xref: /netbsd-src/sys/ufs/ext2fs/ext2fs_vfsops.c (revision 059d80f11ac7e2170fc4c60afd430cc3f0fc061d)
1 /*	$NetBSD: ext2fs_vfsops.c,v 1.228 2024/12/30 09:01:35 hannken 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 1.228 2024/12/30 09:01:35 hannken 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 static void
163 e2fs_cgload(const char *ondisk, struct ext2_gd *inmemory, int cg_size,
164     int shift_cg_entry_size)
165 {
166 
167 	if (shift_cg_entry_size == 6) {
168 		memcpy(inmemory, ondisk, cg_size);
169 		return;
170 	}
171 
172 	const char *iptr = ondisk;
173 	struct ext2_gd *optr = inmemory;
174 	int sh = 1 << shift_cg_entry_size;
175 	int lim = cg_size >> shift_cg_entry_size;
176 	if (shift_cg_entry_size > 6) {
177 		for (int i = 0; i < lim; i++, optr++, iptr += sh) {
178 			memcpy(optr, iptr, sizeof(*optr));
179 		}
180 	} else {
181 		for (int i = 0; i < lim; i++, optr++, iptr += sh) {
182 			memcpy(optr, iptr, E2FS_REV0_GD_SIZE);
183 			memset((char *)optr + E2FS_REV0_GD_SIZE, 0,
184 			    sizeof(*optr) - E2FS_REV0_GD_SIZE);
185 		}
186 	}
187 }
188 
189 static void
190 e2fs_cgsave(const struct ext2_gd *inmemory, char *ondisk, int cg_size,
191     int shift_cg_entry_size)
192 {
193 
194 	if (shift_cg_entry_size == 6) {
195 		memcpy(ondisk, inmemory, cg_size);
196 		return;
197 	}
198 
199 	const struct ext2_gd *iptr = inmemory;
200 	char *optr = ondisk;
201 	int sh = 1 << shift_cg_entry_size;
202 	int lim = cg_size >> shift_cg_entry_size;
203 	if (shift_cg_entry_size > 6) {
204 		for (int i = 0; i < lim; i++, iptr++, optr += sh) {
205 			memcpy(optr, iptr, sizeof(*iptr));
206 			memset(optr + sizeof(*iptr), 0, sh - sizeof(*iptr));
207 		}
208 	} else {
209 		for (int i = 0; i < lim; i++, iptr++, optr += sh) {
210 			memcpy(optr, iptr, E2FS_REV0_GD_SIZE);
211 		}
212 	}
213 }
214 
215 /* Fill in the inode uid/gid from ext2 halves.  */
216 void
217 ext2fs_set_inode_guid(struct inode *ip)
218 {
219 
220 	ip->i_gid = ip->i_e2fs_gid;
221 	ip->i_uid = ip->i_e2fs_uid;
222 	if (ip->i_e2fs->e2fs.e2fs_rev > E2FS_REV0) {
223 		ip->i_gid |= ip->i_e2fs_gid_high << 16;
224 		ip->i_uid |= ip->i_e2fs_uid_high << 16;
225 	}
226 }
227 
228 SYSCTL_SETUP(ext2fs_sysctl_setup, "ext2fs sysctl")
229 {
230 
231 		sysctl_createv(clog, 0, NULL, NULL,
232 			       CTLFLAG_PERMANENT,
233 			       CTLTYPE_NODE, "ext2fs",
234 			       SYSCTL_DESCR("Linux EXT2FS file system"),
235 			       NULL, 0, NULL, 0,
236 			       CTL_VFS, 17, CTL_EOL);
237 		/*
238 		 * XXX the "17" above could be dynamic, thereby eliminating
239 		 * one more instance of the "number to vfs" mapping problem,
240 		 * but "17" is the order as taken from sys/mount.h
241 		 */
242 }
243 
244 static int
245 ext2fs_modcmd(modcmd_t cmd, void *arg)
246 {
247 	int error;
248 
249 	switch (cmd) {
250 	case MODULE_CMD_INIT:
251 		error = vfs_attach(&ext2fs_vfsops);
252 		if (error != 0)
253 			break;
254 		break;
255 	case MODULE_CMD_FINI:
256 		error = vfs_detach(&ext2fs_vfsops);
257 		if (error != 0)
258 			break;
259 		break;
260 	default:
261 		error = ENOTTY;
262 		break;
263 	}
264 
265 	return error;
266 }
267 
268 /*
269  * XXX Same structure as FFS inodes?  Should we share a common pool?
270  */
271 struct pool ext2fs_inode_pool;
272 
273 extern u_long ext2gennumber;
274 
275 void
276 ext2fs_init(void)
277 {
278 
279 	pool_init(&ext2fs_inode_pool, sizeof(struct inode), 0, 0, 0,
280 	    "ext2fsinopl", &pool_allocator_nointr, IPL_NONE);
281 	ufs_init();
282 }
283 
284 void
285 ext2fs_reinit(void)
286 {
287 	ufs_reinit();
288 }
289 
290 void
291 ext2fs_done(void)
292 {
293 
294 	ufs_done();
295 	pool_destroy(&ext2fs_inode_pool);
296 }
297 
298 static void
299 ext2fs_sb_setmountinfo(struct m_ext2fs *fs, struct mount *mp)
300 {
301 	(void)strlcpy(fs->e2fs_fsmnt, mp->mnt_stat.f_mntonname,
302             sizeof(fs->e2fs_fsmnt));
303 	if (fs->e2fs_ronly == 0 && fs->e2fs.e2fs_rev > E2FS_REV0) {
304 		(void)strlcpy(fs->e2fs.e2fs_fsmnt, mp->mnt_stat.f_mntonname,
305 		    sizeof(fs->e2fs.e2fs_fsmnt));
306 
307 		fs->e2fs.e2fs_mtime = time_second;
308 		fs->e2fs.e2fs_mnt_count++;
309 
310 		fs->e2fs_fmod = 1;
311 	}
312 }
313 
314 /*
315  * Called by main() when ext2fs is going to be mounted as root.
316  *
317  * Name is updated by mount(8) after booting.
318  */
319 
320 int
321 ext2fs_mountroot(void)
322 {
323 	extern struct vnode *rootvp;
324 	struct m_ext2fs *fs;
325 	struct mount *mp;
326 	struct ufsmount *ump;
327 	int error;
328 
329 	if (device_class(root_device) != DV_DISK)
330 		return ENODEV;
331 
332 	if ((error = vfs_rootmountalloc(MOUNT_EXT2FS, "root_device", &mp))) {
333 		vrele(rootvp);
334 		return error;
335 	}
336 
337 	if ((error = ext2fs_mountfs(rootvp, mp)) != 0) {
338 		vfs_unbusy(mp);
339 		vfs_rele(mp);
340 		return error;
341 	}
342 	mountlist_append(mp);
343 	ump = VFSTOUFS(mp);
344 	fs = ump->um_e2fs;
345 	ext2fs_sb_setmountinfo(fs, mp);
346 	(void)ext2fs_statvfs(mp, &mp->mnt_stat);
347 	vfs_unbusy(mp);
348 	setrootfstime((time_t)fs->e2fs.e2fs_wtime);
349 	return 0;
350 }
351 
352 /*
353  * VFS Operations.
354  *
355  * mount system call
356  */
357 int
358 ext2fs_mount(struct mount *mp, const char *path, void *data, size_t *data_len)
359 {
360 	struct lwp *l = curlwp;
361 	struct vnode *devvp;
362 	struct ufs_args *args = data;
363 	struct ufsmount *ump = NULL;
364 	struct m_ext2fs *fs;
365 	int error = 0, flags, update;
366 	mode_t accessmode;
367 
368 	if (args == NULL)
369 		return EINVAL;
370 	if (*data_len < sizeof *args)
371 		return EINVAL;
372 
373 	if (mp->mnt_flag & MNT_GETARGS) {
374 		ump = VFSTOUFS(mp);
375 		if (ump == NULL)
376 			return EIO;
377 		memset(args, 0, sizeof *args);
378 		args->fspec = NULL;
379 		*data_len = sizeof *args;
380 		return 0;
381 	}
382 
383 	update = mp->mnt_flag & MNT_UPDATE;
384 
385 	/* Check arguments */
386 	if (args->fspec != NULL) {
387 		/*
388 		 * Look up the name and verify that it's sane.
389 		 */
390 		error = namei_simple_user(args->fspec,
391 					NSM_FOLLOW_NOEMULROOT, &devvp);
392 		if (error != 0)
393 			return error;
394 
395 		if (!update) {
396 			/*
397 			 * Be sure this is a valid block device
398 			 */
399 			if (devvp->v_type != VBLK)
400 				error = ENOTBLK;
401 			else if (bdevsw_lookup(devvp->v_rdev) == NULL)
402 				error = ENXIO;
403 		} else {
404 		        /*
405 			 * Be sure we're still naming the same device
406 			 * used for our initial mount
407 			 */
408 			ump = VFSTOUFS(mp);
409 			if (devvp != ump->um_devvp) {
410 				if (devvp->v_rdev != ump->um_devvp->v_rdev)
411 					error = EINVAL;
412 				else {
413 					vrele(devvp);
414 					devvp = ump->um_devvp;
415 					vref(devvp);
416 				}
417 			}
418 		}
419 	} else {
420 		if (!update) {
421 			/* New mounts must have a filename for the device */
422 			return EINVAL;
423 		} else {
424 			ump = VFSTOUFS(mp);
425 			devvp = ump->um_devvp;
426 			vref(devvp);
427 		}
428 	}
429 
430 	/*
431 	 * If mount by non-root, then verify that user has necessary
432 	 * permissions on the device.
433 	 *
434 	 * Permission to update a mount is checked higher, so here we presume
435 	 * updating the mount is okay (for example, as far as securelevel goes)
436 	 * which leaves us with the normal check.
437 	 */
438 	if (error == 0) {
439 		accessmode = VREAD;
440 		if (update ?
441 		    (mp->mnt_iflag & IMNT_WANTRDWR) != 0 :
442 		    (mp->mnt_flag & MNT_RDONLY) == 0)
443 			accessmode |= VWRITE;
444 		vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
445 		error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT,
446 		    KAUTH_REQ_SYSTEM_MOUNT_DEVICE, mp, devvp,
447 		    KAUTH_ARG(accessmode));
448 		VOP_UNLOCK(devvp);
449 	}
450 
451 	if (error) {
452 		vrele(devvp);
453 		return error;
454 	}
455 
456 	if (!update) {
457 		int xflags;
458 
459 		if (mp->mnt_flag & MNT_RDONLY)
460 			xflags = FREAD;
461 		else
462 			xflags = FREAD|FWRITE;
463 		vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
464 		error = VOP_OPEN(devvp, xflags, FSCRED);
465 		VOP_UNLOCK(devvp);
466 		if (error)
467 			goto fail;
468 		error = ext2fs_mountfs(devvp, mp);
469 		if (error) {
470 			vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
471 			(void)VOP_CLOSE(devvp, xflags, NOCRED);
472 			VOP_UNLOCK(devvp);
473 			goto fail;
474 		}
475 
476 		ump = VFSTOUFS(mp);
477 		fs = ump->um_e2fs;
478 	} else {
479 		/*
480 		 * Update the mount.
481 		 */
482 
483 		/*
484 		 * The initial mount got a reference on this
485 		 * device, so drop the one obtained via
486 		 * namei(), above.
487 		 */
488 		vrele(devvp);
489 
490 		ump = VFSTOUFS(mp);
491 		fs = ump->um_e2fs;
492 		if (fs->e2fs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) {
493 			/*
494 			 * Changing from r/w to r/o
495 			 */
496 			flags = WRITECLOSE;
497 			if (mp->mnt_flag & MNT_FORCE)
498 				flags |= FORCECLOSE;
499 			error = ext2fs_flushfiles(mp, flags);
500 			if (error == 0 &&
501 			    ext2fs_cgupdate(ump, MNT_WAIT) == 0 &&
502 			    (fs->e2fs.e2fs_state & E2FS_ERRORS) == 0) {
503 				fs->e2fs.e2fs_state = E2FS_ISCLEAN;
504 				(void) ext2fs_sbupdate(ump, MNT_WAIT);
505 			}
506 			if (error)
507 				return error;
508 			fs->e2fs_ronly = 1;
509 		}
510 
511 		if (mp->mnt_flag & MNT_RELOAD) {
512 			error = ext2fs_reload(mp, l->l_cred, l);
513 			if (error)
514 				return error;
515 		}
516 
517 		if (fs->e2fs_ronly && (mp->mnt_iflag & IMNT_WANTRDWR)) {
518 			/*
519 			 * Changing from read-only to read/write
520 			 */
521 			fs->e2fs_ronly = 0;
522 			if (fs->e2fs.e2fs_state == E2FS_ISCLEAN)
523 				fs->e2fs.e2fs_state = 0;
524 			else
525 				fs->e2fs.e2fs_state = E2FS_ERRORS;
526 			fs->e2fs_fmod = 1;
527 		}
528 		if (args->fspec == NULL)
529 			return 0;
530 	}
531 
532 	error = set_statvfs_info(path, UIO_USERSPACE, args->fspec,
533 	    UIO_USERSPACE, mp->mnt_op->vfs_name, mp, l);
534 	if (error == 0)
535 		ext2fs_sb_setmountinfo(fs, mp);
536 
537 	if (fs->e2fs_fmod != 0) {	/* XXX */
538 		fs->e2fs_fmod = 0;
539 		if (fs->e2fs.e2fs_state == 0)
540 			fs->e2fs.e2fs_wtime = time_second;
541 		else
542 			printf("%s: file system not clean; please fsck(8)\n",
543 				mp->mnt_stat.f_mntfromname);
544 		(void) ext2fs_cgupdate(ump, MNT_WAIT);
545 	}
546 	return error;
547 
548 fail:
549 	vrele(devvp);
550 	return error;
551 }
552 
553 /*
554  * Sanity check the disk vnode content, and copy it over to inode structure.
555  */
556 static int
557 ext2fs_loadvnode_content(struct m_ext2fs *fs, ino_t ino, struct buf *bp, struct inode *ip)
558 {
559 	struct ext2fs_dinode *din;
560 	int error = 0;
561 
562 	din = (struct ext2fs_dinode *)((char *)bp->b_data +
563 	    (ino_to_fsbo(fs, ino) * EXT2_DINODE_SIZE(fs)));
564 
565 	/* sanity checks - inode data NOT byteswapped at this point */
566 	if (EXT2_DINODE_FITS(din, e2di_extra_isize, EXT2_DINODE_SIZE(fs))
567 	    && (EXT2_DINODE_SIZE(fs) - EXT2_REV0_DINODE_SIZE)
568 	    < fs2h16(din->e2di_extra_isize))
569 	{
570 		printf("ext2fs: inode %"PRIu64" bad extra_isize %u",
571 			ino, din->e2di_extra_isize);
572 		error = EINVAL;
573 		goto bad;
574 	}
575 
576 	/* everything alright, proceed with copy */
577 	if (ip->i_din.e2fs_din == NULL)
578 		ip->i_din.e2fs_din = kmem_alloc(EXT2_DINODE_SIZE(fs), KM_SLEEP);
579 
580 	e2fs_iload(din, ip->i_din.e2fs_din, EXT2_DINODE_SIZE(fs));
581 
582 	ext2fs_set_inode_guid(ip);
583 
584     bad:
585 	return error;
586 }
587 
588 /*
589  * Reload all incore data for a filesystem (used after running fsck on
590  * the root filesystem and finding things to fix). The filesystem must
591  * be mounted read-only.
592  *
593  * Things to do to update the mount:
594  *	1) invalidate all cached meta-data.
595  *	2) re-read superblock from disk.
596  *	3) re-read summary information from disk.
597  *	4) invalidate all inactive vnodes.
598  *	5) invalidate all cached file data.
599  *	6) re-read inode data for all active vnodes.
600  */
601 int
602 ext2fs_reload(struct mount *mp, kauth_cred_t cred, struct lwp *l)
603 {
604 	struct vnode *vp, *devvp;
605 	struct inode *ip;
606 	struct buf *bp;
607 	struct m_ext2fs *fs;
608 	struct ext2fs *newfs;
609 	int i, error;
610 	struct ufsmount *ump;
611 	struct vnode_iterator *marker;
612 
613 	if ((mp->mnt_flag & MNT_RDONLY) == 0)
614 		return EINVAL;
615 
616 	ump = VFSTOUFS(mp);
617 	/*
618 	 * Step 1: invalidate all cached meta-data.
619 	 */
620 	devvp = ump->um_devvp;
621 	vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
622 	error = vinvalbuf(devvp, 0, cred, l, 0, 0);
623 	VOP_UNLOCK(devvp);
624 	if (error)
625 		panic("ext2fs_reload: dirty1");
626 
627 	fs = ump->um_e2fs;
628 	/*
629 	 * Step 2: re-read superblock from disk. Copy in new superblock, and
630 	 * compute in-memory values.
631 	 */
632 	error = bread(devvp, SBLOCK, SBSIZE, 0, &bp);
633 	if (error)
634 		return error;
635 	newfs = (struct ext2fs *)bp->b_data;
636 	e2fs_sbload(newfs, &fs->e2fs);
637 
638 	brelse(bp, 0);
639 
640 	error = ext2fs_sbfill(fs, (mp->mnt_flag & MNT_RDONLY) != 0);
641 	if (error)
642 		return error;
643 
644 	/*
645 	 * Step 3: re-read summary information from disk.
646 	 */
647 	for (i = 0; i < fs->e2fs_ngdb; i++) {
648 		error = bread(devvp ,
649 		    EXT2_FSBTODB(fs, fs->e2fs.e2fs_first_dblock +
650 		    1 /* superblock */ + i),
651 		    fs->e2fs_bsize, 0, &bp);
652 		if (error) {
653 			return error;
654 		}
655 		e2fs_cgload(bp->b_data,
656 		    &fs->e2fs_gd[i *
657 			(fs->e2fs_bsize >> fs->e2fs_group_desc_shift)],
658 		    fs->e2fs_bsize, fs->e2fs_group_desc_shift);
659 		brelse(bp, 0);
660 	}
661 
662 	vfs_vnode_iterator_init(mp, &marker);
663 	while ((vp = vfs_vnode_iterator_next(marker, NULL, NULL))) {
664 		/*
665 		 * Step 4: invalidate all inactive vnodes.
666 		 */
667 		if (vrecycle(vp))
668 			continue;
669 		/*
670 		 * Step 5: invalidate all cached file data.
671 		 */
672 		if (vn_lock(vp, LK_EXCLUSIVE)) {
673 			vrele(vp);
674 			continue;
675 		}
676 		if (vinvalbuf(vp, 0, cred, l, 0, 0))
677 			panic("ext2fs_reload: dirty2");
678 		/*
679 		 * Step 6: re-read inode data for all active vnodes.
680 		 */
681 		ip = VTOI(vp);
682 		error = bread(devvp, EXT2_FSBTODB(fs, ino_to_fsba(fs, ip->i_number)),
683 		    (int)fs->e2fs_bsize, 0, &bp);
684 		if (error) {
685 			vput(vp);
686 			break;
687 		}
688 		error = ext2fs_loadvnode_content(fs, ip->i_number, bp, ip);
689 		brelse(bp, 0);
690 		if (error) {
691 			vput(vp);
692 			break;
693 		}
694 
695 		vput(vp);
696 	}
697 	vfs_vnode_iterator_destroy(marker);
698 	return error;
699 }
700 
701 /*
702  * Common code for mount and mountroot
703  */
704 int
705 ext2fs_mountfs(struct vnode *devvp, struct mount *mp)
706 {
707 	struct lwp *l = curlwp;
708 	struct ufsmount *ump;
709 	struct buf *bp;
710 	struct ext2fs *fs;
711 	struct m_ext2fs *m_fs;
712 	dev_t dev;
713 	int error, i, ronly;
714 	kauth_cred_t cred;
715 
716 	dev = devvp->v_rdev;
717 	cred = l->l_cred;
718 
719 	/* Flush out any old buffers remaining from a previous use. */
720 	vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
721 	error = vinvalbuf(devvp, V_SAVE, cred, l, 0, 0);
722 	VOP_UNLOCK(devvp);
723 	if (error)
724 		return error;
725 
726 	ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
727 
728 	bp = NULL;
729 	ump = NULL;
730 
731 	/* Read the superblock from disk, and swap it directly. */
732 	error = bread(devvp, SBLOCK, SBSIZE, 0, &bp);
733 	if (error)
734 		goto out;
735 	fs = (struct ext2fs *)bp->b_data;
736 	m_fs = kmem_zalloc(sizeof(*m_fs), KM_SLEEP);
737 	e2fs_sbload(fs, &m_fs->e2fs);
738 
739 	brelse(bp, 0);
740 	bp = NULL;
741 
742 	/* Once swapped, validate and fill in the superblock. */
743 	error = ext2fs_sbfill(m_fs, ronly);
744 	if (error) {
745 		kmem_free(m_fs, sizeof(*m_fs));
746 		goto out;
747 	}
748 	m_fs->e2fs_ronly = ronly;
749 
750 	ump = kmem_zalloc(sizeof(*ump), KM_SLEEP);
751 	ump->um_fstype = UFS1;
752 	ump->um_ops = &ext2fs_ufsops;
753 	ump->um_e2fs = m_fs;
754 
755 	if (ronly == 0) {
756 		if (m_fs->e2fs.e2fs_state == E2FS_ISCLEAN)
757 			m_fs->e2fs.e2fs_state = 0;
758 		else
759 			m_fs->e2fs.e2fs_state = E2FS_ERRORS;
760 		m_fs->e2fs_fmod = 1;
761 	}
762 
763 	int32_t sh = m_fs->e2fs_bsize >> m_fs->e2fs_group_desc_shift;
764 	/* XXX: should be added in ext2fs_sbfill()? */
765 	m_fs->e2fs_gd = kmem_alloc(m_fs->e2fs_ngdb * sh
766 	    * sizeof(struct ext2_gd), KM_SLEEP);
767 	for (i = 0; i < m_fs->e2fs_ngdb; i++) {
768 		error = bread(devvp,
769 		    EXT2_FSBTODB(m_fs, m_fs->e2fs.e2fs_first_dblock +
770 		    1 /* superblock */ + i),
771 		    m_fs->e2fs_bsize, 0, &bp);
772 		if (error)
773 			goto out1;
774 		e2fs_cgload(bp->b_data, &m_fs->e2fs_gd[i *
775 			(m_fs->e2fs_bsize >> m_fs->e2fs_group_desc_shift)],
776 		    m_fs->e2fs_bsize, m_fs->e2fs_group_desc_shift);
777 		brelse(bp, 0);
778 		bp = NULL;
779 	}
780 
781 	error = ext2fs_cg_verify_and_initialize(devvp, m_fs, ronly);
782 	if (error)
783 		goto out1;
784 
785 	mp->mnt_data = ump;
786 	mp->mnt_stat.f_fsidx.__fsid_val[0] = (long)dev;
787 	mp->mnt_stat.f_fsidx.__fsid_val[1] = makefstype(MOUNT_EXT2FS);
788 	mp->mnt_stat.f_fsid = mp->mnt_stat.f_fsidx.__fsid_val[0];
789 	mp->mnt_stat.f_namemax = EXT2FS_MAXNAMLEN;
790 	mp->mnt_flag |= MNT_LOCAL;
791 	mp->mnt_dev_bshift = DEV_BSHIFT;	/* XXX */
792 	mp->mnt_fs_bshift = m_fs->e2fs_bshift;
793 	mp->mnt_iflag |= IMNT_DTYPE | IMNT_SHRLOOKUP;
794 	ump->um_flags = 0;
795 	ump->um_mountp = mp;
796 	ump->um_dev = dev;
797 	ump->um_devvp = devvp;
798 	ump->um_nindir = EXT2_NINDIR(m_fs);
799 	ump->um_lognindir = ffs(EXT2_NINDIR(m_fs)) - 1;
800 	ump->um_bptrtodb = m_fs->e2fs_fsbtodb;
801 	ump->um_seqinc = 1; /* no frags */
802 	ump->um_maxsymlinklen = EXT2_MAXSYMLINKLEN;
803 	ump->um_dirblksiz = m_fs->e2fs_bsize;
804 	ump->um_maxfilesize = ((uint64_t)0x80000000 * m_fs->e2fs_bsize - 1);
805 	spec_node_setmountedfs(devvp, mp);
806 	return 0;
807 
808 out1:
809 	kmem_free(m_fs->e2fs_gd, m_fs->e2fs_ngdb * sh * sizeof(struct ext2_gd));
810 out:
811 	if (bp != NULL)
812 		brelse(bp, 0);
813 	if (ump) {
814 		kmem_free(ump->um_e2fs, sizeof(*m_fs));
815 		kmem_free(ump, sizeof(*ump));
816 		mp->mnt_data = NULL;
817 	}
818 	return error;
819 }
820 
821 /*
822  * unmount system call
823  */
824 int
825 ext2fs_unmount(struct mount *mp, int mntflags)
826 {
827 	struct ufsmount *ump;
828 	struct m_ext2fs *fs;
829 	int error, flags;
830 
831 	flags = 0;
832 	if (mntflags & MNT_FORCE)
833 		flags |= FORCECLOSE;
834 	if ((error = ext2fs_flushfiles(mp, flags)) != 0)
835 		return error;
836 	ump = VFSTOUFS(mp);
837 	fs = ump->um_e2fs;
838 	if (fs->e2fs_ronly == 0 &&
839 		ext2fs_cgupdate(ump, MNT_WAIT) == 0 &&
840 		(fs->e2fs.e2fs_state & E2FS_ERRORS) == 0) {
841 		fs->e2fs.e2fs_state = E2FS_ISCLEAN;
842 		(void) ext2fs_sbupdate(ump, MNT_WAIT);
843 	}
844 	if (ump->um_devvp->v_type != VBAD)
845 		spec_node_setmountedfs(ump->um_devvp, NULL);
846 	vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
847 	error = VOP_CLOSE(ump->um_devvp, fs->e2fs_ronly ? FREAD : FREAD|FWRITE,
848 	    NOCRED);
849 	vput(ump->um_devvp);
850 	int32_t sh = fs->e2fs_bsize >> fs->e2fs_group_desc_shift;
851 	kmem_free(fs->e2fs_gd, fs->e2fs_ngdb * sh * sizeof(struct ext2_gd));
852 	kmem_free(fs, sizeof(*fs));
853 	kmem_free(ump, sizeof(*ump));
854 	mp->mnt_data = NULL;
855 	mp->mnt_flag &= ~MNT_LOCAL;
856 	return error;
857 }
858 
859 /*
860  * Flush out all the files in a filesystem.
861  */
862 int
863 ext2fs_flushfiles(struct mount *mp, int flags)
864 {
865 	extern int doforce;
866 	int error;
867 
868 	if (!doforce)
869 		flags &= ~FORCECLOSE;
870 	error = vflush(mp, NULLVP, flags);
871 	return error;
872 }
873 
874 /*
875  * Get file system statistics.
876  */
877 int
878 ext2fs_statvfs(struct mount *mp, struct statvfs *sbp)
879 {
880 	struct ufsmount *ump;
881 	struct m_ext2fs *fs;
882 	uint32_t overhead, overhead_per_group, ngdb;
883 	int i, ngroups;
884 
885 	ump = VFSTOUFS(mp);
886 	fs = ump->um_e2fs;
887 	if (fs->e2fs.e2fs_magic != E2FS_MAGIC)
888 		panic("ext2fs_statvfs");
889 
890 	/*
891 	 * Compute the overhead (FS structures)
892 	 */
893 	overhead_per_group =
894 	    1 /* block bitmap */ +
895 	    1 /* inode bitmap */ +
896 	    fs->e2fs_itpg;
897 	overhead = fs->e2fs.e2fs_first_dblock +
898 	    fs->e2fs_ncg * overhead_per_group;
899 	if (EXT2F_HAS_COMPAT_FEATURE(fs, EXT2F_COMPAT_SPARSESUPER2)) {
900 		/*
901 		 * Superblock and group descriptions is in group zero,
902 		 * then optionally 0, 1 or 2 extra copies.
903 		 */
904 		ngroups = 1
905 			+ (fs->e2fs.e4fs_backup_bgs[0] ? 1 : 0)
906 			+ (fs->e2fs.e4fs_backup_bgs[1] ? 1 : 0);
907 	} else if (EXT2F_HAS_ROCOMPAT_FEATURE(fs, EXT2F_ROCOMPAT_SPARSESUPER)) {
908 		for (i = 0, ngroups = 0; i < fs->e2fs_ncg; i++) {
909 			if (cg_has_sb(i))
910 				ngroups++;
911 		}
912 	} else {
913 		ngroups = fs->e2fs_ncg;
914 	}
915 	ngdb = fs->e2fs_ngdb;
916 	if (EXT2F_HAS_COMPAT_FEATURE(fs, EXT2F_COMPAT_RESIZE))
917 		ngdb += fs->e2fs.e2fs_reserved_ngdb;
918 	overhead += ngroups * (1 /* superblock */ + ngdb);
919 
920 	sbp->f_bsize = fs->e2fs_bsize;
921 	sbp->f_frsize = MINBSIZE << fs->e2fs.e2fs_fsize;
922 	sbp->f_iosize = fs->e2fs_bsize;
923 	sbp->f_blocks = fs->e2fs.e2fs_bcount - overhead;
924 	sbp->f_bfree = fs->e2fs.e2fs_fbcount;
925 	sbp->f_bresvd = fs->e2fs.e2fs_rbcount;
926 	if (sbp->f_bfree > sbp->f_bresvd)
927 		sbp->f_bavail = sbp->f_bfree - sbp->f_bresvd;
928 	else
929 		sbp->f_bavail = 0;
930 	sbp->f_files =  fs->e2fs.e2fs_icount;
931 	sbp->f_ffree = fs->e2fs.e2fs_ficount;
932 	sbp->f_favail = fs->e2fs.e2fs_ficount;
933 	sbp->f_fresvd = 0;
934 	copy_statvfs_info(sbp, mp);
935 	return 0;
936 }
937 
938 static bool
939 ext2fs_sync_selector(void *cl, struct vnode *vp)
940 {
941 	struct inode *ip;
942 
943 	KASSERT(mutex_owned(vp->v_interlock));
944 
945 	ip = VTOI(vp);
946 	/*
947 	 * Skip the vnode/inode if inaccessible.
948 	 */
949 	if (ip == NULL || vp->v_type == VNON)
950 		return false;
951 
952 	if (((ip->i_flag &
953 	      (IN_CHANGE | IN_UPDATE | IN_MODIFIED)) == 0 &&
954 	     LIST_EMPTY(&vp->v_dirtyblkhd) &&
955 	     (vp->v_iflag & VI_ONWORKLST) == 0))
956 		return false;
957 	return true;
958 }
959 
960 /*
961  * Go through the disk queues to initiate sandbagged IO;
962  * go through the inodes to write those that have been modified;
963  * initiate the writing of the super block if it has been modified.
964  */
965 int
966 ext2fs_sync(struct mount *mp, int waitfor, kauth_cred_t cred)
967 {
968 	struct vnode *vp;
969 	struct ufsmount *ump = VFSTOUFS(mp);
970 	struct m_ext2fs *fs;
971 	struct vnode_iterator *marker;
972 	int error, allerror = 0;
973 
974 	fs = ump->um_e2fs;
975 	if (fs->e2fs_fmod != 0 && fs->e2fs_ronly != 0) {	/* XXX */
976 		printf("fs = %s\n", fs->e2fs_fsmnt);
977 		panic("update: rofs mod");
978 	}
979 
980 	/*
981 	 * Write back each (modified) inode.
982 	 */
983 	vfs_vnode_iterator_init(mp, &marker);
984 	while ((vp = vfs_vnode_iterator_next(marker, ext2fs_sync_selector,
985 	    NULL)))
986 	{
987 		error = vn_lock(vp, LK_EXCLUSIVE);
988 		if (error) {
989 			vrele(vp);
990 			continue;
991 		}
992 		if (vp->v_type == VREG && waitfor == MNT_LAZY)
993 			error = ext2fs_update(vp, NULL, NULL, 0);
994 		else
995 			error = VOP_FSYNC(vp, cred,
996 			    waitfor == MNT_WAIT ? FSYNC_WAIT : 0, 0, 0);
997 		if (error)
998 			allerror = error;
999 		vput(vp);
1000 	}
1001 	vfs_vnode_iterator_destroy(marker);
1002 	/*
1003 	 * Force stale file system control information to be flushed.
1004 	 */
1005 	if (waitfor != MNT_LAZY) {
1006 		vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
1007 		if ((error = VOP_FSYNC(ump->um_devvp, cred,
1008 		    waitfor == MNT_WAIT ? FSYNC_WAIT : 0, 0, 0)) != 0)
1009 			allerror = error;
1010 		VOP_UNLOCK(ump->um_devvp);
1011 	}
1012 	/*
1013 	 * Write back modified superblock.
1014 	 */
1015 	if (fs->e2fs_fmod != 0) {
1016 		fs->e2fs_fmod = 0;
1017 		fs->e2fs.e2fs_wtime = time_second;
1018 		if ((error = ext2fs_cgupdate(ump, waitfor)))
1019 			allerror = error;
1020 	}
1021 	return allerror;
1022 }
1023 
1024 /*
1025  * Load inode from disk and initialize vnode.
1026  */
1027 static int
1028 ext2fs_init_vnode(struct ufsmount *ump, struct vnode *vp, ino_t ino)
1029 {
1030 	struct m_ext2fs *fs;
1031 	struct inode *ip;
1032 	struct buf *bp;
1033 	int error;
1034 
1035 	fs = ump->um_e2fs;
1036 
1037 	/* Read in the disk contents for the inode, copy into the inode. */
1038 	error = bread(ump->um_devvp, EXT2_FSBTODB(fs, ino_to_fsba(fs, ino)),
1039 	    (int)fs->e2fs_bsize, 0, &bp);
1040 	if (error)
1041 		return error;
1042 
1043 	/* Allocate and initialize inode. */
1044 	ip = pool_get(&ext2fs_inode_pool, PR_WAITOK);
1045 	memset(ip, 0, sizeof(struct inode));
1046 	ip->i_vnode = vp;
1047 	ip->i_ump = ump;
1048 	ip->i_e2fs = fs;
1049 	ip->i_dev = ump->um_dev;
1050 	ip->i_number = ino;
1051 	ip->i_e2fs_last_lblk = 0;
1052 	ip->i_e2fs_last_blk = 0;
1053 
1054 	error = ext2fs_loadvnode_content(fs, ino, bp, ip);
1055 	brelse(bp, 0);
1056 	if (error) {
1057 		pool_put(&ext2fs_inode_pool, ip);
1058 		return error;
1059 	}
1060 
1061 	/* If the inode was deleted, reset all fields */
1062 	if (ip->i_e2fs_dtime != 0) {
1063 		ip->i_e2fs_mode = 0;
1064 		(void)ext2fs_setsize(ip, 0);
1065 		(void)ext2fs_setnblock(ip, 0);
1066 		memset(ip->i_e2fs_blocks, 0, sizeof(ip->i_e2fs_blocks));
1067 	}
1068 
1069 	/* Initialise vnode with this inode. */
1070 	vp->v_tag = VT_EXT2FS;
1071 	vp->v_op = ext2fs_vnodeop_p;
1072 	vp->v_data = ip;
1073 
1074 	/* Initialize genfs node. */
1075 	genfs_node_init(vp, &ext2fs_genfsops);
1076 
1077 	return 0;
1078 }
1079 
1080 /*
1081  * Read an inode from disk and initialize this vnode / inode pair.
1082  * Caller assures no other thread will try to load this inode.
1083  */
1084 int
1085 ext2fs_loadvnode(struct mount *mp, struct vnode *vp,
1086     const void *key, size_t key_len, const void **new_key)
1087 {
1088 	ino_t ino;
1089 	struct inode *ip;
1090 	struct ufsmount *ump;
1091 	int error;
1092 
1093 	KASSERT(key_len == sizeof(ino));
1094 	memcpy(&ino, key, key_len);
1095 	ump = VFSTOUFS(mp);
1096 
1097 	error = ext2fs_init_vnode(ump, vp, ino);
1098 	if (error)
1099 		return error;
1100 
1101 	ip = VTOI(vp);
1102 
1103 	/* Initialize the vnode from the inode. */
1104 	ext2fs_vinit(mp, ext2fs_specop_p, ext2fs_fifoop_p, &vp);
1105 
1106 	/* Finish inode initialization. */
1107 	ip->i_devvp = ump->um_devvp;
1108 	vref(ip->i_devvp);
1109 
1110 	/*
1111 	 * Set up a generation number for this inode if it does not
1112 	 * already have one. This should only happen on old filesystems.
1113 	 */
1114 
1115 	if (ip->i_e2fs_gen == 0) {
1116 		if (++ext2gennumber < (u_long)time_second)
1117 			ext2gennumber = time_second;
1118 		ip->i_e2fs_gen = ext2gennumber;
1119 		if ((mp->mnt_flag & MNT_RDONLY) == 0)
1120 			ip->i_flag |= IN_MODIFIED;
1121 	}
1122 	uvm_vnp_setsize(vp, ext2fs_size(ip));
1123 	*new_key = &ip->i_number;
1124 	return 0;
1125 }
1126 
1127 /*
1128  * Create a new inode on disk and initialize this vnode / inode pair.
1129  */
1130 int
1131 ext2fs_newvnode(struct mount *mp, struct vnode *dvp, struct vnode *vp,
1132     struct vattr *vap, kauth_cred_t cred, void *extra,
1133     size_t *key_len, const void **new_key)
1134 {
1135 	ino_t ino;
1136 	struct inode *ip, *pdir;
1137 	struct m_ext2fs *fs;
1138 	struct ufsmount *ump;
1139 	int error, mode;
1140 
1141 	KASSERT(dvp->v_mount == mp);
1142 	KASSERT(vap->va_type != VNON);
1143 
1144 	*key_len = sizeof(ino);
1145 
1146 	pdir = VTOI(dvp);
1147 	fs = pdir->i_e2fs;
1148 	ump = VFSTOUFS(mp);
1149 	mode = MAKEIMODE(vap->va_type, vap->va_mode);
1150 
1151 	/* Allocate fresh inode. */
1152 	error = ext2fs_valloc(dvp, mode, cred, &ino);
1153 	if (error)
1154 		return error;
1155 
1156 	/* Attach inode to vnode. */
1157 	error = ext2fs_init_vnode(ump, vp, ino);
1158 	if (error) {
1159 		ext2fs_vfree(dvp, ino, mode);
1160 		return error;
1161 	}
1162 
1163 	ip = VTOI(vp);
1164 
1165 	KASSERT(!E2FS_HAS_GD_CSUM(fs) ||
1166 	    (fs->e2fs_gd[ino_to_cg(fs, ino)].ext2bgd_flags &
1167 	    h2fs16(E2FS_BG_INODE_ZEROED)) != 0);
1168 
1169 	/* check for already used inode; makes sense only for ZEROED itable */
1170 	if (__predict_false(ip->i_e2fs_mode && ip->i_e2fs_nlink != 0)) {
1171 		printf("mode = 0%o, nlinks %d, inum = %llu, fs = %s\n",
1172 		    ip->i_e2fs_mode, ip->i_e2fs_nlink,
1173 		    (unsigned long long)ip->i_number, fs->e2fs_fsmnt);
1174 		panic("ext2fs_valloc: dup alloc");
1175 	}
1176 
1177 	memset(ip->i_din.e2fs_din, 0, EXT2_DINODE_SIZE(fs));
1178 
1179 	/*
1180 	 * Set up a new generation number for this inode.
1181 	 */
1182 	if (++ext2gennumber < time_second)
1183 		ext2gennumber = time_second;
1184 	ip->i_e2fs_gen = ext2gennumber;
1185 
1186 	ip->i_uid = kauth_cred_geteuid(cred);
1187 	ip->i_e2fs_uid = ip->i_uid & 0xffff;
1188 	ip->i_e2fs_gid = pdir->i_e2fs_gid;
1189 	if (ip->i_e2fs->e2fs.e2fs_rev > E2FS_REV0) {
1190 		ip->i_e2fs_uid_high = (ip->i_uid >> 16) & 0xffff;
1191 		ip->i_e2fs_gid_high = pdir->i_e2fs_gid_high;
1192 	} else {
1193 		ip->i_e2fs_uid_high = 0;
1194 		ip->i_e2fs_gid_high = 0;
1195 	}
1196 	ip->i_gid = ip->i_e2fs_gid | (ip->i_e2fs_gid_high << 16);
1197 	ip->i_flag |= IN_ACCESS | IN_CHANGE | IN_UPDATE;
1198 	ip->i_e2fs_mode = mode;
1199 	vp->v_type = IFTOVT(mode);
1200 	ip->i_e2fs_nlink = 1;
1201 
1202 	/* Authorize setting SGID if needed. */
1203 	if (ip->i_e2fs_mode & ISGID) {
1204 		error = kauth_authorize_vnode(cred, KAUTH_VNODE_WRITE_SECURITY,
1205 		    vp, NULL, genfs_can_chmod(vp, cred, ip->i_uid, ip->i_gid,
1206 		    mode));
1207 		if (error)
1208 			ip->i_e2fs_mode &= ~ISGID;
1209 	}
1210 
1211 	/* Initialize extra_isize according to what is set in superblock */
1212 	if (EXT2F_HAS_ROCOMPAT_FEATURE(ip->i_e2fs, EXT2F_ROCOMPAT_EXTRA_ISIZE)
1213 	    && EXT2_DINODE_SIZE(ip->i_e2fs) > EXT2_REV0_DINODE_SIZE) {
1214 		ip->i_din.e2fs_din->e2di_extra_isize =
1215 		    ip->i_e2fs->e2fs.e4fs_want_extra_isize;
1216 	}
1217 
1218 	/* Set create time if possible */
1219 	if (EXT2_DINODE_FITS(ip->i_din.e2fs_din, e2di_crtime,
1220 	    EXT2_DINODE_SIZE(ip->i_e2fs))) {
1221 		struct timespec now;
1222 		vfs_timestamp(&now);
1223 		EXT2_DINODE_TIME_SET(&now, ip->i_din.e2fs_din, e2di_crtime,
1224 		    EXT2_DINODE_SIZE(ip->i_e2fs));
1225 	}
1226 
1227 	/* Initialize the vnode from the inode. */
1228 	ext2fs_vinit(mp, ext2fs_specop_p, ext2fs_fifoop_p, &vp);
1229 
1230 	/* Finish inode initialization. */
1231 	ip->i_devvp = ump->um_devvp;
1232 	vref(ip->i_devvp);
1233 
1234 	uvm_vnp_setsize(vp, ext2fs_size(ip));
1235 	*new_key = &ip->i_number;
1236 	return 0;
1237 }
1238 
1239 /*
1240  * File handle to vnode
1241  *
1242  * Have to be really careful about stale file handles:
1243  * - check that the inode number is valid
1244  * - call ext2fs_vget() to get the locked inode
1245  * - check for an unallocated inode (i_mode == 0)
1246  */
1247 int
1248 ext2fs_fhtovp(struct mount *mp, struct fid *fhp, int lktype, struct vnode **vpp)
1249 {
1250 	struct inode *ip;
1251 	struct vnode *nvp;
1252 	int error;
1253 	struct ufid ufh;
1254 	struct m_ext2fs *fs;
1255 
1256 	if (fhp->fid_len != sizeof(struct ufid))
1257 		return EINVAL;
1258 
1259 	memcpy(&ufh, fhp, sizeof(struct ufid));
1260 	fs = VFSTOUFS(mp)->um_e2fs;
1261 	if ((ufh.ufid_ino < EXT2_FIRSTINO && ufh.ufid_ino != EXT2_ROOTINO) ||
1262 		ufh.ufid_ino >= fs->e2fs_ncg * fs->e2fs.e2fs_ipg)
1263 		return ESTALE;
1264 
1265 	if ((error = VFS_VGET(mp, ufh.ufid_ino, lktype, &nvp)) != 0) {
1266 		*vpp = NULLVP;
1267 		return error;
1268 	}
1269 	ip = VTOI(nvp);
1270 	if (ip->i_e2fs_mode == 0 || ip->i_e2fs_dtime != 0 ||
1271 		ip->i_e2fs_gen != ufh.ufid_gen) {
1272 		vput(nvp);
1273 		*vpp = NULLVP;
1274 		return ESTALE;
1275 	}
1276 	*vpp = nvp;
1277 	return 0;
1278 }
1279 
1280 /*
1281  * Vnode pointer to File handle
1282  */
1283 /* ARGSUSED */
1284 int
1285 ext2fs_vptofh(struct vnode *vp, struct fid *fhp, size_t *fh_size)
1286 {
1287 	struct inode *ip;
1288 	struct ufid ufh;
1289 
1290 	if (*fh_size < sizeof(struct ufid)) {
1291 		*fh_size = sizeof(struct ufid);
1292 		return E2BIG;
1293 	}
1294 	*fh_size = sizeof(struct ufid);
1295 
1296 	ip = VTOI(vp);
1297 	memset(&ufh, 0, sizeof(ufh));
1298 	ufh.ufid_len = sizeof(struct ufid);
1299 	ufh.ufid_ino = ip->i_number;
1300 	ufh.ufid_gen = ip->i_e2fs_gen;
1301 	memcpy(fhp, &ufh, sizeof(ufh));
1302 	return 0;
1303 }
1304 
1305 /*
1306  * Write a superblock and associated information back to disk.
1307  */
1308 int
1309 ext2fs_sbupdate(struct ufsmount *mp, int waitfor)
1310 {
1311 	struct m_ext2fs *fs = mp->um_e2fs;
1312 	struct buf *bp;
1313 	int error = 0;
1314 
1315 	bp = getblk(mp->um_devvp, SBLOCK, SBSIZE, 0, 0);
1316 	e2fs_sbsave(&fs->e2fs, (struct ext2fs*)bp->b_data);
1317 	if (waitfor == MNT_WAIT)
1318 		error = bwrite(bp);
1319 	else
1320 		bawrite(bp);
1321 	return error;
1322 }
1323 
1324 int
1325 ext2fs_cgupdate(struct ufsmount *mp, int waitfor)
1326 {
1327 	struct m_ext2fs *fs = mp->um_e2fs;
1328 	struct buf *bp;
1329 	int i, error = 0, allerror = 0;
1330 
1331 	allerror = ext2fs_sbupdate(mp, waitfor);
1332 	for (i = 0; i < fs->e2fs_ngdb; i++) {
1333 		bp = getblk(mp->um_devvp, EXT2_FSBTODB(fs,
1334 		    fs->e2fs.e2fs_first_dblock +
1335 		    1 /* superblock */ + i), fs->e2fs_bsize, 0, 0);
1336 		e2fs_cgsave(&fs->e2fs_gd[i *
1337 			(fs->e2fs_bsize >> fs->e2fs_group_desc_shift)],
1338 		    bp->b_data, fs->e2fs_bsize, fs->e2fs_group_desc_shift);
1339 		if (waitfor == MNT_WAIT)
1340 			error = bwrite(bp);
1341 		else
1342 			bawrite(bp);
1343 	}
1344 
1345 	if (!allerror && error)
1346 		allerror = error;
1347 	return allerror;
1348 }
1349 
1350 /*
1351  * Fill in the m_fs structure, and validate the fields of the superblock.
1352  * NOTE: here, the superblock is already swapped.
1353  */
1354 static int
1355 ext2fs_sbfill(struct m_ext2fs *m_fs, int ronly)
1356 {
1357 	uint32_t u32;
1358 	struct ext2fs *fs = &m_fs->e2fs;
1359 
1360 	/*
1361 	 * General sanity checks
1362 	 */
1363 	if (fs->e2fs_magic != E2FS_MAGIC)
1364 		return EINVAL;
1365 	if (fs->e2fs_rev > E2FS_REV1) {
1366 		printf("ext2fs: unsupported revision number: %#x\n",
1367 		    fs->e2fs_rev);
1368 		return EINVAL;
1369 	}
1370 	if (fs->e2fs_log_bsize > 2) {
1371 		/* block size = 1024|2048|4096 */
1372 		printf("ext2fs: bad block size: %d\n", fs->e2fs_log_bsize);
1373 		return EINVAL;
1374 	}
1375 	if (fs->e2fs_bpg == 0) {
1376 		printf("ext2fs: zero blocks per group\n");
1377 		return EINVAL;
1378 	}
1379 	if (fs->e2fs_ipg == 0) {
1380 		printf("ext2fs: zero inodes per group\n");
1381 		return EINVAL;
1382 	}
1383 
1384 	if (fs->e2fs_first_dblock >= fs->e2fs_bcount) {
1385 		printf("ext2fs: invalid first data block\n");
1386 		return EINVAL;
1387 	}
1388 	if (fs->e2fs_rbcount > fs->e2fs_bcount ||
1389 	    fs->e2fs_fbcount > fs->e2fs_bcount) {
1390 		printf("ext2fs: invalid block count\n");
1391 		return EINVAL;
1392 	}
1393 
1394 	/*
1395 	 * Compute the fields of the superblock
1396 	 */
1397 	u32 = fs->e2fs_bcount - fs->e2fs_first_dblock; /* > 0 */
1398 	m_fs->e2fs_ncg = howmany(u32, fs->e2fs_bpg);
1399 	if (m_fs->e2fs_ncg == 0) {
1400 		printf("ext2fs: invalid number of cylinder groups\n");
1401 		return EINVAL;
1402 	}
1403 
1404 	m_fs->e2fs_fsbtodb = fs->e2fs_log_bsize + LOG_MINBSIZE - DEV_BSHIFT;
1405 	m_fs->e2fs_bsize = MINBSIZE << fs->e2fs_log_bsize;
1406 	m_fs->e2fs_bshift = LOG_MINBSIZE + fs->e2fs_log_bsize;
1407 	m_fs->e2fs_qbmask = m_fs->e2fs_bsize - 1;
1408 	m_fs->e2fs_bmask = ~m_fs->e2fs_qbmask;
1409 
1410 	if (!(fs->e2fs_features_incompat & EXT2F_INCOMPAT_64BIT) ||
1411 	    (fs->e2fs_rev == E2FS_REV0))
1412 		m_fs->e2fs_group_desc_shift = 5;
1413 	else {
1414 		for (m_fs->e2fs_group_desc_shift = 0;
1415 		     (1 << m_fs->e2fs_group_desc_shift)
1416 		       < fs->e3fs_desc_size;
1417 		     m_fs->e2fs_group_desc_shift++);
1418 	}
1419 
1420 	if ((u32 = (m_fs->e2fs_bsize >> m_fs->e2fs_group_desc_shift)) == 0) {
1421 		/* Unlikely to happen */
1422 		printf("ext2fs: invalid block size\n");
1423 		return EINVAL;
1424 	}
1425 	m_fs->e2fs_ngdb = howmany(m_fs->e2fs_ncg, u32);
1426 	if (m_fs->e2fs_ngdb == 0) {
1427 		printf("ext2fs: invalid number of group descriptor blocks\n");
1428 		return EINVAL;
1429 	}
1430 
1431 	if (m_fs->e2fs_bsize < EXT2_DINODE_SIZE(m_fs)) {
1432 		printf("ext2fs: invalid inode size\n");
1433 		return EINVAL;
1434 	}
1435 	m_fs->e2fs_ipb = m_fs->e2fs_bsize / EXT2_DINODE_SIZE(m_fs);
1436 
1437 	m_fs->e2fs_itpg = fs->e2fs_ipg / m_fs->e2fs_ipb;
1438 
1439 	/*
1440 	 * Revision-specific checks
1441 	 */
1442 	if (fs->e2fs_rev > E2FS_REV0) {
1443 		char buf[256];
1444 		if (fs->e2fs_first_ino != EXT2_FIRSTINO) {
1445 			printf("ext2fs: unsupported first inode position\n");
1446 			return EINVAL;
1447 		}
1448 		u32 = fs->e2fs_features_incompat & ~EXT2F_INCOMPAT_SUPP;
1449 		if (u32) {
1450 			snprintb(buf, sizeof(buf), EXT2F_INCOMPAT_BITS, u32);
1451 			printf("ext2fs: unsupported incompat features: %s\n",
1452 			    buf);
1453 #ifndef EXT2_IGNORE_INCOMPAT_FEATURES
1454 			return EINVAL;
1455 #endif
1456 		}
1457 		u32 = fs->e2fs_features_rocompat & ~EXT2F_ROCOMPAT_SUPP;
1458 		if (!ronly && u32) {
1459 			snprintb(buf, sizeof(buf), EXT2F_ROCOMPAT_BITS, u32);
1460 			printf("ext2fs: unsupported ro-incompat features: %s\n",
1461 			    buf);
1462 #ifndef EXT2_IGNORE_ROCOMPAT_FEATURES
1463 			return EROFS;
1464 #endif
1465 		}
1466 		if (fs->e2fs_inode_size == 0 || !powerof2(fs->e2fs_inode_size) || fs->e2fs_inode_size > m_fs->e2fs_bsize) {
1467 			printf("ext2fs: bad inode size\n");
1468 			return EINVAL;
1469 		}
1470 	}
1471 
1472 	return 0;
1473 }
1474