?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
trace_skb: fix build when CONFIG_NET is not enabled
[linux-2.6/libata-dev.git] / fs / ext3 / super.c
blob524b349c6299ca248b366cf299071939e5843aba
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 #ifdef CONFIG_EXT3_DEFAULTS_TO_ORDERED
48 #define EXT3_MOUNT_DEFAULT_DATA_MODE EXT3_MOUNT_ORDERED_DATA
49 #else
50 #define EXT3_MOUNT_DEFAULT_DATA_MODE EXT3_MOUNT_WRITEBACK_DATA
51 #endif
52
53 static int ext3_load_journal(struct super_block *, struct ext3_super_block *,
54 unsigned long journal_devnum);
55 static int ext3_create_journal(struct super_block *, struct ext3_super_block *,
56 unsigned int);
57 static int ext3_commit_super(struct super_block *sb,
58 struct ext3_super_block *es,
59 int sync);
60 static void ext3_mark_recovery_complete(struct super_block * sb,
61 struct ext3_super_block * es);
62 static void ext3_clear_journal_err(struct super_block * sb,
63 struct ext3_super_block * es);
64 static int ext3_sync_fs(struct super_block *sb, int wait);
65 static const char *ext3_decode_error(struct super_block * sb, int errno,
66 char nbuf[16]);
67 static int ext3_remount (struct super_block * sb, int * flags, char * data);
68 static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf);
69 static int ext3_unfreeze(struct super_block *sb);
70 static int ext3_freeze(struct super_block *sb);
71
72 /*
73 * Wrappers for journal_start/end.
74 *
75 * The only special thing we need to do here is to make sure that all
76 * journal_end calls result in the superblock being marked dirty, so
77 * that sync() will call the filesystem's write_super callback if
78 * appropriate.
79 */
80 handle_t *ext3_journal_start_sb(struct super_block *sb, int nblocks)
81 {
82 journal_t *journal;
83
84 if (sb->s_flags & MS_RDONLY)
85 return ERR_PTR(-EROFS);
86
87 /* Special case here: if the journal has aborted behind our
88 * backs (eg. EIO in the commit thread), then we still need to
89 * take the FS itself readonly cleanly. */
90 journal = EXT3_SB(sb)->s_journal;
91 if (is_journal_aborted(journal)) {
92 ext3_abort(sb, __func__,
93 "Detected aborted journal");
94 return ERR_PTR(-EROFS);
95 }
96
97 return journal_start(journal, nblocks);
98 }
99
100 /*
101 * The only special thing we need to do here is to make sure that all
102 * journal_stop calls result in the superblock being marked dirty, so
103 * that sync() will call the filesystem's write_super callback if
104 * appropriate.
105 */
106 int __ext3_journal_stop(const char *where, handle_t *handle)
107 {
108 struct super_block *sb;
109 int err;
110 int rc;
111
112 sb = handle->h_transaction->t_journal->j_private;
113 err = handle->h_err;
114 rc = journal_stop(handle);
115
116 if (!err)
117 err = rc;
118 if (err)
119 __ext3_std_error(sb, where, err);
120 return err;
121 }
122
123 void ext3_journal_abort_handle(const char *caller, const char *err_fn,
124 struct buffer_head *bh, handle_t *handle, int err)
125 {
126 char nbuf[16];
127 const char *errstr = ext3_decode_error(NULL, err, nbuf);
128
129 if (bh)
130 BUFFER_TRACE(bh, "abort");
131
132 if (!handle->h_err)
133 handle->h_err = err;
134
135 if (is_handle_aborted(handle))
136 return;
137
138 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
139 caller, errstr, err_fn);
140
141 journal_abort_handle(handle);
142 }
143
144 /* Deal with the reporting of failure conditions on a filesystem such as
145 * inconsistencies detected or read IO failures.
146 *
147 * On ext2, we can store the error state of the filesystem in the
148 * superblock. That is not possible on ext3, because we may have other
149 * write ordering constraints on the superblock which prevent us from
150 * writing it out straight away; and given that the journal is about to
151 * be aborted, we can't rely on the current, or future, transactions to
152 * write out the superblock safely.
153 *
154 * We'll just use the journal_abort() error code to record an error in
155 * the journal instead. On recovery, the journal will compain about
156 * that error until we've noted it down and cleared it.
157 */
158
159 static void ext3_handle_error(struct super_block *sb)
160 {
161 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
162
163 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
164 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
165
166 if (sb->s_flags & MS_RDONLY)
167 return;
168
169 if (!test_opt (sb, ERRORS_CONT)) {
170 journal_t *journal = EXT3_SB(sb)->s_journal;
171
172 EXT3_SB(sb)->s_mount_opt |= EXT3_MOUNT_ABORT;
173 if (journal)
174 journal_abort(journal, -EIO);
175 }
176 if (test_opt (sb, ERRORS_RO)) {
177 printk (KERN_CRIT "Remounting filesystem read-only\n");
178 sb->s_flags |= MS_RDONLY;
179 }
180 ext3_commit_super(sb, es, 1);
181 if (test_opt(sb, ERRORS_PANIC))
182 panic("EXT3-fs (device %s): panic forced after error\n",
183 sb->s_id);
184 }
185
186 void ext3_error (struct super_block * sb, const char * function,
187 const char * fmt, ...)
188 {
189 va_list args;
190
191 va_start(args, fmt);
192 printk(KERN_CRIT "EXT3-fs error (device %s): %s: ",sb->s_id, function);
193 vprintk(fmt, args);
194 printk("\n");
195 va_end(args);
196
197 ext3_handle_error(sb);
198 }
199
200 static const char *ext3_decode_error(struct super_block * sb, int errno,
201 char nbuf[16])
202 {
203 char *errstr = NULL;
204
205 switch (errno) {
206 case -EIO:
207 errstr = "IO failure";
208 break;
209 case -ENOMEM:
210 errstr = "Out of memory";
211 break;
212 case -EROFS:
213 if (!sb || EXT3_SB(sb)->s_journal->j_flags & JFS_ABORT)
214 errstr = "Journal has aborted";
215 else
216 errstr = "Readonly filesystem";
217 break;
218 default:
219 /* If the caller passed in an extra buffer for unknown
220 * errors, textualise them now. Else we just return
221 * NULL. */
222 if (nbuf) {
223 /* Check for truncated error codes... */
224 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
225 errstr = nbuf;
226 }
227 break;
228 }
229
230 return errstr;
231 }
232
233 /* __ext3_std_error decodes expected errors from journaling functions
234 * automatically and invokes the appropriate error response. */
235
236 void __ext3_std_error (struct super_block * sb, const char * function,
237 int errno)
238 {
239 char nbuf[16];
240 const char *errstr;
241
242 /* Special case: if the error is EROFS, and we're not already
243 * inside a transaction, then there's really no point in logging
244 * an error. */
245 if (errno == -EROFS && journal_current_handle() == NULL &&
246 (sb->s_flags & MS_RDONLY))
247 return;
248
249 errstr = ext3_decode_error(sb, errno, nbuf);
250 printk (KERN_CRIT "EXT3-fs error (device %s) in %s: %s\n",
251 sb->s_id, function, errstr);
252
253 ext3_handle_error(sb);
254 }
255
256 /*
257 * ext3_abort is a much stronger failure handler than ext3_error. The
258 * abort function may be used to deal with unrecoverable failures such
259 * as journal IO errors or ENOMEM at a critical moment in log management.
260 *
261 * We unconditionally force the filesystem into an ABORT|READONLY state,
262 * unless the error response on the fs has been set to panic in which
263 * case we take the easy way out and panic immediately.
264 */
265
266 void ext3_abort (struct super_block * sb, const char * function,
267 const char * fmt, ...)
268 {
269 va_list args;
270
271 printk (KERN_CRIT "ext3_abort called.\n");
272
273 va_start(args, fmt);
274 printk(KERN_CRIT "EXT3-fs error (device %s): %s: ",sb->s_id, function);
275 vprintk(fmt, args);
276 printk("\n");
277 va_end(args);
278
279 if (test_opt(sb, ERRORS_PANIC))
280 panic("EXT3-fs panic from previous error\n");
281
282 if (sb->s_flags & MS_RDONLY)
283 return;
284
285 printk(KERN_CRIT "Remounting filesystem read-only\n");
286 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
287 sb->s_flags |= MS_RDONLY;
288 EXT3_SB(sb)->s_mount_opt |= EXT3_MOUNT_ABORT;
289 if (EXT3_SB(sb)->s_journal)
290 journal_abort(EXT3_SB(sb)->s_journal, -EIO);
291 }
292
293 void ext3_warning (struct super_block * sb, const char * function,
294 const char * fmt, ...)
295 {
296 va_list args;
297
298 va_start(args, fmt);
299 printk(KERN_WARNING "EXT3-fs warning (device %s): %s: ",
300 sb->s_id, function);
301 vprintk(fmt, args);
302 printk("\n");
303 va_end(args);
304 }
305
306 void ext3_update_dynamic_rev(struct super_block *sb)
307 {
308 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
309
310 if (le32_to_cpu(es->s_rev_level) > EXT3_GOOD_OLD_REV)
311 return;
312
313 ext3_warning(sb, __func__,
314 "updating to rev %d because of new feature flag, "
315 "running e2fsck is recommended",
316 EXT3_DYNAMIC_REV);
317
318 es->s_first_ino = cpu_to_le32(EXT3_GOOD_OLD_FIRST_INO);
319 es->s_inode_size = cpu_to_le16(EXT3_GOOD_OLD_INODE_SIZE);
320 es->s_rev_level = cpu_to_le32(EXT3_DYNAMIC_REV);
321 /* leave es->s_feature_*compat flags alone */
322 /* es->s_uuid will be set by e2fsck if empty */
323
324 /*
325 * The rest of the superblock fields should be zero, and if not it
326 * means they are likely already in use, so leave them alone. We
327 * can leave it up to e2fsck to clean up any inconsistencies there.
328 */
329 }
330
331 /*
332 * Open the external journal device
333 */
334 static struct block_device *ext3_blkdev_get(dev_t dev)
335 {
336 struct block_device *bdev;
337 char b[BDEVNAME_SIZE];
338
339 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
340 if (IS_ERR(bdev))
341 goto fail;
342 return bdev;
343
344 fail:
345 printk(KERN_ERR "EXT3: failed to open journal device %s: %ld\n",
346 __bdevname(dev, b), PTR_ERR(bdev));
347 return NULL;
348 }
349
350 /*
351 * Release the journal device
352 */
353 static int ext3_blkdev_put(struct block_device *bdev)
354 {
355 bd_release(bdev);
356 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
357 }
358
359 static int ext3_blkdev_remove(struct ext3_sb_info *sbi)
360 {
361 struct block_device *bdev;
362 int ret = -ENODEV;
363
364 bdev = sbi->journal_bdev;
365 if (bdev) {
366 ret = ext3_blkdev_put(bdev);
367 sbi->journal_bdev = NULL;
368 }
369 return ret;
370 }
371
372 static inline struct inode *orphan_list_entry(struct list_head *l)
373 {
374 return &list_entry(l, struct ext3_inode_info, i_orphan)->vfs_inode;
375 }
376
377 static void dump_orphan_list(struct super_block *sb, struct ext3_sb_info *sbi)
378 {
379 struct list_head *l;
380
381 printk(KERN_ERR "sb orphan head is %d\n",
382 le32_to_cpu(sbi->s_es->s_last_orphan));
383
384 printk(KERN_ERR "sb_info orphan list:\n");
385 list_for_each(l, &sbi->s_orphan) {
386 struct inode *inode = orphan_list_entry(l);
387 printk(KERN_ERR " "
388 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
389 inode->i_sb->s_id, inode->i_ino, inode,
390 inode->i_mode, inode->i_nlink,
391 NEXT_ORPHAN(inode));
392 }
393 }
394
395 static void ext3_put_super (struct super_block * sb)
396 {
397 struct ext3_sb_info *sbi = EXT3_SB(sb);
398 struct ext3_super_block *es = sbi->s_es;
399 int i, err;
400
401 lock_kernel();
402
403 ext3_xattr_put_super(sb);
404 err = journal_destroy(sbi->s_journal);
405 sbi->s_journal = NULL;
406 if (err < 0)
407 ext3_abort(sb, __func__, "Couldn't clean up the journal");
408
409 if (!(sb->s_flags & MS_RDONLY)) {
410 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
411 es->s_state = cpu_to_le16(sbi->s_mount_state);
412 BUFFER_TRACE(sbi->s_sbh, "marking dirty");
413 mark_buffer_dirty(sbi->s_sbh);
414 ext3_commit_super(sb, es, 1);
415 }
416
417 for (i = 0; i < sbi->s_gdb_count; i++)
418 brelse(sbi->s_group_desc[i]);
419 kfree(sbi->s_group_desc);
420 percpu_counter_destroy(&sbi->s_freeblocks_counter);
421 percpu_counter_destroy(&sbi->s_freeinodes_counter);
422 percpu_counter_destroy(&sbi->s_dirs_counter);
423 brelse(sbi->s_sbh);
424 #ifdef CONFIG_QUOTA
425 for (i = 0; i < MAXQUOTAS; i++)
426 kfree(sbi->s_qf_names[i]);
427 #endif
428
429 /* Debugging code just in case the in-memory inode orphan list
430 * isn't empty. The on-disk one can be non-empty if we've
431 * detected an error and taken the fs readonly, but the
432 * in-memory list had better be clean by this point. */
433 if (!list_empty(&sbi->s_orphan))
434 dump_orphan_list(sb, sbi);
435 J_ASSERT(list_empty(&sbi->s_orphan));
436
437 invalidate_bdev(sb->s_bdev);
438 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
439 /*
440 * Invalidate the journal device's buffers. We don't want them
441 * floating about in memory - the physical journal device may
442 * hotswapped, and it breaks the `ro-after' testing code.
443 */
444 sync_blockdev(sbi->journal_bdev);
445 invalidate_bdev(sbi->journal_bdev);
446 ext3_blkdev_remove(sbi);
447 }
448 sb->s_fs_info = NULL;
449 kfree(sbi->s_blockgroup_lock);
450 kfree(sbi);
451
452 unlock_kernel();
453 }
454
455 static struct kmem_cache *ext3_inode_cachep;
456
457 /*
458 * Called inside transaction, so use GFP_NOFS
459 */
460 static struct inode *ext3_alloc_inode(struct super_block *sb)
461 {
462 struct ext3_inode_info *ei;
463
464 ei = kmem_cache_alloc(ext3_inode_cachep, GFP_NOFS);
465 if (!ei)
466 return NULL;
467 ei->i_block_alloc_info = NULL;
468 ei->vfs_inode.i_version = 1;
469 return &ei->vfs_inode;
470 }
471
472 static void ext3_destroy_inode(struct inode *inode)
473 {
474 if (!list_empty(&(EXT3_I(inode)->i_orphan))) {
475 printk("EXT3 Inode %p: orphan list check failed!\n",
476 EXT3_I(inode));
477 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
478 EXT3_I(inode), sizeof(struct ext3_inode_info),
479 false);
480 dump_stack();
481 }
482 kmem_cache_free(ext3_inode_cachep, EXT3_I(inode));
483 }
484
485 static void init_once(void *foo)
486 {
487 struct ext3_inode_info *ei = (struct ext3_inode_info *) foo;
488
489 INIT_LIST_HEAD(&ei->i_orphan);
490 #ifdef CONFIG_EXT3_FS_XATTR
491 init_rwsem(&ei->xattr_sem);
492 #endif
493 mutex_init(&ei->truncate_mutex);
494 inode_init_once(&ei->vfs_inode);
495 }
496
497 static int init_inodecache(void)
498 {
499 ext3_inode_cachep = kmem_cache_create("ext3_inode_cache",
500 sizeof(struct ext3_inode_info),
501 0, (SLAB_RECLAIM_ACCOUNT|
502 SLAB_MEM_SPREAD),
503 init_once);
504 if (ext3_inode_cachep == NULL)
505 return -ENOMEM;
506 return 0;
507 }
508
509 static void destroy_inodecache(void)
510 {
511 kmem_cache_destroy(ext3_inode_cachep);
512 }
513
514 static void ext3_clear_inode(struct inode *inode)
515 {
516 struct ext3_block_alloc_info *rsv = EXT3_I(inode)->i_block_alloc_info;
517 ext3_discard_reservation(inode);
518 EXT3_I(inode)->i_block_alloc_info = NULL;
519 if (unlikely(rsv))
520 kfree(rsv);
521 }
522
523 static inline void ext3_show_quota_options(struct seq_file *seq, struct super_block *sb)
524 {
525 #if defined(CONFIG_QUOTA)
526 struct ext3_sb_info *sbi = EXT3_SB(sb);
527
528 if (sbi->s_jquota_fmt)
529 seq_printf(seq, ",jqfmt=%s",
530 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold": "vfsv0");
531
532 if (sbi->s_qf_names[USRQUOTA])
533 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
534
535 if (sbi->s_qf_names[GRPQUOTA])
536 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
537
538 if (sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA)
539 seq_puts(seq, ",usrquota");
540
541 if (sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA)
542 seq_puts(seq, ",grpquota");
543 #endif
544 }
545
546 /*
547 * Show an option if
548 * - it's set to a non-default value OR
549 * - if the per-sb default is different from the global default
550 */
551 static int ext3_show_options(struct seq_file *seq, struct vfsmount *vfs)
552 {
553 struct super_block *sb = vfs->mnt_sb;
554 struct ext3_sb_info *sbi = EXT3_SB(sb);
555 struct ext3_super_block *es = sbi->s_es;
556 unsigned long def_mount_opts;
557
558 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
559
560 if (sbi->s_sb_block != 1)
561 seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
562 if (test_opt(sb, MINIX_DF))
563 seq_puts(seq, ",minixdf");
564 if (test_opt(sb, GRPID))
565 seq_puts(seq, ",grpid");
566 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT3_DEFM_BSDGROUPS))
567 seq_puts(seq, ",nogrpid");
568 if (sbi->s_resuid != EXT3_DEF_RESUID ||
569 le16_to_cpu(es->s_def_resuid) != EXT3_DEF_RESUID) {
570 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
571 }
572 if (sbi->s_resgid != EXT3_DEF_RESGID ||
573 le16_to_cpu(es->s_def_resgid) != EXT3_DEF_RESGID) {
574 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
575 }
576 if (test_opt(sb, ERRORS_RO)) {
577 int def_errors = le16_to_cpu(es->s_errors);
578
579 if (def_errors == EXT3_ERRORS_PANIC ||
580 def_errors == EXT3_ERRORS_CONTINUE) {
581 seq_puts(seq, ",errors=remount-ro");
582 }
583 }
584 if (test_opt(sb, ERRORS_CONT))
585 seq_puts(seq, ",errors=continue");
586 if (test_opt(sb, ERRORS_PANIC))
587 seq_puts(seq, ",errors=panic");
588 if (test_opt(sb, NO_UID32))
589 seq_puts(seq, ",nouid32");
590 if (test_opt(sb, DEBUG))
591 seq_puts(seq, ",debug");
592 if (test_opt(sb, OLDALLOC))
593 seq_puts(seq, ",oldalloc");
594 #ifdef CONFIG_EXT3_FS_XATTR
595 if (test_opt(sb, XATTR_USER))
596 seq_puts(seq, ",user_xattr");
597 if (!test_opt(sb, XATTR_USER) &&
598 (def_mount_opts & EXT3_DEFM_XATTR_USER)) {
599 seq_puts(seq, ",nouser_xattr");
600 }
601 #endif
602 #ifdef CONFIG_EXT3_FS_POSIX_ACL
603 if (test_opt(sb, POSIX_ACL))
604 seq_puts(seq, ",acl");
605 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT3_DEFM_ACL))
606 seq_puts(seq, ",noacl");
607 #endif
608 if (!test_opt(sb, RESERVATION))
609 seq_puts(seq, ",noreservation");
610 if (sbi->s_commit_interval) {
611 seq_printf(seq, ",commit=%u",
612 (unsigned) (sbi->s_commit_interval / HZ));
613 }
614 if (test_opt(sb, BARRIER))
615 seq_puts(seq, ",barrier=1");
616 if (test_opt(sb, NOBH))
617 seq_puts(seq, ",nobh");
618
619 if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA)
620 seq_puts(seq, ",data=journal");
621 else if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA)
622 seq_puts(seq, ",data=ordered");
623 else if (test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_WRITEBACK_DATA)
624 seq_puts(seq, ",data=writeback");
625
626 if (test_opt(sb, DATA_ERR_ABORT))
627 seq_puts(seq, ",data_err=abort");
628
629 ext3_show_quota_options(seq, sb);
630
631 return 0;
632 }
633
634
635 static struct inode *ext3_nfs_get_inode(struct super_block *sb,
636 u64 ino, u32 generation)
637 {
638 struct inode *inode;
639
640 if (ino < EXT3_FIRST_INO(sb) && ino != EXT3_ROOT_INO)
641 return ERR_PTR(-ESTALE);
642 if (ino > le32_to_cpu(EXT3_SB(sb)->s_es->s_inodes_count))
643 return ERR_PTR(-ESTALE);
644
645 /* iget isn't really right if the inode is currently unallocated!!
646 *
647 * ext3_read_inode will return a bad_inode if the inode had been
648 * deleted, so we should be safe.
649 *
650 * Currently we don't know the generation for parent directory, so
651 * a generation of 0 means "accept any"
652 */
653 inode = ext3_iget(sb, ino);
654 if (IS_ERR(inode))
655 return ERR_CAST(inode);
656 if (generation && inode->i_generation != generation) {
657 iput(inode);
658 return ERR_PTR(-ESTALE);
659 }
660
661 return inode;
662 }
663
664 static struct dentry *ext3_fh_to_dentry(struct super_block *sb, struct fid *fid,
665 int fh_len, int fh_type)
666 {
667 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
668 ext3_nfs_get_inode);
669 }
670
671 static struct dentry *ext3_fh_to_parent(struct super_block *sb, struct fid *fid,
672 int fh_len, int fh_type)
673 {
674 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
675 ext3_nfs_get_inode);
676 }
677
678 /*
679 * Try to release metadata pages (indirect blocks, directories) which are
680 * mapped via the block device. Since these pages could have journal heads
681 * which would prevent try_to_free_buffers() from freeing them, we must use
682 * jbd layer's try_to_free_buffers() function to release them.
683 */
684 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
685 gfp_t wait)
686 {
687 journal_t *journal = EXT3_SB(sb)->s_journal;
688
689 WARN_ON(PageChecked(page));
690 if (!page_has_buffers(page))
691 return 0;
692 if (journal)
693 return journal_try_to_free_buffers(journal, page,
694 wait & ~__GFP_WAIT);
695 return try_to_free_buffers(page);
696 }
697
698 #ifdef CONFIG_QUOTA
699 #define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
700 #define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
701
702 static int ext3_write_dquot(struct dquot *dquot);
703 static int ext3_acquire_dquot(struct dquot *dquot);
704 static int ext3_release_dquot(struct dquot *dquot);
705 static int ext3_mark_dquot_dirty(struct dquot *dquot);
706 static int ext3_write_info(struct super_block *sb, int type);
707 static int ext3_quota_on(struct super_block *sb, int type, int format_id,
708 char *path, int remount);
709 static int ext3_quota_on_mount(struct super_block *sb, int type);
710 static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
711 size_t len, loff_t off);
712 static ssize_t ext3_quota_write(struct super_block *sb, int type,
713 const char *data, size_t len, loff_t off);
714
715 static struct dquot_operations ext3_quota_operations = {
716 .initialize = dquot_initialize,
717 .drop = dquot_drop,
718 .alloc_space = dquot_alloc_space,
719 .alloc_inode = dquot_alloc_inode,
720 .free_space = dquot_free_space,
721 .free_inode = dquot_free_inode,
722 .transfer = dquot_transfer,
723 .write_dquot = ext3_write_dquot,
724 .acquire_dquot = ext3_acquire_dquot,
725 .release_dquot = ext3_release_dquot,
726 .mark_dirty = ext3_mark_dquot_dirty,
727 .write_info = ext3_write_info,
728 .alloc_dquot = dquot_alloc,
729 .destroy_dquot = dquot_destroy,
730 };
731
732 static struct quotactl_ops ext3_qctl_operations = {
733 .quota_on = ext3_quota_on,
734 .quota_off = vfs_quota_off,
735 .quota_sync = vfs_quota_sync,
736 .get_info = vfs_get_dqinfo,
737 .set_info = vfs_set_dqinfo,
738 .get_dqblk = vfs_get_dqblk,
739 .set_dqblk = vfs_set_dqblk
740 };
741 #endif
742
743 static const struct super_operations ext3_sops = {
744 .alloc_inode = ext3_alloc_inode,
745 .destroy_inode = ext3_destroy_inode,
746 .write_inode = ext3_write_inode,
747 .dirty_inode = ext3_dirty_inode,
748 .delete_inode = ext3_delete_inode,
749 .put_super = ext3_put_super,
750 .sync_fs = ext3_sync_fs,
751 .freeze_fs = ext3_freeze,
752 .unfreeze_fs = ext3_unfreeze,
753 .statfs = ext3_statfs,
754 .remount_fs = ext3_remount,
755 .clear_inode = ext3_clear_inode,
756 .show_options = ext3_show_options,
757 #ifdef CONFIG_QUOTA
758 .quota_read = ext3_quota_read,
759 .quota_write = ext3_quota_write,
760 #endif
761 .bdev_try_to_free_page = bdev_try_to_free_page,
762 };
763
764 static const struct export_operations ext3_export_ops = {
765 .fh_to_dentry = ext3_fh_to_dentry,
766 .fh_to_parent = ext3_fh_to_parent,
767 .get_parent = ext3_get_parent,
768 };
769
770 enum {
771 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
772 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
773 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
774 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
775 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
776 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
777 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
778 Opt_data_err_abort, Opt_data_err_ignore,
779 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
780 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
781 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
782 Opt_grpquota
783 };
784
785 static const match_table_t tokens = {
786 {Opt_bsd_df, "bsddf"},
787 {Opt_minix_df, "minixdf"},
788 {Opt_grpid, "grpid"},
789 {Opt_grpid, "bsdgroups"},
790 {Opt_nogrpid, "nogrpid"},
791 {Opt_nogrpid, "sysvgroups"},
792 {Opt_resgid, "resgid=%u"},
793 {Opt_resuid, "resuid=%u"},
794 {Opt_sb, "sb=%u"},
795 {Opt_err_cont, "errors=continue"},
796 {Opt_err_panic, "errors=panic"},
797 {Opt_err_ro, "errors=remount-ro"},
798 {Opt_nouid32, "nouid32"},
799 {Opt_nocheck, "nocheck"},
800 {Opt_nocheck, "check=none"},
801 {Opt_debug, "debug"},
802 {Opt_oldalloc, "oldalloc"},
803 {Opt_orlov, "orlov"},
804 {Opt_user_xattr, "user_xattr"},
805 {Opt_nouser_xattr, "nouser_xattr"},
806 {Opt_acl, "acl"},
807 {Opt_noacl, "noacl"},
808 {Opt_reservation, "reservation"},
809 {Opt_noreservation, "noreservation"},
810 {Opt_noload, "noload"},
811 {Opt_nobh, "nobh"},
812 {Opt_bh, "bh"},
813 {Opt_commit, "commit=%u"},
814 {Opt_journal_update, "journal=update"},
815 {Opt_journal_inum, "journal=%u"},
816 {Opt_journal_dev, "journal_dev=%u"},
817 {Opt_abort, "abort"},
818 {Opt_data_journal, "data=journal"},
819 {Opt_data_ordered, "data=ordered"},
820 {Opt_data_writeback, "data=writeback"},
821 {Opt_data_err_abort, "data_err=abort"},
822 {Opt_data_err_ignore, "data_err=ignore"},
823 {Opt_offusrjquota, "usrjquota="},
824 {Opt_usrjquota, "usrjquota=%s"},
825 {Opt_offgrpjquota, "grpjquota="},
826 {Opt_grpjquota, "grpjquota=%s"},
827 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
828 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
829 {Opt_grpquota, "grpquota"},
830 {Opt_noquota, "noquota"},
831 {Opt_quota, "quota"},
832 {Opt_usrquota, "usrquota"},
833 {Opt_barrier, "barrier=%u"},
834 {Opt_resize, "resize"},
835 {Opt_err, NULL},
836 };
837
838 static ext3_fsblk_t get_sb_block(void **data)
839 {
840 ext3_fsblk_t sb_block;
841 char *options = (char *) *data;
842
843 if (!options || strncmp(options, "sb=", 3) != 0)
844 return 1; /* Default location */
845 options += 3;
846 /*todo: use simple_strtoll with >32bit ext3 */
847 sb_block = simple_strtoul(options, &options, 0);
848 if (*options && *options != ',') {
849 printk("EXT3-fs: Invalid sb specification: %s\n",
850 (char *) *data);
851 return 1;
852 }
853 if (*options == ',')
854 options++;
855 *data = (void *) options;
856 return sb_block;
857 }
858
859 static int parse_options (char *options, struct super_block *sb,
860 unsigned int *inum, unsigned long *journal_devnum,
861 ext3_fsblk_t *n_blocks_count, int is_remount)
862 {
863 struct ext3_sb_info *sbi = EXT3_SB(sb);
864 char * p;
865 substring_t args[MAX_OPT_ARGS];
866 int data_opt = 0;
867 int option;
868 #ifdef CONFIG_QUOTA
869 int qtype, qfmt;
870 char *qname;
871 #endif
872
873 if (!options)
874 return 1;
875
876 while ((p = strsep (&options, ",")) != NULL) {
877 int token;
878 if (!*p)
879 continue;
880
881 token = match_token(p, tokens, args);
882 switch (token) {
883 case Opt_bsd_df:
884 clear_opt (sbi->s_mount_opt, MINIX_DF);
885 break;
886 case Opt_minix_df:
887 set_opt (sbi->s_mount_opt, MINIX_DF);
888 break;
889 case Opt_grpid:
890 set_opt (sbi->s_mount_opt, GRPID);
891 break;
892 case Opt_nogrpid:
893 clear_opt (sbi->s_mount_opt, GRPID);
894 break;
895 case Opt_resuid:
896 if (match_int(&args[0], &option))
897 return 0;
898 sbi->s_resuid = option;
899 break;
900 case Opt_resgid:
901 if (match_int(&args[0], &option))
902 return 0;
903 sbi->s_resgid = option;
904 break;
905 case Opt_sb:
906 /* handled by get_sb_block() instead of here */
907 /* *sb_block = match_int(&args[0]); */
908 break;
909 case Opt_err_panic:
910 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
911 clear_opt (sbi->s_mount_opt, ERRORS_RO);
912 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
913 break;
914 case Opt_err_ro:
915 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
916 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
917 set_opt (sbi->s_mount_opt, ERRORS_RO);
918 break;
919 case Opt_err_cont:
920 clear_opt (sbi->s_mount_opt, ERRORS_RO);
921 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
922 set_opt (sbi->s_mount_opt, ERRORS_CONT);
923 break;
924 case Opt_nouid32:
925 set_opt (sbi->s_mount_opt, NO_UID32);
926 break;
927 case Opt_nocheck:
928 clear_opt (sbi->s_mount_opt, CHECK);
929 break;
930 case Opt_debug:
931 set_opt (sbi->s_mount_opt, DEBUG);
932 break;
933 case Opt_oldalloc:
934 set_opt (sbi->s_mount_opt, OLDALLOC);
935 break;
936 case Opt_orlov:
937 clear_opt (sbi->s_mount_opt, OLDALLOC);
938 break;
939 #ifdef CONFIG_EXT3_FS_XATTR
940 case Opt_user_xattr:
941 set_opt (sbi->s_mount_opt, XATTR_USER);
942 break;
943 case Opt_nouser_xattr:
944 clear_opt (sbi->s_mount_opt, XATTR_USER);
945 break;
946 #else
947 case Opt_user_xattr:
948 case Opt_nouser_xattr:
949 printk("EXT3 (no)user_xattr options not supported\n");
950 break;
951 #endif
952 #ifdef CONFIG_EXT3_FS_POSIX_ACL
953 case Opt_acl:
954 set_opt(sbi->s_mount_opt, POSIX_ACL);
955 break;
956 case Opt_noacl:
957 clear_opt(sbi->s_mount_opt, POSIX_ACL);
958 break;
959 #else
960 case Opt_acl:
961 case Opt_noacl:
962 printk("EXT3 (no)acl options not supported\n");
963 break;
964 #endif
965 case Opt_reservation:
966 set_opt(sbi->s_mount_opt, RESERVATION);
967 break;
968 case Opt_noreservation:
969 clear_opt(sbi->s_mount_opt, RESERVATION);
970 break;
971 case Opt_journal_update:
972 /* @@@ FIXME */
973 /* Eventually we will want to be able to create
974 a journal file here. For now, only allow the
975 user to specify an existing inode to be the
976 journal file. */
977 if (is_remount) {
978 printk(KERN_ERR "EXT3-fs: cannot specify "
979 "journal on remount\n");
980 return 0;
981 }
982 set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
983 break;
984 case Opt_journal_inum:
985 if (is_remount) {
986 printk(KERN_ERR "EXT3-fs: cannot specify "
987 "journal on remount\n");
988 return 0;
989 }
990 if (match_int(&args[0], &option))
991 return 0;
992 *inum = option;
993 break;
994 case Opt_journal_dev:
995 if (is_remount) {
996 printk(KERN_ERR "EXT3-fs: cannot specify "
997 "journal on remount\n");
998 return 0;
999 }
1000 if (match_int(&args[0], &option))
1001 return 0;
1002 *journal_devnum = option;
1003 break;
1004 case Opt_noload:
1005 set_opt (sbi->s_mount_opt, NOLOAD);
1006 break;
1007 case Opt_commit:
1008 if (match_int(&args[0], &option))
1009 return 0;
1010 if (option < 0)
1011 return 0;
1012 if (option == 0)
1013 option = JBD_DEFAULT_MAX_COMMIT_AGE;
1014 sbi->s_commit_interval = HZ * option;
1015 break;
1016 case Opt_data_journal:
1017 data_opt = EXT3_MOUNT_JOURNAL_DATA;
1018 goto datacheck;
1019 case Opt_data_ordered:
1020 data_opt = EXT3_MOUNT_ORDERED_DATA;
1021 goto datacheck;
1022 case Opt_data_writeback:
1023 data_opt = EXT3_MOUNT_WRITEBACK_DATA;
1024 datacheck:
1025 if (is_remount) {
1026 if ((sbi->s_mount_opt & EXT3_MOUNT_DATA_FLAGS)
1027 != data_opt) {
1028 printk(KERN_ERR
1029 "EXT3-fs: cannot change data "
1030 "mode on remount\n");
1031 return 0;
1032 }
1033 } else {
1034 sbi->s_mount_opt &= ~EXT3_MOUNT_DATA_FLAGS;
1035 sbi->s_mount_opt |= data_opt;
1036 }
1037 break;
1038 case Opt_data_err_abort:
1039 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1040 break;
1041 case Opt_data_err_ignore:
1042 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1043 break;
1044 #ifdef CONFIG_QUOTA
1045 case Opt_usrjquota:
1046 qtype = USRQUOTA;
1047 goto set_qf_name;
1048 case Opt_grpjquota:
1049 qtype = GRPQUOTA;
1050 set_qf_name:
1051 if (sb_any_quota_loaded(sb) &&
1052 !sbi->s_qf_names[qtype]) {
1053 printk(KERN_ERR
1054 "EXT3-fs: Cannot change journaled "
1055 "quota options when quota turned on.\n");
1056 return 0;
1057 }
1058 qname = match_strdup(&args[0]);
1059 if (!qname) {
1060 printk(KERN_ERR
1061 "EXT3-fs: not enough memory for "
1062 "storing quotafile name.\n");
1063 return 0;
1064 }
1065 if (sbi->s_qf_names[qtype] &&
1066 strcmp(sbi->s_qf_names[qtype], qname)) {
1067 printk(KERN_ERR
1068 "EXT3-fs: %s quota file already "
1069 "specified.\n", QTYPE2NAME(qtype));
1070 kfree(qname);
1071 return 0;
1072 }
1073 sbi->s_qf_names[qtype] = qname;
1074 if (strchr(sbi->s_qf_names[qtype], '/')) {
1075 printk(KERN_ERR
1076 "EXT3-fs: quotafile must be on "
1077 "filesystem root.\n");
1078 kfree(sbi->s_qf_names[qtype]);
1079 sbi->s_qf_names[qtype] = NULL;
1080 return 0;
1081 }
1082 set_opt(sbi->s_mount_opt, QUOTA);
1083 break;
1084 case Opt_offusrjquota:
1085 qtype = USRQUOTA;
1086 goto clear_qf_name;
1087 case Opt_offgrpjquota:
1088 qtype = GRPQUOTA;
1089 clear_qf_name:
1090 if (sb_any_quota_loaded(sb) &&
1091 sbi->s_qf_names[qtype]) {
1092 printk(KERN_ERR "EXT3-fs: Cannot change "
1093 "journaled quota options when "
1094 "quota turned on.\n");
1095 return 0;
1096 }
1097 /*
1098 * The space will be released later when all options
1099 * are confirmed to be correct
1100 */
1101 sbi->s_qf_names[qtype] = NULL;
1102 break;
1103 case Opt_jqfmt_vfsold:
1104 qfmt = QFMT_VFS_OLD;
1105 goto set_qf_format;
1106 case Opt_jqfmt_vfsv0:
1107 qfmt = QFMT_VFS_V0;
1108 set_qf_format:
1109 if (sb_any_quota_loaded(sb) &&
1110 sbi->s_jquota_fmt != qfmt) {
1111 printk(KERN_ERR "EXT3-fs: Cannot change "
1112 "journaled quota options when "
1113 "quota turned on.\n");
1114 return 0;
1115 }
1116 sbi->s_jquota_fmt = qfmt;
1117 break;
1118 case Opt_quota:
1119 case Opt_usrquota:
1120 set_opt(sbi->s_mount_opt, QUOTA);
1121 set_opt(sbi->s_mount_opt, USRQUOTA);
1122 break;
1123 case Opt_grpquota:
1124 set_opt(sbi->s_mount_opt, QUOTA);
1125 set_opt(sbi->s_mount_opt, GRPQUOTA);
1126 break;
1127 case Opt_noquota:
1128 if (sb_any_quota_loaded(sb)) {
1129 printk(KERN_ERR "EXT3-fs: Cannot change quota "
1130 "options when quota turned on.\n");
1131 return 0;
1132 }
1133 clear_opt(sbi->s_mount_opt, QUOTA);
1134 clear_opt(sbi->s_mount_opt, USRQUOTA);
1135 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1136 break;
1137 #else
1138 case Opt_quota:
1139 case Opt_usrquota:
1140 case Opt_grpquota:
1141 printk(KERN_ERR
1142 "EXT3-fs: quota options not supported.\n");
1143 break;
1144 case Opt_usrjquota:
1145 case Opt_grpjquota:
1146 case Opt_offusrjquota:
1147 case Opt_offgrpjquota:
1148 case Opt_jqfmt_vfsold:
1149 case Opt_jqfmt_vfsv0:
1150 printk(KERN_ERR
1151 "EXT3-fs: journaled quota options not "
1152 "supported.\n");
1153 break;
1154 case Opt_noquota:
1155 break;
1156 #endif
1157 case Opt_abort:
1158 set_opt(sbi->s_mount_opt, ABORT);
1159 break;
1160 case Opt_barrier:
1161 if (match_int(&args[0], &option))
1162 return 0;
1163 if (option)
1164 set_opt(sbi->s_mount_opt, BARRIER);
1165 else
1166 clear_opt(sbi->s_mount_opt, BARRIER);
1167 break;
1168 case Opt_ignore:
1169 break;
1170 case Opt_resize:
1171 if (!is_remount) {
1172 printk("EXT3-fs: resize option only available "
1173 "for remount\n");
1174 return 0;
1175 }
1176 if (match_int(&args[0], &option) != 0)
1177 return 0;
1178 *n_blocks_count = option;
1179 break;
1180 case Opt_nobh:
1181 set_opt(sbi->s_mount_opt, NOBH);
1182 break;
1183 case Opt_bh:
1184 clear_opt(sbi->s_mount_opt, NOBH);
1185 break;
1186 default:
1187 printk (KERN_ERR
1188 "EXT3-fs: Unrecognized mount option \"%s\" "
1189 "or missing value\n", p);
1190 return 0;
1191 }
1192 }
1193 #ifdef CONFIG_QUOTA
1194 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1195 if ((sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA) &&
1196 sbi->s_qf_names[USRQUOTA])
1197 clear_opt(sbi->s_mount_opt, USRQUOTA);
1198
1199 if ((sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA) &&
1200 sbi->s_qf_names[GRPQUOTA])
1201 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1202
1203 if ((sbi->s_qf_names[USRQUOTA] &&
1204 (sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA)) ||
1205 (sbi->s_qf_names[GRPQUOTA] &&
1206 (sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA))) {
1207 printk(KERN_ERR "EXT3-fs: old and new quota "
1208 "format mixing.\n");
1209 return 0;
1210 }
1211
1212 if (!sbi->s_jquota_fmt) {
1213 printk(KERN_ERR "EXT3-fs: journaled quota format "
1214 "not specified.\n");
1215 return 0;
1216 }
1217 } else {
1218 if (sbi->s_jquota_fmt) {
1219 printk(KERN_ERR "EXT3-fs: journaled quota format "
1220 "specified with no journaling "
1221 "enabled.\n");
1222 return 0;
1223 }
1224 }
1225 #endif
1226 return 1;
1227 }
1228
1229 static int ext3_setup_super(struct super_block *sb, struct ext3_super_block *es,
1230 int read_only)
1231 {
1232 struct ext3_sb_info *sbi = EXT3_SB(sb);
1233 int res = 0;
1234
1235 if (le32_to_cpu(es->s_rev_level) > EXT3_MAX_SUPP_REV) {
1236 printk (KERN_ERR "EXT3-fs warning: revision level too high, "
1237 "forcing read-only mode\n");
1238 res = MS_RDONLY;
1239 }
1240 if (read_only)
1241 return res;
1242 if (!(sbi->s_mount_state & EXT3_VALID_FS))
1243 printk (KERN_WARNING "EXT3-fs warning: mounting unchecked fs, "
1244 "running e2fsck is recommended\n");
1245 else if ((sbi->s_mount_state & EXT3_ERROR_FS))
1246 printk (KERN_WARNING
1247 "EXT3-fs warning: mounting fs with errors, "
1248 "running e2fsck is recommended\n");
1249 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1250 le16_to_cpu(es->s_mnt_count) >=
1251 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1252 printk (KERN_WARNING
1253 "EXT3-fs warning: maximal mount count reached, "
1254 "running e2fsck is recommended\n");
1255 else if (le32_to_cpu(es->s_checkinterval) &&
1256 (le32_to_cpu(es->s_lastcheck) +
1257 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1258 printk (KERN_WARNING
1259 "EXT3-fs warning: checktime reached, "
1260 "running e2fsck is recommended\n");
1261 #if 0
1262 /* @@@ We _will_ want to clear the valid bit if we find
1263 inconsistencies, to force a fsck at reboot. But for
1264 a plain journaled filesystem we can keep it set as
1265 valid forever! :) */
1266 es->s_state &= cpu_to_le16(~EXT3_VALID_FS);
1267 #endif
1268 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1269 es->s_max_mnt_count = cpu_to_le16(EXT3_DFL_MAX_MNT_COUNT);
1270 le16_add_cpu(&es->s_mnt_count, 1);
1271 es->s_mtime = cpu_to_le32(get_seconds());
1272 ext3_update_dynamic_rev(sb);
1273 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
1274
1275 ext3_commit_super(sb, es, 1);
1276 if (test_opt(sb, DEBUG))
1277 printk(KERN_INFO "[EXT3 FS bs=%lu, gc=%lu, "
1278 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1279 sb->s_blocksize,
1280 sbi->s_groups_count,
1281 EXT3_BLOCKS_PER_GROUP(sb),
1282 EXT3_INODES_PER_GROUP(sb),
1283 sbi->s_mount_opt);
1284
1285 printk(KERN_INFO "EXT3 FS on %s, ", sb->s_id);
1286 if (EXT3_SB(sb)->s_journal->j_inode == NULL) {
1287 char b[BDEVNAME_SIZE];
1288
1289 printk("external journal on %s\n",
1290 bdevname(EXT3_SB(sb)->s_journal->j_dev, b));
1291 } else {
1292 printk("internal journal\n");
1293 }
1294 return res;
1295 }
1296
1297 /* Called at mount-time, super-block is locked */
1298 static int ext3_check_descriptors(struct super_block *sb)
1299 {
1300 struct ext3_sb_info *sbi = EXT3_SB(sb);
1301 int i;
1302
1303 ext3_debug ("Checking group descriptors");
1304
1305 for (i = 0; i < sbi->s_groups_count; i++) {
1306 struct ext3_group_desc *gdp = ext3_get_group_desc(sb, i, NULL);
1307 ext3_fsblk_t first_block = ext3_group_first_block_no(sb, i);
1308 ext3_fsblk_t last_block;
1309
1310 if (i == sbi->s_groups_count - 1)
1311 last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
1312 else
1313 last_block = first_block +
1314 (EXT3_BLOCKS_PER_GROUP(sb) - 1);
1315
1316 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
1317 le32_to_cpu(gdp->bg_block_bitmap) > last_block)
1318 {
1319 ext3_error (sb, "ext3_check_descriptors",
1320 "Block bitmap for group %d"
1321 " not in group (block %lu)!",
1322 i, (unsigned long)
1323 le32_to_cpu(gdp->bg_block_bitmap));
1324 return 0;
1325 }
1326 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
1327 le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
1328 {
1329 ext3_error (sb, "ext3_check_descriptors",
1330 "Inode bitmap for group %d"
1331 " not in group (block %lu)!",
1332 i, (unsigned long)
1333 le32_to_cpu(gdp->bg_inode_bitmap));
1334 return 0;
1335 }
1336 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
1337 le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
1338 last_block)
1339 {
1340 ext3_error (sb, "ext3_check_descriptors",
1341 "Inode table for group %d"
1342 " not in group (block %lu)!",
1343 i, (unsigned long)
1344 le32_to_cpu(gdp->bg_inode_table));
1345 return 0;
1346 }
1347 }
1348
1349 sbi->s_es->s_free_blocks_count=cpu_to_le32(ext3_count_free_blocks(sb));
1350 sbi->s_es->s_free_inodes_count=cpu_to_le32(ext3_count_free_inodes(sb));
1351 return 1;
1352 }
1353
1354
1355 /* ext3_orphan_cleanup() walks a singly-linked list of inodes (starting at
1356 * the superblock) which were deleted from all directories, but held open by
1357 * a process at the time of a crash. We walk the list and try to delete these
1358 * inodes at recovery time (only with a read-write filesystem).
1359 *
1360 * In order to keep the orphan inode chain consistent during traversal (in
1361 * case of crash during recovery), we link each inode into the superblock
1362 * orphan list_head and handle it the same way as an inode deletion during
1363 * normal operation (which journals the operations for us).
1364 *
1365 * We only do an iget() and an iput() on each inode, which is very safe if we
1366 * accidentally point at an in-use or already deleted inode. The worst that
1367 * can happen in this case is that we get a "bit already cleared" message from
1368 * ext3_free_inode(). The only reason we would point at a wrong inode is if
1369 * e2fsck was run on this filesystem, and it must have already done the orphan
1370 * inode cleanup for us, so we can safely abort without any further action.
1371 */
1372 static void ext3_orphan_cleanup (struct super_block * sb,
1373 struct ext3_super_block * es)
1374 {
1375 unsigned int s_flags = sb->s_flags;
1376 int nr_orphans = 0, nr_truncates = 0;
1377 #ifdef CONFIG_QUOTA
1378 int i;
1379 #endif
1380 if (!es->s_last_orphan) {
1381 jbd_debug(4, "no orphan inodes to clean up\n");
1382 return;
1383 }
1384
1385 if (bdev_read_only(sb->s_bdev)) {
1386 printk(KERN_ERR "EXT3-fs: write access "
1387 "unavailable, skipping orphan cleanup.\n");
1388 return;
1389 }
1390
1391 if (EXT3_SB(sb)->s_mount_state & EXT3_ERROR_FS) {
1392 if (es->s_last_orphan)
1393 jbd_debug(1, "Errors on filesystem, "
1394 "clearing orphan list.\n");
1395 es->s_last_orphan = 0;
1396 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1397 return;
1398 }
1399
1400 if (s_flags & MS_RDONLY) {
1401 printk(KERN_INFO "EXT3-fs: %s: orphan cleanup on readonly fs\n",
1402 sb->s_id);
1403 sb->s_flags &= ~MS_RDONLY;
1404 }
1405 #ifdef CONFIG_QUOTA
1406 /* Needed for iput() to work correctly and not trash data */
1407 sb->s_flags |= MS_ACTIVE;
1408 /* Turn on quotas so that they are updated correctly */
1409 for (i = 0; i < MAXQUOTAS; i++) {
1410 if (EXT3_SB(sb)->s_qf_names[i]) {
1411 int ret = ext3_quota_on_mount(sb, i);
1412 if (ret < 0)
1413 printk(KERN_ERR
1414 "EXT3-fs: Cannot turn on journaled "
1415 "quota: error %d\n", ret);
1416 }
1417 }
1418 #endif
1419
1420 while (es->s_last_orphan) {
1421 struct inode *inode;
1422
1423 inode = ext3_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1424 if (IS_ERR(inode)) {
1425 es->s_last_orphan = 0;
1426 break;
1427 }
1428
1429 list_add(&EXT3_I(inode)->i_orphan, &EXT3_SB(sb)->s_orphan);
1430 vfs_dq_init(inode);
1431 if (inode->i_nlink) {
1432 printk(KERN_DEBUG
1433 "%s: truncating inode %lu to %Ld bytes\n",
1434 __func__, inode->i_ino, inode->i_size);
1435 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1436 inode->i_ino, inode->i_size);
1437 ext3_truncate(inode);
1438 nr_truncates++;
1439 } else {
1440 printk(KERN_DEBUG
1441 "%s: deleting unreferenced inode %lu\n",
1442 __func__, inode->i_ino);
1443 jbd_debug(2, "deleting unreferenced inode %lu\n",
1444 inode->i_ino);
1445 nr_orphans++;
1446 }
1447 iput(inode); /* The delete magic happens here! */
1448 }
1449
1450 #define PLURAL(x) (x), ((x)==1) ? "" : "s"
1451
1452 if (nr_orphans)
1453 printk(KERN_INFO "EXT3-fs: %s: %d orphan inode%s deleted\n",
1454 sb->s_id, PLURAL(nr_orphans));
1455 if (nr_truncates)
1456 printk(KERN_INFO "EXT3-fs: %s: %d truncate%s cleaned up\n",
1457 sb->s_id, PLURAL(nr_truncates));
1458 #ifdef CONFIG_QUOTA
1459 /* Turn quotas off */
1460 for (i = 0; i < MAXQUOTAS; i++) {
1461 if (sb_dqopt(sb)->files[i])
1462 vfs_quota_off(sb, i, 0);
1463 }
1464 #endif
1465 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1466 }
1467
1468 /*
1469 * Maximal file size. There is a direct, and {,double-,triple-}indirect
1470 * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
1471 * We need to be 1 filesystem block less than the 2^32 sector limit.
1472 */
1473 static loff_t ext3_max_size(int bits)
1474 {
1475 loff_t res = EXT3_NDIR_BLOCKS;
1476 int meta_blocks;
1477 loff_t upper_limit;
1478
1479 /* This is calculated to be the largest file size for a
1480 * dense, file such that the total number of
1481 * sectors in the file, including data and all indirect blocks,
1482 * does not exceed 2^32 -1
1483 * __u32 i_blocks representing the total number of
1484 * 512 bytes blocks of the file
1485 */
1486 upper_limit = (1LL << 32) - 1;
1487
1488 /* total blocks in file system block size */
1489 upper_limit >>= (bits - 9);
1490
1491
1492 /* indirect blocks */
1493 meta_blocks = 1;
1494 /* double indirect blocks */
1495 meta_blocks += 1 + (1LL << (bits-2));
1496 /* tripple indirect blocks */
1497 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1498
1499 upper_limit -= meta_blocks;
1500 upper_limit <<= bits;
1501
1502 res += 1LL << (bits-2);
1503 res += 1LL << (2*(bits-2));
1504 res += 1LL << (3*(bits-2));
1505 res <<= bits;
1506 if (res > upper_limit)
1507 res = upper_limit;
1508
1509 if (res > MAX_LFS_FILESIZE)
1510 res = MAX_LFS_FILESIZE;
1511
1512 return res;
1513 }
1514
1515 static ext3_fsblk_t descriptor_loc(struct super_block *sb,
1516 ext3_fsblk_t logic_sb_block,
1517 int nr)
1518 {
1519 struct ext3_sb_info *sbi = EXT3_SB(sb);
1520 unsigned long bg, first_meta_bg;
1521 int has_super = 0;
1522
1523 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1524
1525 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_META_BG) ||
1526 nr < first_meta_bg)
1527 return (logic_sb_block + nr + 1);
1528 bg = sbi->s_desc_per_block * nr;
1529 if (ext3_bg_has_super(sb, bg))
1530 has_super = 1;
1531 return (has_super + ext3_group_first_block_no(sb, bg));
1532 }
1533
1534
1535 static int ext3_fill_super (struct super_block *sb, void *data, int silent)
1536 {
1537 struct buffer_head * bh;
1538 struct ext3_super_block *es = NULL;
1539 struct ext3_sb_info *sbi;
1540 ext3_fsblk_t block;
1541 ext3_fsblk_t sb_block = get_sb_block(&data);
1542 ext3_fsblk_t logic_sb_block;
1543 unsigned long offset = 0;
1544 unsigned int journal_inum = 0;
1545 unsigned long journal_devnum = 0;
1546 unsigned long def_mount_opts;
1547 struct inode *root;
1548 int blocksize;
1549 int hblock;
1550 int db_count;
1551 int i;
1552 int needs_recovery;
1553 int ret = -EINVAL;
1554 __le32 features;
1555 int err;
1556
1557 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1558 if (!sbi)
1559 return -ENOMEM;
1560
1561 sbi->s_blockgroup_lock =
1562 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
1563 if (!sbi->s_blockgroup_lock) {
1564 kfree(sbi);
1565 return -ENOMEM;
1566 }
1567 sb->s_fs_info = sbi;
1568 sbi->s_mount_opt = 0;
1569 sbi->s_resuid = EXT3_DEF_RESUID;
1570 sbi->s_resgid = EXT3_DEF_RESGID;
1571 sbi->s_sb_block = sb_block;
1572
1573 unlock_kernel();
1574
1575 blocksize = sb_min_blocksize(sb, EXT3_MIN_BLOCK_SIZE);
1576 if (!blocksize) {
1577 printk(KERN_ERR "EXT3-fs: unable to set blocksize\n");
1578 goto out_fail;
1579 }
1580
1581 /*
1582 * The ext3 superblock will not be buffer aligned for other than 1kB
1583 * block sizes. We need to calculate the offset from buffer start.
1584 */
1585 if (blocksize != EXT3_MIN_BLOCK_SIZE) {
1586 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1587 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1588 } else {
1589 logic_sb_block = sb_block;
1590 }
1591
1592 if (!(bh = sb_bread(sb, logic_sb_block))) {
1593 printk (KERN_ERR "EXT3-fs: unable to read superblock\n");
1594 goto out_fail;
1595 }
1596 /*
1597 * Note: s_es must be initialized as soon as possible because
1598 * some ext3 macro-instructions depend on its value
1599 */
1600 es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
1601 sbi->s_es = es;
1602 sb->s_magic = le16_to_cpu(es->s_magic);
1603 if (sb->s_magic != EXT3_SUPER_MAGIC)
1604 goto cantfind_ext3;
1605
1606 /* Set defaults before we parse the mount options */
1607 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1608 if (def_mount_opts & EXT3_DEFM_DEBUG)
1609 set_opt(sbi->s_mount_opt, DEBUG);
1610 if (def_mount_opts & EXT3_DEFM_BSDGROUPS)
1611 set_opt(sbi->s_mount_opt, GRPID);
1612 if (def_mount_opts & EXT3_DEFM_UID16)
1613 set_opt(sbi->s_mount_opt, NO_UID32);
1614 #ifdef CONFIG_EXT3_FS_XATTR
1615 if (def_mount_opts & EXT3_DEFM_XATTR_USER)
1616 set_opt(sbi->s_mount_opt, XATTR_USER);
1617 #endif
1618 #ifdef CONFIG_EXT3_FS_POSIX_ACL
1619 if (def_mount_opts & EXT3_DEFM_ACL)
1620 set_opt(sbi->s_mount_opt, POSIX_ACL);
1621 #endif
1622 if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_DATA)
1623 sbi->s_mount_opt |= EXT3_MOUNT_JOURNAL_DATA;
1624 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_ORDERED)
1625 sbi->s_mount_opt |= EXT3_MOUNT_ORDERED_DATA;
1626 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_WBACK)
1627 sbi->s_mount_opt |= EXT3_MOUNT_WRITEBACK_DATA;
1628
1629 if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_PANIC)
1630 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1631 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_CONTINUE)
1632 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1633 else
1634 set_opt(sbi->s_mount_opt, ERRORS_RO);
1635
1636 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1637 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1638
1639 set_opt(sbi->s_mount_opt, RESERVATION);
1640
1641 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1642 NULL, 0))
1643 goto failed_mount;
1644
1645 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1646 ((sbi->s_mount_opt & EXT3_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1647
1648 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV &&
1649 (EXT3_HAS_COMPAT_FEATURE(sb, ~0U) ||
1650 EXT3_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1651 EXT3_HAS_INCOMPAT_FEATURE(sb, ~0U)))
1652 printk(KERN_WARNING
1653 "EXT3-fs warning: feature flags set on rev 0 fs, "
1654 "running e2fsck is recommended\n");
1655 /*
1656 * Check feature flags regardless of the revision level, since we
1657 * previously didn't change the revision level when setting the flags,
1658 * so there is a chance incompat flags are set on a rev 0 filesystem.
1659 */
1660 features = EXT3_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP);
1661 if (features) {
1662 printk(KERN_ERR "EXT3-fs: %s: couldn't mount because of "
1663 "unsupported optional features (%x).\n",
1664 sb->s_id, le32_to_cpu(features));
1665 goto failed_mount;
1666 }
1667 features = EXT3_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP);
1668 if (!(sb->s_flags & MS_RDONLY) && features) {
1669 printk(KERN_ERR "EXT3-fs: %s: couldn't mount RDWR because of "
1670 "unsupported optional features (%x).\n",
1671 sb->s_id, le32_to_cpu(features));
1672 goto failed_mount;
1673 }
1674 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
1675
1676 if (blocksize < EXT3_MIN_BLOCK_SIZE ||
1677 blocksize > EXT3_MAX_BLOCK_SIZE) {
1678 printk(KERN_ERR
1679 "EXT3-fs: Unsupported filesystem blocksize %d on %s.\n",
1680 blocksize, sb->s_id);
1681 goto failed_mount;
1682 }
1683
1684 hblock = bdev_logical_block_size(sb->s_bdev);
1685 if (sb->s_blocksize != blocksize) {
1686 /*
1687 * Make sure the blocksize for the filesystem is larger
1688 * than the hardware sectorsize for the machine.
1689 */
1690 if (blocksize < hblock) {
1691 printk(KERN_ERR "EXT3-fs: blocksize %d too small for "
1692 "device blocksize %d.\n", blocksize, hblock);
1693 goto failed_mount;
1694 }
1695
1696 brelse (bh);
1697 if (!sb_set_blocksize(sb, blocksize)) {
1698 printk(KERN_ERR "EXT3-fs: bad blocksize %d.\n",
1699 blocksize);
1700 goto out_fail;
1701 }
1702 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1703 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1704 bh = sb_bread(sb, logic_sb_block);
1705 if (!bh) {
1706 printk(KERN_ERR
1707 "EXT3-fs: Can't read superblock on 2nd try.\n");
1708 goto failed_mount;
1709 }
1710 es = (struct ext3_super_block *)(((char *)bh->b_data) + offset);
1711 sbi->s_es = es;
1712 if (es->s_magic != cpu_to_le16(EXT3_SUPER_MAGIC)) {
1713 printk (KERN_ERR
1714 "EXT3-fs: Magic mismatch, very weird !\n");
1715 goto failed_mount;
1716 }
1717 }
1718
1719 sb->s_maxbytes = ext3_max_size(sb->s_blocksize_bits);
1720
1721 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV) {
1722 sbi->s_inode_size = EXT3_GOOD_OLD_INODE_SIZE;
1723 sbi->s_first_ino = EXT3_GOOD_OLD_FIRST_INO;
1724 } else {
1725 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1726 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1727 if ((sbi->s_inode_size < EXT3_GOOD_OLD_INODE_SIZE) ||
1728 (!is_power_of_2(sbi->s_inode_size)) ||
1729 (sbi->s_inode_size > blocksize)) {
1730 printk (KERN_ERR
1731 "EXT3-fs: unsupported inode size: %d\n",
1732 sbi->s_inode_size);
1733 goto failed_mount;
1734 }
1735 }
1736 sbi->s_frag_size = EXT3_MIN_FRAG_SIZE <<
1737 le32_to_cpu(es->s_log_frag_size);
1738 if (blocksize != sbi->s_frag_size) {
1739 printk(KERN_ERR
1740 "EXT3-fs: fragsize %lu != blocksize %u (unsupported)\n",
1741 sbi->s_frag_size, blocksize);
1742 goto failed_mount;
1743 }
1744 sbi->s_frags_per_block = 1;
1745 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1746 sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
1747 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
1748 if (EXT3_INODE_SIZE(sb) == 0 || EXT3_INODES_PER_GROUP(sb) == 0)
1749 goto cantfind_ext3;
1750 sbi->s_inodes_per_block = blocksize / EXT3_INODE_SIZE(sb);
1751 if (sbi->s_inodes_per_block == 0)
1752 goto cantfind_ext3;
1753 sbi->s_itb_per_group = sbi->s_inodes_per_group /
1754 sbi->s_inodes_per_block;
1755 sbi->s_desc_per_block = blocksize / sizeof(struct ext3_group_desc);
1756 sbi->s_sbh = bh;
1757 sbi->s_mount_state = le16_to_cpu(es->s_state);
1758 sbi->s_addr_per_block_bits = ilog2(EXT3_ADDR_PER_BLOCK(sb));
1759 sbi->s_desc_per_block_bits = ilog2(EXT3_DESC_PER_BLOCK(sb));
1760 for (i=0; i < 4; i++)
1761 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
1762 sbi->s_def_hash_version = es->s_def_hash_version;
1763 i = le32_to_cpu(es->s_flags);
1764 if (i & EXT2_FLAGS_UNSIGNED_HASH)
1765 sbi->s_hash_unsigned = 3;
1766 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
1767 #ifdef __CHAR_UNSIGNED__
1768 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
1769 sbi->s_hash_unsigned = 3;
1770 #else
1771 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
1772 #endif
1773 }
1774
1775 if (sbi->s_blocks_per_group > blocksize * 8) {
1776 printk (KERN_ERR
1777 "EXT3-fs: #blocks per group too big: %lu\n",
1778 sbi->s_blocks_per_group);
1779 goto failed_mount;
1780 }
1781 if (sbi->s_frags_per_group > blocksize * 8) {
1782 printk (KERN_ERR
1783 "EXT3-fs: #fragments per group too big: %lu\n",
1784 sbi->s_frags_per_group);
1785 goto failed_mount;
1786 }
1787 if (sbi->s_inodes_per_group > blocksize * 8) {
1788 printk (KERN_ERR
1789 "EXT3-fs: #inodes per group too big: %lu\n",
1790 sbi->s_inodes_per_group);
1791 goto failed_mount;
1792 }
1793
1794 if (le32_to_cpu(es->s_blocks_count) >
1795 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
1796 printk(KERN_ERR "EXT3-fs: filesystem on %s:"
1797 " too large to mount safely\n", sb->s_id);
1798 if (sizeof(sector_t) < 8)
1799 printk(KERN_WARNING "EXT3-fs: CONFIG_LBDAF not "
1800 "enabled\n");
1801 goto failed_mount;
1802 }
1803
1804 if (EXT3_BLOCKS_PER_GROUP(sb) == 0)
1805 goto cantfind_ext3;
1806 sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1807 le32_to_cpu(es->s_first_data_block) - 1)
1808 / EXT3_BLOCKS_PER_GROUP(sb)) + 1;
1809 db_count = (sbi->s_groups_count + EXT3_DESC_PER_BLOCK(sb) - 1) /
1810 EXT3_DESC_PER_BLOCK(sb);
1811 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
1812 GFP_KERNEL);
1813 if (sbi->s_group_desc == NULL) {
1814 printk (KERN_ERR "EXT3-fs: not enough memory\n");
1815 goto failed_mount;
1816 }
1817
1818 bgl_lock_init(sbi->s_blockgroup_lock);
1819
1820 for (i = 0; i < db_count; i++) {
1821 block = descriptor_loc(sb, logic_sb_block, i);
1822 sbi->s_group_desc[i] = sb_bread(sb, block);
1823 if (!sbi->s_group_desc[i]) {
1824 printk (KERN_ERR "EXT3-fs: "
1825 "can't read group descriptor %d\n", i);
1826 db_count = i;
1827 goto failed_mount2;
1828 }
1829 }
1830 if (!ext3_check_descriptors (sb)) {
1831 printk(KERN_ERR "EXT3-fs: group descriptors corrupted!\n");
1832 goto failed_mount2;
1833 }
1834 sbi->s_gdb_count = db_count;
1835 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1836 spin_lock_init(&sbi->s_next_gen_lock);
1837
1838 err = percpu_counter_init(&sbi->s_freeblocks_counter,
1839 ext3_count_free_blocks(sb));
1840 if (!err) {
1841 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1842 ext3_count_free_inodes(sb));
1843 }
1844 if (!err) {
1845 err = percpu_counter_init(&sbi->s_dirs_counter,
1846 ext3_count_dirs(sb));
1847 }
1848 if (err) {
1849 printk(KERN_ERR "EXT3-fs: insufficient memory\n");
1850 goto failed_mount3;
1851 }
1852
1853 /* per fileystem reservation list head & lock */
1854 spin_lock_init(&sbi->s_rsv_window_lock);
1855 sbi->s_rsv_window_root = RB_ROOT;
1856 /* Add a single, static dummy reservation to the start of the
1857 * reservation window list --- it gives us a placeholder for
1858 * append-at-start-of-list which makes the allocation logic
1859 * _much_ simpler. */
1860 sbi->s_rsv_window_head.rsv_start = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1861 sbi->s_rsv_window_head.rsv_end = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1862 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1863 sbi->s_rsv_window_head.rsv_goal_size = 0;
1864 ext3_rsv_window_add(sb, &sbi->s_rsv_window_head);
1865
1866 /*
1867 * set up enough so that it can read an inode
1868 */
1869 sb->s_op = &ext3_sops;
1870 sb->s_export_op = &ext3_export_ops;
1871 sb->s_xattr = ext3_xattr_handlers;
1872 #ifdef CONFIG_QUOTA
1873 sb->s_qcop = &ext3_qctl_operations;
1874 sb->dq_op = &ext3_quota_operations;
1875 #endif
1876 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
1877
1878 sb->s_root = NULL;
1879
1880 needs_recovery = (es->s_last_orphan != 0 ||
1881 EXT3_HAS_INCOMPAT_FEATURE(sb,
1882 EXT3_FEATURE_INCOMPAT_RECOVER));
1883
1884 /*
1885 * The first inode we look at is the journal inode. Don't try
1886 * root first: it may be modified in the journal!
1887 */
1888 if (!test_opt(sb, NOLOAD) &&
1889 EXT3_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL)) {
1890 if (ext3_load_journal(sb, es, journal_devnum))
1891 goto failed_mount3;
1892 } else if (journal_inum) {
1893 if (ext3_create_journal(sb, es, journal_inum))
1894 goto failed_mount3;
1895 } else {
1896 if (!silent)
1897 printk (KERN_ERR
1898 "ext3: No journal on filesystem on %s\n",
1899 sb->s_id);
1900 goto failed_mount3;
1901 }
1902
1903 /* We have now updated the journal if required, so we can
1904 * validate the data journaling mode. */
1905 switch (test_opt(sb, DATA_FLAGS)) {
1906 case 0:
1907 /* No mode set, assume a default based on the journal
1908 capabilities: ORDERED_DATA if the journal can
1909 cope, else JOURNAL_DATA */
1910 if (journal_check_available_features
1911 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE))
1912 set_opt(sbi->s_mount_opt, DEFAULT_DATA_MODE);
1913 else
1914 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
1915 break;
1916
1917 case EXT3_MOUNT_ORDERED_DATA:
1918 case EXT3_MOUNT_WRITEBACK_DATA:
1919 if (!journal_check_available_features
1920 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE)) {
1921 printk(KERN_ERR "EXT3-fs: Journal does not support "
1922 "requested data journaling mode\n");
1923 goto failed_mount4;
1924 }
1925 default:
1926 break;
1927 }
1928
1929 if (test_opt(sb, NOBH)) {
1930 if (!(test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_WRITEBACK_DATA)) {
1931 printk(KERN_WARNING "EXT3-fs: Ignoring nobh option - "
1932 "its supported only with writeback mode\n");
1933 clear_opt(sbi->s_mount_opt, NOBH);
1934 }
1935 }
1936 /*
1937 * The journal_load will have done any necessary log recovery,
1938 * so we can safely mount the rest of the filesystem now.
1939 */
1940
1941 root = ext3_iget(sb, EXT3_ROOT_INO);
1942 if (IS_ERR(root)) {
1943 printk(KERN_ERR "EXT3-fs: get root inode failed\n");
1944 ret = PTR_ERR(root);
1945 goto failed_mount4;
1946 }
1947 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1948 iput(root);
1949 printk(KERN_ERR "EXT3-fs: corrupt root inode, run e2fsck\n");
1950 goto failed_mount4;
1951 }
1952 sb->s_root = d_alloc_root(root);
1953 if (!sb->s_root) {
1954 printk(KERN_ERR "EXT3-fs: get root dentry failed\n");
1955 iput(root);
1956 ret = -ENOMEM;
1957 goto failed_mount4;
1958 }
1959
1960 ext3_setup_super (sb, es, sb->s_flags & MS_RDONLY);
1961 /*
1962 * akpm: core read_super() calls in here with the superblock locked.
1963 * That deadlocks, because orphan cleanup needs to lock the superblock
1964 * in numerous places. Here we just pop the lock - it's relatively
1965 * harmless, because we are now ready to accept write_super() requests,
1966 * and aviro says that's the only reason for hanging onto the
1967 * superblock lock.
1968 */
1969 EXT3_SB(sb)->s_mount_state |= EXT3_ORPHAN_FS;
1970 ext3_orphan_cleanup(sb, es);
1971 EXT3_SB(sb)->s_mount_state &= ~EXT3_ORPHAN_FS;
1972 if (needs_recovery)
1973 printk (KERN_INFO "EXT3-fs: recovery complete.\n");
1974 ext3_mark_recovery_complete(sb, es);
1975 printk (KERN_INFO "EXT3-fs: mounted filesystem with %s data mode.\n",
1976 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA ? "journal":
1977 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA ? "ordered":
1978 "writeback");
1979
1980 lock_kernel();
1981 return 0;
1982
1983 cantfind_ext3:
1984 if (!silent)
1985 printk(KERN_ERR "VFS: Can't find ext3 filesystem on dev %s.\n",
1986 sb->s_id);
1987 goto failed_mount;
1988
1989 failed_mount4:
1990 journal_destroy(sbi->s_journal);
1991 failed_mount3:
1992 percpu_counter_destroy(&sbi->s_freeblocks_counter);
1993 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1994 percpu_counter_destroy(&sbi->s_dirs_counter);
1995 failed_mount2:
1996 for (i = 0; i < db_count; i++)
1997 brelse(sbi->s_group_desc[i]);
1998 kfree(sbi->s_group_desc);
1999 failed_mount:
2000 #ifdef CONFIG_QUOTA
2001 for (i = 0; i < MAXQUOTAS; i++)
2002 kfree(sbi->s_qf_names[i]);
2003 #endif
2004 ext3_blkdev_remove(sbi);
2005 brelse(bh);
2006 out_fail:
2007 sb->s_fs_info = NULL;
2008 kfree(sbi->s_blockgroup_lock);
2009 kfree(sbi);
2010 lock_kernel();
2011 return ret;
2012 }
2013
2014 /*
2015 * Setup any per-fs journal parameters now. We'll do this both on
2016 * initial mount, once the journal has been initialised but before we've
2017 * done any recovery; and again on any subsequent remount.
2018 */
2019 static void ext3_init_journal_params(struct super_block *sb, journal_t *journal)
2020 {
2021 struct ext3_sb_info *sbi = EXT3_SB(sb);
2022
2023 if (sbi->s_commit_interval)
2024 journal->j_commit_interval = sbi->s_commit_interval;
2025 /* We could also set up an ext3-specific default for the commit
2026 * interval here, but for now we'll just fall back to the jbd
2027 * default. */
2028
2029 spin_lock(&journal->j_state_lock);
2030 if (test_opt(sb, BARRIER))
2031 journal->j_flags |= JFS_BARRIER;
2032 else
2033 journal->j_flags &= ~JFS_BARRIER;
2034 if (test_opt(sb, DATA_ERR_ABORT))
2035 journal->j_flags |= JFS_ABORT_ON_SYNCDATA_ERR;
2036 else
2037 journal->j_flags &= ~JFS_ABORT_ON_SYNCDATA_ERR;
2038 spin_unlock(&journal->j_state_lock);
2039 }
2040
2041 static journal_t *ext3_get_journal(struct super_block *sb,
2042 unsigned int journal_inum)
2043 {
2044 struct inode *journal_inode;
2045 journal_t *journal;
2046
2047 /* First, test for the existence of a valid inode on disk. Bad
2048 * things happen if we iget() an unused inode, as the subsequent
2049 * iput() will try to delete it. */
2050
2051 journal_inode = ext3_iget(sb, journal_inum);
2052 if (IS_ERR(journal_inode)) {
2053 printk(KERN_ERR "EXT3-fs: no journal found.\n");
2054 return NULL;
2055 }
2056 if (!journal_inode->i_nlink) {
2057 make_bad_inode(journal_inode);
2058 iput(journal_inode);
2059 printk(KERN_ERR "EXT3-fs: journal inode is deleted.\n");
2060 return NULL;
2061 }
2062
2063 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
2064 journal_inode, journal_inode->i_size);
2065 if (!S_ISREG(journal_inode->i_mode)) {
2066 printk(KERN_ERR "EXT3-fs: invalid journal inode.\n");
2067 iput(journal_inode);
2068 return NULL;
2069 }
2070
2071 journal = journal_init_inode(journal_inode);
2072 if (!journal) {
2073 printk(KERN_ERR "EXT3-fs: Could not load journal inode\n");
2074 iput(journal_inode);
2075 return NULL;
2076 }
2077 journal->j_private = sb;
2078 ext3_init_journal_params(sb, journal);
2079 return journal;
2080 }
2081
2082 static journal_t *ext3_get_dev_journal(struct super_block *sb,
2083 dev_t j_dev)
2084 {
2085 struct buffer_head * bh;
2086 journal_t *journal;
2087 ext3_fsblk_t start;
2088 ext3_fsblk_t len;
2089 int hblock, blocksize;
2090 ext3_fsblk_t sb_block;
2091 unsigned long offset;
2092 struct ext3_super_block * es;
2093 struct block_device *bdev;
2094
2095 bdev = ext3_blkdev_get(j_dev);
2096 if (bdev == NULL)
2097 return NULL;
2098
2099 if (bd_claim(bdev, sb)) {
2100 printk(KERN_ERR
2101 "EXT3: failed to claim external journal device.\n");
2102 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
2103 return NULL;
2104 }
2105
2106 blocksize = sb->s_blocksize;
2107 hblock = bdev_logical_block_size(bdev);
2108 if (blocksize < hblock) {
2109 printk(KERN_ERR
2110 "EXT3-fs: blocksize too small for journal device.\n");
2111 goto out_bdev;
2112 }
2113
2114 sb_block = EXT3_MIN_BLOCK_SIZE / blocksize;
2115 offset = EXT3_MIN_BLOCK_SIZE % blocksize;
2116 set_blocksize(bdev, blocksize);
2117 if (!(bh = __bread(bdev, sb_block, blocksize))) {
2118 printk(KERN_ERR "EXT3-fs: couldn't read superblock of "
2119 "external journal\n");
2120 goto out_bdev;
2121 }
2122
2123 es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
2124 if ((le16_to_cpu(es->s_magic) != EXT3_SUPER_MAGIC) ||
2125 !(le32_to_cpu(es->s_feature_incompat) &
2126 EXT3_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2127 printk(KERN_ERR "EXT3-fs: external journal has "
2128 "bad superblock\n");
2129 brelse(bh);
2130 goto out_bdev;
2131 }
2132
2133 if (memcmp(EXT3_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2134 printk(KERN_ERR "EXT3-fs: journal UUID does not match\n");
2135 brelse(bh);
2136 goto out_bdev;
2137 }
2138
2139 len = le32_to_cpu(es->s_blocks_count);
2140 start = sb_block + 1;
2141 brelse(bh); /* we're done with the superblock */
2142
2143 journal = journal_init_dev(bdev, sb->s_bdev,
2144 start, len, blocksize);
2145 if (!journal) {
2146 printk(KERN_ERR "EXT3-fs: failed to create device journal\n");
2147 goto out_bdev;
2148 }
2149 journal->j_private = sb;
2150 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2151 wait_on_buffer(journal->j_sb_buffer);
2152 if (!buffer_uptodate(journal->j_sb_buffer)) {
2153 printk(KERN_ERR "EXT3-fs: I/O error on journal device\n");
2154 goto out_journal;
2155 }
2156 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
2157 printk(KERN_ERR "EXT3-fs: External journal has more than one "
2158 "user (unsupported) - %d\n",
2159 be32_to_cpu(journal->j_superblock->s_nr_users));
2160 goto out_journal;
2161 }
2162 EXT3_SB(sb)->journal_bdev = bdev;
2163 ext3_init_journal_params(sb, journal);
2164 return journal;
2165 out_journal:
2166 journal_destroy(journal);
2167 out_bdev:
2168 ext3_blkdev_put(bdev);
2169 return NULL;
2170 }
2171
2172 static int ext3_load_journal(struct super_block *sb,
2173 struct ext3_super_block *es,
2174 unsigned long journal_devnum)
2175 {
2176 journal_t *journal;
2177 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2178 dev_t journal_dev;
2179 int err = 0;
2180 int really_read_only;
2181
2182 if (journal_devnum &&
2183 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2184 printk(KERN_INFO "EXT3-fs: external journal device major/minor "
2185 "numbers have changed\n");
2186 journal_dev = new_decode_dev(journal_devnum);
2187 } else
2188 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2189
2190 really_read_only = bdev_read_only(sb->s_bdev);
2191
2192 /*
2193 * Are we loading a blank journal or performing recovery after a
2194 * crash? For recovery, we need to check in advance whether we
2195 * can get read-write access to the device.
2196 */
2197
2198 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER)) {
2199 if (sb->s_flags & MS_RDONLY) {
2200 printk(KERN_INFO "EXT3-fs: INFO: recovery "
2201 "required on readonly filesystem.\n");
2202 if (really_read_only) {
2203 printk(KERN_ERR "EXT3-fs: write access "
2204 "unavailable, cannot proceed.\n");
2205 return -EROFS;
2206 }
2207 printk (KERN_INFO "EXT3-fs: write access will "
2208 "be enabled during recovery.\n");
2209 }
2210 }
2211
2212 if (journal_inum && journal_dev) {
2213 printk(KERN_ERR "EXT3-fs: filesystem has both journal "
2214 "and inode journals!\n");
2215 return -EINVAL;
2216 }
2217
2218 if (journal_inum) {
2219 if (!(journal = ext3_get_journal(sb, journal_inum)))
2220 return -EINVAL;
2221 } else {
2222 if (!(journal = ext3_get_dev_journal(sb, journal_dev)))
2223 return -EINVAL;
2224 }
2225
2226 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2227 err = journal_update_format(journal);
2228 if (err) {
2229 printk(KERN_ERR "EXT3-fs: error updating journal.\n");
2230 journal_destroy(journal);
2231 return err;
2232 }
2233 }
2234
2235 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER))
2236 err = journal_wipe(journal, !really_read_only);
2237 if (!err)
2238 err = journal_load(journal);
2239
2240 if (err) {
2241 printk(KERN_ERR "EXT3-fs: error loading journal.\n");
2242 journal_destroy(journal);
2243 return err;
2244 }
2245
2246 EXT3_SB(sb)->s_journal = journal;
2247 ext3_clear_journal_err(sb, es);
2248
2249 if (journal_devnum &&
2250 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2251 es->s_journal_dev = cpu_to_le32(journal_devnum);
2252
2253 /* Make sure we flush the recovery flag to disk. */
2254 ext3_commit_super(sb, es, 1);
2255 }
2256
2257 return 0;
2258 }
2259
2260 static int ext3_create_journal(struct super_block * sb,
2261 struct ext3_super_block * es,
2262 unsigned int journal_inum)
2263 {
2264 journal_t *journal;
2265 int err;
2266
2267 if (sb->s_flags & MS_RDONLY) {
2268 printk(KERN_ERR "EXT3-fs: readonly filesystem when trying to "
2269 "create journal.\n");
2270 return -EROFS;
2271 }
2272
2273 journal = ext3_get_journal(sb, journal_inum);
2274 if (!journal)
2275 return -EINVAL;
2276
2277 printk(KERN_INFO "EXT3-fs: creating new journal on inode %u\n",
2278 journal_inum);
2279
2280 err = journal_create(journal);
2281 if (err) {
2282 printk(KERN_ERR "EXT3-fs: error creating journal.\n");
2283 journal_destroy(journal);
2284 return -EIO;
2285 }
2286
2287 EXT3_SB(sb)->s_journal = journal;
2288
2289 ext3_update_dynamic_rev(sb);
2290 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2291 EXT3_SET_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL);
2292
2293 es->s_journal_inum = cpu_to_le32(journal_inum);
2294
2295 /* Make sure we flush the recovery flag to disk. */
2296 ext3_commit_super(sb, es, 1);
2297
2298 return 0;
2299 }
2300
2301 static int ext3_commit_super(struct super_block *sb,
2302 struct ext3_super_block *es,
2303 int sync)
2304 {
2305 struct buffer_head *sbh = EXT3_SB(sb)->s_sbh;
2306 int error = 0;
2307
2308 if (!sbh)
2309 return error;
2310 es->s_wtime = cpu_to_le32(get_seconds());
2311 es->s_free_blocks_count = cpu_to_le32(ext3_count_free_blocks(sb));
2312 es->s_free_inodes_count = cpu_to_le32(ext3_count_free_inodes(sb));
2313 BUFFER_TRACE(sbh, "marking dirty");
2314 mark_buffer_dirty(sbh);
2315 if (sync)
2316 error = sync_dirty_buffer(sbh);
2317 return error;
2318 }
2319
2320
2321 /*
2322 * Have we just finished recovery? If so, and if we are mounting (or
2323 * remounting) the filesystem readonly, then we will end up with a
2324 * consistent fs on disk. Record that fact.
2325 */
2326 static void ext3_mark_recovery_complete(struct super_block * sb,
2327 struct ext3_super_block * es)
2328 {
2329 journal_t *journal = EXT3_SB(sb)->s_journal;
2330
2331 journal_lock_updates(journal);
2332 if (journal_flush(journal) < 0)
2333 goto out;
2334
2335 lock_super(sb);
2336 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER) &&
2337 sb->s_flags & MS_RDONLY) {
2338 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2339 ext3_commit_super(sb, es, 1);
2340 }
2341 unlock_super(sb);
2342
2343 out:
2344 journal_unlock_updates(journal);
2345 }
2346
2347 /*
2348 * If we are mounting (or read-write remounting) a filesystem whose journal
2349 * has recorded an error from a previous lifetime, move that error to the
2350 * main filesystem now.
2351 */
2352 static void ext3_clear_journal_err(struct super_block * sb,
2353 struct ext3_super_block * es)
2354 {
2355 journal_t *journal;
2356 int j_errno;
2357 const char *errstr;
2358
2359 journal = EXT3_SB(sb)->s_journal;
2360
2361 /*
2362 * Now check for any error status which may have been recorded in the
2363 * journal by a prior ext3_error() or ext3_abort()
2364 */
2365
2366 j_errno = journal_errno(journal);
2367 if (j_errno) {
2368 char nbuf[16];
2369
2370 errstr = ext3_decode_error(sb, j_errno, nbuf);
2371 ext3_warning(sb, __func__, "Filesystem error recorded "
2372 "from previous mount: %s", errstr);
2373 ext3_warning(sb, __func__, "Marking fs in need of "
2374 "filesystem check.");
2375
2376 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
2377 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
2378 ext3_commit_super (sb, es, 1);
2379
2380 journal_clear_err(journal);
2381 }
2382 }
2383
2384 /*
2385 * Force the running and committing transactions to commit,
2386 * and wait on the commit.
2387 */
2388 int ext3_force_commit(struct super_block *sb)
2389 {
2390 journal_t *journal;
2391 int ret;
2392
2393 if (sb->s_flags & MS_RDONLY)
2394 return 0;
2395
2396 journal = EXT3_SB(sb)->s_journal;
2397 ret = ext3_journal_force_commit(journal);
2398 return ret;
2399 }
2400
2401 static int ext3_sync_fs(struct super_block *sb, int wait)
2402 {
2403 tid_t target;
2404
2405 if (journal_start_commit(EXT3_SB(sb)->s_journal, &target)) {
2406 if (wait)
2407 log_wait_commit(EXT3_SB(sb)->s_journal, target);
2408 }
2409 return 0;
2410 }
2411
2412 /*
2413 * LVM calls this function before a (read-only) snapshot is created. This
2414 * gives us a chance to flush the journal completely and mark the fs clean.
2415 */
2416 static int ext3_freeze(struct super_block *sb)
2417 {
2418 int error = 0;
2419 journal_t *journal;
2420
2421 if (!(sb->s_flags & MS_RDONLY)) {
2422 journal = EXT3_SB(sb)->s_journal;
2423
2424 /* Now we set up the journal barrier. */
2425 journal_lock_updates(journal);
2426
2427 /*
2428 * We don't want to clear needs_recovery flag when we failed
2429 * to flush the journal.
2430 */
2431 error = journal_flush(journal);
2432 if (error < 0)
2433 goto out;
2434
2435 /* Journal blocked and flushed, clear needs_recovery flag. */
2436 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2437 error = ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2438 if (error)
2439 goto out;
2440 }
2441 return 0;
2442
2443 out:
2444 journal_unlock_updates(journal);
2445 return error;
2446 }
2447
2448 /*
2449 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2450 * flag here, even though the filesystem is not technically dirty yet.
2451 */
2452 static int ext3_unfreeze(struct super_block *sb)
2453 {
2454 if (!(sb->s_flags & MS_RDONLY)) {
2455 lock_super(sb);
2456 /* Reser the needs_recovery flag before the fs is unlocked. */
2457 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2458 ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2459 unlock_super(sb);
2460 journal_unlock_updates(EXT3_SB(sb)->s_journal);
2461 }
2462 return 0;
2463 }
2464
2465 static int ext3_remount (struct super_block * sb, int * flags, char * data)
2466 {
2467 struct ext3_super_block * es;
2468 struct ext3_sb_info *sbi = EXT3_SB(sb);
2469 ext3_fsblk_t n_blocks_count = 0;
2470 unsigned long old_sb_flags;
2471 struct ext3_mount_options old_opts;
2472 int err;
2473 #ifdef CONFIG_QUOTA
2474 int i;
2475 #endif
2476
2477 lock_kernel();
2478
2479 /* Store the original options */
2480 lock_super(sb);
2481 old_sb_flags = sb->s_flags;
2482 old_opts.s_mount_opt = sbi->s_mount_opt;
2483 old_opts.s_resuid = sbi->s_resuid;
2484 old_opts.s_resgid = sbi->s_resgid;
2485 old_opts.s_commit_interval = sbi->s_commit_interval;
2486 #ifdef CONFIG_QUOTA
2487 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2488 for (i = 0; i < MAXQUOTAS; i++)
2489 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2490 #endif
2491
2492 /*
2493 * Allow the "check" option to be passed as a remount option.
2494 */
2495 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
2496 err = -EINVAL;
2497 goto restore_opts;
2498 }
2499
2500 if (sbi->s_mount_opt & EXT3_MOUNT_ABORT)
2501 ext3_abort(sb, __func__, "Abort forced by user");
2502
2503 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2504 ((sbi->s_mount_opt & EXT3_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2505
2506 es = sbi->s_es;
2507
2508 ext3_init_journal_params(sb, sbi->s_journal);
2509
2510 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
2511 n_blocks_count > le32_to_cpu(es->s_blocks_count)) {
2512 if (sbi->s_mount_opt & EXT3_MOUNT_ABORT) {
2513 err = -EROFS;
2514 goto restore_opts;
2515 }
2516
2517 if (*flags & MS_RDONLY) {
2518 /*
2519 * First of all, the unconditional stuff we have to do
2520 * to disable replay of the journal when we next remount
2521 */
2522 sb->s_flags |= MS_RDONLY;
2523
2524 /*
2525 * OK, test if we are remounting a valid rw partition
2526 * readonly, and if so set the rdonly flag and then
2527 * mark the partition as valid again.
2528 */
2529 if (!(es->s_state & cpu_to_le16(EXT3_VALID_FS)) &&
2530 (sbi->s_mount_state & EXT3_VALID_FS))
2531 es->s_state = cpu_to_le16(sbi->s_mount_state);
2532
2533 /*
2534 * We have to unlock super so that we can wait for
2535 * transactions.
2536 */
2537 unlock_super(sb);
2538 ext3_mark_recovery_complete(sb, es);
2539 lock_super(sb);
2540 } else {
2541 __le32 ret;
2542 if ((ret = EXT3_HAS_RO_COMPAT_FEATURE(sb,
2543 ~EXT3_FEATURE_RO_COMPAT_SUPP))) {
2544 printk(KERN_WARNING "EXT3-fs: %s: couldn't "
2545 "remount RDWR because of unsupported "
2546 "optional features (%x).\n",
2547 sb->s_id, le32_to_cpu(ret));
2548 err = -EROFS;
2549 goto restore_opts;
2550 }
2551
2552 /*
2553 * If we have an unprocessed orphan list hanging
2554 * around from a previously readonly bdev mount,
2555 * require a full umount/remount for now.
2556 */
2557 if (es->s_last_orphan) {
2558 printk(KERN_WARNING "EXT3-fs: %s: couldn't "
2559 "remount RDWR because of unprocessed "
2560 "orphan inode list. Please "
2561 "umount/remount instead.\n",
2562 sb->s_id);
2563 err = -EINVAL;
2564 goto restore_opts;
2565 }
2566
2567 /*
2568 * Mounting a RDONLY partition read-write, so reread
2569 * and store the current valid flag. (It may have
2570 * been changed by e2fsck since we originally mounted
2571 * the partition.)
2572 */
2573 ext3_clear_journal_err(sb, es);
2574 sbi->s_mount_state = le16_to_cpu(es->s_state);
2575 if ((err = ext3_group_extend(sb, es, n_blocks_count)))
2576 goto restore_opts;
2577 if (!ext3_setup_super (sb, es, 0))
2578 sb->s_flags &= ~MS_RDONLY;
2579 }
2580 }
2581 #ifdef CONFIG_QUOTA
2582 /* Release old quota file names */
2583 for (i = 0; i < MAXQUOTAS; i++)
2584 if (old_opts.s_qf_names[i] &&
2585 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2586 kfree(old_opts.s_qf_names[i]);
2587 #endif
2588 unlock_super(sb);
2589 unlock_kernel();
2590 return 0;
2591 restore_opts:
2592 sb->s_flags = old_sb_flags;
2593 sbi->s_mount_opt = old_opts.s_mount_opt;
2594 sbi->s_resuid = old_opts.s_resuid;
2595 sbi->s_resgid = old_opts.s_resgid;
2596 sbi->s_commit_interval = old_opts.s_commit_interval;
2597 #ifdef CONFIG_QUOTA
2598 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
2599 for (i = 0; i < MAXQUOTAS; i++) {
2600 if (sbi->s_qf_names[i] &&
2601 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2602 kfree(sbi->s_qf_names[i]);
2603 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
2604 }
2605 #endif
2606 unlock_super(sb);
2607 unlock_kernel();
2608 return err;
2609 }
2610
2611 static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf)
2612 {
2613 struct super_block *sb = dentry->d_sb;
2614 struct ext3_sb_info *sbi = EXT3_SB(sb);
2615 struct ext3_super_block *es = sbi->s_es;
2616 u64 fsid;
2617
2618 if (test_opt(sb, MINIX_DF)) {
2619 sbi->s_overhead_last = 0;
2620 } else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
2621 unsigned long ngroups = sbi->s_groups_count, i;
2622 ext3_fsblk_t overhead = 0;
2623 smp_rmb();
2624
2625 /*
2626 * Compute the overhead (FS structures). This is constant
2627 * for a given filesystem unless the number of block groups
2628 * changes so we cache the previous value until it does.
2629 */
2630
2631 /*
2632 * All of the blocks before first_data_block are
2633 * overhead
2634 */
2635 overhead = le32_to_cpu(es->s_first_data_block);
2636
2637 /*
2638 * Add the overhead attributed to the superblock and
2639 * block group descriptors. If the sparse superblocks
2640 * feature is turned on, then not all groups have this.
2641 */
2642 for (i = 0; i < ngroups; i++) {
2643 overhead += ext3_bg_has_super(sb, i) +
2644 ext3_bg_num_gdb(sb, i);
2645 cond_resched();
2646 }
2647
2648 /*
2649 * Every block group has an inode bitmap, a block
2650 * bitmap, and an inode table.
2651 */
2652 overhead += ngroups * (2 + sbi->s_itb_per_group);
2653 sbi->s_overhead_last = overhead;
2654 smp_wmb();
2655 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
2656 }
2657
2658 buf->f_type = EXT3_SUPER_MAGIC;
2659 buf->f_bsize = sb->s_blocksize;
2660 buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
2661 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter);
2662 es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
2663 buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
2664 if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
2665 buf->f_bavail = 0;
2666 buf->f_files = le32_to_cpu(es->s_inodes_count);
2667 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
2668 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
2669 buf->f_namelen = EXT3_NAME_LEN;
2670 fsid = le64_to_cpup((void *)es->s_uuid) ^
2671 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
2672 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
2673 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
2674 return 0;
2675 }
2676
2677 /* Helper function for writing quotas on sync - we need to start transaction before quota file
2678 * is locked for write. Otherwise the are possible deadlocks:
2679 * Process 1 Process 2
2680 * ext3_create() quota_sync()
2681 * journal_start() write_dquot()
2682 * vfs_dq_init() down(dqio_mutex)
2683 * down(dqio_mutex) journal_start()
2684 *
2685 */
2686
2687 #ifdef CONFIG_QUOTA
2688
2689 static inline struct inode *dquot_to_inode(struct dquot *dquot)
2690 {
2691 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
2692 }
2693
2694 static int ext3_write_dquot(struct dquot *dquot)
2695 {
2696 int ret, err;
2697 handle_t *handle;
2698 struct inode *inode;
2699
2700 inode = dquot_to_inode(dquot);
2701 handle = ext3_journal_start(inode,
2702 EXT3_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
2703 if (IS_ERR(handle))
2704 return PTR_ERR(handle);
2705 ret = dquot_commit(dquot);
2706 err = ext3_journal_stop(handle);
2707 if (!ret)
2708 ret = err;
2709 return ret;
2710 }
2711
2712 static int ext3_acquire_dquot(struct dquot *dquot)
2713 {
2714 int ret, err;
2715 handle_t *handle;
2716
2717 handle = ext3_journal_start(dquot_to_inode(dquot),
2718 EXT3_QUOTA_INIT_BLOCKS(dquot->dq_sb));
2719 if (IS_ERR(handle))
2720 return PTR_ERR(handle);
2721 ret = dquot_acquire(dquot);
2722 err = ext3_journal_stop(handle);
2723 if (!ret)
2724 ret = err;
2725 return ret;
2726 }
2727
2728 static int ext3_release_dquot(struct dquot *dquot)
2729 {
2730 int ret, err;
2731 handle_t *handle;
2732
2733 handle = ext3_journal_start(dquot_to_inode(dquot),
2734 EXT3_QUOTA_DEL_BLOCKS(dquot->dq_sb));
2735 if (IS_ERR(handle)) {
2736 /* Release dquot anyway to avoid endless cycle in dqput() */
2737 dquot_release(dquot);
2738 return PTR_ERR(handle);
2739 }
2740 ret = dquot_release(dquot);
2741 err = ext3_journal_stop(handle);
2742 if (!ret)
2743 ret = err;
2744 return ret;
2745 }
2746
2747 static int ext3_mark_dquot_dirty(struct dquot *dquot)
2748 {
2749 /* Are we journaling quotas? */
2750 if (EXT3_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
2751 EXT3_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
2752 dquot_mark_dquot_dirty(dquot);
2753 return ext3_write_dquot(dquot);
2754 } else {
2755 return dquot_mark_dquot_dirty(dquot);
2756 }
2757 }
2758
2759 static int ext3_write_info(struct super_block *sb, int type)
2760 {
2761 int ret, err;
2762 handle_t *handle;
2763
2764 /* Data block + inode block */
2765 handle = ext3_journal_start(sb->s_root->d_inode, 2);
2766 if (IS_ERR(handle))
2767 return PTR_ERR(handle);
2768 ret = dquot_commit_info(sb, type);
2769 err = ext3_journal_stop(handle);
2770 if (!ret)
2771 ret = err;
2772 return ret;
2773 }
2774
2775 /*
2776 * Turn on quotas during mount time - we need to find
2777 * the quota file and such...
2778 */
2779 static int ext3_quota_on_mount(struct super_block *sb, int type)
2780 {
2781 return vfs_quota_on_mount(sb, EXT3_SB(sb)->s_qf_names[type],
2782 EXT3_SB(sb)->s_jquota_fmt, type);
2783 }
2784
2785 /*
2786 * Standard function to be called on quota_on
2787 */
2788 static int ext3_quota_on(struct super_block *sb, int type, int format_id,
2789 char *name, int remount)
2790 {
2791 int err;
2792 struct path path;
2793
2794 if (!test_opt(sb, QUOTA))
2795 return -EINVAL;
2796 /* When remounting, no checks are needed and in fact, name is NULL */
2797 if (remount)
2798 return vfs_quota_on(sb, type, format_id, name, remount);
2799
2800 err = kern_path(name, LOOKUP_FOLLOW, &path);
2801 if (err)
2802 return err;
2803
2804 /* Quotafile not on the same filesystem? */
2805 if (path.mnt->mnt_sb != sb) {
2806 path_put(&path);
2807 return -EXDEV;
2808 }
2809 /* Journaling quota? */
2810 if (EXT3_SB(sb)->s_qf_names[type]) {
2811 /* Quotafile not of fs root? */
2812 if (path.dentry->d_parent != sb->s_root)
2813 printk(KERN_WARNING
2814 "EXT3-fs: Quota file not on filesystem root. "
2815 "Journaled quota will not work.\n");
2816 }
2817
2818 /*
2819 * When we journal data on quota file, we have to flush journal to see
2820 * all updates to the file when we bypass pagecache...
2821 */
2822 if (ext3_should_journal_data(path.dentry->d_inode)) {
2823 /*
2824 * We don't need to lock updates but journal_flush() could
2825 * otherwise be livelocked...
2826 */
2827 journal_lock_updates(EXT3_SB(sb)->s_journal);
2828 err = journal_flush(EXT3_SB(sb)->s_journal);
2829 journal_unlock_updates(EXT3_SB(sb)->s_journal);
2830 if (err) {
2831 path_put(&path);
2832 return err;
2833 }
2834 }
2835
2836 err = vfs_quota_on_path(sb, type, format_id, &path);
2837 path_put(&path);
2838 return err;
2839 }
2840
2841 /* Read data from quotafile - avoid pagecache and such because we cannot afford
2842 * acquiring the locks... As quota files are never truncated and quota code
2843 * itself serializes the operations (and noone else should touch the files)
2844 * we don't have to be afraid of races */
2845 static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
2846 size_t len, loff_t off)
2847 {
2848 struct inode *inode = sb_dqopt(sb)->files[type];
2849 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2850 int err = 0;
2851 int offset = off & (sb->s_blocksize - 1);
2852 int tocopy;
2853 size_t toread;
2854 struct buffer_head *bh;
2855 loff_t i_size = i_size_read(inode);
2856
2857 if (off > i_size)
2858 return 0;
2859 if (off+len > i_size)
2860 len = i_size-off;
2861 toread = len;
2862 while (toread > 0) {
2863 tocopy = sb->s_blocksize - offset < toread ?
2864 sb->s_blocksize - offset : toread;
2865 bh = ext3_bread(NULL, inode, blk, 0, &err);
2866 if (err)
2867 return err;
2868 if (!bh) /* A hole? */
2869 memset(data, 0, tocopy);
2870 else
2871 memcpy(data, bh->b_data+offset, tocopy);
2872 brelse(bh);
2873 offset = 0;
2874 toread -= tocopy;
2875 data += tocopy;
2876 blk++;
2877 }
2878 return len;
2879 }
2880
2881 /* Write to quotafile (we know the transaction is already started and has
2882 * enough credits) */
2883 static ssize_t ext3_quota_write(struct super_block *sb, int type,
2884 const char *data, size_t len, loff_t off)
2885 {
2886 struct inode *inode = sb_dqopt(sb)->files[type];
2887 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2888 int err = 0;
2889 int offset = off & (sb->s_blocksize - 1);
2890 int tocopy;
2891 int journal_quota = EXT3_SB(sb)->s_qf_names[type] != NULL;
2892 size_t towrite = len;
2893 struct buffer_head *bh;
2894 handle_t *handle = journal_current_handle();
2895
2896 if (!handle) {
2897 printk(KERN_WARNING "EXT3-fs: Quota write (off=%Lu, len=%Lu)"
2898 " cancelled because transaction is not started.\n",
2899 (unsigned long long)off, (unsigned long long)len);
2900 return -EIO;
2901 }
2902 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
2903 while (towrite > 0) {
2904 tocopy = sb->s_blocksize - offset < towrite ?
2905 sb->s_blocksize - offset : towrite;
2906 bh = ext3_bread(handle, inode, blk, 1, &err);
2907 if (!bh)
2908 goto out;
2909 if (journal_quota) {
2910 err = ext3_journal_get_write_access(handle, bh);
2911 if (err) {
2912 brelse(bh);
2913 goto out;
2914 }
2915 }
2916 lock_buffer(bh);
2917 memcpy(bh->b_data+offset, data, tocopy);
2918 flush_dcache_page(bh->b_page);
2919 unlock_buffer(bh);
2920 if (journal_quota)
2921 err = ext3_journal_dirty_metadata(handle, bh);
2922 else {
2923 /* Always do at least ordered writes for quotas */
2924 err = ext3_journal_dirty_data(handle, bh);
2925 mark_buffer_dirty(bh);
2926 }
2927 brelse(bh);
2928 if (err)
2929 goto out;
2930 offset = 0;
2931 towrite -= tocopy;
2932 data += tocopy;
2933 blk++;
2934 }
2935 out:
2936 if (len == towrite) {
2937 mutex_unlock(&inode->i_mutex);
2938 return err;
2939 }
2940 if (inode->i_size < off+len-towrite) {
2941 i_size_write(inode, off+len-towrite);
2942 EXT3_I(inode)->i_disksize = inode->i_size;
2943 }
2944 inode->i_version++;
2945 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
2946 ext3_mark_inode_dirty(handle, inode);
2947 mutex_unlock(&inode->i_mutex);
2948 return len - towrite;
2949 }
2950
2951 #endif
2952
2953 static int ext3_get_sb(struct file_system_type *fs_type,
2954 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
2955 {
2956 return get_sb_bdev(fs_type, flags, dev_name, data, ext3_fill_super, mnt);
2957 }
2958
2959 static struct file_system_type ext3_fs_type = {
2960 .owner = THIS_MODULE,
2961 .name = "ext3",
2962 .get_sb = ext3_get_sb,
2963 .kill_sb = kill_block_super,
2964 .fs_flags = FS_REQUIRES_DEV,
2965 };
2966
2967 static int __init init_ext3_fs(void)
2968 {
2969 int err = init_ext3_xattr();
2970 if (err)
2971 return err;
2972 err = init_inodecache();
2973 if (err)
2974 goto out1;
2975 err = register_filesystem(&ext3_fs_type);
2976 if (err)
2977 goto out;
2978 return 0;
2979 out:
2980 destroy_inodecache();
2981 out1:
2982 exit_ext3_xattr();
2983 return err;
2984 }
2985
2986 static void __exit exit_ext3_fs(void)
2987 {
2988 unregister_filesystem(&ext3_fs_type);
2989 destroy_inodecache();
2990 exit_ext3_xattr();
2991 }
2992
2993 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
2994 MODULE_DESCRIPTION("Second Extended Filesystem with journaling extensions");
2995 MODULE_LICENSE("GPL");
2996 module_init(init_ext3_fs)
2997 module_exit(exit_ext3_fs)
Linux 2.6 libata-dev (mirror)