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: releng/10.0/sys/fs/ext2fs/ext2_vfsops.c 254260 2013-08-12 21:34:48Z pfg $
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 = NULL;
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)
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 uint32_t e2fs_descpb;
322
323 fs->e2fs_bshift = EXT2_MIN_BLOCK_LOG_SIZE + es->e2fs_log_bsize;
324 fs->e2fs_bsize = 1U << fs->e2fs_bshift;
325 fs->e2fs_fsbtodb = es->e2fs_log_bsize + 1;
326 fs->e2fs_qbmask = fs->e2fs_bsize - 1;
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_isize = E2FS_REV0_INODE_SIZE ;
335 } else {
336 fs->e2fs_isize = es->e2fs_inode_size;
337
338 /*
339 * Simple sanity check for superblock inode size value.
340 */
341 if (EXT2_INODE_SIZE(fs) < E2FS_REV0_INODE_SIZE ||
342 EXT2_INODE_SIZE(fs) > fs->e2fs_bsize ||
343 (fs->e2fs_isize & (fs->e2fs_isize - 1)) != 0) {
344 printf("ext2fs: invalid inode size %d\n",
345 fs->e2fs_isize);
346 return (EIO);
347 }
348 }
349 /* Check for extra isize in big inodes. */
350 if (EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_EXTRA_ISIZE) &&
351 EXT2_INODE_SIZE(fs) < sizeof(struct ext2fs_dinode)) {
352 printf("ext2fs: no space for extra inode timestamps\n");
353 return (EINVAL);
354 }
355
356 fs->e2fs_ipb = fs->e2fs_bsize / EXT2_INODE_SIZE(fs);
357 fs->e2fs_itpg = fs->e2fs_ipg /fs->e2fs_ipb;
358 /* s_resuid / s_resgid ? */
359 fs->e2fs_gcount = (es->e2fs_bcount - es->e2fs_first_dblock +
360 EXT2_BLOCKS_PER_GROUP(fs) - 1) / EXT2_BLOCKS_PER_GROUP(fs);
361 e2fs_descpb = fs->e2fs_bsize / sizeof(struct ext2_gd);
362 db_count = (fs->e2fs_gcount + e2fs_descpb - 1) / e2fs_descpb;
363 fs->e2fs_gdbcount = db_count;
364 fs->e2fs_gd = malloc(db_count * fs->e2fs_bsize,
365 M_EXT2MNT, M_WAITOK);
366 fs->e2fs_contigdirs = malloc(fs->e2fs_gcount *
367 sizeof(*fs->e2fs_contigdirs), M_EXT2MNT, M_WAITOK);
368
369 /*
370 * Adjust logic_sb_block.
371 * Godmar thinks: if the blocksize is greater than 1024, then
372 * the superblock is logically part of block zero.
373 */
374 if(fs->e2fs_bsize > SBSIZE)
375 logic_sb_block = 0;
376 for (i = 0; i < db_count; i++) {
377 error = bread(devvp ,
378 fsbtodb(fs, logic_sb_block + i + 1 ),
379 fs->e2fs_bsize, NOCRED, &bp);
380 if (error) {
381 free(fs->e2fs_gd, M_EXT2MNT);
382 brelse(bp);
383 return (error);
384 }
385 e2fs_cgload((struct ext2_gd *)bp->b_data,
386 &fs->e2fs_gd[
387 i * fs->e2fs_bsize / sizeof(struct ext2_gd)],
388 fs->e2fs_bsize);
389 brelse(bp);
390 bp = NULL;
391 }
392 fs->e2fs_total_dir = 0;
393 for (i=0; i < fs->e2fs_gcount; i++){
394 fs->e2fs_total_dir += fs->e2fs_gd[i].ext2bgd_ndirs;
395 fs->e2fs_contigdirs[i] = 0;
396 }
397 if (es->e2fs_rev == E2FS_REV0 ||
398 !EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_LARGEFILE))
399 fs->e2fs_maxfilesize = 0x7fffffff;
400 else {
401 fs->e2fs_maxfilesize = 0xffffffffffff;
402 if (EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_HUGE_FILE))
403 fs->e2fs_maxfilesize = 0x7fffffffffffffff;
404 }
405 if (es->e4fs_flags & E2FS_UNSIGNED_HASH) {
406 fs->e2fs_uhash = 3;
407 } else if ((es->e4fs_flags & E2FS_SIGNED_HASH) == 0) {
408 #ifdef __CHAR_UNSIGNED__
409 es->e4fs_flags |= E2FS_UNSIGNED_HASH;
410 fs->e2fs_uhash = 3;
411 #else
412 es->e4fs_flags |= E2FS_SIGNED_HASH;
413 #endif
414 }
415
416 return (0);
417 }
418
419 /*
420 * Reload all incore data for a filesystem (used after running fsck on
421 * the root filesystem and finding things to fix). The filesystem must
422 * be mounted read-only.
423 *
424 * Things to do to update the mount:
425 * 1) invalidate all cached meta-data.
426 * 2) re-read superblock from disk.
427 * 3) invalidate all cluster summary information.
428 * 4) invalidate all inactive vnodes.
429 * 5) invalidate all cached file data.
430 * 6) re-read inode data for all active vnodes.
431 * XXX we are missing some steps, in particular # 3, this has to be reviewed.
432 */
433 static int
434 ext2_reload(struct mount *mp, struct thread *td)
435 {
436 struct vnode *vp, *mvp, *devvp;
437 struct inode *ip;
438 struct buf *bp;
439 struct ext2fs *es;
440 struct m_ext2fs *fs;
441 struct csum *sump;
442 int error, i;
443 int32_t *lp;
444
445 if ((mp->mnt_flag & MNT_RDONLY) == 0)
446 return (EINVAL);
447 /*
448 * Step 1: invalidate all cached meta-data.
449 */
450 devvp = VFSTOEXT2(mp)->um_devvp;
451 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
452 if (vinvalbuf(devvp, 0, 0, 0) != 0)
453 panic("ext2_reload: dirty1");
454 VOP_UNLOCK(devvp, 0);
455
456 /*
457 * Step 2: re-read superblock from disk.
458 * constants have been adjusted for ext2
459 */
460 if ((error = bread(devvp, SBLOCK, SBSIZE, NOCRED, &bp)) != 0)
461 return (error);
462 es = (struct ext2fs *)bp->b_data;
463 if (ext2_check_sb_compat(es, devvp->v_rdev, 0) != 0) {
464 brelse(bp);
465 return (EIO); /* XXX needs translation */
466 }
467 fs = VFSTOEXT2(mp)->um_e2fs;
468 bcopy(bp->b_data, fs->e2fs, sizeof(struct ext2fs));
469
470 if((error = compute_sb_data(devvp, es, fs)) != 0) {
471 brelse(bp);
472 return (error);
473 }
474 #ifdef UNKLAR
475 if (fs->fs_sbsize < SBSIZE)
476 bp->b_flags |= B_INVAL;
477 #endif
478 brelse(bp);
479
480 /*
481 * Step 3: invalidate all cluster summary information.
482 */
483 if (fs->e2fs_contigsumsize > 0) {
484 lp = fs->e2fs_maxcluster;
485 sump = fs->e2fs_clustersum;
486 for (i = 0; i < fs->e2fs_gcount; i++, sump++) {
487 *lp++ = fs->e2fs_contigsumsize;
488 sump->cs_init = 0;
489 bzero(sump->cs_sum, fs->e2fs_contigsumsize + 1);
490 }
491 }
492
493 loop:
494 MNT_VNODE_FOREACH_ALL(vp, mp, mvp) {
495 /*
496 * Step 4: invalidate all cached file data.
497 */
498 if (vget(vp, LK_EXCLUSIVE | LK_INTERLOCK, td)) {
499 MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp);
500 goto loop;
501 }
502 if (vinvalbuf(vp, 0, 0, 0))
503 panic("ext2_reload: dirty2");
504
505 /*
506 * Step 5: re-read inode data for all active vnodes.
507 */
508 ip = VTOI(vp);
509 error = bread(devvp, fsbtodb(fs, ino_to_fsba(fs, ip->i_number)),
510 (int)fs->e2fs_bsize, NOCRED, &bp);
511 if (error) {
512 VOP_UNLOCK(vp, 0);
513 vrele(vp);
514 MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp);
515 return (error);
516 }
517 ext2_ei2i((struct ext2fs_dinode *) ((char *)bp->b_data +
518 EXT2_INODE_SIZE(fs) * ino_to_fsbo(fs, ip->i_number)), ip);
519 brelse(bp);
520 VOP_UNLOCK(vp, 0);
521 vrele(vp);
522 }
523 return (0);
524 }
525
526 /*
527 * Common code for mount and mountroot.
528 */
529 static int
530 ext2_mountfs(struct vnode *devvp, struct mount *mp)
531 {
532 struct ext2mount *ump;
533 struct buf *bp;
534 struct m_ext2fs *fs;
535 struct ext2fs *es;
536 struct cdev *dev = devvp->v_rdev;
537 struct g_consumer *cp;
538 struct bufobj *bo;
539 struct csum *sump;
540 int error;
541 int ronly;
542 int i, size;
543 int32_t *lp;
544 int32_t e2fs_maxcontig;
545
546 ronly = vfs_flagopt(mp->mnt_optnew, "ro", NULL, 0);
547 /* XXX: use VOP_ACESS to check FS perms */
548 DROP_GIANT();
549 g_topology_lock();
550 error = g_vfs_open(devvp, &cp, "ext2fs", ronly ? 0 : 1);
551 g_topology_unlock();
552 PICKUP_GIANT();
553 VOP_UNLOCK(devvp, 0);
554 if (error)
555 return (error);
556
557 /* XXX: should we check for some sectorsize or 512 instead? */
558 if (((SBSIZE % cp->provider->sectorsize) != 0) ||
559 (SBSIZE < cp->provider->sectorsize)) {
560 DROP_GIANT();
561 g_topology_lock();
562 g_vfs_close(cp);
563 g_topology_unlock();
564 PICKUP_GIANT();
565 return (EINVAL);
566 }
567
568 bo = &devvp->v_bufobj;
569 bo->bo_private = cp;
570 bo->bo_ops = g_vfs_bufops;
571 if (devvp->v_rdev->si_iosize_max != 0)
572 mp->mnt_iosize_max = devvp->v_rdev->si_iosize_max;
573 if (mp->mnt_iosize_max > MAXPHYS)
574 mp->mnt_iosize_max = MAXPHYS;
575
576 bp = NULL;
577 ump = NULL;
578 if ((error = bread(devvp, SBLOCK, SBSIZE, NOCRED, &bp)) != 0)
579 goto out;
580 es = (struct ext2fs *)bp->b_data;
581 if (ext2_check_sb_compat(es, dev, ronly) != 0) {
582 error = EINVAL; /* XXX needs translation */
583 goto out;
584 }
585 if ((es->e2fs_state & E2FS_ISCLEAN) == 0 ||
586 (es->e2fs_state & E2FS_ERRORS)) {
587 if (ronly || (mp->mnt_flag & MNT_FORCE)) {
588 printf(
589 "WARNING: Filesystem was not properly dismounted\n");
590 } else {
591 printf(
592 "WARNING: R/W mount denied. Filesystem is not clean - run fsck\n");
593 error = EPERM;
594 goto out;
595 }
596 }
597 ump = malloc(sizeof(*ump), M_EXT2MNT, M_WAITOK | M_ZERO);
598
599 /*
600 * I don't know whether this is the right strategy. Note that
601 * we dynamically allocate both an ext2_sb_info and an ext2_super_block
602 * while Linux keeps the super block in a locked buffer.
603 */
604 ump->um_e2fs = malloc(sizeof(struct m_ext2fs),
605 M_EXT2MNT, M_WAITOK);
606 ump->um_e2fs->e2fs = malloc(sizeof(struct ext2fs),
607 M_EXT2MNT, M_WAITOK);
608 mtx_init(EXT2_MTX(ump), "EXT2FS", "EXT2FS Lock", MTX_DEF);
609 bcopy(es, ump->um_e2fs->e2fs, (u_int)sizeof(struct ext2fs));
610 if ((error = compute_sb_data(devvp, ump->um_e2fs->e2fs, ump->um_e2fs)))
611 goto out;
612
613 /*
614 * Calculate the maximum contiguous blocks and size of cluster summary
615 * array. In FFS this is done by newfs; however, the superblock
616 * in ext2fs doesn't have these variables, so we can calculate
617 * them here.
618 */
619 e2fs_maxcontig = MAX(1, MAXPHYS / ump->um_e2fs->e2fs_bsize);
620 ump->um_e2fs->e2fs_contigsumsize = MIN(e2fs_maxcontig, EXT2_MAXCONTIG);
621 if (ump->um_e2fs->e2fs_contigsumsize > 0) {
622 size = ump->um_e2fs->e2fs_gcount * sizeof(int32_t);
623 ump->um_e2fs->e2fs_maxcluster = malloc(size, M_EXT2MNT, M_WAITOK);
624 size = ump->um_e2fs->e2fs_gcount * sizeof(struct csum);
625 ump->um_e2fs->e2fs_clustersum = malloc(size, M_EXT2MNT, M_WAITOK);
626 lp = ump->um_e2fs->e2fs_maxcluster;
627 sump = ump->um_e2fs->e2fs_clustersum;
628 for (i = 0; i < ump->um_e2fs->e2fs_gcount; i++, sump++) {
629 *lp++ = ump->um_e2fs->e2fs_contigsumsize;
630 sump->cs_init = 0;
631 sump->cs_sum = malloc((ump->um_e2fs->e2fs_contigsumsize + 1) *
632 sizeof(int32_t), M_EXT2MNT, M_WAITOK | M_ZERO);
633 }
634 }
635
636 brelse(bp);
637 bp = NULL;
638 fs = ump->um_e2fs;
639 fs->e2fs_ronly = ronly; /* ronly is set according to mnt_flags */
640
641 /*
642 * If the fs is not mounted read-only, make sure the super block is
643 * always written back on a sync().
644 */
645 fs->e2fs_wasvalid = fs->e2fs->e2fs_state & E2FS_ISCLEAN ? 1 : 0;
646 if (ronly == 0) {
647 fs->e2fs_fmod = 1; /* mark it modified */
648 fs->e2fs->e2fs_state &= ~E2FS_ISCLEAN; /* set fs invalid */
649 }
650 mp->mnt_data = ump;
651 mp->mnt_stat.f_fsid.val[0] = dev2udev(dev);
652 mp->mnt_stat.f_fsid.val[1] = mp->mnt_vfc->vfc_typenum;
653 mp->mnt_maxsymlinklen = EXT2_MAXSYMLINKLEN;
654 MNT_ILOCK(mp);
655 mp->mnt_flag |= MNT_LOCAL;
656 MNT_IUNLOCK(mp);
657 ump->um_mountp = mp;
658 ump->um_dev = dev;
659 ump->um_devvp = devvp;
660 ump->um_bo = &devvp->v_bufobj;
661 ump->um_cp = cp;
662
663 /*
664 * Setting those two parameters allowed us to use
665 * ufs_bmap w/o changse!
666 */
667 ump->um_nindir = EXT2_ADDR_PER_BLOCK(fs);
668 ump->um_bptrtodb = fs->e2fs->e2fs_log_bsize + 1;
669 ump->um_seqinc = EXT2_FRAGS_PER_BLOCK(fs);
670 if (ronly == 0)
671 ext2_sbupdate(ump, MNT_WAIT);
672 /*
673 * Initialize filesystem stat information in mount struct.
674 */
675 MNT_ILOCK(mp);
676 mp->mnt_kern_flag |= MNTK_LOOKUP_SHARED | MNTK_EXTENDED_SHARED;
677 MNT_IUNLOCK(mp);
678 return (0);
679 out:
680 if (bp)
681 brelse(bp);
682 if (cp != NULL) {
683 DROP_GIANT();
684 g_topology_lock();
685 g_vfs_close(cp);
686 g_topology_unlock();
687 PICKUP_GIANT();
688 }
689 if (ump) {
690 mtx_destroy(EXT2_MTX(ump));
691 free(ump->um_e2fs->e2fs_gd, M_EXT2MNT);
692 free(ump->um_e2fs->e2fs_contigdirs, M_EXT2MNT);
693 free(ump->um_e2fs->e2fs, M_EXT2MNT);
694 free(ump->um_e2fs, M_EXT2MNT);
695 free(ump, M_EXT2MNT);
696 mp->mnt_data = NULL;
697 }
698 return (error);
699 }
700
701 /*
702 * Unmount system call.
703 */
704 static int
705 ext2_unmount(struct mount *mp, int mntflags)
706 {
707 struct ext2mount *ump;
708 struct m_ext2fs *fs;
709 struct csum *sump;
710 int error, flags, i, ronly;
711
712 flags = 0;
713 if (mntflags & MNT_FORCE) {
714 if (mp->mnt_flag & MNT_ROOTFS)
715 return (EINVAL);
716 flags |= FORCECLOSE;
717 }
718 if ((error = ext2_flushfiles(mp, flags, curthread)) != 0)
719 return (error);
720 ump = VFSTOEXT2(mp);
721 fs = ump->um_e2fs;
722 ronly = fs->e2fs_ronly;
723 if (ronly == 0 && ext2_cgupdate(ump, MNT_WAIT) == 0) {
724 if (fs->e2fs_wasvalid)
725 fs->e2fs->e2fs_state |= E2FS_ISCLEAN;
726 ext2_sbupdate(ump, MNT_WAIT);
727 }
728
729 DROP_GIANT();
730 g_topology_lock();
731 g_vfs_close(ump->um_cp);
732 g_topology_unlock();
733 PICKUP_GIANT();
734 vrele(ump->um_devvp);
735 sump = fs->e2fs_clustersum;
736 for (i = 0; i < fs->e2fs_gcount; i++, sump++)
737 free(sump->cs_sum, M_EXT2MNT);
738 free(fs->e2fs_clustersum, M_EXT2MNT);
739 free(fs->e2fs_maxcluster, M_EXT2MNT);
740 free(fs->e2fs_gd, M_EXT2MNT);
741 free(fs->e2fs_contigdirs, M_EXT2MNT);
742 free(fs->e2fs, M_EXT2MNT);
743 free(fs, M_EXT2MNT);
744 free(ump, M_EXT2MNT);
745 mp->mnt_data = NULL;
746 MNT_ILOCK(mp);
747 mp->mnt_flag &= ~MNT_LOCAL;
748 MNT_IUNLOCK(mp);
749 return (error);
750 }
751
752 /*
753 * Flush out all the files in a filesystem.
754 */
755 static int
756 ext2_flushfiles(struct mount *mp, int flags, struct thread *td)
757 {
758 int error;
759
760 error = vflush(mp, 0, flags, td);
761 return (error);
762 }
763 /*
764 * Get filesystem statistics.
765 */
766 int
767 ext2_statfs(struct mount *mp, struct statfs *sbp)
768 {
769 struct ext2mount *ump;
770 struct m_ext2fs *fs;
771 uint32_t overhead, overhead_per_group, ngdb;
772 int i, ngroups;
773
774 ump = VFSTOEXT2(mp);
775 fs = ump->um_e2fs;
776 if (fs->e2fs->e2fs_magic != E2FS_MAGIC)
777 panic("ext2_statfs");
778
779 /*
780 * Compute the overhead (FS structures)
781 */
782 overhead_per_group =
783 1 /* block bitmap */ +
784 1 /* inode bitmap */ +
785 fs->e2fs_itpg;
786 overhead = fs->e2fs->e2fs_first_dblock +
787 fs->e2fs_gcount * overhead_per_group;
788 if (fs->e2fs->e2fs_rev > E2FS_REV0 &&
789 fs->e2fs->e2fs_features_rocompat & EXT2F_ROCOMPAT_SPARSESUPER) {
790 for (i = 0, ngroups = 0; i < fs->e2fs_gcount; i++) {
791 if (cg_has_sb(i))
792 ngroups++;
793 }
794 } else {
795 ngroups = fs->e2fs_gcount;
796 }
797 ngdb = fs->e2fs_gdbcount;
798 if (fs->e2fs->e2fs_rev > E2FS_REV0 &&
799 fs->e2fs->e2fs_features_compat & EXT2F_COMPAT_RESIZE)
800 ngdb += fs->e2fs->e2fs_reserved_ngdb;
801 overhead += ngroups * (1 /* superblock */ + ngdb);
802
803 sbp->f_bsize = EXT2_FRAG_SIZE(fs);
804 sbp->f_iosize = EXT2_BLOCK_SIZE(fs);
805 sbp->f_blocks = fs->e2fs->e2fs_bcount - overhead;
806 sbp->f_bfree = fs->e2fs->e2fs_fbcount;
807 sbp->f_bavail = sbp->f_bfree - fs->e2fs->e2fs_rbcount;
808 sbp->f_files = fs->e2fs->e2fs_icount;
809 sbp->f_ffree = fs->e2fs->e2fs_ficount;
810 return (0);
811 }
812
813 /*
814 * Go through the disk queues to initiate sandbagged IO;
815 * go through the inodes to write those that have been modified;
816 * initiate the writing of the super block if it has been modified.
817 *
818 * Note: we are always called with the filesystem marked `MPBUSY'.
819 */
820 static int
821 ext2_sync(struct mount *mp, int waitfor)
822 {
823 struct vnode *mvp, *vp;
824 struct thread *td;
825 struct inode *ip;
826 struct ext2mount *ump = VFSTOEXT2(mp);
827 struct m_ext2fs *fs;
828 int error, allerror = 0;
829
830 td = curthread;
831 fs = ump->um_e2fs;
832 if (fs->e2fs_fmod != 0 && fs->e2fs_ronly != 0) { /* XXX */
833 printf("fs = %s\n", fs->e2fs_fsmnt);
834 panic("ext2_sync: rofs mod");
835 }
836
837 /*
838 * Write back each (modified) inode.
839 */
840 loop:
841 MNT_VNODE_FOREACH_ALL(vp, mp, mvp) {
842 if (vp->v_type == VNON) {
843 VI_UNLOCK(vp);
844 continue;
845 }
846 ip = VTOI(vp);
847 if ((ip->i_flag &
848 (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 &&
849 (vp->v_bufobj.bo_dirty.bv_cnt == 0 ||
850 waitfor == MNT_LAZY)) {
851 VI_UNLOCK(vp);
852 continue;
853 }
854 error = vget(vp, LK_EXCLUSIVE | LK_NOWAIT | LK_INTERLOCK, td);
855 if (error) {
856 if (error == ENOENT) {
857 MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp);
858 goto loop;
859 }
860 continue;
861 }
862 if ((error = VOP_FSYNC(vp, waitfor, td)) != 0)
863 allerror = error;
864 VOP_UNLOCK(vp, 0);
865 vrele(vp);
866 }
867
868 /*
869 * Force stale filesystem control information to be flushed.
870 */
871 if (waitfor != MNT_LAZY) {
872 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
873 if ((error = VOP_FSYNC(ump->um_devvp, waitfor, td)) != 0)
874 allerror = error;
875 VOP_UNLOCK(ump->um_devvp, 0);
876 }
877
878 /*
879 * Write back modified superblock.
880 */
881 if (fs->e2fs_fmod != 0) {
882 fs->e2fs_fmod = 0;
883 fs->e2fs->e2fs_wtime = time_second;
884 if ((error = ext2_cgupdate(ump, waitfor)) != 0)
885 allerror = error;
886 }
887 return (allerror);
888 }
889
890 /*
891 * Look up an EXT2FS dinode number to find its incore vnode, otherwise read it
892 * in from disk. If it is in core, wait for the lock bit to clear, then
893 * return the inode locked. Detection and handling of mount points must be
894 * done by the calling routine.
895 */
896 static int
897 ext2_vget(struct mount *mp, ino_t ino, int flags, struct vnode **vpp)
898 {
899 struct m_ext2fs *fs;
900 struct inode *ip;
901 struct ext2mount *ump;
902 struct buf *bp;
903 struct vnode *vp;
904 struct cdev *dev;
905 struct thread *td;
906 int i, error;
907 int used_blocks;
908
909 td = curthread;
910 error = vfs_hash_get(mp, ino, flags, td, vpp, NULL, NULL);
911 if (error || *vpp != NULL)
912 return (error);
913
914 ump = VFSTOEXT2(mp);
915 dev = ump->um_dev;
916 ip = malloc(sizeof(struct inode), M_EXT2NODE, M_WAITOK | M_ZERO);
917
918 /* Allocate a new vnode/inode. */
919 if ((error = getnewvnode("ext2fs", mp, &ext2_vnodeops, &vp)) != 0) {
920 *vpp = NULL;
921 free(ip, M_EXT2NODE);
922 return (error);
923 }
924 vp->v_data = ip;
925 ip->i_vnode = vp;
926 ip->i_e2fs = fs = ump->um_e2fs;
927 ip->i_ump = ump;
928 ip->i_number = ino;
929
930 lockmgr(vp->v_vnlock, LK_EXCLUSIVE, NULL);
931 error = insmntque(vp, mp);
932 if (error != 0) {
933 free(ip, M_EXT2NODE);
934 *vpp = NULL;
935 return (error);
936 }
937 error = vfs_hash_insert(vp, ino, flags, td, vpp, NULL, NULL);
938 if (error || *vpp != NULL)
939 return (error);
940
941 /* Read in the disk contents for the inode, copy into the inode. */
942 if ((error = bread(ump->um_devvp, fsbtodb(fs, ino_to_fsba(fs, ino)),
943 (int)fs->e2fs_bsize, NOCRED, &bp)) != 0) {
944 /*
945 * The inode does not contain anything useful, so it would
946 * be misleading to leave it on its hash chain. With mode
947 * still zero, it will be unlinked and returned to the free
948 * list by vput().
949 */
950 brelse(bp);
951 vput(vp);
952 *vpp = NULL;
953 return (error);
954 }
955 /* convert ext2 inode to dinode */
956 ext2_ei2i((struct ext2fs_dinode *) ((char *)bp->b_data + EXT2_INODE_SIZE(fs) *
957 ino_to_fsbo(fs, ino)), ip);
958 ip->i_block_group = ino_to_cg(fs, ino);
959 ip->i_next_alloc_block = 0;
960 ip->i_next_alloc_goal = 0;
961
962 /*
963 * Now we want to make sure that block pointers for unused
964 * blocks are zeroed out - ext2_balloc depends on this
965 * although for regular files and directories only
966 *
967 * If EXT4_EXTENTS flag is enabled, unused blocks aren't
968 * zeroed out because we could corrupt the extent tree.
969 */
970 if (!(ip->i_flags & EXT4_EXTENTS) &&
971 (S_ISDIR(ip->i_mode) || S_ISREG(ip->i_mode))) {
972 used_blocks = (ip->i_size+fs->e2fs_bsize-1) / fs->e2fs_bsize;
973 for (i = used_blocks; i < EXT2_NDIR_BLOCKS; i++)
974 ip->i_db[i] = 0;
975 }
976 /*
977 ext2_print_inode(ip);
978 */
979 bqrelse(bp);
980
981 /*
982 * Initialize the vnode from the inode, check for aliases.
983 * Note that the underlying vnode may have changed.
984 */
985 if ((error = ext2_vinit(mp, &ext2_fifoops, &vp)) != 0) {
986 vput(vp);
987 *vpp = NULL;
988 return (error);
989 }
990
991 /*
992 * Finish inode initialization.
993 */
994
995 /*
996 * Set up a generation number for this inode if it does not
997 * already have one. This should only happen on old filesystems.
998 */
999 if (ip->i_gen == 0) {
1000 ip->i_gen = random() + 1;
1001 if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0)
1002 ip->i_flag |= IN_MODIFIED;
1003 }
1004 *vpp = vp;
1005 return (0);
1006 }
1007
1008 /*
1009 * File handle to vnode
1010 *
1011 * Have to be really careful about stale file handles:
1012 * - check that the inode number is valid
1013 * - call ext2_vget() to get the locked inode
1014 * - check for an unallocated inode (i_mode == 0)
1015 * - check that the given client host has export rights and return
1016 * those rights via. exflagsp and credanonp
1017 */
1018 static int
1019 ext2_fhtovp(struct mount *mp, struct fid *fhp, int flags, struct vnode **vpp)
1020 {
1021 struct inode *ip;
1022 struct ufid *ufhp;
1023 struct vnode *nvp;
1024 struct m_ext2fs *fs;
1025 int error;
1026
1027 ufhp = (struct ufid *)fhp;
1028 fs = VFSTOEXT2(mp)->um_e2fs;
1029 if (ufhp->ufid_ino < EXT2_ROOTINO ||
1030 ufhp->ufid_ino > fs->e2fs_gcount * fs->e2fs->e2fs_ipg)
1031 return (ESTALE);
1032
1033 error = VFS_VGET(mp, ufhp->ufid_ino, LK_EXCLUSIVE, &nvp);
1034 if (error) {
1035 *vpp = NULLVP;
1036 return (error);
1037 }
1038 ip = VTOI(nvp);
1039 if (ip->i_mode == 0 ||
1040 ip->i_gen != ufhp->ufid_gen || ip->i_nlink <= 0) {
1041 vput(nvp);
1042 *vpp = NULLVP;
1043 return (ESTALE);
1044 }
1045 *vpp = nvp;
1046 vnode_create_vobject(*vpp, 0, curthread);
1047 return (0);
1048 }
1049
1050 /*
1051 * Write a superblock and associated information back to disk.
1052 */
1053 static int
1054 ext2_sbupdate(struct ext2mount *mp, int waitfor)
1055 {
1056 struct m_ext2fs *fs = mp->um_e2fs;
1057 struct ext2fs *es = fs->e2fs;
1058 struct buf *bp;
1059 int error = 0;
1060
1061 bp = getblk(mp->um_devvp, SBLOCK, SBSIZE, 0, 0, 0);
1062 bcopy((caddr_t)es, bp->b_data, (u_int)sizeof(struct ext2fs));
1063 if (waitfor == MNT_WAIT)
1064 error = bwrite(bp);
1065 else
1066 bawrite(bp);
1067
1068 /*
1069 * The buffers for group descriptors, inode bitmaps and block bitmaps
1070 * are not busy at this point and are (hopefully) written by the
1071 * usual sync mechanism. No need to write them here.
1072 */
1073 return (error);
1074 }
1075 int
1076 ext2_cgupdate(struct ext2mount *mp, int waitfor)
1077 {
1078 struct m_ext2fs *fs = mp->um_e2fs;
1079 struct buf *bp;
1080 int i, error = 0, allerror = 0;
1081
1082 allerror = ext2_sbupdate(mp, waitfor);
1083 for (i = 0; i < fs->e2fs_gdbcount; i++) {
1084 bp = getblk(mp->um_devvp, fsbtodb(fs,
1085 fs->e2fs->e2fs_first_dblock +
1086 1 /* superblock */ + i), fs->e2fs_bsize, 0, 0, 0);
1087 e2fs_cgsave(&fs->e2fs_gd[
1088 i * fs->e2fs_bsize / sizeof(struct ext2_gd)],
1089 (struct ext2_gd *)bp->b_data, fs->e2fs_bsize);
1090 if (waitfor == MNT_WAIT)
1091 error = bwrite(bp);
1092 else
1093 bawrite(bp);
1094 }
1095
1096 if (!allerror && error)
1097 allerror = error;
1098 return (allerror);
1099 }
1100 /*
1101 * Return the root of a filesystem.
1102 */
1103 static int
1104 ext2_root(struct mount *mp, int flags, struct vnode **vpp)
1105 {
1106 struct vnode *nvp;
1107 int error;
1108
1109 error = VFS_VGET(mp, EXT2_ROOTINO, LK_EXCLUSIVE, &nvp);
1110 if (error)
1111 return (error);
1112 *vpp = nvp;
1113 return (0);
1114 }
Cache object: 661450fe2d5a82f469e681b1470c939e
|