO;
299 		} else {
300 		        /*
301 			 * Be sure we're still naming the same device
302 			 * used for our initial mount
303 			 */
304 			ump = VFSTOUFS(mp);
305 			if (devvp != ump->um_devvp)
306 				error = EINVAL;
307 		}
308 	} else {
309 		if (!update) {
310 			/* New mounts must have a filename for the device */
311 			return (EINVAL);
312 		} else {
313 			ump = VFSTOUFS(mp);
314 			devvp = ump->um_devvp;
315 			vref(devvp);
316 		}
317 	}
318 
319 	/*
320 	 * If mount by non-root, then verify that user has necessary
321 	 * permissions on the device.
322 	 */
323 	if (error == 0 && kauth_authorize_generic(l->l_cred,
324 	    KAUTH_GENERIC_ISSUSER, NULL) != 0) {
325 		accessmode = VREAD;
326 		if (update ?
327 		    (mp->mnt_iflag & IMNT_WANTRDWR) != 0 :
328 		    (mp->mnt_flag & MNT_RDONLY) == 0)
329 			accessmode |= VWRITE;
330 		vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
331 		error = VOP_ACCESS(devvp, accessmode, l->l_cred);
332 		VOP_UNLOCK(devvp, 0);
333 	}
334 
335 	if (error) {
336 		vrele(devvp);
337 		return (error);
338 	}
339 
340 	if (!update) {
341 		int xflags;
342 
343 		if (mp->mnt_flag & MNT_RDONLY)
344 			xflags = FREAD;
345 		else
346 			xflags = FREAD|FWRITE;
347 		error = VOP_OPEN(devvp, xflags, FSCRED);
348 		if (error)
349 			goto fail;
350 		error = ext2fs_mountfs(devvp, mp);
351 		if (error) {
352 			vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
353 			(void)VOP_CLOSE(devvp, xflags, NOCRED);
354 			VOP_UNLOCK(devvp, 0);
355 			goto fail;
356 		}
357 
358 		ump = VFSTOUFS(mp);
359 		fs = ump->um_e2fs;
360 	} else {
361 		/*
362 		 * Update the mount.
363 		 */
364 
365 		/*
366 		 * The initial mount got a reference on this
367 		 * device, so drop the one obtained via
368 		 * namei(), above.
369 		 */
370 		vrele(devvp);
371 
372 		ump = VFSTOUFS(mp);
373 		fs = ump->um_e2fs;
374 		if (fs->e2fs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) {
375 			/*
376 			 * Changing from r/w to r/o
377 			 */
378 			flags = WRITECLOSE;
379 			if (mp->mnt_flag & MNT_FORCE)
380 				flags |= FORCECLOSE;
381 			error = ext2fs_flushfiles(mp, flags);
382 			if (error == 0 &&
383 			    ext2fs_cgupdate(ump, MNT_WAIT) == 0 &&
384 			    (fs->e2fs.e2fs_state & E2FS_ERRORS) == 0) {
385 				fs->e2fs.e2fs_state = E2FS_ISCLEAN;
386 				(void) ext2fs_sbupdate(ump, MNT_WAIT);
387 			}
388 			if (error)
389 				return (error);
390 			fs->e2fs_ronly = 1;
391 		}
392 
393 		if (mp->mnt_flag & MNT_RELOAD) {
394 			error = ext2fs_reload(mp, l->l_cred);
395 			if (error)
396 				return (error);
397 		}
398 
399 		if (fs->e2fs_ronly && (mp->mnt_iflag & IMNT_WANTRDWR)) {
400 			/*
401 			 * Changing from read-only to read/write
402 			 */
403 			fs->e2fs_ronly = 0;
404 			if (fs->e2fs.e2fs_state == E2FS_ISCLEAN)
405 				fs->e2fs.e2fs_state = 0;
406 			else
407 				fs->e2fs.e2fs_state = E2FS_ERRORS;
408 			fs->e2fs_fmod = 1;
409 		}
410 		if (args->fspec == NULL)
411 			return EINVAL;
412 	}
413 
414 	error = set_statvfs_info(path, UIO_USERSPACE, args->fspec,
415 	    UIO_USERSPACE, mp->mnt_op->vfs_name, mp, l);
416 	(void) copystr(mp->mnt_stat.f_mntonname, fs->e2fs_fsmnt,
417 	    sizeof(fs->e2fs_fsmnt) - 1, &size);
418 	memset(fs->e2fs_fsmnt + size, 0, sizeof(fs->e2fs_fsmnt) - size);
419 	if (fs->e2fs.e2fs_rev > E2FS_REV0) {
420 		(void) copystr(mp->mnt_stat.f_mntonname, fs->e2fs.e2fs_fsmnt,
421 		    sizeof(fs->e2fs.e2fs_fsmnt) - 1, &size);
422 		memset(fs->e2fs.e2fs_fsmnt, 0,
423 		    sizeof(fs->e2fs.e2fs_fsmnt) - size);
424 	}
425 	if (fs->e2fs_fmod != 0) {	/* XXX */
426 		fs->e2fs_fmod = 0;
427 		if (fs->e2fs.e2fs_state == 0)
428 			fs->e2fs.e2fs_wtime = time_second;
429 		else
430 			printf("%s: file system not clean; please fsck(8)\n",
431 				mp->mnt_stat.f_mntfromname);
432 		(void) ext2fs_cgupdate(ump, MNT_WAIT);
433 	}
434 	return (error);
435 
436 fail:
437 	vrele(devvp);
438 	return (error);
439 }
440 
441 /*
442  * Reload all incore data for a filesystem (used after running fsck on
443  * the root filesystem and finding things to fix). The filesystem must
444  * be mounted read-only.
445  *
446  * Things to do to update the mount:
447  *	1) invalidate all cached meta-data.
448  *	2) re-read superblock from disk.
449  *	3) re-read summary information from disk.
450  *	4) invalidate all inactive vnodes.
451  *	5) invalidate all cached file data.
452  *	6) re-read inode data for all active vnodes.
453  */
454 int
455 ext2fs_reload(struct mount *mountp, kauth_cred_t cred)
456 {
457 	struct lwp *l = curlwp;
458 	struct vnode *vp, *mvp, *devvp;
459 	struct inode *ip;
460 	struct buf *bp;
461 	struct m_ext2fs *fs;
462 	struct ext2fs *newfs;
463 	struct partinfo dpart;
464 	int i, size, error;
465 	void *cp;
466 
467 	if ((mountp->mnt_flag & MNT_RDONLY) == 0)
468 		return (EINVAL);
469 
470 	/*
471 	 * Step 1: invalidate all cached meta-data.
472 	 */
473 	devvp = VFSTOUFS(mountp)->um_devvp;
474 	vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
475 	error = vinvalbuf(devvp, 0, cred, l, 0, 0);
476 	VOP_UNLOCK(devvp, 0);
477 	if (error)
478 		panic("ext2fs_reload: dirty1");
479 	/*
480 	 * Step 2: re-read superblock from disk.
481 	 */
482 	if (VOP_IOCTL(devvp, DIOCGPART, &dpart, FREAD, NOCRED) != 0)
483 		size = DEV_BSIZE;
484 	else
485 		size = dpart.disklab->d_secsize;
486 	error = bread(devvp, (daddr_t)(SBOFF / size), SBSIZE, NOCRED, &bp);
487 	if (error) {
488 		brelse(bp, 0);
489 		return (error);
490 	}
491 	newfs = (struct ext2fs *)bp->b_data;
492 	error = ext2fs_checksb(newfs, (mountp->mnt_flag & MNT_RDONLY) != 0);
493 	if (error) {
494 		brelse(bp, 0);
495 		return (error);
496 	}
497 
498 	fs = VFSTOUFS(mountp)->um_e2fs;
499 	/*
500 	 * copy in new superblock, and compute in-memory values
501 	 */
502 	e2fs_sbload(newfs, &fs->e2fs);
503 	fs->e2fs_ncg =
504 	    howmany(fs->e2fs.e2fs_bcount - fs->e2fs.e2fs_first_dblock,
505 	    fs->e2fs.e2fs_bpg);
506 	/* XXX assume hw bsize = 512 */
507 	fs->e2fs_fsbtodb = fs->e2fs.e2fs_log_bsize + 1;
508 	fs->e2fs_bsize = MINBSIZE << fs->e2fs.e2fs_log_bsize;
509 	fs->e2fs_bshift = LOG_MINBSIZE + fs->e2fs.e2fs_log_bsize;
510 	fs->e2fs_qbmask = fs->e2fs_bsize - 1;
511 	fs->e2fs_bmask = ~fs->e2fs_qbmask;
512 	fs->e2fs_ngdb =
513 	    howmany(fs->e2fs_ncg, fs->e2fs_bsize / sizeof(struct ext2_gd));
514 	fs->e2fs_ipb = fs->e2fs_bsize / EXT2_DINODE_SIZE;
515 	fs->e2fs_itpg = fs->e2fs.e2fs_ipg / fs->e2fs_ipb;
516 
517 	/*
518 	 * Step 3: re-read summary information from disk.
519 	 */
520 
521 	for (i = 0; i < fs->e2fs_ngdb; i++) {
522 		error = bread(devvp ,
523 		    fsbtodb(fs, fs->e2fs.e2fs_first_dblock +
524 		    1 /* superblock */ + i),
525 		    fs->e2fs_bsize, NOCRED, &bp);
526 		if (error) {
527 			brelse(bp, 0);
528 			return (error);
529 		}
530 		e2fs_cgload((struct ext2_gd *)bp->b_data,
531 		    &fs->e2fs_gd[i * fs->e2fs_bsize / sizeof(struct ext2_gd)],
532 		    fs->e2fs_bsize);
533 		brelse(bp, 0);
534 	}
535 
536 	/* Allocate a marker vnode. */
537 	if ((mvp = vnalloc(mountp)) == NULL)
538 		return (ENOMEM);
539 	/*
540 	 * NOTE: not using the TAILQ_FOREACH here since in this loop vgone()
541 	 * and vclean() can be called indirectly
542 	 */
543 	mutex_enter(&mntvnode_lock);
544 loop:
545 	for (vp = TAILQ_FIRST(&mountp->mnt_vnodelist); vp; vp = vunmark(mvp)) {
546 		vmark(mvp, vp);
547 		if (vp->v_mount != mountp || vismarker(vp))
548 			continue;
549 		/*
550 		 * Step 4: invalidate all inactive vnodes.
551 		 */
552 		if (vrecycle(vp, &mntvnode_lock, l)) {
553 			mutex_enter(&mntvnode_lock);
554 			(void)vunmark(mvp);
555 			goto loop;
556 		}
557 		/*
558 		 * Step 5: invalidate all cached file data.
559 		 */
560 		mutex_enter(&vp->v_interlock);
561 		mutex_exit(&mntvnode_lock);
562 		if (vget(vp, LK_EXCLUSIVE | LK_INTERLOCK)) {
563 			mutex_enter(&mntvnode_lock);
564 			(void)vunmark(mvp);
565 			goto loop;
566 		}
567 		if (vinvalbuf(vp, 0, cred, l, 0, 0))
568 			panic("ext2fs_reload: dirty2");
569 		/*
570 		 * Step 6: re-read inode data for all active vnodes.
571 		 */
572 		ip = VTOI(vp);
573 		error = bread(devvp, fsbtodb(fs, ino_to_fsba(fs, ip->i_number)),
574 		    (int)fs->e2fs_bsize, NOCRED, &bp);
575 		if (error) {
576 			vput(vp);
577 			mutex_enter(&mntvnode_lock);
578 			(void)vunmark(mvp);
579 			break;
580 		}
581 		cp = (char *)bp->b_data +
582 		    (ino_to_fsbo(fs, ip->i_number) * EXT2_DINODE_SIZE);
583 		e2fs_iload((struct ext2fs_dinode *)cp, ip->i_din.e2fs_din);
584 		brelse(bp, 0);
585 		vput(vp);
586 		mutex_enter(&mntvnode_lock);
587 	}
588 	mutex_exit(&mntvnode_lock);
589 	vnfree(mvp);
590 	return (error);
591 }
592 
593 /*
594  * Common code for mount and mountroot
595  */
596 int
597 ext2fs_mountfs(struct vnode *devvp, struct mount *mp)
598 {
599 	struct lwp *l = curlwp;
600 	struct ufsmount *ump;
601 	struct buf *bp;
602 	struct ext2fs *fs;
603 	struct m_ext2fs *m_fs;
604 	dev_t dev;
605 	struct partinfo dpart;
606 	int error, i, size, ronly;
607 	kauth_cred_t cred;
608 	struct proc *p;
609 
610 	dev = devvp->v_rdev;
611 	p = l ? l->l_proc : NULL;
612 	cred = l ? l->l_cred : NOCRED;
613 
614 	/* Flush out any old buffers remaining from a previous use. */
615 	vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
616 	error = vinvalbuf(devvp, V_SAVE, cred, l, 0, 0);
617 	VOP_UNLOCK(devvp, 0);
618 	if (error)
619 		return (error);
620 
621 	ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
622 	if (VOP_IOCTL(devvp, DIOCGPART, &dpart, FREAD, cred) != 0)
623 		size = DEV_BSIZE;
624 	else
625 		size = dpart.disklab->d_secsize;
626 
627 	bp = NULL;
628 	ump = NULL;
629 
630 #ifdef DEBUG_EXT2
631 	printf("sb size: %d ino size %d\n", sizeof(struct ext2fs),
632 	    EXT2_DINODE_SIZE);
633 #endif
634 	error = bread(devvp, (SBOFF / size), SBSIZE, cred, &bp);
635 	if (error)
636 		goto out;
637 	fs = (struct ext2fs *)bp->b_data;
638 	error = ext2fs_checksb(fs, ronly);
639 	if (error)
640 		goto out;
641 	ump = malloc(sizeof(*ump), M_UFSMNT, M_WAITOK);
642 	memset(ump, 0, sizeof(*ump));
643 	ump->um_fstype = UFS1;
644 	ump->um_ops = &ext2fs_ufsops;
645 	ump->um_e2fs = malloc(sizeof(struct m_ext2fs), M_UFSMNT, M_WAITOK);
646 	memset(ump->um_e2fs, 0, sizeof(struct m_ext2fs));
647 	e2fs_sbload((struct ext2fs *)bp->b_data, &ump->um_e2fs->e2fs);
648 	brelse(bp, 0);
649 	bp = NULL;
650 	m_fs = ump->um_e2fs;
651 	m_fs->e2fs_ronly = ronly;
652 	if (ronly == 0) {
653 		if (m_fs->e2fs.e2fs_state == E2FS_ISCLEAN)
654 			m_fs->e2fs.e2fs_state = 0;
655 		else
656 			m_fs->e2fs.e2fs_state = E2FS_ERRORS;
657 		m_fs->e2fs_fmod = 1;
658 	}
659 
660 	/* compute dynamic sb infos */
661 	m_fs->e2fs_ncg =
662 	    howmany(m_fs->e2fs.e2fs_bcount - m_fs->e2fs.e2fs_first_dblock,
663 	    m_fs->e2fs.e2fs_bpg);
664 	/* XXX assume hw bsize = 512 */
665 	m_fs->e2fs_fsbtodb = m_fs->e2fs.e2fs_log_bsize + 1;
666 	m_fs->e2fs_bsize = MINBSIZE << m_fs->e2fs.e2fs_log_bsize;
667 	m_fs->e2fs_bshift = LOG_MINBSIZE + m_fs->e2fs.e2fs_log_bsize;
668 	m_fs->e2fs_qbmask = m_fs->e2fs_bsize - 1;
669 	m_fs->e2fs_bmask = ~m_fs->e2fs_qbmask;
670 	m_fs->e2fs_ngdb =
671 	    howmany(m_fs->e2fs_ncg, m_fs->e2fs_bsize / sizeof(struct ext2_gd));
672 	m_fs->e2fs_ipb = m_fs->e2fs_bsize / EXT2_DINODE_SIZE;
673 	m_fs->e2fs_itpg = m_fs->e2fs.e2fs_ipg / m_fs->e2fs_ipb;
674 
675 	m_fs->e2fs_gd = malloc(m_fs->e2fs_ngdb * m_fs->e2fs_bsize,
676 	    M_UFSMNT, M_WAITOK);
677 	for (i = 0; i < m_fs->e2fs_ngdb; i++) {
678 		error = bread(devvp ,
679 		    fsbtodb(m_fs, m_fs->e2fs.e2fs_first_dblock +
680 		    1 /* superblock */ + i),
681 		    m_fs->e2fs_bsize, NOCRED, &bp);
682 		if (error) {
683 			free(m_fs->e2fs_gd, M_UFSMNT);
684 			goto out;
685 		}
686 		e2fs_cgload((struct ext2_gd *)bp->b_data,
687 		    &m_fs->e2fs_gd[
688 			i * m_fs->e2fs_bsize / sizeof(struct ext2_gd)],
689 		    m_fs->e2fs_bsize);
690 		brelse(bp, 0);
691 		bp = NULL;
692 	}
693 
694 	mp->mnt_data = ump;
695 	mp->mnt_stat.f_fsidx.__fsid_val[0] = (long)dev;
696 	mp->mnt_stat.f_fsidx.__fsid_val[1] = makefstype(MOUNT_EXT2FS);
697 	mp->mnt_stat.f_fsid = mp->mnt_stat.f_fsidx.__fsid_val[0];
698 	mp->mnt_stat.f_namemax = EXT2FS_MAXNAMLEN;
699 	mp->mnt_flag |= MNT_LOCAL;
700 	mp->mnt_dev_bshift = DEV_BSHIFT;	/* XXX */
701 	mp->mnt_fs_bshift = m_fs->e2fs_bshift;
702 	mp->mnt_iflag |= IMNT_DTYPE;
703 	ump->um_flags = 0;
704 	ump->um_mountp = mp;
705 	ump->um_dev = dev;
706 	ump->um_devvp = devvp;
707 	ump->um_nindir = NINDIR(m_fs);
708 	ump->um_lognindir = ffs(NINDIR(m_fs)) - 1;
709 	ump->um_bptrtodb = m_fs->e2fs_fsbtodb;
710 	ump->um_seqinc = 1; /* no frags */
711 	ump->um_maxsymlinklen = EXT2_MAXSYMLINKLEN;
712 	ump->um_dirblksiz = m_fs->e2fs_bsize;
713 	ump->um_maxfilesize = ((uint64_t)0x80000000 * m_fs->e2fs_bsize - 1);
714 	devvp->v_specmountpoint = mp;
715 	return (0);
716 
717 out:
718 	KASSERT(bp != NULL);
719 	brelse(bp, 0);
720 	if (ump) {
721 		free(ump->um_e2fs, M_UFSMNT);
722 		free(ump, M_UFSMNT);
723 		mp->mnt_data = NULL;
724 	}
725 	return (error);
726 }
727 
728 /*
729  * unmount system call
730  */
731 int
732 ext2fs_unmount(struct mount *mp, int mntflags)
733 {
734 	struct ufsmount *ump;
735 	struct m_ext2fs *fs;
736 	int error, flags;
737 
738 	flags = 0;
739 	if (mntflags & MNT_FORCE)
740 		flags |= FORCECLOSE;
741 	if ((error = ext2fs_flushfiles(mp, flags)) != 0)
742 		return (error);
743 	ump = VFSTOUFS(mp);
744 	fs = ump->um_e2fs;
745 	if (fs->e2fs_ronly == 0 &&
746 		ext2fs_cgupdate(ump, MNT_WAIT) == 0 &&
747 		(fs->e2fs.e2fs_state & E2FS_ERRORS) == 0) {
748 		fs->e2fs.e2fs_state = E2FS_ISCLEAN;
749 		(void) ext2fs_sbupdate(ump, MNT_WAIT);
750 	}
751 	if (ump->um_devvp->v_type != VBAD)
752 		ump->um_devvp->v_specmountpoint = NULL;
753 	vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
754 	error = VOP_CLOSE(ump->um_devvp, fs->e2fs_ronly ? FREAD : FREAD|FWRITE,
755 	    NOCRED);
756 	vput(ump->um_devvp);
757 	free(fs->e2fs_gd, M_UFSMNT);
758 	free(fs, M_UFSMNT);
759 	free(ump, M_UFSMNT);
760 	mp->mnt_data = NULL;
761 	mp->mnt_flag &= ~MNT_LOCAL;
762 	return (error);
763 }
764 
765 /*
766  * Flush out all the files in a filesystem.
767  */
768 int
769 ext2fs_flushfiles(struct mount *mp, int flags)
770 {
771 	extern int doforce;
772 	int error;
773 
774 	if (!doforce)
775 		flags &= ~FORCECLOSE;
776 	error = vflush(mp, NULLVP, flags);
777 	return (error);
778 }
779 
780 /*
781  * Get file system statistics.
782  */
783 int
784 ext2fs_statvfs(struct mount *mp, struct statvfs *sbp)
785 {
786 	struct ufsmount *ump;
787 	struct m_ext2fs *fs;
788 	uint32_t overhead, overhead_per_group, ngdb;
789 	int i, ngroups;
790 
791 	ump = VFSTOUFS(mp);
792 	fs = ump->um_e2fs;
793 	if (fs->e2fs.e2fs_magic != E2FS_MAGIC)
794 		panic("ext2fs_statvfs");
795 
796 	/*
797 	 * Compute the overhead (FS structures)
798 	 */
799 	overhead_per_group =
800 	    1 /* block bitmap */ +
801 	    1 /* inode bitmap */ +
802 	    fs->e2fs_itpg;
803 	overhead = fs->e2fs.e2fs_first_dblock +
804 	    fs->e2fs_ncg * overhead_per_group;
805 	if (fs->e2fs.e2fs_rev > E2FS_REV0 &&
806 	    fs->e2fs.e2fs_features_rocompat & EXT2F_ROCOMPAT_SPARSESUPER) {
807 		for (i = 0, ngroups = 0; i < fs->e2fs_ncg; i++) {
808 			if (cg_has_sb(i))
809 				ngroups++;
810 		}
811 	} else {
812 		ngroups = fs->e2fs_ncg;
813 	}
814 	ngdb = fs->e2fs_ngdb;
815 	if (fs->e2fs.e2fs_rev > E2FS_REV0 &&
816 	    fs->e2fs.e2fs_features_compat & EXT2F_COMPAT_RESIZE)
817 		ngdb += fs->e2fs.e2fs_reserved_ngdb;
818 	overhead += ngroups * (1 /* superblock */ + ngdb);
819 
820 	sbp->f_bsize = fs->e2fs_bsize;
821 	sbp->f_frsize = MINBSIZE << fs->e2fs.e2fs_fsize;
822 	sbp->f_iosize = fs->e2fs_bsize;
823 	sbp->f_blocks = fs->e2fs.e2fs_bcount - overhead;
824 	sbp->f_bfree = fs->e2fs.e2fs_fbcount;
825 	sbp->f_bresvd = fs->e2fs.e2fs_rbcount;
826 	if (sbp->f_bfree > sbp->f_bresvd)
827 		sbp->f_bavail = sbp->f_bfree - sbp->f_bresvd;
828 	else
829 		sbp->f_bavail = 0;
830 	sbp->f_files =  fs->e2fs.e2fs_icount;
831 	sbp->f_ffree = fs->e2fs.e2fs_ficount;
832 	sbp->f_favail = fs->e2fs.e2fs_ficount;
833 	sbp->f_fresvd = 0;
834 	copy_statvfs_info(sbp, mp);
835 	return (0);
836 }
837 
838 /*
839  * Go through the disk queues to initiate sandbagged IO;
840  * go through the inodes to write those that have been modified;
841  * initiate the writing of the super block if it has been modified.
842  *
843  * Note: we are always called with the filesystem marked `MPBUSY'.
844  */
845 int
846 ext2fs_sync(struct mount *mp, int waitfor, kauth_cred_t cred)
847 {
848 	struct vnode *vp, *mvp;
849 	struct inode *ip;
850 	struct ufsmount *ump = VFSTOUFS(mp);
851 	struct m_ext2fs *fs;
852 	int error, allerror = 0;
853 
854 	fs = ump->um_e2fs;
855 	if (fs->e2fs_fmod != 0 && fs->e2fs_ronly != 0) {	/* XXX */
856 		printf("fs = %s\n", fs->e2fs_fsmnt);
857 		panic("update: rofs mod");
858 	}
859 
860 	/* Allocate a marker vnode. */
861 	if ((mvp = vnalloc(mp)) == NULL)
862 		return (ENOMEM);
863 
864 	/*
865 	 * Write back each (modified) inode.
866 	 */
867 	mutex_enter(&mntvnode_lock);
868 loop:
869 	/*
870 	 * NOTE: not using the TAILQ_FOREACH here since in this loop vgone()
871 	 * and vclean() can be called indirectly
872 	 */
873 	for (vp = TAILQ_FIRST(&mp->mnt_vnodelist); vp; vp = vunmark(mvp)) {
874 		vmark(mvp, vp);
875 		if (vp->v_mount != mp || vismarker(vp))
876 			continue;
877 		mutex_enter(&vp->v_interlock);
878 		ip = VTOI(vp);
879 		if (ip == NULL || (vp->v_iflag & (VI_XLOCK|VI_CLEAN)) != 0 ||
880 		    vp->v_type == VNON ||
881 		    ((ip->i_flag &
882 		      (IN_CHANGE | IN_UPDATE | IN_MODIFIED)) == 0 &&
883 		     LIST_EMPTY(&vp->v_dirtyblkhd) &&
884 		     UVM_OBJ_IS_CLEAN(&vp->v_uobj)))
885 		{
886 			mutex_exit(&vp->v_interlock);
887 			continue;
888 		}
889 		mutex_exit(&mntvnode_lock);
890 		error = vget(vp, LK_EXCLUSIVE | LK_NOWAIT | LK_INTERLOCK);
891 		if (error) {
892 			mutex_enter(&mntvnode_lock);
893 			if (error == ENOENT) {
894 				mutex_enter(&mntvnode_lock);
895 				(void)vunmark(mvp);
896 				goto loop;
897 			}
898 			continue;
899 		}
900 		if (vp->v_type == VREG && waitfor == MNT_LAZY)
901 			error = ext2fs_update(vp, NULL, NULL, 0);
902 		else
903 			error = VOP_FSYNC(vp, cred,
904 			    waitfor == MNT_WAIT ? FSYNC_WAIT : 0, 0, 0);
905 		if (error)
906 			allerror = error;
907 		vput(vp);
908 		mutex_enter(&mntvnode_lock);
909 	}
910 	mutex_exit(&mntvnode_lock);
911 	vnfree(mvp);
912 	/*
913 	 * Force stale file system control information to be flushed.
914 	 */
915 	if (waitfor != MNT_LAZY) {
916 		vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
917 		if ((error = VOP_FSYNC(ump->um_devvp, cred,
918 		    waitfor == MNT_WAIT ? FSYNC_WAIT : 0, 0, 0)) != 0)
919 			allerror = error;
920 		VOP_UNLOCK(ump->um_devvp, 0);
921 	}
922 	/*
923 	 * Write back modified superblock.
924 	 */
925 	if (fs->e2fs_fmod != 0) {
926 		fs->e2fs_fmod = 0;
927 		fs->e2fs.e2fs_wtime = time_second;
928 		if ((error = ext2fs_cgupdate(ump, waitfor)))
929 			allerror = error;
930 	}
931 	return (allerror);
932 }
933 
934 /*
935  * Look up a EXT2FS dinode number to find its incore vnode, otherwise read it
936  * in from disk.  If it is in core, wait for the lock bit to clear, then
937  * return the inode locked.  Detection and handling of mount points must be
938  * done by the calling routine.
939  */
940 int
941 ext2fs_vget(struct mount *mp, ino_t ino, struct vnode **vpp)
942 {
943 	struct m_ext2fs *fs;
944 	struct inode *ip;
945 	struct ufsmount *ump;
946 	struct buf *bp;
947 	struct vnode *vp;
948 	dev_t dev;
949 	int error;
950 	void *cp;
951 
952 	ump = VFSTOUFS(mp);
953 	dev = ump->um_dev;
954 retry:
955 	if ((*vpp = ufs_ihashget(dev, ino, LK_EXCLUSIVE)) != NULL)
956 		return (0);
957 
958 	/* Allocate a new vnode/inode. */
959 	if ((error = getnewvnode(VT_EXT2FS, mp, ext2fs_vnodeop_p, &vp)) != 0) {
960 		*vpp = NULL;
961 		return (error);
962 	}
963 	ip = pool_get(&ext2fs_inode_pool, PR_WAITOK);
964 
965 	mutex_enter(&ufs_hashlock);
966 	if ((*vpp = ufs_ihashget(dev, ino, 0)) != NULL) {
967 		mutex_exit(&ufs_hashlock);
968 		ungetnewvnode(vp);
969 		pool_put(&ext2fs_inode_pool, ip);
970 		goto retry;
971 	}
972 
973 	vp->v_vflag |= VV_LOCKSWORK;
974 
975 	memset(ip, 0, sizeof(struct inode));
976 	vp->v_data = ip;
977 	ip->i_vnode = vp;
978 	ip->i_ump = ump;
979 	ip->i_e2fs = fs = ump->um_e2fs;
980 	ip->i_dev = dev;
981 	ip->i_number = ino;
982 	ip->i_e2fs_last_lblk = 0;
983 	ip->i_e2fs_last_blk = 0;
984 	genfs_node_init(vp, &ext2fs_genfsops);
985 
986 	/*
987 	 * Put it onto its hash chain and lock it so that other requests for
988 	 * this inode will block if they arrive while we are sleeping waiting
989 	 * for old data structures to be purged or for the contents of the
990 	 * disk portion of this inode to be read.
991 	 */
992 
993 	ufs_ihashins(ip);
994 	mutex_exit(&ufs_hashlock);
995 
996 	/* Read in the disk contents for the inode, copy into the inode. */
997 	error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)),
998 	    (int)fs->e2fs_bsize, NOCRED, &bp);
999 	if (error) {
1000 
1001 		/*
1002 		 * The inode does not contain anything useful, so it would
1003 		 * be misleading to leave it on its hash chain. With mode
1004 		 * still zero, it will be unlinked and returned to the free
1005 		 * list by vput().
1006 		 */
1007 
1008 		vput(vp);
1009 		brelse(bp, 0);
1010 		*vpp = NULL;
1011 		return (error);
1012 	}
1013 	cp = (char *)bp->b_data + (ino_to_fsbo(fs, ino) * EXT2_DINODE_SIZE);
1014 	ip->i_din.e2fs_din = pool_get(&ext2fs_dinode_pool, PR_WAITOK);
1015 	e2fs_iload((struct ext2fs_dinode *)cp, ip->i_din.e2fs_din);
1016 	brelse(bp, 0);
1017 
1018 	/* If the inode was deleted, reset all fields */
1019 	if (ip->i_e2fs_dtime != 0) {
1020 		ip->i_e2fs_mode = ip->i_e2fs_nblock = 0;
1021 		(void)ext2fs_setsize(ip, 0);
1022 		memset(ip->i_e2fs_blocks, 0, sizeof(ip->i_e2fs_blocks));
1023 	}
1024 
1025 	/*
1026 	 * Initialize the vnode from the inode, check for aliases.
1027 	 * Note that the underlying vnode may have changed.
1028 	 */
1029 
1030 	error = ext2fs_vinit(mp, ext2fs_specop_p, ext2fs_fifoop_p, &vp);
1031 	if (error) {
1032 		vput(vp);
1033 		*vpp = NULL;
1034 		return (error);
1035 	}
1036 	/*
1037 	 * Finish inode initialization now that aliasing has been resolved.
1038 	 */
1039 
1040 	ip->i_devvp = ump->um_devvp;
1041 	VREF(ip->i_devvp);
1042 
1043 	/*
1044 	 * Set up a generation number for this inode if it does not
1045 	 * already have one. This should only happen on old filesystems.
1046 	 */
1047 
1048 	if (ip->i_e2fs_gen == 0) {
1049 		if (++ext2gennumber < (u_long)time_second)
1050 			ext2gennumber = time_second;
1051 		ip->i_e2fs_gen = ext2gennumber;
1052 		if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0)
1053 			ip->i_flag |= IN_MODIFIED;
1054 	}
1055 	uvm_vnp_setsize(vp, ext2fs_size(ip));
1056 	*vpp = vp;
1057 	return (0);
1058 }
1059 
1060 /*
1061  * File handle to vnode
1062  *
1063  * Have to be really careful about stale file handles:
1064  * - check that the inode number is valid
1065  * - call ext2fs_vget() to get the locked inode
1066  * - check for an unallocated inode (i_mode == 0)
1067  */
1068 int
1069 ext2fs_fhtovp(struct mount *mp, struct fid *fhp, struct vnode **vpp)
1070 {
1071 	struct inode *ip;
1072 	struct vnode *nvp;
1073 	int error;
1074 	struct ufid ufh;
1075 	struct m_ext2fs *fs;
1076 
1077 	if (fhp->fid_len != sizeof(struct ufid))
1078 		return EINVAL;
1079 
1080 	memcpy(&ufh, fhp, sizeof(struct ufid));
1081 	fs = VFSTOUFS(mp)->um_e2fs;
1082 	if ((ufh.ufid_ino < EXT2_FIRSTINO && ufh.ufid_ino != EXT2_ROOTINO) ||
1083 		ufh.ufid_ino >= fs->e2fs_ncg * fs->e2fs.e2fs_ipg)
1084 		return (ESTALE);
1085 
1086 	if ((error = VFS_VGET(mp, ufh.ufid_ino, &nvp)) != 0) {
1087 		*vpp = NULLVP;
1088 		return (error);
1089 	}
1090 	ip = VTOI(nvp);
1091 	if (ip->i_e2fs_mode == 0 || ip->i_e2fs_dtime != 0 ||
1092 		ip->i_e2fs_gen != ufh.ufid_gen) {
1093 		vput(nvp);
1094 		*vpp = NULLVP;
1095 		return (ESTALE);
1096 	}
1097 	*vpp = nvp;
1098 	return (0);
1099 }
1100 
1101 /*
1102  * Vnode pointer to File handle
1103  */
1104 /* ARGSUSED */
1105 int
1106 ext2fs_vptofh(struct vnode *vp, struct fid *fhp, size_t *fh_size)
1107 {
1108 	struct inode *ip;
1109 	struct ufid ufh;
1110 
1111 	if (*fh_size < sizeof(struct ufid)) {
1112 		*fh_size = sizeof(struct ufid);
1113 		return E2BIG;
1114 	}
1115 	*fh_size = sizeof(struct ufid);
1116 
1117 	ip = VTOI(vp);
1118 	memset(&ufh, 0, sizeof(ufh));
1119 	ufh.ufid_len = sizeof(struct ufid);
1120 	ufh.ufid_ino = ip->i_number;
1121 	ufh.ufid_gen = ip->i_e2fs_gen;
1122 	memcpy(fhp, &ufh, sizeof(ufh));
1123 	return (0);
1124 }
1125 
1126 SYSCTL_SETUP(sysctl_vfs_ext2fs_setup, "sysctl vfs.ext2fs subtree setup")
1127 {
1128 
1129 	sysctl_createv(clog, 0, NULL, NULL,
1130 		       CTLFLAG_PERMANENT,
1131 		       CTLTYPE_NODE, "vfs", NULL,
1132 		       NULL, 0, NULL, 0,
1133 		       CTL_VFS, CTL_EOL);
1134 	sysctl_createv(clog, 0, NULL, NULL,
1135 		       CTLFLAG_PERMANENT,
1136 		       CTLTYPE_NODE, "ext2fs",
1137 		       SYSCTL_DESCR("Linux EXT2FS file system"),
1138 		       NULL, 0, NULL, 0,
1139 		       CTL_VFS, 17, CTL_EOL);
1140 	/*
1141 	 * XXX the "17" above could be dynamic, thereby eliminating
1142 	 * one more instance of the "number to vfs" mapping problem,
1143 	 * but "17" is the order as taken from sys/mount.h
1144 	 */
1145 }
1146 
1147 /*
1148  * Write a superblock and associated information back to disk.
1149  */
1150 int
1151 ext2fs_sbupdate(struct ufsmount *mp, int waitfor)
1152 {
1153 	struct m_ext2fs *fs = mp->um_e2fs;
1154 	struct buf *bp;
1155 	int error = 0;
1156 
1157 	bp = getblk(mp->um_devvp, SBLOCK, SBSIZE, 0, 0);
1158 	e2fs_sbsave(&fs->e2fs, (struct ext2fs*)bp->b_data);
1159 	if (waitfor == MNT_WAIT)
1160 		error = bwrite(bp);
1161 	else
1162 		bawrite(bp);
1163 	return (error);
1164 }
1165 
1166 int
1167 ext2fs_cgupdate(struct ufsmount *mp, int waitfor)
1168 {
1169 	struct m_ext2fs *fs = mp->um_e2fs;
1170 	struct buf *bp;
1171 	int i, error = 0, allerror = 0;
1172 
1173 	allerror = ext2fs_sbupdate(mp, waitfor);
1174 	for (i = 0; i < fs->e2fs_ngdb; i++) {
1175 		bp = getblk(mp->um_devvp, fsbtodb(fs,
1176 		    fs->e2fs.e2fs_first_dblock +
1177 		    1 /* superblock */ + i), fs->e2fs_bsize, 0, 0);
1178 		e2fs_cgsave(&fs->e2fs_gd[
1179 		    i * fs->e2fs_bsize / sizeof(struct ext2_gd)],
1180 		    (struct ext2_gd *)bp->b_data, fs->e2fs_bsize);
1181 		if (waitfor == MNT_WAIT)
1182 			error = bwrite(bp);
1183 		else
1184 			bawrite(bp);
1185 	}
1186 
1187 	if (!allerror && error)
1188 		allerror = error;
1189 	return (allerror);
1190 }
1191 
1192 static int
1193 ext2fs_checksb(struct ext2fs *fs, int ronly)
1194 {
1195 
1196 	if (fs2h16(fs->e2fs_magic) != E2FS_MAGIC) {
1197 		return (EINVAL);		/* XXX needs translation */
1198 	}
1199 	if (fs2h32(fs->e2fs_rev) > E2FS_REV1) {
1200 #ifdef DIAGNOSTIC
1201 		printf("Ext2 fs: unsupported revision number: %x\n",
1202 		    fs2h32(fs->e2fs_rev));
1203 #endif
1204 		return (EINVAL);		/* XXX needs translation */
1205 	}
1206 	if (fs2h32(fs->e2fs_log_bsize) > 2) { /* block size = 1024|2048|4096 */
1207 #ifdef DIAGNOSTIC
1208 		printf("Ext2 fs: bad block size: %d "
1209 		    "(expected <= 2 for ext2 fs)\n",
1210 		    fs2h32(fs->e2fs_log_bsize));
1211 #endif
1212 		return (EINVAL);	   /* XXX needs translation */
1213 	}
1214 	if (fs2h32(fs->e2fs_rev) > E2FS_REV0) {
1215 		if (fs2h32(fs->e2fs_first_ino) != EXT2_FIRSTINO ||
1216 		    fs2h16(fs->e2fs_inode_size) != EXT2_DINODE_SIZE) {
1217 			printf("Ext2 fs: unsupported inode size\n");
1218 			return (EINVAL);      /* XXX needs translation */
1219 		}
1220 		if (fs2h32(fs->e2fs_features_incompat) &
1221 		    ~EXT2F_INCOMPAT_SUPP) {
1222 			printf("Ext2 fs: unsupported optional feature\n");
1223 			return (EINVAL);      /* XXX needs translation */
1224 		}
1225 		if (!ronly && fs2h32(fs->e2fs_features_rocompat) &
1226 		    ~EXT2F_ROCOMPAT_SUPP) {
1227 			return (EROFS);      /* XXX needs translation */
1228 		}
1229 	}
1230 	return (0);
1231 }
1232