nfs41: implement DESTROY_CLIENTID operation
[linux-btrfs-devel.git] / fs / ext3 / super.c
blob7beb69ae0015996140fe916712f4b1e13860faa0
1 /*
2 * linux/fs/ext3/super.c
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)
9 * from
11 * linux/fs/minix/inode.c
13 * Copyright (C) 1991, 1992 Linus Torvalds
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
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/buffer_head.h>
31 #include <linux/exportfs.h>
32 #include <linux/vfs.h>
33 #include <linux/random.h>
34 #include <linux/mount.h>
35 #include <linux/namei.h>
36 #include <linux/quotaops.h>
37 #include <linux/seq_file.h>
38 #include <linux/log2.h>
39 #include <linux/cleancache.h>
41 #include <asm/uaccess.h>
43 #include "xattr.h"
44 #include "acl.h"
45 #include "namei.h"
47 #define CREATE_TRACE_POINTS
48 #include <trace/events/ext3.h>
50 #ifdef CONFIG_EXT3_DEFAULTS_TO_ORDERED
51 #define EXT3_MOUNT_DEFAULT_DATA_MODE EXT3_MOUNT_ORDERED_DATA
52 #else
53 #define EXT3_MOUNT_DEFAULT_DATA_MODE EXT3_MOUNT_WRITEBACK_DATA
54 #endif
56 static int ext3_load_journal(struct super_block *, struct ext3_super_block *,
57 unsigned long journal_devnum);
58 static int ext3_create_journal(struct super_block *, struct ext3_super_block *,
59 unsigned int);
60 static int ext3_commit_super(struct super_block *sb,
61 struct ext3_super_block *es,
62 int sync);
63 static void ext3_mark_recovery_complete(struct super_block * sb,
64 struct ext3_super_block * es);
65 static void ext3_clear_journal_err(struct super_block * sb,
66 struct ext3_super_block * es);
67 static int ext3_sync_fs(struct super_block *sb, int wait);
68 static const char *ext3_decode_error(struct super_block * sb, int errno,
69 char nbuf[16]);
70 static int ext3_remount (struct super_block * sb, int * flags, char * data);
71 static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf);
72 static int ext3_unfreeze(struct super_block *sb);
73 static int ext3_freeze(struct super_block *sb);
76 * Wrappers for journal_start/end.
78 * The only special thing we need to do here is to make sure that all
79 * journal_end calls result in the superblock being marked dirty, so
80 * that sync() will call the filesystem's write_super callback if
81 * appropriate.
83 handle_t *ext3_journal_start_sb(struct super_block *sb, int nblocks)
85 journal_t *journal;
87 if (sb->s_flags & MS_RDONLY)
88 return ERR_PTR(-EROFS);
90 /* Special case here: if the journal has aborted behind our
91 * backs (eg. EIO in the commit thread), then we still need to
92 * take the FS itself readonly cleanly. */
93 journal = EXT3_SB(sb)->s_journal;
94 if (is_journal_aborted(journal)) {
95 ext3_abort(sb, __func__,
96 "Detected aborted journal");
97 return ERR_PTR(-EROFS);
100 return journal_start(journal, nblocks);
104 * The only special thing we need to do here is to make sure that all
105 * journal_stop calls result in the superblock being marked dirty, so
106 * that sync() will call the filesystem's write_super callback if
107 * appropriate.
109 int __ext3_journal_stop(const char *where, handle_t *handle)
111 struct super_block *sb;
112 int err;
113 int rc;
115 sb = handle->h_transaction->t_journal->j_private;
116 err = handle->h_err;
117 rc = journal_stop(handle);
119 if (!err)
120 err = rc;
121 if (err)
122 __ext3_std_error(sb, where, err);
123 return err;
126 void ext3_journal_abort_handle(const char *caller, const char *err_fn,
127 struct buffer_head *bh, handle_t *handle, int err)
129 char nbuf[16];
130 const char *errstr = ext3_decode_error(NULL, err, nbuf);
132 if (bh)
133 BUFFER_TRACE(bh, "abort");
135 if (!handle->h_err)
136 handle->h_err = err;
138 if (is_handle_aborted(handle))
139 return;
141 printk(KERN_ERR "EXT3-fs: %s: aborting transaction: %s in %s\n",
142 caller, errstr, err_fn);
144 journal_abort_handle(handle);
147 void ext3_msg(struct super_block *sb, const char *prefix,
148 const char *fmt, ...)
150 struct va_format vaf;
151 va_list args;
153 va_start(args, fmt);
155 vaf.fmt = fmt;
156 vaf.va = &args;
158 printk("%sEXT3-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
160 va_end(args);
163 /* Deal with the reporting of failure conditions on a filesystem such as
164 * inconsistencies detected or read IO failures.
166 * On ext2, we can store the error state of the filesystem in the
167 * superblock. That is not possible on ext3, because we may have other
168 * write ordering constraints on the superblock which prevent us from
169 * writing it out straight away; and given that the journal is about to
170 * be aborted, we can't rely on the current, or future, transactions to
171 * write out the superblock safely.
173 * We'll just use the journal_abort() error code to record an error in
174 * the journal instead. On recovery, the journal will complain about
175 * that error until we've noted it down and cleared it.
178 static void ext3_handle_error(struct super_block *sb)
180 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
182 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
183 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
185 if (sb->s_flags & MS_RDONLY)
186 return;
188 if (!test_opt (sb, ERRORS_CONT)) {
189 journal_t *journal = EXT3_SB(sb)->s_journal;
191 set_opt(EXT3_SB(sb)->s_mount_opt, ABORT);
192 if (journal)
193 journal_abort(journal, -EIO);
195 if (test_opt (sb, ERRORS_RO)) {
196 ext3_msg(sb, KERN_CRIT,
197 "error: remounting filesystem read-only");
198 sb->s_flags |= MS_RDONLY;
200 ext3_commit_super(sb, es, 1);
201 if (test_opt(sb, ERRORS_PANIC))
202 panic("EXT3-fs (%s): panic forced after error\n",
203 sb->s_id);
206 void ext3_error(struct super_block *sb, const char *function,
207 const char *fmt, ...)
209 struct va_format vaf;
210 va_list args;
212 va_start(args, fmt);
214 vaf.fmt = fmt;
215 vaf.va = &args;
217 printk(KERN_CRIT "EXT3-fs error (device %s): %s: %pV\n",
218 sb->s_id, function, &vaf);
220 va_end(args);
222 ext3_handle_error(sb);
225 static const char *ext3_decode_error(struct super_block * sb, int errno,
226 char nbuf[16])
228 char *errstr = NULL;
230 switch (errno) {
231 case -EIO:
232 errstr = "IO failure";
233 break;
234 case -ENOMEM:
235 errstr = "Out of memory";
236 break;
237 case -EROFS:
238 if (!sb || EXT3_SB(sb)->s_journal->j_flags & JFS_ABORT)
239 errstr = "Journal has aborted";
240 else
241 errstr = "Readonly filesystem";
242 break;
243 default:
244 /* If the caller passed in an extra buffer for unknown
245 * errors, textualise them now. Else we just return
246 * NULL. */
247 if (nbuf) {
248 /* Check for truncated error codes... */
249 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
250 errstr = nbuf;
252 break;
255 return errstr;
258 /* __ext3_std_error decodes expected errors from journaling functions
259 * automatically and invokes the appropriate error response. */
261 void __ext3_std_error (struct super_block * sb, const char * function,
262 int errno)
264 char nbuf[16];
265 const char *errstr;
267 /* Special case: if the error is EROFS, and we're not already
268 * inside a transaction, then there's really no point in logging
269 * an error. */
270 if (errno == -EROFS && journal_current_handle() == NULL &&
271 (sb->s_flags & MS_RDONLY))
272 return;
274 errstr = ext3_decode_error(sb, errno, nbuf);
275 ext3_msg(sb, KERN_CRIT, "error in %s: %s", function, errstr);
277 ext3_handle_error(sb);
281 * ext3_abort is a much stronger failure handler than ext3_error. The
282 * abort function may be used to deal with unrecoverable failures such
283 * as journal IO errors or ENOMEM at a critical moment in log management.
285 * We unconditionally force the filesystem into an ABORT|READONLY state,
286 * unless the error response on the fs has been set to panic in which
287 * case we take the easy way out and panic immediately.
290 void ext3_abort(struct super_block *sb, const char *function,
291 const char *fmt, ...)
293 struct va_format vaf;
294 va_list args;
296 va_start(args, fmt);
298 vaf.fmt = fmt;
299 vaf.va = &args;
301 printk(KERN_CRIT "EXT3-fs (%s): error: %s: %pV\n",
302 sb->s_id, function, &vaf);
304 va_end(args);
306 if (test_opt(sb, ERRORS_PANIC))
307 panic("EXT3-fs: panic from previous error\n");
309 if (sb->s_flags & MS_RDONLY)
310 return;
312 ext3_msg(sb, KERN_CRIT,
313 "error: remounting filesystem read-only");
314 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
315 sb->s_flags |= MS_RDONLY;
316 set_opt(EXT3_SB(sb)->s_mount_opt, ABORT);
317 if (EXT3_SB(sb)->s_journal)
318 journal_abort(EXT3_SB(sb)->s_journal, -EIO);
321 void ext3_warning(struct super_block *sb, const char *function,
322 const char *fmt, ...)
324 struct va_format vaf;
325 va_list args;
327 va_start(args, fmt);
329 vaf.fmt = fmt;
330 vaf.va = &args;
332 printk(KERN_WARNING "EXT3-fs (%s): warning: %s: %pV\n",
333 sb->s_id, function, &vaf);
335 va_end(args);
338 void ext3_update_dynamic_rev(struct super_block *sb)
340 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
342 if (le32_to_cpu(es->s_rev_level) > EXT3_GOOD_OLD_REV)
343 return;
345 ext3_msg(sb, KERN_WARNING,
346 "warning: updating to rev %d because of "
347 "new feature flag, running e2fsck is recommended",
348 EXT3_DYNAMIC_REV);
350 es->s_first_ino = cpu_to_le32(EXT3_GOOD_OLD_FIRST_INO);
351 es->s_inode_size = cpu_to_le16(EXT3_GOOD_OLD_INODE_SIZE);
352 es->s_rev_level = cpu_to_le32(EXT3_DYNAMIC_REV);
353 /* leave es->s_feature_*compat flags alone */
354 /* es->s_uuid will be set by e2fsck if empty */
357 * The rest of the superblock fields should be zero, and if not it
358 * means they are likely already in use, so leave them alone. We
359 * can leave it up to e2fsck to clean up any inconsistencies there.
364 * Open the external journal device
366 static struct block_device *ext3_blkdev_get(dev_t dev, struct super_block *sb)
368 struct block_device *bdev;
369 char b[BDEVNAME_SIZE];
371 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
372 if (IS_ERR(bdev))
373 goto fail;
374 return bdev;
376 fail:
377 ext3_msg(sb, "error: failed to open journal device %s: %ld",
378 __bdevname(dev, b), PTR_ERR(bdev));
380 return NULL;
384 * Release the journal device
386 static int ext3_blkdev_put(struct block_device *bdev)
388 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
391 static int ext3_blkdev_remove(struct ext3_sb_info *sbi)
393 struct block_device *bdev;
394 int ret = -ENODEV;
396 bdev = sbi->journal_bdev;
397 if (bdev) {
398 ret = ext3_blkdev_put(bdev);
399 sbi->journal_bdev = NULL;
401 return ret;
404 static inline struct inode *orphan_list_entry(struct list_head *l)
406 return &list_entry(l, struct ext3_inode_info, i_orphan)->vfs_inode;
409 static void dump_orphan_list(struct super_block *sb, struct ext3_sb_info *sbi)
411 struct list_head *l;
413 ext3_msg(sb, KERN_ERR, "error: sb orphan head is %d",
414 le32_to_cpu(sbi->s_es->s_last_orphan));
416 ext3_msg(sb, KERN_ERR, "sb_info orphan list:");
417 list_for_each(l, &sbi->s_orphan) {
418 struct inode *inode = orphan_list_entry(l);
419 ext3_msg(sb, KERN_ERR, " "
420 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
421 inode->i_sb->s_id, inode->i_ino, inode,
422 inode->i_mode, inode->i_nlink,
423 NEXT_ORPHAN(inode));
427 static void ext3_put_super (struct super_block * sb)
429 struct ext3_sb_info *sbi = EXT3_SB(sb);
430 struct ext3_super_block *es = sbi->s_es;
431 int i, err;
433 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
434 ext3_xattr_put_super(sb);
435 err = journal_destroy(sbi->s_journal);
436 sbi->s_journal = NULL;
437 if (err < 0)
438 ext3_abort(sb, __func__, "Couldn't clean up the journal");
440 if (!(sb->s_flags & MS_RDONLY)) {
441 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
442 es->s_state = cpu_to_le16(sbi->s_mount_state);
443 BUFFER_TRACE(sbi->s_sbh, "marking dirty");
444 mark_buffer_dirty(sbi->s_sbh);
445 ext3_commit_super(sb, es, 1);
448 for (i = 0; i < sbi->s_gdb_count; i++)
449 brelse(sbi->s_group_desc[i]);
450 kfree(sbi->s_group_desc);
451 percpu_counter_destroy(&sbi->s_freeblocks_counter);
452 percpu_counter_destroy(&sbi->s_freeinodes_counter);
453 percpu_counter_destroy(&sbi->s_dirs_counter);
454 brelse(sbi->s_sbh);
455 #ifdef CONFIG_QUOTA
456 for (i = 0; i < MAXQUOTAS; i++)
457 kfree(sbi->s_qf_names[i]);
458 #endif
460 /* Debugging code just in case the in-memory inode orphan list
461 * isn't empty. The on-disk one can be non-empty if we've
462 * detected an error and taken the fs readonly, but the
463 * in-memory list had better be clean by this point. */
464 if (!list_empty(&sbi->s_orphan))
465 dump_orphan_list(sb, sbi);
466 J_ASSERT(list_empty(&sbi->s_orphan));
468 invalidate_bdev(sb->s_bdev);
469 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
471 * Invalidate the journal device's buffers. We don't want them
472 * floating about in memory - the physical journal device may
473 * hotswapped, and it breaks the `ro-after' testing code.
475 sync_blockdev(sbi->journal_bdev);
476 invalidate_bdev(sbi->journal_bdev);
477 ext3_blkdev_remove(sbi);
479 sb->s_fs_info = NULL;
480 kfree(sbi->s_blockgroup_lock);
481 kfree(sbi);
484 static struct kmem_cache *ext3_inode_cachep;
487 * Called inside transaction, so use GFP_NOFS
489 static struct inode *ext3_alloc_inode(struct super_block *sb)
491 struct ext3_inode_info *ei;
493 ei = kmem_cache_alloc(ext3_inode_cachep, GFP_NOFS);
494 if (!ei)
495 return NULL;
496 ei->i_block_alloc_info = NULL;
497 ei->vfs_inode.i_version = 1;
498 atomic_set(&ei->i_datasync_tid, 0);
499 atomic_set(&ei->i_sync_tid, 0);
500 return &ei->vfs_inode;
503 static int ext3_drop_inode(struct inode *inode)
505 int drop = generic_drop_inode(inode);
507 trace_ext3_drop_inode(inode, drop);
508 return drop;
511 static void ext3_i_callback(struct rcu_head *head)
513 struct inode *inode = container_of(head, struct inode, i_rcu);
514 INIT_LIST_HEAD(&inode->i_dentry);
515 kmem_cache_free(ext3_inode_cachep, EXT3_I(inode));
518 static void ext3_destroy_inode(struct inode *inode)
520 if (!list_empty(&(EXT3_I(inode)->i_orphan))) {
521 printk("EXT3 Inode %p: orphan list check failed!\n",
522 EXT3_I(inode));
523 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
524 EXT3_I(inode), sizeof(struct ext3_inode_info),
525 false);
526 dump_stack();
528 call_rcu(&inode->i_rcu, ext3_i_callback);
531 static void init_once(void *foo)
533 struct ext3_inode_info *ei = (struct ext3_inode_info *) foo;
535 INIT_LIST_HEAD(&ei->i_orphan);
536 #ifdef CONFIG_EXT3_FS_XATTR
537 init_rwsem(&ei->xattr_sem);
538 #endif
539 mutex_init(&ei->truncate_mutex);
540 inode_init_once(&ei->vfs_inode);
543 static int init_inodecache(void)
545 ext3_inode_cachep = kmem_cache_create("ext3_inode_cache",
546 sizeof(struct ext3_inode_info),
547 0, (SLAB_RECLAIM_ACCOUNT|
548 SLAB_MEM_SPREAD),
549 init_once);
550 if (ext3_inode_cachep == NULL)
551 return -ENOMEM;
552 return 0;
555 static void destroy_inodecache(void)
557 kmem_cache_destroy(ext3_inode_cachep);
560 static inline void ext3_show_quota_options(struct seq_file *seq, struct super_block *sb)
562 #if defined(CONFIG_QUOTA)
563 struct ext3_sb_info *sbi = EXT3_SB(sb);
565 if (sbi->s_jquota_fmt) {
566 char *fmtname = "";
568 switch (sbi->s_jquota_fmt) {
569 case QFMT_VFS_OLD:
570 fmtname = "vfsold";
571 break;
572 case QFMT_VFS_V0:
573 fmtname = "vfsv0";
574 break;
575 case QFMT_VFS_V1:
576 fmtname = "vfsv1";
577 break;
579 seq_printf(seq, ",jqfmt=%s", fmtname);
582 if (sbi->s_qf_names[USRQUOTA])
583 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
585 if (sbi->s_qf_names[GRPQUOTA])
586 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
588 if (test_opt(sb, USRQUOTA))
589 seq_puts(seq, ",usrquota");
591 if (test_opt(sb, GRPQUOTA))
592 seq_puts(seq, ",grpquota");
593 #endif
596 static char *data_mode_string(unsigned long mode)
598 switch (mode) {
599 case EXT3_MOUNT_JOURNAL_DATA:
600 return "journal";
601 case EXT3_MOUNT_ORDERED_DATA:
602 return "ordered";
603 case EXT3_MOUNT_WRITEBACK_DATA:
604 return "writeback";
606 return "unknown";
610 * Show an option if
611 * - it's set to a non-default value OR
612 * - if the per-sb default is different from the global default
614 static int ext3_show_options(struct seq_file *seq, struct vfsmount *vfs)
616 struct super_block *sb = vfs->mnt_sb;
617 struct ext3_sb_info *sbi = EXT3_SB(sb);
618 struct ext3_super_block *es = sbi->s_es;
619 unsigned long def_mount_opts;
621 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
623 if (sbi->s_sb_block != 1)
624 seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
625 if (test_opt(sb, MINIX_DF))
626 seq_puts(seq, ",minixdf");
627 if (test_opt(sb, GRPID))
628 seq_puts(seq, ",grpid");
629 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT3_DEFM_BSDGROUPS))
630 seq_puts(seq, ",nogrpid");
631 if (sbi->s_resuid != EXT3_DEF_RESUID ||
632 le16_to_cpu(es->s_def_resuid) != EXT3_DEF_RESUID) {
633 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
635 if (sbi->s_resgid != EXT3_DEF_RESGID ||
636 le16_to_cpu(es->s_def_resgid) != EXT3_DEF_RESGID) {
637 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
639 if (test_opt(sb, ERRORS_RO)) {
640 int def_errors = le16_to_cpu(es->s_errors);
642 if (def_errors == EXT3_ERRORS_PANIC ||
643 def_errors == EXT3_ERRORS_CONTINUE) {
644 seq_puts(seq, ",errors=remount-ro");
647 if (test_opt(sb, ERRORS_CONT))
648 seq_puts(seq, ",errors=continue");
649 if (test_opt(sb, ERRORS_PANIC))
650 seq_puts(seq, ",errors=panic");
651 if (test_opt(sb, NO_UID32))
652 seq_puts(seq, ",nouid32");
653 if (test_opt(sb, DEBUG))
654 seq_puts(seq, ",debug");
655 if (test_opt(sb, OLDALLOC))
656 seq_puts(seq, ",oldalloc");
657 #ifdef CONFIG_EXT3_FS_XATTR
658 if (test_opt(sb, XATTR_USER))
659 seq_puts(seq, ",user_xattr");
660 if (!test_opt(sb, XATTR_USER) &&
661 (def_mount_opts & EXT3_DEFM_XATTR_USER)) {
662 seq_puts(seq, ",nouser_xattr");
664 #endif
665 #ifdef CONFIG_EXT3_FS_POSIX_ACL
666 if (test_opt(sb, POSIX_ACL))
667 seq_puts(seq, ",acl");
668 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT3_DEFM_ACL))
669 seq_puts(seq, ",noacl");
670 #endif
671 if (!test_opt(sb, RESERVATION))
672 seq_puts(seq, ",noreservation");
673 if (sbi->s_commit_interval) {
674 seq_printf(seq, ",commit=%u",
675 (unsigned) (sbi->s_commit_interval / HZ));
679 * Always display barrier state so it's clear what the status is.
681 seq_puts(seq, ",barrier=");
682 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
683 seq_printf(seq, ",data=%s", data_mode_string(test_opt(sb, DATA_FLAGS)));
684 if (test_opt(sb, DATA_ERR_ABORT))
685 seq_puts(seq, ",data_err=abort");
687 if (test_opt(sb, NOLOAD))
688 seq_puts(seq, ",norecovery");
690 ext3_show_quota_options(seq, sb);
692 return 0;
696 static struct inode *ext3_nfs_get_inode(struct super_block *sb,
697 u64 ino, u32 generation)
699 struct inode *inode;
701 if (ino < EXT3_FIRST_INO(sb) && ino != EXT3_ROOT_INO)
702 return ERR_PTR(-ESTALE);
703 if (ino > le32_to_cpu(EXT3_SB(sb)->s_es->s_inodes_count))
704 return ERR_PTR(-ESTALE);
706 /* iget isn't really right if the inode is currently unallocated!!
708 * ext3_read_inode will return a bad_inode if the inode had been
709 * deleted, so we should be safe.
711 * Currently we don't know the generation for parent directory, so
712 * a generation of 0 means "accept any"
714 inode = ext3_iget(sb, ino);
715 if (IS_ERR(inode))
716 return ERR_CAST(inode);
717 if (generation && inode->i_generation != generation) {
718 iput(inode);
719 return ERR_PTR(-ESTALE);
722 return inode;
725 static struct dentry *ext3_fh_to_dentry(struct super_block *sb, struct fid *fid,
726 int fh_len, int fh_type)
728 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
729 ext3_nfs_get_inode);
732 static struct dentry *ext3_fh_to_parent(struct super_block *sb, struct fid *fid,
733 int fh_len, int fh_type)
735 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
736 ext3_nfs_get_inode);
740 * Try to release metadata pages (indirect blocks, directories) which are
741 * mapped via the block device. Since these pages could have journal heads
742 * which would prevent try_to_free_buffers() from freeing them, we must use
743 * jbd layer's try_to_free_buffers() function to release them.
745 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
746 gfp_t wait)
748 journal_t *journal = EXT3_SB(sb)->s_journal;
750 WARN_ON(PageChecked(page));
751 if (!page_has_buffers(page))
752 return 0;
753 if (journal)
754 return journal_try_to_free_buffers(journal, page,
755 wait & ~__GFP_WAIT);
756 return try_to_free_buffers(page);
759 #ifdef CONFIG_QUOTA
760 #define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
761 #define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
763 static int ext3_write_dquot(struct dquot *dquot);
764 static int ext3_acquire_dquot(struct dquot *dquot);
765 static int ext3_release_dquot(struct dquot *dquot);
766 static int ext3_mark_dquot_dirty(struct dquot *dquot);
767 static int ext3_write_info(struct super_block *sb, int type);
768 static int ext3_quota_on(struct super_block *sb, int type, int format_id,
769 struct path *path);
770 static int ext3_quota_on_mount(struct super_block *sb, int type);
771 static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
772 size_t len, loff_t off);
773 static ssize_t ext3_quota_write(struct super_block *sb, int type,
774 const char *data, size_t len, loff_t off);
776 static const struct dquot_operations ext3_quota_operations = {
777 .write_dquot = ext3_write_dquot,
778 .acquire_dquot = ext3_acquire_dquot,
779 .release_dquot = ext3_release_dquot,
780 .mark_dirty = ext3_mark_dquot_dirty,
781 .write_info = ext3_write_info,
782 .alloc_dquot = dquot_alloc,
783 .destroy_dquot = dquot_destroy,
786 static const struct quotactl_ops ext3_qctl_operations = {
787 .quota_on = ext3_quota_on,
788 .quota_off = dquot_quota_off,
789 .quota_sync = dquot_quota_sync,
790 .get_info = dquot_get_dqinfo,
791 .set_info = dquot_set_dqinfo,
792 .get_dqblk = dquot_get_dqblk,
793 .set_dqblk = dquot_set_dqblk
795 #endif
797 static const struct super_operations ext3_sops = {
798 .alloc_inode = ext3_alloc_inode,
799 .destroy_inode = ext3_destroy_inode,
800 .write_inode = ext3_write_inode,
801 .dirty_inode = ext3_dirty_inode,
802 .drop_inode = ext3_drop_inode,
803 .evict_inode = ext3_evict_inode,
804 .put_super = ext3_put_super,
805 .sync_fs = ext3_sync_fs,
806 .freeze_fs = ext3_freeze,
807 .unfreeze_fs = ext3_unfreeze,
808 .statfs = ext3_statfs,
809 .remount_fs = ext3_remount,
810 .show_options = ext3_show_options,
811 #ifdef CONFIG_QUOTA
812 .quota_read = ext3_quota_read,
813 .quota_write = ext3_quota_write,
814 #endif
815 .bdev_try_to_free_page = bdev_try_to_free_page,
818 static const struct export_operations ext3_export_ops = {
819 .fh_to_dentry = ext3_fh_to_dentry,
820 .fh_to_parent = ext3_fh_to_parent,
821 .get_parent = ext3_get_parent,
824 enum {
825 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
826 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
827 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
828 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
829 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
830 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
831 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
832 Opt_data_err_abort, Opt_data_err_ignore,
833 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
834 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
835 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
836 Opt_resize, Opt_usrquota, Opt_grpquota
839 static const match_table_t tokens = {
840 {Opt_bsd_df, "bsddf"},
841 {Opt_minix_df, "minixdf"},
842 {Opt_grpid, "grpid"},
843 {Opt_grpid, "bsdgroups"},
844 {Opt_nogrpid, "nogrpid"},
845 {Opt_nogrpid, "sysvgroups"},
846 {Opt_resgid, "resgid=%u"},
847 {Opt_resuid, "resuid=%u"},
848 {Opt_sb, "sb=%u"},
849 {Opt_err_cont, "errors=continue"},
850 {Opt_err_panic, "errors=panic"},
851 {Opt_err_ro, "errors=remount-ro"},
852 {Opt_nouid32, "nouid32"},
853 {Opt_nocheck, "nocheck"},
854 {Opt_nocheck, "check=none"},
855 {Opt_debug, "debug"},
856 {Opt_oldalloc, "oldalloc"},
857 {Opt_orlov, "orlov"},
858 {Opt_user_xattr, "user_xattr"},
859 {Opt_nouser_xattr, "nouser_xattr"},
860 {Opt_acl, "acl"},
861 {Opt_noacl, "noacl"},
862 {Opt_reservation, "reservation"},
863 {Opt_noreservation, "noreservation"},
864 {Opt_noload, "noload"},
865 {Opt_noload, "norecovery"},
866 {Opt_nobh, "nobh"},
867 {Opt_bh, "bh"},
868 {Opt_commit, "commit=%u"},
869 {Opt_journal_update, "journal=update"},
870 {Opt_journal_inum, "journal=%u"},
871 {Opt_journal_dev, "journal_dev=%u"},
872 {Opt_abort, "abort"},
873 {Opt_data_journal, "data=journal"},
874 {Opt_data_ordered, "data=ordered"},
875 {Opt_data_writeback, "data=writeback"},
876 {Opt_data_err_abort, "data_err=abort"},
877 {Opt_data_err_ignore, "data_err=ignore"},
878 {Opt_offusrjquota, "usrjquota="},
879 {Opt_usrjquota, "usrjquota=%s"},
880 {Opt_offgrpjquota, "grpjquota="},
881 {Opt_grpjquota, "grpjquota=%s"},
882 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
883 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
884 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
885 {Opt_grpquota, "grpquota"},
886 {Opt_noquota, "noquota"},
887 {Opt_quota, "quota"},
888 {Opt_usrquota, "usrquota"},
889 {Opt_barrier, "barrier=%u"},
890 {Opt_barrier, "barrier"},
891 {Opt_nobarrier, "nobarrier"},
892 {Opt_resize, "resize"},
893 {Opt_err, NULL},
896 static ext3_fsblk_t get_sb_block(void **data, struct super_block *sb)
898 ext3_fsblk_t sb_block;
899 char *options = (char *) *data;
901 if (!options || strncmp(options, "sb=", 3) != 0)
902 return 1; /* Default location */
903 options += 3;
904 /*todo: use simple_strtoll with >32bit ext3 */
905 sb_block = simple_strtoul(options, &options, 0);
906 if (*options && *options != ',') {
907 ext3_msg(sb, "error: invalid sb specification: %s",
908 (char *) *data);
909 return 1;
911 if (*options == ',')
912 options++;
913 *data = (void *) options;
914 return sb_block;
917 #ifdef CONFIG_QUOTA
918 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
920 struct ext3_sb_info *sbi = EXT3_SB(sb);
921 char *qname;
923 if (sb_any_quota_loaded(sb) &&
924 !sbi->s_qf_names[qtype]) {
925 ext3_msg(sb, KERN_ERR,
926 "Cannot change journaled "
927 "quota options when quota turned on");
928 return 0;
930 qname = match_strdup(args);
931 if (!qname) {
932 ext3_msg(sb, KERN_ERR,
933 "Not enough memory for storing quotafile name");
934 return 0;
936 if (sbi->s_qf_names[qtype] &&
937 strcmp(sbi->s_qf_names[qtype], qname)) {
938 ext3_msg(sb, KERN_ERR,
939 "%s quota file already specified", QTYPE2NAME(qtype));
940 kfree(qname);
941 return 0;
943 sbi->s_qf_names[qtype] = qname;
944 if (strchr(sbi->s_qf_names[qtype], '/')) {
945 ext3_msg(sb, KERN_ERR,
946 "quotafile must be on filesystem root");
947 kfree(sbi->s_qf_names[qtype]);
948 sbi->s_qf_names[qtype] = NULL;
949 return 0;
951 set_opt(sbi->s_mount_opt, QUOTA);
952 return 1;
955 static int clear_qf_name(struct super_block *sb, int qtype) {
957 struct ext3_sb_info *sbi = EXT3_SB(sb);
959 if (sb_any_quota_loaded(sb) &&
960 sbi->s_qf_names[qtype]) {
961 ext3_msg(sb, KERN_ERR, "Cannot change journaled quota options"
962 " when quota turned on");
963 return 0;
966 * The space will be released later when all options are confirmed
967 * to be correct
969 sbi->s_qf_names[qtype] = NULL;
970 return 1;
972 #endif
974 static int parse_options (char *options, struct super_block *sb,
975 unsigned int *inum, unsigned long *journal_devnum,
976 ext3_fsblk_t *n_blocks_count, int is_remount)
978 struct ext3_sb_info *sbi = EXT3_SB(sb);
979 char * p;
980 substring_t args[MAX_OPT_ARGS];
981 int data_opt = 0;
982 int option;
983 #ifdef CONFIG_QUOTA
984 int qfmt;
985 #endif
987 if (!options)
988 return 1;
990 while ((p = strsep (&options, ",")) != NULL) {
991 int token;
992 if (!*p)
993 continue;
995 * Initialize args struct so we know whether arg was
996 * found; some options take optional arguments.
998 args[0].to = args[0].from = 0;
999 token = match_token(p, tokens, args);
1000 switch (token) {
1001 case Opt_bsd_df:
1002 clear_opt (sbi->s_mount_opt, MINIX_DF);
1003 break;
1004 case Opt_minix_df:
1005 set_opt (sbi->s_mount_opt, MINIX_DF);
1006 break;
1007 case Opt_grpid:
1008 set_opt (sbi->s_mount_opt, GRPID);
1009 break;
1010 case Opt_nogrpid:
1011 clear_opt (sbi->s_mount_opt, GRPID);
1012 break;
1013 case Opt_resuid:
1014 if (match_int(&args[0], &option))
1015 return 0;
1016 sbi->s_resuid = option;
1017 break;
1018 case Opt_resgid:
1019 if (match_int(&args[0], &option))
1020 return 0;
1021 sbi->s_resgid = option;
1022 break;
1023 case Opt_sb:
1024 /* handled by get_sb_block() instead of here */
1025 /* *sb_block = match_int(&args[0]); */
1026 break;
1027 case Opt_err_panic:
1028 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1029 clear_opt (sbi->s_mount_opt, ERRORS_RO);
1030 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
1031 break;
1032 case Opt_err_ro:
1033 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1034 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1035 set_opt (sbi->s_mount_opt, ERRORS_RO);
1036 break;
1037 case Opt_err_cont:
1038 clear_opt (sbi->s_mount_opt, ERRORS_RO);
1039 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1040 set_opt (sbi->s_mount_opt, ERRORS_CONT);
1041 break;
1042 case Opt_nouid32:
1043 set_opt (sbi->s_mount_opt, NO_UID32);
1044 break;
1045 case Opt_nocheck:
1046 clear_opt (sbi->s_mount_opt, CHECK);
1047 break;
1048 case Opt_debug:
1049 set_opt (sbi->s_mount_opt, DEBUG);
1050 break;
1051 case Opt_oldalloc:
1052 set_opt (sbi->s_mount_opt, OLDALLOC);
1053 break;
1054 case Opt_orlov:
1055 clear_opt (sbi->s_mount_opt, OLDALLOC);
1056 break;
1057 #ifdef CONFIG_EXT3_FS_XATTR
1058 case Opt_user_xattr:
1059 set_opt (sbi->s_mount_opt, XATTR_USER);
1060 break;
1061 case Opt_nouser_xattr:
1062 clear_opt (sbi->s_mount_opt, XATTR_USER);
1063 break;
1064 #else
1065 case Opt_user_xattr:
1066 case Opt_nouser_xattr:
1067 ext3_msg(sb, KERN_INFO,
1068 "(no)user_xattr options not supported");
1069 break;
1070 #endif
1071 #ifdef CONFIG_EXT3_FS_POSIX_ACL
1072 case Opt_acl:
1073 set_opt(sbi->s_mount_opt, POSIX_ACL);
1074 break;
1075 case Opt_noacl:
1076 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1077 break;
1078 #else
1079 case Opt_acl:
1080 case Opt_noacl:
1081 ext3_msg(sb, KERN_INFO,
1082 "(no)acl options not supported");
1083 break;
1084 #endif
1085 case Opt_reservation:
1086 set_opt(sbi->s_mount_opt, RESERVATION);
1087 break;
1088 case Opt_noreservation:
1089 clear_opt(sbi->s_mount_opt, RESERVATION);
1090 break;
1091 case Opt_journal_update:
1092 /* @@@ FIXME */
1093 /* Eventually we will want to be able to create
1094 a journal file here. For now, only allow the
1095 user to specify an existing inode to be the
1096 journal file. */
1097 if (is_remount) {
1098 ext3_msg(sb, KERN_ERR, "error: cannot specify "
1099 "journal on remount");
1100 return 0;
1102 set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
1103 break;
1104 case Opt_journal_inum:
1105 if (is_remount) {
1106 ext3_msg(sb, KERN_ERR, "error: cannot specify "
1107 "journal on remount");
1108 return 0;
1110 if (match_int(&args[0], &option))
1111 return 0;
1112 *inum = option;
1113 break;
1114 case Opt_journal_dev:
1115 if (is_remount) {
1116 ext3_msg(sb, KERN_ERR, "error: cannot specify "
1117 "journal on remount");
1118 return 0;
1120 if (match_int(&args[0], &option))
1121 return 0;
1122 *journal_devnum = option;
1123 break;
1124 case Opt_noload:
1125 set_opt (sbi->s_mount_opt, NOLOAD);
1126 break;
1127 case Opt_commit:
1128 if (match_int(&args[0], &option))
1129 return 0;
1130 if (option < 0)
1131 return 0;
1132 if (option == 0)
1133 option = JBD_DEFAULT_MAX_COMMIT_AGE;
1134 sbi->s_commit_interval = HZ * option;
1135 break;
1136 case Opt_data_journal:
1137 data_opt = EXT3_MOUNT_JOURNAL_DATA;
1138 goto datacheck;
1139 case Opt_data_ordered:
1140 data_opt = EXT3_MOUNT_ORDERED_DATA;
1141 goto datacheck;
1142 case Opt_data_writeback:
1143 data_opt = EXT3_MOUNT_WRITEBACK_DATA;
1144 datacheck:
1145 if (is_remount) {
1146 if (test_opt(sb, DATA_FLAGS) == data_opt)
1147 break;
1148 ext3_msg(sb, KERN_ERR,
1149 "error: cannot change "
1150 "data mode on remount. The filesystem "
1151 "is mounted in data=%s mode and you "
1152 "try to remount it in data=%s mode.",
1153 data_mode_string(test_opt(sb,
1154 DATA_FLAGS)),
1155 data_mode_string(data_opt));
1156 return 0;
1157 } else {
1158 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
1159 sbi->s_mount_opt |= data_opt;
1161 break;
1162 case Opt_data_err_abort:
1163 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1164 break;
1165 case Opt_data_err_ignore:
1166 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1167 break;
1168 #ifdef CONFIG_QUOTA
1169 case Opt_usrjquota:
1170 if (!set_qf_name(sb, USRQUOTA, &args[0]))
1171 return 0;
1172 break;
1173 case Opt_grpjquota:
1174 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
1175 return 0;
1176 break;
1177 case Opt_offusrjquota:
1178 if (!clear_qf_name(sb, USRQUOTA))
1179 return 0;
1180 break;
1181 case Opt_offgrpjquota:
1182 if (!clear_qf_name(sb, GRPQUOTA))
1183 return 0;
1184 break;
1185 case Opt_jqfmt_vfsold:
1186 qfmt = QFMT_VFS_OLD;
1187 goto set_qf_format;
1188 case Opt_jqfmt_vfsv0:
1189 qfmt = QFMT_VFS_V0;
1190 goto set_qf_format;
1191 case Opt_jqfmt_vfsv1:
1192 qfmt = QFMT_VFS_V1;
1193 set_qf_format:
1194 if (sb_any_quota_loaded(sb) &&
1195 sbi->s_jquota_fmt != qfmt) {
1196 ext3_msg(sb, KERN_ERR, "error: cannot change "
1197 "journaled quota options when "
1198 "quota turned on.");
1199 return 0;
1201 sbi->s_jquota_fmt = qfmt;
1202 break;
1203 case Opt_quota:
1204 case Opt_usrquota:
1205 set_opt(sbi->s_mount_opt, QUOTA);
1206 set_opt(sbi->s_mount_opt, USRQUOTA);
1207 break;
1208 case Opt_grpquota:
1209 set_opt(sbi->s_mount_opt, QUOTA);
1210 set_opt(sbi->s_mount_opt, GRPQUOTA);
1211 break;
1212 case Opt_noquota:
1213 if (sb_any_quota_loaded(sb)) {
1214 ext3_msg(sb, KERN_ERR, "error: cannot change "
1215 "quota options when quota turned on.");
1216 return 0;
1218 clear_opt(sbi->s_mount_opt, QUOTA);
1219 clear_opt(sbi->s_mount_opt, USRQUOTA);
1220 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1221 break;
1222 #else
1223 case Opt_quota:
1224 case Opt_usrquota:
1225 case Opt_grpquota:
1226 ext3_msg(sb, KERN_ERR,
1227 "error: quota options not supported.");
1228 break;
1229 case Opt_usrjquota:
1230 case Opt_grpjquota:
1231 case Opt_offusrjquota:
1232 case Opt_offgrpjquota:
1233 case Opt_jqfmt_vfsold:
1234 case Opt_jqfmt_vfsv0:
1235 case Opt_jqfmt_vfsv1:
1236 ext3_msg(sb, KERN_ERR,
1237 "error: journaled quota options not "
1238 "supported.");
1239 break;
1240 case Opt_noquota:
1241 break;
1242 #endif
1243 case Opt_abort:
1244 set_opt(sbi->s_mount_opt, ABORT);
1245 break;
1246 case Opt_nobarrier:
1247 clear_opt(sbi->s_mount_opt, BARRIER);
1248 break;
1249 case Opt_barrier:
1250 if (args[0].from) {
1251 if (match_int(&args[0], &option))
1252 return 0;
1253 } else
1254 option = 1; /* No argument, default to 1 */
1255 if (option)
1256 set_opt(sbi->s_mount_opt, BARRIER);
1257 else
1258 clear_opt(sbi->s_mount_opt, BARRIER);
1259 break;
1260 case Opt_ignore:
1261 break;
1262 case Opt_resize:
1263 if (!is_remount) {
1264 ext3_msg(sb, KERN_ERR,
1265 "error: resize option only available "
1266 "for remount");
1267 return 0;
1269 if (match_int(&args[0], &option) != 0)
1270 return 0;
1271 *n_blocks_count = option;
1272 break;
1273 case Opt_nobh:
1274 ext3_msg(sb, KERN_WARNING,
1275 "warning: ignoring deprecated nobh option");
1276 break;
1277 case Opt_bh:
1278 ext3_msg(sb, KERN_WARNING,
1279 "warning: ignoring deprecated bh option");
1280 break;
1281 default:
1282 ext3_msg(sb, KERN_ERR,
1283 "error: unrecognized mount option \"%s\" "
1284 "or missing value", p);
1285 return 0;
1288 #ifdef CONFIG_QUOTA
1289 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1290 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1291 clear_opt(sbi->s_mount_opt, USRQUOTA);
1292 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1293 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1295 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1296 ext3_msg(sb, KERN_ERR, "error: old and new quota "
1297 "format mixing.");
1298 return 0;
1301 if (!sbi->s_jquota_fmt) {
1302 ext3_msg(sb, KERN_ERR, "error: journaled quota format "
1303 "not specified.");
1304 return 0;
1306 } else {
1307 if (sbi->s_jquota_fmt) {
1308 ext3_msg(sb, KERN_ERR, "error: journaled quota format "
1309 "specified with no journaling "
1310 "enabled.");
1311 return 0;
1314 #endif
1315 return 1;
1318 static int ext3_setup_super(struct super_block *sb, struct ext3_super_block *es,
1319 int read_only)
1321 struct ext3_sb_info *sbi = EXT3_SB(sb);
1322 int res = 0;
1324 if (le32_to_cpu(es->s_rev_level) > EXT3_MAX_SUPP_REV) {
1325 ext3_msg(sb, KERN_ERR,
1326 "error: revision level too high, "
1327 "forcing read-only mode");
1328 res = MS_RDONLY;
1330 if (read_only)
1331 return res;
1332 if (!(sbi->s_mount_state & EXT3_VALID_FS))
1333 ext3_msg(sb, KERN_WARNING,
1334 "warning: mounting unchecked fs, "
1335 "running e2fsck is recommended");
1336 else if ((sbi->s_mount_state & EXT3_ERROR_FS))
1337 ext3_msg(sb, KERN_WARNING,
1338 "warning: mounting fs with errors, "
1339 "running e2fsck is recommended");
1340 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1341 le16_to_cpu(es->s_mnt_count) >=
1342 le16_to_cpu(es->s_max_mnt_count))
1343 ext3_msg(sb, KERN_WARNING,
1344 "warning: maximal mount count reached, "
1345 "running e2fsck is recommended");
1346 else if (le32_to_cpu(es->s_checkinterval) &&
1347 (le32_to_cpu(es->s_lastcheck) +
1348 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1349 ext3_msg(sb, KERN_WARNING,
1350 "warning: checktime reached, "
1351 "running e2fsck is recommended");
1352 #if 0
1353 /* @@@ We _will_ want to clear the valid bit if we find
1354 inconsistencies, to force a fsck at reboot. But for
1355 a plain journaled filesystem we can keep it set as
1356 valid forever! :) */
1357 es->s_state &= cpu_to_le16(~EXT3_VALID_FS);
1358 #endif
1359 if (!le16_to_cpu(es->s_max_mnt_count))
1360 es->s_max_mnt_count = cpu_to_le16(EXT3_DFL_MAX_MNT_COUNT);
1361 le16_add_cpu(&es->s_mnt_count, 1);
1362 es->s_mtime = cpu_to_le32(get_seconds());
1363 ext3_update_dynamic_rev(sb);
1364 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
1366 ext3_commit_super(sb, es, 1);
1367 if (test_opt(sb, DEBUG))
1368 ext3_msg(sb, KERN_INFO, "[bs=%lu, gc=%lu, "
1369 "bpg=%lu, ipg=%lu, mo=%04lx]",
1370 sb->s_blocksize,
1371 sbi->s_groups_count,
1372 EXT3_BLOCKS_PER_GROUP(sb),
1373 EXT3_INODES_PER_GROUP(sb),
1374 sbi->s_mount_opt);
1376 if (EXT3_SB(sb)->s_journal->j_inode == NULL) {
1377 char b[BDEVNAME_SIZE];
1378 ext3_msg(sb, KERN_INFO, "using external journal on %s",
1379 bdevname(EXT3_SB(sb)->s_journal->j_dev, b));
1380 } else {
1381 ext3_msg(sb, KERN_INFO, "using internal journal");
1383 cleancache_init_fs(sb);
1384 return res;
1387 /* Called at mount-time, super-block is locked */
1388 static int ext3_check_descriptors(struct super_block *sb)
1390 struct ext3_sb_info *sbi = EXT3_SB(sb);
1391 int i;
1393 ext3_debug ("Checking group descriptors");
1395 for (i = 0; i < sbi->s_groups_count; i++) {
1396 struct ext3_group_desc *gdp = ext3_get_group_desc(sb, i, NULL);
1397 ext3_fsblk_t first_block = ext3_group_first_block_no(sb, i);
1398 ext3_fsblk_t last_block;
1400 if (i == sbi->s_groups_count - 1)
1401 last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
1402 else
1403 last_block = first_block +
1404 (EXT3_BLOCKS_PER_GROUP(sb) - 1);
1406 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
1407 le32_to_cpu(gdp->bg_block_bitmap) > last_block)
1409 ext3_error (sb, "ext3_check_descriptors",
1410 "Block bitmap for group %d"
1411 " not in group (block %lu)!",
1412 i, (unsigned long)
1413 le32_to_cpu(gdp->bg_block_bitmap));
1414 return 0;
1416 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
1417 le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
1419 ext3_error (sb, "ext3_check_descriptors",
1420 "Inode bitmap for group %d"
1421 " not in group (block %lu)!",
1422 i, (unsigned long)
1423 le32_to_cpu(gdp->bg_inode_bitmap));
1424 return 0;
1426 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
1427 le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
1428 last_block)
1430 ext3_error (sb, "ext3_check_descriptors",
1431 "Inode table for group %d"
1432 " not in group (block %lu)!",
1433 i, (unsigned long)
1434 le32_to_cpu(gdp->bg_inode_table));
1435 return 0;
1439 sbi->s_es->s_free_blocks_count=cpu_to_le32(ext3_count_free_blocks(sb));
1440 sbi->s_es->s_free_inodes_count=cpu_to_le32(ext3_count_free_inodes(sb));
1441 return 1;
1445 /* ext3_orphan_cleanup() walks a singly-linked list of inodes (starting at
1446 * the superblock) which were deleted from all directories, but held open by
1447 * a process at the time of a crash. We walk the list and try to delete these
1448 * inodes at recovery time (only with a read-write filesystem).
1450 * In order to keep the orphan inode chain consistent during traversal (in
1451 * case of crash during recovery), we link each inode into the superblock
1452 * orphan list_head and handle it the same way as an inode deletion during
1453 * normal operation (which journals the operations for us).
1455 * We only do an iget() and an iput() on each inode, which is very safe if we
1456 * accidentally point at an in-use or already deleted inode. The worst that
1457 * can happen in this case is that we get a "bit already cleared" message from
1458 * ext3_free_inode(). The only reason we would point at a wrong inode is if
1459 * e2fsck was run on this filesystem, and it must have already done the orphan
1460 * inode cleanup for us, so we can safely abort without any further action.
1462 static void ext3_orphan_cleanup (struct super_block * sb,
1463 struct ext3_super_block * es)
1465 unsigned int s_flags = sb->s_flags;
1466 int nr_orphans = 0, nr_truncates = 0;
1467 #ifdef CONFIG_QUOTA
1468 int i;
1469 #endif
1470 if (!es->s_last_orphan) {
1471 jbd_debug(4, "no orphan inodes to clean up\n");
1472 return;
1475 if (bdev_read_only(sb->s_bdev)) {
1476 ext3_msg(sb, KERN_ERR, "error: write access "
1477 "unavailable, skipping orphan cleanup.");
1478 return;
1481 /* Check if feature set allows readwrite operations */
1482 if (EXT3_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP)) {
1483 ext3_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
1484 "unknown ROCOMPAT features");
1485 return;
1488 if (EXT3_SB(sb)->s_mount_state & EXT3_ERROR_FS) {
1489 if (es->s_last_orphan)
1490 jbd_debug(1, "Errors on filesystem, "
1491 "clearing orphan list.\n");
1492 es->s_last_orphan = 0;
1493 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1494 return;
1497 if (s_flags & MS_RDONLY) {
1498 ext3_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
1499 sb->s_flags &= ~MS_RDONLY;
1501 #ifdef CONFIG_QUOTA
1502 /* Needed for iput() to work correctly and not trash data */
1503 sb->s_flags |= MS_ACTIVE;
1504 /* Turn on quotas so that they are updated correctly */
1505 for (i = 0; i < MAXQUOTAS; i++) {
1506 if (EXT3_SB(sb)->s_qf_names[i]) {
1507 int ret = ext3_quota_on_mount(sb, i);
1508 if (ret < 0)
1509 ext3_msg(sb, KERN_ERR,
1510 "error: cannot turn on journaled "
1511 "quota: %d", ret);
1514 #endif
1516 while (es->s_last_orphan) {
1517 struct inode *inode;
1519 inode = ext3_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1520 if (IS_ERR(inode)) {
1521 es->s_last_orphan = 0;
1522 break;
1525 list_add(&EXT3_I(inode)->i_orphan, &EXT3_SB(sb)->s_orphan);
1526 dquot_initialize(inode);
1527 if (inode->i_nlink) {
1528 printk(KERN_DEBUG
1529 "%s: truncating inode %lu to %Ld bytes\n",
1530 __func__, inode->i_ino, inode->i_size);
1531 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1532 inode->i_ino, inode->i_size);
1533 ext3_truncate(inode);
1534 nr_truncates++;
1535 } else {
1536 printk(KERN_DEBUG
1537 "%s: deleting unreferenced inode %lu\n",
1538 __func__, inode->i_ino);
1539 jbd_debug(2, "deleting unreferenced inode %lu\n",
1540 inode->i_ino);
1541 nr_orphans++;
1543 iput(inode); /* The delete magic happens here! */
1546 #define PLURAL(x) (x), ((x)==1) ? "" : "s"
1548 if (nr_orphans)
1549 ext3_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
1550 PLURAL(nr_orphans));
1551 if (nr_truncates)
1552 ext3_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
1553 PLURAL(nr_truncates));
1554 #ifdef CONFIG_QUOTA
1555 /* Turn quotas off */
1556 for (i = 0; i < MAXQUOTAS; i++) {
1557 if (sb_dqopt(sb)->files[i])
1558 dquot_quota_off(sb, i);
1560 #endif
1561 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1565 * Maximal file size. There is a direct, and {,double-,triple-}indirect
1566 * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
1567 * We need to be 1 filesystem block less than the 2^32 sector limit.
1569 static loff_t ext3_max_size(int bits)
1571 loff_t res = EXT3_NDIR_BLOCKS;
1572 int meta_blocks;
1573 loff_t upper_limit;
1575 /* This is calculated to be the largest file size for a
1576 * dense, file such that the total number of
1577 * sectors in the file, including data and all indirect blocks,
1578 * does not exceed 2^32 -1
1579 * __u32 i_blocks representing the total number of
1580 * 512 bytes blocks of the file
1582 upper_limit = (1LL << 32) - 1;
1584 /* total blocks in file system block size */
1585 upper_limit >>= (bits - 9);
1588 /* indirect blocks */
1589 meta_blocks = 1;
1590 /* double indirect blocks */
1591 meta_blocks += 1 + (1LL << (bits-2));
1592 /* tripple indirect blocks */
1593 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1595 upper_limit -= meta_blocks;
1596 upper_limit <<= bits;
1598 res += 1LL << (bits-2);
1599 res += 1LL << (2*(bits-2));
1600 res += 1LL << (3*(bits-2));
1601 res <<= bits;
1602 if (res > upper_limit)
1603 res = upper_limit;
1605 if (res > MAX_LFS_FILESIZE)
1606 res = MAX_LFS_FILESIZE;
1608 return res;
1611 static ext3_fsblk_t descriptor_loc(struct super_block *sb,
1612 ext3_fsblk_t logic_sb_block,
1613 int nr)
1615 struct ext3_sb_info *sbi = EXT3_SB(sb);
1616 unsigned long bg, first_meta_bg;
1617 int has_super = 0;
1619 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1621 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_META_BG) ||
1622 nr < first_meta_bg)
1623 return (logic_sb_block + nr + 1);
1624 bg = sbi->s_desc_per_block * nr;
1625 if (ext3_bg_has_super(sb, bg))
1626 has_super = 1;
1627 return (has_super + ext3_group_first_block_no(sb, bg));
1631 static int ext3_fill_super (struct super_block *sb, void *data, int silent)
1633 struct buffer_head * bh;
1634 struct ext3_super_block *es = NULL;
1635 struct ext3_sb_info *sbi;
1636 ext3_fsblk_t block;
1637 ext3_fsblk_t sb_block = get_sb_block(&data, sb);
1638 ext3_fsblk_t logic_sb_block;
1639 unsigned long offset = 0;
1640 unsigned int journal_inum = 0;
1641 unsigned long journal_devnum = 0;
1642 unsigned long def_mount_opts;
1643 struct inode *root;
1644 int blocksize;
1645 int hblock;
1646 int db_count;
1647 int i;
1648 int needs_recovery;
1649 int ret = -EINVAL;
1650 __le32 features;
1651 int err;
1653 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1654 if (!sbi)
1655 return -ENOMEM;
1657 sbi->s_blockgroup_lock =
1658 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
1659 if (!sbi->s_blockgroup_lock) {
1660 kfree(sbi);
1661 return -ENOMEM;
1663 sb->s_fs_info = sbi;
1664 sbi->s_mount_opt = 0;
1665 sbi->s_resuid = EXT3_DEF_RESUID;
1666 sbi->s_resgid = EXT3_DEF_RESGID;
1667 sbi->s_sb_block = sb_block;
1669 blocksize = sb_min_blocksize(sb, EXT3_MIN_BLOCK_SIZE);
1670 if (!blocksize) {
1671 ext3_msg(sb, KERN_ERR, "error: unable to set blocksize");
1672 goto out_fail;
1676 * The ext3 superblock will not be buffer aligned for other than 1kB
1677 * block sizes. We need to calculate the offset from buffer start.
1679 if (blocksize != EXT3_MIN_BLOCK_SIZE) {
1680 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1681 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1682 } else {
1683 logic_sb_block = sb_block;
1686 if (!(bh = sb_bread(sb, logic_sb_block))) {
1687 ext3_msg(sb, KERN_ERR, "error: unable to read superblock");
1688 goto out_fail;
1691 * Note: s_es must be initialized as soon as possible because
1692 * some ext3 macro-instructions depend on its value
1694 es = (struct ext3_super_block *) (bh->b_data + offset);
1695 sbi->s_es = es;
1696 sb->s_magic = le16_to_cpu(es->s_magic);
1697 if (sb->s_magic != EXT3_SUPER_MAGIC)
1698 goto cantfind_ext3;
1700 /* Set defaults before we parse the mount options */
1701 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1702 if (def_mount_opts & EXT3_DEFM_DEBUG)
1703 set_opt(sbi->s_mount_opt, DEBUG);
1704 if (def_mount_opts & EXT3_DEFM_BSDGROUPS)
1705 set_opt(sbi->s_mount_opt, GRPID);
1706 if (def_mount_opts & EXT3_DEFM_UID16)
1707 set_opt(sbi->s_mount_opt, NO_UID32);
1708 #ifdef CONFIG_EXT3_FS_XATTR
1709 if (def_mount_opts & EXT3_DEFM_XATTR_USER)
1710 set_opt(sbi->s_mount_opt, XATTR_USER);
1711 #endif
1712 #ifdef CONFIG_EXT3_FS_POSIX_ACL
1713 if (def_mount_opts & EXT3_DEFM_ACL)
1714 set_opt(sbi->s_mount_opt, POSIX_ACL);
1715 #endif
1716 if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_DATA)
1717 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
1718 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_ORDERED)
1719 set_opt(sbi->s_mount_opt, ORDERED_DATA);
1720 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_WBACK)
1721 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
1723 if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_PANIC)
1724 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1725 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_CONTINUE)
1726 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1727 else
1728 set_opt(sbi->s_mount_opt, ERRORS_RO);
1730 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1731 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1733 /* enable barriers by default */
1734 set_opt(sbi->s_mount_opt, BARRIER);
1735 set_opt(sbi->s_mount_opt, RESERVATION);
1737 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1738 NULL, 0))
1739 goto failed_mount;
1741 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1742 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
1744 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV &&
1745 (EXT3_HAS_COMPAT_FEATURE(sb, ~0U) ||
1746 EXT3_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1747 EXT3_HAS_INCOMPAT_FEATURE(sb, ~0U)))
1748 ext3_msg(sb, KERN_WARNING,
1749 "warning: feature flags set on rev 0 fs, "
1750 "running e2fsck is recommended");
1752 * Check feature flags regardless of the revision level, since we
1753 * previously didn't change the revision level when setting the flags,
1754 * so there is a chance incompat flags are set on a rev 0 filesystem.
1756 features = EXT3_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP);
1757 if (features) {
1758 ext3_msg(sb, KERN_ERR,
1759 "error: couldn't mount because of unsupported "
1760 "optional features (%x)", le32_to_cpu(features));
1761 goto failed_mount;
1763 features = EXT3_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP);
1764 if (!(sb->s_flags & MS_RDONLY) && features) {
1765 ext3_msg(sb, KERN_ERR,
1766 "error: couldn't mount RDWR because of unsupported "
1767 "optional features (%x)", le32_to_cpu(features));
1768 goto failed_mount;
1770 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
1772 if (blocksize < EXT3_MIN_BLOCK_SIZE ||
1773 blocksize > EXT3_MAX_BLOCK_SIZE) {
1774 ext3_msg(sb, KERN_ERR,
1775 "error: couldn't mount because of unsupported "
1776 "filesystem blocksize %d", blocksize);
1777 goto failed_mount;
1780 hblock = bdev_logical_block_size(sb->s_bdev);
1781 if (sb->s_blocksize != blocksize) {
1783 * Make sure the blocksize for the filesystem is larger
1784 * than the hardware sectorsize for the machine.
1786 if (blocksize < hblock) {
1787 ext3_msg(sb, KERN_ERR,
1788 "error: fsblocksize %d too small for "
1789 "hardware sectorsize %d", blocksize, hblock);
1790 goto failed_mount;
1793 brelse (bh);
1794 if (!sb_set_blocksize(sb, blocksize)) {
1795 ext3_msg(sb, KERN_ERR,
1796 "error: bad blocksize %d", blocksize);
1797 goto out_fail;
1799 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1800 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1801 bh = sb_bread(sb, logic_sb_block);
1802 if (!bh) {
1803 ext3_msg(sb, KERN_ERR,
1804 "error: can't read superblock on 2nd try");
1805 goto failed_mount;
1807 es = (struct ext3_super_block *)(bh->b_data + offset);
1808 sbi->s_es = es;
1809 if (es->s_magic != cpu_to_le16(EXT3_SUPER_MAGIC)) {
1810 ext3_msg(sb, KERN_ERR,
1811 "error: magic mismatch");
1812 goto failed_mount;
1816 sb->s_maxbytes = ext3_max_size(sb->s_blocksize_bits);
1818 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV) {
1819 sbi->s_inode_size = EXT3_GOOD_OLD_INODE_SIZE;
1820 sbi->s_first_ino = EXT3_GOOD_OLD_FIRST_INO;
1821 } else {
1822 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1823 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1824 if ((sbi->s_inode_size < EXT3_GOOD_OLD_INODE_SIZE) ||
1825 (!is_power_of_2(sbi->s_inode_size)) ||
1826 (sbi->s_inode_size > blocksize)) {
1827 ext3_msg(sb, KERN_ERR,
1828 "error: unsupported inode size: %d",
1829 sbi->s_inode_size);
1830 goto failed_mount;
1833 sbi->s_frag_size = EXT3_MIN_FRAG_SIZE <<
1834 le32_to_cpu(es->s_log_frag_size);
1835 if (blocksize != sbi->s_frag_size) {
1836 ext3_msg(sb, KERN_ERR,
1837 "error: fragsize %lu != blocksize %u (unsupported)",
1838 sbi->s_frag_size, blocksize);
1839 goto failed_mount;
1841 sbi->s_frags_per_block = 1;
1842 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1843 sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
1844 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
1845 if (EXT3_INODE_SIZE(sb) == 0 || EXT3_INODES_PER_GROUP(sb) == 0)
1846 goto cantfind_ext3;
1847 sbi->s_inodes_per_block = blocksize / EXT3_INODE_SIZE(sb);
1848 if (sbi->s_inodes_per_block == 0)
1849 goto cantfind_ext3;
1850 sbi->s_itb_per_group = sbi->s_inodes_per_group /
1851 sbi->s_inodes_per_block;
1852 sbi->s_desc_per_block = blocksize / sizeof(struct ext3_group_desc);
1853 sbi->s_sbh = bh;
1854 sbi->s_mount_state = le16_to_cpu(es->s_state);
1855 sbi->s_addr_per_block_bits = ilog2(EXT3_ADDR_PER_BLOCK(sb));
1856 sbi->s_desc_per_block_bits = ilog2(EXT3_DESC_PER_BLOCK(sb));
1857 for (i=0; i < 4; i++)
1858 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
1859 sbi->s_def_hash_version = es->s_def_hash_version;
1860 i = le32_to_cpu(es->s_flags);
1861 if (i & EXT2_FLAGS_UNSIGNED_HASH)
1862 sbi->s_hash_unsigned = 3;
1863 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
1864 #ifdef __CHAR_UNSIGNED__
1865 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
1866 sbi->s_hash_unsigned = 3;
1867 #else
1868 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
1869 #endif
1872 if (sbi->s_blocks_per_group > blocksize * 8) {
1873 ext3_msg(sb, KERN_ERR,
1874 "#blocks per group too big: %lu",
1875 sbi->s_blocks_per_group);
1876 goto failed_mount;
1878 if (sbi->s_frags_per_group > blocksize * 8) {
1879 ext3_msg(sb, KERN_ERR,
1880 "error: #fragments per group too big: %lu",
1881 sbi->s_frags_per_group);
1882 goto failed_mount;
1884 if (sbi->s_inodes_per_group > blocksize * 8) {
1885 ext3_msg(sb, KERN_ERR,
1886 "error: #inodes per group too big: %lu",
1887 sbi->s_inodes_per_group);
1888 goto failed_mount;
1891 err = generic_check_addressable(sb->s_blocksize_bits,
1892 le32_to_cpu(es->s_blocks_count));
1893 if (err) {
1894 ext3_msg(sb, KERN_ERR,
1895 "error: filesystem is too large to mount safely");
1896 if (sizeof(sector_t) < 8)
1897 ext3_msg(sb, KERN_ERR,
1898 "error: CONFIG_LBDAF not enabled");
1899 ret = err;
1900 goto failed_mount;
1903 if (EXT3_BLOCKS_PER_GROUP(sb) == 0)
1904 goto cantfind_ext3;
1905 sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1906 le32_to_cpu(es->s_first_data_block) - 1)
1907 / EXT3_BLOCKS_PER_GROUP(sb)) + 1;
1908 db_count = DIV_ROUND_UP(sbi->s_groups_count, EXT3_DESC_PER_BLOCK(sb));
1909 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
1910 GFP_KERNEL);
1911 if (sbi->s_group_desc == NULL) {
1912 ext3_msg(sb, KERN_ERR,
1913 "error: not enough memory");
1914 ret = -ENOMEM;
1915 goto failed_mount;
1918 bgl_lock_init(sbi->s_blockgroup_lock);
1920 for (i = 0; i < db_count; i++) {
1921 block = descriptor_loc(sb, logic_sb_block, i);
1922 sbi->s_group_desc[i] = sb_bread(sb, block);
1923 if (!sbi->s_group_desc[i]) {
1924 ext3_msg(sb, KERN_ERR,
1925 "error: can't read group descriptor %d", i);
1926 db_count = i;
1927 goto failed_mount2;
1930 if (!ext3_check_descriptors (sb)) {
1931 ext3_msg(sb, KERN_ERR,
1932 "error: group descriptors corrupted");
1933 goto failed_mount2;
1935 sbi->s_gdb_count = db_count;
1936 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1937 spin_lock_init(&sbi->s_next_gen_lock);
1939 /* per fileystem reservation list head & lock */
1940 spin_lock_init(&sbi->s_rsv_window_lock);
1941 sbi->s_rsv_window_root = RB_ROOT;
1942 /* Add a single, static dummy reservation to the start of the
1943 * reservation window list --- it gives us a placeholder for
1944 * append-at-start-of-list which makes the allocation logic
1945 * _much_ simpler. */
1946 sbi->s_rsv_window_head.rsv_start = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1947 sbi->s_rsv_window_head.rsv_end = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1948 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1949 sbi->s_rsv_window_head.rsv_goal_size = 0;
1950 ext3_rsv_window_add(sb, &sbi->s_rsv_window_head);
1953 * set up enough so that it can read an inode
1955 sb->s_op = &ext3_sops;
1956 sb->s_export_op = &ext3_export_ops;
1957 sb->s_xattr = ext3_xattr_handlers;
1958 #ifdef CONFIG_QUOTA
1959 sb->s_qcop = &ext3_qctl_operations;
1960 sb->dq_op = &ext3_quota_operations;
1961 #endif
1962 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
1963 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
1964 mutex_init(&sbi->s_orphan_lock);
1965 mutex_init(&sbi->s_resize_lock);
1967 sb->s_root = NULL;
1969 needs_recovery = (es->s_last_orphan != 0 ||
1970 EXT3_HAS_INCOMPAT_FEATURE(sb,
1971 EXT3_FEATURE_INCOMPAT_RECOVER));
1974 * The first inode we look at is the journal inode. Don't try
1975 * root first: it may be modified in the journal!
1977 if (!test_opt(sb, NOLOAD) &&
1978 EXT3_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL)) {
1979 if (ext3_load_journal(sb, es, journal_devnum))
1980 goto failed_mount2;
1981 } else if (journal_inum) {
1982 if (ext3_create_journal(sb, es, journal_inum))
1983 goto failed_mount2;
1984 } else {
1985 if (!silent)
1986 ext3_msg(sb, KERN_ERR,
1987 "error: no journal found. "
1988 "mounting ext3 over ext2?");
1989 goto failed_mount2;
1991 err = percpu_counter_init(&sbi->s_freeblocks_counter,
1992 ext3_count_free_blocks(sb));
1993 if (!err) {
1994 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1995 ext3_count_free_inodes(sb));
1997 if (!err) {
1998 err = percpu_counter_init(&sbi->s_dirs_counter,
1999 ext3_count_dirs(sb));
2001 if (err) {
2002 ext3_msg(sb, KERN_ERR, "error: insufficient memory");
2003 ret = err;
2004 goto failed_mount3;
2007 /* We have now updated the journal if required, so we can
2008 * validate the data journaling mode. */
2009 switch (test_opt(sb, DATA_FLAGS)) {
2010 case 0:
2011 /* No mode set, assume a default based on the journal
2012 capabilities: ORDERED_DATA if the journal can
2013 cope, else JOURNAL_DATA */
2014 if (journal_check_available_features
2015 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE))
2016 set_opt(sbi->s_mount_opt, DEFAULT_DATA_MODE);
2017 else
2018 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2019 break;
2021 case EXT3_MOUNT_ORDERED_DATA:
2022 case EXT3_MOUNT_WRITEBACK_DATA:
2023 if (!journal_check_available_features
2024 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE)) {
2025 ext3_msg(sb, KERN_ERR,
2026 "error: journal does not support "
2027 "requested data journaling mode");
2028 goto failed_mount3;
2030 default:
2031 break;
2035 * The journal_load will have done any necessary log recovery,
2036 * so we can safely mount the rest of the filesystem now.
2039 root = ext3_iget(sb, EXT3_ROOT_INO);
2040 if (IS_ERR(root)) {
2041 ext3_msg(sb, KERN_ERR, "error: get root inode failed");
2042 ret = PTR_ERR(root);
2043 goto failed_mount3;
2045 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2046 iput(root);
2047 ext3_msg(sb, KERN_ERR, "error: corrupt root inode, run e2fsck");
2048 goto failed_mount3;
2050 sb->s_root = d_alloc_root(root);
2051 if (!sb->s_root) {
2052 ext3_msg(sb, KERN_ERR, "error: get root dentry failed");
2053 iput(root);
2054 ret = -ENOMEM;
2055 goto failed_mount3;
2058 ext3_setup_super (sb, es, sb->s_flags & MS_RDONLY);
2060 EXT3_SB(sb)->s_mount_state |= EXT3_ORPHAN_FS;
2061 ext3_orphan_cleanup(sb, es);
2062 EXT3_SB(sb)->s_mount_state &= ~EXT3_ORPHAN_FS;
2063 if (needs_recovery)
2064 ext3_msg(sb, KERN_INFO, "recovery complete");
2065 ext3_mark_recovery_complete(sb, es);
2066 ext3_msg(sb, KERN_INFO, "mounted filesystem with %s data mode",
2067 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA ? "journal":
2068 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA ? "ordered":
2069 "writeback");
2071 return 0;
2073 cantfind_ext3:
2074 if (!silent)
2075 ext3_msg(sb, KERN_INFO,
2076 "error: can't find ext3 filesystem on dev %s.",
2077 sb->s_id);
2078 goto failed_mount;
2080 failed_mount3:
2081 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2082 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2083 percpu_counter_destroy(&sbi->s_dirs_counter);
2084 journal_destroy(sbi->s_journal);
2085 failed_mount2:
2086 for (i = 0; i < db_count; i++)
2087 brelse(sbi->s_group_desc[i]);
2088 kfree(sbi->s_group_desc);
2089 failed_mount:
2090 #ifdef CONFIG_QUOTA
2091 for (i = 0; i < MAXQUOTAS; i++)
2092 kfree(sbi->s_qf_names[i]);
2093 #endif
2094 ext3_blkdev_remove(sbi);
2095 brelse(bh);
2096 out_fail:
2097 sb->s_fs_info = NULL;
2098 kfree(sbi->s_blockgroup_lock);
2099 kfree(sbi);
2100 return ret;
2104 * Setup any per-fs journal parameters now. We'll do this both on
2105 * initial mount, once the journal has been initialised but before we've
2106 * done any recovery; and again on any subsequent remount.
2108 static void ext3_init_journal_params(struct super_block *sb, journal_t *journal)
2110 struct ext3_sb_info *sbi = EXT3_SB(sb);
2112 if (sbi->s_commit_interval)
2113 journal->j_commit_interval = sbi->s_commit_interval;
2114 /* We could also set up an ext3-specific default for the commit
2115 * interval here, but for now we'll just fall back to the jbd
2116 * default. */
2118 spin_lock(&journal->j_state_lock);
2119 if (test_opt(sb, BARRIER))
2120 journal->j_flags |= JFS_BARRIER;
2121 else
2122 journal->j_flags &= ~JFS_BARRIER;
2123 if (test_opt(sb, DATA_ERR_ABORT))
2124 journal->j_flags |= JFS_ABORT_ON_SYNCDATA_ERR;
2125 else
2126 journal->j_flags &= ~JFS_ABORT_ON_SYNCDATA_ERR;
2127 spin_unlock(&journal->j_state_lock);
2130 static journal_t *ext3_get_journal(struct super_block *sb,
2131 unsigned int journal_inum)
2133 struct inode *journal_inode;
2134 journal_t *journal;
2136 /* First, test for the existence of a valid inode on disk. Bad
2137 * things happen if we iget() an unused inode, as the subsequent
2138 * iput() will try to delete it. */
2140 journal_inode = ext3_iget(sb, journal_inum);
2141 if (IS_ERR(journal_inode)) {
2142 ext3_msg(sb, KERN_ERR, "error: no journal found");
2143 return NULL;
2145 if (!journal_inode->i_nlink) {
2146 make_bad_inode(journal_inode);
2147 iput(journal_inode);
2148 ext3_msg(sb, KERN_ERR, "error: journal inode is deleted");
2149 return NULL;
2152 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
2153 journal_inode, journal_inode->i_size);
2154 if (!S_ISREG(journal_inode->i_mode)) {
2155 ext3_msg(sb, KERN_ERR, "error: invalid journal inode");
2156 iput(journal_inode);
2157 return NULL;
2160 journal = journal_init_inode(journal_inode);
2161 if (!journal) {
2162 ext3_msg(sb, KERN_ERR, "error: could not load journal inode");
2163 iput(journal_inode);
2164 return NULL;
2166 journal->j_private = sb;
2167 ext3_init_journal_params(sb, journal);
2168 return journal;
2171 static journal_t *ext3_get_dev_journal(struct super_block *sb,
2172 dev_t j_dev)
2174 struct buffer_head * bh;
2175 journal_t *journal;
2176 ext3_fsblk_t start;
2177 ext3_fsblk_t len;
2178 int hblock, blocksize;
2179 ext3_fsblk_t sb_block;
2180 unsigned long offset;
2181 struct ext3_super_block * es;
2182 struct block_device *bdev;
2184 bdev = ext3_blkdev_get(j_dev, sb);
2185 if (bdev == NULL)
2186 return NULL;
2188 blocksize = sb->s_blocksize;
2189 hblock = bdev_logical_block_size(bdev);
2190 if (blocksize < hblock) {
2191 ext3_msg(sb, KERN_ERR,
2192 "error: blocksize too small for journal device");
2193 goto out_bdev;
2196 sb_block = EXT3_MIN_BLOCK_SIZE / blocksize;
2197 offset = EXT3_MIN_BLOCK_SIZE % blocksize;
2198 set_blocksize(bdev, blocksize);
2199 if (!(bh = __bread(bdev, sb_block, blocksize))) {
2200 ext3_msg(sb, KERN_ERR, "error: couldn't read superblock of "
2201 "external journal");
2202 goto out_bdev;
2205 es = (struct ext3_super_block *) (bh->b_data + offset);
2206 if ((le16_to_cpu(es->s_magic) != EXT3_SUPER_MAGIC) ||
2207 !(le32_to_cpu(es->s_feature_incompat) &
2208 EXT3_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2209 ext3_msg(sb, KERN_ERR, "error: external journal has "
2210 "bad superblock");
2211 brelse(bh);
2212 goto out_bdev;
2215 if (memcmp(EXT3_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2216 ext3_msg(sb, KERN_ERR, "error: journal UUID does not match");
2217 brelse(bh);
2218 goto out_bdev;
2221 len = le32_to_cpu(es->s_blocks_count);
2222 start = sb_block + 1;
2223 brelse(bh); /* we're done with the superblock */
2225 journal = journal_init_dev(bdev, sb->s_bdev,
2226 start, len, blocksize);
2227 if (!journal) {
2228 ext3_msg(sb, KERN_ERR,
2229 "error: failed to create device journal");
2230 goto out_bdev;
2232 journal->j_private = sb;
2233 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2234 wait_on_buffer(journal->j_sb_buffer);
2235 if (!buffer_uptodate(journal->j_sb_buffer)) {
2236 ext3_msg(sb, KERN_ERR, "I/O error on journal device");
2237 goto out_journal;
2239 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
2240 ext3_msg(sb, KERN_ERR,
2241 "error: external journal has more than one "
2242 "user (unsupported) - %d",
2243 be32_to_cpu(journal->j_superblock->s_nr_users));
2244 goto out_journal;
2246 EXT3_SB(sb)->journal_bdev = bdev;
2247 ext3_init_journal_params(sb, journal);
2248 return journal;
2249 out_journal:
2250 journal_destroy(journal);
2251 out_bdev:
2252 ext3_blkdev_put(bdev);
2253 return NULL;
2256 static int ext3_load_journal(struct super_block *sb,
2257 struct ext3_super_block *es,
2258 unsigned long journal_devnum)
2260 journal_t *journal;
2261 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2262 dev_t journal_dev;
2263 int err = 0;
2264 int really_read_only;
2266 if (journal_devnum &&
2267 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2268 ext3_msg(sb, KERN_INFO, "external journal device major/minor "
2269 "numbers have changed");
2270 journal_dev = new_decode_dev(journal_devnum);
2271 } else
2272 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2274 really_read_only = bdev_read_only(sb->s_bdev);
2277 * Are we loading a blank journal or performing recovery after a
2278 * crash? For recovery, we need to check in advance whether we
2279 * can get read-write access to the device.
2282 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER)) {
2283 if (sb->s_flags & MS_RDONLY) {
2284 ext3_msg(sb, KERN_INFO,
2285 "recovery required on readonly filesystem");
2286 if (really_read_only) {
2287 ext3_msg(sb, KERN_ERR, "error: write access "
2288 "unavailable, cannot proceed");
2289 return -EROFS;
2291 ext3_msg(sb, KERN_INFO,
2292 "write access will be enabled during recovery");
2296 if (journal_inum && journal_dev) {
2297 ext3_msg(sb, KERN_ERR, "error: filesystem has both journal "
2298 "and inode journals");
2299 return -EINVAL;
2302 if (journal_inum) {
2303 if (!(journal = ext3_get_journal(sb, journal_inum)))
2304 return -EINVAL;
2305 } else {
2306 if (!(journal = ext3_get_dev_journal(sb, journal_dev)))
2307 return -EINVAL;
2310 if (!(journal->j_flags & JFS_BARRIER))
2311 printk(KERN_INFO "EXT3-fs: barriers not enabled\n");
2313 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2314 err = journal_update_format(journal);
2315 if (err) {
2316 ext3_msg(sb, KERN_ERR, "error updating journal");
2317 journal_destroy(journal);
2318 return err;
2322 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER))
2323 err = journal_wipe(journal, !really_read_only);
2324 if (!err)
2325 err = journal_load(journal);
2327 if (err) {
2328 ext3_msg(sb, KERN_ERR, "error loading journal");
2329 journal_destroy(journal);
2330 return err;
2333 EXT3_SB(sb)->s_journal = journal;
2334 ext3_clear_journal_err(sb, es);
2336 if (!really_read_only && journal_devnum &&
2337 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2338 es->s_journal_dev = cpu_to_le32(journal_devnum);
2340 /* Make sure we flush the recovery flag to disk. */
2341 ext3_commit_super(sb, es, 1);
2344 return 0;
2347 static int ext3_create_journal(struct super_block *sb,
2348 struct ext3_super_block *es,
2349 unsigned int journal_inum)
2351 journal_t *journal;
2352 int err;
2354 if (sb->s_flags & MS_RDONLY) {
2355 ext3_msg(sb, KERN_ERR,
2356 "error: readonly filesystem when trying to "
2357 "create journal");
2358 return -EROFS;
2361 journal = ext3_get_journal(sb, journal_inum);
2362 if (!journal)
2363 return -EINVAL;
2365 ext3_msg(sb, KERN_INFO, "creating new journal on inode %u",
2366 journal_inum);
2368 err = journal_create(journal);
2369 if (err) {
2370 ext3_msg(sb, KERN_ERR, "error creating journal");
2371 journal_destroy(journal);
2372 return -EIO;
2375 EXT3_SB(sb)->s_journal = journal;
2377 ext3_update_dynamic_rev(sb);
2378 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2379 EXT3_SET_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL);
2381 es->s_journal_inum = cpu_to_le32(journal_inum);
2383 /* Make sure we flush the recovery flag to disk. */
2384 ext3_commit_super(sb, es, 1);
2386 return 0;
2389 static int ext3_commit_super(struct super_block *sb,
2390 struct ext3_super_block *es,
2391 int sync)
2393 struct buffer_head *sbh = EXT3_SB(sb)->s_sbh;
2394 int error = 0;
2396 if (!sbh)
2397 return error;
2399 if (buffer_write_io_error(sbh)) {
2401 * Oh, dear. A previous attempt to write the
2402 * superblock failed. This could happen because the
2403 * USB device was yanked out. Or it could happen to
2404 * be a transient write error and maybe the block will
2405 * be remapped. Nothing we can do but to retry the
2406 * write and hope for the best.
2408 ext3_msg(sb, KERN_ERR, "previous I/O error to "
2409 "superblock detected");
2410 clear_buffer_write_io_error(sbh);
2411 set_buffer_uptodate(sbh);
2414 * If the file system is mounted read-only, don't update the
2415 * superblock write time. This avoids updating the superblock
2416 * write time when we are mounting the root file system
2417 * read/only but we need to replay the journal; at that point,
2418 * for people who are east of GMT and who make their clock
2419 * tick in localtime for Windows bug-for-bug compatibility,
2420 * the clock is set in the future, and this will cause e2fsck
2421 * to complain and force a full file system check.
2423 if (!(sb->s_flags & MS_RDONLY))
2424 es->s_wtime = cpu_to_le32(get_seconds());
2425 es->s_free_blocks_count = cpu_to_le32(ext3_count_free_blocks(sb));
2426 es->s_free_inodes_count = cpu_to_le32(ext3_count_free_inodes(sb));
2427 BUFFER_TRACE(sbh, "marking dirty");
2428 mark_buffer_dirty(sbh);
2429 if (sync) {
2430 error = sync_dirty_buffer(sbh);
2431 if (buffer_write_io_error(sbh)) {
2432 ext3_msg(sb, KERN_ERR, "I/O error while writing "
2433 "superblock");
2434 clear_buffer_write_io_error(sbh);
2435 set_buffer_uptodate(sbh);
2438 return error;
2443 * Have we just finished recovery? If so, and if we are mounting (or
2444 * remounting) the filesystem readonly, then we will end up with a
2445 * consistent fs on disk. Record that fact.
2447 static void ext3_mark_recovery_complete(struct super_block * sb,
2448 struct ext3_super_block * es)
2450 journal_t *journal = EXT3_SB(sb)->s_journal;
2452 journal_lock_updates(journal);
2453 if (journal_flush(journal) < 0)
2454 goto out;
2456 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER) &&
2457 sb->s_flags & MS_RDONLY) {
2458 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2459 ext3_commit_super(sb, es, 1);
2462 out:
2463 journal_unlock_updates(journal);
2467 * If we are mounting (or read-write remounting) a filesystem whose journal
2468 * has recorded an error from a previous lifetime, move that error to the
2469 * main filesystem now.
2471 static void ext3_clear_journal_err(struct super_block *sb,
2472 struct ext3_super_block *es)
2474 journal_t *journal;
2475 int j_errno;
2476 const char *errstr;
2478 journal = EXT3_SB(sb)->s_journal;
2481 * Now check for any error status which may have been recorded in the
2482 * journal by a prior ext3_error() or ext3_abort()
2485 j_errno = journal_errno(journal);
2486 if (j_errno) {
2487 char nbuf[16];
2489 errstr = ext3_decode_error(sb, j_errno, nbuf);
2490 ext3_warning(sb, __func__, "Filesystem error recorded "
2491 "from previous mount: %s", errstr);
2492 ext3_warning(sb, __func__, "Marking fs in need of "
2493 "filesystem check.");
2495 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
2496 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
2497 ext3_commit_super (sb, es, 1);
2499 journal_clear_err(journal);
2504 * Force the running and committing transactions to commit,
2505 * and wait on the commit.
2507 int ext3_force_commit(struct super_block *sb)
2509 journal_t *journal;
2510 int ret;
2512 if (sb->s_flags & MS_RDONLY)
2513 return 0;
2515 journal = EXT3_SB(sb)->s_journal;
2516 ret = ext3_journal_force_commit(journal);
2517 return ret;
2520 static int ext3_sync_fs(struct super_block *sb, int wait)
2522 tid_t target;
2524 trace_ext3_sync_fs(sb, wait);
2525 if (journal_start_commit(EXT3_SB(sb)->s_journal, &target)) {
2526 if (wait)
2527 log_wait_commit(EXT3_SB(sb)->s_journal, target);
2529 return 0;
2533 * LVM calls this function before a (read-only) snapshot is created. This
2534 * gives us a chance to flush the journal completely and mark the fs clean.
2536 static int ext3_freeze(struct super_block *sb)
2538 int error = 0;
2539 journal_t *journal;
2541 if (!(sb->s_flags & MS_RDONLY)) {
2542 journal = EXT3_SB(sb)->s_journal;
2544 /* Now we set up the journal barrier. */
2545 journal_lock_updates(journal);
2548 * We don't want to clear needs_recovery flag when we failed
2549 * to flush the journal.
2551 error = journal_flush(journal);
2552 if (error < 0)
2553 goto out;
2555 /* Journal blocked and flushed, clear needs_recovery flag. */
2556 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2557 error = ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2558 if (error)
2559 goto out;
2561 return 0;
2563 out:
2564 journal_unlock_updates(journal);
2565 return error;
2569 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2570 * flag here, even though the filesystem is not technically dirty yet.
2572 static int ext3_unfreeze(struct super_block *sb)
2574 if (!(sb->s_flags & MS_RDONLY)) {
2575 lock_super(sb);
2576 /* Reser the needs_recovery flag before the fs is unlocked. */
2577 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2578 ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2579 unlock_super(sb);
2580 journal_unlock_updates(EXT3_SB(sb)->s_journal);
2582 return 0;
2585 static int ext3_remount (struct super_block * sb, int * flags, char * data)
2587 struct ext3_super_block * es;
2588 struct ext3_sb_info *sbi = EXT3_SB(sb);
2589 ext3_fsblk_t n_blocks_count = 0;
2590 unsigned long old_sb_flags;
2591 struct ext3_mount_options old_opts;
2592 int enable_quota = 0;
2593 int err;
2594 #ifdef CONFIG_QUOTA
2595 int i;
2596 #endif
2598 /* Store the original options */
2599 lock_super(sb);
2600 old_sb_flags = sb->s_flags;
2601 old_opts.s_mount_opt = sbi->s_mount_opt;
2602 old_opts.s_resuid = sbi->s_resuid;
2603 old_opts.s_resgid = sbi->s_resgid;
2604 old_opts.s_commit_interval = sbi->s_commit_interval;
2605 #ifdef CONFIG_QUOTA
2606 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2607 for (i = 0; i < MAXQUOTAS; i++)
2608 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2609 #endif
2612 * Allow the "check" option to be passed as a remount option.
2614 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
2615 err = -EINVAL;
2616 goto restore_opts;
2619 if (test_opt(sb, ABORT))
2620 ext3_abort(sb, __func__, "Abort forced by user");
2622 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2623 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
2625 es = sbi->s_es;
2627 ext3_init_journal_params(sb, sbi->s_journal);
2629 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
2630 n_blocks_count > le32_to_cpu(es->s_blocks_count)) {
2631 if (test_opt(sb, ABORT)) {
2632 err = -EROFS;
2633 goto restore_opts;
2636 if (*flags & MS_RDONLY) {
2637 err = dquot_suspend(sb, -1);
2638 if (err < 0)
2639 goto restore_opts;
2642 * First of all, the unconditional stuff we have to do
2643 * to disable replay of the journal when we next remount
2645 sb->s_flags |= MS_RDONLY;
2648 * OK, test if we are remounting a valid rw partition
2649 * readonly, and if so set the rdonly flag and then
2650 * mark the partition as valid again.
2652 if (!(es->s_state & cpu_to_le16(EXT3_VALID_FS)) &&
2653 (sbi->s_mount_state & EXT3_VALID_FS))
2654 es->s_state = cpu_to_le16(sbi->s_mount_state);
2656 ext3_mark_recovery_complete(sb, es);
2657 } else {
2658 __le32 ret;
2659 if ((ret = EXT3_HAS_RO_COMPAT_FEATURE(sb,
2660 ~EXT3_FEATURE_RO_COMPAT_SUPP))) {
2661 ext3_msg(sb, KERN_WARNING,
2662 "warning: couldn't remount RDWR "
2663 "because of unsupported optional "
2664 "features (%x)", le32_to_cpu(ret));
2665 err = -EROFS;
2666 goto restore_opts;
2670 * If we have an unprocessed orphan list hanging
2671 * around from a previously readonly bdev mount,
2672 * require a full umount/remount for now.
2674 if (es->s_last_orphan) {
2675 ext3_msg(sb, KERN_WARNING, "warning: couldn't "
2676 "remount RDWR because of unprocessed "
2677 "orphan inode list. Please "
2678 "umount/remount instead.");
2679 err = -EINVAL;
2680 goto restore_opts;
2684 * Mounting a RDONLY partition read-write, so reread
2685 * and store the current valid flag. (It may have
2686 * been changed by e2fsck since we originally mounted
2687 * the partition.)
2689 ext3_clear_journal_err(sb, es);
2690 sbi->s_mount_state = le16_to_cpu(es->s_state);
2691 if ((err = ext3_group_extend(sb, es, n_blocks_count)))
2692 goto restore_opts;
2693 if (!ext3_setup_super (sb, es, 0))
2694 sb->s_flags &= ~MS_RDONLY;
2695 enable_quota = 1;
2698 #ifdef CONFIG_QUOTA
2699 /* Release old quota file names */
2700 for (i = 0; i < MAXQUOTAS; i++)
2701 if (old_opts.s_qf_names[i] &&
2702 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2703 kfree(old_opts.s_qf_names[i]);
2704 #endif
2705 unlock_super(sb);
2707 if (enable_quota)
2708 dquot_resume(sb, -1);
2709 return 0;
2710 restore_opts:
2711 sb->s_flags = old_sb_flags;
2712 sbi->s_mount_opt = old_opts.s_mount_opt;
2713 sbi->s_resuid = old_opts.s_resuid;
2714 sbi->s_resgid = old_opts.s_resgid;
2715 sbi->s_commit_interval = old_opts.s_commit_interval;
2716 #ifdef CONFIG_QUOTA
2717 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
2718 for (i = 0; i < MAXQUOTAS; i++) {
2719 if (sbi->s_qf_names[i] &&
2720 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2721 kfree(sbi->s_qf_names[i]);
2722 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
2724 #endif
2725 unlock_super(sb);
2726 return err;
2729 static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf)
2731 struct super_block *sb = dentry->d_sb;
2732 struct ext3_sb_info *sbi = EXT3_SB(sb);
2733 struct ext3_super_block *es = sbi->s_es;
2734 u64 fsid;
2736 if (test_opt(sb, MINIX_DF)) {
2737 sbi->s_overhead_last = 0;
2738 } else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
2739 unsigned long ngroups = sbi->s_groups_count, i;
2740 ext3_fsblk_t overhead = 0;
2741 smp_rmb();
2744 * Compute the overhead (FS structures). This is constant
2745 * for a given filesystem unless the number of block groups
2746 * changes so we cache the previous value until it does.
2750 * All of the blocks before first_data_block are
2751 * overhead
2753 overhead = le32_to_cpu(es->s_first_data_block);
2756 * Add the overhead attributed to the superblock and
2757 * block group descriptors. If the sparse superblocks
2758 * feature is turned on, then not all groups have this.
2760 for (i = 0; i < ngroups; i++) {
2761 overhead += ext3_bg_has_super(sb, i) +
2762 ext3_bg_num_gdb(sb, i);
2763 cond_resched();
2767 * Every block group has an inode bitmap, a block
2768 * bitmap, and an inode table.
2770 overhead += ngroups * (2 + sbi->s_itb_per_group);
2771 sbi->s_overhead_last = overhead;
2772 smp_wmb();
2773 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
2776 buf->f_type = EXT3_SUPER_MAGIC;
2777 buf->f_bsize = sb->s_blocksize;
2778 buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
2779 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter);
2780 buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
2781 if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
2782 buf->f_bavail = 0;
2783 buf->f_files = le32_to_cpu(es->s_inodes_count);
2784 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
2785 buf->f_namelen = EXT3_NAME_LEN;
2786 fsid = le64_to_cpup((void *)es->s_uuid) ^
2787 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
2788 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
2789 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
2790 return 0;
2793 /* Helper function for writing quotas on sync - we need to start transaction before quota file
2794 * is locked for write. Otherwise the are possible deadlocks:
2795 * Process 1 Process 2
2796 * ext3_create() quota_sync()
2797 * journal_start() write_dquot()
2798 * dquot_initialize() down(dqio_mutex)
2799 * down(dqio_mutex) journal_start()
2803 #ifdef CONFIG_QUOTA
2805 static inline struct inode *dquot_to_inode(struct dquot *dquot)
2807 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
2810 static int ext3_write_dquot(struct dquot *dquot)
2812 int ret, err;
2813 handle_t *handle;
2814 struct inode *inode;
2816 inode = dquot_to_inode(dquot);
2817 handle = ext3_journal_start(inode,
2818 EXT3_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
2819 if (IS_ERR(handle))
2820 return PTR_ERR(handle);
2821 ret = dquot_commit(dquot);
2822 err = ext3_journal_stop(handle);
2823 if (!ret)
2824 ret = err;
2825 return ret;
2828 static int ext3_acquire_dquot(struct dquot *dquot)
2830 int ret, err;
2831 handle_t *handle;
2833 handle = ext3_journal_start(dquot_to_inode(dquot),
2834 EXT3_QUOTA_INIT_BLOCKS(dquot->dq_sb));
2835 if (IS_ERR(handle))
2836 return PTR_ERR(handle);
2837 ret = dquot_acquire(dquot);
2838 err = ext3_journal_stop(handle);
2839 if (!ret)
2840 ret = err;
2841 return ret;
2844 static int ext3_release_dquot(struct dquot *dquot)
2846 int ret, err;
2847 handle_t *handle;
2849 handle = ext3_journal_start(dquot_to_inode(dquot),
2850 EXT3_QUOTA_DEL_BLOCKS(dquot->dq_sb));
2851 if (IS_ERR(handle)) {
2852 /* Release dquot anyway to avoid endless cycle in dqput() */
2853 dquot_release(dquot);
2854 return PTR_ERR(handle);
2856 ret = dquot_release(dquot);
2857 err = ext3_journal_stop(handle);
2858 if (!ret)
2859 ret = err;
2860 return ret;
2863 static int ext3_mark_dquot_dirty(struct dquot *dquot)
2865 /* Are we journaling quotas? */
2866 if (EXT3_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
2867 EXT3_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
2868 dquot_mark_dquot_dirty(dquot);
2869 return ext3_write_dquot(dquot);
2870 } else {
2871 return dquot_mark_dquot_dirty(dquot);
2875 static int ext3_write_info(struct super_block *sb, int type)
2877 int ret, err;
2878 handle_t *handle;
2880 /* Data block + inode block */
2881 handle = ext3_journal_start(sb->s_root->d_inode, 2);
2882 if (IS_ERR(handle))
2883 return PTR_ERR(handle);
2884 ret = dquot_commit_info(sb, type);
2885 err = ext3_journal_stop(handle);
2886 if (!ret)
2887 ret = err;
2888 return ret;
2892 * Turn on quotas during mount time - we need to find
2893 * the quota file and such...
2895 static int ext3_quota_on_mount(struct super_block *sb, int type)
2897 return dquot_quota_on_mount(sb, EXT3_SB(sb)->s_qf_names[type],
2898 EXT3_SB(sb)->s_jquota_fmt, type);
2902 * Standard function to be called on quota_on
2904 static int ext3_quota_on(struct super_block *sb, int type, int format_id,
2905 struct path *path)
2907 int err;
2909 if (!test_opt(sb, QUOTA))
2910 return -EINVAL;
2912 /* Quotafile not on the same filesystem? */
2913 if (path->mnt->mnt_sb != sb)
2914 return -EXDEV;
2915 /* Journaling quota? */
2916 if (EXT3_SB(sb)->s_qf_names[type]) {
2917 /* Quotafile not of fs root? */
2918 if (path->dentry->d_parent != sb->s_root)
2919 ext3_msg(sb, KERN_WARNING,
2920 "warning: Quota file not on filesystem root. "
2921 "Journaled quota will not work.");
2925 * When we journal data on quota file, we have to flush journal to see
2926 * all updates to the file when we bypass pagecache...
2928 if (ext3_should_journal_data(path->dentry->d_inode)) {
2930 * We don't need to lock updates but journal_flush() could
2931 * otherwise be livelocked...
2933 journal_lock_updates(EXT3_SB(sb)->s_journal);
2934 err = journal_flush(EXT3_SB(sb)->s_journal);
2935 journal_unlock_updates(EXT3_SB(sb)->s_journal);
2936 if (err)
2937 return err;
2940 return dquot_quota_on(sb, type, format_id, path);
2943 /* Read data from quotafile - avoid pagecache and such because we cannot afford
2944 * acquiring the locks... As quota files are never truncated and quota code
2945 * itself serializes the operations (and no one else should touch the files)
2946 * we don't have to be afraid of races */
2947 static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
2948 size_t len, loff_t off)
2950 struct inode *inode = sb_dqopt(sb)->files[type];
2951 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2952 int err = 0;
2953 int offset = off & (sb->s_blocksize - 1);
2954 int tocopy;
2955 size_t toread;
2956 struct buffer_head *bh;
2957 loff_t i_size = i_size_read(inode);
2959 if (off > i_size)
2960 return 0;
2961 if (off+len > i_size)
2962 len = i_size-off;
2963 toread = len;
2964 while (toread > 0) {
2965 tocopy = sb->s_blocksize - offset < toread ?
2966 sb->s_blocksize - offset : toread;
2967 bh = ext3_bread(NULL, inode, blk, 0, &err);
2968 if (err)
2969 return err;
2970 if (!bh) /* A hole? */
2971 memset(data, 0, tocopy);
2972 else
2973 memcpy(data, bh->b_data+offset, tocopy);
2974 brelse(bh);
2975 offset = 0;
2976 toread -= tocopy;
2977 data += tocopy;
2978 blk++;
2980 return len;
2983 /* Write to quotafile (we know the transaction is already started and has
2984 * enough credits) */
2985 static ssize_t ext3_quota_write(struct super_block *sb, int type,
2986 const char *data, size_t len, loff_t off)
2988 struct inode *inode = sb_dqopt(sb)->files[type];
2989 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2990 int err = 0;
2991 int offset = off & (sb->s_blocksize - 1);
2992 int journal_quota = EXT3_SB(sb)->s_qf_names[type] != NULL;
2993 struct buffer_head *bh;
2994 handle_t *handle = journal_current_handle();
2996 if (!handle) {
2997 ext3_msg(sb, KERN_WARNING,
2998 "warning: quota write (off=%llu, len=%llu)"
2999 " cancelled because transaction is not started.",
3000 (unsigned long long)off, (unsigned long long)len);
3001 return -EIO;
3005 * Since we account only one data block in transaction credits,
3006 * then it is impossible to cross a block boundary.
3008 if (sb->s_blocksize - offset < len) {
3009 ext3_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
3010 " cancelled because not block aligned",
3011 (unsigned long long)off, (unsigned long long)len);
3012 return -EIO;
3014 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3015 bh = ext3_bread(handle, inode, blk, 1, &err);
3016 if (!bh)
3017 goto out;
3018 if (journal_quota) {
3019 err = ext3_journal_get_write_access(handle, bh);
3020 if (err) {
3021 brelse(bh);
3022 goto out;
3025 lock_buffer(bh);
3026 memcpy(bh->b_data+offset, data, len);
3027 flush_dcache_page(bh->b_page);
3028 unlock_buffer(bh);
3029 if (journal_quota)
3030 err = ext3_journal_dirty_metadata(handle, bh);
3031 else {
3032 /* Always do at least ordered writes for quotas */
3033 err = ext3_journal_dirty_data(handle, bh);
3034 mark_buffer_dirty(bh);
3036 brelse(bh);
3037 out:
3038 if (err) {
3039 mutex_unlock(&inode->i_mutex);
3040 return err;
3042 if (inode->i_size < off + len) {
3043 i_size_write(inode, off + len);
3044 EXT3_I(inode)->i_disksize = inode->i_size;
3046 inode->i_version++;
3047 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3048 ext3_mark_inode_dirty(handle, inode);
3049 mutex_unlock(&inode->i_mutex);
3050 return len;
3053 #endif
3055 static struct dentry *ext3_mount(struct file_system_type *fs_type,
3056 int flags, const char *dev_name, void *data)
3058 return mount_bdev(fs_type, flags, dev_name, data, ext3_fill_super);
3061 static struct file_system_type ext3_fs_type = {
3062 .owner = THIS_MODULE,
3063 .name = "ext3",
3064 .mount = ext3_mount,
3065 .kill_sb = kill_block_super,
3066 .fs_flags = FS_REQUIRES_DEV,
3069 static int __init init_ext3_fs(void)
3071 int err = init_ext3_xattr();
3072 if (err)
3073 return err;
3074 err = init_inodecache();
3075 if (err)
3076 goto out1;
3077 err = register_filesystem(&ext3_fs_type);
3078 if (err)
3079 goto out;
3080 return 0;
3081 out:
3082 destroy_inodecache();
3083 out1:
3084 exit_ext3_xattr();
3085 return err;
3088 static void __exit exit_ext3_fs(void)
3090 unregister_filesystem(&ext3_fs_type);
3091 destroy_inodecache();
3092 exit_ext3_xattr();
3095 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
3096 MODULE_DESCRIPTION("Second Extended Filesystem with journaling extensions");
3097 MODULE_LICENSE("GPL");
3098 module_init(init_ext3_fs)
3099 module_exit(exit_ext3_fs)