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 * SPDX-License-Identifier: BSD-3-Clause
9 *
10 * Copyright (c) 1989, 1991, 1993, 1994
11 * The Regents of the University of California. All rights reserved.
12 *
13 * Redistribution and use in source and binary forms, with or without
14 * modification, are permitted provided that the following conditions
15 * are met:
16 * 1. Redistributions of source code must retain the above copyright
17 * notice, this list of conditions and the following disclaimer.
18 * 2. Redistributions in binary form must reproduce the above copyright
19 * notice, this list of conditions and the following disclaimer in the
20 * documentation and/or other materials provided with the distribution.
21 * 3. Neither the name of the University nor the names of its contributors
22 * may be used to endorse or promote products derived from this software
23 * without specific prior written permission.
24 *
25 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
26 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
27 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
28 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
30 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
31 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
32 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
33 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
34 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
35 * SUCH DAMAGE.
36 *
37 * @(#)ffs_vfsops.c 8.8 (Berkeley) 4/18/94
38 * $FreeBSD$
39 */
40
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/namei.h>
44 #include <sys/caps.h>
45 #include <sys/proc.h>
46 #include <sys/kernel.h>
47 #include <sys/vnode.h>
48 #include <sys/mount.h>
49 #include <sys/bio.h>
50 #include <sys/buf2.h>
51 #include <sys/conf.h>
52 #include <sys/endian.h>
53 #include <sys/fcntl.h>
54 #include <sys/malloc.h>
55 #include <sys/stat.h>
56 #include <sys/mutex2.h>
57 #include <sys/nlookup.h>
58
59 #include <vfs/ext2fs/fs.h>
60 #include <vfs/ext2fs/ext2_mount.h>
61 #include <vfs/ext2fs/inode.h>
62
63 #include <vfs/ext2fs/ext2fs.h>
64 #include <vfs/ext2fs/ext2_dinode.h>
65 #include <vfs/ext2fs/ext2_extern.h>
66 #include <vfs/ext2fs/ext2_extents.h>
67
68 SDT_PROVIDER_DECLARE(ext2fs);
69 /*
70 * ext2fs trace probe:
71 * arg0: verbosity. Higher numbers give more verbose messages
72 * arg1: Textual message
73 */
74 SDT_PROBE_DEFINE2(ext2fs, , vfsops, trace, "int", "char*");
75 SDT_PROBE_DEFINE2(ext2fs, , vfsops, ext2_cg_validate_error, "char*", "int");
76 SDT_PROBE_DEFINE1(ext2fs, , vfsops, ext2_compute_sb_data_error, "char*");
77
78 static int ext2_flushfiles(struct mount *mp, int flags);
79 static int ext2_mountfs(struct vnode *, struct mount *);
80 static int ext2_reload(struct mount *mp);
81 static int ext2_sbupdate(struct ext2mount *, int);
82 static int ext2_cgupdate(struct ext2mount *, int);
83 static int ext2_init(struct vfsconf *);
84 static int ext2_uninit(struct vfsconf *);
85 static vfs_unmount_t ext2_unmount;
86 static vfs_root_t ext2_root;
87 static vfs_statfs_t ext2_statfs;
88 static vfs_statvfs_t ext2_statvfs;
89 static vfs_sync_t ext2_sync;
90 static vfs_vget_t ext2_vget;
91 static vfs_fhtovp_t ext2_fhtovp;
92 static vfs_vptofh_t ext2_vptofh;
93 static vfs_checkexp_t ext2_check_export;
94 static vfs_mount_t ext2_mount;
95
96 MALLOC_DEFINE(M_EXT2NODE, "ext2_node", "EXT2 vnode private part");
97 static MALLOC_DEFINE(M_EXT2MNT, "ext2_mount", "EXT2 mount structure");
98
99 static struct vfsops ext2fs_vfsops = {
100 .vfs_flags = 0,
101 .vfs_mount = ext2_mount,
102 .vfs_unmount = ext2_unmount,
103 .vfs_root = ext2_root, /* root inode via vget */
104 .vfs_statfs = ext2_statfs,
105 .vfs_statvfs = ext2_statvfs,
106 .vfs_sync = ext2_sync,
107 .vfs_vget = ext2_vget,
108 .vfs_fhtovp = ext2_fhtovp,
109 .vfs_vptofh = ext2_vptofh,
110 .vfs_checkexp = ext2_check_export,
111 .vfs_init = ext2_init,
112 .vfs_uninit = ext2_uninit
113 };
114
115 VFS_SET(ext2fs_vfsops, ext2fs, VFCF_MPSAFE);
116 MODULE_VERSION(ext2fs, 1);
117
118 static int ext2_check_sb_compat(struct ext2fs *es, struct cdev *dev,
119 int ronly);
120 static int ext2_compute_sb_data(struct vnode * devvp,
121 struct ext2fs * es, struct m_ext2fs * fs);
122
123 static int ext2fs_inode_hash_lock;
124
125 /*
126 * VFS Operations.
127 *
128 * mount system call
129 */
130 static int
ext2_mount(struct mount * mp,char * path,caddr_t data,struct ucred * cred)131 ext2_mount(struct mount *mp, char *path, caddr_t data, struct ucred *cred)
132 {
133 struct ext2_args args;
134 struct vnode *devvp;
135 struct ext2mount *ump = NULL;
136 struct m_ext2fs *fs;
137 struct nlookupdata nd;
138 mode_t accmode;
139 int error, flags;
140 size_t size;
141
142 if ((error = copyin(data, (caddr_t)&args, sizeof (struct ext2_args))) != 0)
143 return (error);
144
145 /*
146 * If updating, check whether changing from read-only to
147 * read/write; if there is no device name, that's all we do.
148 */
149 if (mp->mnt_flag & MNT_UPDATE) {
150 ump = VFSTOEXT2(mp);
151 fs = ump->um_e2fs;
152 devvp = ump->um_devvp;
153 error = 0;
154 if (fs->e2fs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) {
155 error = VFS_SYNC(mp, MNT_WAIT);
156 if (error)
157 return (error);
158 flags = WRITECLOSE;
159 if (mp->mnt_flag & MNT_FORCE)
160 flags |= FORCECLOSE;
161 if (vfs_busy(mp, LK_NOWAIT))
162 return (EBUSY);
163 error = ext2_flushfiles(mp, flags);
164 vfs_unbusy(mp);
165 if (error == 0 && fs->e2fs_wasvalid &&
166 ext2_cgupdate(ump, MNT_WAIT) == 0) {
167 fs->e2fs->e2fs_state =
168 htole16((le16toh(fs->e2fs->e2fs_state) |
169 E2FS_ISCLEAN));
170 ext2_sbupdate(ump, MNT_WAIT);
171 }
172 fs->e2fs_ronly = 1;
173 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
174 VOP_OPEN(devvp, FREAD, FSCRED, NULL);
175 VOP_CLOSE(devvp, FREAD | FWRITE, NULL);
176 vn_unlock(devvp);
177 }
178 if (!error && (mp->mnt_flag & MNT_RELOAD))
179 error = ext2_reload(mp);
180 if (error)
181 return (error);
182 devvp = ump->um_devvp;
183 if (fs->e2fs_ronly && (mp->mnt_kern_flag & MNTK_WANTRDWR)) {
184 if (ext2_check_sb_compat(fs->e2fs, devvp->v_rdev, 0))
185 return (EPERM);
186
187 /*
188 * If upgrade to read-write by non-root, then verify
189 * that user has necessary permissions on the device.
190 */
191 if (cred->cr_uid != 0) {
192 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
193 error = VOP_EACCESS(devvp, VREAD | VWRITE, cred);
194 if (error) {
195 vn_unlock(devvp);
196 return (error);
197 }
198 vn_unlock(devvp);
199 }
200
201 if ((le16toh(fs->e2fs->e2fs_state) & E2FS_ISCLEAN) == 0 ||
202 (le16toh(fs->e2fs->e2fs_state) & E2FS_ERRORS)) {
203 if (mp->mnt_flag & MNT_FORCE) {
204 printf(
205 "WARNING: %s was not properly dismounted\n", fs->e2fs_fsmnt);
206 } else {
207 printf(
208 "WARNING: R/W mount of %s denied. Filesystem is not clean - run fsck\n",
209 fs->e2fs_fsmnt);
210 return (EPERM);
211 }
212 }
213 fs->e2fs->e2fs_state =
214 htole16(le16toh(fs->e2fs->e2fs_state) & ~E2FS_ISCLEAN);
215 (void)ext2_cgupdate(ump, MNT_WAIT);
216 fs->e2fs_ronly = 0;
217 mp->mnt_flag &= ~MNT_RDONLY;
218
219 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
220 VOP_OPEN(devvp, FREAD | FWRITE, FSCRED, NULL);
221 VOP_CLOSE(devvp, FREAD, NULL);
222 vn_unlock(devvp);
223 }
224 if (args.fspec == NULL) {
225 /*
226 * Process export requests.
227 */
228 return (vfs_export(mp, &ump->um_export, &args.export));
229 }
230 }
231
232 /*
233 * Not an update, or updating the name: look up the name
234 * and verify that it refers to a sensible disk device.
235 */
236 devvp = NULL;
237 error = nlookup_init(&nd, args.fspec, UIO_USERSPACE, NLC_FOLLOW);
238 if (error == 0)
239 error = nlookup(&nd);
240 if (error == 0)
241 error = cache_vref(&nd.nl_nch, nd.nl_cred, &devvp);
242 nlookup_done(&nd);
243 if (error)
244 return (error);
245
246 if (!vn_isdisk(devvp, &error)) {
247 vrele(devvp);
248 return (error);
249 }
250
251 /*
252 * If mount by non-root, then verify that user has necessary
253 * permissions on the device.
254 *
255 * XXXRW: VOP_ACCESS() enough?
256 */
257 if (cred->cr_uid != 0) {
258 accmode = VREAD;
259 if ((mp->mnt_flag & MNT_RDONLY) == 0)
260 accmode |= VWRITE;
261 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
262 if ((error = VOP_EACCESS(devvp, accmode, cred)) != 0) {
263 vput(devvp);
264 return (error);
265 }
266 vn_unlock(devvp);
267 }
268
269 if ((mp->mnt_flag & MNT_UPDATE) == 0) {
270 error = ext2_mountfs(devvp, mp);
271 } else {
272 if (devvp != ump->um_devvp)
273 error = EINVAL; /* needs translation */
274 else
275 vrele(devvp);
276 }
277 if (error) {
278 vrele(devvp);
279 return (error);
280 }
281 ump = VFSTOEXT2(mp);
282 fs = ump->um_e2fs;
283
284 /*
285 * Note that this strncpy() is ok because of a check at the start
286 * of ext2_mount().
287 */
288 copyinstr(path, fs->e2fs_fsmnt, sizeof(fs->e2fs_fsmnt) - 1, &size);
289 bzero(fs->e2fs_fsmnt + size, sizeof(fs->e2fs_fsmnt) - size);
290 copyinstr(args.fspec, mp->mnt_stat.f_mntfromname, MNAMELEN - 1, &size);
291 bzero(mp->mnt_stat.f_mntfromname + size, MNAMELEN - size);
292 ext2_statfs(mp, &mp->mnt_stat, cred);
293 return (0);
294 }
295
296 static int
ext2_check_sb_compat(struct ext2fs * es,struct cdev * dev,int ronly)297 ext2_check_sb_compat(struct ext2fs *es, struct cdev *dev, int ronly)
298 {
299 uint32_t i, mask;
300
301 if (le16toh(es->e2fs_magic) != E2FS_MAGIC) {
302 printf("ext2fs: %s: wrong magic number %#x (expected %#x)\n",
303 devtoname(dev), le16toh(es->e2fs_magic), E2FS_MAGIC);
304 return (1);
305 }
306 if (le32toh(es->e2fs_rev) > E2FS_REV0) {
307 mask = le32toh(es->e2fs_features_incompat) & ~(EXT2F_INCOMPAT_SUPP);
308 if (mask) {
309 printf("WARNING: mount of %s denied due to "
310 "unsupported optional features:\n", devtoname(dev));
311 for (i = 0;
312 i < sizeof(incompat)/sizeof(struct ext2_feature);
313 i++)
314 if (mask & incompat[i].mask)
315 printf("%s ", incompat[i].name);
316 printf("\n");
317 return (1);
318 }
319 mask = le32toh(es->e2fs_features_rocompat) & ~EXT2F_ROCOMPAT_SUPP;
320 if (!ronly && mask) {
321 printf("WARNING: R/W mount of %s denied due to "
322 "unsupported optional features:\n", devtoname(dev));
323 for (i = 0;
324 i < sizeof(ro_compat)/sizeof(struct ext2_feature);
325 i++)
326 if (mask & ro_compat[i].mask)
327 printf("%s ", ro_compat[i].name);
328 printf("\n");
329 return (1);
330 }
331 }
332 return (0);
333 }
334
335 static e4fs_daddr_t
ext2_cg_location(struct m_ext2fs * fs,int number)336 ext2_cg_location(struct m_ext2fs *fs, int number)
337 {
338 int cg, descpb, logical_sb, has_super = 0;
339
340 /*
341 * Adjust logical superblock block number.
342 * Godmar thinks: if the blocksize is greater than 1024, then
343 * the superblock is logically part of block zero.
344 */
345 logical_sb = fs->e2fs_bsize > SBSIZE ? 0 : 1;
346
347 if (!EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_META_BG) ||
348 number < le32toh(fs->e2fs->e3fs_first_meta_bg))
349 return (logical_sb + number + 1);
350
351 if (EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_64BIT))
352 descpb = fs->e2fs_bsize / sizeof(struct ext2_gd);
353 else
354 descpb = fs->e2fs_bsize / E2FS_REV0_GD_SIZE;
355
356 cg = descpb * number;
357
358 if (ext2_cg_has_sb(fs, cg))
359 has_super = 1;
360
361 return (has_super + cg * (e4fs_daddr_t)EXT2_BLOCKS_PER_GROUP(fs) +
362 le32toh(fs->e2fs->e2fs_first_dblock));
363 }
364
365 static int
ext2_cg_validate(struct m_ext2fs * fs)366 ext2_cg_validate(struct m_ext2fs *fs)
367 {
368 uint64_t b_bitmap;
369 uint64_t i_bitmap;
370 uint64_t i_tables;
371 uint64_t first_block, last_block, last_cg_block;
372 struct ext2_gd *gd;
373 unsigned int i, cg_count;
374
375 first_block = le32toh(fs->e2fs->e2fs_first_dblock);
376 last_cg_block = ext2_cg_number_gdb(fs, 0);
377 cg_count = fs->e2fs_gcount;
378
379 for (i = 0; i < fs->e2fs_gcount; i++) {
380 gd = &fs->e2fs_gd[i];
381
382 if (EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_FLEX_BG) ||
383 i == fs->e2fs_gcount - 1) {
384 last_block = fs->e2fs_bcount - 1;
385 } else {
386 last_block = first_block +
387 (EXT2_BLOCKS_PER_GROUP(fs) - 1);
388 }
389
390 if ((cg_count == fs->e2fs_gcount) &&
391 !(le16toh(gd->ext4bgd_flags) & EXT2_BG_INODE_ZEROED))
392 cg_count = i;
393
394 b_bitmap = e2fs_gd_get_b_bitmap(gd);
395 if (b_bitmap == 0) {
396 SDT_PROBE2(ext2fs, , vfsops, ext2_cg_validate_error,
397 "block bitmap is zero", i);
398 return (EINVAL);
399 }
400 if (b_bitmap <= last_cg_block) {
401 SDT_PROBE2(ext2fs, , vfsops, ext2_cg_validate_error,
402 "block bitmap overlaps gds", i);
403 return (EINVAL);
404 }
405 if (b_bitmap < first_block || b_bitmap > last_block) {
406 SDT_PROBE2(ext2fs, , vfsops, ext2_cg_validate_error,
407 "block bitmap not in group", i);
408 return (EINVAL);
409 }
410
411 i_bitmap = e2fs_gd_get_i_bitmap(gd);
412 if (i_bitmap == 0) {
413 SDT_PROBE2(ext2fs, , vfsops, ext2_cg_validate_error,
414 "inode bitmap is zero", i);
415 return (EINVAL);
416 }
417 if (i_bitmap <= last_cg_block) {
418 SDT_PROBE2(ext2fs, , vfsops, ext2_cg_validate_error,
419 "inode bitmap overlaps gds", i);
420 return (EINVAL);
421 }
422 if (i_bitmap < first_block || i_bitmap > last_block) {
423 SDT_PROBE2(ext2fs, , vfsops, ext2_cg_validate_error,
424 "inode bitmap not in group blk", i);
425 return (EINVAL);
426 }
427
428 i_tables = e2fs_gd_get_i_tables(gd);
429 if (i_tables == 0) {
430 SDT_PROBE2(ext2fs, , vfsops, ext2_cg_validate_error,
431 "inode table is zero", i);
432 return (EINVAL);
433 }
434 if (i_tables <= last_cg_block) {
435 SDT_PROBE2(ext2fs, , vfsops, ext2_cg_validate_error,
436 "inode tables overlaps gds", i);
437 return (EINVAL);
438 }
439 if (i_tables < first_block ||
440 i_tables + fs->e2fs_itpg - 1 > last_block) {
441 SDT_PROBE2(ext2fs, , vfsops, ext2_cg_validate_error,
442 "inode tables not in group blk", i);
443 return (EINVAL);
444 }
445
446 if (!EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_FLEX_BG))
447 first_block += EXT2_BLOCKS_PER_GROUP(fs);
448 }
449
450 return (0);
451 }
452
453 /*
454 * This computes the fields of the m_ext2fs structure from the
455 * data in the ext2fs structure read in.
456 */
457 static int
ext2_compute_sb_data(struct vnode * devvp,struct ext2fs * es,struct m_ext2fs * fs)458 ext2_compute_sb_data(struct vnode *devvp, struct ext2fs *es,
459 struct m_ext2fs *fs)
460 {
461 struct buf *bp;
462 uint32_t e2fs_descpb, e2fs_gdbcount_alloc;
463 int i, j;
464 int g_count = 0;
465 int error;
466
467 /* Check checksum features */
468 if (EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_GDT_CSUM) &&
469 EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_METADATA_CKSUM)) {
470 SDT_PROBE1(ext2fs, , vfsops, ext2_compute_sb_data_error,
471 "incorrect checksum features combination");
472 return (EINVAL);
473 }
474
475 /* Precompute checksum seed for all metadata */
476 ext2_sb_csum_set_seed(fs);
477
478 /* Verify sb csum if possible */
479 if (EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_METADATA_CKSUM)) {
480 error = ext2_sb_csum_verify(fs);
481 if (error) {
482 return (error);
483 }
484 }
485
486 /* Check for block size = 1K|2K|4K */
487 if (le32toh(es->e2fs_log_bsize) > 2) {
488 SDT_PROBE1(ext2fs, , vfsops, ext2_compute_sb_data_error,
489 "bad block size");
490 return (EINVAL);
491 }
492
493 fs->e2fs_bshift = EXT2_MIN_BLOCK_LOG_SIZE + le32toh(es->e2fs_log_bsize);
494 fs->e2fs_bsize = 1U << fs->e2fs_bshift;
495 fs->e2fs_fsbtodb = le32toh(es->e2fs_log_bsize) + 1;
496 fs->e2fs_qbmask = fs->e2fs_bsize - 1;
497
498 /* Check for fragment size */
499 if (le32toh(es->e2fs_log_fsize) >
500 (EXT2_MAX_FRAG_LOG_SIZE - EXT2_MIN_BLOCK_LOG_SIZE)) {
501 SDT_PROBE1(ext2fs, , vfsops, ext2_compute_sb_data_error,
502 "invalid log cluster size");
503 return (EINVAL);
504 }
505
506 fs->e2fs_fsize = EXT2_MIN_FRAG_SIZE << le32toh(es->e2fs_log_fsize);
507 if (fs->e2fs_fsize != fs->e2fs_bsize) {
508 SDT_PROBE1(ext2fs, , vfsops, ext2_compute_sb_data_error,
509 "fragment size != block size");
510 return (EINVAL);
511 }
512
513 fs->e2fs_fpb = fs->e2fs_bsize / fs->e2fs_fsize;
514
515 /* Check reserved gdt blocks for future filesystem expansion */
516 if (le16toh(es->e2fs_reserved_ngdb) > (fs->e2fs_bsize / 4)) {
517 SDT_PROBE1(ext2fs, , vfsops, ext2_compute_sb_data_error,
518 "number of reserved GDT blocks too large");
519 return (EINVAL);
520 }
521
522 if (le32toh(es->e2fs_rev) == E2FS_REV0) {
523 fs->e2fs_isize = E2FS_REV0_INODE_SIZE;
524 } else {
525 fs->e2fs_isize = le16toh(es->e2fs_inode_size);
526
527 /*
528 * Check first ino.
529 */
530 if (le32toh(es->e2fs_first_ino) < EXT2_FIRSTINO) {
531 SDT_PROBE1(ext2fs, , vfsops, ext2_compute_sb_data_error,
532 "invalid first ino");
533 return (EINVAL);
534 }
535
536 /*
537 * Simple sanity check for superblock inode size value.
538 */
539 if (EXT2_INODE_SIZE(fs) < E2FS_REV0_INODE_SIZE ||
540 EXT2_INODE_SIZE(fs) > fs->e2fs_bsize ||
541 (fs->e2fs_isize & (fs->e2fs_isize - 1)) != 0) {
542 SDT_PROBE1(ext2fs, , vfsops, ext2_compute_sb_data_error,
543 "invalid inode size");
544 return (EINVAL);
545 }
546 }
547
548 /* Check group descriptors */
549 if (EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_64BIT) &&
550 le16toh(es->e3fs_desc_size) != E2FS_64BIT_GD_SIZE) {
551 SDT_PROBE1(ext2fs, , vfsops, ext2_compute_sb_data_error,
552 "unsupported 64bit descriptor size");
553 return (EINVAL);
554 }
555
556 fs->e2fs_bpg = le32toh(es->e2fs_bpg);
557 fs->e2fs_fpg = le32toh(es->e2fs_fpg);
558 if (fs->e2fs_bpg == 0 || fs->e2fs_fpg == 0) {
559 SDT_PROBE1(ext2fs, , vfsops, ext2_compute_sb_data_error,
560 "zero blocks/fragments per group");
561 return (EINVAL);
562 } else if (fs->e2fs_bpg != fs->e2fs_fpg) {
563 SDT_PROBE1(ext2fs, , vfsops, ext2_compute_sb_data_error,
564 "blocks per group not equal fragments per group");
565 return (EINVAL);
566 }
567
568 if (fs->e2fs_bpg != fs->e2fs_bsize * 8) {
569 SDT_PROBE1(ext2fs, , vfsops, ext2_compute_sb_data_error,
570 "non-standard group size unsupported");
571 return (EINVAL);
572 }
573
574 fs->e2fs_ipb = fs->e2fs_bsize / EXT2_INODE_SIZE(fs);
575 if (fs->e2fs_ipb == 0 ||
576 fs->e2fs_ipb > fs->e2fs_bsize / E2FS_REV0_INODE_SIZE) {
577 SDT_PROBE1(ext2fs, , vfsops, ext2_compute_sb_data_error,
578 "bad inodes per block size");
579 return (EINVAL);
580 }
581
582 fs->e2fs_ipg = le32toh(es->e2fs_ipg);
583 if (fs->e2fs_ipg < fs->e2fs_ipb || fs->e2fs_ipg > fs->e2fs_bsize * 8) {
584 SDT_PROBE1(ext2fs, , vfsops, ext2_compute_sb_data_error,
585 "invalid inodes per group");
586 return (EINVAL);
587 }
588
589 fs->e2fs_itpg = fs->e2fs_ipg / fs->e2fs_ipb;
590
591 fs->e2fs_bcount = le32toh(es->e2fs_bcount);
592 fs->e2fs_rbcount = le32toh(es->e2fs_rbcount);
593 fs->e2fs_fbcount = le32toh(es->e2fs_fbcount);
594 if (EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_64BIT)) {
595 fs->e2fs_bcount |= (uint64_t)(le32toh(es->e4fs_bcount_hi)) << 32;
596 fs->e2fs_rbcount |= (uint64_t)(le32toh(es->e4fs_rbcount_hi)) << 32;
597 fs->e2fs_fbcount |= (uint64_t)(le32toh(es->e4fs_fbcount_hi)) << 32;
598 }
599 if (fs->e2fs_rbcount > fs->e2fs_bcount ||
600 fs->e2fs_fbcount > fs->e2fs_bcount) {
601 SDT_PROBE1(ext2fs, , vfsops, ext2_compute_sb_data_error,
602 "invalid block count");
603 return (EINVAL);
604 }
605
606 fs->e2fs_ficount = le32toh(es->e2fs_ficount);
607 if (fs->e2fs_ficount > le32toh(es->e2fs_icount)) {
608 SDT_PROBE1(ext2fs, , vfsops, ext2_compute_sb_data_error,
609 "invalid number of free inodes");
610 return (EINVAL);
611 }
612
613 if (le32toh(es->e2fs_first_dblock) != (fs->e2fs_bsize > 1024 ? 0 : 1) ||
614 le32toh(es->e2fs_first_dblock) >= fs->e2fs_bcount) {
615 SDT_PROBE1(ext2fs, , vfsops, ext2_compute_sb_data_error,
616 "first data block out of range");
617 return (EINVAL);
618 }
619
620 fs->e2fs_gcount = howmany(fs->e2fs_bcount -
621 le32toh(es->e2fs_first_dblock), EXT2_BLOCKS_PER_GROUP(fs));
622 if (fs->e2fs_gcount > ((uint64_t)1 << 32) - EXT2_DESCS_PER_BLOCK(fs)) {
623 SDT_PROBE1(ext2fs, , vfsops, ext2_compute_sb_data_error,
624 "groups count too large");
625 return (EINVAL);
626 }
627
628 /* Check for extra isize in big inodes. */
629 if (EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_EXTRA_ISIZE) &&
630 EXT2_INODE_SIZE(fs) < sizeof(struct ext2fs_dinode)) {
631 SDT_PROBE1(ext2fs, , vfsops, ext2_compute_sb_data_error,
632 "no space for extra inode timestamps");
633 return (EINVAL);
634 }
635
636 /* s_resuid / s_resgid ? */
637
638 if (EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_64BIT)) {
639 e2fs_descpb = fs->e2fs_bsize / E2FS_64BIT_GD_SIZE;
640 e2fs_gdbcount_alloc = howmany(fs->e2fs_gcount, e2fs_descpb);
641 } else {
642 e2fs_descpb = fs->e2fs_bsize / E2FS_REV0_GD_SIZE;
643 e2fs_gdbcount_alloc = howmany(fs->e2fs_gcount,
644 fs->e2fs_bsize / sizeof(struct ext2_gd));
645 }
646 fs->e2fs_gdbcount = howmany(fs->e2fs_gcount, e2fs_descpb);
647 fs->e2fs_gd = malloc(e2fs_gdbcount_alloc * fs->e2fs_bsize,
648 M_EXT2MNT, M_WAITOK | M_ZERO);
649 fs->e2fs_contigdirs = malloc(fs->e2fs_gcount *
650 sizeof(*fs->e2fs_contigdirs), M_EXT2MNT, M_WAITOK | M_ZERO);
651
652 for (i = 0; i < fs->e2fs_gdbcount; i++) {
653 error = bread(devvp, fsbtodoff(fs, ext2_cg_location(fs, i)),
654 fs->e2fs_bsize, &bp);
655 if (error) {
656 /*
657 * fs->e2fs_gd and fs->e2fs_contigdirs
658 * will be freed later by the caller,
659 * because this function could be called from
660 * MNT_UPDATE path.
661 */
662 brelse(bp);
663 return (error);
664 }
665 if (EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_64BIT)) {
666 memcpy(&fs->e2fs_gd[
667 i * fs->e2fs_bsize / sizeof(struct ext2_gd)],
668 bp->b_data, fs->e2fs_bsize);
669 } else {
670 for (j = 0; j < e2fs_descpb &&
671 g_count < fs->e2fs_gcount; j++, g_count++)
672 memcpy(&fs->e2fs_gd[g_count],
673 bp->b_data + j * E2FS_REV0_GD_SIZE,
674 E2FS_REV0_GD_SIZE);
675 }
676 brelse(bp);
677 bp = NULL;
678 }
679
680 /* Validate cgs consistency */
681 error = ext2_cg_validate(fs);
682 if (error)
683 return (error);
684
685 /* Verfy cgs csum */
686 if (EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_GDT_CSUM) ||
687 EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_METADATA_CKSUM)) {
688 error = ext2_gd_csum_verify(fs, devvp->v_rdev);
689 if (error)
690 return (error);
691 }
692 /* Initialization for the ext2 Orlov allocator variant. */
693 fs->e2fs_total_dir = 0;
694 for (i = 0; i < fs->e2fs_gcount; i++)
695 fs->e2fs_total_dir += e2fs_gd_get_ndirs(&fs->e2fs_gd[i]);
696
697 if (le32toh(es->e2fs_rev) == E2FS_REV0 ||
698 !EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_LARGEFILE))
699 fs->e2fs_maxfilesize = 0x7fffffff;
700 else {
701 fs->e2fs_maxfilesize = 0xffffffffffff;
702 if (EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_HUGE_FILE))
703 fs->e2fs_maxfilesize = 0x7fffffffffffffff;
704 }
705 if (le32toh(es->e4fs_flags) & E2FS_UNSIGNED_HASH) {
706 fs->e2fs_uhash = 3;
707 } else if ((le32toh(es->e4fs_flags) & E2FS_SIGNED_HASH) == 0) {
708 #ifdef __CHAR_UNSIGNED__
709 es->e4fs_flags = htole32(le32toh(es->e4fs_flags) | E2FS_UNSIGNED_HASH);
710 fs->e2fs_uhash = 3;
711 #else
712 es->e4fs_flags = htole32(le32toh(es->e4fs_flags) | E2FS_SIGNED_HASH);
713 #endif
714 }
715 if (EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_METADATA_CKSUM))
716 error = ext2_sb_csum_verify(fs);
717
718 return (error);
719 }
720
721 struct scaninfo {
722 int rescan;
723 int allerror;
724 int waitfor;
725 struct vnode *devvp;
726 struct m_ext2fs *fs;
727 };
728
729 static int
ext2_reload_scan(struct mount * mp,struct vnode * vp,void * data)730 ext2_reload_scan(struct mount *mp, struct vnode *vp, void *data)
731 {
732 struct scaninfo *info = data;
733 struct inode *ip;
734 struct buf *bp;
735 int error;
736
737 /*
738 * Try to recycle
739 */
740 if (vrecycle(vp))
741 return (0);
742
743 /*
744 * Step 1: invalidate all cached file data.
745 */
746 if (vinvalbuf(vp, 0, 0, 0))
747 panic("ext2_reload: dirty2");
748 /*
749 * Step 2: re-read inode data for all active vnodes.
750 */
751 ip = VTOI(vp);
752 error = bread(info->devvp,
753 fsbtodoff(info->fs, ino_to_fsba(info->fs, ip->i_number)),
754 (int)info->fs->e2fs_bsize, &bp);
755 if (error) {
756 brelse(bp);
757 return (error);
758 }
759
760 error = ext2_ei2i((struct ext2fs_dinode *)((char *)bp->b_data +
761 EXT2_INODE_SIZE(info->fs) * ino_to_fsbo(info->fs, ip->i_number)),
762 ip);
763
764 brelse(bp);
765 return (error);
766 }
767
768 /*
769 * Reload all incore data for a filesystem (used after running fsck on
770 * the root filesystem and finding things to fix). The filesystem must
771 * be mounted read-only.
772 *
773 * Things to do to update the mount:
774 * 1) invalidate all cached meta-data.
775 * 2) re-read superblock from disk.
776 * 3) invalidate all cluster summary information.
777 * 4) invalidate all inactive vnodes.
778 * 5) invalidate all cached file data.
779 * 6) re-read inode data for all active vnodes.
780 * XXX we are missing some steps, in particular # 3, this has to be reviewed.
781 */
782 static int
ext2_reload(struct mount * mp)783 ext2_reload(struct mount *mp)
784 {
785 struct vnode *devvp;
786 struct buf *bp;
787 struct ext2fs *es;
788 struct m_ext2fs *fs;
789 struct csum *sump;
790 struct scaninfo scaninfo;
791 int error, i;
792 int32_t *lp;
793
794 if ((mp->mnt_flag & MNT_RDONLY) == 0)
795 return (EINVAL);
796 /*
797 * Step 1: invalidate all cached meta-data.
798 */
799 devvp = VFSTOEXT2(mp)->um_devvp;
800 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
801 if (vinvalbuf(devvp, 0, 0, 0) != 0)
802 panic("ext2_reload: dirty1");
803 vn_unlock(devvp);
804
805 /*
806 * Step 2: re-read superblock from disk.
807 * constants have been adjusted for ext2
808 */
809 if ((error = bread(devvp, SBOFF, SBSIZE, &bp)) != 0) {
810 brelse(bp);
811 return (error);
812 }
813 es = (struct ext2fs *)bp->b_data;
814 if (ext2_check_sb_compat(es, devvp->v_rdev, 0) != 0) {
815 brelse(bp);
816 return (EIO); /* XXX needs translation */
817 }
818 fs = VFSTOEXT2(mp)->um_e2fs;
819 bcopy(bp->b_data, fs->e2fs, sizeof(struct ext2fs));
820
821 if ((error = ext2_compute_sb_data(devvp, es, fs)) != 0) {
822 brelse(bp);
823 return (error);
824 }
825 #ifdef UNKLAR
826 if (fs->fs_sbsize < SBSIZE)
827 bp->b_flags |= B_INVAL;
828 #endif
829 brelse(bp);
830
831 /*
832 * Step 3: invalidate all cluster summary information.
833 */
834 if (fs->e2fs_contigsumsize > 0) {
835 lp = fs->e2fs_maxcluster;
836 sump = fs->e2fs_clustersum;
837 for (i = 0; i < fs->e2fs_gcount; i++, sump++) {
838 *lp++ = fs->e2fs_contigsumsize;
839 sump->cs_init = 0;
840 bzero(sump->cs_sum, fs->e2fs_contigsumsize + 1);
841 }
842 }
843
844 scaninfo.rescan = 1;
845 scaninfo.devvp = devvp;
846 scaninfo.fs = fs;
847 while (error == 0 && scaninfo.rescan) {
848 scaninfo.rescan = 0;
849 error = vmntvnodescan(mp, VMSC_GETVX, NULL, ext2_reload_scan,
850 &scaninfo);
851 }
852 return (error);
853 }
854
855 /*
856 * Common code for mount and mountroot.
857 */
858 static int
ext2_mountfs(struct vnode * devvp,struct mount * mp)859 ext2_mountfs(struct vnode *devvp, struct mount *mp)
860 {
861 struct ext2mount *ump;
862 struct buf *bp;
863 struct m_ext2fs *fs;
864 struct ext2fs *es;
865 struct cdev *dev = devvp->v_rdev;
866 struct csum *sump;
867 int error;
868 int ronly;
869 int i;
870 u_long size;
871 int32_t *lp;
872 int32_t e2fs_maxcontig;
873
874 /*
875 * Disallow multiple mounts of the same device.
876 * Disallow mounting of a device that is currently in use
877 * (except for root, which might share swap device for miniroot).
878 * Flush out any old buffers remaining from a previous use.
879 */
880 if ((error = vfs_mountedon(devvp)) != 0)
881 return (error);
882 if (vcount(devvp) > 0)
883 return (EBUSY);
884 if ((error = vinvalbuf(devvp, V_SAVE, 0, 0)) != 0)
885 return (error);
886 #ifdef READONLY
887 /* Turn on this to force it to be read-only. */
888 mp->mnt_flag |= MNT_RDONLY;
889 #endif
890 ronly = (mp->mnt_flag & MNT_RDONLY) != 0;
891 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
892 error = VOP_OPEN(devvp, ronly ? FREAD : FREAD | FWRITE, FSCRED, NULL);
893 vn_unlock(devvp);
894 if (error)
895 return (error);
896
897 if (devvp->v_rdev->si_iosize_max != 0)
898 mp->mnt_iosize_max = devvp->v_rdev->si_iosize_max;
899 if (mp->mnt_iosize_max > MAXPHYS)
900 mp->mnt_iosize_max = MAXPHYS;
901
902 bp = NULL;
903 ump = NULL;
904 if ((error = bread(devvp, SBOFF, SBSIZE, &bp)) != 0)
905 goto out;
906 es = (struct ext2fs *)bp->b_data;
907 if (ext2_check_sb_compat(es, dev, ronly) != 0) {
908 error = EINVAL; /* XXX needs translation */
909 goto out;
910 }
911 if ((le16toh(es->e2fs_state) & E2FS_ISCLEAN) == 0 ||
912 (le16toh(es->e2fs_state) & E2FS_ERRORS)) {
913 if (ronly || (mp->mnt_flag & MNT_FORCE)) {
914 printf(
915 "WARNING: Filesystem was not properly dismounted\n");
916 } else {
917 printf(
918 "WARNING: R/W mount denied. Filesystem is not clean - run fsck\n");
919 error = EPERM;
920 goto out;
921 }
922 }
923 ump = malloc(sizeof(*ump), M_EXT2MNT, M_WAITOK | M_ZERO);
924
925 /*
926 * I don't know whether this is the right strategy. Note that
927 * we dynamically allocate both an m_ext2fs and an ext2fs
928 * while Linux keeps the super block in a locked buffer.
929 */
930 ump->um_e2fs = malloc(sizeof(struct m_ext2fs),
931 M_EXT2MNT, M_WAITOK | M_ZERO);
932 ump->um_e2fs->e2fs = malloc(sizeof(struct ext2fs),
933 M_EXT2MNT, M_WAITOK);
934 mtx_init(EXT2_MTX(ump), "EXT2FS Lock");
935 bcopy(es, ump->um_e2fs->e2fs, (u_int)sizeof(struct ext2fs));
936 if ((error = ext2_compute_sb_data(devvp, ump->um_e2fs->e2fs, ump->um_e2fs)))
937 goto out;
938
939 /*
940 * Calculate the maximum contiguous blocks and size of cluster summary
941 * array. In FFS this is done by newfs; however, the superblock
942 * in ext2fs doesn't have these variables, so we can calculate
943 * them here.
944 */
945 e2fs_maxcontig = MAX(1, MAXPHYS / ump->um_e2fs->e2fs_bsize);
946 ump->um_e2fs->e2fs_contigsumsize = MIN(e2fs_maxcontig, EXT2_MAXCONTIG);
947 if (ump->um_e2fs->e2fs_contigsumsize > 0) {
948 size = ump->um_e2fs->e2fs_gcount * sizeof(int32_t);
949 ump->um_e2fs->e2fs_maxcluster = malloc(size, M_EXT2MNT, M_WAITOK);
950 size = ump->um_e2fs->e2fs_gcount * sizeof(struct csum);
951 ump->um_e2fs->e2fs_clustersum = malloc(size, M_EXT2MNT, M_WAITOK);
952 lp = ump->um_e2fs->e2fs_maxcluster;
953 sump = ump->um_e2fs->e2fs_clustersum;
954 for (i = 0; i < ump->um_e2fs->e2fs_gcount; i++, sump++) {
955 *lp++ = ump->um_e2fs->e2fs_contigsumsize;
956 sump->cs_init = 0;
957 sump->cs_sum = malloc((ump->um_e2fs->e2fs_contigsumsize + 1) *
958 sizeof(int32_t), M_EXT2MNT, M_WAITOK | M_ZERO);
959 }
960 }
961
962 brelse(bp);
963 bp = NULL;
964 fs = ump->um_e2fs;
965 fs->e2fs_ronly = ronly; /* ronly is set according to mnt_flags */
966
967 /*
968 * If the fs is not mounted read-only, make sure the super block is
969 * always written back on a sync().
970 */
971 fs->e2fs_wasvalid = le16toh(fs->e2fs->e2fs_state) & E2FS_ISCLEAN ? 1 : 0;
972 if (ronly == 0) {
973 fs->e2fs_fmod = 1; /* mark it modified and set fs invalid */
974 fs->e2fs->e2fs_state =
975 htole16(le16toh(fs->e2fs->e2fs_state) & ~E2FS_ISCLEAN);
976 }
977 mp->mnt_data = (qaddr_t)ump;
978 mp->mnt_stat.f_fsid.val[0] = devid_from_dev(dev);
979 mp->mnt_stat.f_fsid.val[1] = mp->mnt_vfc->vfc_typenum;
980 mp->mnt_maxsymlinklen = EXT2_MAXSYMLINKLEN;
981 mp->mnt_flag |= MNT_LOCAL;
982 mp->mnt_kern_flag |= MNTK_ALL_MPSAFE;
983 ump->um_mountp = mp;
984 ump->um_dev = dev;
985 ump->um_devvp = devvp;
986
987 /*
988 * Setting those two parameters allowed us to use
989 * ufs_bmap w/o changse!
990 */
991 ump->um_nindir = EXT2_ADDR_PER_BLOCK(fs);
992 ump->um_bptrtodb = le32toh(fs->e2fs->e2fs_log_bsize) + 1;
993 ump->um_seqinc = EXT2_FRAGS_PER_BLOCK(fs);
994 dev->si_mountpoint = mp;
995
996 vfs_add_vnodeops(mp, &ext2_vnodeops, &mp->mnt_vn_norm_ops);
997 vfs_add_vnodeops(mp, &ext2_specops, &mp->mnt_vn_spec_ops);
998 vfs_add_vnodeops(mp, &ext2_fifoops, &mp->mnt_vn_fifo_ops);
999
1000 if (ronly == 0)
1001 ext2_sbupdate(ump, MNT_WAIT);
1002 return (0);
1003 out:
1004 if (bp)
1005 brelse(bp);
1006 vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
1007 VOP_CLOSE(devvp, ronly ? FREAD : FREAD | FWRITE, NULL);
1008 vn_unlock(devvp);
1009 if (ump) {
1010 mtx_uninit(EXT2_MTX(ump));
1011 free(ump->um_e2fs->e2fs_gd, M_EXT2MNT);
1012 free(ump->um_e2fs->e2fs_contigdirs, M_EXT2MNT);
1013 free(ump->um_e2fs->e2fs, M_EXT2MNT);
1014 free(ump->um_e2fs, M_EXT2MNT);
1015 free(ump, M_EXT2MNT);
1016 mp->mnt_data = NULL;
1017 }
1018 return (error);
1019 }
1020
1021 /*
1022 * Unmount system call.
1023 */
1024 static int
ext2_unmount(struct mount * mp,int mntflags)1025 ext2_unmount(struct mount *mp, int mntflags)
1026 {
1027 struct ext2mount *ump;
1028 struct m_ext2fs *fs;
1029 struct csum *sump;
1030 int error, flags, i, ronly;
1031
1032 flags = 0;
1033 if (mntflags & MNT_FORCE) {
1034 if (mp->mnt_flag & MNT_ROOTFS)
1035 return (EINVAL);
1036 flags |= FORCECLOSE;
1037 }
1038 if ((error = ext2_flushfiles(mp, flags)) != 0)
1039 return (error);
1040 ump = VFSTOEXT2(mp);
1041 fs = ump->um_e2fs;
1042 ronly = fs->e2fs_ronly;
1043 if (ronly == 0 && ext2_cgupdate(ump, MNT_WAIT) == 0) {
1044 if (fs->e2fs_wasvalid)
1045 fs->e2fs->e2fs_state =
1046 htole16(le16toh(fs->e2fs->e2fs_state) | E2FS_ISCLEAN);
1047 ext2_sbupdate(ump, MNT_WAIT);
1048 }
1049
1050 ump->um_devvp->v_rdev->si_mountpoint = NULL;
1051
1052 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
1053 error = VOP_CLOSE(ump->um_devvp, ronly ? FREAD : FREAD | FWRITE, NULL);
1054 vn_unlock(ump->um_devvp);
1055
1056 vrele(ump->um_devvp);
1057 sump = fs->e2fs_clustersum;
1058 for (i = 0; i < fs->e2fs_gcount; i++, sump++)
1059 free(sump->cs_sum, M_EXT2MNT);
1060 free(fs->e2fs_clustersum, M_EXT2MNT);
1061 free(fs->e2fs_maxcluster, M_EXT2MNT);
1062 free(fs->e2fs_gd, M_EXT2MNT);
1063 free(fs->e2fs_contigdirs, M_EXT2MNT);
1064 free(fs->e2fs, M_EXT2MNT);
1065 free(fs, M_EXT2MNT);
1066 free(ump, M_EXT2MNT);
1067 mp->mnt_data = NULL;
1068 mp->mnt_flag &= ~MNT_LOCAL;
1069 return (error);
1070 }
1071
1072 /*
1073 * Flush out all the files in a filesystem.
1074 */
1075 static int
ext2_flushfiles(struct mount * mp,int flags)1076 ext2_flushfiles(struct mount *mp, int flags)
1077 {
1078 int error;
1079
1080 error = vflush(mp, 0, flags);
1081 return (error);
1082 }
1083
1084 /*
1085 * Get filesystem statistics.
1086 */
1087 static int
ext2_statfs(struct mount * mp,struct statfs * sbp,struct ucred * cred)1088 ext2_statfs(struct mount *mp, struct statfs *sbp, struct ucred *cred)
1089 {
1090 struct ext2mount *ump;
1091 struct m_ext2fs *fs;
1092 uint32_t overhead, overhead_per_group, ngdb;
1093 int i, ngroups;
1094
1095 ump = VFSTOEXT2(mp);
1096 fs = ump->um_e2fs;
1097 if (le16toh(fs->e2fs->e2fs_magic) != E2FS_MAGIC)
1098 panic("ext2_statfs");
1099
1100 /*
1101 * Compute the overhead (FS structures)
1102 */
1103 overhead_per_group =
1104 1 /* block bitmap */ +
1105 1 /* inode bitmap */ +
1106 fs->e2fs_itpg;
1107 overhead = le32toh(fs->e2fs->e2fs_first_dblock) +
1108 fs->e2fs_gcount * overhead_per_group;
1109 if (le32toh(fs->e2fs->e2fs_rev) > E2FS_REV0 &&
1110 le32toh(fs->e2fs->e2fs_features_rocompat) & EXT2F_ROCOMPAT_SPARSESUPER) {
1111 for (i = 0, ngroups = 0; i < fs->e2fs_gcount; i++) {
1112 if (ext2_cg_has_sb(fs, i))
1113 ngroups++;
1114 }
1115 } else {
1116 ngroups = fs->e2fs_gcount;
1117 }
1118 ngdb = fs->e2fs_gdbcount;
1119 if (le32toh(fs->e2fs->e2fs_rev) > E2FS_REV0 &&
1120 le32toh(fs->e2fs->e2fs_features_compat) & EXT2F_COMPAT_RESIZE)
1121 ngdb += le16toh(fs->e2fs->e2fs_reserved_ngdb);
1122 overhead += ngroups * (1 /* superblock */ + ngdb);
1123
1124 sbp->f_type = mp->mnt_vfc->vfc_typenum;
1125 sbp->f_bsize = EXT2_FRAG_SIZE(fs);
1126 sbp->f_iosize = EXT2_BLOCK_SIZE(fs);
1127 sbp->f_blocks = fs->e2fs_bcount - overhead;
1128 sbp->f_bfree = fs->e2fs_fbcount;
1129 sbp->f_bavail = sbp->f_bfree - fs->e2fs_rbcount;
1130 sbp->f_files = le32toh(fs->e2fs->e2fs_icount);
1131 sbp->f_ffree = fs->e2fs_ficount;
1132 if (sbp != &mp->mnt_stat) {
1133 bcopy((caddr_t)mp->mnt_stat.f_mntfromname,
1134 (caddr_t)&sbp->f_mntfromname[0], MNAMELEN);
1135 }
1136 return (0);
1137 }
1138
1139 static int
ext2_statvfs(struct mount * mp,struct statvfs * sbp,struct ucred * cred)1140 ext2_statvfs(struct mount *mp, struct statvfs *sbp, struct ucred *cred)
1141 {
1142 struct ext2mount *ump;
1143 struct m_ext2fs *fs;
1144 uint32_t overhead, overhead_per_group, ngdb;
1145 int i, ngroups;
1146
1147 ump = VFSTOEXT2(mp);
1148 fs = ump->um_e2fs;
1149 if (le16toh(fs->e2fs->e2fs_magic) != E2FS_MAGIC)
1150 panic("ext2_statfs");
1151
1152 /*
1153 * Compute the overhead (FS structures)
1154 */
1155 overhead_per_group =
1156 1 /* block bitmap */ +
1157 1 /* inode bitmap */ +
1158 fs->e2fs_itpg;
1159 overhead = le32toh(fs->e2fs->e2fs_first_dblock) +
1160 fs->e2fs_gcount * overhead_per_group;
1161 if (le32toh(fs->e2fs->e2fs_rev) > E2FS_REV0 &&
1162 le32toh(fs->e2fs->e2fs_features_rocompat) & EXT2F_ROCOMPAT_SPARSESUPER) {
1163 for (i = 0, ngroups = 0; i < fs->e2fs_gcount; i++) {
1164 if (ext2_cg_has_sb(fs, i))
1165 ngroups++;
1166 }
1167 } else {
1168 ngroups = fs->e2fs_gcount;
1169 }
1170 ngdb = fs->e2fs_gdbcount;
1171 if (le32toh(fs->e2fs->e2fs_rev) > E2FS_REV0 &&
1172 le32toh(fs->e2fs->e2fs_features_compat) & EXT2F_COMPAT_RESIZE)
1173 ngdb += le16toh(fs->e2fs->e2fs_reserved_ngdb);
1174 overhead += ngroups * (1 /* superblock */ + ngdb);
1175
1176 sbp->f_type = mp->mnt_vfc->vfc_typenum;
1177 sbp->f_bsize = EXT2_FRAG_SIZE(fs);
1178 sbp->f_frsize = EXT2_BLOCK_SIZE(fs);
1179 sbp->f_blocks = fs->e2fs_bcount - overhead;
1180 sbp->f_bfree = fs->e2fs_fbcount;
1181 sbp->f_bavail = sbp->f_bfree - fs->e2fs_rbcount;
1182 sbp->f_files = le32toh(fs->e2fs->e2fs_icount);
1183 sbp->f_ffree = fs->e2fs_ficount;
1184 return (0);
1185 }
1186
1187 static int
ext2_sync_scan(struct mount * mp,struct vnode * vp,void * data)1188 ext2_sync_scan(struct mount *mp, struct vnode *vp, void *data)
1189 {
1190 struct scaninfo *info = data;
1191 struct inode *ip;
1192 int error;
1193
1194 ip = VTOI(vp);
1195 if (vp->v_type == VNON ||
1196 ((ip->i_flag &
1197 (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) == 0 &&
1198 (RB_EMPTY(&vp->v_rbdirty_tree) || (info->waitfor & MNT_LAZY)))) {
1199 return (0);
1200 }
1201 if ((error = VOP_FSYNC(vp, info->waitfor, 0)) != 0)
1202 info->allerror = error;
1203 return (0);
1204 }
1205
1206 /*
1207 * Go through the disk queues to initiate sandbagged IO;
1208 * go through the inodes to write those that have been modified;
1209 * initiate the writing of the super block if it has been modified.
1210 *
1211 * Note: we are always called with the filesystem marked `MPBUSY'.
1212 */
1213 static int
ext2_sync(struct mount * mp,int waitfor)1214 ext2_sync(struct mount *mp, int waitfor)
1215 {
1216 struct ext2mount *ump = VFSTOEXT2(mp);
1217 struct m_ext2fs *fs;
1218 struct scaninfo scaninfo;
1219 int error;
1220
1221 fs = ump->um_e2fs;
1222 if (fs->e2fs_fmod != 0 && fs->e2fs_ronly != 0) { /* XXX */
1223 panic("ext2_sync: rofs mod fs=%s", fs->e2fs_fsmnt);
1224 }
1225
1226 /*
1227 * Write back each (modified) inode.
1228 */
1229 scaninfo.allerror = 0;
1230 scaninfo.rescan = 1;
1231 scaninfo.waitfor = waitfor;
1232 while (scaninfo.rescan) {
1233 scaninfo.rescan = 0;
1234 vmntvnodescan(mp, VMSC_GETVP | VMSC_NOWAIT,
1235 NULL, ext2_sync_scan, &scaninfo);
1236 }
1237
1238 /*
1239 * Force stale filesystem control information to be flushed.
1240 */
1241 if ((waitfor & MNT_LAZY) == 0) {
1242 vn_lock(ump->um_devvp, LK_EXCLUSIVE | LK_RETRY);
1243 if ((error = VOP_FSYNC(ump->um_devvp, waitfor, 0)) != 0)
1244 scaninfo.allerror = error;
1245 vn_unlock(ump->um_devvp);
1246 }
1247
1248 /*
1249 * Write back modified superblock.
1250 */
1251 if (fs->e2fs_fmod != 0) {
1252 fs->e2fs_fmod = 0;
1253 fs->e2fs->e2fs_wtime = htole32(time_second);
1254 if ((error = ext2_cgupdate(ump, waitfor)) != 0)
1255 scaninfo.allerror = error;
1256 }
1257 return (scaninfo.allerror);
1258 }
1259
1260 int
ext2_alloc_vnode(struct mount * mp,ino_t ino,struct vnode ** vpp)1261 ext2_alloc_vnode(struct mount *mp, ino_t ino, struct vnode **vpp)
1262 {
1263 struct ext2mount *ump;
1264 struct vnode *vp;
1265 struct inode *ip;
1266 int error;
1267
1268 ump = VFSTOEXT2(mp);
1269 /*
1270 * Lock out the creation of new entries in the FFS hash table in
1271 * case getnewvnode() or MALLOC() blocks, otherwise a duplicate
1272 * may occur!
1273 */
1274 if (ext2fs_inode_hash_lock) {
1275 while (ext2fs_inode_hash_lock) {
1276 ext2fs_inode_hash_lock = -1;
1277 tsleep(&ext2fs_inode_hash_lock, 0, "e2vget", 0);
1278 }
1279 return (-1);
1280 }
1281 ext2fs_inode_hash_lock = 1;
1282
1283 ip = malloc(sizeof(struct inode), M_EXT2NODE, M_WAITOK | M_ZERO);
1284
1285 /* Allocate a new vnode/inode. */
1286 if ((error = getnewvnode(VT_EXT2FS, mp, &vp, VLKTIMEOUT,
1287 LK_CANRECURSE)) != 0) {
1288 if (ext2fs_inode_hash_lock < 0)
1289 wakeup(&ext2fs_inode_hash_lock);
1290 ext2fs_inode_hash_lock = 0;
1291 *vpp = NULL;
1292 free(ip, M_EXT2NODE);
1293 return (error);
1294 }
1295 //lockmgr(vp->v_vnlock, LK_EXCLUSIVE, NULL);
1296 vp->v_data = ip;
1297 ip->i_vnode = vp;
1298 ip->i_e2fs = ump->um_e2fs;
1299 ip->i_dev = ump->um_dev;
1300 ip->i_ump = ump;
1301 ip->i_number = ino;
1302 ip->i_block_group = ino_to_cg(ip->i_e2fs, ino);
1303 ip->i_next_alloc_block = 0;
1304 ip->i_next_alloc_goal = 0;
1305
1306 /*
1307 * Put it onto its hash chain. Since our vnode is locked, other
1308 * requests for this inode will block if they arrive while we are
1309 * sleeping waiting for old data structures to be purged or for the
1310 * contents of the disk portion of this inode to be read.
1311 */
1312 if (ext2_ihashins(ip)) {
1313 printf("ext2_alloc_vnode: ihashins collision, retrying inode %ld\n",
1314 (long)ip->i_number);
1315 *vpp = NULL;
1316 vp->v_type = VBAD;
1317 vx_put(vp);
1318 free(ip, M_EXT2NODE);
1319 return (-1);
1320 }
1321
1322 if (ext2fs_inode_hash_lock < 0)
1323 wakeup(&ext2fs_inode_hash_lock);
1324 ext2fs_inode_hash_lock = 0;
1325 *vpp = vp;
1326
1327 return (0);
1328 }
1329
1330 /*
1331 * Look up an EXT2FS dinode number to find its incore vnode, otherwise read it
1332 * in from disk. If it is in core, wait for the lock bit to clear, then
1333 * return the inode locked. Detection and handling of mount points must be
1334 * done by the calling routine.
1335 */
1336 static int
ext2_vget(struct mount * mp,struct vnode * dvp,ino_t ino,struct vnode ** vpp)1337 ext2_vget(struct mount *mp, struct vnode *dvp, ino_t ino, struct vnode **vpp)
1338 {
1339 struct m_ext2fs *fs;
1340 struct inode *ip;
1341 struct ext2mount *ump;
1342 struct buf *bp;
1343 struct vnode *vp;
1344 unsigned int i, used_blocks;
1345 int error;
1346
1347 ump = VFSTOEXT2(mp);
1348 restart:
1349 if ((*vpp = ext2_ihashget(ump->um_dev, ino)) != NULL)
1350 return (0);
1351 if (ext2_alloc_vnode(mp, ino, &vp) == -1)
1352 goto restart;
1353 ip = VTOI(vp);
1354 fs = ip->i_e2fs;
1355
1356 /* Read in the disk contents for the inode, copy into the inode. */
1357 if ((error = bread(ump->um_devvp, fsbtodoff(fs, ino_to_fsba(fs, ino)),
1358 (int)fs->e2fs_bsize, &bp)) != 0) {
1359 /*
1360 * The inode does not contain anything useful, so it would
1361 * be misleading to leave it on its hash chain. With mode
1362 * still zero, it will be unlinked and returned to the free
1363 * list by vput().
1364 */
1365 vp->v_type = VBAD;
1366 brelse(bp);
1367 vx_put(vp);
1368 *vpp = NULL;
1369 return (error);
1370 }
1371 /* convert ext2 inode to dinode */
1372 error = ext2_ei2i((struct ext2fs_dinode *)((char *)bp->b_data +
1373 EXT2_INODE_SIZE(fs) * ino_to_fsbo(fs, ino)), ip);
1374 if (error) {
1375 brelse(bp);
1376 vx_put(vp);
1377 *vpp = NULL;
1378 return (error);
1379 }
1380
1381 /*
1382 * Now we want to make sure that block pointers for unused
1383 * blocks are zeroed out - ext2_balloc depends on this
1384 * although for regular files and directories only
1385 *
1386 * If IN_E4EXTENTS is enabled, unused blocks are not zeroed
1387 * out because we could corrupt the extent tree.
1388 */
1389 if (!(ip->i_flag & IN_E4EXTENTS) &&
1390 (S_ISDIR(ip->i_mode) || S_ISREG(ip->i_mode))) {
1391 used_blocks = howmany(ip->i_size, fs->e2fs_bsize);
1392 for (i = used_blocks; i < EXT2_NDIR_BLOCKS; i++)
1393 ip->i_db[i] = 0;
1394 }
1395 #ifdef EXT2FS_PRINT_EXTENTS
1396 ext2_print_inode(ip);
1397 ext4_ext_print_extent_tree_status(ip);
1398 #endif
1399 bqrelse(bp);
1400
1401 /*
1402 * Initialize the vnode from the inode, check for aliases.
1403 * Note that the underlying vnode may have changed.
1404 */
1405 if ((error = ext2_vinit(mp, &vp)) != 0) {
1406 vx_put(vp);
1407 *vpp = NULL;
1408 return (error);
1409 }
1410
1411 /*
1412 * Finish inode initialization now that aliasing has been resolved.
1413 */
1414 vref(ip->i_devvp);
1415 /*
1416 * Set up a generation number for this inode if it does not
1417 * already have one. This should only happen on old filesystems.
1418 */
1419 if (ip->i_gen == 0) {
1420 ip->i_gen = krandom() / 2 + 1;
1421 if ((vp->v_mount->mnt_flag & MNT_RDONLY) == 0)
1422 ip->i_flag |= IN_MODIFIED;
1423 }
1424 /*
1425 * Return the locked and refd vnode.
1426 */
1427 vx_downgrade(vp); /* downgrade VX lock to VN lock */
1428 *vpp = vp;
1429
1430 return (0);
1431 }
1432
1433 /*
1434 * File handle to vnode
1435 *
1436 * Have to be really careful about stale file handles:
1437 * - check that the inode number is valid
1438 * - call ext2_vget() to get the locked inode
1439 * - check for an unallocated inode (i_mode == 0)
1440 * - check that the given client host has export rights and return
1441 * those rights via. exflagsp and credanonp
1442 */
1443 static int
ext2_fhtovp(struct mount * mp,struct vnode * rootvp,struct fid * fhp,struct vnode ** vpp)1444 ext2_fhtovp(struct mount *mp, struct vnode *rootvp, struct fid *fhp,
1445 struct vnode **vpp)
1446 {
1447 struct inode *ip;
1448 struct ufid *ufhp;
1449 struct vnode *nvp;
1450 struct m_ext2fs *fs;
1451 int error;
1452
1453 ufhp = (struct ufid *)fhp;
1454 fs = VFSTOEXT2(mp)->um_e2fs;
1455 if (ufhp->ufid_ino < EXT2_ROOTINO ||
1456 ufhp->ufid_ino > fs->e2fs_gcount * fs->e2fs_ipg)
1457 return (ESTALE);
1458
1459 error = VFS_VGET(mp, NULL, ufhp->ufid_ino, &nvp);
1460 if (error) {
1461 *vpp = NULLVP;
1462 return (error);
1463 }
1464 ip = VTOI(nvp);
1465 if (ip->i_mode == 0 ||
1466 ip->i_gen != ufhp->ufid_gen || ip->i_nlink <= 0) {
1467 vput(nvp);
1468 *vpp = NULLVP;
1469 return (ESTALE);
1470 }
1471 *vpp = nvp;
1472 return (0);
1473 }
1474
1475 /*
1476 * Vnode pointer to File handle
1477 */
1478 /* ARGSUSED */
1479 static int
ext2_vptofh(struct vnode * vp,struct fid * fhp)1480 ext2_vptofh(struct vnode *vp, struct fid *fhp)
1481 {
1482 struct inode *ip;
1483 struct ufid *ufhp;
1484
1485 ip = VTOI(vp);
1486 ufhp = (struct ufid *)fhp;
1487 ufhp->ufid_len = sizeof(struct ufid);
1488 ufhp->ufid_ino = ip->i_number;
1489 ufhp->ufid_gen = ip->i_gen;
1490 return (0);
1491 }
1492
1493 /*
1494 * This is the generic part of fhtovp called after the underlying
1495 * filesystem has validated the file handle.
1496 *
1497 * Verify that a host should have access to a filesystem.
1498 */
1499 static int
ext2_check_export(struct mount * mp,struct sockaddr * nam,int * exflagsp,struct ucred ** credanonp)1500 ext2_check_export(struct mount *mp, struct sockaddr *nam, int *exflagsp,
1501 struct ucred **credanonp)
1502 {
1503 struct netcred *np;
1504 struct ext2mount *ump;
1505
1506 ump = VFSTOEXT2(mp);
1507 /*
1508 * Get the export permission structure for this <mp, client> tuple.
1509 */
1510 np = vfs_export_lookup(mp, &ump->um_export, nam);
1511 if (np == NULL)
1512 return (EACCES);
1513
1514 *exflagsp = np->netc_exflags;
1515 *credanonp = &np->netc_anon;
1516 return (0);
1517 }
1518
1519 /*
1520 * Write a superblock and associated information back to disk.
1521 */
1522 static int
ext2_sbupdate(struct ext2mount * mp,int waitfor)1523 ext2_sbupdate(struct ext2mount *mp, int waitfor)
1524 {
1525 struct m_ext2fs *fs = mp->um_e2fs;
1526 struct ext2fs *es = fs->e2fs;
1527 struct buf *bp;
1528 int error = 0;
1529
1530 es->e2fs_bcount = htole32(fs->e2fs_bcount & 0xffffffff);
1531 es->e2fs_rbcount = htole32(fs->e2fs_rbcount & 0xffffffff);
1532 es->e2fs_fbcount = htole32(fs->e2fs_fbcount & 0xffffffff);
1533 if (EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_64BIT)) {
1534 es->e4fs_bcount_hi = htole32(fs->e2fs_bcount >> 32);
1535 es->e4fs_rbcount_hi = htole32(fs->e2fs_rbcount >> 32);
1536 es->e4fs_fbcount_hi = htole32(fs->e2fs_fbcount >> 32);
1537 }
1538
1539 es->e2fs_ficount = htole32(fs->e2fs_ficount);
1540
1541 if (EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_METADATA_CKSUM))
1542 ext2_sb_csum_set(fs);
1543
1544 bp = getblk(mp->um_devvp, SBOFF, SBSIZE, 0, 0);
1545 bcopy((caddr_t)es, bp->b_data, (u_int)sizeof(struct ext2fs));
1546 if (waitfor == MNT_WAIT)
1547 error = bwrite(bp);
1548 else
1549 bawrite(bp);
1550
1551 /*
1552 * The buffers for group descriptors, inode bitmaps and block bitmaps
1553 * are not busy at this point and are (hopefully) written by the
1554 * usual sync mechanism. No need to write them here.
1555 */
1556 return (error);
1557 }
1558
1559 static int
ext2_cgupdate(struct ext2mount * mp,int waitfor)1560 ext2_cgupdate(struct ext2mount *mp, int waitfor)
1561 {
1562 struct m_ext2fs *fs = mp->um_e2fs;
1563 struct buf *bp;
1564 int i, j, g_count = 0, error = 0, allerror = 0;
1565
1566 allerror = ext2_sbupdate(mp, waitfor);
1567
1568 /* Update gd csums */
1569 if (EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_GDT_CSUM) ||
1570 EXT2_HAS_RO_COMPAT_FEATURE(fs, EXT2F_ROCOMPAT_METADATA_CKSUM))
1571 ext2_gd_csum_set(fs);
1572
1573 for (i = 0; i < fs->e2fs_gdbcount; i++) {
1574 bp = getblk(mp->um_devvp, fsbtodoff(fs,
1575 ext2_cg_location(fs, i)),
1576 fs->e2fs_bsize, 0, 0);
1577 if (EXT2_HAS_INCOMPAT_FEATURE(fs, EXT2F_INCOMPAT_64BIT)) {
1578 memcpy(bp->b_data, &fs->e2fs_gd[
1579 i * fs->e2fs_bsize / sizeof(struct ext2_gd)],
1580 fs->e2fs_bsize);
1581 } else {
1582 for (j = 0; j < fs->e2fs_bsize / E2FS_REV0_GD_SIZE &&
1583 g_count < fs->e2fs_gcount; j++, g_count++)
1584 memcpy(bp->b_data + j * E2FS_REV0_GD_SIZE,
1585 &fs->e2fs_gd[g_count], E2FS_REV0_GD_SIZE);
1586 }
1587 if (waitfor == MNT_WAIT)
1588 error = bwrite(bp);
1589 else
1590 bawrite(bp);
1591 }
1592
1593 if (!allerror && error)
1594 allerror = error;
1595 return (allerror);
1596 }
1597
1598 /*
1599 * Return the root of a filesystem.
1600 */
1601 static int
ext2_root(struct mount * mp,struct vnode ** vpp)1602 ext2_root(struct mount *mp, struct vnode **vpp)
1603 {
1604 struct vnode *nvp;
1605 int error;
1606
1607 error = VFS_VGET(mp, NULL, (ino_t)EXT2_ROOTINO, &nvp);
1608 if (error)
1609 return (error);
1610 *vpp = nvp;
1611 return (0);
1612 }
1613
1614 /*
1615 * Initialize ext2 filesystems, done only once.
1616 */
1617 static int
ext2_init(struct vfsconf * vfsp)1618 ext2_init(struct vfsconf *vfsp)
1619 {
1620 static int done;
1621
1622 if (done)
1623 return (0);
1624 done = 1;
1625 ext2_ihashinit();
1626
1627 return (0);
1628 }
1629
1630 static int
ext2_uninit(struct vfsconf * vfsp)1631 ext2_uninit(struct vfsconf *vfsp)
1632 {
1633
1634 ext2_ihashuninit();
1635
1636 return (0);
1637 }
1638