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/10.1/sys/fs/ext2fs/ext2_vfsops.c 262723 2014-03-04 03:10:31Z 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 EXT4F_RO_INCOMPAT_SUPP)) {
295 printf(
296 "WARNING: mount of %s denied due to unsupported optional features\n",
297 devtoname(dev));
298 return (1);
299 }
300 if (!ronly &&
301 (es->e2fs_features_rocompat & ~EXT2F_ROCOMPAT_SUPP)) {
302 printf("WARNING: R/W mount of %s denied due to "
303 "unsupported optional features\n", devtoname(dev));
304 return (1);
305 }
306 }
307 return (0);
308 }
309
310 /*
311 * This computes the fields of the ext2_sb_info structure from the
312 * data in the ext2_super_block structure read in.
313 */
314 static int
315 compute_sb_data(struct vnode *devvp, struct ext2fs *es,
316 struct m_ext2fs *fs)
317 {
318 int db_count, error;
319 int i;
320 int logic_sb_block = 1; /* XXX for now */
321 struct buf *bp;
322 uint32_t e2fs_descpb;
323
324 fs->e2fs_bshift = EXT2_MIN_BLOCK_LOG_SIZE + es->e2fs_log_bsize;
325 fs->e2fs_bsize = 1U << fs->e2fs_bshift;
326 fs->e2fs_fsbtodb = es->e2fs_log_bsize + 1;
327 fs->e2fs_qbmask = fs->e2fs_bsize - 1;
328 fs->e2fs_fsize = EXT2_MIN_FRAG_SIZE << es->e2fs_log_fsize;
329 if (fs->e2fs_fsize)
330 fs->e2fs_fpb = fs->e2fs_bsize / fs->e2fs_fsize;
331 fs->e2fs_bpg = es->e2fs_bpg;
332 fs->e2fs_fpg = es->e2fs_fpg;
333 fs->e2fs_ipg = es->e2fs_ipg;
334 if (es->e2fs_rev == E2FS_REV0) {
335 fs->e2fs_isize = E2FS_REV0_INODE_SIZE ;
336 } else {
337 fs->e2fs_isize = es->e2fs_inode_size;
338
339 /*
340 * Simple sanity check for superblock inode size value.
341 */
342 if (EXT2_INODE_SIZE(fs) < E2FS_REV0_INODE_SIZE ||
343 EXT2_INODE_SIZE(fs) > fs->e2fs_bsize ||
344 (fs->e2fs_isize & (fs->e2fs_isize - 1)) != 0) {
345 printf("ext2fs: invalid inode size %d\n",
346 fs->e2fs_isize);
347 return (EIO);
348 }
349 }
350 /* Check for extra isize in big inodes. */
351 if (EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_EXTRA_ISIZE) &&
352 EXT2_INODE_SIZE(fs) < sizeof(struct ext2fs_dinode)) {
353 printf("ext2fs: no space for extra inode timestamps\n");
354 return (EINVAL);
355 }
356
357 fs->e2fs_ipb = fs->e2fs_bsize / EXT2_INODE_SIZE(fs);
358 fs->e2fs_itpg = fs->e2fs_ipg /fs->e2fs_ipb;
359 /* s_resuid / s_resgid ? */
360 fs->e2fs_gcount = (es->e2fs_bcount - es->e2fs_first_dblock +
361 EXT2_BLOCKS_PER_GROUP(fs) - 1) / EXT2_BLOCKS_PER_GROUP(fs);
362 e2fs_descpb = fs->e2fs_bsize / sizeof(struct ext2_gd);
363 db_count = (fs->e2fs_gcount + e2fs_descpb - 1) / e2fs_descpb;
364 fs->e2fs_gdbcount = db_count;
365 fs->e2fs_gd = malloc(db_count * fs->e2fs_bsize,
366 M_EXT2MNT, M_WAITOK);
367 fs->e2fs_contigdirs = malloc(fs->e2fs_gcount *
368 sizeof(*fs->e2fs_contigdirs), M_EXT2MNT, M_WAITOK);
369
370 /*
371 * Adjust logic_sb_block.
372 * Godmar thinks: if the blocksize is greater than 1024, then
373 * the superblock is logically part of block zero.
374 */
375 if(fs->e2fs_bsize > SBSIZE)
376 logic_sb_block = 0;
377 for (i = 0; i < db_count; i++) {
378 error = bread(devvp ,
379 fsbtodb(fs, logic_sb_block + i + 1 ),
380 fs->e2fs_bsize, NOCRED, &bp);
381 if (error) {
382 free(fs->e2fs_gd, M_EXT2MNT);
383 brelse(bp);
384 return (error);
385 }
386 e2fs_cgload((struct ext2_gd *)bp->b_data,
387 &fs->e2fs_gd[
388 i * fs->e2fs_bsize / sizeof(struct ext2_gd)],
389 fs->e2fs_bsize);
390 brelse(bp);
391 bp = NULL;
392 }
393 fs->e2fs_total_dir = 0;
394 for (i=0; i < fs->e2fs_gcount; i++){
395 fs->e2fs_total_dir += fs->e2fs_gd[i].ext2bgd_ndirs;
396 fs->e2fs_contigdirs[i] = 0;
397 }
398 if (es->e2fs_rev == E2FS_REV0 ||
399 !EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_LARGEFILE))
400 fs->e2fs_maxfilesize = 0x7fffffff;
401 else {
402 fs->e2fs_maxfilesize = 0xffffffffffff;
403 if (EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_HUGE_FILE))
404 fs->e2fs_maxfilesize = 0x7fffffffffffffff;
405 }
406 if (es->e4fs_flags & E2FS_UNSIGNED_HASH) {
407 fs->e2fs_uhash = 3;
408 } else if ((es->e4fs_flags & E2FS_SIGNED_HASH) == 0) {
409 #ifdef __CHAR_UNSIGNED__
410 es->e4fs_flags |= E2FS_UNSIGNED_HASH;
411 fs->e2fs_uhash = 3;
412 #else
413 es->e4fs_flags |= E2FS_SIGNED_HASH;
414 #endif
415 }
416
417 return (0);
418 }
419
420 /*
421 * Reload all incore data for a filesystem (used after running fsck on
422 * the root filesystem and finding things to fix). The filesystem must
423 * be mounted read-only.
424 *
425 * Things to do to update the mount:
426 * 1) invalidate all cached meta-data.
427 * 2) re-read superblock from disk.
428 * 3) invalidate all cluster summary information.
429 * 4) invalidate all inactive vnodes.
430 * 5) invalidate all cached file data.
431 * 6) re-read inode data for all active vnodes.
432 * XXX we are missing some steps, in particular # 3, this has to be reviewed.
433 */
434 static int
435 ext2_reload(struct mount *mp, struct thread *td)
436 {
437 struct vnode *vp, *mvp, *devvp;
438 struct inode *ip;
439 struct buf *bp;
440 struct ext2fs *es;
441 struct m_ext2fs *fs;
442 struct csum *sump;
443 int error, i;
444 int32_t *lp;
445
446 if ((mp->mnt_flag & MNT_RDONLY) == 0)
447 return (EINVAL);
448 /*
449 * Step 1: invalidate all cached meta-data.
450 */
451 devvp = VFSTOEXT2(mp)->um_devvp;
452 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
453 if (vinvalbuf(devvp, 0, 0, 0) != 0)
454 panic("ext2_reload: dirty1");
455 VOP_UNLOCK(devvp, 0);
456
457 /*
458 * Step 2: re-read superblock from disk.
459 * constants have been adjusted for ext2
460 */
461 if ((error = bread(devvp, SBLOCK, SBSIZE, NOCRED, &bp)) != 0)
462 return (error);
463 es = (struct ext2fs *)bp->b_data;
464 if (ext2_check_sb_compat(es, devvp->v_rdev, 0) != 0) {
465 brelse(bp);
466 return (EIO); /* XXX needs translation */
467 }
468 fs = VFSTOEXT2(mp)->um_e2fs;
469 bcopy(bp->b_data, fs->e2fs, sizeof(struct ext2fs));
470
471 if((error = compute_sb_data(devvp, es, fs)) != 0) {
472 brelse(bp);
473 return (error);
474 }
475 #ifdef UNKLAR
476 if (fs->fs_sbsize < SBSIZE)
477 bp->b_flags |= B_INVAL;
478 #endif
479 brelse(bp);
480
481 /*
482 * Step 3: invalidate all cluster summary information.
483 */
484 if (fs->e2fs_contigsumsize > 0) {
485 lp = fs->e2fs_maxcluster;
486 sump = fs->e2fs_clustersum;
487 for (i = 0; i < fs->e2fs_gcount; i++, sump++) {
488 *lp++ = fs->e2fs_contigsumsize;
489 sump->cs_init = 0;
490 bzero(sump->cs_sum, fs->e2fs_contigsumsize + 1);
491 }
492 }
493
494 loop:
495 MNT_VNODE_FOREACH_ALL(vp, mp, mvp) {
496 /*
497 * Step 4: invalidate all cached file data.
498 */
499 if (vget(vp, LK_EXCLUSIVE | LK_INTERLOCK, td)) {
500 MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp);
501 goto loop;
502 }
503 if (vinvalbuf(vp, 0, 0, 0))
504 panic("ext2_reload: dirty2");
505
506 /*
507 * Step 5: re-read inode data for all active vnodes.
508 */
509 ip = VTOI(vp);
510 error = bread(devvp, fsbtodb(fs, ino_to_fsba(fs, ip->i_number)),
511 (int)fs->e2fs_bsize, NOCRED, &bp);
512 if (error) {
513 VOP_UNLOCK(vp, 0);
514 vrele(vp);
515 MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp);
516 return (error);
517 }
518 ext2_ei2i((struct ext2fs_dinode *) ((char *)bp->b_data +
519 EXT2_INODE_SIZE(fs) * ino_to_fsbo(fs, ip->i_number)), ip);
520 brelse(bp);
521 VOP_UNLOCK(vp, 0);
522 vrele(vp);
523 }
524 return (0);
525 }
526
527 /*
528 * Common code for mount and mountroot.
529 */
530 static int
531 ext2_mountfs(struct vnode *devvp, struct mount *mp)
532 {
533 struct ext2mount *ump;
534 struct buf *bp;
535 struct m_ext2fs *fs;
536 struct ext2fs *es;
537 struct cdev *dev = devvp->v_rdev;
538 struct g_consumer *cp;
539 struct bufobj *bo;
540 struct csum *sump;
541 int error;
542 int ronly;
543 int i, size;
544 int32_t *lp;
545 int32_t e2fs_maxcontig;
546
547 ronly = vfs_flagopt(mp->mnt_optnew, "ro", NULL, 0);
548 /* XXX: use VOP_ACESS to check FS perms */
549 DROP_GIANT();
550 g_topology_lock();
551 error = g_vfs_open(devvp, &cp, "ext2fs", ronly ? 0 : 1);
552 g_topology_unlock();
553 PICKUP_GIANT();
554 VOP_UNLOCK(devvp, 0);
555 if (error)
556 return (error);
557
558 /* XXX: should we check for some sectorsize or 512 instead? */
559 if (((SBSIZE % cp->provider->sectorsize) != 0) ||
560 (SBSIZE < cp->provider->sectorsize)) {
561 DROP_GIANT();
562 g_topology_lock();
563 g_vfs_close(cp);
564 g_topology_unlock();
565 PICKUP_GIANT();
566 return (EINVAL);
567 }
568
569 bo = &devvp->v_bufobj;
570 bo->bo_private = cp;
571 bo->bo_ops = g_vfs_bufops;
572 if (devvp->v_rdev->si_iosize_max != 0)
573 mp->mnt_iosize_max = devvp->v_rdev->si_iosize_max;
574 if (mp->mnt_iosize_max > MAXPHYS)
575 mp->mnt_iosize_max = MAXPHYS;
576
577 bp = NULL;
578 ump = NULL;
579 if ((error = bread(devvp, SBLOCK, SBSIZE, NOCRED, &bp)) != 0)
580 goto out;
581 es = (struct ext2fs *)bp->b_data;
582 if (ext2_check_sb_compat(es, dev, ronly) != 0) {
583 error = EINVAL; /* XXX needs translation */
584 goto out;
585 }
586 if ((es->e2fs_state & E2FS_ISCLEAN) == 0 ||
587 (es->e2fs_state & E2FS_ERRORS)) {
588 if (ronly || (mp->mnt_flag & MNT_FORCE)) {
589 printf(
590 "WARNING: Filesystem was not properly dismounted\n");
591 } else {
592 printf(
593 "WARNING: R/W mount denied. Filesystem is not clean - run fsck\n");
594 error = EPERM;
595 goto out;
596 }
597 }
598 ump = malloc(sizeof(*ump), M_EXT2MNT, M_WAITOK | M_ZERO);
599
600 /*
601 * I don't know whether this is the right strategy. Note that
602 * we dynamically allocate both an ext2_sb_info and an ext2_super_block
603 * while Linux keeps the super block in a locked buffer.
604 */
605 ump->um_e2fs = malloc(sizeof(struct m_ext2fs),
606 M_EXT2MNT, M_WAITOK);
607 ump->um_e2fs->e2fs = malloc(sizeof(struct ext2fs),
608 M_EXT2MNT, M_WAITOK);
609 mtx_init(EXT2_MTX(ump), "EXT2FS", "EXT2FS Lock", MTX_DEF);
610 bcopy(es, ump->um_e2fs->e2fs, (u_int)sizeof(struct ext2fs));
611 if ((error = compute_sb_data(devvp, ump->um_e2fs->e2fs, ump->um_e2fs)))
612 goto out;
613
614 /*
615 * Calculate the maximum contiguous blocks and size of cluster summary
616 * array. In FFS this is done by newfs; however, the superblock
617 * in ext2fs doesn't have these variables, so we can calculate
618 * them here.
619 */
620 e2fs_maxcontig = MAX(1, MAXPHYS / ump->um_e2fs->e2fs_bsize);
621 ump->um_e2fs->e2fs_contigsumsize = MIN(e2fs_maxcontig, EXT2_MAXCONTIG);
622 if (ump->um_e2fs->e2fs_contigsumsize > 0) {
623 size = ump->um_e2fs->e2fs_gcount * sizeof(int32_t);
624 ump->um_e2fs->e2fs_maxcluster = malloc(size, M_EXT2MNT, M_WAITOK);
625 size = ump->um_e2fs->e2fs_gcount * sizeof(struct csum);
626 ump->um_e2fs->e2fs_clustersum = malloc(size, M_EXT2MNT, M_WAITOK);
627 lp = ump->um_e2fs->e2fs_maxcluster;
628 sump = ump->um_e2fs->e2fs_clustersum;
629 for (i = 0; i < ump->um_e2fs->e2fs_gcount; i++, sump++) {
630 *lp++ = ump->um_e2fs->e2fs_contigsumsize;
631 sump->cs_init = 0;
632 sump->cs_sum = malloc((ump->um_e2fs->e2fs_contigsumsize + 1) *
633 sizeof(int32_t), M_EXT2MNT, M_WAITOK | M_ZERO);
634 }
635 }
636
637 brelse(bp);
638 bp = NULL;
639 fs = ump->um_e2fs;
640 fs->e2fs_ronly = ronly; /* ronly is set according to mnt_flags */
641
642 /*
643 * If the fs is not mounted read-only, make sure the super block is
644 * always written back on a sync().
645 */
646 fs->e2fs_wasvalid = fs->e2fs->e2fs_state & E2FS_ISCLEAN ? 1 : 0;
647 if (ronly == 0) {
648 fs->e2fs_fmod = 1; /* mark it modified */
649 fs->e2fs->e2fs_state &= ~E2FS_ISCLEAN; /* set fs invalid */
650 }
651 mp->mnt_data = ump;
652 mp->mnt_stat.f_fsid.val[0] = dev2udev(dev);
653 mp->mnt_stat.f_fsid.val[1] = mp->mnt_vfc->vfc_typenum;
654 mp->mnt_maxsymlinklen = EXT2_MAXSYMLINKLEN;
655 MNT_ILOCK(mp);
656 mp->mnt_flag |= MNT_LOCAL;
657 MNT_IUNLOCK(mp);
658 ump->um_mountp = mp;
659 ump->um_dev = dev;
660 ump->um_devvp = devvp;
661 ump->um_bo = &devvp->v_bufobj;
662 ump->um_cp = cp;
663
664 /*
665 * Setting those two parameters allowed us to use
666 * ufs_bmap w/o changse!
667 */
668 ump->um_nindir = EXT2_ADDR_PER_BLOCK(fs);
669 ump->um_bptrtodb = fs->e2fs->e2fs_log_bsize + 1;
670 ump->um_seqinc = EXT2_FRAGS_PER_BLOCK(fs);
671 if (ronly == 0)
672 ext2_sbupdate(ump, MNT_WAIT);
673 /*
674 * Initialize filesystem stat information in mount struct.
675 */
676 MNT_ILOCK(mp);
677 mp->mnt_kern_flag |= MNTK_LOOKUP_SHARED | MNTK_EXTENDED_SHARED;
678 MNT_IUNLOCK(mp);
679 return (0);
680 out:
681 if (bp)
682 brelse(bp);
683 if (cp != NULL) {
684 DROP_GIANT();
685 g_topology_lock();
686 g_vfs_close(cp);
687 g_topology_unlock();
688 PICKUP_GIANT();
689 }
690 if (ump) {
691 mtx_destroy(EXT2_MTX(ump));
692 free(ump->um_e2fs->e2fs_gd, M_EXT2MNT);
693 free(ump->um_e2fs->e2fs_contigdirs, M_EXT2MNT);
694 free(ump->um_e2fs->e2fs, M_EXT2MNT);
695 free(ump->um_e2fs, M_EXT2MNT);
696 free(ump, M_EXT2MNT);
697 mp->mnt_data = NULL;
698 }
699 return (error);
700 }
701
702 /*
703 * Unmount system call.
704 */
705 static int
706 ext2_unmount(struct mount *mp, int mntflags)
707 {
708 struct ext2mount *ump;
709 struct m_ext2fs *fs;
710 struct csum *sump;
711 int error, flags, i, ronly;
712
713 flags = 0;
714 if (mntflags & MNT_FORCE) {
715 if (mp->mnt_flag & MNT_ROOTFS)
716 return (EINVAL);
717 flags |= FORCECLOSE;
718 }
719 if ((error = ext2_flushfiles(mp, flags, curthread)) != 0)
720 return (error);
721 ump = VFSTOEXT2(mp);
722 fs = ump->um_e2fs;
723 ronly = fs->e2fs_ronly;
724 if (ronly == 0 && ext2_cgupdate(ump, MNT_WAIT) == 0) {
725 if (fs->e2fs_wasvalid)
726 fs->e2fs->e2fs_state |= E2FS_ISCLEAN;
727 ext2_sbupdate(ump, MNT_WAIT);
728 }
729
730 DROP_GIANT();
731 g_topology_lock();
732 g_vfs_close(ump->um_cp);
733 g_topology_unlock();
734 PICKUP_GIANT();
735 vrele(ump->um_devvp);
736 sump = fs->e2fs_clustersum;
737 for (i = 0; i < fs->e2fs_gcount; i++, sump++)
738 free(sump->cs_sum, M_EXT2MNT);
739 free(fs->e2fs_clustersum, M_EXT2MNT);
740 free(fs->e2fs_maxcluster, M_EXT2MNT);
741 free(fs->e2fs_gd, M_EXT2MNT);
742 free(fs->e2fs_contigdirs, M_EXT2MNT);
743 free(fs->e2fs, M_EXT2MNT);
744 free(fs, M_EXT2MNT);
745 free(ump, M_EXT2MNT);
746 mp->mnt_data = NULL;
747 MNT_ILOCK(mp);
748 mp->mnt_flag &= ~MNT_LOCAL;
749 MNT_IUNLOCK(mp);
750 return (error);
751 }
752
753 /*
754 * Flush out all the files in a filesystem.
755 */
756 static int
757 ext2_flushfiles(struct mount *mp, int flags, struct thread *td)
758 {
759 int error;
760
761 error = vflush(mp, 0, flags, td);
762 return (error);
763 }
764 /*
765 * Get filesystem statistics.
766 */
767 int
768 ext2_statfs(struct mount *mp, struct statfs *sbp)
769 {
770 struct ext2mount *ump;
771 struct m_ext2fs *fs;
772 uint32_t overhead, overhead_per_group, ngdb;
773 int i, ngroups;
774
775 ump = VFSTOEXT2(mp);
776 fs = ump->um_e2fs;
777 if (fs->e2fs->e2fs_magic != E2FS_MAGIC)
778 panic("ext2_statfs");
779
780 /*
781 * Compute the overhead (FS structures)
782 */
783 overhead_per_group =
784 1 /* block bitmap */ +
785 1 /* inode bitmap */ +
786 fs->e2fs_itpg;
787 overhead = fs->e2fs->e2fs_first_dblock +
788 fs->e2fs_gcount * overhead_per_group;
789 if (fs->e2fs->e2fs_rev > E2FS_REV0 &&
790 fs->e2fs->e2fs_features_rocompat & EXT2F_ROCOMPAT_SPARSESUPER) {
791 for (i = 0, ngroups = 0; i < fs->e2fs_gcount; i++) {
792 if (cg_has_sb(i))
793 ngroups++;
794 }
795 } else {
796 ngroups = fs->e2fs_gcount;
797 }
798 ngdb = fs->e2fs_gdbcount;
799 if (fs->e2fs->e2fs_rev > E2FS_REV0 &&
800 fs->e2fs->e2fs_features_compat & EXT2F_COMPAT_RESIZE)
801 ngdb += fs->e2fs->e2fs_reserved_ngdb;
802 overhead += ngroups * (1 /* superblock */ + ngdb);
803
804 sbp->f_bsize = EXT2_FRAG_SIZE(fs);
805 sbp->f_iosize = EXT2_BLOCK_SIZE(fs);
806 sbp->f_blocks = fs->e2fs->e2fs_bcount - overhead;
807 sbp->f_bfree = fs->e2fs->e2fs_fbcount;
808 sbp->f_bavail = sbp->f_bfree - fs->e2fs->e2fs_rbcount;
809 sbp->f_files = fs->e2fs->e2fs_icount;
810 sbp->f_ffree = fs->e2fs->e2fs_ficount;
811 return (0);
812 }
813
814 /*
815 * Go through the disk queues to initiate sandbagged IO;
816 * go through the inodes to write those that have been modified;
817 * initiate the writing of the super block if it has been modified.
818 *
819 * Note: we are always called with the filesystem marked `MPBUSY'.
820 */
821 static int
822 ext2_sync(struct mount *mp, int waitfor)
823 {
824 struct vnode *mvp, *vp;
825 struct thread *td;
826 struct inode *ip;
827 struct ext2mount *ump = VFSTOEXT2(mp);
828 struct m_ext2fs *fs;
829 int error, allerror = 0;
830
831 td = curthread;
832 fs = ump->um_e2fs;
833 if (fs->e2fs_fmod != 0 && fs->e2fs_ronly != 0) { /* XXX */
834 printf("fs = %s\n", fs->e2fs_fsmnt);
835 panic("ext2_sync: rofs mod");
836 }
837
838 /*
839 * Write back each (modified) inode.
840 */
841 loop:
842 MNT_VNODE_FOREACH_ALL(vp, mp, mvp) {
843 if (vp->v_type == VNON) {
844 VI_UNLOCK(vp);
845 continue;
846 }
847 ip = VTOI(vp);
848 if ((ip->i_flag &
849 (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 &&
850 (vp->v_bufobj.bo_dirty.bv_cnt == 0 ||
851 waitfor == MNT_LAZY)) {
852 VI_UNLOCK(vp);
853 continue;
854 }
855 error = vget(vp, LK_EXCLUSIVE | LK_NOWAIT | LK_INTERLOCK, td);
856 if (error) {
857 if (error == ENOENT) {
858 MNT_VNODE_FOREACH_ALL_ABORT(mp, mvp);
859 goto loop;
860 }
861 continue;
862 }
863 if ((error = VOP_FSYNC(vp, waitfor, td)) != 0)
864 allerror = error;
865 VOP_UNLOCK(vp, 0);
866 vrele(vp);
867 }
868
869 /*
870 * Force stale filesystem control information to be flushed.
871 */
872 if (waitfor != MNT_LAZY) {
873 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
874 if ((error = VOP_FSYNC(ump->um_devvp, waitfor, td)) != 0)
875 allerror = error;
876 VOP_UNLOCK(ump->um_devvp, 0);
877 }
878
879 /*
880 * Write back modified superblock.
881 */
882 if (fs->e2fs_fmod != 0) {
883 fs->e2fs_fmod = 0;
884 fs->e2fs->e2fs_wtime = time_second;
885 if ((error = ext2_cgupdate(ump, waitfor)) != 0)
886 allerror = error;
887 }
888 return (allerror);
889 }
890
891 /*
892 * Look up an EXT2FS dinode number to find its incore vnode, otherwise read it
893 * in from disk. If it is in core, wait for the lock bit to clear, then
894 * return the inode locked. Detection and handling of mount points must be
895 * done by the calling routine.
896 */
897 static int
898 ext2_vget(struct mount *mp, ino_t ino, int flags, struct vnode **vpp)
899 {
900 struct m_ext2fs *fs;
901 struct inode *ip;
902 struct ext2mount *ump;
903 struct buf *bp;
904 struct vnode *vp;
905 struct cdev *dev;
906 struct thread *td;
907 int i, error;
908 int used_blocks;
909
910 td = curthread;
911 error = vfs_hash_get(mp, ino, flags, td, vpp, NULL, NULL);
912 if (error || *vpp != NULL)
913 return (error);
914
915 ump = VFSTOEXT2(mp);
916 dev = ump->um_dev;
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 IN_E4EXTENTS is enabled, unused blocks are not zeroed
969 * out because we could corrupt the extent tree.
970 */
971 if (!(ip->i_flag & IN_E4EXTENTS) &&
972 (S_ISDIR(ip->i_mode) || S_ISREG(ip->i_mode))) {
973 used_blocks = (ip->i_size+fs->e2fs_bsize-1) / fs->e2fs_bsize;
974 for (i = used_blocks; i < EXT2_NDIR_BLOCKS; i++)
975 ip->i_db[i] = 0;
976 }
977 /*
978 ext2_print_inode(ip);
979 */
980 bqrelse(bp);
981
982 /*
983 * Initialize the vnode from the inode, check for aliases.
984 * Note that the underlying vnode may have changed.
985 */
986 if ((error = ext2_vinit(mp, &ext2_fifoops, &vp)) != 0) {
987 vput(vp);
988 *vpp = NULL;
989 return (error);
990 }
991
992 /*
993 * Finish inode initialization.
994 */
995
996 /*
997 * Set up a generation number for this inode if it does not
998 * already have one. This should only happen on old filesystems.
999 */
1000 if (ip->i_gen == 0) {
1001 ip->i_gen = random() + 1;
1002 if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0)
1003 ip->i_flag |= IN_MODIFIED;
1004 }
1005 *vpp = vp;
1006 return (0);
1007 }
1008
1009 /*
1010 * File handle to vnode
1011 *
1012 * Have to be really careful about stale file handles:
1013 * - check that the inode number is valid
1014 * - call ext2_vget() to get the locked inode
1015 * - check for an unallocated inode (i_mode == 0)
1016 * - check that the given client host has export rights and return
1017 * those rights via. exflagsp and credanonp
1018 */
1019 static int
1020 ext2_fhtovp(struct mount *mp, struct fid *fhp, int flags, struct vnode **vpp)
1021 {
1022 struct inode *ip;
1023 struct ufid *ufhp;
1024 struct vnode *nvp;
1025 struct m_ext2fs *fs;
1026 int error;
1027
1028 ufhp = (struct ufid *)fhp;
1029 fs = VFSTOEXT2(mp)->um_e2fs;
1030 if (ufhp->ufid_ino < EXT2_ROOTINO ||
1031 ufhp->ufid_ino > fs->e2fs_gcount * fs->e2fs->e2fs_ipg)
1032 return (ESTALE);
1033
1034 error = VFS_VGET(mp, ufhp->ufid_ino, LK_EXCLUSIVE, &nvp);
1035 if (error) {
1036 *vpp = NULLVP;
1037 return (error);
1038 }
1039 ip = VTOI(nvp);
1040 if (ip->i_mode == 0 ||
1041 ip->i_gen != ufhp->ufid_gen || ip->i_nlink <= 0) {
1042 vput(nvp);
1043 *vpp = NULLVP;
1044 return (ESTALE);
1045 }
1046 *vpp = nvp;
1047 vnode_create_vobject(*vpp, 0, curthread);
1048 return (0);
1049 }
1050
1051 /*
1052 * Write a superblock and associated information back to disk.
1053 */
1054 static int
1055 ext2_sbupdate(struct ext2mount *mp, int waitfor)
1056 {
1057 struct m_ext2fs *fs = mp->um_e2fs;
1058 struct ext2fs *es = fs->e2fs;
1059 struct buf *bp;
1060 int error = 0;
1061
1062 bp = getblk(mp->um_devvp, SBLOCK, SBSIZE, 0, 0, 0);
1063 bcopy((caddr_t)es, bp->b_data, (u_int)sizeof(struct ext2fs));
1064 if (waitfor == MNT_WAIT)
1065 error = bwrite(bp);
1066 else
1067 bawrite(bp);
1068
1069 /*
1070 * The buffers for group descriptors, inode bitmaps and block bitmaps
1071 * are not busy at this point and are (hopefully) written by the
1072 * usual sync mechanism. No need to write them here.
1073 */
1074 return (error);
1075 }
1076 int
1077 ext2_cgupdate(struct ext2mount *mp, int waitfor)
1078 {
1079 struct m_ext2fs *fs = mp->um_e2fs;
1080 struct buf *bp;
1081 int i, error = 0, allerror = 0;
1082
1083 allerror = ext2_sbupdate(mp, waitfor);
1084 for (i = 0; i < fs->e2fs_gdbcount; i++) {
1085 bp = getblk(mp->um_devvp, fsbtodb(fs,
1086 fs->e2fs->e2fs_first_dblock +
1087 1 /* superblock */ + i), fs->e2fs_bsize, 0, 0, 0);
1088 e2fs_cgsave(&fs->e2fs_gd[
1089 i * fs->e2fs_bsize / sizeof(struct ext2_gd)],
1090 (struct ext2_gd *)bp->b_data, fs->e2fs_bsize);
1091 if (waitfor == MNT_WAIT)
1092 error = bwrite(bp);
1093 else
1094 bawrite(bp);
1095 }
1096
1097 if (!allerror && error)
1098 allerror = error;
1099 return (allerror);
1100 }
1101 /*
1102 * Return the root of a filesystem.
1103 */
1104 static int
1105 ext2_root(struct mount *mp, int flags, struct vnode **vpp)
1106 {
1107 struct vnode *nvp;
1108 int error;
1109
1110 error = VFS_VGET(mp, EXT2_ROOTINO, LK_EXCLUSIVE, &nvp);
1111 if (error)
1112 return (error);
1113 *vpp = nvp;
1114 return (0);
1115 }
Cache object: 354c3626a191aa70b1531e814132c708
|