?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Linux 2.6.28.1
[linux/fpc-iii.git] / fs / ext3 / super.c
blobf6c94f232ec1e8fee05e47dfcc54b234c82e7cf0
1 /*
2 * linux/fs/ext3/super.c
3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
19 #include <linux/module.h>
20 #include <linux/string.h>
21 #include <linux/fs.h>
22 #include <linux/time.h>
23 #include <linux/jbd.h>
24 #include <linux/ext3_fs.h>
25 #include <linux/ext3_jbd.h>
26 #include <linux/slab.h>
27 #include <linux/init.h>
28 #include <linux/blkdev.h>
29 #include <linux/parser.h>
30 #include <linux/smp_lock.h>
31 #include <linux/buffer_head.h>
32 #include <linux/exportfs.h>
33 #include <linux/vfs.h>
34 #include <linux/random.h>
35 #include <linux/mount.h>
36 #include <linux/namei.h>
37 #include <linux/quotaops.h>
38 #include <linux/seq_file.h>
39 #include <linux/log2.h>
40
41 #include <asm/uaccess.h>
42
43 #include "xattr.h"
44 #include "acl.h"
45 #include "namei.h"
46
47 static int ext3_load_journal(struct super_block *, struct ext3_super_block *,
48 unsigned long journal_devnum);
49 static int ext3_create_journal(struct super_block *, struct ext3_super_block *,
50 unsigned int);
51 static void ext3_commit_super (struct super_block * sb,
52 struct ext3_super_block * es,
53 int sync);
54 static void ext3_mark_recovery_complete(struct super_block * sb,
55 struct ext3_super_block * es);
56 static void ext3_clear_journal_err(struct super_block * sb,
57 struct ext3_super_block * es);
58 static int ext3_sync_fs(struct super_block *sb, int wait);
59 static const char *ext3_decode_error(struct super_block * sb, int errno,
60 char nbuf[16]);
61 static int ext3_remount (struct super_block * sb, int * flags, char * data);
62 static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf);
63 static void ext3_unlockfs(struct super_block *sb);
64 static void ext3_write_super (struct super_block * sb);
65 static void ext3_write_super_lockfs(struct super_block *sb);
66
67 /*
68 * Wrappers for journal_start/end.
69 *
70 * The only special thing we need to do here is to make sure that all
71 * journal_end calls result in the superblock being marked dirty, so
72 * that sync() will call the filesystem's write_super callback if
73 * appropriate.
74 */
75 handle_t *ext3_journal_start_sb(struct super_block *sb, int nblocks)
76 {
77 journal_t *journal;
78
79 if (sb->s_flags & MS_RDONLY)
80 return ERR_PTR(-EROFS);
81
82 /* Special case here: if the journal has aborted behind our
83 * backs (eg. EIO in the commit thread), then we still need to
84 * take the FS itself readonly cleanly. */
85 journal = EXT3_SB(sb)->s_journal;
86 if (is_journal_aborted(journal)) {
87 ext3_abort(sb, __func__,
88 "Detected aborted journal");
89 return ERR_PTR(-EROFS);
90 }
91
92 return journal_start(journal, nblocks);
93 }
94
95 /*
96 * The only special thing we need to do here is to make sure that all
97 * journal_stop calls result in the superblock being marked dirty, so
98 * that sync() will call the filesystem's write_super callback if
99 * appropriate.
100 */
101 int __ext3_journal_stop(const char *where, handle_t *handle)
102 {
103 struct super_block *sb;
104 int err;
105 int rc;
106
107 sb = handle->h_transaction->t_journal->j_private;
108 err = handle->h_err;
109 rc = journal_stop(handle);
110
111 if (!err)
112 err = rc;
113 if (err)
114 __ext3_std_error(sb, where, err);
115 return err;
116 }
117
118 void ext3_journal_abort_handle(const char *caller, const char *err_fn,
119 struct buffer_head *bh, handle_t *handle, int err)
120 {
121 char nbuf[16];
122 const char *errstr = ext3_decode_error(NULL, err, nbuf);
123
124 if (bh)
125 BUFFER_TRACE(bh, "abort");
126
127 if (!handle->h_err)
128 handle->h_err = err;
129
130 if (is_handle_aborted(handle))
131 return;
132
133 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
134 caller, errstr, err_fn);
135
136 journal_abort_handle(handle);
137 }
138
139 /* Deal with the reporting of failure conditions on a filesystem such as
140 * inconsistencies detected or read IO failures.
141 *
142 * On ext2, we can store the error state of the filesystem in the
143 * superblock. That is not possible on ext3, because we may have other
144 * write ordering constraints on the superblock which prevent us from
145 * writing it out straight away; and given that the journal is about to
146 * be aborted, we can't rely on the current, or future, transactions to
147 * write out the superblock safely.
148 *
149 * We'll just use the journal_abort() error code to record an error in
150 * the journal instead. On recovery, the journal will compain about
151 * that error until we've noted it down and cleared it.
152 */
153
154 static void ext3_handle_error(struct super_block *sb)
155 {
156 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
157
158 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
159 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
160
161 if (sb->s_flags & MS_RDONLY)
162 return;
163
164 if (!test_opt (sb, ERRORS_CONT)) {
165 journal_t *journal = EXT3_SB(sb)->s_journal;
166
167 EXT3_SB(sb)->s_mount_opt |= EXT3_MOUNT_ABORT;
168 if (journal)
169 journal_abort(journal, -EIO);
170 }
171 if (test_opt (sb, ERRORS_RO)) {
172 printk (KERN_CRIT "Remounting filesystem read-only\n");
173 sb->s_flags |= MS_RDONLY;
174 }
175 ext3_commit_super(sb, es, 1);
176 if (test_opt(sb, ERRORS_PANIC))
177 panic("EXT3-fs (device %s): panic forced after error\n",
178 sb->s_id);
179 }
180
181 void ext3_error (struct super_block * sb, const char * function,
182 const char * fmt, ...)
183 {
184 va_list args;
185
186 va_start(args, fmt);
187 printk(KERN_CRIT "EXT3-fs error (device %s): %s: ",sb->s_id, function);
188 vprintk(fmt, args);
189 printk("\n");
190 va_end(args);
191
192 ext3_handle_error(sb);
193 }
194
195 static const char *ext3_decode_error(struct super_block * sb, int errno,
196 char nbuf[16])
197 {
198 char *errstr = NULL;
199
200 switch (errno) {
201 case -EIO:
202 errstr = "IO failure";
203 break;
204 case -ENOMEM:
205 errstr = "Out of memory";
206 break;
207 case -EROFS:
208 if (!sb || EXT3_SB(sb)->s_journal->j_flags & JFS_ABORT)
209 errstr = "Journal has aborted";
210 else
211 errstr = "Readonly filesystem";
212 break;
213 default:
214 /* If the caller passed in an extra buffer for unknown
215 * errors, textualise them now. Else we just return
216 * NULL. */
217 if (nbuf) {
218 /* Check for truncated error codes... */
219 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
220 errstr = nbuf;
221 }
222 break;
223 }
224
225 return errstr;
226 }
227
228 /* __ext3_std_error decodes expected errors from journaling functions
229 * automatically and invokes the appropriate error response. */
230
231 void __ext3_std_error (struct super_block * sb, const char * function,
232 int errno)
233 {
234 char nbuf[16];
235 const char *errstr;
236
237 /* Special case: if the error is EROFS, and we're not already
238 * inside a transaction, then there's really no point in logging
239 * an error. */
240 if (errno == -EROFS && journal_current_handle() == NULL &&
241 (sb->s_flags & MS_RDONLY))
242 return;
243
244 errstr = ext3_decode_error(sb, errno, nbuf);
245 printk (KERN_CRIT "EXT3-fs error (device %s) in %s: %s\n",
246 sb->s_id, function, errstr);
247
248 ext3_handle_error(sb);
249 }
250
251 /*
252 * ext3_abort is a much stronger failure handler than ext3_error. The
253 * abort function may be used to deal with unrecoverable failures such
254 * as journal IO errors or ENOMEM at a critical moment in log management.
255 *
256 * We unconditionally force the filesystem into an ABORT|READONLY state,
257 * unless the error response on the fs has been set to panic in which
258 * case we take the easy way out and panic immediately.
259 */
260
261 void ext3_abort (struct super_block * sb, const char * function,
262 const char * fmt, ...)
263 {
264 va_list args;
265
266 printk (KERN_CRIT "ext3_abort called.\n");
267
268 va_start(args, fmt);
269 printk(KERN_CRIT "EXT3-fs error (device %s): %s: ",sb->s_id, function);
270 vprintk(fmt, args);
271 printk("\n");
272 va_end(args);
273
274 if (test_opt(sb, ERRORS_PANIC))
275 panic("EXT3-fs panic from previous error\n");
276
277 if (sb->s_flags & MS_RDONLY)
278 return;
279
280 printk(KERN_CRIT "Remounting filesystem read-only\n");
281 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
282 sb->s_flags |= MS_RDONLY;
283 EXT3_SB(sb)->s_mount_opt |= EXT3_MOUNT_ABORT;
284 if (EXT3_SB(sb)->s_journal)
285 journal_abort(EXT3_SB(sb)->s_journal, -EIO);
286 }
287
288 void ext3_warning (struct super_block * sb, const char * function,
289 const char * fmt, ...)
290 {
291 va_list args;
292
293 va_start(args, fmt);
294 printk(KERN_WARNING "EXT3-fs warning (device %s): %s: ",
295 sb->s_id, function);
296 vprintk(fmt, args);
297 printk("\n");
298 va_end(args);
299 }
300
301 void ext3_update_dynamic_rev(struct super_block *sb)
302 {
303 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
304
305 if (le32_to_cpu(es->s_rev_level) > EXT3_GOOD_OLD_REV)
306 return;
307
308 ext3_warning(sb, __func__,
309 "updating to rev %d because of new feature flag, "
310 "running e2fsck is recommended",
311 EXT3_DYNAMIC_REV);
312
313 es->s_first_ino = cpu_to_le32(EXT3_GOOD_OLD_FIRST_INO);
314 es->s_inode_size = cpu_to_le16(EXT3_GOOD_OLD_INODE_SIZE);
315 es->s_rev_level = cpu_to_le32(EXT3_DYNAMIC_REV);
316 /* leave es->s_feature_*compat flags alone */
317 /* es->s_uuid will be set by e2fsck if empty */
318
319 /*
320 * The rest of the superblock fields should be zero, and if not it
321 * means they are likely already in use, so leave them alone. We
322 * can leave it up to e2fsck to clean up any inconsistencies there.
323 */
324 }
325
326 /*
327 * Open the external journal device
328 */
329 static struct block_device *ext3_blkdev_get(dev_t dev)
330 {
331 struct block_device *bdev;
332 char b[BDEVNAME_SIZE];
333
334 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
335 if (IS_ERR(bdev))
336 goto fail;
337 return bdev;
338
339 fail:
340 printk(KERN_ERR "EXT3: failed to open journal device %s: %ld\n",
341 __bdevname(dev, b), PTR_ERR(bdev));
342 return NULL;
343 }
344
345 /*
346 * Release the journal device
347 */
348 static int ext3_blkdev_put(struct block_device *bdev)
349 {
350 bd_release(bdev);
351 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
352 }
353
354 static int ext3_blkdev_remove(struct ext3_sb_info *sbi)
355 {
356 struct block_device *bdev;
357 int ret = -ENODEV;
358
359 bdev = sbi->journal_bdev;
360 if (bdev) {
361 ret = ext3_blkdev_put(bdev);
362 sbi->journal_bdev = NULL;
363 }
364 return ret;
365 }
366
367 static inline struct inode *orphan_list_entry(struct list_head *l)
368 {
369 return &list_entry(l, struct ext3_inode_info, i_orphan)->vfs_inode;
370 }
371
372 static void dump_orphan_list(struct super_block *sb, struct ext3_sb_info *sbi)
373 {
374 struct list_head *l;
375
376 printk(KERN_ERR "sb orphan head is %d\n",
377 le32_to_cpu(sbi->s_es->s_last_orphan));
378
379 printk(KERN_ERR "sb_info orphan list:\n");
380 list_for_each(l, &sbi->s_orphan) {
381 struct inode *inode = orphan_list_entry(l);
382 printk(KERN_ERR " "
383 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
384 inode->i_sb->s_id, inode->i_ino, inode,
385 inode->i_mode, inode->i_nlink,
386 NEXT_ORPHAN(inode));
387 }
388 }
389
390 static void ext3_put_super (struct super_block * sb)
391 {
392 struct ext3_sb_info *sbi = EXT3_SB(sb);
393 struct ext3_super_block *es = sbi->s_es;
394 int i, err;
395
396 ext3_xattr_put_super(sb);
397 err = journal_destroy(sbi->s_journal);
398 sbi->s_journal = NULL;
399 if (err < 0)
400 ext3_abort(sb, __func__, "Couldn't clean up the journal");
401
402 if (!(sb->s_flags & MS_RDONLY)) {
403 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
404 es->s_state = cpu_to_le16(sbi->s_mount_state);
405 BUFFER_TRACE(sbi->s_sbh, "marking dirty");
406 mark_buffer_dirty(sbi->s_sbh);
407 ext3_commit_super(sb, es, 1);
408 }
409
410 for (i = 0; i < sbi->s_gdb_count; i++)
411 brelse(sbi->s_group_desc[i]);
412 kfree(sbi->s_group_desc);
413 percpu_counter_destroy(&sbi->s_freeblocks_counter);
414 percpu_counter_destroy(&sbi->s_freeinodes_counter);
415 percpu_counter_destroy(&sbi->s_dirs_counter);
416 brelse(sbi->s_sbh);
417 #ifdef CONFIG_QUOTA
418 for (i = 0; i < MAXQUOTAS; i++)
419 kfree(sbi->s_qf_names[i]);
420 #endif
421
422 /* Debugging code just in case the in-memory inode orphan list
423 * isn't empty. The on-disk one can be non-empty if we've
424 * detected an error and taken the fs readonly, but the
425 * in-memory list had better be clean by this point. */
426 if (!list_empty(&sbi->s_orphan))
427 dump_orphan_list(sb, sbi);
428 J_ASSERT(list_empty(&sbi->s_orphan));
429
430 invalidate_bdev(sb->s_bdev);
431 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
432 /*
433 * Invalidate the journal device's buffers. We don't want them
434 * floating about in memory - the physical journal device may
435 * hotswapped, and it breaks the `ro-after' testing code.
436 */
437 sync_blockdev(sbi->journal_bdev);
438 invalidate_bdev(sbi->journal_bdev);
439 ext3_blkdev_remove(sbi);
440 }
441 sb->s_fs_info = NULL;
442 kfree(sbi);
443 return;
444 }
445
446 static struct kmem_cache *ext3_inode_cachep;
447
448 /*
449 * Called inside transaction, so use GFP_NOFS
450 */
451 static struct inode *ext3_alloc_inode(struct super_block *sb)
452 {
453 struct ext3_inode_info *ei;
454
455 ei = kmem_cache_alloc(ext3_inode_cachep, GFP_NOFS);
456 if (!ei)
457 return NULL;
458 #ifdef CONFIG_EXT3_FS_POSIX_ACL
459 ei->i_acl = EXT3_ACL_NOT_CACHED;
460 ei->i_default_acl = EXT3_ACL_NOT_CACHED;
461 #endif
462 ei->i_block_alloc_info = NULL;
463 ei->vfs_inode.i_version = 1;
464 return &ei->vfs_inode;
465 }
466
467 static void ext3_destroy_inode(struct inode *inode)
468 {
469 if (!list_empty(&(EXT3_I(inode)->i_orphan))) {
470 printk("EXT3 Inode %p: orphan list check failed!\n",
471 EXT3_I(inode));
472 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
473 EXT3_I(inode), sizeof(struct ext3_inode_info),
474 false);
475 dump_stack();
476 }
477 kmem_cache_free(ext3_inode_cachep, EXT3_I(inode));
478 }
479
480 static void init_once(void *foo)
481 {
482 struct ext3_inode_info *ei = (struct ext3_inode_info *) foo;
483
484 INIT_LIST_HEAD(&ei->i_orphan);
485 #ifdef CONFIG_EXT3_FS_XATTR
486 init_rwsem(&ei->xattr_sem);
487 #endif
488 mutex_init(&ei->truncate_mutex);
489 inode_init_once(&ei->vfs_inode);
490 }
491
492 static int init_inodecache(void)
493 {
494 ext3_inode_cachep = kmem_cache_create("ext3_inode_cache",
495 sizeof(struct ext3_inode_info),
496 0, (SLAB_RECLAIM_ACCOUNT|
497 SLAB_MEM_SPREAD),
498 init_once);
499 if (ext3_inode_cachep == NULL)
500 return -ENOMEM;
501 return 0;
502 }
503
504 static void destroy_inodecache(void)
505 {
506 kmem_cache_destroy(ext3_inode_cachep);
507 }
508
509 static void ext3_clear_inode(struct inode *inode)
510 {
511 struct ext3_block_alloc_info *rsv = EXT3_I(inode)->i_block_alloc_info;
512 #ifdef CONFIG_EXT3_FS_POSIX_ACL
513 if (EXT3_I(inode)->i_acl &&
514 EXT3_I(inode)->i_acl != EXT3_ACL_NOT_CACHED) {
515 posix_acl_release(EXT3_I(inode)->i_acl);
516 EXT3_I(inode)->i_acl = EXT3_ACL_NOT_CACHED;
517 }
518 if (EXT3_I(inode)->i_default_acl &&
519 EXT3_I(inode)->i_default_acl != EXT3_ACL_NOT_CACHED) {
520 posix_acl_release(EXT3_I(inode)->i_default_acl);
521 EXT3_I(inode)->i_default_acl = EXT3_ACL_NOT_CACHED;
522 }
523 #endif
524 ext3_discard_reservation(inode);
525 EXT3_I(inode)->i_block_alloc_info = NULL;
526 if (unlikely(rsv))
527 kfree(rsv);
528 }
529
530 static inline void ext3_show_quota_options(struct seq_file *seq, struct super_block *sb)
531 {
532 #if defined(CONFIG_QUOTA)
533 struct ext3_sb_info *sbi = EXT3_SB(sb);
534
535 if (sbi->s_jquota_fmt)
536 seq_printf(seq, ",jqfmt=%s",
537 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold": "vfsv0");
538
539 if (sbi->s_qf_names[USRQUOTA])
540 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
541
542 if (sbi->s_qf_names[GRPQUOTA])
543 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
544
545 if (sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA)
546 seq_puts(seq, ",usrquota");
547
548 if (sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA)
549 seq_puts(seq, ",grpquota");
550 #endif
551 }
552
553 /*
554 * Show an option if
555 * - it's set to a non-default value OR
556 * - if the per-sb default is different from the global default
557 */
558 static int ext3_show_options(struct seq_file *seq, struct vfsmount *vfs)
559 {
560 struct super_block *sb = vfs->mnt_sb;
561 struct ext3_sb_info *sbi = EXT3_SB(sb);
562 struct ext3_super_block *es = sbi->s_es;
563 unsigned long def_mount_opts;
564
565 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
566
567 if (sbi->s_sb_block != 1)
568 seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
569 if (test_opt(sb, MINIX_DF))
570 seq_puts(seq, ",minixdf");
571 if (test_opt(sb, GRPID))
572 seq_puts(seq, ",grpid");
573 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT3_DEFM_BSDGROUPS))
574 seq_puts(seq, ",nogrpid");
575 if (sbi->s_resuid != EXT3_DEF_RESUID ||
576 le16_to_cpu(es->s_def_resuid) != EXT3_DEF_RESUID) {
577 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
578 }
579 if (sbi->s_resgid != EXT3_DEF_RESGID ||
580 le16_to_cpu(es->s_def_resgid) != EXT3_DEF_RESGID) {
581 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
582 }
583 if (test_opt(sb, ERRORS_RO)) {
584 int def_errors = le16_to_cpu(es->s_errors);
585
586 if (def_errors == EXT3_ERRORS_PANIC ||
587 def_errors == EXT3_ERRORS_CONTINUE) {
588 seq_puts(seq, ",errors=remount-ro");
589 }
590 }
591 if (test_opt(sb, ERRORS_CONT))
592 seq_puts(seq, ",errors=continue");
593 if (test_opt(sb, ERRORS_PANIC))
594 seq_puts(seq, ",errors=panic");
595 if (test_opt(sb, NO_UID32))
596 seq_puts(seq, ",nouid32");
597 if (test_opt(sb, DEBUG))
598 seq_puts(seq, ",debug");
599 if (test_opt(sb, OLDALLOC))
600 seq_puts(seq, ",oldalloc");
601 #ifdef CONFIG_EXT3_FS_XATTR
602 if (test_opt(sb, XATTR_USER))
603 seq_puts(seq, ",user_xattr");
604 if (!test_opt(sb, XATTR_USER) &&
605 (def_mount_opts & EXT3_DEFM_XATTR_USER)) {
606 seq_puts(seq, ",nouser_xattr");
607 }
608 #endif
609 #ifdef CONFIG_EXT3_FS_POSIX_ACL
610 if (test_opt(sb, POSIX_ACL))
611 seq_puts(seq, ",acl");
612 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT3_DEFM_ACL))
613 seq_puts(seq, ",noacl");
614 #endif
615 if (!test_opt(sb, RESERVATION))
616 seq_puts(seq, ",noreservation");
617 if (sbi->s_commit_interval) {
618 seq_printf(seq, ",commit=%u",
619 (unsigned) (sbi->s_commit_interval / HZ));
620 }
621 if (test_opt(sb, BARRIER))
622 seq_puts(seq, ",barrier=1");
623 if (test_opt(sb, NOBH))
624 seq_puts(seq, ",nobh");
625
626 if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA)
627 seq_puts(seq, ",data=journal");
628 else if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA)
629 seq_puts(seq, ",data=ordered");
630 else if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_WRITEBACK_DATA)
631 seq_puts(seq, ",data=writeback");
632
633 if (test_opt(sb, DATA_ERR_ABORT))
634 seq_puts(seq, ",data_err=abort");
635
636 ext3_show_quota_options(seq, sb);
637
638 return 0;
639 }
640
641
642 static struct inode *ext3_nfs_get_inode(struct super_block *sb,
643 u64 ino, u32 generation)
644 {
645 struct inode *inode;
646
647 if (ino < EXT3_FIRST_INO(sb) && ino != EXT3_ROOT_INO)
648 return ERR_PTR(-ESTALE);
649 if (ino > le32_to_cpu(EXT3_SB(sb)->s_es->s_inodes_count))
650 return ERR_PTR(-ESTALE);
651
652 /* iget isn't really right if the inode is currently unallocated!!
653 *
654 * ext3_read_inode will return a bad_inode if the inode had been
655 * deleted, so we should be safe.
656 *
657 * Currently we don't know the generation for parent directory, so
658 * a generation of 0 means "accept any"
659 */
660 inode = ext3_iget(sb, ino);
661 if (IS_ERR(inode))
662 return ERR_CAST(inode);
663 if (generation && inode->i_generation != generation) {
664 iput(inode);
665 return ERR_PTR(-ESTALE);
666 }
667
668 return inode;
669 }
670
671 static struct dentry *ext3_fh_to_dentry(struct super_block *sb, struct fid *fid,
672 int fh_len, int fh_type)
673 {
674 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
675 ext3_nfs_get_inode);
676 }
677
678 static struct dentry *ext3_fh_to_parent(struct super_block *sb, struct fid *fid,
679 int fh_len, int fh_type)
680 {
681 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
682 ext3_nfs_get_inode);
683 }
684
685 #ifdef CONFIG_QUOTA
686 #define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
687 #define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
688
689 static int ext3_dquot_initialize(struct inode *inode, int type);
690 static int ext3_dquot_drop(struct inode *inode);
691 static int ext3_write_dquot(struct dquot *dquot);
692 static int ext3_acquire_dquot(struct dquot *dquot);
693 static int ext3_release_dquot(struct dquot *dquot);
694 static int ext3_mark_dquot_dirty(struct dquot *dquot);
695 static int ext3_write_info(struct super_block *sb, int type);
696 static int ext3_quota_on(struct super_block *sb, int type, int format_id,
697 char *path, int remount);
698 static int ext3_quota_on_mount(struct super_block *sb, int type);
699 static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
700 size_t len, loff_t off);
701 static ssize_t ext3_quota_write(struct super_block *sb, int type,
702 const char *data, size_t len, loff_t off);
703
704 static struct dquot_operations ext3_quota_operations = {
705 .initialize = ext3_dquot_initialize,
706 .drop = ext3_dquot_drop,
707 .alloc_space = dquot_alloc_space,
708 .alloc_inode = dquot_alloc_inode,
709 .free_space = dquot_free_space,
710 .free_inode = dquot_free_inode,
711 .transfer = dquot_transfer,
712 .write_dquot = ext3_write_dquot,
713 .acquire_dquot = ext3_acquire_dquot,
714 .release_dquot = ext3_release_dquot,
715 .mark_dirty = ext3_mark_dquot_dirty,
716 .write_info = ext3_write_info
717 };
718
719 static struct quotactl_ops ext3_qctl_operations = {
720 .quota_on = ext3_quota_on,
721 .quota_off = vfs_quota_off,
722 .quota_sync = vfs_quota_sync,
723 .get_info = vfs_get_dqinfo,
724 .set_info = vfs_set_dqinfo,
725 .get_dqblk = vfs_get_dqblk,
726 .set_dqblk = vfs_set_dqblk
727 };
728 #endif
729
730 static const struct super_operations ext3_sops = {
731 .alloc_inode = ext3_alloc_inode,
732 .destroy_inode = ext3_destroy_inode,
733 .write_inode = ext3_write_inode,
734 .dirty_inode = ext3_dirty_inode,
735 .delete_inode = ext3_delete_inode,
736 .put_super = ext3_put_super,
737 .write_super = ext3_write_super,
738 .sync_fs = ext3_sync_fs,
739 .write_super_lockfs = ext3_write_super_lockfs,
740 .unlockfs = ext3_unlockfs,
741 .statfs = ext3_statfs,
742 .remount_fs = ext3_remount,
743 .clear_inode = ext3_clear_inode,
744 .show_options = ext3_show_options,
745 #ifdef CONFIG_QUOTA
746 .quota_read = ext3_quota_read,
747 .quota_write = ext3_quota_write,
748 #endif
749 };
750
751 static const struct export_operations ext3_export_ops = {
752 .fh_to_dentry = ext3_fh_to_dentry,
753 .fh_to_parent = ext3_fh_to_parent,
754 .get_parent = ext3_get_parent,
755 };
756
757 enum {
758 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
759 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
760 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
761 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
762 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
763 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
764 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
765 Opt_data_err_abort, Opt_data_err_ignore,
766 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
767 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
768 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
769 Opt_grpquota
770 };
771
772 static const match_table_t tokens = {
773 {Opt_bsd_df, "bsddf"},
774 {Opt_minix_df, "minixdf"},
775 {Opt_grpid, "grpid"},
776 {Opt_grpid, "bsdgroups"},
777 {Opt_nogrpid, "nogrpid"},
778 {Opt_nogrpid, "sysvgroups"},
779 {Opt_resgid, "resgid=%u"},
780 {Opt_resuid, "resuid=%u"},
781 {Opt_sb, "sb=%u"},
782 {Opt_err_cont, "errors=continue"},
783 {Opt_err_panic, "errors=panic"},
784 {Opt_err_ro, "errors=remount-ro"},
785 {Opt_nouid32, "nouid32"},
786 {Opt_nocheck, "nocheck"},
787 {Opt_nocheck, "check=none"},
788 {Opt_debug, "debug"},
789 {Opt_oldalloc, "oldalloc"},
790 {Opt_orlov, "orlov"},
791 {Opt_user_xattr, "user_xattr"},
792 {Opt_nouser_xattr, "nouser_xattr"},
793 {Opt_acl, "acl"},
794 {Opt_noacl, "noacl"},
795 {Opt_reservation, "reservation"},
796 {Opt_noreservation, "noreservation"},
797 {Opt_noload, "noload"},
798 {Opt_nobh, "nobh"},
799 {Opt_bh, "bh"},
800 {Opt_commit, "commit=%u"},
801 {Opt_journal_update, "journal=update"},
802 {Opt_journal_inum, "journal=%u"},
803 {Opt_journal_dev, "journal_dev=%u"},
804 {Opt_abort, "abort"},
805 {Opt_data_journal, "data=journal"},
806 {Opt_data_ordered, "data=ordered"},
807 {Opt_data_writeback, "data=writeback"},
808 {Opt_data_err_abort, "data_err=abort"},
809 {Opt_data_err_ignore, "data_err=ignore"},
810 {Opt_offusrjquota, "usrjquota="},
811 {Opt_usrjquota, "usrjquota=%s"},
812 {Opt_offgrpjquota, "grpjquota="},
813 {Opt_grpjquota, "grpjquota=%s"},
814 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
815 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
816 {Opt_grpquota, "grpquota"},
817 {Opt_noquota, "noquota"},
818 {Opt_quota, "quota"},
819 {Opt_usrquota, "usrquota"},
820 {Opt_barrier, "barrier=%u"},
821 {Opt_resize, "resize"},
822 {Opt_err, NULL},
823 };
824
825 static ext3_fsblk_t get_sb_block(void **data)
826 {
827 ext3_fsblk_t sb_block;
828 char *options = (char *) *data;
829
830 if (!options || strncmp(options, "sb=", 3) != 0)
831 return 1; /* Default location */
832 options += 3;
833 /*todo: use simple_strtoll with >32bit ext3 */
834 sb_block = simple_strtoul(options, &options, 0);
835 if (*options && *options != ',') {
836 printk("EXT3-fs: Invalid sb specification: %s\n",
837 (char *) *data);
838 return 1;
839 }
840 if (*options == ',')
841 options++;
842 *data = (void *) options;
843 return sb_block;
844 }
845
846 static int parse_options (char *options, struct super_block *sb,
847 unsigned int *inum, unsigned long *journal_devnum,
848 ext3_fsblk_t *n_blocks_count, int is_remount)
849 {
850 struct ext3_sb_info *sbi = EXT3_SB(sb);
851 char * p;
852 substring_t args[MAX_OPT_ARGS];
853 int data_opt = 0;
854 int option;
855 #ifdef CONFIG_QUOTA
856 int qtype, qfmt;
857 char *qname;
858 #endif
859
860 if (!options)
861 return 1;
862
863 while ((p = strsep (&options, ",")) != NULL) {
864 int token;
865 if (!*p)
866 continue;
867
868 token = match_token(p, tokens, args);
869 switch (token) {
870 case Opt_bsd_df:
871 clear_opt (sbi->s_mount_opt, MINIX_DF);
872 break;
873 case Opt_minix_df:
874 set_opt (sbi->s_mount_opt, MINIX_DF);
875 break;
876 case Opt_grpid:
877 set_opt (sbi->s_mount_opt, GRPID);
878 break;
879 case Opt_nogrpid:
880 clear_opt (sbi->s_mount_opt, GRPID);
881 break;
882 case Opt_resuid:
883 if (match_int(&args[0], &option))
884 return 0;
885 sbi->s_resuid = option;
886 break;
887 case Opt_resgid:
888 if (match_int(&args[0], &option))
889 return 0;
890 sbi->s_resgid = option;
891 break;
892 case Opt_sb:
893 /* handled by get_sb_block() instead of here */
894 /* *sb_block = match_int(&args[0]); */
895 break;
896 case Opt_err_panic:
897 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
898 clear_opt (sbi->s_mount_opt, ERRORS_RO);
899 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
900 break;
901 case Opt_err_ro:
902 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
903 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
904 set_opt (sbi->s_mount_opt, ERRORS_RO);
905 break;
906 case Opt_err_cont:
907 clear_opt (sbi->s_mount_opt, ERRORS_RO);
908 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
909 set_opt (sbi->s_mount_opt, ERRORS_CONT);
910 break;
911 case Opt_nouid32:
912 set_opt (sbi->s_mount_opt, NO_UID32);
913 break;
914 case Opt_nocheck:
915 clear_opt (sbi->s_mount_opt, CHECK);
916 break;
917 case Opt_debug:
918 set_opt (sbi->s_mount_opt, DEBUG);
919 break;
920 case Opt_oldalloc:
921 set_opt (sbi->s_mount_opt, OLDALLOC);
922 break;
923 case Opt_orlov:
924 clear_opt (sbi->s_mount_opt, OLDALLOC);
925 break;
926 #ifdef CONFIG_EXT3_FS_XATTR
927 case Opt_user_xattr:
928 set_opt (sbi->s_mount_opt, XATTR_USER);
929 break;
930 case Opt_nouser_xattr:
931 clear_opt (sbi->s_mount_opt, XATTR_USER);
932 break;
933 #else
934 case Opt_user_xattr:
935 case Opt_nouser_xattr:
936 printk("EXT3 (no)user_xattr options not supported\n");
937 break;
938 #endif
939 #ifdef CONFIG_EXT3_FS_POSIX_ACL
940 case Opt_acl:
941 set_opt(sbi->s_mount_opt, POSIX_ACL);
942 break;
943 case Opt_noacl:
944 clear_opt(sbi->s_mount_opt, POSIX_ACL);
945 break;
946 #else
947 case Opt_acl:
948 case Opt_noacl:
949 printk("EXT3 (no)acl options not supported\n");
950 break;
951 #endif
952 case Opt_reservation:
953 set_opt(sbi->s_mount_opt, RESERVATION);
954 break;
955 case Opt_noreservation:
956 clear_opt(sbi->s_mount_opt, RESERVATION);
957 break;
958 case Opt_journal_update:
959 /* @@@ FIXME */
960 /* Eventually we will want to be able to create
961 a journal file here. For now, only allow the
962 user to specify an existing inode to be the
963 journal file. */
964 if (is_remount) {
965 printk(KERN_ERR "EXT3-fs: cannot specify "
966 "journal on remount\n");
967 return 0;
968 }
969 set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
970 break;
971 case Opt_journal_inum:
972 if (is_remount) {
973 printk(KERN_ERR "EXT3-fs: cannot specify "
974 "journal on remount\n");
975 return 0;
976 }
977 if (match_int(&args[0], &option))
978 return 0;
979 *inum = option;
980 break;
981 case Opt_journal_dev:
982 if (is_remount) {
983 printk(KERN_ERR "EXT3-fs: cannot specify "
984 "journal on remount\n");
985 return 0;
986 }
987 if (match_int(&args[0], &option))
988 return 0;
989 *journal_devnum = option;
990 break;
991 case Opt_noload:
992 set_opt (sbi->s_mount_opt, NOLOAD);
993 break;
994 case Opt_commit:
995 if (match_int(&args[0], &option))
996 return 0;
997 if (option < 0)
998 return 0;
999 if (option == 0)
1000 option = JBD_DEFAULT_MAX_COMMIT_AGE;
1001 sbi->s_commit_interval = HZ * option;
1002 break;
1003 case Opt_data_journal:
1004 data_opt = EXT3_MOUNT_JOURNAL_DATA;
1005 goto datacheck;
1006 case Opt_data_ordered:
1007 data_opt = EXT3_MOUNT_ORDERED_DATA;
1008 goto datacheck;
1009 case Opt_data_writeback:
1010 data_opt = EXT3_MOUNT_WRITEBACK_DATA;
1011 datacheck:
1012 if (is_remount) {
1013 if ((sbi->s_mount_opt & EXT3_MOUNT_DATA_FLAGS)
1014 != data_opt) {
1015 printk(KERN_ERR
1016 "EXT3-fs: cannot change data "
1017 "mode on remount\n");
1018 return 0;
1019 }
1020 } else {
1021 sbi->s_mount_opt &= ~EXT3_MOUNT_DATA_FLAGS;
1022 sbi->s_mount_opt |= data_opt;
1023 }
1024 break;
1025 case Opt_data_err_abort:
1026 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1027 break;
1028 case Opt_data_err_ignore:
1029 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1030 break;
1031 #ifdef CONFIG_QUOTA
1032 case Opt_usrjquota:
1033 qtype = USRQUOTA;
1034 goto set_qf_name;
1035 case Opt_grpjquota:
1036 qtype = GRPQUOTA;
1037 set_qf_name:
1038 if ((sb_any_quota_enabled(sb) ||
1039 sb_any_quota_suspended(sb)) &&
1040 !sbi->s_qf_names[qtype]) {
1041 printk(KERN_ERR
1042 "EXT3-fs: Cannot change journaled "
1043 "quota options when quota turned on.\n");
1044 return 0;
1045 }
1046 qname = match_strdup(&args[0]);
1047 if (!qname) {
1048 printk(KERN_ERR
1049 "EXT3-fs: not enough memory for "
1050 "storing quotafile name.\n");
1051 return 0;
1052 }
1053 if (sbi->s_qf_names[qtype] &&
1054 strcmp(sbi->s_qf_names[qtype], qname)) {
1055 printk(KERN_ERR
1056 "EXT3-fs: %s quota file already "
1057 "specified.\n", QTYPE2NAME(qtype));
1058 kfree(qname);
1059 return 0;
1060 }
1061 sbi->s_qf_names[qtype] = qname;
1062 if (strchr(sbi->s_qf_names[qtype], '/')) {
1063 printk(KERN_ERR
1064 "EXT3-fs: quotafile must be on "
1065 "filesystem root.\n");
1066 kfree(sbi->s_qf_names[qtype]);
1067 sbi->s_qf_names[qtype] = NULL;
1068 return 0;
1069 }
1070 set_opt(sbi->s_mount_opt, QUOTA);
1071 break;
1072 case Opt_offusrjquota:
1073 qtype = USRQUOTA;
1074 goto clear_qf_name;
1075 case Opt_offgrpjquota:
1076 qtype = GRPQUOTA;
1077 clear_qf_name:
1078 if ((sb_any_quota_enabled(sb) ||
1079 sb_any_quota_suspended(sb)) &&
1080 sbi->s_qf_names[qtype]) {
1081 printk(KERN_ERR "EXT3-fs: Cannot change "
1082 "journaled quota options when "
1083 "quota turned on.\n");
1084 return 0;
1085 }
1086 /*
1087 * The space will be released later when all options
1088 * are confirmed to be correct
1089 */
1090 sbi->s_qf_names[qtype] = NULL;
1091 break;
1092 case Opt_jqfmt_vfsold:
1093 qfmt = QFMT_VFS_OLD;
1094 goto set_qf_format;
1095 case Opt_jqfmt_vfsv0:
1096 qfmt = QFMT_VFS_V0;
1097 set_qf_format:
1098 if ((sb_any_quota_enabled(sb) ||
1099 sb_any_quota_suspended(sb)) &&
1100 sbi->s_jquota_fmt != qfmt) {
1101 printk(KERN_ERR "EXT3-fs: Cannot change "
1102 "journaled quota options when "
1103 "quota turned on.\n");
1104 return 0;
1105 }
1106 sbi->s_jquota_fmt = qfmt;
1107 break;
1108 case Opt_quota:
1109 case Opt_usrquota:
1110 set_opt(sbi->s_mount_opt, QUOTA);
1111 set_opt(sbi->s_mount_opt, USRQUOTA);
1112 break;
1113 case Opt_grpquota:
1114 set_opt(sbi->s_mount_opt, QUOTA);
1115 set_opt(sbi->s_mount_opt, GRPQUOTA);
1116 break;
1117 case Opt_noquota:
1118 if (sb_any_quota_enabled(sb) ||
1119 sb_any_quota_suspended(sb)) {
1120 printk(KERN_ERR "EXT3-fs: Cannot change quota "
1121 "options when quota turned on.\n");
1122 return 0;
1123 }
1124 clear_opt(sbi->s_mount_opt, QUOTA);
1125 clear_opt(sbi->s_mount_opt, USRQUOTA);
1126 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1127 break;
1128 #else
1129 case Opt_quota:
1130 case Opt_usrquota:
1131 case Opt_grpquota:
1132 printk(KERN_ERR
1133 "EXT3-fs: quota options not supported.\n");
1134 break;
1135 case Opt_usrjquota:
1136 case Opt_grpjquota:
1137 case Opt_offusrjquota:
1138 case Opt_offgrpjquota:
1139 case Opt_jqfmt_vfsold:
1140 case Opt_jqfmt_vfsv0:
1141 printk(KERN_ERR
1142 "EXT3-fs: journaled quota options not "
1143 "supported.\n");
1144 break;
1145 case Opt_noquota:
1146 break;
1147 #endif
1148 case Opt_abort:
1149 set_opt(sbi->s_mount_opt, ABORT);
1150 break;
1151 case Opt_barrier:
1152 if (match_int(&args[0], &option))
1153 return 0;
1154 if (option)
1155 set_opt(sbi->s_mount_opt, BARRIER);
1156 else
1157 clear_opt(sbi->s_mount_opt, BARRIER);
1158 break;
1159 case Opt_ignore:
1160 break;
1161 case Opt_resize:
1162 if (!is_remount) {
1163 printk("EXT3-fs: resize option only available "
1164 "for remount\n");
1165 return 0;
1166 }
1167 if (match_int(&args[0], &option) != 0)
1168 return 0;
1169 *n_blocks_count = option;
1170 break;
1171 case Opt_nobh:
1172 set_opt(sbi->s_mount_opt, NOBH);
1173 break;
1174 case Opt_bh:
1175 clear_opt(sbi->s_mount_opt, NOBH);
1176 break;
1177 default:
1178 printk (KERN_ERR
1179 "EXT3-fs: Unrecognized mount option \"%s\" "
1180 "or missing value\n", p);
1181 return 0;
1182 }
1183 }
1184 #ifdef CONFIG_QUOTA
1185 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1186 if ((sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA) &&
1187 sbi->s_qf_names[USRQUOTA])
1188 clear_opt(sbi->s_mount_opt, USRQUOTA);
1189
1190 if ((sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA) &&
1191 sbi->s_qf_names[GRPQUOTA])
1192 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1193
1194 if ((sbi->s_qf_names[USRQUOTA] &&
1195 (sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA)) ||
1196 (sbi->s_qf_names[GRPQUOTA] &&
1197 (sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA))) {
1198 printk(KERN_ERR "EXT3-fs: old and new quota "
1199 "format mixing.\n");
1200 return 0;
1201 }
1202
1203 if (!sbi->s_jquota_fmt) {
1204 printk(KERN_ERR "EXT3-fs: journaled quota format "
1205 "not specified.\n");
1206 return 0;
1207 }
1208 } else {
1209 if (sbi->s_jquota_fmt) {
1210 printk(KERN_ERR "EXT3-fs: journaled quota format "
1211 "specified with no journaling "
1212 "enabled.\n");
1213 return 0;
1214 }
1215 }
1216 #endif
1217 return 1;
1218 }
1219
1220 static int ext3_setup_super(struct super_block *sb, struct ext3_super_block *es,
1221 int read_only)
1222 {
1223 struct ext3_sb_info *sbi = EXT3_SB(sb);
1224 int res = 0;
1225
1226 if (le32_to_cpu(es->s_rev_level) > EXT3_MAX_SUPP_REV) {
1227 printk (KERN_ERR "EXT3-fs warning: revision level too high, "
1228 "forcing read-only mode\n");
1229 res = MS_RDONLY;
1230 }
1231 if (read_only)
1232 return res;
1233 if (!(sbi->s_mount_state & EXT3_VALID_FS))
1234 printk (KERN_WARNING "EXT3-fs warning: mounting unchecked fs, "
1235 "running e2fsck is recommended\n");
1236 else if ((sbi->s_mount_state & EXT3_ERROR_FS))
1237 printk (KERN_WARNING
1238 "EXT3-fs warning: mounting fs with errors, "
1239 "running e2fsck is recommended\n");
1240 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1241 le16_to_cpu(es->s_mnt_count) >=
1242 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1243 printk (KERN_WARNING
1244 "EXT3-fs warning: maximal mount count reached, "
1245 "running e2fsck is recommended\n");
1246 else if (le32_to_cpu(es->s_checkinterval) &&
1247 (le32_to_cpu(es->s_lastcheck) +
1248 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1249 printk (KERN_WARNING
1250 "EXT3-fs warning: checktime reached, "
1251 "running e2fsck is recommended\n");
1252 #if 0
1253 /* @@@ We _will_ want to clear the valid bit if we find
1254 inconsistencies, to force a fsck at reboot. But for
1255 a plain journaled filesystem we can keep it set as
1256 valid forever! :) */
1257 es->s_state &= cpu_to_le16(~EXT3_VALID_FS);
1258 #endif
1259 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1260 es->s_max_mnt_count = cpu_to_le16(EXT3_DFL_MAX_MNT_COUNT);
1261 le16_add_cpu(&es->s_mnt_count, 1);
1262 es->s_mtime = cpu_to_le32(get_seconds());
1263 ext3_update_dynamic_rev(sb);
1264 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
1265
1266 ext3_commit_super(sb, es, 1);
1267 if (test_opt(sb, DEBUG))
1268 printk(KERN_INFO "[EXT3 FS bs=%lu, gc=%lu, "
1269 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1270 sb->s_blocksize,
1271 sbi->s_groups_count,
1272 EXT3_BLOCKS_PER_GROUP(sb),
1273 EXT3_INODES_PER_GROUP(sb),
1274 sbi->s_mount_opt);
1275
1276 printk(KERN_INFO "EXT3 FS on %s, ", sb->s_id);
1277 if (EXT3_SB(sb)->s_journal->j_inode == NULL) {
1278 char b[BDEVNAME_SIZE];
1279
1280 printk("external journal on %s\n",
1281 bdevname(EXT3_SB(sb)->s_journal->j_dev, b));
1282 } else {
1283 printk("internal journal\n");
1284 }
1285 return res;
1286 }
1287
1288 /* Called at mount-time, super-block is locked */
1289 static int ext3_check_descriptors(struct super_block *sb)
1290 {
1291 struct ext3_sb_info *sbi = EXT3_SB(sb);
1292 int i;
1293
1294 ext3_debug ("Checking group descriptors");
1295
1296 for (i = 0; i < sbi->s_groups_count; i++) {
1297 struct ext3_group_desc *gdp = ext3_get_group_desc(sb, i, NULL);
1298 ext3_fsblk_t first_block = ext3_group_first_block_no(sb, i);
1299 ext3_fsblk_t last_block;
1300
1301 if (i == sbi->s_groups_count - 1)
1302 last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
1303 else
1304 last_block = first_block +
1305 (EXT3_BLOCKS_PER_GROUP(sb) - 1);
1306
1307 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
1308 le32_to_cpu(gdp->bg_block_bitmap) > last_block)
1309 {
1310 ext3_error (sb, "ext3_check_descriptors",
1311 "Block bitmap for group %d"
1312 " not in group (block %lu)!",
1313 i, (unsigned long)
1314 le32_to_cpu(gdp->bg_block_bitmap));
1315 return 0;
1316 }
1317 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
1318 le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
1319 {
1320 ext3_error (sb, "ext3_check_descriptors",
1321 "Inode bitmap for group %d"
1322 " not in group (block %lu)!",
1323 i, (unsigned long)
1324 le32_to_cpu(gdp->bg_inode_bitmap));
1325 return 0;
1326 }
1327 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
1328 le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
1329 last_block)
1330 {
1331 ext3_error (sb, "ext3_check_descriptors",
1332 "Inode table for group %d"
1333 " not in group (block %lu)!",
1334 i, (unsigned long)
1335 le32_to_cpu(gdp->bg_inode_table));
1336 return 0;
1337 }
1338 }
1339
1340 sbi->s_es->s_free_blocks_count=cpu_to_le32(ext3_count_free_blocks(sb));
1341 sbi->s_es->s_free_inodes_count=cpu_to_le32(ext3_count_free_inodes(sb));
1342 return 1;
1343 }
1344
1345
1346 /* ext3_orphan_cleanup() walks a singly-linked list of inodes (starting at
1347 * the superblock) which were deleted from all directories, but held open by
1348 * a process at the time of a crash. We walk the list and try to delete these
1349 * inodes at recovery time (only with a read-write filesystem).
1350 *
1351 * In order to keep the orphan inode chain consistent during traversal (in
1352 * case of crash during recovery), we link each inode into the superblock
1353 * orphan list_head and handle it the same way as an inode deletion during
1354 * normal operation (which journals the operations for us).
1355 *
1356 * We only do an iget() and an iput() on each inode, which is very safe if we
1357 * accidentally point at an in-use or already deleted inode. The worst that
1358 * can happen in this case is that we get a "bit already cleared" message from
1359 * ext3_free_inode(). The only reason we would point at a wrong inode is if
1360 * e2fsck was run on this filesystem, and it must have already done the orphan
1361 * inode cleanup for us, so we can safely abort without any further action.
1362 */
1363 static void ext3_orphan_cleanup (struct super_block * sb,
1364 struct ext3_super_block * es)
1365 {
1366 unsigned int s_flags = sb->s_flags;
1367 int nr_orphans = 0, nr_truncates = 0;
1368 #ifdef CONFIG_QUOTA
1369 int i;
1370 #endif
1371 if (!es->s_last_orphan) {
1372 jbd_debug(4, "no orphan inodes to clean up\n");
1373 return;
1374 }
1375
1376 if (bdev_read_only(sb->s_bdev)) {
1377 printk(KERN_ERR "EXT3-fs: write access "
1378 "unavailable, skipping orphan cleanup.\n");
1379 return;
1380 }
1381
1382 if (EXT3_SB(sb)->s_mount_state & EXT3_ERROR_FS) {
1383 if (es->s_last_orphan)
1384 jbd_debug(1, "Errors on filesystem, "
1385 "clearing orphan list.\n");
1386 es->s_last_orphan = 0;
1387 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1388 return;
1389 }
1390
1391 if (s_flags & MS_RDONLY) {
1392 printk(KERN_INFO "EXT3-fs: %s: orphan cleanup on readonly fs\n",
1393 sb->s_id);
1394 sb->s_flags &= ~MS_RDONLY;
1395 }
1396 #ifdef CONFIG_QUOTA
1397 /* Needed for iput() to work correctly and not trash data */
1398 sb->s_flags |= MS_ACTIVE;
1399 /* Turn on quotas so that they are updated correctly */
1400 for (i = 0; i < MAXQUOTAS; i++) {
1401 if (EXT3_SB(sb)->s_qf_names[i]) {
1402 int ret = ext3_quota_on_mount(sb, i);
1403 if (ret < 0)
1404 printk(KERN_ERR
1405 "EXT3-fs: Cannot turn on journaled "
1406 "quota: error %d\n", ret);
1407 }
1408 }
1409 #endif
1410
1411 while (es->s_last_orphan) {
1412 struct inode *inode;
1413
1414 inode = ext3_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1415 if (IS_ERR(inode)) {
1416 es->s_last_orphan = 0;
1417 break;
1418 }
1419
1420 list_add(&EXT3_I(inode)->i_orphan, &EXT3_SB(sb)->s_orphan);
1421 DQUOT_INIT(inode);
1422 if (inode->i_nlink) {
1423 printk(KERN_DEBUG
1424 "%s: truncating inode %lu to %Ld bytes\n",
1425 __func__, inode->i_ino, inode->i_size);
1426 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1427 inode->i_ino, inode->i_size);
1428 ext3_truncate(inode);
1429 nr_truncates++;
1430 } else {
1431 printk(KERN_DEBUG
1432 "%s: deleting unreferenced inode %lu\n",
1433 __func__, inode->i_ino);
1434 jbd_debug(2, "deleting unreferenced inode %lu\n",
1435 inode->i_ino);
1436 nr_orphans++;
1437 }
1438 iput(inode); /* The delete magic happens here! */
1439 }
1440
1441 #define PLURAL(x) (x), ((x)==1) ? "" : "s"
1442
1443 if (nr_orphans)
1444 printk(KERN_INFO "EXT3-fs: %s: %d orphan inode%s deleted\n",
1445 sb->s_id, PLURAL(nr_orphans));
1446 if (nr_truncates)
1447 printk(KERN_INFO "EXT3-fs: %s: %d truncate%s cleaned up\n",
1448 sb->s_id, PLURAL(nr_truncates));
1449 #ifdef CONFIG_QUOTA
1450 /* Turn quotas off */
1451 for (i = 0; i < MAXQUOTAS; i++) {
1452 if (sb_dqopt(sb)->files[i])
1453 vfs_quota_off(sb, i, 0);
1454 }
1455 #endif
1456 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1457 }
1458
1459 /*
1460 * Maximal file size. There is a direct, and {,double-,triple-}indirect
1461 * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
1462 * We need to be 1 filesystem block less than the 2^32 sector limit.
1463 */
1464 static loff_t ext3_max_size(int bits)
1465 {
1466 loff_t res = EXT3_NDIR_BLOCKS;
1467 int meta_blocks;
1468 loff_t upper_limit;
1469
1470 /* This is calculated to be the largest file size for a
1471 * dense, file such that the total number of
1472 * sectors in the file, including data and all indirect blocks,
1473 * does not exceed 2^32 -1
1474 * __u32 i_blocks representing the total number of
1475 * 512 bytes blocks of the file
1476 */
1477 upper_limit = (1LL << 32) - 1;
1478
1479 /* total blocks in file system block size */
1480 upper_limit >>= (bits - 9);
1481
1482
1483 /* indirect blocks */
1484 meta_blocks = 1;
1485 /* double indirect blocks */
1486 meta_blocks += 1 + (1LL << (bits-2));
1487 /* tripple indirect blocks */
1488 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1489
1490 upper_limit -= meta_blocks;
1491 upper_limit <<= bits;
1492
1493 res += 1LL << (bits-2);
1494 res += 1LL << (2*(bits-2));
1495 res += 1LL << (3*(bits-2));
1496 res <<= bits;
1497 if (res > upper_limit)
1498 res = upper_limit;
1499
1500 if (res > MAX_LFS_FILESIZE)
1501 res = MAX_LFS_FILESIZE;
1502
1503 return res;
1504 }
1505
1506 static ext3_fsblk_t descriptor_loc(struct super_block *sb,
1507 ext3_fsblk_t logic_sb_block,
1508 int nr)
1509 {
1510 struct ext3_sb_info *sbi = EXT3_SB(sb);
1511 unsigned long bg, first_meta_bg;
1512 int has_super = 0;
1513
1514 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1515
1516 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_META_BG) ||
1517 nr < first_meta_bg)
1518 return (logic_sb_block + nr + 1);
1519 bg = sbi->s_desc_per_block * nr;
1520 if (ext3_bg_has_super(sb, bg))
1521 has_super = 1;
1522 return (has_super + ext3_group_first_block_no(sb, bg));
1523 }
1524
1525
1526 static int ext3_fill_super (struct super_block *sb, void *data, int silent)
1527 {
1528 struct buffer_head * bh;
1529 struct ext3_super_block *es = NULL;
1530 struct ext3_sb_info *sbi;
1531 ext3_fsblk_t block;
1532 ext3_fsblk_t sb_block = get_sb_block(&data);
1533 ext3_fsblk_t logic_sb_block;
1534 unsigned long offset = 0;
1535 unsigned int journal_inum = 0;
1536 unsigned long journal_devnum = 0;
1537 unsigned long def_mount_opts;
1538 struct inode *root;
1539 int blocksize;
1540 int hblock;
1541 int db_count;
1542 int i;
1543 int needs_recovery;
1544 int ret = -EINVAL;
1545 __le32 features;
1546 int err;
1547
1548 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1549 if (!sbi)
1550 return -ENOMEM;
1551 sb->s_fs_info = sbi;
1552 sbi->s_mount_opt = 0;
1553 sbi->s_resuid = EXT3_DEF_RESUID;
1554 sbi->s_resgid = EXT3_DEF_RESGID;
1555 sbi->s_sb_block = sb_block;
1556
1557 unlock_kernel();
1558
1559 blocksize = sb_min_blocksize(sb, EXT3_MIN_BLOCK_SIZE);
1560 if (!blocksize) {
1561 printk(KERN_ERR "EXT3-fs: unable to set blocksize\n");
1562 goto out_fail;
1563 }
1564
1565 /*
1566 * The ext3 superblock will not be buffer aligned for other than 1kB
1567 * block sizes. We need to calculate the offset from buffer start.
1568 */
1569 if (blocksize != EXT3_MIN_BLOCK_SIZE) {
1570 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1571 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1572 } else {
1573 logic_sb_block = sb_block;
1574 }
1575
1576 if (!(bh = sb_bread(sb, logic_sb_block))) {
1577 printk (KERN_ERR "EXT3-fs: unable to read superblock\n");
1578 goto out_fail;
1579 }
1580 /*
1581 * Note: s_es must be initialized as soon as possible because
1582 * some ext3 macro-instructions depend on its value
1583 */
1584 es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
1585 sbi->s_es = es;
1586 sb->s_magic = le16_to_cpu(es->s_magic);
1587 if (sb->s_magic != EXT3_SUPER_MAGIC)
1588 goto cantfind_ext3;
1589
1590 /* Set defaults before we parse the mount options */
1591 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1592 if (def_mount_opts & EXT3_DEFM_DEBUG)
1593 set_opt(sbi->s_mount_opt, DEBUG);
1594 if (def_mount_opts & EXT3_DEFM_BSDGROUPS)
1595 set_opt(sbi->s_mount_opt, GRPID);
1596 if (def_mount_opts & EXT3_DEFM_UID16)
1597 set_opt(sbi->s_mount_opt, NO_UID32);
1598 #ifdef CONFIG_EXT3_FS_XATTR
1599 if (def_mount_opts & EXT3_DEFM_XATTR_USER)
1600 set_opt(sbi->s_mount_opt, XATTR_USER);
1601 #endif
1602 #ifdef CONFIG_EXT3_FS_POSIX_ACL
1603 if (def_mount_opts & EXT3_DEFM_ACL)
1604 set_opt(sbi->s_mount_opt, POSIX_ACL);
1605 #endif
1606 if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_DATA)
1607 sbi->s_mount_opt |= EXT3_MOUNT_JOURNAL_DATA;
1608 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_ORDERED)
1609 sbi->s_mount_opt |= EXT3_MOUNT_ORDERED_DATA;
1610 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_WBACK)
1611 sbi->s_mount_opt |= EXT3_MOUNT_WRITEBACK_DATA;
1612
1613 if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_PANIC)
1614 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1615 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_CONTINUE)
1616 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1617 else
1618 set_opt(sbi->s_mount_opt, ERRORS_RO);
1619
1620 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1621 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1622
1623 set_opt(sbi->s_mount_opt, RESERVATION);
1624
1625 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1626 NULL, 0))
1627 goto failed_mount;
1628
1629 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1630 ((sbi->s_mount_opt & EXT3_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1631
1632 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV &&
1633 (EXT3_HAS_COMPAT_FEATURE(sb, ~0U) ||
1634 EXT3_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1635 EXT3_HAS_INCOMPAT_FEATURE(sb, ~0U)))
1636 printk(KERN_WARNING
1637 "EXT3-fs warning: feature flags set on rev 0 fs, "
1638 "running e2fsck is recommended\n");
1639 /*
1640 * Check feature flags regardless of the revision level, since we
1641 * previously didn't change the revision level when setting the flags,
1642 * so there is a chance incompat flags are set on a rev 0 filesystem.
1643 */
1644 features = EXT3_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP);
1645 if (features) {
1646 printk(KERN_ERR "EXT3-fs: %s: couldn't mount because of "
1647 "unsupported optional features (%x).\n",
1648 sb->s_id, le32_to_cpu(features));
1649 goto failed_mount;
1650 }
1651 features = EXT3_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP);
1652 if (!(sb->s_flags & MS_RDONLY) && features) {
1653 printk(KERN_ERR "EXT3-fs: %s: couldn't mount RDWR because of "
1654 "unsupported optional features (%x).\n",
1655 sb->s_id, le32_to_cpu(features));
1656 goto failed_mount;
1657 }
1658 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
1659
1660 if (blocksize < EXT3_MIN_BLOCK_SIZE ||
1661 blocksize > EXT3_MAX_BLOCK_SIZE) {
1662 printk(KERN_ERR
1663 "EXT3-fs: Unsupported filesystem blocksize %d on %s.\n",
1664 blocksize, sb->s_id);
1665 goto failed_mount;
1666 }
1667
1668 hblock = bdev_hardsect_size(sb->s_bdev);
1669 if (sb->s_blocksize != blocksize) {
1670 /*
1671 * Make sure the blocksize for the filesystem is larger
1672 * than the hardware sectorsize for the machine.
1673 */
1674 if (blocksize < hblock) {
1675 printk(KERN_ERR "EXT3-fs: blocksize %d too small for "
1676 "device blocksize %d.\n", blocksize, hblock);
1677 goto failed_mount;
1678 }
1679
1680 brelse (bh);
1681 if (!sb_set_blocksize(sb, blocksize)) {
1682 printk(KERN_ERR "EXT3-fs: bad blocksize %d.\n",
1683 blocksize);
1684 goto out_fail;
1685 }
1686 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1687 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1688 bh = sb_bread(sb, logic_sb_block);
1689 if (!bh) {
1690 printk(KERN_ERR
1691 "EXT3-fs: Can't read superblock on 2nd try.\n");
1692 goto failed_mount;
1693 }
1694 es = (struct ext3_super_block *)(((char *)bh->b_data) + offset);
1695 sbi->s_es = es;
1696 if (es->s_magic != cpu_to_le16(EXT3_SUPER_MAGIC)) {
1697 printk (KERN_ERR
1698 "EXT3-fs: Magic mismatch, very weird !\n");
1699 goto failed_mount;
1700 }
1701 }
1702
1703 sb->s_maxbytes = ext3_max_size(sb->s_blocksize_bits);
1704
1705 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV) {
1706 sbi->s_inode_size = EXT3_GOOD_OLD_INODE_SIZE;
1707 sbi->s_first_ino = EXT3_GOOD_OLD_FIRST_INO;
1708 } else {
1709 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1710 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1711 if ((sbi->s_inode_size < EXT3_GOOD_OLD_INODE_SIZE) ||
1712 (!is_power_of_2(sbi->s_inode_size)) ||
1713 (sbi->s_inode_size > blocksize)) {
1714 printk (KERN_ERR
1715 "EXT3-fs: unsupported inode size: %d\n",
1716 sbi->s_inode_size);
1717 goto failed_mount;
1718 }
1719 }
1720 sbi->s_frag_size = EXT3_MIN_FRAG_SIZE <<
1721 le32_to_cpu(es->s_log_frag_size);
1722 if (blocksize != sbi->s_frag_size) {
1723 printk(KERN_ERR
1724 "EXT3-fs: fragsize %lu != blocksize %u (unsupported)\n",
1725 sbi->s_frag_size, blocksize);
1726 goto failed_mount;
1727 }
1728 sbi->s_frags_per_block = 1;
1729 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1730 sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
1731 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
1732 if (EXT3_INODE_SIZE(sb) == 0 || EXT3_INODES_PER_GROUP(sb) == 0)
1733 goto cantfind_ext3;
1734 sbi->s_inodes_per_block = blocksize / EXT3_INODE_SIZE(sb);
1735 if (sbi->s_inodes_per_block == 0)
1736 goto cantfind_ext3;
1737 sbi->s_itb_per_group = sbi->s_inodes_per_group /
1738 sbi->s_inodes_per_block;
1739 sbi->s_desc_per_block = blocksize / sizeof(struct ext3_group_desc);
1740 sbi->s_sbh = bh;
1741 sbi->s_mount_state = le16_to_cpu(es->s_state);
1742 sbi->s_addr_per_block_bits = ilog2(EXT3_ADDR_PER_BLOCK(sb));
1743 sbi->s_desc_per_block_bits = ilog2(EXT3_DESC_PER_BLOCK(sb));
1744 for (i=0; i < 4; i++)
1745 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
1746 sbi->s_def_hash_version = es->s_def_hash_version;
1747
1748 if (sbi->s_blocks_per_group > blocksize * 8) {
1749 printk (KERN_ERR
1750 "EXT3-fs: #blocks per group too big: %lu\n",
1751 sbi->s_blocks_per_group);
1752 goto failed_mount;
1753 }
1754 if (sbi->s_frags_per_group > blocksize * 8) {
1755 printk (KERN_ERR
1756 "EXT3-fs: #fragments per group too big: %lu\n",
1757 sbi->s_frags_per_group);
1758 goto failed_mount;
1759 }
1760 if (sbi->s_inodes_per_group > blocksize * 8) {
1761 printk (KERN_ERR
1762 "EXT3-fs: #inodes per group too big: %lu\n",
1763 sbi->s_inodes_per_group);
1764 goto failed_mount;
1765 }
1766
1767 if (le32_to_cpu(es->s_blocks_count) >
1768 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
1769 printk(KERN_ERR "EXT3-fs: filesystem on %s:"
1770 " too large to mount safely\n", sb->s_id);
1771 if (sizeof(sector_t) < 8)
1772 printk(KERN_WARNING "EXT3-fs: CONFIG_LBD not "
1773 "enabled\n");
1774 goto failed_mount;
1775 }
1776
1777 if (EXT3_BLOCKS_PER_GROUP(sb) == 0)
1778 goto cantfind_ext3;
1779 sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1780 le32_to_cpu(es->s_first_data_block) - 1)
1781 / EXT3_BLOCKS_PER_GROUP(sb)) + 1;
1782 db_count = (sbi->s_groups_count + EXT3_DESC_PER_BLOCK(sb) - 1) /
1783 EXT3_DESC_PER_BLOCK(sb);
1784 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
1785 GFP_KERNEL);
1786 if (sbi->s_group_desc == NULL) {
1787 printk (KERN_ERR "EXT3-fs: not enough memory\n");
1788 goto failed_mount;
1789 }
1790
1791 bgl_lock_init(&sbi->s_blockgroup_lock);
1792
1793 for (i = 0; i < db_count; i++) {
1794 block = descriptor_loc(sb, logic_sb_block, i);
1795 sbi->s_group_desc[i] = sb_bread(sb, block);
1796 if (!sbi->s_group_desc[i]) {
1797 printk (KERN_ERR "EXT3-fs: "
1798 "can't read group descriptor %d\n", i);
1799 db_count = i;
1800 goto failed_mount2;
1801 }
1802 }
1803 if (!ext3_check_descriptors (sb)) {
1804 printk(KERN_ERR "EXT3-fs: group descriptors corrupted!\n");
1805 goto failed_mount2;
1806 }
1807 sbi->s_gdb_count = db_count;
1808 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1809 spin_lock_init(&sbi->s_next_gen_lock);
1810
1811 err = percpu_counter_init(&sbi->s_freeblocks_counter,
1812 ext3_count_free_blocks(sb));
1813 if (!err) {
1814 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1815 ext3_count_free_inodes(sb));
1816 }
1817 if (!err) {
1818 err = percpu_counter_init(&sbi->s_dirs_counter,
1819 ext3_count_dirs(sb));
1820 }
1821 if (err) {
1822 printk(KERN_ERR "EXT3-fs: insufficient memory\n");
1823 goto failed_mount3;
1824 }
1825
1826 /* per fileystem reservation list head & lock */
1827 spin_lock_init(&sbi->s_rsv_window_lock);
1828 sbi->s_rsv_window_root = RB_ROOT;
1829 /* Add a single, static dummy reservation to the start of the
1830 * reservation window list --- it gives us a placeholder for
1831 * append-at-start-of-list which makes the allocation logic
1832 * _much_ simpler. */
1833 sbi->s_rsv_window_head.rsv_start = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1834 sbi->s_rsv_window_head.rsv_end = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1835 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1836 sbi->s_rsv_window_head.rsv_goal_size = 0;
1837 ext3_rsv_window_add(sb, &sbi->s_rsv_window_head);
1838
1839 /*
1840 * set up enough so that it can read an inode
1841 */
1842 sb->s_op = &ext3_sops;
1843 sb->s_export_op = &ext3_export_ops;
1844 sb->s_xattr = ext3_xattr_handlers;
1845 #ifdef CONFIG_QUOTA
1846 sb->s_qcop = &ext3_qctl_operations;
1847 sb->dq_op = &ext3_quota_operations;
1848 #endif
1849 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
1850
1851 sb->s_root = NULL;
1852
1853 needs_recovery = (es->s_last_orphan != 0 ||
1854 EXT3_HAS_INCOMPAT_FEATURE(sb,
1855 EXT3_FEATURE_INCOMPAT_RECOVER));
1856
1857 /*
1858 * The first inode we look at is the journal inode. Don't try
1859 * root first: it may be modified in the journal!
1860 */
1861 if (!test_opt(sb, NOLOAD) &&
1862 EXT3_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL)) {
1863 if (ext3_load_journal(sb, es, journal_devnum))
1864 goto failed_mount3;
1865 } else if (journal_inum) {
1866 if (ext3_create_journal(sb, es, journal_inum))
1867 goto failed_mount3;
1868 } else {
1869 if (!silent)
1870 printk (KERN_ERR
1871 "ext3: No journal on filesystem on %s\n",
1872 sb->s_id);
1873 goto failed_mount3;
1874 }
1875
1876 /* We have now updated the journal if required, so we can
1877 * validate the data journaling mode. */
1878 switch (test_opt(sb, DATA_FLAGS)) {
1879 case 0:
1880 /* No mode set, assume a default based on the journal
1881 capabilities: ORDERED_DATA if the journal can
1882 cope, else JOURNAL_DATA */
1883 if (journal_check_available_features
1884 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE))
1885 set_opt(sbi->s_mount_opt, ORDERED_DATA);
1886 else
1887 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
1888 break;
1889
1890 case EXT3_MOUNT_ORDERED_DATA:
1891 case EXT3_MOUNT_WRITEBACK_DATA:
1892 if (!journal_check_available_features
1893 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE)) {
1894 printk(KERN_ERR "EXT3-fs: Journal does not support "
1895 "requested data journaling mode\n");
1896 goto failed_mount4;
1897 }
1898 default:
1899 break;
1900 }
1901
1902 if (test_opt(sb, NOBH)) {
1903 if (!(test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_WRITEBACK_DATA)) {
1904 printk(KERN_WARNING "EXT3-fs: Ignoring nobh option - "
1905 "its supported only with writeback mode\n");
1906 clear_opt(sbi->s_mount_opt, NOBH);
1907 }
1908 }
1909 /*
1910 * The journal_load will have done any necessary log recovery,
1911 * so we can safely mount the rest of the filesystem now.
1912 */
1913
1914 root = ext3_iget(sb, EXT3_ROOT_INO);
1915 if (IS_ERR(root)) {
1916 printk(KERN_ERR "EXT3-fs: get root inode failed\n");
1917 ret = PTR_ERR(root);
1918 goto failed_mount4;
1919 }
1920 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1921 iput(root);
1922 printk(KERN_ERR "EXT3-fs: corrupt root inode, run e2fsck\n");
1923 goto failed_mount4;
1924 }
1925 sb->s_root = d_alloc_root(root);
1926 if (!sb->s_root) {
1927 printk(KERN_ERR "EXT3-fs: get root dentry failed\n");
1928 iput(root);
1929 ret = -ENOMEM;
1930 goto failed_mount4;
1931 }
1932
1933 ext3_setup_super (sb, es, sb->s_flags & MS_RDONLY);
1934 /*
1935 * akpm: core read_super() calls in here with the superblock locked.
1936 * That deadlocks, because orphan cleanup needs to lock the superblock
1937 * in numerous places. Here we just pop the lock - it's relatively
1938 * harmless, because we are now ready to accept write_super() requests,
1939 * and aviro says that's the only reason for hanging onto the
1940 * superblock lock.
1941 */
1942 EXT3_SB(sb)->s_mount_state |= EXT3_ORPHAN_FS;
1943 ext3_orphan_cleanup(sb, es);
1944 EXT3_SB(sb)->s_mount_state &= ~EXT3_ORPHAN_FS;
1945 if (needs_recovery)
1946 printk (KERN_INFO "EXT3-fs: recovery complete.\n");
1947 ext3_mark_recovery_complete(sb, es);
1948 printk (KERN_INFO "EXT3-fs: mounted filesystem with %s data mode.\n",
1949 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA ? "journal":
1950 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA ? "ordered":
1951 "writeback");
1952
1953 lock_kernel();
1954 return 0;
1955
1956 cantfind_ext3:
1957 if (!silent)
1958 printk(KERN_ERR "VFS: Can't find ext3 filesystem on dev %s.\n",
1959 sb->s_id);
1960 goto failed_mount;
1961
1962 failed_mount4:
1963 journal_destroy(sbi->s_journal);
1964 failed_mount3:
1965 percpu_counter_destroy(&sbi->s_freeblocks_counter);
1966 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1967 percpu_counter_destroy(&sbi->s_dirs_counter);
1968 failed_mount2:
1969 for (i = 0; i < db_count; i++)
1970 brelse(sbi->s_group_desc[i]);
1971 kfree(sbi->s_group_desc);
1972 failed_mount:
1973 #ifdef CONFIG_QUOTA
1974 for (i = 0; i < MAXQUOTAS; i++)
1975 kfree(sbi->s_qf_names[i]);
1976 #endif
1977 ext3_blkdev_remove(sbi);
1978 brelse(bh);
1979 out_fail:
1980 sb->s_fs_info = NULL;
1981 kfree(sbi);
1982 lock_kernel();
1983 return ret;
1984 }
1985
1986 /*
1987 * Setup any per-fs journal parameters now. We'll do this both on
1988 * initial mount, once the journal has been initialised but before we've
1989 * done any recovery; and again on any subsequent remount.
1990 */
1991 static void ext3_init_journal_params(struct super_block *sb, journal_t *journal)
1992 {
1993 struct ext3_sb_info *sbi = EXT3_SB(sb);
1994
1995 if (sbi->s_commit_interval)
1996 journal->j_commit_interval = sbi->s_commit_interval;
1997 /* We could also set up an ext3-specific default for the commit
1998 * interval here, but for now we'll just fall back to the jbd
1999 * default. */
2000
2001 spin_lock(&journal->j_state_lock);
2002 if (test_opt(sb, BARRIER))
2003 journal->j_flags |= JFS_BARRIER;
2004 else
2005 journal->j_flags &= ~JFS_BARRIER;
2006 if (test_opt(sb, DATA_ERR_ABORT))
2007 journal->j_flags |= JFS_ABORT_ON_SYNCDATA_ERR;
2008 else
2009 journal->j_flags &= ~JFS_ABORT_ON_SYNCDATA_ERR;
2010 spin_unlock(&journal->j_state_lock);
2011 }
2012
2013 static journal_t *ext3_get_journal(struct super_block *sb,
2014 unsigned int journal_inum)
2015 {
2016 struct inode *journal_inode;
2017 journal_t *journal;
2018
2019 /* First, test for the existence of a valid inode on disk. Bad
2020 * things happen if we iget() an unused inode, as the subsequent
2021 * iput() will try to delete it. */
2022
2023 journal_inode = ext3_iget(sb, journal_inum);
2024 if (IS_ERR(journal_inode)) {
2025 printk(KERN_ERR "EXT3-fs: no journal found.\n");
2026 return NULL;
2027 }
2028 if (!journal_inode->i_nlink) {
2029 make_bad_inode(journal_inode);
2030 iput(journal_inode);
2031 printk(KERN_ERR "EXT3-fs: journal inode is deleted.\n");
2032 return NULL;
2033 }
2034
2035 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
2036 journal_inode, journal_inode->i_size);
2037 if (!S_ISREG(journal_inode->i_mode)) {
2038 printk(KERN_ERR "EXT3-fs: invalid journal inode.\n");
2039 iput(journal_inode);
2040 return NULL;
2041 }
2042
2043 journal = journal_init_inode(journal_inode);
2044 if (!journal) {
2045 printk(KERN_ERR "EXT3-fs: Could not load journal inode\n");
2046 iput(journal_inode);
2047 return NULL;
2048 }
2049 journal->j_private = sb;
2050 ext3_init_journal_params(sb, journal);
2051 return journal;
2052 }
2053
2054 static journal_t *ext3_get_dev_journal(struct super_block *sb,
2055 dev_t j_dev)
2056 {
2057 struct buffer_head * bh;
2058 journal_t *journal;
2059 ext3_fsblk_t start;
2060 ext3_fsblk_t len;
2061 int hblock, blocksize;
2062 ext3_fsblk_t sb_block;
2063 unsigned long offset;
2064 struct ext3_super_block * es;
2065 struct block_device *bdev;
2066
2067 bdev = ext3_blkdev_get(j_dev);
2068 if (bdev == NULL)
2069 return NULL;
2070
2071 if (bd_claim(bdev, sb)) {
2072 printk(KERN_ERR
2073 "EXT3: failed to claim external journal device.\n");
2074 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
2075 return NULL;
2076 }
2077
2078 blocksize = sb->s_blocksize;
2079 hblock = bdev_hardsect_size(bdev);
2080 if (blocksize < hblock) {
2081 printk(KERN_ERR
2082 "EXT3-fs: blocksize too small for journal device.\n");
2083 goto out_bdev;
2084 }
2085
2086 sb_block = EXT3_MIN_BLOCK_SIZE / blocksize;
2087 offset = EXT3_MIN_BLOCK_SIZE % blocksize;
2088 set_blocksize(bdev, blocksize);
2089 if (!(bh = __bread(bdev, sb_block, blocksize))) {
2090 printk(KERN_ERR "EXT3-fs: couldn't read superblock of "
2091 "external journal\n");
2092 goto out_bdev;
2093 }
2094
2095 es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
2096 if ((le16_to_cpu(es->s_magic) != EXT3_SUPER_MAGIC) ||
2097 !(le32_to_cpu(es->s_feature_incompat) &
2098 EXT3_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2099 printk(KERN_ERR "EXT3-fs: external journal has "
2100 "bad superblock\n");
2101 brelse(bh);
2102 goto out_bdev;
2103 }
2104
2105 if (memcmp(EXT3_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2106 printk(KERN_ERR "EXT3-fs: journal UUID does not match\n");
2107 brelse(bh);
2108 goto out_bdev;
2109 }
2110
2111 len = le32_to_cpu(es->s_blocks_count);
2112 start = sb_block + 1;
2113 brelse(bh); /* we're done with the superblock */
2114
2115 journal = journal_init_dev(bdev, sb->s_bdev,
2116 start, len, blocksize);
2117 if (!journal) {
2118 printk(KERN_ERR "EXT3-fs: failed to create device journal\n");
2119 goto out_bdev;
2120 }
2121 journal->j_private = sb;
2122 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2123 wait_on_buffer(journal->j_sb_buffer);
2124 if (!buffer_uptodate(journal->j_sb_buffer)) {
2125 printk(KERN_ERR "EXT3-fs: I/O error on journal device\n");
2126 goto out_journal;
2127 }
2128 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
2129 printk(KERN_ERR "EXT3-fs: External journal has more than one "
2130 "user (unsupported) - %d\n",
2131 be32_to_cpu(journal->j_superblock->s_nr_users));
2132 goto out_journal;
2133 }
2134 EXT3_SB(sb)->journal_bdev = bdev;
2135 ext3_init_journal_params(sb, journal);
2136 return journal;
2137 out_journal:
2138 journal_destroy(journal);
2139 out_bdev:
2140 ext3_blkdev_put(bdev);
2141 return NULL;
2142 }
2143
2144 static int ext3_load_journal(struct super_block *sb,
2145 struct ext3_super_block *es,
2146 unsigned long journal_devnum)
2147 {
2148 journal_t *journal;
2149 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2150 dev_t journal_dev;
2151 int err = 0;
2152 int really_read_only;
2153
2154 if (journal_devnum &&
2155 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2156 printk(KERN_INFO "EXT3-fs: external journal device major/minor "
2157 "numbers have changed\n");
2158 journal_dev = new_decode_dev(journal_devnum);
2159 } else
2160 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2161
2162 really_read_only = bdev_read_only(sb->s_bdev);
2163
2164 /*
2165 * Are we loading a blank journal or performing recovery after a
2166 * crash? For recovery, we need to check in advance whether we
2167 * can get read-write access to the device.
2168 */
2169
2170 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER)) {
2171 if (sb->s_flags & MS_RDONLY) {
2172 printk(KERN_INFO "EXT3-fs: INFO: recovery "
2173 "required on readonly filesystem.\n");
2174 if (really_read_only) {
2175 printk(KERN_ERR "EXT3-fs: write access "
2176 "unavailable, cannot proceed.\n");
2177 return -EROFS;
2178 }
2179 printk (KERN_INFO "EXT3-fs: write access will "
2180 "be enabled during recovery.\n");
2181 }
2182 }
2183
2184 if (journal_inum && journal_dev) {
2185 printk(KERN_ERR "EXT3-fs: filesystem has both journal "
2186 "and inode journals!\n");
2187 return -EINVAL;
2188 }
2189
2190 if (journal_inum) {
2191 if (!(journal = ext3_get_journal(sb, journal_inum)))
2192 return -EINVAL;
2193 } else {
2194 if (!(journal = ext3_get_dev_journal(sb, journal_dev)))
2195 return -EINVAL;
2196 }
2197
2198 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2199 err = journal_update_format(journal);
2200 if (err) {
2201 printk(KERN_ERR "EXT3-fs: error updating journal.\n");
2202 journal_destroy(journal);
2203 return err;
2204 }
2205 }
2206
2207 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER))
2208 err = journal_wipe(journal, !really_read_only);
2209 if (!err)
2210 err = journal_load(journal);
2211
2212 if (err) {
2213 printk(KERN_ERR "EXT3-fs: error loading journal.\n");
2214 journal_destroy(journal);
2215 return err;
2216 }
2217
2218 EXT3_SB(sb)->s_journal = journal;
2219 ext3_clear_journal_err(sb, es);
2220
2221 if (journal_devnum &&
2222 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2223 es->s_journal_dev = cpu_to_le32(journal_devnum);
2224 sb->s_dirt = 1;
2225
2226 /* Make sure we flush the recovery flag to disk. */
2227 ext3_commit_super(sb, es, 1);
2228 }
2229
2230 return 0;
2231 }
2232
2233 static int ext3_create_journal(struct super_block * sb,
2234 struct ext3_super_block * es,
2235 unsigned int journal_inum)
2236 {
2237 journal_t *journal;
2238 int err;
2239
2240 if (sb->s_flags & MS_RDONLY) {
2241 printk(KERN_ERR "EXT3-fs: readonly filesystem when trying to "
2242 "create journal.\n");
2243 return -EROFS;
2244 }
2245
2246 journal = ext3_get_journal(sb, journal_inum);
2247 if (!journal)
2248 return -EINVAL;
2249
2250 printk(KERN_INFO "EXT3-fs: creating new journal on inode %u\n",
2251 journal_inum);
2252
2253 err = journal_create(journal);
2254 if (err) {
2255 printk(KERN_ERR "EXT3-fs: error creating journal.\n");
2256 journal_destroy(journal);
2257 return -EIO;
2258 }
2259
2260 EXT3_SB(sb)->s_journal = journal;
2261
2262 ext3_update_dynamic_rev(sb);
2263 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2264 EXT3_SET_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL);
2265
2266 es->s_journal_inum = cpu_to_le32(journal_inum);
2267 sb->s_dirt = 1;
2268
2269 /* Make sure we flush the recovery flag to disk. */
2270 ext3_commit_super(sb, es, 1);
2271
2272 return 0;
2273 }
2274
2275 static void ext3_commit_super (struct super_block * sb,
2276 struct ext3_super_block * es,
2277 int sync)
2278 {
2279 struct buffer_head *sbh = EXT3_SB(sb)->s_sbh;
2280
2281 if (!sbh)
2282 return;
2283 es->s_wtime = cpu_to_le32(get_seconds());
2284 es->s_free_blocks_count = cpu_to_le32(ext3_count_free_blocks(sb));
2285 es->s_free_inodes_count = cpu_to_le32(ext3_count_free_inodes(sb));
2286 BUFFER_TRACE(sbh, "marking dirty");
2287 mark_buffer_dirty(sbh);
2288 if (sync)
2289 sync_dirty_buffer(sbh);
2290 }
2291
2292
2293 /*
2294 * Have we just finished recovery? If so, and if we are mounting (or
2295 * remounting) the filesystem readonly, then we will end up with a
2296 * consistent fs on disk. Record that fact.
2297 */
2298 static void ext3_mark_recovery_complete(struct super_block * sb,
2299 struct ext3_super_block * es)
2300 {
2301 journal_t *journal = EXT3_SB(sb)->s_journal;
2302
2303 journal_lock_updates(journal);
2304 if (journal_flush(journal) < 0)
2305 goto out;
2306
2307 lock_super(sb);
2308 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER) &&
2309 sb->s_flags & MS_RDONLY) {
2310 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2311 sb->s_dirt = 0;
2312 ext3_commit_super(sb, es, 1);
2313 }
2314 unlock_super(sb);
2315
2316 out:
2317 journal_unlock_updates(journal);
2318 }
2319
2320 /*
2321 * If we are mounting (or read-write remounting) a filesystem whose journal
2322 * has recorded an error from a previous lifetime, move that error to the
2323 * main filesystem now.
2324 */
2325 static void ext3_clear_journal_err(struct super_block * sb,
2326 struct ext3_super_block * es)
2327 {
2328 journal_t *journal;
2329 int j_errno;
2330 const char *errstr;
2331
2332 journal = EXT3_SB(sb)->s_journal;
2333
2334 /*
2335 * Now check for any error status which may have been recorded in the
2336 * journal by a prior ext3_error() or ext3_abort()
2337 */
2338
2339 j_errno = journal_errno(journal);
2340 if (j_errno) {
2341 char nbuf[16];
2342
2343 errstr = ext3_decode_error(sb, j_errno, nbuf);
2344 ext3_warning(sb, __func__, "Filesystem error recorded "
2345 "from previous mount: %s", errstr);
2346 ext3_warning(sb, __func__, "Marking fs in need of "
2347 "filesystem check.");
2348
2349 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
2350 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
2351 ext3_commit_super (sb, es, 1);
2352
2353 journal_clear_err(journal);
2354 }
2355 }
2356
2357 /*
2358 * Force the running and committing transactions to commit,
2359 * and wait on the commit.
2360 */
2361 int ext3_force_commit(struct super_block *sb)
2362 {
2363 journal_t *journal;
2364 int ret;
2365
2366 if (sb->s_flags & MS_RDONLY)
2367 return 0;
2368
2369 journal = EXT3_SB(sb)->s_journal;
2370 sb->s_dirt = 0;
2371 ret = ext3_journal_force_commit(journal);
2372 return ret;
2373 }
2374
2375 /*
2376 * Ext3 always journals updates to the superblock itself, so we don't
2377 * have to propagate any other updates to the superblock on disk at this
2378 * point. (We can probably nuke this function altogether, and remove
2379 * any mention to sb->s_dirt in all of fs/ext3; eventual cleanup...)
2380 */
2381 static void ext3_write_super (struct super_block * sb)
2382 {
2383 if (mutex_trylock(&sb->s_lock) != 0)
2384 BUG();
2385 sb->s_dirt = 0;
2386 }
2387
2388 static int ext3_sync_fs(struct super_block *sb, int wait)
2389 {
2390 sb->s_dirt = 0;
2391 if (wait)
2392 ext3_force_commit(sb);
2393 else
2394 journal_start_commit(EXT3_SB(sb)->s_journal, NULL);
2395
2396 return 0;
2397 }
2398
2399 /*
2400 * LVM calls this function before a (read-only) snapshot is created. This
2401 * gives us a chance to flush the journal completely and mark the fs clean.
2402 */
2403 static void ext3_write_super_lockfs(struct super_block *sb)
2404 {
2405 sb->s_dirt = 0;
2406
2407 if (!(sb->s_flags & MS_RDONLY)) {
2408 journal_t *journal = EXT3_SB(sb)->s_journal;
2409
2410 /* Now we set up the journal barrier. */
2411 journal_lock_updates(journal);
2412
2413 /*
2414 * We don't want to clear needs_recovery flag when we failed
2415 * to flush the journal.
2416 */
2417 if (journal_flush(journal) < 0)
2418 return;
2419
2420 /* Journal blocked and flushed, clear needs_recovery flag. */
2421 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2422 ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2423 }
2424 }
2425
2426 /*
2427 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2428 * flag here, even though the filesystem is not technically dirty yet.
2429 */
2430 static void ext3_unlockfs(struct super_block *sb)
2431 {
2432 if (!(sb->s_flags & MS_RDONLY)) {
2433 lock_super(sb);
2434 /* Reser the needs_recovery flag before the fs is unlocked. */
2435 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2436 ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2437 unlock_super(sb);
2438 journal_unlock_updates(EXT3_SB(sb)->s_journal);
2439 }
2440 }
2441
2442 static int ext3_remount (struct super_block * sb, int * flags, char * data)
2443 {
2444 struct ext3_super_block * es;
2445 struct ext3_sb_info *sbi = EXT3_SB(sb);
2446 ext3_fsblk_t n_blocks_count = 0;
2447 unsigned long old_sb_flags;
2448 struct ext3_mount_options old_opts;
2449 int err;
2450 #ifdef CONFIG_QUOTA
2451 int i;
2452 #endif
2453
2454 /* Store the original options */
2455 old_sb_flags = sb->s_flags;
2456 old_opts.s_mount_opt = sbi->s_mount_opt;
2457 old_opts.s_resuid = sbi->s_resuid;
2458 old_opts.s_resgid = sbi->s_resgid;
2459 old_opts.s_commit_interval = sbi->s_commit_interval;
2460 #ifdef CONFIG_QUOTA
2461 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2462 for (i = 0; i < MAXQUOTAS; i++)
2463 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2464 #endif
2465
2466 /*
2467 * Allow the "check" option to be passed as a remount option.
2468 */
2469 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
2470 err = -EINVAL;
2471 goto restore_opts;
2472 }
2473
2474 if (sbi->s_mount_opt & EXT3_MOUNT_ABORT)
2475 ext3_abort(sb, __func__, "Abort forced by user");
2476
2477 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2478 ((sbi->s_mount_opt & EXT3_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2479
2480 es = sbi->s_es;
2481
2482 ext3_init_journal_params(sb, sbi->s_journal);
2483
2484 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
2485 n_blocks_count > le32_to_cpu(es->s_blocks_count)) {
2486 if (sbi->s_mount_opt & EXT3_MOUNT_ABORT) {
2487 err = -EROFS;
2488 goto restore_opts;
2489 }
2490
2491 if (*flags & MS_RDONLY) {
2492 /*
2493 * First of all, the unconditional stuff we have to do
2494 * to disable replay of the journal when we next remount
2495 */
2496 sb->s_flags |= MS_RDONLY;
2497
2498 /*
2499 * OK, test if we are remounting a valid rw partition
2500 * readonly, and if so set the rdonly flag and then
2501 * mark the partition as valid again.
2502 */
2503 if (!(es->s_state & cpu_to_le16(EXT3_VALID_FS)) &&
2504 (sbi->s_mount_state & EXT3_VALID_FS))
2505 es->s_state = cpu_to_le16(sbi->s_mount_state);
2506
2507 /*
2508 * We have to unlock super so that we can wait for
2509 * transactions.
2510 */
2511 unlock_super(sb);
2512 ext3_mark_recovery_complete(sb, es);
2513 lock_super(sb);
2514 } else {
2515 __le32 ret;
2516 if ((ret = EXT3_HAS_RO_COMPAT_FEATURE(sb,
2517 ~EXT3_FEATURE_RO_COMPAT_SUPP))) {
2518 printk(KERN_WARNING "EXT3-fs: %s: couldn't "
2519 "remount RDWR because of unsupported "
2520 "optional features (%x).\n",
2521 sb->s_id, le32_to_cpu(ret));
2522 err = -EROFS;
2523 goto restore_opts;
2524 }
2525
2526 /*
2527 * If we have an unprocessed orphan list hanging
2528 * around from a previously readonly bdev mount,
2529 * require a full umount/remount for now.
2530 */
2531 if (es->s_last_orphan) {
2532 printk(KERN_WARNING "EXT3-fs: %s: couldn't "
2533 "remount RDWR because of unprocessed "
2534 "orphan inode list. Please "
2535 "umount/remount instead.\n",
2536 sb->s_id);
2537 err = -EINVAL;
2538 goto restore_opts;
2539 }
2540
2541 /*
2542 * Mounting a RDONLY partition read-write, so reread
2543 * and store the current valid flag. (It may have
2544 * been changed by e2fsck since we originally mounted
2545 * the partition.)
2546 */
2547 ext3_clear_journal_err(sb, es);
2548 sbi->s_mount_state = le16_to_cpu(es->s_state);
2549 if ((err = ext3_group_extend(sb, es, n_blocks_count)))
2550 goto restore_opts;
2551 if (!ext3_setup_super (sb, es, 0))
2552 sb->s_flags &= ~MS_RDONLY;
2553 }
2554 }
2555 #ifdef CONFIG_QUOTA
2556 /* Release old quota file names */
2557 for (i = 0; i < MAXQUOTAS; i++)
2558 if (old_opts.s_qf_names[i] &&
2559 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2560 kfree(old_opts.s_qf_names[i]);
2561 #endif
2562 return 0;
2563 restore_opts:
2564 sb->s_flags = old_sb_flags;
2565 sbi->s_mount_opt = old_opts.s_mount_opt;
2566 sbi->s_resuid = old_opts.s_resuid;
2567 sbi->s_resgid = old_opts.s_resgid;
2568 sbi->s_commit_interval = old_opts.s_commit_interval;
2569 #ifdef CONFIG_QUOTA
2570 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
2571 for (i = 0; i < MAXQUOTAS; i++) {
2572 if (sbi->s_qf_names[i] &&
2573 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2574 kfree(sbi->s_qf_names[i]);
2575 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
2576 }
2577 #endif
2578 return err;
2579 }
2580
2581 static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf)
2582 {
2583 struct super_block *sb = dentry->d_sb;
2584 struct ext3_sb_info *sbi = EXT3_SB(sb);
2585 struct ext3_super_block *es = sbi->s_es;
2586 u64 fsid;
2587
2588 if (test_opt(sb, MINIX_DF)) {
2589 sbi->s_overhead_last = 0;
2590 } else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
2591 unsigned long ngroups = sbi->s_groups_count, i;
2592 ext3_fsblk_t overhead = 0;
2593 smp_rmb();
2594
2595 /*
2596 * Compute the overhead (FS structures). This is constant
2597 * for a given filesystem unless the number of block groups
2598 * changes so we cache the previous value until it does.
2599 */
2600
2601 /*
2602 * All of the blocks before first_data_block are
2603 * overhead
2604 */
2605 overhead = le32_to_cpu(es->s_first_data_block);
2606
2607 /*
2608 * Add the overhead attributed to the superblock and
2609 * block group descriptors. If the sparse superblocks
2610 * feature is turned on, then not all groups have this.
2611 */
2612 for (i = 0; i < ngroups; i++) {
2613 overhead += ext3_bg_has_super(sb, i) +
2614 ext3_bg_num_gdb(sb, i);
2615 cond_resched();
2616 }
2617
2618 /*
2619 * Every block group has an inode bitmap, a block
2620 * bitmap, and an inode table.
2621 */
2622 overhead += ngroups * (2 + sbi->s_itb_per_group);
2623 sbi->s_overhead_last = overhead;
2624 smp_wmb();
2625 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
2626 }
2627
2628 buf->f_type = EXT3_SUPER_MAGIC;
2629 buf->f_bsize = sb->s_blocksize;
2630 buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
2631 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter);
2632 es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
2633 buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
2634 if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
2635 buf->f_bavail = 0;
2636 buf->f_files = le32_to_cpu(es->s_inodes_count);
2637 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
2638 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
2639 buf->f_namelen = EXT3_NAME_LEN;
2640 fsid = le64_to_cpup((void *)es->s_uuid) ^
2641 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
2642 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
2643 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
2644 return 0;
2645 }
2646
2647 /* Helper function for writing quotas on sync - we need to start transaction before quota file
2648 * is locked for write. Otherwise the are possible deadlocks:
2649 * Process 1 Process 2
2650 * ext3_create() quota_sync()
2651 * journal_start() write_dquot()
2652 * DQUOT_INIT() down(dqio_mutex)
2653 * down(dqio_mutex) journal_start()
2654 *
2655 */
2656
2657 #ifdef CONFIG_QUOTA
2658
2659 static inline struct inode *dquot_to_inode(struct dquot *dquot)
2660 {
2661 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
2662 }
2663
2664 static int ext3_dquot_initialize(struct inode *inode, int type)
2665 {
2666 handle_t *handle;
2667 int ret, err;
2668
2669 /* We may create quota structure so we need to reserve enough blocks */
2670 handle = ext3_journal_start(inode, 2*EXT3_QUOTA_INIT_BLOCKS(inode->i_sb));
2671 if (IS_ERR(handle))
2672 return PTR_ERR(handle);
2673 ret = dquot_initialize(inode, type);
2674 err = ext3_journal_stop(handle);
2675 if (!ret)
2676 ret = err;
2677 return ret;
2678 }
2679
2680 static int ext3_dquot_drop(struct inode *inode)
2681 {
2682 handle_t *handle;
2683 int ret, err;
2684
2685 /* We may delete quota structure so we need to reserve enough blocks */
2686 handle = ext3_journal_start(inode, 2*EXT3_QUOTA_DEL_BLOCKS(inode->i_sb));
2687 if (IS_ERR(handle)) {
2688 /*
2689 * We call dquot_drop() anyway to at least release references
2690 * to quota structures so that umount does not hang.
2691 */
2692 dquot_drop(inode);
2693 return PTR_ERR(handle);
2694 }
2695 ret = dquot_drop(inode);
2696 err = ext3_journal_stop(handle);
2697 if (!ret)
2698 ret = err;
2699 return ret;
2700 }
2701
2702 static int ext3_write_dquot(struct dquot *dquot)
2703 {
2704 int ret, err;
2705 handle_t *handle;
2706 struct inode *inode;
2707
2708 inode = dquot_to_inode(dquot);
2709 handle = ext3_journal_start(inode,
2710 EXT3_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
2711 if (IS_ERR(handle))
2712 return PTR_ERR(handle);
2713 ret = dquot_commit(dquot);
2714 err = ext3_journal_stop(handle);
2715 if (!ret)
2716 ret = err;
2717 return ret;
2718 }
2719
2720 static int ext3_acquire_dquot(struct dquot *dquot)
2721 {
2722 int ret, err;
2723 handle_t *handle;
2724
2725 handle = ext3_journal_start(dquot_to_inode(dquot),
2726 EXT3_QUOTA_INIT_BLOCKS(dquot->dq_sb));
2727 if (IS_ERR(handle))
2728 return PTR_ERR(handle);
2729 ret = dquot_acquire(dquot);
2730 err = ext3_journal_stop(handle);
2731 if (!ret)
2732 ret = err;
2733 return ret;
2734 }
2735
2736 static int ext3_release_dquot(struct dquot *dquot)
2737 {
2738 int ret, err;
2739 handle_t *handle;
2740
2741 handle = ext3_journal_start(dquot_to_inode(dquot),
2742 EXT3_QUOTA_DEL_BLOCKS(dquot->dq_sb));
2743 if (IS_ERR(handle)) {
2744 /* Release dquot anyway to avoid endless cycle in dqput() */
2745 dquot_release(dquot);
2746 return PTR_ERR(handle);
2747 }
2748 ret = dquot_release(dquot);
2749 err = ext3_journal_stop(handle);
2750 if (!ret)
2751 ret = err;
2752 return ret;
2753 }
2754
2755 static int ext3_mark_dquot_dirty(struct dquot *dquot)
2756 {
2757 /* Are we journaling quotas? */
2758 if (EXT3_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
2759 EXT3_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
2760 dquot_mark_dquot_dirty(dquot);
2761 return ext3_write_dquot(dquot);
2762 } else {
2763 return dquot_mark_dquot_dirty(dquot);
2764 }
2765 }
2766
2767 static int ext3_write_info(struct super_block *sb, int type)
2768 {
2769 int ret, err;
2770 handle_t *handle;
2771
2772 /* Data block + inode block */
2773 handle = ext3_journal_start(sb->s_root->d_inode, 2);
2774 if (IS_ERR(handle))
2775 return PTR_ERR(handle);
2776 ret = dquot_commit_info(sb, type);
2777 err = ext3_journal_stop(handle);
2778 if (!ret)
2779 ret = err;
2780 return ret;
2781 }
2782
2783 /*
2784 * Turn on quotas during mount time - we need to find
2785 * the quota file and such...
2786 */
2787 static int ext3_quota_on_mount(struct super_block *sb, int type)
2788 {
2789 return vfs_quota_on_mount(sb, EXT3_SB(sb)->s_qf_names[type],
2790 EXT3_SB(sb)->s_jquota_fmt, type);
2791 }
2792
2793 /*
2794 * Standard function to be called on quota_on
2795 */
2796 static int ext3_quota_on(struct super_block *sb, int type, int format_id,
2797 char *name, int remount)
2798 {
2799 int err;
2800 struct path path;
2801
2802 if (!test_opt(sb, QUOTA))
2803 return -EINVAL;
2804 /* When remounting, no checks are needed and in fact, name is NULL */
2805 if (remount)
2806 return vfs_quota_on(sb, type, format_id, name, remount);
2807
2808 err = kern_path(name, LOOKUP_FOLLOW, &path);
2809 if (err)
2810 return err;
2811
2812 /* Quotafile not on the same filesystem? */
2813 if (path.mnt->mnt_sb != sb) {
2814 path_put(&path);
2815 return -EXDEV;
2816 }
2817 /* Journaling quota? */
2818 if (EXT3_SB(sb)->s_qf_names[type]) {
2819 /* Quotafile not of fs root? */
2820 if (path.dentry->d_parent != sb->s_root)
2821 printk(KERN_WARNING
2822 "EXT3-fs: Quota file not on filesystem root. "
2823 "Journaled quota will not work.\n");
2824 }
2825
2826 /*
2827 * When we journal data on quota file, we have to flush journal to see
2828 * all updates to the file when we bypass pagecache...
2829 */
2830 if (ext3_should_journal_data(path.dentry->d_inode)) {
2831 /*
2832 * We don't need to lock updates but journal_flush() could
2833 * otherwise be livelocked...
2834 */
2835 journal_lock_updates(EXT3_SB(sb)->s_journal);
2836 err = journal_flush(EXT3_SB(sb)->s_journal);
2837 journal_unlock_updates(EXT3_SB(sb)->s_journal);
2838 if (err) {
2839 path_put(&path);
2840 return err;
2841 }
2842 }
2843
2844 err = vfs_quota_on_path(sb, type, format_id, &path);
2845 path_put(&path);
2846 return err;
2847 }
2848
2849 /* Read data from quotafile - avoid pagecache and such because we cannot afford
2850 * acquiring the locks... As quota files are never truncated and quota code
2851 * itself serializes the operations (and noone else should touch the files)
2852 * we don't have to be afraid of races */
2853 static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
2854 size_t len, loff_t off)
2855 {
2856 struct inode *inode = sb_dqopt(sb)->files[type];
2857 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2858 int err = 0;
2859 int offset = off & (sb->s_blocksize - 1);
2860 int tocopy;
2861 size_t toread;
2862 struct buffer_head *bh;
2863 loff_t i_size = i_size_read(inode);
2864
2865 if (off > i_size)
2866 return 0;
2867 if (off+len > i_size)
2868 len = i_size-off;
2869 toread = len;
2870 while (toread > 0) {
2871 tocopy = sb->s_blocksize - offset < toread ?
2872 sb->s_blocksize - offset : toread;
2873 bh = ext3_bread(NULL, inode, blk, 0, &err);
2874 if (err)
2875 return err;
2876 if (!bh) /* A hole? */
2877 memset(data, 0, tocopy);
2878 else
2879 memcpy(data, bh->b_data+offset, tocopy);
2880 brelse(bh);
2881 offset = 0;
2882 toread -= tocopy;
2883 data += tocopy;
2884 blk++;
2885 }
2886 return len;
2887 }
2888
2889 /* Write to quotafile (we know the transaction is already started and has
2890 * enough credits) */
2891 static ssize_t ext3_quota_write(struct super_block *sb, int type,
2892 const char *data, size_t len, loff_t off)
2893 {
2894 struct inode *inode = sb_dqopt(sb)->files[type];
2895 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2896 int err = 0;
2897 int offset = off & (sb->s_blocksize - 1);
2898 int tocopy;
2899 int journal_quota = EXT3_SB(sb)->s_qf_names[type] != NULL;
2900 size_t towrite = len;
2901 struct buffer_head *bh;
2902 handle_t *handle = journal_current_handle();
2903
2904 if (!handle) {
2905 printk(KERN_WARNING "EXT3-fs: Quota write (off=%Lu, len=%Lu)"
2906 " cancelled because transaction is not started.\n",
2907 (unsigned long long)off, (unsigned long long)len);
2908 return -EIO;
2909 }
2910 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
2911 while (towrite > 0) {
2912 tocopy = sb->s_blocksize - offset < towrite ?
2913 sb->s_blocksize - offset : towrite;
2914 bh = ext3_bread(handle, inode, blk, 1, &err);
2915 if (!bh)
2916 goto out;
2917 if (journal_quota) {
2918 err = ext3_journal_get_write_access(handle, bh);
2919 if (err) {
2920 brelse(bh);
2921 goto out;
2922 }
2923 }
2924 lock_buffer(bh);
2925 memcpy(bh->b_data+offset, data, tocopy);
2926 flush_dcache_page(bh->b_page);
2927 unlock_buffer(bh);
2928 if (journal_quota)
2929 err = ext3_journal_dirty_metadata(handle, bh);
2930 else {
2931 /* Always do at least ordered writes for quotas */
2932 err = ext3_journal_dirty_data(handle, bh);
2933 mark_buffer_dirty(bh);
2934 }
2935 brelse(bh);
2936 if (err)
2937 goto out;
2938 offset = 0;
2939 towrite -= tocopy;
2940 data += tocopy;
2941 blk++;
2942 }
2943 out:
2944 if (len == towrite) {
2945 mutex_unlock(&inode->i_mutex);
2946 return err;
2947 }
2948 if (inode->i_size < off+len-towrite) {
2949 i_size_write(inode, off+len-towrite);
2950 EXT3_I(inode)->i_disksize = inode->i_size;
2951 }
2952 inode->i_version++;
2953 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
2954 ext3_mark_inode_dirty(handle, inode);
2955 mutex_unlock(&inode->i_mutex);
2956 return len - towrite;
2957 }
2958
2959 #endif
2960
2961 static int ext3_get_sb(struct file_system_type *fs_type,
2962 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
2963 {
2964 return get_sb_bdev(fs_type, flags, dev_name, data, ext3_fill_super, mnt);
2965 }
2966
2967 static struct file_system_type ext3_fs_type = {
2968 .owner = THIS_MODULE,
2969 .name = "ext3",
2970 .get_sb = ext3_get_sb,
2971 .kill_sb = kill_block_super,
2972 .fs_flags = FS_REQUIRES_DEV,
2973 };
2974
2975 static int __init init_ext3_fs(void)
2976 {
2977 int err = init_ext3_xattr();
2978 if (err)
2979 return err;
2980 err = init_inodecache();
2981 if (err)
2982 goto out1;
2983 err = register_filesystem(&ext3_fs_type);
2984 if (err)
2985 goto out;
2986 return 0;
2987 out:
2988 destroy_inodecache();
2989 out1:
2990 exit_ext3_xattr();
2991 return err;
2992 }
2993
2994 static void __exit exit_ext3_fs(void)
2995 {
2996 unregister_filesystem(&ext3_fs_type);
2997 destroy_inodecache();
2998 exit_ext3_xattr();
2999 }
3000
3001 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
3002 MODULE_DESCRIPTION("Second Extended Filesystem with journaling extensions");
3003 MODULE_LICENSE("GPL");
3004 module_init(init_ext3_fs)
3005 module_exit(exit_ext3_fs)
Linux with added FPC-III support