?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
const: make struct super_block::dq_op const
[linux-2.6.git] / fs / ext3 / super.c
blobe7a4e11352d2137e968754aeda93679ac8b907e6
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 static char *data_mode_string(unsigned long mode)
547 {
548 switch (mode) {
549 case EXT3_MOUNT_JOURNAL_DATA:
550 return "journal";
551 case EXT3_MOUNT_ORDERED_DATA:
552 return "ordered";
553 case EXT3_MOUNT_WRITEBACK_DATA:
554 return "writeback";
555 }
556 return "unknown";
557 }
558
559 /*
560 * Show an option if
561 * - it's set to a non-default value OR
562 * - if the per-sb default is different from the global default
563 */
564 static int ext3_show_options(struct seq_file *seq, struct vfsmount *vfs)
565 {
566 struct super_block *sb = vfs->mnt_sb;
567 struct ext3_sb_info *sbi = EXT3_SB(sb);
568 struct ext3_super_block *es = sbi->s_es;
569 unsigned long def_mount_opts;
570
571 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
572
573 if (sbi->s_sb_block != 1)
574 seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
575 if (test_opt(sb, MINIX_DF))
576 seq_puts(seq, ",minixdf");
577 if (test_opt(sb, GRPID))
578 seq_puts(seq, ",grpid");
579 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT3_DEFM_BSDGROUPS))
580 seq_puts(seq, ",nogrpid");
581 if (sbi->s_resuid != EXT3_DEF_RESUID ||
582 le16_to_cpu(es->s_def_resuid) != EXT3_DEF_RESUID) {
583 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
584 }
585 if (sbi->s_resgid != EXT3_DEF_RESGID ||
586 le16_to_cpu(es->s_def_resgid) != EXT3_DEF_RESGID) {
587 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
588 }
589 if (test_opt(sb, ERRORS_RO)) {
590 int def_errors = le16_to_cpu(es->s_errors);
591
592 if (def_errors == EXT3_ERRORS_PANIC ||
593 def_errors == EXT3_ERRORS_CONTINUE) {
594 seq_puts(seq, ",errors=remount-ro");
595 }
596 }
597 if (test_opt(sb, ERRORS_CONT))
598 seq_puts(seq, ",errors=continue");
599 if (test_opt(sb, ERRORS_PANIC))
600 seq_puts(seq, ",errors=panic");
601 if (test_opt(sb, NO_UID32))
602 seq_puts(seq, ",nouid32");
603 if (test_opt(sb, DEBUG))
604 seq_puts(seq, ",debug");
605 if (test_opt(sb, OLDALLOC))
606 seq_puts(seq, ",oldalloc");
607 #ifdef CONFIG_EXT3_FS_XATTR
608 if (test_opt(sb, XATTR_USER))
609 seq_puts(seq, ",user_xattr");
610 if (!test_opt(sb, XATTR_USER) &&
611 (def_mount_opts & EXT3_DEFM_XATTR_USER)) {
612 seq_puts(seq, ",nouser_xattr");
613 }
614 #endif
615 #ifdef CONFIG_EXT3_FS_POSIX_ACL
616 if (test_opt(sb, POSIX_ACL))
617 seq_puts(seq, ",acl");
618 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT3_DEFM_ACL))
619 seq_puts(seq, ",noacl");
620 #endif
621 if (!test_opt(sb, RESERVATION))
622 seq_puts(seq, ",noreservation");
623 if (sbi->s_commit_interval) {
624 seq_printf(seq, ",commit=%u",
625 (unsigned) (sbi->s_commit_interval / HZ));
626 }
627 if (test_opt(sb, BARRIER))
628 seq_puts(seq, ",barrier=1");
629 if (test_opt(sb, NOBH))
630 seq_puts(seq, ",nobh");
631
632 seq_printf(seq, ",data=%s", data_mode_string(sbi->s_mount_opt &
633 EXT3_MOUNT_DATA_FLAGS));
634 if (test_opt(sb, DATA_ERR_ABORT))
635 seq_puts(seq, ",data_err=abort");
636
637 ext3_show_quota_options(seq, sb);
638
639 return 0;
640 }
641
642
643 static struct inode *ext3_nfs_get_inode(struct super_block *sb,
644 u64 ino, u32 generation)
645 {
646 struct inode *inode;
647
648 if (ino < EXT3_FIRST_INO(sb) && ino != EXT3_ROOT_INO)
649 return ERR_PTR(-ESTALE);
650 if (ino > le32_to_cpu(EXT3_SB(sb)->s_es->s_inodes_count))
651 return ERR_PTR(-ESTALE);
652
653 /* iget isn't really right if the inode is currently unallocated!!
654 *
655 * ext3_read_inode will return a bad_inode if the inode had been
656 * deleted, so we should be safe.
657 *
658 * Currently we don't know the generation for parent directory, so
659 * a generation of 0 means "accept any"
660 */
661 inode = ext3_iget(sb, ino);
662 if (IS_ERR(inode))
663 return ERR_CAST(inode);
664 if (generation && inode->i_generation != generation) {
665 iput(inode);
666 return ERR_PTR(-ESTALE);
667 }
668
669 return inode;
670 }
671
672 static struct dentry *ext3_fh_to_dentry(struct super_block *sb, struct fid *fid,
673 int fh_len, int fh_type)
674 {
675 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
676 ext3_nfs_get_inode);
677 }
678
679 static struct dentry *ext3_fh_to_parent(struct super_block *sb, struct fid *fid,
680 int fh_len, int fh_type)
681 {
682 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
683 ext3_nfs_get_inode);
684 }
685
686 /*
687 * Try to release metadata pages (indirect blocks, directories) which are
688 * mapped via the block device. Since these pages could have journal heads
689 * which would prevent try_to_free_buffers() from freeing them, we must use
690 * jbd layer's try_to_free_buffers() function to release them.
691 */
692 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
693 gfp_t wait)
694 {
695 journal_t *journal = EXT3_SB(sb)->s_journal;
696
697 WARN_ON(PageChecked(page));
698 if (!page_has_buffers(page))
699 return 0;
700 if (journal)
701 return journal_try_to_free_buffers(journal, page,
702 wait & ~__GFP_WAIT);
703 return try_to_free_buffers(page);
704 }
705
706 #ifdef CONFIG_QUOTA
707 #define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
708 #define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
709
710 static int ext3_write_dquot(struct dquot *dquot);
711 static int ext3_acquire_dquot(struct dquot *dquot);
712 static int ext3_release_dquot(struct dquot *dquot);
713 static int ext3_mark_dquot_dirty(struct dquot *dquot);
714 static int ext3_write_info(struct super_block *sb, int type);
715 static int ext3_quota_on(struct super_block *sb, int type, int format_id,
716 char *path, int remount);
717 static int ext3_quota_on_mount(struct super_block *sb, int type);
718 static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
719 size_t len, loff_t off);
720 static ssize_t ext3_quota_write(struct super_block *sb, int type,
721 const char *data, size_t len, loff_t off);
722
723 static const struct dquot_operations ext3_quota_operations = {
724 .initialize = dquot_initialize,
725 .drop = dquot_drop,
726 .alloc_space = dquot_alloc_space,
727 .alloc_inode = dquot_alloc_inode,
728 .free_space = dquot_free_space,
729 .free_inode = dquot_free_inode,
730 .transfer = dquot_transfer,
731 .write_dquot = ext3_write_dquot,
732 .acquire_dquot = ext3_acquire_dquot,
733 .release_dquot = ext3_release_dquot,
734 .mark_dirty = ext3_mark_dquot_dirty,
735 .write_info = ext3_write_info,
736 .alloc_dquot = dquot_alloc,
737 .destroy_dquot = dquot_destroy,
738 };
739
740 static struct quotactl_ops ext3_qctl_operations = {
741 .quota_on = ext3_quota_on,
742 .quota_off = vfs_quota_off,
743 .quota_sync = vfs_quota_sync,
744 .get_info = vfs_get_dqinfo,
745 .set_info = vfs_set_dqinfo,
746 .get_dqblk = vfs_get_dqblk,
747 .set_dqblk = vfs_set_dqblk
748 };
749 #endif
750
751 static const struct super_operations ext3_sops = {
752 .alloc_inode = ext3_alloc_inode,
753 .destroy_inode = ext3_destroy_inode,
754 .write_inode = ext3_write_inode,
755 .dirty_inode = ext3_dirty_inode,
756 .delete_inode = ext3_delete_inode,
757 .put_super = ext3_put_super,
758 .sync_fs = ext3_sync_fs,
759 .freeze_fs = ext3_freeze,
760 .unfreeze_fs = ext3_unfreeze,
761 .statfs = ext3_statfs,
762 .remount_fs = ext3_remount,
763 .clear_inode = ext3_clear_inode,
764 .show_options = ext3_show_options,
765 #ifdef CONFIG_QUOTA
766 .quota_read = ext3_quota_read,
767 .quota_write = ext3_quota_write,
768 #endif
769 .bdev_try_to_free_page = bdev_try_to_free_page,
770 };
771
772 static const struct export_operations ext3_export_ops = {
773 .fh_to_dentry = ext3_fh_to_dentry,
774 .fh_to_parent = ext3_fh_to_parent,
775 .get_parent = ext3_get_parent,
776 };
777
778 enum {
779 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
780 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
781 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
782 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
783 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
784 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
785 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
786 Opt_data_err_abort, Opt_data_err_ignore,
787 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
788 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
789 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
790 Opt_grpquota
791 };
792
793 static const match_table_t tokens = {
794 {Opt_bsd_df, "bsddf"},
795 {Opt_minix_df, "minixdf"},
796 {Opt_grpid, "grpid"},
797 {Opt_grpid, "bsdgroups"},
798 {Opt_nogrpid, "nogrpid"},
799 {Opt_nogrpid, "sysvgroups"},
800 {Opt_resgid, "resgid=%u"},
801 {Opt_resuid, "resuid=%u"},
802 {Opt_sb, "sb=%u"},
803 {Opt_err_cont, "errors=continue"},
804 {Opt_err_panic, "errors=panic"},
805 {Opt_err_ro, "errors=remount-ro"},
806 {Opt_nouid32, "nouid32"},
807 {Opt_nocheck, "nocheck"},
808 {Opt_nocheck, "check=none"},
809 {Opt_debug, "debug"},
810 {Opt_oldalloc, "oldalloc"},
811 {Opt_orlov, "orlov"},
812 {Opt_user_xattr, "user_xattr"},
813 {Opt_nouser_xattr, "nouser_xattr"},
814 {Opt_acl, "acl"},
815 {Opt_noacl, "noacl"},
816 {Opt_reservation, "reservation"},
817 {Opt_noreservation, "noreservation"},
818 {Opt_noload, "noload"},
819 {Opt_nobh, "nobh"},
820 {Opt_bh, "bh"},
821 {Opt_commit, "commit=%u"},
822 {Opt_journal_update, "journal=update"},
823 {Opt_journal_inum, "journal=%u"},
824 {Opt_journal_dev, "journal_dev=%u"},
825 {Opt_abort, "abort"},
826 {Opt_data_journal, "data=journal"},
827 {Opt_data_ordered, "data=ordered"},
828 {Opt_data_writeback, "data=writeback"},
829 {Opt_data_err_abort, "data_err=abort"},
830 {Opt_data_err_ignore, "data_err=ignore"},
831 {Opt_offusrjquota, "usrjquota="},
832 {Opt_usrjquota, "usrjquota=%s"},
833 {Opt_offgrpjquota, "grpjquota="},
834 {Opt_grpjquota, "grpjquota=%s"},
835 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
836 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
837 {Opt_grpquota, "grpquota"},
838 {Opt_noquota, "noquota"},
839 {Opt_quota, "quota"},
840 {Opt_usrquota, "usrquota"},
841 {Opt_barrier, "barrier=%u"},
842 {Opt_resize, "resize"},
843 {Opt_err, NULL},
844 };
845
846 static ext3_fsblk_t get_sb_block(void **data)
847 {
848 ext3_fsblk_t sb_block;
849 char *options = (char *) *data;
850
851 if (!options || strncmp(options, "sb=", 3) != 0)
852 return 1; /* Default location */
853 options += 3;
854 /*todo: use simple_strtoll with >32bit ext3 */
855 sb_block = simple_strtoul(options, &options, 0);
856 if (*options && *options != ',') {
857 printk("EXT3-fs: Invalid sb specification: %s\n",
858 (char *) *data);
859 return 1;
860 }
861 if (*options == ',')
862 options++;
863 *data = (void *) options;
864 return sb_block;
865 }
866
867 static int parse_options (char *options, struct super_block *sb,
868 unsigned int *inum, unsigned long *journal_devnum,
869 ext3_fsblk_t *n_blocks_count, int is_remount)
870 {
871 struct ext3_sb_info *sbi = EXT3_SB(sb);
872 char * p;
873 substring_t args[MAX_OPT_ARGS];
874 int data_opt = 0;
875 int option;
876 #ifdef CONFIG_QUOTA
877 int qtype, qfmt;
878 char *qname;
879 #endif
880
881 if (!options)
882 return 1;
883
884 while ((p = strsep (&options, ",")) != NULL) {
885 int token;
886 if (!*p)
887 continue;
888
889 token = match_token(p, tokens, args);
890 switch (token) {
891 case Opt_bsd_df:
892 clear_opt (sbi->s_mount_opt, MINIX_DF);
893 break;
894 case Opt_minix_df:
895 set_opt (sbi->s_mount_opt, MINIX_DF);
896 break;
897 case Opt_grpid:
898 set_opt (sbi->s_mount_opt, GRPID);
899 break;
900 case Opt_nogrpid:
901 clear_opt (sbi->s_mount_opt, GRPID);
902 break;
903 case Opt_resuid:
904 if (match_int(&args[0], &option))
905 return 0;
906 sbi->s_resuid = option;
907 break;
908 case Opt_resgid:
909 if (match_int(&args[0], &option))
910 return 0;
911 sbi->s_resgid = option;
912 break;
913 case Opt_sb:
914 /* handled by get_sb_block() instead of here */
915 /* *sb_block = match_int(&args[0]); */
916 break;
917 case Opt_err_panic:
918 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
919 clear_opt (sbi->s_mount_opt, ERRORS_RO);
920 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
921 break;
922 case Opt_err_ro:
923 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
924 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
925 set_opt (sbi->s_mount_opt, ERRORS_RO);
926 break;
927 case Opt_err_cont:
928 clear_opt (sbi->s_mount_opt, ERRORS_RO);
929 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
930 set_opt (sbi->s_mount_opt, ERRORS_CONT);
931 break;
932 case Opt_nouid32:
933 set_opt (sbi->s_mount_opt, NO_UID32);
934 break;
935 case Opt_nocheck:
936 clear_opt (sbi->s_mount_opt, CHECK);
937 break;
938 case Opt_debug:
939 set_opt (sbi->s_mount_opt, DEBUG);
940 break;
941 case Opt_oldalloc:
942 set_opt (sbi->s_mount_opt, OLDALLOC);
943 break;
944 case Opt_orlov:
945 clear_opt (sbi->s_mount_opt, OLDALLOC);
946 break;
947 #ifdef CONFIG_EXT3_FS_XATTR
948 case Opt_user_xattr:
949 set_opt (sbi->s_mount_opt, XATTR_USER);
950 break;
951 case Opt_nouser_xattr:
952 clear_opt (sbi->s_mount_opt, XATTR_USER);
953 break;
954 #else
955 case Opt_user_xattr:
956 case Opt_nouser_xattr:
957 printk("EXT3 (no)user_xattr options not supported\n");
958 break;
959 #endif
960 #ifdef CONFIG_EXT3_FS_POSIX_ACL
961 case Opt_acl:
962 set_opt(sbi->s_mount_opt, POSIX_ACL);
963 break;
964 case Opt_noacl:
965 clear_opt(sbi->s_mount_opt, POSIX_ACL);
966 break;
967 #else
968 case Opt_acl:
969 case Opt_noacl:
970 printk("EXT3 (no)acl options not supported\n");
971 break;
972 #endif
973 case Opt_reservation:
974 set_opt(sbi->s_mount_opt, RESERVATION);
975 break;
976 case Opt_noreservation:
977 clear_opt(sbi->s_mount_opt, RESERVATION);
978 break;
979 case Opt_journal_update:
980 /* @@@ FIXME */
981 /* Eventually we will want to be able to create
982 a journal file here. For now, only allow the
983 user to specify an existing inode to be the
984 journal file. */
985 if (is_remount) {
986 printk(KERN_ERR "EXT3-fs: cannot specify "
987 "journal on remount\n");
988 return 0;
989 }
990 set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
991 break;
992 case Opt_journal_inum:
993 if (is_remount) {
994 printk(KERN_ERR "EXT3-fs: cannot specify "
995 "journal on remount\n");
996 return 0;
997 }
998 if (match_int(&args[0], &option))
999 return 0;
1000 *inum = option;
1001 break;
1002 case Opt_journal_dev:
1003 if (is_remount) {
1004 printk(KERN_ERR "EXT3-fs: cannot specify "
1005 "journal on remount\n");
1006 return 0;
1007 }
1008 if (match_int(&args[0], &option))
1009 return 0;
1010 *journal_devnum = option;
1011 break;
1012 case Opt_noload:
1013 set_opt (sbi->s_mount_opt, NOLOAD);
1014 break;
1015 case Opt_commit:
1016 if (match_int(&args[0], &option))
1017 return 0;
1018 if (option < 0)
1019 return 0;
1020 if (option == 0)
1021 option = JBD_DEFAULT_MAX_COMMIT_AGE;
1022 sbi->s_commit_interval = HZ * option;
1023 break;
1024 case Opt_data_journal:
1025 data_opt = EXT3_MOUNT_JOURNAL_DATA;
1026 goto datacheck;
1027 case Opt_data_ordered:
1028 data_opt = EXT3_MOUNT_ORDERED_DATA;
1029 goto datacheck;
1030 case Opt_data_writeback:
1031 data_opt = EXT3_MOUNT_WRITEBACK_DATA;
1032 datacheck:
1033 if (is_remount) {
1034 if ((sbi->s_mount_opt & EXT3_MOUNT_DATA_FLAGS)
1035 == data_opt)
1036 break;
1037 printk(KERN_ERR
1038 "EXT3-fs (device %s): Cannot change "
1039 "data mode on remount. The filesystem "
1040 "is mounted in data=%s mode and you "
1041 "try to remount it in data=%s mode.\n",
1042 sb->s_id,
1043 data_mode_string(sbi->s_mount_opt &
1044 EXT3_MOUNT_DATA_FLAGS),
1045 data_mode_string(data_opt));
1046 return 0;
1047 } else {
1048 sbi->s_mount_opt &= ~EXT3_MOUNT_DATA_FLAGS;
1049 sbi->s_mount_opt |= data_opt;
1050 }
1051 break;
1052 case Opt_data_err_abort:
1053 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1054 break;
1055 case Opt_data_err_ignore:
1056 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1057 break;
1058 #ifdef CONFIG_QUOTA
1059 case Opt_usrjquota:
1060 qtype = USRQUOTA;
1061 goto set_qf_name;
1062 case Opt_grpjquota:
1063 qtype = GRPQUOTA;
1064 set_qf_name:
1065 if (sb_any_quota_loaded(sb) &&
1066 !sbi->s_qf_names[qtype]) {
1067 printk(KERN_ERR
1068 "EXT3-fs: Cannot change journaled "
1069 "quota options when quota turned on.\n");
1070 return 0;
1071 }
1072 qname = match_strdup(&args[0]);
1073 if (!qname) {
1074 printk(KERN_ERR
1075 "EXT3-fs: not enough memory for "
1076 "storing quotafile name.\n");
1077 return 0;
1078 }
1079 if (sbi->s_qf_names[qtype] &&
1080 strcmp(sbi->s_qf_names[qtype], qname)) {
1081 printk(KERN_ERR
1082 "EXT3-fs: %s quota file already "
1083 "specified.\n", QTYPE2NAME(qtype));
1084 kfree(qname);
1085 return 0;
1086 }
1087 sbi->s_qf_names[qtype] = qname;
1088 if (strchr(sbi->s_qf_names[qtype], '/')) {
1089 printk(KERN_ERR
1090 "EXT3-fs: quotafile must be on "
1091 "filesystem root.\n");
1092 kfree(sbi->s_qf_names[qtype]);
1093 sbi->s_qf_names[qtype] = NULL;
1094 return 0;
1095 }
1096 set_opt(sbi->s_mount_opt, QUOTA);
1097 break;
1098 case Opt_offusrjquota:
1099 qtype = USRQUOTA;
1100 goto clear_qf_name;
1101 case Opt_offgrpjquota:
1102 qtype = GRPQUOTA;
1103 clear_qf_name:
1104 if (sb_any_quota_loaded(sb) &&
1105 sbi->s_qf_names[qtype]) {
1106 printk(KERN_ERR "EXT3-fs: Cannot change "
1107 "journaled quota options when "
1108 "quota turned on.\n");
1109 return 0;
1110 }
1111 /*
1112 * The space will be released later when all options
1113 * are confirmed to be correct
1114 */
1115 sbi->s_qf_names[qtype] = NULL;
1116 break;
1117 case Opt_jqfmt_vfsold:
1118 qfmt = QFMT_VFS_OLD;
1119 goto set_qf_format;
1120 case Opt_jqfmt_vfsv0:
1121 qfmt = QFMT_VFS_V0;
1122 set_qf_format:
1123 if (sb_any_quota_loaded(sb) &&
1124 sbi->s_jquota_fmt != qfmt) {
1125 printk(KERN_ERR "EXT3-fs: Cannot change "
1126 "journaled quota options when "
1127 "quota turned on.\n");
1128 return 0;
1129 }
1130 sbi->s_jquota_fmt = qfmt;
1131 break;
1132 case Opt_quota:
1133 case Opt_usrquota:
1134 set_opt(sbi->s_mount_opt, QUOTA);
1135 set_opt(sbi->s_mount_opt, USRQUOTA);
1136 break;
1137 case Opt_grpquota:
1138 set_opt(sbi->s_mount_opt, QUOTA);
1139 set_opt(sbi->s_mount_opt, GRPQUOTA);
1140 break;
1141 case Opt_noquota:
1142 if (sb_any_quota_loaded(sb)) {
1143 printk(KERN_ERR "EXT3-fs: Cannot change quota "
1144 "options when quota turned on.\n");
1145 return 0;
1146 }
1147 clear_opt(sbi->s_mount_opt, QUOTA);
1148 clear_opt(sbi->s_mount_opt, USRQUOTA);
1149 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1150 break;
1151 #else
1152 case Opt_quota:
1153 case Opt_usrquota:
1154 case Opt_grpquota:
1155 printk(KERN_ERR
1156 "EXT3-fs: quota options not supported.\n");
1157 break;
1158 case Opt_usrjquota:
1159 case Opt_grpjquota:
1160 case Opt_offusrjquota:
1161 case Opt_offgrpjquota:
1162 case Opt_jqfmt_vfsold:
1163 case Opt_jqfmt_vfsv0:
1164 printk(KERN_ERR
1165 "EXT3-fs: journaled quota options not "
1166 "supported.\n");
1167 break;
1168 case Opt_noquota:
1169 break;
1170 #endif
1171 case Opt_abort:
1172 set_opt(sbi->s_mount_opt, ABORT);
1173 break;
1174 case Opt_barrier:
1175 if (match_int(&args[0], &option))
1176 return 0;
1177 if (option)
1178 set_opt(sbi->s_mount_opt, BARRIER);
1179 else
1180 clear_opt(sbi->s_mount_opt, BARRIER);
1181 break;
1182 case Opt_ignore:
1183 break;
1184 case Opt_resize:
1185 if (!is_remount) {
1186 printk("EXT3-fs: resize option only available "
1187 "for remount\n");
1188 return 0;
1189 }
1190 if (match_int(&args[0], &option) != 0)
1191 return 0;
1192 *n_blocks_count = option;
1193 break;
1194 case Opt_nobh:
1195 set_opt(sbi->s_mount_opt, NOBH);
1196 break;
1197 case Opt_bh:
1198 clear_opt(sbi->s_mount_opt, NOBH);
1199 break;
1200 default:
1201 printk (KERN_ERR
1202 "EXT3-fs: Unrecognized mount option \"%s\" "
1203 "or missing value\n", p);
1204 return 0;
1205 }
1206 }
1207 #ifdef CONFIG_QUOTA
1208 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1209 if ((sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA) &&
1210 sbi->s_qf_names[USRQUOTA])
1211 clear_opt(sbi->s_mount_opt, USRQUOTA);
1212
1213 if ((sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA) &&
1214 sbi->s_qf_names[GRPQUOTA])
1215 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1216
1217 if ((sbi->s_qf_names[USRQUOTA] &&
1218 (sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA)) ||
1219 (sbi->s_qf_names[GRPQUOTA] &&
1220 (sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA))) {
1221 printk(KERN_ERR "EXT3-fs: old and new quota "
1222 "format mixing.\n");
1223 return 0;
1224 }
1225
1226 if (!sbi->s_jquota_fmt) {
1227 printk(KERN_ERR "EXT3-fs: journaled quota format "
1228 "not specified.\n");
1229 return 0;
1230 }
1231 } else {
1232 if (sbi->s_jquota_fmt) {
1233 printk(KERN_ERR "EXT3-fs: journaled quota format "
1234 "specified with no journaling "
1235 "enabled.\n");
1236 return 0;
1237 }
1238 }
1239 #endif
1240 return 1;
1241 }
1242
1243 static int ext3_setup_super(struct super_block *sb, struct ext3_super_block *es,
1244 int read_only)
1245 {
1246 struct ext3_sb_info *sbi = EXT3_SB(sb);
1247 int res = 0;
1248
1249 if (le32_to_cpu(es->s_rev_level) > EXT3_MAX_SUPP_REV) {
1250 printk (KERN_ERR "EXT3-fs warning: revision level too high, "
1251 "forcing read-only mode\n");
1252 res = MS_RDONLY;
1253 }
1254 if (read_only)
1255 return res;
1256 if (!(sbi->s_mount_state & EXT3_VALID_FS))
1257 printk (KERN_WARNING "EXT3-fs warning: mounting unchecked fs, "
1258 "running e2fsck is recommended\n");
1259 else if ((sbi->s_mount_state & EXT3_ERROR_FS))
1260 printk (KERN_WARNING
1261 "EXT3-fs warning: mounting fs with errors, "
1262 "running e2fsck is recommended\n");
1263 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1264 le16_to_cpu(es->s_mnt_count) >=
1265 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1266 printk (KERN_WARNING
1267 "EXT3-fs warning: maximal mount count reached, "
1268 "running e2fsck is recommended\n");
1269 else if (le32_to_cpu(es->s_checkinterval) &&
1270 (le32_to_cpu(es->s_lastcheck) +
1271 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1272 printk (KERN_WARNING
1273 "EXT3-fs warning: checktime reached, "
1274 "running e2fsck is recommended\n");
1275 #if 0
1276 /* @@@ We _will_ want to clear the valid bit if we find
1277 inconsistencies, to force a fsck at reboot. But for
1278 a plain journaled filesystem we can keep it set as
1279 valid forever! :) */
1280 es->s_state &= cpu_to_le16(~EXT3_VALID_FS);
1281 #endif
1282 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1283 es->s_max_mnt_count = cpu_to_le16(EXT3_DFL_MAX_MNT_COUNT);
1284 le16_add_cpu(&es->s_mnt_count, 1);
1285 es->s_mtime = cpu_to_le32(get_seconds());
1286 ext3_update_dynamic_rev(sb);
1287 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
1288
1289 ext3_commit_super(sb, es, 1);
1290 if (test_opt(sb, DEBUG))
1291 printk(KERN_INFO "[EXT3 FS bs=%lu, gc=%lu, "
1292 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1293 sb->s_blocksize,
1294 sbi->s_groups_count,
1295 EXT3_BLOCKS_PER_GROUP(sb),
1296 EXT3_INODES_PER_GROUP(sb),
1297 sbi->s_mount_opt);
1298
1299 printk(KERN_INFO "EXT3 FS on %s, ", sb->s_id);
1300 if (EXT3_SB(sb)->s_journal->j_inode == NULL) {
1301 char b[BDEVNAME_SIZE];
1302
1303 printk("external journal on %s\n",
1304 bdevname(EXT3_SB(sb)->s_journal->j_dev, b));
1305 } else {
1306 printk("internal journal\n");
1307 }
1308 return res;
1309 }
1310
1311 /* Called at mount-time, super-block is locked */
1312 static int ext3_check_descriptors(struct super_block *sb)
1313 {
1314 struct ext3_sb_info *sbi = EXT3_SB(sb);
1315 int i;
1316
1317 ext3_debug ("Checking group descriptors");
1318
1319 for (i = 0; i < sbi->s_groups_count; i++) {
1320 struct ext3_group_desc *gdp = ext3_get_group_desc(sb, i, NULL);
1321 ext3_fsblk_t first_block = ext3_group_first_block_no(sb, i);
1322 ext3_fsblk_t last_block;
1323
1324 if (i == sbi->s_groups_count - 1)
1325 last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
1326 else
1327 last_block = first_block +
1328 (EXT3_BLOCKS_PER_GROUP(sb) - 1);
1329
1330 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
1331 le32_to_cpu(gdp->bg_block_bitmap) > last_block)
1332 {
1333 ext3_error (sb, "ext3_check_descriptors",
1334 "Block bitmap for group %d"
1335 " not in group (block %lu)!",
1336 i, (unsigned long)
1337 le32_to_cpu(gdp->bg_block_bitmap));
1338 return 0;
1339 }
1340 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
1341 le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
1342 {
1343 ext3_error (sb, "ext3_check_descriptors",
1344 "Inode bitmap for group %d"
1345 " not in group (block %lu)!",
1346 i, (unsigned long)
1347 le32_to_cpu(gdp->bg_inode_bitmap));
1348 return 0;
1349 }
1350 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
1351 le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
1352 last_block)
1353 {
1354 ext3_error (sb, "ext3_check_descriptors",
1355 "Inode table for group %d"
1356 " not in group (block %lu)!",
1357 i, (unsigned long)
1358 le32_to_cpu(gdp->bg_inode_table));
1359 return 0;
1360 }
1361 }
1362
1363 sbi->s_es->s_free_blocks_count=cpu_to_le32(ext3_count_free_blocks(sb));
1364 sbi->s_es->s_free_inodes_count=cpu_to_le32(ext3_count_free_inodes(sb));
1365 return 1;
1366 }
1367
1368
1369 /* ext3_orphan_cleanup() walks a singly-linked list of inodes (starting at
1370 * the superblock) which were deleted from all directories, but held open by
1371 * a process at the time of a crash. We walk the list and try to delete these
1372 * inodes at recovery time (only with a read-write filesystem).
1373 *
1374 * In order to keep the orphan inode chain consistent during traversal (in
1375 * case of crash during recovery), we link each inode into the superblock
1376 * orphan list_head and handle it the same way as an inode deletion during
1377 * normal operation (which journals the operations for us).
1378 *
1379 * We only do an iget() and an iput() on each inode, which is very safe if we
1380 * accidentally point at an in-use or already deleted inode. The worst that
1381 * can happen in this case is that we get a "bit already cleared" message from
1382 * ext3_free_inode(). The only reason we would point at a wrong inode is if
1383 * e2fsck was run on this filesystem, and it must have already done the orphan
1384 * inode cleanup for us, so we can safely abort without any further action.
1385 */
1386 static void ext3_orphan_cleanup (struct super_block * sb,
1387 struct ext3_super_block * es)
1388 {
1389 unsigned int s_flags = sb->s_flags;
1390 int nr_orphans = 0, nr_truncates = 0;
1391 #ifdef CONFIG_QUOTA
1392 int i;
1393 #endif
1394 if (!es->s_last_orphan) {
1395 jbd_debug(4, "no orphan inodes to clean up\n");
1396 return;
1397 }
1398
1399 if (bdev_read_only(sb->s_bdev)) {
1400 printk(KERN_ERR "EXT3-fs: write access "
1401 "unavailable, skipping orphan cleanup.\n");
1402 return;
1403 }
1404
1405 if (EXT3_SB(sb)->s_mount_state & EXT3_ERROR_FS) {
1406 if (es->s_last_orphan)
1407 jbd_debug(1, "Errors on filesystem, "
1408 "clearing orphan list.\n");
1409 es->s_last_orphan = 0;
1410 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1411 return;
1412 }
1413
1414 if (s_flags & MS_RDONLY) {
1415 printk(KERN_INFO "EXT3-fs: %s: orphan cleanup on readonly fs\n",
1416 sb->s_id);
1417 sb->s_flags &= ~MS_RDONLY;
1418 }
1419 #ifdef CONFIG_QUOTA
1420 /* Needed for iput() to work correctly and not trash data */
1421 sb->s_flags |= MS_ACTIVE;
1422 /* Turn on quotas so that they are updated correctly */
1423 for (i = 0; i < MAXQUOTAS; i++) {
1424 if (EXT3_SB(sb)->s_qf_names[i]) {
1425 int ret = ext3_quota_on_mount(sb, i);
1426 if (ret < 0)
1427 printk(KERN_ERR
1428 "EXT3-fs: Cannot turn on journaled "
1429 "quota: error %d\n", ret);
1430 }
1431 }
1432 #endif
1433
1434 while (es->s_last_orphan) {
1435 struct inode *inode;
1436
1437 inode = ext3_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1438 if (IS_ERR(inode)) {
1439 es->s_last_orphan = 0;
1440 break;
1441 }
1442
1443 list_add(&EXT3_I(inode)->i_orphan, &EXT3_SB(sb)->s_orphan);
1444 vfs_dq_init(inode);
1445 if (inode->i_nlink) {
1446 printk(KERN_DEBUG
1447 "%s: truncating inode %lu to %Ld bytes\n",
1448 __func__, inode->i_ino, inode->i_size);
1449 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1450 inode->i_ino, inode->i_size);
1451 ext3_truncate(inode);
1452 nr_truncates++;
1453 } else {
1454 printk(KERN_DEBUG
1455 "%s: deleting unreferenced inode %lu\n",
1456 __func__, inode->i_ino);
1457 jbd_debug(2, "deleting unreferenced inode %lu\n",
1458 inode->i_ino);
1459 nr_orphans++;
1460 }
1461 iput(inode); /* The delete magic happens here! */
1462 }
1463
1464 #define PLURAL(x) (x), ((x)==1) ? "" : "s"
1465
1466 if (nr_orphans)
1467 printk(KERN_INFO "EXT3-fs: %s: %d orphan inode%s deleted\n",
1468 sb->s_id, PLURAL(nr_orphans));
1469 if (nr_truncates)
1470 printk(KERN_INFO "EXT3-fs: %s: %d truncate%s cleaned up\n",
1471 sb->s_id, PLURAL(nr_truncates));
1472 #ifdef CONFIG_QUOTA
1473 /* Turn quotas off */
1474 for (i = 0; i < MAXQUOTAS; i++) {
1475 if (sb_dqopt(sb)->files[i])
1476 vfs_quota_off(sb, i, 0);
1477 }
1478 #endif
1479 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1480 }
1481
1482 /*
1483 * Maximal file size. There is a direct, and {,double-,triple-}indirect
1484 * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
1485 * We need to be 1 filesystem block less than the 2^32 sector limit.
1486 */
1487 static loff_t ext3_max_size(int bits)
1488 {
1489 loff_t res = EXT3_NDIR_BLOCKS;
1490 int meta_blocks;
1491 loff_t upper_limit;
1492
1493 /* This is calculated to be the largest file size for a
1494 * dense, file such that the total number of
1495 * sectors in the file, including data and all indirect blocks,
1496 * does not exceed 2^32 -1
1497 * __u32 i_blocks representing the total number of
1498 * 512 bytes blocks of the file
1499 */
1500 upper_limit = (1LL << 32) - 1;
1501
1502 /* total blocks in file system block size */
1503 upper_limit >>= (bits - 9);
1504
1505
1506 /* indirect blocks */
1507 meta_blocks = 1;
1508 /* double indirect blocks */
1509 meta_blocks += 1 + (1LL << (bits-2));
1510 /* tripple indirect blocks */
1511 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1512
1513 upper_limit -= meta_blocks;
1514 upper_limit <<= bits;
1515
1516 res += 1LL << (bits-2);
1517 res += 1LL << (2*(bits-2));
1518 res += 1LL << (3*(bits-2));
1519 res <<= bits;
1520 if (res > upper_limit)
1521 res = upper_limit;
1522
1523 if (res > MAX_LFS_FILESIZE)
1524 res = MAX_LFS_FILESIZE;
1525
1526 return res;
1527 }
1528
1529 static ext3_fsblk_t descriptor_loc(struct super_block *sb,
1530 ext3_fsblk_t logic_sb_block,
1531 int nr)
1532 {
1533 struct ext3_sb_info *sbi = EXT3_SB(sb);
1534 unsigned long bg, first_meta_bg;
1535 int has_super = 0;
1536
1537 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1538
1539 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_META_BG) ||
1540 nr < first_meta_bg)
1541 return (logic_sb_block + nr + 1);
1542 bg = sbi->s_desc_per_block * nr;
1543 if (ext3_bg_has_super(sb, bg))
1544 has_super = 1;
1545 return (has_super + ext3_group_first_block_no(sb, bg));
1546 }
1547
1548
1549 static int ext3_fill_super (struct super_block *sb, void *data, int silent)
1550 {
1551 struct buffer_head * bh;
1552 struct ext3_super_block *es = NULL;
1553 struct ext3_sb_info *sbi;
1554 ext3_fsblk_t block;
1555 ext3_fsblk_t sb_block = get_sb_block(&data);
1556 ext3_fsblk_t logic_sb_block;
1557 unsigned long offset = 0;
1558 unsigned int journal_inum = 0;
1559 unsigned long journal_devnum = 0;
1560 unsigned long def_mount_opts;
1561 struct inode *root;
1562 int blocksize;
1563 int hblock;
1564 int db_count;
1565 int i;
1566 int needs_recovery;
1567 int ret = -EINVAL;
1568 __le32 features;
1569 int err;
1570
1571 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1572 if (!sbi)
1573 return -ENOMEM;
1574
1575 sbi->s_blockgroup_lock =
1576 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
1577 if (!sbi->s_blockgroup_lock) {
1578 kfree(sbi);
1579 return -ENOMEM;
1580 }
1581 sb->s_fs_info = sbi;
1582 sbi->s_mount_opt = 0;
1583 sbi->s_resuid = EXT3_DEF_RESUID;
1584 sbi->s_resgid = EXT3_DEF_RESGID;
1585 sbi->s_sb_block = sb_block;
1586
1587 unlock_kernel();
1588
1589 blocksize = sb_min_blocksize(sb, EXT3_MIN_BLOCK_SIZE);
1590 if (!blocksize) {
1591 printk(KERN_ERR "EXT3-fs: unable to set blocksize\n");
1592 goto out_fail;
1593 }
1594
1595 /*
1596 * The ext3 superblock will not be buffer aligned for other than 1kB
1597 * block sizes. We need to calculate the offset from buffer start.
1598 */
1599 if (blocksize != EXT3_MIN_BLOCK_SIZE) {
1600 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1601 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1602 } else {
1603 logic_sb_block = sb_block;
1604 }
1605
1606 if (!(bh = sb_bread(sb, logic_sb_block))) {
1607 printk (KERN_ERR "EXT3-fs: unable to read superblock\n");
1608 goto out_fail;
1609 }
1610 /*
1611 * Note: s_es must be initialized as soon as possible because
1612 * some ext3 macro-instructions depend on its value
1613 */
1614 es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
1615 sbi->s_es = es;
1616 sb->s_magic = le16_to_cpu(es->s_magic);
1617 if (sb->s_magic != EXT3_SUPER_MAGIC)
1618 goto cantfind_ext3;
1619
1620 /* Set defaults before we parse the mount options */
1621 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1622 if (def_mount_opts & EXT3_DEFM_DEBUG)
1623 set_opt(sbi->s_mount_opt, DEBUG);
1624 if (def_mount_opts & EXT3_DEFM_BSDGROUPS)
1625 set_opt(sbi->s_mount_opt, GRPID);
1626 if (def_mount_opts & EXT3_DEFM_UID16)
1627 set_opt(sbi->s_mount_opt, NO_UID32);
1628 #ifdef CONFIG_EXT3_FS_XATTR
1629 if (def_mount_opts & EXT3_DEFM_XATTR_USER)
1630 set_opt(sbi->s_mount_opt, XATTR_USER);
1631 #endif
1632 #ifdef CONFIG_EXT3_FS_POSIX_ACL
1633 if (def_mount_opts & EXT3_DEFM_ACL)
1634 set_opt(sbi->s_mount_opt, POSIX_ACL);
1635 #endif
1636 if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_DATA)
1637 sbi->s_mount_opt |= EXT3_MOUNT_JOURNAL_DATA;
1638 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_ORDERED)
1639 sbi->s_mount_opt |= EXT3_MOUNT_ORDERED_DATA;
1640 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_WBACK)
1641 sbi->s_mount_opt |= EXT3_MOUNT_WRITEBACK_DATA;
1642
1643 if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_PANIC)
1644 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1645 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_CONTINUE)
1646 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1647 else
1648 set_opt(sbi->s_mount_opt, ERRORS_RO);
1649
1650 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1651 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1652
1653 set_opt(sbi->s_mount_opt, RESERVATION);
1654
1655 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1656 NULL, 0))
1657 goto failed_mount;
1658
1659 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1660 ((sbi->s_mount_opt & EXT3_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1661
1662 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV &&
1663 (EXT3_HAS_COMPAT_FEATURE(sb, ~0U) ||
1664 EXT3_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1665 EXT3_HAS_INCOMPAT_FEATURE(sb, ~0U)))
1666 printk(KERN_WARNING
1667 "EXT3-fs warning: feature flags set on rev 0 fs, "
1668 "running e2fsck is recommended\n");
1669 /*
1670 * Check feature flags regardless of the revision level, since we
1671 * previously didn't change the revision level when setting the flags,
1672 * so there is a chance incompat flags are set on a rev 0 filesystem.
1673 */
1674 features = EXT3_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP);
1675 if (features) {
1676 printk(KERN_ERR "EXT3-fs: %s: couldn't mount because of "
1677 "unsupported optional features (%x).\n",
1678 sb->s_id, le32_to_cpu(features));
1679 goto failed_mount;
1680 }
1681 features = EXT3_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP);
1682 if (!(sb->s_flags & MS_RDONLY) && features) {
1683 printk(KERN_ERR "EXT3-fs: %s: couldn't mount RDWR because of "
1684 "unsupported optional features (%x).\n",
1685 sb->s_id, le32_to_cpu(features));
1686 goto failed_mount;
1687 }
1688 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
1689
1690 if (blocksize < EXT3_MIN_BLOCK_SIZE ||
1691 blocksize > EXT3_MAX_BLOCK_SIZE) {
1692 printk(KERN_ERR
1693 "EXT3-fs: Unsupported filesystem blocksize %d on %s.\n",
1694 blocksize, sb->s_id);
1695 goto failed_mount;
1696 }
1697
1698 hblock = bdev_logical_block_size(sb->s_bdev);
1699 if (sb->s_blocksize != blocksize) {
1700 /*
1701 * Make sure the blocksize for the filesystem is larger
1702 * than the hardware sectorsize for the machine.
1703 */
1704 if (blocksize < hblock) {
1705 printk(KERN_ERR "EXT3-fs: blocksize %d too small for "
1706 "device blocksize %d.\n", blocksize, hblock);
1707 goto failed_mount;
1708 }
1709
1710 brelse (bh);
1711 if (!sb_set_blocksize(sb, blocksize)) {
1712 printk(KERN_ERR "EXT3-fs: bad blocksize %d.\n",
1713 blocksize);
1714 goto out_fail;
1715 }
1716 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1717 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1718 bh = sb_bread(sb, logic_sb_block);
1719 if (!bh) {
1720 printk(KERN_ERR
1721 "EXT3-fs: Can't read superblock on 2nd try.\n");
1722 goto failed_mount;
1723 }
1724 es = (struct ext3_super_block *)(((char *)bh->b_data) + offset);
1725 sbi->s_es = es;
1726 if (es->s_magic != cpu_to_le16(EXT3_SUPER_MAGIC)) {
1727 printk (KERN_ERR
1728 "EXT3-fs: Magic mismatch, very weird !\n");
1729 goto failed_mount;
1730 }
1731 }
1732
1733 sb->s_maxbytes = ext3_max_size(sb->s_blocksize_bits);
1734
1735 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV) {
1736 sbi->s_inode_size = EXT3_GOOD_OLD_INODE_SIZE;
1737 sbi->s_first_ino = EXT3_GOOD_OLD_FIRST_INO;
1738 } else {
1739 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1740 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1741 if ((sbi->s_inode_size < EXT3_GOOD_OLD_INODE_SIZE) ||
1742 (!is_power_of_2(sbi->s_inode_size)) ||
1743 (sbi->s_inode_size > blocksize)) {
1744 printk (KERN_ERR
1745 "EXT3-fs: unsupported inode size: %d\n",
1746 sbi->s_inode_size);
1747 goto failed_mount;
1748 }
1749 }
1750 sbi->s_frag_size = EXT3_MIN_FRAG_SIZE <<
1751 le32_to_cpu(es->s_log_frag_size);
1752 if (blocksize != sbi->s_frag_size) {
1753 printk(KERN_ERR
1754 "EXT3-fs: fragsize %lu != blocksize %u (unsupported)\n",
1755 sbi->s_frag_size, blocksize);
1756 goto failed_mount;
1757 }
1758 sbi->s_frags_per_block = 1;
1759 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1760 sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
1761 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
1762 if (EXT3_INODE_SIZE(sb) == 0 || EXT3_INODES_PER_GROUP(sb) == 0)
1763 goto cantfind_ext3;
1764 sbi->s_inodes_per_block = blocksize / EXT3_INODE_SIZE(sb);
1765 if (sbi->s_inodes_per_block == 0)
1766 goto cantfind_ext3;
1767 sbi->s_itb_per_group = sbi->s_inodes_per_group /
1768 sbi->s_inodes_per_block;
1769 sbi->s_desc_per_block = blocksize / sizeof(struct ext3_group_desc);
1770 sbi->s_sbh = bh;
1771 sbi->s_mount_state = le16_to_cpu(es->s_state);
1772 sbi->s_addr_per_block_bits = ilog2(EXT3_ADDR_PER_BLOCK(sb));
1773 sbi->s_desc_per_block_bits = ilog2(EXT3_DESC_PER_BLOCK(sb));
1774 for (i=0; i < 4; i++)
1775 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
1776 sbi->s_def_hash_version = es->s_def_hash_version;
1777 i = le32_to_cpu(es->s_flags);
1778 if (i & EXT2_FLAGS_UNSIGNED_HASH)
1779 sbi->s_hash_unsigned = 3;
1780 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
1781 #ifdef __CHAR_UNSIGNED__
1782 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
1783 sbi->s_hash_unsigned = 3;
1784 #else
1785 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
1786 #endif
1787 }
1788
1789 if (sbi->s_blocks_per_group > blocksize * 8) {
1790 printk (KERN_ERR
1791 "EXT3-fs: #blocks per group too big: %lu\n",
1792 sbi->s_blocks_per_group);
1793 goto failed_mount;
1794 }
1795 if (sbi->s_frags_per_group > blocksize * 8) {
1796 printk (KERN_ERR
1797 "EXT3-fs: #fragments per group too big: %lu\n",
1798 sbi->s_frags_per_group);
1799 goto failed_mount;
1800 }
1801 if (sbi->s_inodes_per_group > blocksize * 8) {
1802 printk (KERN_ERR
1803 "EXT3-fs: #inodes per group too big: %lu\n",
1804 sbi->s_inodes_per_group);
1805 goto failed_mount;
1806 }
1807
1808 if (le32_to_cpu(es->s_blocks_count) >
1809 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
1810 printk(KERN_ERR "EXT3-fs: filesystem on %s:"
1811 " too large to mount safely\n", sb->s_id);
1812 if (sizeof(sector_t) < 8)
1813 printk(KERN_WARNING "EXT3-fs: CONFIG_LBDAF not "
1814 "enabled\n");
1815 goto failed_mount;
1816 }
1817
1818 if (EXT3_BLOCKS_PER_GROUP(sb) == 0)
1819 goto cantfind_ext3;
1820 sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1821 le32_to_cpu(es->s_first_data_block) - 1)
1822 / EXT3_BLOCKS_PER_GROUP(sb)) + 1;
1823 db_count = (sbi->s_groups_count + EXT3_DESC_PER_BLOCK(sb) - 1) /
1824 EXT3_DESC_PER_BLOCK(sb);
1825 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
1826 GFP_KERNEL);
1827 if (sbi->s_group_desc == NULL) {
1828 printk (KERN_ERR "EXT3-fs: not enough memory\n");
1829 goto failed_mount;
1830 }
1831
1832 bgl_lock_init(sbi->s_blockgroup_lock);
1833
1834 for (i = 0; i < db_count; i++) {
1835 block = descriptor_loc(sb, logic_sb_block, i);
1836 sbi->s_group_desc[i] = sb_bread(sb, block);
1837 if (!sbi->s_group_desc[i]) {
1838 printk (KERN_ERR "EXT3-fs: "
1839 "can't read group descriptor %d\n", i);
1840 db_count = i;
1841 goto failed_mount2;
1842 }
1843 }
1844 if (!ext3_check_descriptors (sb)) {
1845 printk(KERN_ERR "EXT3-fs: group descriptors corrupted!\n");
1846 goto failed_mount2;
1847 }
1848 sbi->s_gdb_count = db_count;
1849 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1850 spin_lock_init(&sbi->s_next_gen_lock);
1851
1852 err = percpu_counter_init(&sbi->s_freeblocks_counter,
1853 ext3_count_free_blocks(sb));
1854 if (!err) {
1855 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1856 ext3_count_free_inodes(sb));
1857 }
1858 if (!err) {
1859 err = percpu_counter_init(&sbi->s_dirs_counter,
1860 ext3_count_dirs(sb));
1861 }
1862 if (err) {
1863 printk(KERN_ERR "EXT3-fs: insufficient memory\n");
1864 goto failed_mount3;
1865 }
1866
1867 /* per fileystem reservation list head & lock */
1868 spin_lock_init(&sbi->s_rsv_window_lock);
1869 sbi->s_rsv_window_root = RB_ROOT;
1870 /* Add a single, static dummy reservation to the start of the
1871 * reservation window list --- it gives us a placeholder for
1872 * append-at-start-of-list which makes the allocation logic
1873 * _much_ simpler. */
1874 sbi->s_rsv_window_head.rsv_start = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1875 sbi->s_rsv_window_head.rsv_end = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1876 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1877 sbi->s_rsv_window_head.rsv_goal_size = 0;
1878 ext3_rsv_window_add(sb, &sbi->s_rsv_window_head);
1879
1880 /*
1881 * set up enough so that it can read an inode
1882 */
1883 sb->s_op = &ext3_sops;
1884 sb->s_export_op = &ext3_export_ops;
1885 sb->s_xattr = ext3_xattr_handlers;
1886 #ifdef CONFIG_QUOTA
1887 sb->s_qcop = &ext3_qctl_operations;
1888 sb->dq_op = &ext3_quota_operations;
1889 #endif
1890 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
1891
1892 sb->s_root = NULL;
1893
1894 needs_recovery = (es->s_last_orphan != 0 ||
1895 EXT3_HAS_INCOMPAT_FEATURE(sb,
1896 EXT3_FEATURE_INCOMPAT_RECOVER));
1897
1898 /*
1899 * The first inode we look at is the journal inode. Don't try
1900 * root first: it may be modified in the journal!
1901 */
1902 if (!test_opt(sb, NOLOAD) &&
1903 EXT3_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL)) {
1904 if (ext3_load_journal(sb, es, journal_devnum))
1905 goto failed_mount3;
1906 } else if (journal_inum) {
1907 if (ext3_create_journal(sb, es, journal_inum))
1908 goto failed_mount3;
1909 } else {
1910 if (!silent)
1911 printk (KERN_ERR
1912 "ext3: No journal on filesystem on %s\n",
1913 sb->s_id);
1914 goto failed_mount3;
1915 }
1916
1917 /* We have now updated the journal if required, so we can
1918 * validate the data journaling mode. */
1919 switch (test_opt(sb, DATA_FLAGS)) {
1920 case 0:
1921 /* No mode set, assume a default based on the journal
1922 capabilities: ORDERED_DATA if the journal can
1923 cope, else JOURNAL_DATA */
1924 if (journal_check_available_features
1925 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE))
1926 set_opt(sbi->s_mount_opt, DEFAULT_DATA_MODE);
1927 else
1928 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
1929 break;
1930
1931 case EXT3_MOUNT_ORDERED_DATA:
1932 case EXT3_MOUNT_WRITEBACK_DATA:
1933 if (!journal_check_available_features
1934 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE)) {
1935 printk(KERN_ERR "EXT3-fs: Journal does not support "
1936 "requested data journaling mode\n");
1937 goto failed_mount4;
1938 }
1939 default:
1940 break;
1941 }
1942
1943 if (test_opt(sb, NOBH)) {
1944 if (!(test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_WRITEBACK_DATA)) {
1945 printk(KERN_WARNING "EXT3-fs: Ignoring nobh option - "
1946 "its supported only with writeback mode\n");
1947 clear_opt(sbi->s_mount_opt, NOBH);
1948 }
1949 }
1950 /*
1951 * The journal_load will have done any necessary log recovery,
1952 * so we can safely mount the rest of the filesystem now.
1953 */
1954
1955 root = ext3_iget(sb, EXT3_ROOT_INO);
1956 if (IS_ERR(root)) {
1957 printk(KERN_ERR "EXT3-fs: get root inode failed\n");
1958 ret = PTR_ERR(root);
1959 goto failed_mount4;
1960 }
1961 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1962 iput(root);
1963 printk(KERN_ERR "EXT3-fs: corrupt root inode, run e2fsck\n");
1964 goto failed_mount4;
1965 }
1966 sb->s_root = d_alloc_root(root);
1967 if (!sb->s_root) {
1968 printk(KERN_ERR "EXT3-fs: get root dentry failed\n");
1969 iput(root);
1970 ret = -ENOMEM;
1971 goto failed_mount4;
1972 }
1973
1974 ext3_setup_super (sb, es, sb->s_flags & MS_RDONLY);
1975 /*
1976 * akpm: core read_super() calls in here with the superblock locked.
1977 * That deadlocks, because orphan cleanup needs to lock the superblock
1978 * in numerous places. Here we just pop the lock - it's relatively
1979 * harmless, because we are now ready to accept write_super() requests,
1980 * and aviro says that's the only reason for hanging onto the
1981 * superblock lock.
1982 */
1983 EXT3_SB(sb)->s_mount_state |= EXT3_ORPHAN_FS;
1984 ext3_orphan_cleanup(sb, es);
1985 EXT3_SB(sb)->s_mount_state &= ~EXT3_ORPHAN_FS;
1986 if (needs_recovery)
1987 printk (KERN_INFO "EXT3-fs: recovery complete.\n");
1988 ext3_mark_recovery_complete(sb, es);
1989 printk (KERN_INFO "EXT3-fs: mounted filesystem with %s data mode.\n",
1990 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA ? "journal":
1991 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA ? "ordered":
1992 "writeback");
1993
1994 lock_kernel();
1995 return 0;
1996
1997 cantfind_ext3:
1998 if (!silent)
1999 printk(KERN_ERR "VFS: Can't find ext3 filesystem on dev %s.\n",
2000 sb->s_id);
2001 goto failed_mount;
2002
2003 failed_mount4:
2004 journal_destroy(sbi->s_journal);
2005 failed_mount3:
2006 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2007 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2008 percpu_counter_destroy(&sbi->s_dirs_counter);
2009 failed_mount2:
2010 for (i = 0; i < db_count; i++)
2011 brelse(sbi->s_group_desc[i]);
2012 kfree(sbi->s_group_desc);
2013 failed_mount:
2014 #ifdef CONFIG_QUOTA
2015 for (i = 0; i < MAXQUOTAS; i++)
2016 kfree(sbi->s_qf_names[i]);
2017 #endif
2018 ext3_blkdev_remove(sbi);
2019 brelse(bh);
2020 out_fail:
2021 sb->s_fs_info = NULL;
2022 kfree(sbi->s_blockgroup_lock);
2023 kfree(sbi);
2024 lock_kernel();
2025 return ret;
2026 }
2027
2028 /*
2029 * Setup any per-fs journal parameters now. We'll do this both on
2030 * initial mount, once the journal has been initialised but before we've
2031 * done any recovery; and again on any subsequent remount.
2032 */
2033 static void ext3_init_journal_params(struct super_block *sb, journal_t *journal)
2034 {
2035 struct ext3_sb_info *sbi = EXT3_SB(sb);
2036
2037 if (sbi->s_commit_interval)
2038 journal->j_commit_interval = sbi->s_commit_interval;
2039 /* We could also set up an ext3-specific default for the commit
2040 * interval here, but for now we'll just fall back to the jbd
2041 * default. */
2042
2043 spin_lock(&journal->j_state_lock);
2044 if (test_opt(sb, BARRIER))
2045 journal->j_flags |= JFS_BARRIER;
2046 else
2047 journal->j_flags &= ~JFS_BARRIER;
2048 if (test_opt(sb, DATA_ERR_ABORT))
2049 journal->j_flags |= JFS_ABORT_ON_SYNCDATA_ERR;
2050 else
2051 journal->j_flags &= ~JFS_ABORT_ON_SYNCDATA_ERR;
2052 spin_unlock(&journal->j_state_lock);
2053 }
2054
2055 static journal_t *ext3_get_journal(struct super_block *sb,
2056 unsigned int journal_inum)
2057 {
2058 struct inode *journal_inode;
2059 journal_t *journal;
2060
2061 /* First, test for the existence of a valid inode on disk. Bad
2062 * things happen if we iget() an unused inode, as the subsequent
2063 * iput() will try to delete it. */
2064
2065 journal_inode = ext3_iget(sb, journal_inum);
2066 if (IS_ERR(journal_inode)) {
2067 printk(KERN_ERR "EXT3-fs: no journal found.\n");
2068 return NULL;
2069 }
2070 if (!journal_inode->i_nlink) {
2071 make_bad_inode(journal_inode);
2072 iput(journal_inode);
2073 printk(KERN_ERR "EXT3-fs: journal inode is deleted.\n");
2074 return NULL;
2075 }
2076
2077 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
2078 journal_inode, journal_inode->i_size);
2079 if (!S_ISREG(journal_inode->i_mode)) {
2080 printk(KERN_ERR "EXT3-fs: invalid journal inode.\n");
2081 iput(journal_inode);
2082 return NULL;
2083 }
2084
2085 journal = journal_init_inode(journal_inode);
2086 if (!journal) {
2087 printk(KERN_ERR "EXT3-fs: Could not load journal inode\n");
2088 iput(journal_inode);
2089 return NULL;
2090 }
2091 journal->j_private = sb;
2092 ext3_init_journal_params(sb, journal);
2093 return journal;
2094 }
2095
2096 static journal_t *ext3_get_dev_journal(struct super_block *sb,
2097 dev_t j_dev)
2098 {
2099 struct buffer_head * bh;
2100 journal_t *journal;
2101 ext3_fsblk_t start;
2102 ext3_fsblk_t len;
2103 int hblock, blocksize;
2104 ext3_fsblk_t sb_block;
2105 unsigned long offset;
2106 struct ext3_super_block * es;
2107 struct block_device *bdev;
2108
2109 bdev = ext3_blkdev_get(j_dev);
2110 if (bdev == NULL)
2111 return NULL;
2112
2113 if (bd_claim(bdev, sb)) {
2114 printk(KERN_ERR
2115 "EXT3: failed to claim external journal device.\n");
2116 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
2117 return NULL;
2118 }
2119
2120 blocksize = sb->s_blocksize;
2121 hblock = bdev_logical_block_size(bdev);
2122 if (blocksize < hblock) {
2123 printk(KERN_ERR
2124 "EXT3-fs: blocksize too small for journal device.\n");
2125 goto out_bdev;
2126 }
2127
2128 sb_block = EXT3_MIN_BLOCK_SIZE / blocksize;
2129 offset = EXT3_MIN_BLOCK_SIZE % blocksize;
2130 set_blocksize(bdev, blocksize);
2131 if (!(bh = __bread(bdev, sb_block, blocksize))) {
2132 printk(KERN_ERR "EXT3-fs: couldn't read superblock of "
2133 "external journal\n");
2134 goto out_bdev;
2135 }
2136
2137 es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
2138 if ((le16_to_cpu(es->s_magic) != EXT3_SUPER_MAGIC) ||
2139 !(le32_to_cpu(es->s_feature_incompat) &
2140 EXT3_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2141 printk(KERN_ERR "EXT3-fs: external journal has "
2142 "bad superblock\n");
2143 brelse(bh);
2144 goto out_bdev;
2145 }
2146
2147 if (memcmp(EXT3_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2148 printk(KERN_ERR "EXT3-fs: journal UUID does not match\n");
2149 brelse(bh);
2150 goto out_bdev;
2151 }
2152
2153 len = le32_to_cpu(es->s_blocks_count);
2154 start = sb_block + 1;
2155 brelse(bh); /* we're done with the superblock */
2156
2157 journal = journal_init_dev(bdev, sb->s_bdev,
2158 start, len, blocksize);
2159 if (!journal) {
2160 printk(KERN_ERR "EXT3-fs: failed to create device journal\n");
2161 goto out_bdev;
2162 }
2163 journal->j_private = sb;
2164 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2165 wait_on_buffer(journal->j_sb_buffer);
2166 if (!buffer_uptodate(journal->j_sb_buffer)) {
2167 printk(KERN_ERR "EXT3-fs: I/O error on journal device\n");
2168 goto out_journal;
2169 }
2170 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
2171 printk(KERN_ERR "EXT3-fs: External journal has more than one "
2172 "user (unsupported) - %d\n",
2173 be32_to_cpu(journal->j_superblock->s_nr_users));
2174 goto out_journal;
2175 }
2176 EXT3_SB(sb)->journal_bdev = bdev;
2177 ext3_init_journal_params(sb, journal);
2178 return journal;
2179 out_journal:
2180 journal_destroy(journal);
2181 out_bdev:
2182 ext3_blkdev_put(bdev);
2183 return NULL;
2184 }
2185
2186 static int ext3_load_journal(struct super_block *sb,
2187 struct ext3_super_block *es,
2188 unsigned long journal_devnum)
2189 {
2190 journal_t *journal;
2191 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2192 dev_t journal_dev;
2193 int err = 0;
2194 int really_read_only;
2195
2196 if (journal_devnum &&
2197 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2198 printk(KERN_INFO "EXT3-fs: external journal device major/minor "
2199 "numbers have changed\n");
2200 journal_dev = new_decode_dev(journal_devnum);
2201 } else
2202 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2203
2204 really_read_only = bdev_read_only(sb->s_bdev);
2205
2206 /*
2207 * Are we loading a blank journal or performing recovery after a
2208 * crash? For recovery, we need to check in advance whether we
2209 * can get read-write access to the device.
2210 */
2211
2212 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER)) {
2213 if (sb->s_flags & MS_RDONLY) {
2214 printk(KERN_INFO "EXT3-fs: INFO: recovery "
2215 "required on readonly filesystem.\n");
2216 if (really_read_only) {
2217 printk(KERN_ERR "EXT3-fs: write access "
2218 "unavailable, cannot proceed.\n");
2219 return -EROFS;
2220 }
2221 printk (KERN_INFO "EXT3-fs: write access will "
2222 "be enabled during recovery.\n");
2223 }
2224 }
2225
2226 if (journal_inum && journal_dev) {
2227 printk(KERN_ERR "EXT3-fs: filesystem has both journal "
2228 "and inode journals!\n");
2229 return -EINVAL;
2230 }
2231
2232 if (journal_inum) {
2233 if (!(journal = ext3_get_journal(sb, journal_inum)))
2234 return -EINVAL;
2235 } else {
2236 if (!(journal = ext3_get_dev_journal(sb, journal_dev)))
2237 return -EINVAL;
2238 }
2239
2240 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2241 err = journal_update_format(journal);
2242 if (err) {
2243 printk(KERN_ERR "EXT3-fs: error updating journal.\n");
2244 journal_destroy(journal);
2245 return err;
2246 }
2247 }
2248
2249 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER))
2250 err = journal_wipe(journal, !really_read_only);
2251 if (!err)
2252 err = journal_load(journal);
2253
2254 if (err) {
2255 printk(KERN_ERR "EXT3-fs: error loading journal.\n");
2256 journal_destroy(journal);
2257 return err;
2258 }
2259
2260 EXT3_SB(sb)->s_journal = journal;
2261 ext3_clear_journal_err(sb, es);
2262
2263 if (journal_devnum &&
2264 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2265 es->s_journal_dev = cpu_to_le32(journal_devnum);
2266
2267 /* Make sure we flush the recovery flag to disk. */
2268 ext3_commit_super(sb, es, 1);
2269 }
2270
2271 return 0;
2272 }
2273
2274 static int ext3_create_journal(struct super_block * sb,
2275 struct ext3_super_block * es,
2276 unsigned int journal_inum)
2277 {
2278 journal_t *journal;
2279 int err;
2280
2281 if (sb->s_flags & MS_RDONLY) {
2282 printk(KERN_ERR "EXT3-fs: readonly filesystem when trying to "
2283 "create journal.\n");
2284 return -EROFS;
2285 }
2286
2287 journal = ext3_get_journal(sb, journal_inum);
2288 if (!journal)
2289 return -EINVAL;
2290
2291 printk(KERN_INFO "EXT3-fs: creating new journal on inode %u\n",
2292 journal_inum);
2293
2294 err = journal_create(journal);
2295 if (err) {
2296 printk(KERN_ERR "EXT3-fs: error creating journal.\n");
2297 journal_destroy(journal);
2298 return -EIO;
2299 }
2300
2301 EXT3_SB(sb)->s_journal = journal;
2302
2303 ext3_update_dynamic_rev(sb);
2304 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2305 EXT3_SET_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL);
2306
2307 es->s_journal_inum = cpu_to_le32(journal_inum);
2308
2309 /* Make sure we flush the recovery flag to disk. */
2310 ext3_commit_super(sb, es, 1);
2311
2312 return 0;
2313 }
2314
2315 static int ext3_commit_super(struct super_block *sb,
2316 struct ext3_super_block *es,
2317 int sync)
2318 {
2319 struct buffer_head *sbh = EXT3_SB(sb)->s_sbh;
2320 int error = 0;
2321
2322 if (!sbh)
2323 return error;
2324 es->s_wtime = cpu_to_le32(get_seconds());
2325 es->s_free_blocks_count = cpu_to_le32(ext3_count_free_blocks(sb));
2326 es->s_free_inodes_count = cpu_to_le32(ext3_count_free_inodes(sb));
2327 BUFFER_TRACE(sbh, "marking dirty");
2328 mark_buffer_dirty(sbh);
2329 if (sync)
2330 error = sync_dirty_buffer(sbh);
2331 return error;
2332 }
2333
2334
2335 /*
2336 * Have we just finished recovery? If so, and if we are mounting (or
2337 * remounting) the filesystem readonly, then we will end up with a
2338 * consistent fs on disk. Record that fact.
2339 */
2340 static void ext3_mark_recovery_complete(struct super_block * sb,
2341 struct ext3_super_block * es)
2342 {
2343 journal_t *journal = EXT3_SB(sb)->s_journal;
2344
2345 journal_lock_updates(journal);
2346 if (journal_flush(journal) < 0)
2347 goto out;
2348
2349 lock_super(sb);
2350 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER) &&
2351 sb->s_flags & MS_RDONLY) {
2352 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2353 ext3_commit_super(sb, es, 1);
2354 }
2355 unlock_super(sb);
2356
2357 out:
2358 journal_unlock_updates(journal);
2359 }
2360
2361 /*
2362 * If we are mounting (or read-write remounting) a filesystem whose journal
2363 * has recorded an error from a previous lifetime, move that error to the
2364 * main filesystem now.
2365 */
2366 static void ext3_clear_journal_err(struct super_block * sb,
2367 struct ext3_super_block * es)
2368 {
2369 journal_t *journal;
2370 int j_errno;
2371 const char *errstr;
2372
2373 journal = EXT3_SB(sb)->s_journal;
2374
2375 /*
2376 * Now check for any error status which may have been recorded in the
2377 * journal by a prior ext3_error() or ext3_abort()
2378 */
2379
2380 j_errno = journal_errno(journal);
2381 if (j_errno) {
2382 char nbuf[16];
2383
2384 errstr = ext3_decode_error(sb, j_errno, nbuf);
2385 ext3_warning(sb, __func__, "Filesystem error recorded "
2386 "from previous mount: %s", errstr);
2387 ext3_warning(sb, __func__, "Marking fs in need of "
2388 "filesystem check.");
2389
2390 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
2391 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
2392 ext3_commit_super (sb, es, 1);
2393
2394 journal_clear_err(journal);
2395 }
2396 }
2397
2398 /*
2399 * Force the running and committing transactions to commit,
2400 * and wait on the commit.
2401 */
2402 int ext3_force_commit(struct super_block *sb)
2403 {
2404 journal_t *journal;
2405 int ret;
2406
2407 if (sb->s_flags & MS_RDONLY)
2408 return 0;
2409
2410 journal = EXT3_SB(sb)->s_journal;
2411 ret = ext3_journal_force_commit(journal);
2412 return ret;
2413 }
2414
2415 static int ext3_sync_fs(struct super_block *sb, int wait)
2416 {
2417 tid_t target;
2418
2419 if (journal_start_commit(EXT3_SB(sb)->s_journal, &target)) {
2420 if (wait)
2421 log_wait_commit(EXT3_SB(sb)->s_journal, target);
2422 }
2423 return 0;
2424 }
2425
2426 /*
2427 * LVM calls this function before a (read-only) snapshot is created. This
2428 * gives us a chance to flush the journal completely and mark the fs clean.
2429 */
2430 static int ext3_freeze(struct super_block *sb)
2431 {
2432 int error = 0;
2433 journal_t *journal;
2434
2435 if (!(sb->s_flags & MS_RDONLY)) {
2436 journal = EXT3_SB(sb)->s_journal;
2437
2438 /* Now we set up the journal barrier. */
2439 journal_lock_updates(journal);
2440
2441 /*
2442 * We don't want to clear needs_recovery flag when we failed
2443 * to flush the journal.
2444 */
2445 error = journal_flush(journal);
2446 if (error < 0)
2447 goto out;
2448
2449 /* Journal blocked and flushed, clear needs_recovery flag. */
2450 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2451 error = ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2452 if (error)
2453 goto out;
2454 }
2455 return 0;
2456
2457 out:
2458 journal_unlock_updates(journal);
2459 return error;
2460 }
2461
2462 /*
2463 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2464 * flag here, even though the filesystem is not technically dirty yet.
2465 */
2466 static int ext3_unfreeze(struct super_block *sb)
2467 {
2468 if (!(sb->s_flags & MS_RDONLY)) {
2469 lock_super(sb);
2470 /* Reser the needs_recovery flag before the fs is unlocked. */
2471 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2472 ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2473 unlock_super(sb);
2474 journal_unlock_updates(EXT3_SB(sb)->s_journal);
2475 }
2476 return 0;
2477 }
2478
2479 static int ext3_remount (struct super_block * sb, int * flags, char * data)
2480 {
2481 struct ext3_super_block * es;
2482 struct ext3_sb_info *sbi = EXT3_SB(sb);
2483 ext3_fsblk_t n_blocks_count = 0;
2484 unsigned long old_sb_flags;
2485 struct ext3_mount_options old_opts;
2486 int err;
2487 #ifdef CONFIG_QUOTA
2488 int i;
2489 #endif
2490
2491 lock_kernel();
2492
2493 /* Store the original options */
2494 lock_super(sb);
2495 old_sb_flags = sb->s_flags;
2496 old_opts.s_mount_opt = sbi->s_mount_opt;
2497 old_opts.s_resuid = sbi->s_resuid;
2498 old_opts.s_resgid = sbi->s_resgid;
2499 old_opts.s_commit_interval = sbi->s_commit_interval;
2500 #ifdef CONFIG_QUOTA
2501 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2502 for (i = 0; i < MAXQUOTAS; i++)
2503 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2504 #endif
2505
2506 /*
2507 * Allow the "check" option to be passed as a remount option.
2508 */
2509 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
2510 err = -EINVAL;
2511 goto restore_opts;
2512 }
2513
2514 if (sbi->s_mount_opt & EXT3_MOUNT_ABORT)
2515 ext3_abort(sb, __func__, "Abort forced by user");
2516
2517 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2518 ((sbi->s_mount_opt & EXT3_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2519
2520 es = sbi->s_es;
2521
2522 ext3_init_journal_params(sb, sbi->s_journal);
2523
2524 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
2525 n_blocks_count > le32_to_cpu(es->s_blocks_count)) {
2526 if (sbi->s_mount_opt & EXT3_MOUNT_ABORT) {
2527 err = -EROFS;
2528 goto restore_opts;
2529 }
2530
2531 if (*flags & MS_RDONLY) {
2532 /*
2533 * First of all, the unconditional stuff we have to do
2534 * to disable replay of the journal when we next remount
2535 */
2536 sb->s_flags |= MS_RDONLY;
2537
2538 /*
2539 * OK, test if we are remounting a valid rw partition
2540 * readonly, and if so set the rdonly flag and then
2541 * mark the partition as valid again.
2542 */
2543 if (!(es->s_state & cpu_to_le16(EXT3_VALID_FS)) &&
2544 (sbi->s_mount_state & EXT3_VALID_FS))
2545 es->s_state = cpu_to_le16(sbi->s_mount_state);
2546
2547 /*
2548 * We have to unlock super so that we can wait for
2549 * transactions.
2550 */
2551 unlock_super(sb);
2552 ext3_mark_recovery_complete(sb, es);
2553 lock_super(sb);
2554 } else {
2555 __le32 ret;
2556 if ((ret = EXT3_HAS_RO_COMPAT_FEATURE(sb,
2557 ~EXT3_FEATURE_RO_COMPAT_SUPP))) {
2558 printk(KERN_WARNING "EXT3-fs: %s: couldn't "
2559 "remount RDWR because of unsupported "
2560 "optional features (%x).\n",
2561 sb->s_id, le32_to_cpu(ret));
2562 err = -EROFS;
2563 goto restore_opts;
2564 }
2565
2566 /*
2567 * If we have an unprocessed orphan list hanging
2568 * around from a previously readonly bdev mount,
2569 * require a full umount/remount for now.
2570 */
2571 if (es->s_last_orphan) {
2572 printk(KERN_WARNING "EXT3-fs: %s: couldn't "
2573 "remount RDWR because of unprocessed "
2574 "orphan inode list. Please "
2575 "umount/remount instead.\n",
2576 sb->s_id);
2577 err = -EINVAL;
2578 goto restore_opts;
2579 }
2580
2581 /*
2582 * Mounting a RDONLY partition read-write, so reread
2583 * and store the current valid flag. (It may have
2584 * been changed by e2fsck since we originally mounted
2585 * the partition.)
2586 */
2587 ext3_clear_journal_err(sb, es);
2588 sbi->s_mount_state = le16_to_cpu(es->s_state);
2589 if ((err = ext3_group_extend(sb, es, n_blocks_count)))
2590 goto restore_opts;
2591 if (!ext3_setup_super (sb, es, 0))
2592 sb->s_flags &= ~MS_RDONLY;
2593 }
2594 }
2595 #ifdef CONFIG_QUOTA
2596 /* Release old quota file names */
2597 for (i = 0; i < MAXQUOTAS; i++)
2598 if (old_opts.s_qf_names[i] &&
2599 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2600 kfree(old_opts.s_qf_names[i]);
2601 #endif
2602 unlock_super(sb);
2603 unlock_kernel();
2604 return 0;
2605 restore_opts:
2606 sb->s_flags = old_sb_flags;
2607 sbi->s_mount_opt = old_opts.s_mount_opt;
2608 sbi->s_resuid = old_opts.s_resuid;
2609 sbi->s_resgid = old_opts.s_resgid;
2610 sbi->s_commit_interval = old_opts.s_commit_interval;
2611 #ifdef CONFIG_QUOTA
2612 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
2613 for (i = 0; i < MAXQUOTAS; i++) {
2614 if (sbi->s_qf_names[i] &&
2615 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2616 kfree(sbi->s_qf_names[i]);
2617 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
2618 }
2619 #endif
2620 unlock_super(sb);
2621 unlock_kernel();
2622 return err;
2623 }
2624
2625 static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf)
2626 {
2627 struct super_block *sb = dentry->d_sb;
2628 struct ext3_sb_info *sbi = EXT3_SB(sb);
2629 struct ext3_super_block *es = sbi->s_es;
2630 u64 fsid;
2631
2632 if (test_opt(sb, MINIX_DF)) {
2633 sbi->s_overhead_last = 0;
2634 } else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
2635 unsigned long ngroups = sbi->s_groups_count, i;
2636 ext3_fsblk_t overhead = 0;
2637 smp_rmb();
2638
2639 /*
2640 * Compute the overhead (FS structures). This is constant
2641 * for a given filesystem unless the number of block groups
2642 * changes so we cache the previous value until it does.
2643 */
2644
2645 /*
2646 * All of the blocks before first_data_block are
2647 * overhead
2648 */
2649 overhead = le32_to_cpu(es->s_first_data_block);
2650
2651 /*
2652 * Add the overhead attributed to the superblock and
2653 * block group descriptors. If the sparse superblocks
2654 * feature is turned on, then not all groups have this.
2655 */
2656 for (i = 0; i < ngroups; i++) {
2657 overhead += ext3_bg_has_super(sb, i) +
2658 ext3_bg_num_gdb(sb, i);
2659 cond_resched();
2660 }
2661
2662 /*
2663 * Every block group has an inode bitmap, a block
2664 * bitmap, and an inode table.
2665 */
2666 overhead += ngroups * (2 + sbi->s_itb_per_group);
2667 sbi->s_overhead_last = overhead;
2668 smp_wmb();
2669 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
2670 }
2671
2672 buf->f_type = EXT3_SUPER_MAGIC;
2673 buf->f_bsize = sb->s_blocksize;
2674 buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
2675 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter);
2676 es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
2677 buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
2678 if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
2679 buf->f_bavail = 0;
2680 buf->f_files = le32_to_cpu(es->s_inodes_count);
2681 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
2682 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
2683 buf->f_namelen = EXT3_NAME_LEN;
2684 fsid = le64_to_cpup((void *)es->s_uuid) ^
2685 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
2686 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
2687 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
2688 return 0;
2689 }
2690
2691 /* Helper function for writing quotas on sync - we need to start transaction before quota file
2692 * is locked for write. Otherwise the are possible deadlocks:
2693 * Process 1 Process 2
2694 * ext3_create() quota_sync()
2695 * journal_start() write_dquot()
2696 * vfs_dq_init() down(dqio_mutex)
2697 * down(dqio_mutex) journal_start()
2698 *
2699 */
2700
2701 #ifdef CONFIG_QUOTA
2702
2703 static inline struct inode *dquot_to_inode(struct dquot *dquot)
2704 {
2705 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
2706 }
2707
2708 static int ext3_write_dquot(struct dquot *dquot)
2709 {
2710 int ret, err;
2711 handle_t *handle;
2712 struct inode *inode;
2713
2714 inode = dquot_to_inode(dquot);
2715 handle = ext3_journal_start(inode,
2716 EXT3_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
2717 if (IS_ERR(handle))
2718 return PTR_ERR(handle);
2719 ret = dquot_commit(dquot);
2720 err = ext3_journal_stop(handle);
2721 if (!ret)
2722 ret = err;
2723 return ret;
2724 }
2725
2726 static int ext3_acquire_dquot(struct dquot *dquot)
2727 {
2728 int ret, err;
2729 handle_t *handle;
2730
2731 handle = ext3_journal_start(dquot_to_inode(dquot),
2732 EXT3_QUOTA_INIT_BLOCKS(dquot->dq_sb));
2733 if (IS_ERR(handle))
2734 return PTR_ERR(handle);
2735 ret = dquot_acquire(dquot);
2736 err = ext3_journal_stop(handle);
2737 if (!ret)
2738 ret = err;
2739 return ret;
2740 }
2741
2742 static int ext3_release_dquot(struct dquot *dquot)
2743 {
2744 int ret, err;
2745 handle_t *handle;
2746
2747 handle = ext3_journal_start(dquot_to_inode(dquot),
2748 EXT3_QUOTA_DEL_BLOCKS(dquot->dq_sb));
2749 if (IS_ERR(handle)) {
2750 /* Release dquot anyway to avoid endless cycle in dqput() */
2751 dquot_release(dquot);
2752 return PTR_ERR(handle);
2753 }
2754 ret = dquot_release(dquot);
2755 err = ext3_journal_stop(handle);
2756 if (!ret)
2757 ret = err;
2758 return ret;
2759 }
2760
2761 static int ext3_mark_dquot_dirty(struct dquot *dquot)
2762 {
2763 /* Are we journaling quotas? */
2764 if (EXT3_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
2765 EXT3_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
2766 dquot_mark_dquot_dirty(dquot);
2767 return ext3_write_dquot(dquot);
2768 } else {
2769 return dquot_mark_dquot_dirty(dquot);
2770 }
2771 }
2772
2773 static int ext3_write_info(struct super_block *sb, int type)
2774 {
2775 int ret, err;
2776 handle_t *handle;
2777
2778 /* Data block + inode block */
2779 handle = ext3_journal_start(sb->s_root->d_inode, 2);
2780 if (IS_ERR(handle))
2781 return PTR_ERR(handle);
2782 ret = dquot_commit_info(sb, type);
2783 err = ext3_journal_stop(handle);
2784 if (!ret)
2785 ret = err;
2786 return ret;
2787 }
2788
2789 /*
2790 * Turn on quotas during mount time - we need to find
2791 * the quota file and such...
2792 */
2793 static int ext3_quota_on_mount(struct super_block *sb, int type)
2794 {
2795 return vfs_quota_on_mount(sb, EXT3_SB(sb)->s_qf_names[type],
2796 EXT3_SB(sb)->s_jquota_fmt, type);
2797 }
2798
2799 /*
2800 * Standard function to be called on quota_on
2801 */
2802 static int ext3_quota_on(struct super_block *sb, int type, int format_id,
2803 char *name, int remount)
2804 {
2805 int err;
2806 struct path path;
2807
2808 if (!test_opt(sb, QUOTA))
2809 return -EINVAL;
2810 /* When remounting, no checks are needed and in fact, name is NULL */
2811 if (remount)
2812 return vfs_quota_on(sb, type, format_id, name, remount);
2813
2814 err = kern_path(name, LOOKUP_FOLLOW, &path);
2815 if (err)
2816 return err;
2817
2818 /* Quotafile not on the same filesystem? */
2819 if (path.mnt->mnt_sb != sb) {
2820 path_put(&path);
2821 return -EXDEV;
2822 }
2823 /* Journaling quota? */
2824 if (EXT3_SB(sb)->s_qf_names[type]) {
2825 /* Quotafile not of fs root? */
2826 if (path.dentry->d_parent != sb->s_root)
2827 printk(KERN_WARNING
2828 "EXT3-fs: Quota file not on filesystem root. "
2829 "Journaled quota will not work.\n");
2830 }
2831
2832 /*
2833 * When we journal data on quota file, we have to flush journal to see
2834 * all updates to the file when we bypass pagecache...
2835 */
2836 if (ext3_should_journal_data(path.dentry->d_inode)) {
2837 /*
2838 * We don't need to lock updates but journal_flush() could
2839 * otherwise be livelocked...
2840 */
2841 journal_lock_updates(EXT3_SB(sb)->s_journal);
2842 err = journal_flush(EXT3_SB(sb)->s_journal);
2843 journal_unlock_updates(EXT3_SB(sb)->s_journal);
2844 if (err) {
2845 path_put(&path);
2846 return err;
2847 }
2848 }
2849
2850 err = vfs_quota_on_path(sb, type, format_id, &path);
2851 path_put(&path);
2852 return err;
2853 }
2854
2855 /* Read data from quotafile - avoid pagecache and such because we cannot afford
2856 * acquiring the locks... As quota files are never truncated and quota code
2857 * itself serializes the operations (and noone else should touch the files)
2858 * we don't have to be afraid of races */
2859 static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
2860 size_t len, loff_t off)
2861 {
2862 struct inode *inode = sb_dqopt(sb)->files[type];
2863 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2864 int err = 0;
2865 int offset = off & (sb->s_blocksize - 1);
2866 int tocopy;
2867 size_t toread;
2868 struct buffer_head *bh;
2869 loff_t i_size = i_size_read(inode);
2870
2871 if (off > i_size)
2872 return 0;
2873 if (off+len > i_size)
2874 len = i_size-off;
2875 toread = len;
2876 while (toread > 0) {
2877 tocopy = sb->s_blocksize - offset < toread ?
2878 sb->s_blocksize - offset : toread;
2879 bh = ext3_bread(NULL, inode, blk, 0, &err);
2880 if (err)
2881 return err;
2882 if (!bh) /* A hole? */
2883 memset(data, 0, tocopy);
2884 else
2885 memcpy(data, bh->b_data+offset, tocopy);
2886 brelse(bh);
2887 offset = 0;
2888 toread -= tocopy;
2889 data += tocopy;
2890 blk++;
2891 }
2892 return len;
2893 }
2894
2895 /* Write to quotafile (we know the transaction is already started and has
2896 * enough credits) */
2897 static ssize_t ext3_quota_write(struct super_block *sb, int type,
2898 const char *data, size_t len, loff_t off)
2899 {
2900 struct inode *inode = sb_dqopt(sb)->files[type];
2901 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2902 int err = 0;
2903 int offset = off & (sb->s_blocksize - 1);
2904 int tocopy;
2905 int journal_quota = EXT3_SB(sb)->s_qf_names[type] != NULL;
2906 size_t towrite = len;
2907 struct buffer_head *bh;
2908 handle_t *handle = journal_current_handle();
2909
2910 if (!handle) {
2911 printk(KERN_WARNING "EXT3-fs: Quota write (off=%Lu, len=%Lu)"
2912 " cancelled because transaction is not started.\n",
2913 (unsigned long long)off, (unsigned long long)len);
2914 return -EIO;
2915 }
2916 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
2917 while (towrite > 0) {
2918 tocopy = sb->s_blocksize - offset < towrite ?
2919 sb->s_blocksize - offset : towrite;
2920 bh = ext3_bread(handle, inode, blk, 1, &err);
2921 if (!bh)
2922 goto out;
2923 if (journal_quota) {
2924 err = ext3_journal_get_write_access(handle, bh);
2925 if (err) {
2926 brelse(bh);
2927 goto out;
2928 }
2929 }
2930 lock_buffer(bh);
2931 memcpy(bh->b_data+offset, data, tocopy);
2932 flush_dcache_page(bh->b_page);
2933 unlock_buffer(bh);
2934 if (journal_quota)
2935 err = ext3_journal_dirty_metadata(handle, bh);
2936 else {
2937 /* Always do at least ordered writes for quotas */
2938 err = ext3_journal_dirty_data(handle, bh);
2939 mark_buffer_dirty(bh);
2940 }
2941 brelse(bh);
2942 if (err)
2943 goto out;
2944 offset = 0;
2945 towrite -= tocopy;
2946 data += tocopy;
2947 blk++;
2948 }
2949 out:
2950 if (len == towrite) {
2951 mutex_unlock(&inode->i_mutex);
2952 return err;
2953 }
2954 if (inode->i_size < off+len-towrite) {
2955 i_size_write(inode, off+len-towrite);
2956 EXT3_I(inode)->i_disksize = inode->i_size;
2957 }
2958 inode->i_version++;
2959 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
2960 ext3_mark_inode_dirty(handle, inode);
2961 mutex_unlock(&inode->i_mutex);
2962 return len - towrite;
2963 }
2964
2965 #endif
2966
2967 static int ext3_get_sb(struct file_system_type *fs_type,
2968 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
2969 {
2970 return get_sb_bdev(fs_type, flags, dev_name, data, ext3_fill_super, mnt);
2971 }
2972
2973 static struct file_system_type ext3_fs_type = {
2974 .owner = THIS_MODULE,
2975 .name = "ext3",
2976 .get_sb = ext3_get_sb,
2977 .kill_sb = kill_block_super,
2978 .fs_flags = FS_REQUIRES_DEV,
2979 };
2980
2981 static int __init init_ext3_fs(void)
2982 {
2983 int err = init_ext3_xattr();
2984 if (err)
2985 return err;
2986 err = init_inodecache();
2987 if (err)
2988 goto out1;
2989 err = register_filesystem(&ext3_fs_type);
2990 if (err)
2991 goto out;
2992 return 0;
2993 out:
2994 destroy_inodecache();
2995 out1:
2996 exit_ext3_xattr();
2997 return err;
2998 }
2999
3000 static void __exit exit_ext3_fs(void)
3001 {
3002 unregister_filesystem(&ext3_fs_type);
3003 destroy_inodecache();
3004 exit_ext3_xattr();
3005 }
3006
3007 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
3008 MODULE_DESCRIPTION("Second Extended Filesystem with journaling extensions");
3009 MODULE_LICENSE("GPL");
3010 module_init(init_ext3_fs)
3011 module_exit(exit_ext3_fs)
Linus' kernel tree