xref: /freebsd/sys/fs/ext2fs/ext2_vfsops.c (revision aa64588d28258aef88cc33b8043112e8856948d0)
1 /*-
2  *  modified for EXT2FS support in Lites 1.1
3  *
4  *  Aug 1995, Godmar Back (gback@cs.utah.edu)
5  *  University of Utah, Department of Computer Science
6  */
7 /*-
8  * Copyright (c) 1989, 1991, 1993, 1994
9  *	The Regents of the University of California.  All rights reserved.
10  *
11  * Redistribution and use in source and binary forms, with or without
12  * modification, are permitted provided that the following conditions
13  * are met:
14  * 1. Redistributions of source code must retain the above copyright
15  *    notice, this list of conditions and the following disclaimer.
16  * 2. Redistributions in binary form must reproduce the above copyright
17  *    notice, this list of conditions and the following disclaimer in the
18  *    documentation and/or other materials provided with the distribution.
19  * 4. Neither the name of the University nor the names of its contributors
20  *    may be used to endorse or promote products derived from this software
21  *    without specific prior written permission.
22  *
23  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
24  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
25  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
26  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
28  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
29  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
30  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
31  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
32  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33  * SUCH DAMAGE.
34  *
35  *	@(#)ffs_vfsops.c	8.8 (Berkeley) 4/18/94
36  * $FreeBSD$
37  */
38 
39 #include <sys/param.h>
40 #include <sys/systm.h>
41 #include <sys/namei.h>
42 #include <sys/priv.h>
43 #include <sys/proc.h>
44 #include <sys/kernel.h>
45 #include <sys/vnode.h>
46 #include <sys/mount.h>
47 #include <sys/bio.h>
48 #include <sys/buf.h>
49 #include <sys/conf.h>
50 #include <sys/fcntl.h>
51 #include <sys/malloc.h>
52 #include <sys/stat.h>
53 #include <sys/mutex.h>
54 
55 #include <geom/geom.h>
56 #include <geom/geom_vfs.h>
57 
58 #include <fs/ext2fs/ext2_mount.h>
59 #include <fs/ext2fs/inode.h>
60 
61 #include <fs/ext2fs/fs.h>
62 #include <fs/ext2fs/ext2_extern.h>
63 #include <fs/ext2fs/ext2fs.h>
64 
65 static int	ext2_flushfiles(struct mount *mp, int flags, struct thread *td);
66 static int	ext2_mountfs(struct vnode *, struct mount *);
67 static int	ext2_reload(struct mount *mp, struct thread *td);
68 static int	ext2_sbupdate(struct ext2mount *, int);
69 static int	ext2_cgupdate(struct ext2mount *, int);
70 static vfs_unmount_t		ext2_unmount;
71 static vfs_root_t		ext2_root;
72 static vfs_statfs_t		ext2_statfs;
73 static vfs_sync_t		ext2_sync;
74 static vfs_vget_t		ext2_vget;
75 static vfs_fhtovp_t		ext2_fhtovp;
76 static vfs_mount_t		ext2_mount;
77 
78 MALLOC_DEFINE(M_EXT2NODE, "ext2_node", "EXT2 vnode private part");
79 static MALLOC_DEFINE(M_EXT2MNT, "ext2_mount", "EXT2 mount structure");
80 
81 static struct vfsops ext2fs_vfsops = {
82 	.vfs_fhtovp =		ext2_fhtovp,
83 	.vfs_mount =		ext2_mount,
84 	.vfs_root =		ext2_root,	/* root inode via vget */
85 	.vfs_statfs =		ext2_statfs,
86 	.vfs_sync =		ext2_sync,
87 	.vfs_unmount =		ext2_unmount,
88 	.vfs_vget =		ext2_vget,
89 };
90 
91 VFS_SET(ext2fs_vfsops, ext2fs, 0);
92 
93 static int	ext2_check_sb_compat(struct ext2fs *es, struct cdev *dev,
94 		    int ronly);
95 static int	compute_sb_data(struct vnode * devvp,
96 		    struct ext2fs * es, struct m_ext2fs * fs);
97 
98 static const char *ext2_opts[] = { "from", "export", "acls", "noexec",
99     "noatime", "union", "suiddir", "multilabel", "nosymfollow",
100     "noclusterr", "noclusterw", "force", NULL };
101 
102 /*
103  * VFS Operations.
104  *
105  * mount system call
106  */
107 static int
108 ext2_mount(struct mount *mp)
109 {
110 	struct vfsoptlist *opts;
111 	struct vnode *devvp;
112 	struct thread *td;
113 	struct ext2mount *ump = 0;
114 	struct m_ext2fs *fs;
115 	struct nameidata nd, *ndp = &nd;
116 	accmode_t accmode;
117 	char *path, *fspec;
118 	int error, flags, len;
119 
120 	td = curthread;
121 	opts = mp->mnt_optnew;
122 
123 	if (vfs_filteropt(opts, ext2_opts))
124 		return (EINVAL);
125 
126 	vfs_getopt(opts, "fspath", (void **)&path, NULL);
127 	/* Double-check the length of path.. */
128 	if (strlen(path) >= MAXMNTLEN - 1)
129 		return (ENAMETOOLONG);
130 
131 	fspec = NULL;
132 	error = vfs_getopt(opts, "from", (void **)&fspec, &len);
133 	if (!error && fspec[len - 1] != '\0')
134 		return (EINVAL);
135 
136 	/*
137 	 * If updating, check whether changing from read-only to
138 	 * read/write; if there is no device name, that's all we do.
139 	 */
140 	if (mp->mnt_flag & MNT_UPDATE) {
141 		ump = VFSTOEXT2(mp);
142 		fs = ump->um_e2fs;
143 		error = 0;
144 		if (fs->e2fs_ronly == 0 &&
145 		    vfs_flagopt(opts, "ro", NULL, 0)) {
146 			error = VFS_SYNC(mp, MNT_WAIT);
147 			if (error)
148 				return (error);
149 			flags = WRITECLOSE;
150 			if (mp->mnt_flag & MNT_FORCE)
151 				flags |= FORCECLOSE;
152 			error = ext2_flushfiles(mp, flags, td);
153 			if ( error == 0 && fs->e2fs_wasvalid && ext2_cgupdate(ump, MNT_WAIT) == 0) {
154 				fs->e2fs->e2fs_state |= E2FS_ISCLEAN;
155 				ext2_sbupdate(ump, MNT_WAIT);
156 			}
157 			fs->e2fs_ronly = 1;
158 			vfs_flagopt(opts, "ro", &mp->mnt_flag, MNT_RDONLY);
159 			DROP_GIANT();
160 			g_topology_lock();
161 			g_access(ump->um_cp, 0, -1, 0);
162 			g_topology_unlock();
163 			PICKUP_GIANT();
164 		}
165 		if (!error && (mp->mnt_flag & MNT_RELOAD))
166 			error = ext2_reload(mp, td);
167 		if (error)
168 			return (error);
169 		devvp = ump->um_devvp;
170 		if (fs->e2fs_ronly && !vfs_flagopt(opts, "ro", NULL, 0)) {
171 			if (ext2_check_sb_compat(fs->e2fs, devvp->v_rdev, 0))
172 				return (EPERM);
173 
174 			/*
175 			 * If upgrade to read-write by non-root, then verify
176 			 * that user has necessary permissions on the device.
177 			 */
178 			vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
179 			error = VOP_ACCESS(devvp, VREAD | VWRITE,
180 			    td->td_ucred, td);
181 			if (error)
182 				error = priv_check(td, PRIV_VFS_MOUNT_PERM);
183 			if (error) {
184 				VOP_UNLOCK(devvp, 0);
185 				return (error);
186 			}
187 			VOP_UNLOCK(devvp, 0);
188 			DROP_GIANT();
189 			g_topology_lock();
190 			error = g_access(ump->um_cp, 0, 1, 0);
191 			g_topology_unlock();
192 			PICKUP_GIANT();
193 			if (error)
194 				return (error);
195 
196 			if ((fs->e2fs->e2fs_state & E2FS_ISCLEAN) == 0 ||
197 			    (fs->e2fs->e2fs_state & E2FS_ERRORS)) {
198 				if (mp->mnt_flag & MNT_FORCE) {
199 					printf(
200 "WARNING: %s was not properly dismounted\n", fs->e2fs_fsmnt);
201 				} else {
202 					printf(
203 "WARNING: R/W mount of %s denied.  Filesystem is not clean - run fsck\n",
204 					    fs->e2fs_fsmnt);
205 					return (EPERM);
206 				}
207 			}
208 			fs->e2fs->e2fs_state &= ~E2FS_ISCLEAN;
209 			(void)ext2_cgupdate(ump, MNT_WAIT);
210 			fs->e2fs_ronly = 0;
211 			MNT_ILOCK(mp);
212 			mp->mnt_flag &= ~MNT_RDONLY;
213 			MNT_IUNLOCK(mp);
214 		}
215 		if (vfs_flagopt(opts, "export", NULL, 0)) {
216 			/* Process export requests in vfs_mount.c. */
217 			return (error);
218 		}
219 	}
220 
221 	/*
222 	 * Not an update, or updating the name: look up the name
223 	 * and verify that it refers to a sensible disk device.
224 	 */
225 	if (fspec == NULL)
226 		return (EINVAL);
227 	NDINIT(ndp, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, fspec, td);
228 	if ((error = namei(ndp)) != 0)
229 		return (error);
230 	NDFREE(ndp, NDF_ONLY_PNBUF);
231 	devvp = ndp->ni_vp;
232 
233 	if (!vn_isdisk(devvp, &error)) {
234 		vput(devvp);
235 		return (error);
236 	}
237 
238 	/*
239 	 * If mount by non-root, then verify that user has necessary
240 	 * permissions on the device.
241 	 *
242 	 * XXXRW: VOP_ACCESS() enough?
243 	 */
244 	accmode = VREAD;
245 	if ((mp->mnt_flag & MNT_RDONLY) == 0)
246 		accmode |= VWRITE;
247 	error = VOP_ACCESS(devvp, accmode, td->td_ucred, td);
248 	if (error)
249 		error = priv_check(td, PRIV_VFS_MOUNT_PERM);
250 	if (error) {
251 		vput(devvp);
252 		return (error);
253 	}
254 
255 	if ((mp->mnt_flag & MNT_UPDATE) == 0) {
256 		error = ext2_mountfs(devvp, mp);
257 	} else {
258 		if (devvp != ump->um_devvp) {
259 			vput(devvp);
260 			return (EINVAL);	/* needs translation */
261 		} else
262 			vput(devvp);
263 	}
264 	if (error) {
265 		vrele(devvp);
266 		return (error);
267 	}
268 	ump = VFSTOEXT2(mp);
269 	fs = ump->um_e2fs;
270 
271 	/*
272 	 * Note that this strncpy() is ok because of a check at the start
273 	 * of ext2_mount().
274 	 */
275 	strncpy(fs->e2fs_fsmnt, path, MAXMNTLEN);
276 	fs->e2fs_fsmnt[MAXMNTLEN - 1] = '\0';
277 	vfs_mountedfrom(mp, fspec);
278 	return (0);
279 }
280 
281 static int
282 ext2_check_sb_compat(struct ext2fs *es, struct cdev *dev, int ronly)
283 {
284 
285 	if (es->e2fs_magic != E2FS_MAGIC) {
286 		printf("ext2fs: %s: wrong magic number %#x (expected %#x)\n",
287 		    devtoname(dev), es->e2fs_magic, E2FS_MAGIC);
288 		return (1);
289 	}
290 	if (es->e2fs_rev > E2FS_REV0) {
291 		if (es->e2fs_features_incompat & ~EXT2F_INCOMPAT_SUPP) {
292 			printf(
293 "WARNING: mount of %s denied due to unsupported optional features\n",
294 			    devtoname(dev));
295 			return (1);
296 		}
297 		if (!ronly &&
298 		    (es->e2fs_features_rocompat & ~EXT2F_ROCOMPAT_SUPP)) {
299 			printf("WARNING: R/W mount of %s denied due to "
300 			    "unsupported optional features\n", devtoname(dev));
301 			return (1);
302 		}
303 	}
304 	return (0);
305 }
306 
307 /*
308  * This computes the fields of the  ext2_sb_info structure from the
309  * data in the ext2_super_block structure read in.
310  */
311 static int
312 compute_sb_data(struct vnode *devvp, struct ext2fs *es,
313     struct m_ext2fs *fs)
314 {
315 	int db_count, error;
316 	int i;
317 	int logic_sb_block = 1;	/* XXX for now */
318 	struct buf *bp;
319 
320 	fs->e2fs_bsize = EXT2_MIN_BLOCK_SIZE << es->e2fs_log_bsize;
321 	fs->e2fs_bshift = EXT2_MIN_BLOCK_LOG_SIZE + es->e2fs_log_bsize;
322 	fs->e2fs_fsbtodb = es->e2fs_log_bsize + 1;
323 	fs->e2fs_qbmask = fs->e2fs_bsize - 1;
324 	fs->e2fs_blocksize_bits = es->e2fs_log_bsize + 10;
325 	fs->e2fs_fsize = EXT2_MIN_FRAG_SIZE << es->e2fs_log_fsize;
326 	if (fs->e2fs_fsize)
327 		fs->e2fs_fpb = fs->e2fs_bsize / fs->e2fs_fsize;
328 	fs->e2fs_bpg = es->e2fs_bpg;
329 	fs->e2fs_fpg = es->e2fs_fpg;
330 	fs->e2fs_ipg = es->e2fs_ipg;
331 	if (es->e2fs_rev == E2FS_REV0) {
332 		fs->e2fs_first_inode = E2FS_REV0_FIRST_INO;
333 		fs->e2fs_isize = E2FS_REV0_INODE_SIZE ;
334 	} else {
335 		fs->e2fs_first_inode = es->e2fs_first_ino;
336 		fs->e2fs_isize = es->e2fs_inode_size;
337 
338 		/*
339 		 * Simple sanity check for superblock inode size value.
340 		 */
341 		if (fs->e2fs_isize < E2FS_REV0_INODE_SIZE  ||
342 		    fs->e2fs_isize > fs->e2fs_bsize ||
343 		    (fs->e2fs_isize & (fs->e2fs_isize - 1)) != 0) {
344 			printf("EXT2-fs: invalid inode size %d\n",
345 			    fs->e2fs_isize);
346 			return (EIO);
347 		}
348 	}
349 	fs->e2fs_ipb = fs->e2fs_bsize / EXT2_INODE_SIZE(fs);
350 	fs->e2fs_itpg = fs->e2fs_ipg /fs->e2fs_ipb;
351 	fs->e2fs_descpb = fs->e2fs_bsize / sizeof (struct ext2_gd);
352 	/* s_resuid / s_resgid ? */
353 	fs->e2fs_gcount = (es->e2fs_bcount - es->e2fs_first_dblock +
354 	    EXT2_BLOCKS_PER_GROUP(fs) - 1) / EXT2_BLOCKS_PER_GROUP(fs);
355 	db_count = (fs->e2fs_gcount + EXT2_DESC_PER_BLOCK(fs) - 1) /
356 	    EXT2_DESC_PER_BLOCK(fs);
357 	fs->e2fs_gdbcount = db_count;
358 	fs->e2fs_gd = malloc(db_count * fs->e2fs_bsize,
359 	    M_EXT2MNT, M_WAITOK);
360 	fs->e2fs_contigdirs = malloc(fs->e2fs_gcount * sizeof(*fs->e2fs_contigdirs),
361 	    M_EXT2MNT, M_WAITOK);
362 
363 	/*
364 	 * Adjust logic_sb_block.
365 	 * Godmar thinks: if the blocksize is greater than 1024, then
366 	 * the superblock is logically part of block zero.
367 	 */
368 	if(fs->e2fs_bsize > SBSIZE)
369 		logic_sb_block = 0;
370 	for (i = 0; i < db_count; i++) {
371 		error = bread(devvp ,
372 			 fsbtodb(fs, logic_sb_block + i + 1 ),
373 			fs->e2fs_bsize, NOCRED, &bp);
374 		if (error) {
375 			free(fs->e2fs_gd, M_EXT2MNT);
376 			brelse(bp);
377 			return (error);
378 		}
379 		e2fs_cgload((struct ext2_gd *)bp->b_data,
380 		    &fs->e2fs_gd[
381 			i * fs->e2fs_bsize / sizeof(struct ext2_gd)],
382 		    fs->e2fs_bsize);
383 		brelse(bp);
384 		bp = NULL;
385 	}
386 	fs->e2fs_total_dir = 0;
387 	for (i=0; i < fs->e2fs_gcount; i++){
388 		fs->e2fs_total_dir += fs->e2fs_gd[i].ext2bgd_ndirs;
389 		fs->e2fs_contigdirs[i] = 0;
390 	}
391 	if (es->e2fs_rev == E2FS_REV0 ||
392 	    (es->e2fs_features_rocompat & EXT2F_ROCOMPAT_LARGEFILE) == 0)
393 		fs->e2fs_maxfilesize = 0x7fffffff;
394 	else
395 		fs->e2fs_maxfilesize = 0x7fffffffffffffff;
396 	return (0);
397 }
398 
399 /*
400  * Reload all incore data for a filesystem (used after running fsck on
401  * the root filesystem and finding things to fix). The filesystem must
402  * be mounted read-only.
403  *
404  * Things to do to update the mount:
405  *	1) invalidate all cached meta-data.
406  *	2) re-read superblock from disk.
407  *	3) re-read summary information from disk.
408  *	4) invalidate all inactive vnodes.
409  *	5) invalidate all cached file data.
410  *	6) re-read inode data for all active vnodes.
411  * XXX we are missing some steps, in particular # 3, this has to be reviewed.
412  */
413 static int
414 ext2_reload(struct mount *mp, struct thread *td)
415 {
416 	struct vnode *vp, *mvp, *devvp;
417 	struct inode *ip;
418 	struct buf *bp;
419 	struct ext2fs *es;
420 	struct m_ext2fs *fs;
421 	int error;
422 
423 	if ((mp->mnt_flag & MNT_RDONLY) == 0)
424 		return (EINVAL);
425 	/*
426 	 * Step 1: invalidate all cached meta-data.
427 	 */
428 	devvp = VFSTOEXT2(mp)->um_devvp;
429 	vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
430 	if (vinvalbuf(devvp, 0, 0, 0) != 0)
431 		panic("ext2_reload: dirty1");
432 	VOP_UNLOCK(devvp, 0);
433 
434 	/*
435 	 * Step 2: re-read superblock from disk.
436 	 * constants have been adjusted for ext2
437 	 */
438 	if ((error = bread(devvp, SBLOCK, SBSIZE, NOCRED, &bp)) != 0)
439 		return (error);
440 	es = (struct ext2fs *)bp->b_data;
441 	if (ext2_check_sb_compat(es, devvp->v_rdev, 0) != 0) {
442 		brelse(bp);
443 		return (EIO);		/* XXX needs translation */
444 	}
445 	fs = VFSTOEXT2(mp)->um_e2fs;
446 	bcopy(bp->b_data, fs->e2fs, sizeof(struct ext2fs));
447 
448 	if((error = compute_sb_data(devvp, es, fs)) != 0) {
449 		brelse(bp);
450 		return (error);
451 	}
452 #ifdef UNKLAR
453 	if (fs->fs_sbsize < SBSIZE)
454 		bp->b_flags |= B_INVAL;
455 #endif
456 	brelse(bp);
457 
458 loop:
459 	MNT_ILOCK(mp);
460 	MNT_VNODE_FOREACH(vp, mp, mvp) {
461 		VI_LOCK(vp);
462 		if (vp->v_iflag & VI_DOOMED) {
463 			VI_UNLOCK(vp);
464 			continue;
465 		}
466 		MNT_IUNLOCK(mp);
467 		/*
468 		 * Step 4: invalidate all cached file data.
469 		 */
470 		if (vget(vp, LK_EXCLUSIVE | LK_INTERLOCK, td)) {
471 			MNT_VNODE_FOREACH_ABORT(mp, mvp);
472 			goto loop;
473 		}
474 		if (vinvalbuf(vp, 0, 0, 0))
475 			panic("ext2_reload: dirty2");
476 
477 		/*
478 		 * Step 5: re-read inode data for all active vnodes.
479 		 */
480 		ip = VTOI(vp);
481 		error = bread(devvp, fsbtodb(fs, ino_to_fsba(fs, ip->i_number)),
482 		    (int)fs->e2fs_bsize, NOCRED, &bp);
483 		if (error) {
484 			VOP_UNLOCK(vp, 0);
485 			vrele(vp);
486 			MNT_VNODE_FOREACH_ABORT(mp, mvp);
487 			return (error);
488 		}
489 		ext2_ei2i((struct ext2fs_dinode *) ((char *)bp->b_data +
490 		    EXT2_INODE_SIZE(fs) * ino_to_fsbo(fs, ip->i_number)), ip);
491 		brelse(bp);
492 		VOP_UNLOCK(vp, 0);
493 		vrele(vp);
494 		MNT_ILOCK(mp);
495 	}
496 	MNT_IUNLOCK(mp);
497 	return (0);
498 }
499 
500 /*
501  * Common code for mount and mountroot.
502  */
503 static int
504 ext2_mountfs(struct vnode *devvp, struct mount *mp)
505 {
506 	struct ext2mount *ump;
507 	struct buf *bp;
508 	struct m_ext2fs *fs;
509 	struct ext2fs *es;
510 	struct cdev *dev = devvp->v_rdev;
511 	struct g_consumer *cp;
512 	struct bufobj *bo;
513 	int error;
514 	int ronly;
515 
516 	ronly = vfs_flagopt(mp->mnt_optnew, "ro", NULL, 0);
517 	/* XXX: use VOP_ACESS to check FS perms */
518 	DROP_GIANT();
519 	g_topology_lock();
520 	error = g_vfs_open(devvp, &cp, "ext2fs", ronly ? 0 : 1);
521 	g_topology_unlock();
522 	PICKUP_GIANT();
523 	VOP_UNLOCK(devvp, 0);
524 	if (error)
525 		return (error);
526 
527 	/* XXX: should we check for some sectorsize or 512 instead? */
528 	if (((SBSIZE % cp->provider->sectorsize) != 0) ||
529 	    (SBSIZE < cp->provider->sectorsize)) {
530 		DROP_GIANT();
531 		g_topology_lock();
532 		g_vfs_close(cp);
533 		g_topology_unlock();
534 		PICKUP_GIANT();
535 		return (EINVAL);
536 	}
537 
538 	bo = &devvp->v_bufobj;
539 	bo->bo_private = cp;
540 	bo->bo_ops = g_vfs_bufops;
541 	if (devvp->v_rdev->si_iosize_max != 0)
542 		mp->mnt_iosize_max = devvp->v_rdev->si_iosize_max;
543 	if (mp->mnt_iosize_max > MAXPHYS)
544 		mp->mnt_iosize_max = MAXPHYS;
545 
546 	bp = NULL;
547 	ump = NULL;
548 	if ((error = bread(devvp, SBLOCK, SBSIZE, NOCRED, &bp)) != 0)
549 		goto out;
550 	es = (struct ext2fs *)bp->b_data;
551 	if (ext2_check_sb_compat(es, dev, ronly) != 0) {
552 		error = EINVAL;		/* XXX needs translation */
553 		goto out;
554 	}
555 	if ((es->e2fs_state & E2FS_ISCLEAN) == 0 ||
556 	    (es->e2fs_state & E2FS_ERRORS)) {
557 		if (ronly || (mp->mnt_flag & MNT_FORCE)) {
558 			printf(
559 "WARNING: Filesystem was not properly dismounted\n");
560 		} else {
561 			printf(
562 "WARNING: R/W mount denied.  Filesystem is not clean - run fsck\n");
563 			error = EPERM;
564 			goto out;
565 		}
566 	}
567 	ump = malloc(sizeof *ump, M_EXT2MNT, M_WAITOK);
568 	bzero((caddr_t)ump, sizeof *ump);
569 
570 	/*
571 	 * I don't know whether this is the right strategy. Note that
572 	 * we dynamically allocate both an ext2_sb_info and an ext2_super_block
573 	 * while Linux keeps the super block in a locked buffer.
574 	 */
575 	ump->um_e2fs = malloc(sizeof(struct m_ext2fs),
576 		M_EXT2MNT, M_WAITOK);
577 	ump->um_e2fs->e2fs = malloc(sizeof(struct ext2fs),
578 		M_EXT2MNT, M_WAITOK);
579 	mtx_init(EXT2_MTX(ump), "EXT2FS", "EXT2FS Lock", MTX_DEF);
580 	bcopy(es, ump->um_e2fs->e2fs, (u_int)sizeof(struct ext2fs));
581 	if ((error = compute_sb_data(devvp, ump->um_e2fs->e2fs, ump->um_e2fs)))
582 		goto out;
583 
584 	brelse(bp);
585 	bp = NULL;
586 	fs = ump->um_e2fs;
587 	fs->e2fs_ronly = ronly;	/* ronly is set according to mnt_flags */
588 
589 	/*
590 	 * If the fs is not mounted read-only, make sure the super block is
591 	 * always written back on a sync().
592 	 */
593 	fs->e2fs_wasvalid = fs->e2fs->e2fs_state & E2FS_ISCLEAN ? 1 : 0;
594 	if (ronly == 0) {
595 		fs->e2fs_fmod = 1;		/* mark it modified */
596 		fs->e2fs->e2fs_state &= ~E2FS_ISCLEAN;	/* set fs invalid */
597 	}
598 	mp->mnt_data = ump;
599 	mp->mnt_stat.f_fsid.val[0] = dev2udev(dev);
600 	mp->mnt_stat.f_fsid.val[1] = mp->mnt_vfc->vfc_typenum;
601 	mp->mnt_maxsymlinklen = EXT2_MAXSYMLINKLEN;
602 	MNT_ILOCK(mp);
603 	mp->mnt_flag |= MNT_LOCAL;
604 	MNT_IUNLOCK(mp);
605 	ump->um_mountp = mp;
606 	ump->um_dev = dev;
607 	ump->um_devvp = devvp;
608 	ump->um_bo = &devvp->v_bufobj;
609 	ump->um_cp = cp;
610 
611 	/*
612 	 * Setting those two parameters allowed us to use
613 	 * ufs_bmap w/o changse!
614 	 */
615 	ump->um_nindir = EXT2_ADDR_PER_BLOCK(fs);
616 	ump->um_bptrtodb = fs->e2fs->e2fs_log_bsize + 1;
617 	ump->um_seqinc = EXT2_FRAGS_PER_BLOCK(fs);
618 	if (ronly == 0)
619 		ext2_sbupdate(ump, MNT_WAIT);
620 	/*
621 	 * Initialize filesystem stat information in mount struct.
622 	 */
623 	MNT_ILOCK(mp);
624  	mp->mnt_kern_flag |= MNTK_MPSAFE | MNTK_LOOKUP_SHARED |
625             MNTK_EXTENDED_SHARED;
626 	MNT_IUNLOCK(mp);
627 	return (0);
628 out:
629 	if (bp)
630 		brelse(bp);
631 	if (cp != NULL) {
632 		DROP_GIANT();
633 		g_topology_lock();
634 		g_vfs_close(cp);
635 		g_topology_unlock();
636 		PICKUP_GIANT();
637 	}
638 	if (ump) {
639 	  	mtx_destroy(EXT2_MTX(ump));
640 		free(ump->um_e2fs->e2fs_gd, M_EXT2MNT);
641 		free(ump->um_e2fs->e2fs_contigdirs, M_EXT2MNT);
642 		free(ump->um_e2fs->e2fs, M_EXT2MNT);
643 		free(ump->um_e2fs, M_EXT2MNT);
644 		free(ump, M_EXT2MNT);
645 		mp->mnt_data = NULL;
646 	}
647 	return (error);
648 }
649 
650 /*
651  * Unmount system call.
652  */
653 static int
654 ext2_unmount(struct mount *mp, int mntflags)
655 {
656 	struct ext2mount *ump;
657 	struct m_ext2fs *fs;
658 	int error, flags, ronly;
659 
660 	flags = 0;
661 	if (mntflags & MNT_FORCE) {
662 		if (mp->mnt_flag & MNT_ROOTFS)
663 			return (EINVAL);
664 		flags |= FORCECLOSE;
665 	}
666 	if ((error = ext2_flushfiles(mp, flags, curthread)) != 0)
667 		return (error);
668 	ump = VFSTOEXT2(mp);
669 	fs = ump->um_e2fs;
670 	ronly = fs->e2fs_ronly;
671 	if (ronly == 0 && ext2_cgupdate(ump, MNT_WAIT) == 0) {
672 		if (fs->e2fs_wasvalid)
673  			fs->e2fs->e2fs_state |= E2FS_ISCLEAN;
674  		ext2_sbupdate(ump, MNT_WAIT);
675 	}
676 
677 	DROP_GIANT();
678 	g_topology_lock();
679 	g_vfs_close(ump->um_cp);
680 	g_topology_unlock();
681 	PICKUP_GIANT();
682 	vrele(ump->um_devvp);
683 	free(fs->e2fs_gd, M_EXT2MNT);
684 	free(fs->e2fs_contigdirs, M_EXT2MNT);
685 	free(fs->e2fs, M_EXT2MNT);
686 	free(fs, M_EXT2MNT);
687 	free(ump, M_EXT2MNT);
688 	mp->mnt_data = NULL;
689 	MNT_ILOCK(mp);
690 	mp->mnt_flag &= ~MNT_LOCAL;
691 	MNT_IUNLOCK(mp);
692 	return (error);
693 }
694 
695 /*
696  * Flush out all the files in a filesystem.
697  */
698 static int
699 ext2_flushfiles(struct mount *mp, int flags, struct thread *td)
700 {
701 	int error;
702 
703 	error = vflush(mp, 0, flags, td);
704 	return (error);
705 }
706 /*
707  * Get file system statistics.
708  */
709 int
710 ext2_statfs(struct mount *mp, struct statfs *sbp)
711 {
712 	struct ext2mount *ump;
713 	struct m_ext2fs *fs;
714 	uint32_t overhead, overhead_per_group, ngdb;
715 	int i, ngroups;
716 
717 	ump = VFSTOEXT2(mp);
718 	fs = ump->um_e2fs;
719 	if (fs->e2fs->e2fs_magic != E2FS_MAGIC)
720 		panic("ext2fs_statvfs");
721 
722 	/*
723 	 * Compute the overhead (FS structures)
724 	 */
725 	overhead_per_group =
726 	    1 /* block bitmap */ +
727 	    1 /* inode bitmap */ +
728 	    fs->e2fs_itpg;
729 	overhead = fs->e2fs->e2fs_first_dblock +
730 	    fs->e2fs_gcount * overhead_per_group;
731 	if (fs->e2fs->e2fs_rev > E2FS_REV0 &&
732 	    fs->e2fs->e2fs_features_rocompat & EXT2F_ROCOMPAT_SPARSESUPER) {
733 		for (i = 0, ngroups = 0; i < fs->e2fs_gcount; i++) {
734 			if (cg_has_sb(i))
735 				ngroups++;
736 		}
737 	} else {
738 		ngroups = fs->e2fs_gcount;
739 	}
740 	ngdb = fs->e2fs_gdbcount;
741 	if (fs->e2fs->e2fs_rev > E2FS_REV0 &&
742 	    fs->e2fs->e2fs_features_compat & EXT2F_COMPAT_RESIZE)
743 		ngdb += fs->e2fs->e2fs_reserved_ngdb;
744 	overhead += ngroups * (1 /* superblock */ + ngdb);
745 
746 	sbp->f_bsize = EXT2_FRAG_SIZE(fs);
747 	sbp->f_iosize = EXT2_BLOCK_SIZE(fs);
748 	sbp->f_blocks = fs->e2fs->e2fs_bcount - overhead;
749 	sbp->f_bfree = fs->e2fs->e2fs_fbcount;
750 	sbp->f_bavail = sbp->f_bfree - fs->e2fs->e2fs_rbcount;
751 	sbp->f_files = fs->e2fs->e2fs_icount;
752 	sbp->f_ffree = fs->e2fs->e2fs_ficount;
753 	return (0);
754 }
755 
756 /*
757  * Go through the disk queues to initiate sandbagged IO;
758  * go through the inodes to write those that have been modified;
759  * initiate the writing of the super block if it has been modified.
760  *
761  * Note: we are always called with the filesystem marked `MPBUSY'.
762  */
763 static int
764 ext2_sync(struct mount *mp, int waitfor)
765 {
766 	struct vnode *mvp, *vp;
767 	struct thread *td;
768 	struct inode *ip;
769 	struct ext2mount *ump = VFSTOEXT2(mp);
770 	struct m_ext2fs *fs;
771 	int error, allerror = 0;
772 
773 	td = curthread;
774 	fs = ump->um_e2fs;
775 	if (fs->e2fs_fmod != 0 && fs->e2fs_ronly != 0) {		/* XXX */
776 		printf("fs = %s\n", fs->e2fs_fsmnt);
777 		panic("ext2_sync: rofs mod");
778 	}
779 
780 	/*
781 	 * Write back each (modified) inode.
782 	 */
783 	MNT_ILOCK(mp);
784 loop:
785 	MNT_VNODE_FOREACH(vp, mp, mvp) {
786 		VI_LOCK(vp);
787 		if (vp->v_type == VNON || (vp->v_iflag & VI_DOOMED)) {
788 			VI_UNLOCK(vp);
789 			continue;
790 		}
791 		MNT_IUNLOCK(mp);
792 		ip = VTOI(vp);
793 		if ((ip->i_flag &
794 		    (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 &&
795 		    (vp->v_bufobj.bo_dirty.bv_cnt == 0 ||
796 		    waitfor == MNT_LAZY)) {
797 			VI_UNLOCK(vp);
798 			MNT_ILOCK(mp);
799 			continue;
800 		}
801 		error = vget(vp, LK_EXCLUSIVE | LK_NOWAIT | LK_INTERLOCK, td);
802 		if (error) {
803 			MNT_ILOCK(mp);
804 			if (error == ENOENT) {
805 				MNT_VNODE_FOREACH_ABORT_ILOCKED(mp, mvp);
806 				goto loop;
807 			}
808 			continue;
809 		}
810 		if ((error = VOP_FSYNC(vp, waitfor, td)) != 0)
811 			allerror = error;
812 		VOP_UNLOCK(vp, 0);
813 		vrele(vp);
814 		MNT_ILOCK(mp);
815 	}
816 	MNT_IUNLOCK(mp);
817 
818 	/*
819 	 * Force stale file system control information to be flushed.
820 	 */
821 	if (waitfor != MNT_LAZY) {
822 		vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
823 		if ((error = VOP_FSYNC(ump->um_devvp, waitfor, td)) != 0)
824 			allerror = error;
825 		VOP_UNLOCK(ump->um_devvp, 0);
826 	}
827 
828 	/*
829 	 * Write back modified superblock.
830 	 */
831 	if (fs->e2fs_fmod != 0) {
832 		fs->e2fs_fmod = 0;
833 		fs->e2fs->e2fs_wtime = time_second;
834 		if ((error = ext2_cgupdate(ump, waitfor)) != 0)
835 			allerror = error;
836 	}
837 	return (allerror);
838 }
839 
840 /*
841  * Look up an EXT2FS dinode number to find its incore vnode, otherwise read it
842  * in from disk.  If it is in core, wait for the lock bit to clear, then
843  * return the inode locked.  Detection and handling of mount points must be
844  * done by the calling routine.
845  */
846 static int
847 ext2_vget(struct mount *mp, ino_t ino, int flags, struct vnode **vpp)
848 {
849 	struct m_ext2fs *fs;
850 	struct inode *ip;
851 	struct ext2mount *ump;
852 	struct buf *bp;
853 	struct vnode *vp;
854 	struct cdev *dev;
855 	struct thread *td;
856 	int i, error;
857 	int used_blocks;
858 
859 	td = curthread;
860 	error = vfs_hash_get(mp, ino, flags, td, vpp, NULL, NULL);
861 	if (error || *vpp != NULL)
862 		return (error);
863 
864 	ump = VFSTOEXT2(mp);
865 	dev = ump->um_dev;
866 
867 	/*
868 	 * If this malloc() is performed after the getnewvnode()
869 	 * it might block, leaving a vnode with a NULL v_data to be
870 	 * found by ext2_sync() if a sync happens to fire right then,
871 	 * which will cause a panic because ext2_sync() blindly
872 	 * dereferences vp->v_data (as well it should).
873 	 */
874 	ip = malloc(sizeof(struct inode), M_EXT2NODE, M_WAITOK | M_ZERO);
875 
876 	/* Allocate a new vnode/inode. */
877 	if ((error = getnewvnode("ext2fs", mp, &ext2_vnodeops, &vp)) != 0) {
878 		*vpp = NULL;
879 		free(ip, M_EXT2NODE);
880 		return (error);
881 	}
882 	vp->v_data = ip;
883 	ip->i_vnode = vp;
884 	ip->i_e2fs = fs = ump->um_e2fs;
885 	ip->i_ump  = ump;
886 	ip->i_number = ino;
887 
888 	lockmgr(vp->v_vnlock, LK_EXCLUSIVE, NULL);
889 	error = insmntque(vp, mp);
890 	if (error != 0) {
891 		free(ip, M_EXT2NODE);
892 		*vpp = NULL;
893 		return (error);
894 	}
895 	error = vfs_hash_insert(vp, ino, flags, td, vpp, NULL, NULL);
896 	if (error || *vpp != NULL)
897 		return (error);
898 
899 	/* Read in the disk contents for the inode, copy into the inode. */
900 	if ((error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)),
901 	    (int)fs->e2fs_bsize, NOCRED, &bp)) != 0) {
902 		/*
903 		 * The inode does not contain anything useful, so it would
904 		 * be misleading to leave it on its hash chain. With mode
905 		 * still zero, it will be unlinked and returned to the free
906 		 * list by vput().
907 		 */
908 		brelse(bp);
909 		vput(vp);
910 		*vpp = NULL;
911 		return (error);
912 	}
913 	/* convert ext2 inode to dinode */
914 	ext2_ei2i((struct ext2fs_dinode *) ((char *)bp->b_data + EXT2_INODE_SIZE(fs) *
915 			ino_to_fsbo(fs, ino)), ip);
916 	ip->i_block_group = ino_to_cg(fs, ino);
917 	ip->i_next_alloc_block = 0;
918 	ip->i_next_alloc_goal = 0;
919 	ip->i_prealloc_count = 0;
920 	ip->i_prealloc_block = 0;
921 
922 	/*
923 	 * Now we want to make sure that block pointers for unused
924 	 * blocks are zeroed out - ext2_balloc depends on this
925 	 * although for regular files and directories only
926 	 */
927 	if(S_ISDIR(ip->i_mode) || S_ISREG(ip->i_mode)) {
928 		used_blocks = (ip->i_size+fs->e2fs_bsize-1) / fs->e2fs_bsize;
929 		for(i = used_blocks; i < EXT2_NDIR_BLOCKS; i++)
930 			ip->i_db[i] = 0;
931 	}
932 /*
933 	ext2_print_inode(ip);
934 */
935 	bqrelse(bp);
936 
937 	/*
938 	 * Initialize the vnode from the inode, check for aliases.
939 	 * Note that the underlying vnode may have changed.
940 	 */
941 	if ((error = ext2_vinit(mp, &ext2_fifoops, &vp)) != 0) {
942 		vput(vp);
943 		*vpp = NULL;
944 		return (error);
945 	}
946 
947 	/*
948 	 * Finish inode initialization now that aliasing has been resolved.
949 	 */
950 	ip->i_devvp = ump->um_devvp;
951 
952 	/*
953 	 * Set up a generation number for this inode if it does not
954 	 * already have one. This should only happen on old filesystems.
955 	 */
956 	if (ip->i_gen == 0) {
957 		ip->i_gen = random() / 2 + 1;
958 		if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0)
959 			ip->i_flag |= IN_MODIFIED;
960 	}
961 	*vpp = vp;
962 	return (0);
963 }
964 
965 /*
966  * File handle to vnode
967  *
968  * Have to be really careful about stale file handles:
969  * - check that the inode number is valid
970  * - call ext2_vget() to get the locked inode
971  * - check for an unallocated inode (i_mode == 0)
972  * - check that the given client host has export rights and return
973  *   those rights via. exflagsp and credanonp
974  */
975 static int
976 ext2_fhtovp(struct mount *mp, struct fid *fhp, struct vnode **vpp)
977 {
978 	struct inode *ip;
979 	struct ufid *ufhp;
980 	struct vnode *nvp;
981 	struct m_ext2fs *fs;
982 	int error;
983 
984 	ufhp = (struct ufid *)fhp;
985 	fs = VFSTOEXT2(mp)->um_e2fs;
986 	if (ufhp->ufid_ino < ROOTINO ||
987 	    ufhp->ufid_ino > fs->e2fs_gcount * fs->e2fs->e2fs_ipg)
988 		return (ESTALE);
989 
990 	error = VFS_VGET(mp, ufhp->ufid_ino, LK_EXCLUSIVE, &nvp);
991 	if (error) {
992 		*vpp = NULLVP;
993 		return (error);
994 	}
995 	ip = VTOI(nvp);
996 	if (ip->i_mode == 0 ||
997 	    ip->i_gen != ufhp->ufid_gen || ip->i_nlink <= 0) {
998 		vput(nvp);
999 		*vpp = NULLVP;
1000 		return (ESTALE);
1001 	}
1002 	*vpp = nvp;
1003 	vnode_create_vobject(*vpp, 0, curthread);
1004 	return (0);
1005 }
1006 
1007 /*
1008  * Write a superblock and associated information back to disk.
1009  */
1010 static int
1011 ext2_sbupdate(struct ext2mount *mp, int waitfor)
1012 {
1013 	struct m_ext2fs *fs = mp->um_e2fs;
1014 	struct ext2fs *es = fs->e2fs;
1015 	struct buf *bp;
1016 	int error = 0;
1017 
1018 	bp = getblk(mp->um_devvp, SBLOCK, SBSIZE, 0, 0, 0);
1019 	bcopy((caddr_t)es, bp->b_data, (u_int)sizeof(struct ext2fs));
1020 	if (waitfor == MNT_WAIT)
1021 		error = bwrite(bp);
1022 	else
1023 		bawrite(bp);
1024 
1025 	/*
1026 	 * The buffers for group descriptors, inode bitmaps and block bitmaps
1027 	 * are not busy at this point and are (hopefully) written by the
1028 	 * usual sync mechanism. No need to write them here.
1029 	 */
1030 	return (error);
1031 }
1032 int
1033 ext2_cgupdate(struct ext2mount *mp, int waitfor)
1034 {
1035 	struct m_ext2fs *fs = mp->um_e2fs;
1036 	struct buf *bp;
1037 	int i, error = 0, allerror = 0;
1038 
1039 	allerror = ext2_sbupdate(mp, waitfor);
1040 	for (i = 0; i < fs->e2fs_gdbcount; i++) {
1041 		bp = getblk(mp->um_devvp, fsbtodb(fs,
1042 		    fs->e2fs->e2fs_first_dblock +
1043 		    1 /* superblock */ + i), fs->e2fs_bsize, 0, 0, 0);
1044 		e2fs_cgsave(&fs->e2fs_gd[
1045 		    i * fs->e2fs_bsize / sizeof(struct ext2_gd)],
1046 		    (struct ext2_gd *)bp->b_data, fs->e2fs_bsize);
1047 		if (waitfor == MNT_WAIT)
1048 			error = bwrite(bp);
1049 		else
1050 			bawrite(bp);
1051 	}
1052 
1053 	if (!allerror && error)
1054 		allerror = error;
1055 	return (allerror);
1056 }
1057 /*
1058  * Return the root of a filesystem.
1059  */
1060 static int
1061 ext2_root(struct mount *mp, int flags, struct vnode **vpp)
1062 {
1063 	struct vnode *nvp;
1064 	int error;
1065 
1066 	error = VFS_VGET(mp, (ino_t)ROOTINO, LK_EXCLUSIVE, &nvp);
1067 	if (error)
1068 		return (error);
1069 	*vpp = nvp;
1070 	return (0);
1071 }
1072