xref: /freebsd/sys/fs/ext2fs/ext2_vfsops.c (revision 7750ad47a9a7dbc83f87158464170c8640723293)
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/endian.h>
51 #include <sys/fcntl.h>
52 #include <sys/malloc.h>
53 #include <sys/stat.h>
54 #include <sys/mutex.h>
55 
56 #include <geom/geom.h>
57 #include <geom/geom_vfs.h>
58 
59 #include <fs/ext2fs/ext2_mount.h>
60 #include <fs/ext2fs/inode.h>
61 
62 #include <fs/ext2fs/fs.h>
63 #include <fs/ext2fs/ext2fs.h>
64 #include <fs/ext2fs/ext2_dinode.h>
65 #include <fs/ext2fs/ext2_extern.h>
66 
67 static int	ext2_flushfiles(struct mount *mp, int flags, struct thread *td);
68 static int	ext2_mountfs(struct vnode *, struct mount *);
69 static int	ext2_reload(struct mount *mp, struct thread *td);
70 static int	ext2_sbupdate(struct ext2mount *, int);
71 static int	ext2_cgupdate(struct ext2mount *, int);
72 static vfs_unmount_t		ext2_unmount;
73 static vfs_root_t		ext2_root;
74 static vfs_statfs_t		ext2_statfs;
75 static vfs_sync_t		ext2_sync;
76 static vfs_vget_t		ext2_vget;
77 static vfs_fhtovp_t		ext2_fhtovp;
78 static vfs_mount_t		ext2_mount;
79 
80 MALLOC_DEFINE(M_EXT2NODE, "ext2_node", "EXT2 vnode private part");
81 static MALLOC_DEFINE(M_EXT2MNT, "ext2_mount", "EXT2 mount structure");
82 
83 static struct vfsops ext2fs_vfsops = {
84 	.vfs_fhtovp =		ext2_fhtovp,
85 	.vfs_mount =		ext2_mount,
86 	.vfs_root =		ext2_root,	/* root inode via vget */
87 	.vfs_statfs =		ext2_statfs,
88 	.vfs_sync =		ext2_sync,
89 	.vfs_unmount =		ext2_unmount,
90 	.vfs_vget =		ext2_vget,
91 };
92 
93 VFS_SET(ext2fs_vfsops, ext2fs, 0);
94 
95 static int	ext2_check_sb_compat(struct ext2fs *es, struct cdev *dev,
96 		    int ronly);
97 static int	compute_sb_data(struct vnode * devvp,
98 		    struct ext2fs * es, struct m_ext2fs * fs);
99 
100 static const char *ext2_opts[] = { "acls", "async", "noatime", "noclusterr",
101     "noclusterw", "noexec", "export", "force", "from", "multilabel",
102     "suiddir", "nosymfollow", "sync", "union", NULL };
103 
104 /*
105  * VFS Operations.
106  *
107  * mount system call
108  */
109 static int
110 ext2_mount(struct mount *mp)
111 {
112 	struct vfsoptlist *opts;
113 	struct vnode *devvp;
114 	struct thread *td;
115 	struct ext2mount *ump = 0;
116 	struct m_ext2fs *fs;
117 	struct nameidata nd, *ndp = &nd;
118 	accmode_t accmode;
119 	char *path, *fspec;
120 	int error, flags, len;
121 
122 	td = curthread;
123 	opts = mp->mnt_optnew;
124 
125 	if (vfs_filteropt(opts, ext2_opts))
126 		return (EINVAL);
127 
128 	vfs_getopt(opts, "fspath", (void **)&path, NULL);
129 	/* Double-check the length of path.. */
130 	if (strlen(path) >= MAXMNTLEN - 1)
131 		return (ENAMETOOLONG);
132 
133 	fspec = NULL;
134 	error = vfs_getopt(opts, "from", (void **)&fspec, &len);
135 	if (!error && fspec[len - 1] != '\0')
136 		return (EINVAL);
137 
138 	/*
139 	 * If updating, check whether changing from read-only to
140 	 * read/write; if there is no device name, that's all we do.
141 	 */
142 	if (mp->mnt_flag & MNT_UPDATE) {
143 		ump = VFSTOEXT2(mp);
144 		fs = ump->um_e2fs;
145 		error = 0;
146 		if (fs->e2fs_ronly == 0 &&
147 		    vfs_flagopt(opts, "ro", NULL, 0)) {
148 			error = VFS_SYNC(mp, MNT_WAIT);
149 			if (error)
150 				return (error);
151 			flags = WRITECLOSE;
152 			if (mp->mnt_flag & MNT_FORCE)
153 				flags |= FORCECLOSE;
154 			error = ext2_flushfiles(mp, flags, td);
155 			if ( error == 0 && fs->e2fs_wasvalid && ext2_cgupdate(ump, MNT_WAIT) == 0) {
156 				fs->e2fs->e2fs_state |= E2FS_ISCLEAN;
157 				ext2_sbupdate(ump, MNT_WAIT);
158 			}
159 			fs->e2fs_ronly = 1;
160 			vfs_flagopt(opts, "ro", &mp->mnt_flag, MNT_RDONLY);
161 			DROP_GIANT();
162 			g_topology_lock();
163 			g_access(ump->um_cp, 0, -1, 0);
164 			g_topology_unlock();
165 			PICKUP_GIANT();
166 		}
167 		if (!error && (mp->mnt_flag & MNT_RELOAD))
168 			error = ext2_reload(mp, td);
169 		if (error)
170 			return (error);
171 		devvp = ump->um_devvp;
172 		if (fs->e2fs_ronly && !vfs_flagopt(opts, "ro", NULL, 0)) {
173 			if (ext2_check_sb_compat(fs->e2fs, devvp->v_rdev, 0))
174 				return (EPERM);
175 
176 			/*
177 			 * If upgrade to read-write by non-root, then verify
178 			 * that user has necessary permissions on the device.
179 			 */
180 			vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
181 			error = VOP_ACCESS(devvp, VREAD | VWRITE,
182 			    td->td_ucred, td);
183 			if (error)
184 				error = priv_check(td, PRIV_VFS_MOUNT_PERM);
185 			if (error) {
186 				VOP_UNLOCK(devvp, 0);
187 				return (error);
188 			}
189 			VOP_UNLOCK(devvp, 0);
190 			DROP_GIANT();
191 			g_topology_lock();
192 			error = g_access(ump->um_cp, 0, 1, 0);
193 			g_topology_unlock();
194 			PICKUP_GIANT();
195 			if (error)
196 				return (error);
197 
198 			if ((fs->e2fs->e2fs_state & E2FS_ISCLEAN) == 0 ||
199 			    (fs->e2fs->e2fs_state & E2FS_ERRORS)) {
200 				if (mp->mnt_flag & MNT_FORCE) {
201 					printf(
202 "WARNING: %s was not properly dismounted\n", fs->e2fs_fsmnt);
203 				} else {
204 					printf(
205 "WARNING: R/W mount of %s denied.  Filesystem is not clean - run fsck\n",
206 					    fs->e2fs_fsmnt);
207 					return (EPERM);
208 				}
209 			}
210 			fs->e2fs->e2fs_state &= ~E2FS_ISCLEAN;
211 			(void)ext2_cgupdate(ump, MNT_WAIT);
212 			fs->e2fs_ronly = 0;
213 			MNT_ILOCK(mp);
214 			mp->mnt_flag &= ~MNT_RDONLY;
215 			MNT_IUNLOCK(mp);
216 		}
217 		if (vfs_flagopt(opts, "export", NULL, 0)) {
218 			/* Process export requests in vfs_mount.c. */
219 			return (error);
220 		}
221 	}
222 
223 	/*
224 	 * Not an update, or updating the name: look up the name
225 	 * and verify that it refers to a sensible disk device.
226 	 */
227 	if (fspec == NULL)
228 		return (EINVAL);
229 	NDINIT(ndp, LOOKUP, FOLLOW | LOCKLEAF, UIO_SYSSPACE, fspec, td);
230 	if ((error = namei(ndp)) != 0)
231 		return (error);
232 	NDFREE(ndp, NDF_ONLY_PNBUF);
233 	devvp = ndp->ni_vp;
234 
235 	if (!vn_isdisk(devvp, &error)) {
236 		vput(devvp);
237 		return (error);
238 	}
239 
240 	/*
241 	 * If mount by non-root, then verify that user has necessary
242 	 * permissions on the device.
243 	 *
244 	 * XXXRW: VOP_ACCESS() enough?
245 	 */
246 	accmode = VREAD;
247 	if ((mp->mnt_flag & MNT_RDONLY) == 0)
248 		accmode |= VWRITE;
249 	error = VOP_ACCESS(devvp, accmode, td->td_ucred, td);
250 	if (error)
251 		error = priv_check(td, PRIV_VFS_MOUNT_PERM);
252 	if (error) {
253 		vput(devvp);
254 		return (error);
255 	}
256 
257 	if ((mp->mnt_flag & MNT_UPDATE) == 0) {
258 		error = ext2_mountfs(devvp, mp);
259 	} else {
260 		if (devvp != ump->um_devvp) {
261 			vput(devvp);
262 			return (EINVAL);	/* needs translation */
263 		} else
264 			vput(devvp);
265 	}
266 	if (error) {
267 		vrele(devvp);
268 		return (error);
269 	}
270 	ump = VFSTOEXT2(mp);
271 	fs = ump->um_e2fs;
272 
273 	/*
274 	 * Note that this strncpy() is ok because of a check at the start
275 	 * of ext2_mount().
276 	 */
277 	strncpy(fs->e2fs_fsmnt, path, MAXMNTLEN);
278 	fs->e2fs_fsmnt[MAXMNTLEN - 1] = '\0';
279 	vfs_mountedfrom(mp, fspec);
280 	return (0);
281 }
282 
283 static int
284 ext2_check_sb_compat(struct ext2fs *es, struct cdev *dev, int ronly)
285 {
286 
287 	if (es->e2fs_magic != E2FS_MAGIC) {
288 		printf("ext2fs: %s: wrong magic number %#x (expected %#x)\n",
289 		    devtoname(dev), es->e2fs_magic, E2FS_MAGIC);
290 		return (1);
291 	}
292 	if (es->e2fs_rev > E2FS_REV0) {
293 		if (es->e2fs_features_incompat & ~EXT2F_INCOMPAT_SUPP) {
294 			printf(
295 "WARNING: mount of %s denied due to unsupported optional features\n",
296 			    devtoname(dev));
297 			return (1);
298 		}
299 		if (!ronly &&
300 		    (es->e2fs_features_rocompat & ~EXT2F_ROCOMPAT_SUPP)) {
301 			printf("WARNING: R/W mount of %s denied due to "
302 			    "unsupported optional features\n", devtoname(dev));
303 			return (1);
304 		}
305 	}
306 	return (0);
307 }
308 
309 /*
310  * This computes the fields of the  ext2_sb_info structure from the
311  * data in the ext2_super_block structure read in.
312  */
313 static int
314 compute_sb_data(struct vnode *devvp, struct ext2fs *es,
315     struct m_ext2fs *fs)
316 {
317 	int db_count, error;
318 	int i;
319 	int logic_sb_block = 1;	/* XXX for now */
320 	struct buf *bp;
321 
322 	fs->e2fs_bsize = EXT2_MIN_BLOCK_SIZE << es->e2fs_log_bsize;
323 	fs->e2fs_bshift = EXT2_MIN_BLOCK_LOG_SIZE + es->e2fs_log_bsize;
324 	fs->e2fs_fsbtodb = es->e2fs_log_bsize + 1;
325 	fs->e2fs_qbmask = fs->e2fs_bsize - 1;
326 	fs->e2fs_blocksize_bits = es->e2fs_log_bsize + 10;
327 	fs->e2fs_fsize = EXT2_MIN_FRAG_SIZE << es->e2fs_log_fsize;
328 	if (fs->e2fs_fsize)
329 		fs->e2fs_fpb = fs->e2fs_bsize / fs->e2fs_fsize;
330 	fs->e2fs_bpg = es->e2fs_bpg;
331 	fs->e2fs_fpg = es->e2fs_fpg;
332 	fs->e2fs_ipg = es->e2fs_ipg;
333 	if (es->e2fs_rev == E2FS_REV0) {
334 		fs->e2fs_first_inode = EXT2_FIRSTINO;
335 		fs->e2fs_isize = E2FS_REV0_INODE_SIZE ;
336 	} else {
337 		fs->e2fs_first_inode = es->e2fs_first_ino;
338 		fs->e2fs_isize = es->e2fs_inode_size;
339 
340 		/*
341 		 * Simple sanity check for superblock inode size value.
342 		 */
343 		if (EXT2_INODE_SIZE(fs) < E2FS_REV0_INODE_SIZE ||
344 		    EXT2_INODE_SIZE(fs) > fs->e2fs_bsize ||
345 		    (fs->e2fs_isize & (fs->e2fs_isize - 1)) != 0) {
346 			printf("ext2fs: invalid inode size %d\n",
347 			    fs->e2fs_isize);
348 			return (EIO);
349 		}
350 	}
351 	/* Check for extra isize in big inodes. */
352 	if (EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT4F_ROCOMPAT_EXTRA_ISIZE) &&
353 	    EXT2_INODE_SIZE(fs) < sizeof(struct ext2fs_dinode)) {
354 		printf("ext2fs: no space for extra inode timestamps\n");
355 		return (EINVAL);
356 	}
357 
358 	fs->e2fs_ipb = fs->e2fs_bsize / EXT2_INODE_SIZE(fs);
359 	fs->e2fs_itpg = fs->e2fs_ipg /fs->e2fs_ipb;
360 	fs->e2fs_descpb = fs->e2fs_bsize / sizeof(struct ext2_gd);
361 	/* s_resuid / s_resgid ? */
362 	fs->e2fs_gcount = (es->e2fs_bcount - es->e2fs_first_dblock +
363 	    EXT2_BLOCKS_PER_GROUP(fs) - 1) / EXT2_BLOCKS_PER_GROUP(fs);
364 	db_count = (fs->e2fs_gcount + EXT2_DESC_PER_BLOCK(fs) - 1) /
365 	    EXT2_DESC_PER_BLOCK(fs);
366 	fs->e2fs_gdbcount = db_count;
367 	fs->e2fs_gd = malloc(db_count * fs->e2fs_bsize,
368 	    M_EXT2MNT, M_WAITOK);
369 	fs->e2fs_contigdirs = malloc(fs->e2fs_gcount *
370 	    sizeof(*fs->e2fs_contigdirs), M_EXT2MNT, M_WAITOK);
371 
372 	/*
373 	 * Adjust logic_sb_block.
374 	 * Godmar thinks: if the blocksize is greater than 1024, then
375 	 * the superblock is logically part of block zero.
376 	 */
377 	if(fs->e2fs_bsize > SBSIZE)
378 		logic_sb_block = 0;
379 	for (i = 0; i < db_count; i++) {
380 		error = bread(devvp ,
381 			 fsbtodb(fs, logic_sb_block + i + 1 ),
382 			fs->e2fs_bsize, NOCRED, &bp);
383 		if (error) {
384 			free(fs->e2fs_gd, M_EXT2MNT);
385 			brelse(bp);
386 			return (error);
387 		}
388 		e2fs_cgload((struct ext2_gd *)bp->b_data,
389 		    &fs->e2fs_gd[
390 			i * fs->e2fs_bsize / sizeof(struct ext2_gd)],
391 		    fs->e2fs_bsize);
392 		brelse(bp);
393 		bp = NULL;
394 	}
395 	fs->e2fs_total_dir = 0;
396 	for (i=0; i < fs->e2fs_gcount; i++){
397 		fs->e2fs_total_dir += fs->e2fs_gd[i].ext2bgd_ndirs;
398 		fs->e2fs_contigdirs[i] = 0;
399 	}
400 	if (es->e2fs_rev == E2FS_REV0 ||
401 	    !EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_LARGEFILE))
402 		fs->e2fs_maxfilesize = 0x7fffffff;
403 	else
404 		fs->e2fs_maxfilesize = 0x7fffffffffffffff;
405 	return (0);
406 }
407 
408 /*
409  * Reload all incore data for a filesystem (used after running fsck on
410  * the root filesystem and finding things to fix). The filesystem must
411  * be mounted read-only.
412  *
413  * Things to do to update the mount:
414  *	1) invalidate all cached meta-data.
415  *	2) re-read superblock from disk.
416  *	3) invalidate all cluster summary information.
417  *	4) invalidate all inactive vnodes.
418  *	5) invalidate all cached file data.
419  *	6) re-read inode data for all active vnodes.
420  * XXX we are missing some steps, in particular # 3, this has to be reviewed.
421  */
422 static int
423 ext2_reload(struct mount *mp, struct thread *td)
424 {
425 	struct vnode *vp, *mvp, *devvp;
426 	struct inode *ip;
427 	struct buf *bp;
428 	struct ext2fs *es;
429 	struct m_ext2fs *fs;
430 	struct csum *sump;
431 	int error, i;
432 	int32_t *lp;
433 
434 	if ((mp->mnt_flag & MNT_RDONLY) == 0)
435 		return (EINVAL);
436 	/*
437 	 * Step 1: invalidate all cached meta-data.
438 	 */
439 	devvp = VFSTOEXT2(mp)->um_devvp;
440 	vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
441 	if (vinvalbuf(devvp, 0, 0, 0) != 0)
442 		panic("ext2_reload: dirty1");
443 	VOP_UNLOCK(devvp, 0);
444 
445 	/*
446 	 * Step 2: re-read superblock from disk.
447 	 * constants have been adjusted for ext2
448 	 */
449 	if ((error = bread(devvp, SBLOCK, SBSIZE, NOCRED, &bp)) != 0)
450 		return (error);
451 	es = (struct ext2fs *)bp->b_data;
452 	if (ext2_check_sb_compat(es, devvp->v_rdev, 0) != 0) {
453 		brelse(bp);
454 		return (EIO);		/* XXX needs translation */
455 	}
456 	fs = VFSTOEXT2(mp)->um_e2fs;
457 	bcopy(bp->b_data, fs->e2fs, sizeof(struct ext2fs));
458 
459 	if((error = compute_sb_data(devvp, es, fs)) != 0) {
460 		brelse(bp);
461 		return (error);
462 	}
463 #ifdef UNKLAR
464 	if (fs->fs_sbsize < SBSIZE)
465 		bp->b_flags |= B_INVAL;
466 #endif
467 	brelse(bp);
468 
469 	/*
470 	 * Step 3: invalidate all cluster summary information.
471 	 */
472 	if (fs->e2fs_contigsumsize > 0) {
473 		lp = fs->e2fs_maxcluster;
474 		sump = fs->e2fs_clustersum;
475 		for (i = 0; i < fs->e2fs_gcount; i++, sump++) {
476 			*lp++ = fs->e2fs_contigsumsize;
477 			sump->cs_init = 0;
478 			bzero(sump->cs_sum, fs->e2fs_contigsumsize + 1);
479 		}
480 	}
481 
482 loop:
483 	MNT_VNODE_FOREACH_ALL(vp, mp, mvp) {
484 		/*
485 		 * Step 4: invalidate all cached file data.
486 		 */
487 		if (vget(vp, LK_EXCLUSIVE | LK_INTERLOCK, td)) {
488 			MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp);
489 			goto loop;
490 		}
491 		if (vinvalbuf(vp, 0, 0, 0))
492 			panic("ext2_reload: dirty2");
493 
494 		/*
495 		 * Step 5: re-read inode data for all active vnodes.
496 		 */
497 		ip = VTOI(vp);
498 		error = bread(devvp, fsbtodb(fs, ino_to_fsba(fs, ip->i_number)),
499 		    (int)fs->e2fs_bsize, NOCRED, &bp);
500 		if (error) {
501 			VOP_UNLOCK(vp, 0);
502 			vrele(vp);
503 			MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp);
504 			return (error);
505 		}
506 		ext2_ei2i((struct ext2fs_dinode *) ((char *)bp->b_data +
507 		    EXT2_INODE_SIZE(fs) * ino_to_fsbo(fs, ip->i_number)), ip);
508 		brelse(bp);
509 		VOP_UNLOCK(vp, 0);
510 		vrele(vp);
511 	}
512 	return (0);
513 }
514 
515 /*
516  * Common code for mount and mountroot.
517  */
518 static int
519 ext2_mountfs(struct vnode *devvp, struct mount *mp)
520 {
521 	struct ext2mount *ump;
522 	struct buf *bp;
523 	struct m_ext2fs *fs;
524 	struct ext2fs *es;
525 	struct cdev *dev = devvp->v_rdev;
526 	struct g_consumer *cp;
527 	struct bufobj *bo;
528 	struct csum *sump;
529 	int error;
530 	int ronly;
531 	int i, size;
532 	int32_t *lp;
533 
534 	ronly = vfs_flagopt(mp->mnt_optnew, "ro", NULL, 0);
535 	/* XXX: use VOP_ACESS to check FS perms */
536 	DROP_GIANT();
537 	g_topology_lock();
538 	error = g_vfs_open(devvp, &cp, "ext2fs", ronly ? 0 : 1);
539 	g_topology_unlock();
540 	PICKUP_GIANT();
541 	VOP_UNLOCK(devvp, 0);
542 	if (error)
543 		return (error);
544 
545 	/* XXX: should we check for some sectorsize or 512 instead? */
546 	if (((SBSIZE % cp->provider->sectorsize) != 0) ||
547 	    (SBSIZE < cp->provider->sectorsize)) {
548 		DROP_GIANT();
549 		g_topology_lock();
550 		g_vfs_close(cp);
551 		g_topology_unlock();
552 		PICKUP_GIANT();
553 		return (EINVAL);
554 	}
555 
556 	bo = &devvp->v_bufobj;
557 	bo->bo_private = cp;
558 	bo->bo_ops = g_vfs_bufops;
559 	if (devvp->v_rdev->si_iosize_max != 0)
560 		mp->mnt_iosize_max = devvp->v_rdev->si_iosize_max;
561 	if (mp->mnt_iosize_max > MAXPHYS)
562 		mp->mnt_iosize_max = MAXPHYS;
563 
564 	bp = NULL;
565 	ump = NULL;
566 	if ((error = bread(devvp, SBLOCK, SBSIZE, NOCRED, &bp)) != 0)
567 		goto out;
568 	es = (struct ext2fs *)bp->b_data;
569 	if (ext2_check_sb_compat(es, dev, ronly) != 0) {
570 		error = EINVAL;		/* XXX needs translation */
571 		goto out;
572 	}
573 	if ((es->e2fs_state & E2FS_ISCLEAN) == 0 ||
574 	    (es->e2fs_state & E2FS_ERRORS)) {
575 		if (ronly || (mp->mnt_flag & MNT_FORCE)) {
576 			printf(
577 "WARNING: Filesystem was not properly dismounted\n");
578 		} else {
579 			printf(
580 "WARNING: R/W mount denied.  Filesystem is not clean - run fsck\n");
581 			error = EPERM;
582 			goto out;
583 		}
584 	}
585 	ump = malloc(sizeof(*ump), M_EXT2MNT, M_WAITOK | M_ZERO);
586 
587 	/*
588 	 * I don't know whether this is the right strategy. Note that
589 	 * we dynamically allocate both an ext2_sb_info and an ext2_super_block
590 	 * while Linux keeps the super block in a locked buffer.
591 	 */
592 	ump->um_e2fs = malloc(sizeof(struct m_ext2fs),
593 		M_EXT2MNT, M_WAITOK);
594 	ump->um_e2fs->e2fs = malloc(sizeof(struct ext2fs),
595 		M_EXT2MNT, M_WAITOK);
596 	mtx_init(EXT2_MTX(ump), "EXT2FS", "EXT2FS Lock", MTX_DEF);
597 	bcopy(es, ump->um_e2fs->e2fs, (u_int)sizeof(struct ext2fs));
598 	if ((error = compute_sb_data(devvp, ump->um_e2fs->e2fs, ump->um_e2fs)))
599 		goto out;
600 
601 	/*
602 	 * Calculate the maximum contiguous blocks and size of cluster summary
603 	 * array.  In FFS this is done by newfs; however, the superblock
604 	 * in ext2fs doesn't have these variables, so we can calculate
605 	 * them here.
606 	 */
607 	ump->um_e2fs->e2fs_maxcontig = MAX(1, MAXPHYS / ump->um_e2fs->e2fs_bsize);
608 	if (ump->um_e2fs->e2fs_maxcontig > 0)
609 		ump->um_e2fs->e2fs_contigsumsize =
610 		    MIN(ump->um_e2fs->e2fs_maxcontig, EXT2_MAXCONTIG);
611 	else
612 		ump->um_e2fs->e2fs_contigsumsize = 0;
613 	if (ump->um_e2fs->e2fs_contigsumsize > 0) {
614 		size = ump->um_e2fs->e2fs_gcount * sizeof(int32_t);
615 		ump->um_e2fs->e2fs_maxcluster = malloc(size, M_EXT2MNT, M_WAITOK);
616 		size = ump->um_e2fs->e2fs_gcount * sizeof(struct csum);
617 		ump->um_e2fs->e2fs_clustersum = malloc(size, M_EXT2MNT, M_WAITOK);
618 		lp = ump->um_e2fs->e2fs_maxcluster;
619 		sump = ump->um_e2fs->e2fs_clustersum;
620 		for (i = 0; i < ump->um_e2fs->e2fs_gcount; i++, sump++) {
621 			*lp++ = ump->um_e2fs->e2fs_contigsumsize;
622 			sump->cs_init = 0;
623 			sump->cs_sum = malloc((ump->um_e2fs->e2fs_contigsumsize + 1) *
624 			    sizeof(int32_t), M_EXT2MNT, M_WAITOK | M_ZERO);
625 		}
626 	}
627 
628 	brelse(bp);
629 	bp = NULL;
630 	fs = ump->um_e2fs;
631 	fs->e2fs_ronly = ronly;	/* ronly is set according to mnt_flags */
632 
633 	/*
634 	 * If the fs is not mounted read-only, make sure the super block is
635 	 * always written back on a sync().
636 	 */
637 	fs->e2fs_wasvalid = fs->e2fs->e2fs_state & E2FS_ISCLEAN ? 1 : 0;
638 	if (ronly == 0) {
639 		fs->e2fs_fmod = 1;		/* mark it modified */
640 		fs->e2fs->e2fs_state &= ~E2FS_ISCLEAN;	/* set fs invalid */
641 	}
642 	mp->mnt_data = ump;
643 	mp->mnt_stat.f_fsid.val[0] = dev2udev(dev);
644 	mp->mnt_stat.f_fsid.val[1] = mp->mnt_vfc->vfc_typenum;
645 	mp->mnt_maxsymlinklen = EXT2_MAXSYMLINKLEN;
646 	MNT_ILOCK(mp);
647 	mp->mnt_flag |= MNT_LOCAL;
648 	MNT_IUNLOCK(mp);
649 	ump->um_mountp = mp;
650 	ump->um_dev = dev;
651 	ump->um_devvp = devvp;
652 	ump->um_bo = &devvp->v_bufobj;
653 	ump->um_cp = cp;
654 
655 	/*
656 	 * Setting those two parameters allowed us to use
657 	 * ufs_bmap w/o changse!
658 	 */
659 	ump->um_nindir = EXT2_ADDR_PER_BLOCK(fs);
660 	ump->um_bptrtodb = fs->e2fs->e2fs_log_bsize + 1;
661 	ump->um_seqinc = EXT2_FRAGS_PER_BLOCK(fs);
662 	if (ronly == 0)
663 		ext2_sbupdate(ump, MNT_WAIT);
664 	/*
665 	 * Initialize filesystem stat information in mount struct.
666 	 */
667 	MNT_ILOCK(mp);
668  	mp->mnt_kern_flag |= MNTK_MPSAFE | MNTK_LOOKUP_SHARED |
669             MNTK_EXTENDED_SHARED;
670 	MNT_IUNLOCK(mp);
671 	return (0);
672 out:
673 	if (bp)
674 		brelse(bp);
675 	if (cp != NULL) {
676 		DROP_GIANT();
677 		g_topology_lock();
678 		g_vfs_close(cp);
679 		g_topology_unlock();
680 		PICKUP_GIANT();
681 	}
682 	if (ump) {
683 	  	mtx_destroy(EXT2_MTX(ump));
684 		free(ump->um_e2fs->e2fs_gd, M_EXT2MNT);
685 		free(ump->um_e2fs->e2fs_contigdirs, M_EXT2MNT);
686 		free(ump->um_e2fs->e2fs, M_EXT2MNT);
687 		free(ump->um_e2fs, M_EXT2MNT);
688 		free(ump, M_EXT2MNT);
689 		mp->mnt_data = NULL;
690 	}
691 	return (error);
692 }
693 
694 /*
695  * Unmount system call.
696  */
697 static int
698 ext2_unmount(struct mount *mp, int mntflags)
699 {
700 	struct ext2mount *ump;
701 	struct m_ext2fs *fs;
702 	struct csum *sump;
703 	int error, flags, i, ronly;
704 
705 	flags = 0;
706 	if (mntflags & MNT_FORCE) {
707 		if (mp->mnt_flag & MNT_ROOTFS)
708 			return (EINVAL);
709 		flags |= FORCECLOSE;
710 	}
711 	if ((error = ext2_flushfiles(mp, flags, curthread)) != 0)
712 		return (error);
713 	ump = VFSTOEXT2(mp);
714 	fs = ump->um_e2fs;
715 	ronly = fs->e2fs_ronly;
716 	if (ronly == 0 && ext2_cgupdate(ump, MNT_WAIT) == 0) {
717 		if (fs->e2fs_wasvalid)
718  			fs->e2fs->e2fs_state |= E2FS_ISCLEAN;
719  		ext2_sbupdate(ump, MNT_WAIT);
720 	}
721 
722 	DROP_GIANT();
723 	g_topology_lock();
724 	g_vfs_close(ump->um_cp);
725 	g_topology_unlock();
726 	PICKUP_GIANT();
727 	vrele(ump->um_devvp);
728 	sump = fs->e2fs_clustersum;
729 	for (i = 0; i < fs->e2fs_gcount; i++, sump++)
730 		free(sump->cs_sum, M_EXT2MNT);
731 	free(fs->e2fs_clustersum, M_EXT2MNT);
732 	free(fs->e2fs_maxcluster, M_EXT2MNT);
733 	free(fs->e2fs_gd, M_EXT2MNT);
734 	free(fs->e2fs_contigdirs, M_EXT2MNT);
735 	free(fs->e2fs, M_EXT2MNT);
736 	free(fs, M_EXT2MNT);
737 	free(ump, M_EXT2MNT);
738 	mp->mnt_data = NULL;
739 	MNT_ILOCK(mp);
740 	mp->mnt_flag &= ~MNT_LOCAL;
741 	MNT_IUNLOCK(mp);
742 	return (error);
743 }
744 
745 /*
746  * Flush out all the files in a filesystem.
747  */
748 static int
749 ext2_flushfiles(struct mount *mp, int flags, struct thread *td)
750 {
751 	int error;
752 
753 	error = vflush(mp, 0, flags, td);
754 	return (error);
755 }
756 /*
757  * Get file system statistics.
758  */
759 int
760 ext2_statfs(struct mount *mp, struct statfs *sbp)
761 {
762 	struct ext2mount *ump;
763 	struct m_ext2fs *fs;
764 	uint32_t overhead, overhead_per_group, ngdb;
765 	int i, ngroups;
766 
767 	ump = VFSTOEXT2(mp);
768 	fs = ump->um_e2fs;
769 	if (fs->e2fs->e2fs_magic != E2FS_MAGIC)
770 		panic("ext2fs_statvfs");
771 
772 	/*
773 	 * Compute the overhead (FS structures)
774 	 */
775 	overhead_per_group =
776 	    1 /* block bitmap */ +
777 	    1 /* inode bitmap */ +
778 	    fs->e2fs_itpg;
779 	overhead = fs->e2fs->e2fs_first_dblock +
780 	    fs->e2fs_gcount * overhead_per_group;
781 	if (fs->e2fs->e2fs_rev > E2FS_REV0 &&
782 	    fs->e2fs->e2fs_features_rocompat & EXT2F_ROCOMPAT_SPARSESUPER) {
783 		for (i = 0, ngroups = 0; i < fs->e2fs_gcount; i++) {
784 			if (cg_has_sb(i))
785 				ngroups++;
786 		}
787 	} else {
788 		ngroups = fs->e2fs_gcount;
789 	}
790 	ngdb = fs->e2fs_gdbcount;
791 	if (fs->e2fs->e2fs_rev > E2FS_REV0 &&
792 	    fs->e2fs->e2fs_features_compat & EXT2F_COMPAT_RESIZE)
793 		ngdb += fs->e2fs->e2fs_reserved_ngdb;
794 	overhead += ngroups * (1 /* superblock */ + ngdb);
795 
796 	sbp->f_bsize = EXT2_FRAG_SIZE(fs);
797 	sbp->f_iosize = EXT2_BLOCK_SIZE(fs);
798 	sbp->f_blocks = fs->e2fs->e2fs_bcount - overhead;
799 	sbp->f_bfree = fs->e2fs->e2fs_fbcount;
800 	sbp->f_bavail = sbp->f_bfree - fs->e2fs->e2fs_rbcount;
801 	sbp->f_files = fs->e2fs->e2fs_icount;
802 	sbp->f_ffree = fs->e2fs->e2fs_ficount;
803 	return (0);
804 }
805 
806 /*
807  * Go through the disk queues to initiate sandbagged IO;
808  * go through the inodes to write those that have been modified;
809  * initiate the writing of the super block if it has been modified.
810  *
811  * Note: we are always called with the filesystem marked `MPBUSY'.
812  */
813 static int
814 ext2_sync(struct mount *mp, int waitfor)
815 {
816 	struct vnode *mvp, *vp;
817 	struct thread *td;
818 	struct inode *ip;
819 	struct ext2mount *ump = VFSTOEXT2(mp);
820 	struct m_ext2fs *fs;
821 	int error, allerror = 0;
822 
823 	td = curthread;
824 	fs = ump->um_e2fs;
825 	if (fs->e2fs_fmod != 0 && fs->e2fs_ronly != 0) {		/* XXX */
826 		printf("fs = %s\n", fs->e2fs_fsmnt);
827 		panic("ext2_sync: rofs mod");
828 	}
829 
830 	/*
831 	 * Write back each (modified) inode.
832 	 */
833 loop:
834 	MNT_VNODE_FOREACH_ALL(vp, mp, mvp) {
835 		if (vp->v_type == VNON) {
836 			VI_UNLOCK(vp);
837 			continue;
838 		}
839 		ip = VTOI(vp);
840 		if ((ip->i_flag &
841 		    (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 &&
842 		    (vp->v_bufobj.bo_dirty.bv_cnt == 0 ||
843 		    waitfor == MNT_LAZY)) {
844 			VI_UNLOCK(vp);
845 			continue;
846 		}
847 		error = vget(vp, LK_EXCLUSIVE | LK_NOWAIT | LK_INTERLOCK, td);
848 		if (error) {
849 			if (error == ENOENT) {
850 				MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp);
851 				goto loop;
852 			}
853 			continue;
854 		}
855 		if ((error = VOP_FSYNC(vp, waitfor, td)) != 0)
856 			allerror = error;
857 		VOP_UNLOCK(vp, 0);
858 		vrele(vp);
859 	}
860 
861 	/*
862 	 * Force stale file system control information to be flushed.
863 	 */
864 	if (waitfor != MNT_LAZY) {
865 		vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
866 		if ((error = VOP_FSYNC(ump->um_devvp, waitfor, td)) != 0)
867 			allerror = error;
868 		VOP_UNLOCK(ump->um_devvp, 0);
869 	}
870 
871 	/*
872 	 * Write back modified superblock.
873 	 */
874 	if (fs->e2fs_fmod != 0) {
875 		fs->e2fs_fmod = 0;
876 		fs->e2fs->e2fs_wtime = time_second;
877 		if ((error = ext2_cgupdate(ump, waitfor)) != 0)
878 			allerror = error;
879 	}
880 	return (allerror);
881 }
882 
883 /*
884  * Look up an EXT2FS dinode number to find its incore vnode, otherwise read it
885  * in from disk.  If it is in core, wait for the lock bit to clear, then
886  * return the inode locked.  Detection and handling of mount points must be
887  * done by the calling routine.
888  */
889 static int
890 ext2_vget(struct mount *mp, ino_t ino, int flags, struct vnode **vpp)
891 {
892 	struct m_ext2fs *fs;
893 	struct inode *ip;
894 	struct ext2mount *ump;
895 	struct buf *bp;
896 	struct vnode *vp;
897 	struct cdev *dev;
898 	struct thread *td;
899 	int i, error;
900 	int used_blocks;
901 
902 	td = curthread;
903 	error = vfs_hash_get(mp, ino, flags, td, vpp, NULL, NULL);
904 	if (error || *vpp != NULL)
905 		return (error);
906 
907 	ump = VFSTOEXT2(mp);
908 	dev = ump->um_dev;
909 
910 	/*
911 	 * If this malloc() is performed after the getnewvnode()
912 	 * it might block, leaving a vnode with a NULL v_data to be
913 	 * found by ext2_sync() if a sync happens to fire right then,
914 	 * which will cause a panic because ext2_sync() blindly
915 	 * dereferences vp->v_data (as well it should).
916 	 */
917 	ip = malloc(sizeof(struct inode), M_EXT2NODE, M_WAITOK | M_ZERO);
918 
919 	/* Allocate a new vnode/inode. */
920 	if ((error = getnewvnode("ext2fs", mp, &ext2_vnodeops, &vp)) != 0) {
921 		*vpp = NULL;
922 		free(ip, M_EXT2NODE);
923 		return (error);
924 	}
925 	vp->v_data = ip;
926 	ip->i_vnode = vp;
927 	ip->i_e2fs = fs = ump->um_e2fs;
928 	ip->i_ump  = ump;
929 	ip->i_number = ino;
930 
931 	lockmgr(vp->v_vnlock, LK_EXCLUSIVE, NULL);
932 	error = insmntque(vp, mp);
933 	if (error != 0) {
934 		free(ip, M_EXT2NODE);
935 		*vpp = NULL;
936 		return (error);
937 	}
938 	error = vfs_hash_insert(vp, ino, flags, td, vpp, NULL, NULL);
939 	if (error || *vpp != NULL)
940 		return (error);
941 
942 	/* Read in the disk contents for the inode, copy into the inode. */
943 	if ((error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)),
944 	    (int)fs->e2fs_bsize, NOCRED, &bp)) != 0) {
945 		/*
946 		 * The inode does not contain anything useful, so it would
947 		 * be misleading to leave it on its hash chain. With mode
948 		 * still zero, it will be unlinked and returned to the free
949 		 * list by vput().
950 		 */
951 		brelse(bp);
952 		vput(vp);
953 		*vpp = NULL;
954 		return (error);
955 	}
956 	/* convert ext2 inode to dinode */
957 	ext2_ei2i((struct ext2fs_dinode *) ((char *)bp->b_data + EXT2_INODE_SIZE(fs) *
958 			ino_to_fsbo(fs, ino)), ip);
959 	ip->i_block_group = ino_to_cg(fs, ino);
960 	ip->i_next_alloc_block = 0;
961 	ip->i_next_alloc_goal = 0;
962 
963 	/*
964 	 * Now we want to make sure that block pointers for unused
965 	 * blocks are zeroed out - ext2_balloc depends on this
966 	 * although for regular files and directories only
967 	 */
968 	if(S_ISDIR(ip->i_mode) || S_ISREG(ip->i_mode)) {
969 		used_blocks = (ip->i_size+fs->e2fs_bsize-1) / fs->e2fs_bsize;
970 		for (i = used_blocks; i < EXT2_NDIR_BLOCKS; i++)
971 			ip->i_db[i] = 0;
972 	}
973 /*
974 	ext2_print_inode(ip);
975 */
976 	bqrelse(bp);
977 
978 	/*
979 	 * Initialize the vnode from the inode, check for aliases.
980 	 * Note that the underlying vnode may have changed.
981 	 */
982 	if ((error = ext2_vinit(mp, &ext2_fifoops, &vp)) != 0) {
983 		vput(vp);
984 		*vpp = NULL;
985 		return (error);
986 	}
987 
988 	/*
989 	 * Finish inode initialization.
990 	 */
991 
992 	/*
993 	 * Set up a generation number for this inode if it does not
994 	 * already have one. This should only happen on old filesystems.
995 	 */
996 	if (ip->i_gen == 0) {
997 		ip->i_gen = random() / 2 + 1;
998 		if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0)
999 			ip->i_flag |= IN_MODIFIED;
1000 	}
1001 	*vpp = vp;
1002 	return (0);
1003 }
1004 
1005 /*
1006  * File handle to vnode
1007  *
1008  * Have to be really careful about stale file handles:
1009  * - check that the inode number is valid
1010  * - call ext2_vget() to get the locked inode
1011  * - check for an unallocated inode (i_mode == 0)
1012  * - check that the given client host has export rights and return
1013  *   those rights via. exflagsp and credanonp
1014  */
1015 static int
1016 ext2_fhtovp(struct mount *mp, struct fid *fhp, int flags, struct vnode **vpp)
1017 {
1018 	struct inode *ip;
1019 	struct ufid *ufhp;
1020 	struct vnode *nvp;
1021 	struct m_ext2fs *fs;
1022 	int error;
1023 
1024 	ufhp = (struct ufid *)fhp;
1025 	fs = VFSTOEXT2(mp)->um_e2fs;
1026 	if (ufhp->ufid_ino < EXT2_ROOTINO ||
1027 	    ufhp->ufid_ino > fs->e2fs_gcount * fs->e2fs->e2fs_ipg)
1028 		return (ESTALE);
1029 
1030 	error = VFS_VGET(mp, ufhp->ufid_ino, LK_EXCLUSIVE, &nvp);
1031 	if (error) {
1032 		*vpp = NULLVP;
1033 		return (error);
1034 	}
1035 	ip = VTOI(nvp);
1036 	if (ip->i_mode == 0 ||
1037 	    ip->i_gen != ufhp->ufid_gen || ip->i_nlink <= 0) {
1038 		vput(nvp);
1039 		*vpp = NULLVP;
1040 		return (ESTALE);
1041 	}
1042 	*vpp = nvp;
1043 	vnode_create_vobject(*vpp, 0, curthread);
1044 	return (0);
1045 }
1046 
1047 /*
1048  * Write a superblock and associated information back to disk.
1049  */
1050 static int
1051 ext2_sbupdate(struct ext2mount *mp, int waitfor)
1052 {
1053 	struct m_ext2fs *fs = mp->um_e2fs;
1054 	struct ext2fs *es = fs->e2fs;
1055 	struct buf *bp;
1056 	int error = 0;
1057 
1058 	bp = getblk(mp->um_devvp, SBLOCK, SBSIZE, 0, 0, 0);
1059 	bcopy((caddr_t)es, bp->b_data, (u_int)sizeof(struct ext2fs));
1060 	if (waitfor == MNT_WAIT)
1061 		error = bwrite(bp);
1062 	else
1063 		bawrite(bp);
1064 
1065 	/*
1066 	 * The buffers for group descriptors, inode bitmaps and block bitmaps
1067 	 * are not busy at this point and are (hopefully) written by the
1068 	 * usual sync mechanism. No need to write them here.
1069 	 */
1070 	return (error);
1071 }
1072 int
1073 ext2_cgupdate(struct ext2mount *mp, int waitfor)
1074 {
1075 	struct m_ext2fs *fs = mp->um_e2fs;
1076 	struct buf *bp;
1077 	int i, error = 0, allerror = 0;
1078 
1079 	allerror = ext2_sbupdate(mp, waitfor);
1080 	for (i = 0; i < fs->e2fs_gdbcount; i++) {
1081 		bp = getblk(mp->um_devvp, fsbtodb(fs,
1082 		    fs->e2fs->e2fs_first_dblock +
1083 		    1 /* superblock */ + i), fs->e2fs_bsize, 0, 0, 0);
1084 		e2fs_cgsave(&fs->e2fs_gd[
1085 		    i * fs->e2fs_bsize / sizeof(struct ext2_gd)],
1086 		    (struct ext2_gd *)bp->b_data, fs->e2fs_bsize);
1087 		if (waitfor == MNT_WAIT)
1088 			error = bwrite(bp);
1089 		else
1090 			bawrite(bp);
1091 	}
1092 
1093 	if (!allerror && error)
1094 		allerror = error;
1095 	return (allerror);
1096 }
1097 /*
1098  * Return the root of a filesystem.
1099  */
1100 static int
1101 ext2_root(struct mount *mp, int flags, struct vnode **vpp)
1102 {
1103 	struct vnode *nvp;
1104 	int error;
1105 
1106 	error = VFS_VGET(mp, EXT2_ROOTINO, LK_EXCLUSIVE, &nvp);
1107 	if (error)
1108 		return (error);
1109 	*vpp = nvp;
1110 	return (0);
1111 }
1112