quota: Fix deadlock during path resolution
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / fs / ext3 / super.c
blob0e0d391626be3e1aacb7e6b227699ff623b546c8
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>
40 #include <asm/uaccess.h>
42 #include "xattr.h"
43 #include "acl.h"
44 #include "namei.h"
46 #ifdef CONFIG_EXT3_DEFAULTS_TO_ORDERED
47 #define EXT3_MOUNT_DEFAULT_DATA_MODE EXT3_MOUNT_ORDERED_DATA
48 #else
49 #define EXT3_MOUNT_DEFAULT_DATA_MODE EXT3_MOUNT_WRITEBACK_DATA
50 #endif
52 static int ext3_load_journal(struct super_block *, struct ext3_super_block *,
53 unsigned long journal_devnum);
54 static int ext3_create_journal(struct super_block *, struct ext3_super_block *,
55 unsigned int);
56 static int ext3_commit_super(struct super_block *sb,
57 struct ext3_super_block *es,
58 int sync);
59 static void ext3_mark_recovery_complete(struct super_block * sb,
60 struct ext3_super_block * es);
61 static void ext3_clear_journal_err(struct super_block * sb,
62 struct ext3_super_block * es);
63 static int ext3_sync_fs(struct super_block *sb, int wait);
64 static const char *ext3_decode_error(struct super_block * sb, int errno,
65 char nbuf[16]);
66 static int ext3_remount (struct super_block * sb, int * flags, char * data);
67 static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf);
68 static int ext3_unfreeze(struct super_block *sb);
69 static int ext3_freeze(struct super_block *sb);
72 * Wrappers for journal_start/end.
74 * The only special thing we need to do here is to make sure that all
75 * journal_end calls result in the superblock being marked dirty, so
76 * that sync() will call the filesystem's write_super callback if
77 * appropriate.
79 handle_t *ext3_journal_start_sb(struct super_block *sb, int nblocks)
81 journal_t *journal;
83 if (sb->s_flags & MS_RDONLY)
84 return ERR_PTR(-EROFS);
86 /* Special case here: if the journal has aborted behind our
87 * backs (eg. EIO in the commit thread), then we still need to
88 * take the FS itself readonly cleanly. */
89 journal = EXT3_SB(sb)->s_journal;
90 if (is_journal_aborted(journal)) {
91 ext3_abort(sb, __func__,
92 "Detected aborted journal");
93 return ERR_PTR(-EROFS);
96 return journal_start(journal, nblocks);
100 * The only special thing we need to do here is to make sure that all
101 * journal_stop calls result in the superblock being marked dirty, so
102 * that sync() will call the filesystem's write_super callback if
103 * appropriate.
105 int __ext3_journal_stop(const char *where, handle_t *handle)
107 struct super_block *sb;
108 int err;
109 int rc;
111 sb = handle->h_transaction->t_journal->j_private;
112 err = handle->h_err;
113 rc = journal_stop(handle);
115 if (!err)
116 err = rc;
117 if (err)
118 __ext3_std_error(sb, where, err);
119 return err;
122 void ext3_journal_abort_handle(const char *caller, const char *err_fn,
123 struct buffer_head *bh, handle_t *handle, int err)
125 char nbuf[16];
126 const char *errstr = ext3_decode_error(NULL, err, nbuf);
128 if (bh)
129 BUFFER_TRACE(bh, "abort");
131 if (!handle->h_err)
132 handle->h_err = err;
134 if (is_handle_aborted(handle))
135 return;
137 printk(KERN_ERR "EXT3-fs: %s: aborting transaction: %s in %s\n",
138 caller, errstr, err_fn);
140 journal_abort_handle(handle);
143 void ext3_msg(struct super_block *sb, const char *prefix,
144 const char *fmt, ...)
146 struct va_format vaf;
147 va_list args;
149 va_start(args, fmt);
151 vaf.fmt = fmt;
152 vaf.va = &args;
154 printk("%sEXT3-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
156 va_end(args);
159 /* Deal with the reporting of failure conditions on a filesystem such as
160 * inconsistencies detected or read IO failures.
162 * On ext2, we can store the error state of the filesystem in the
163 * superblock. That is not possible on ext3, because we may have other
164 * write ordering constraints on the superblock which prevent us from
165 * writing it out straight away; and given that the journal is about to
166 * be aborted, we can't rely on the current, or future, transactions to
167 * write out the superblock safely.
169 * We'll just use the journal_abort() error code to record an error in
170 * the journal instead. On recovery, the journal will complain about
171 * that error until we've noted it down and cleared it.
174 static void ext3_handle_error(struct super_block *sb)
176 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
178 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
179 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
181 if (sb->s_flags & MS_RDONLY)
182 return;
184 if (!test_opt (sb, ERRORS_CONT)) {
185 journal_t *journal = EXT3_SB(sb)->s_journal;
187 set_opt(EXT3_SB(sb)->s_mount_opt, ABORT);
188 if (journal)
189 journal_abort(journal, -EIO);
191 if (test_opt (sb, ERRORS_RO)) {
192 ext3_msg(sb, KERN_CRIT,
193 "error: remounting filesystem read-only");
194 sb->s_flags |= MS_RDONLY;
196 ext3_commit_super(sb, es, 1);
197 if (test_opt(sb, ERRORS_PANIC))
198 panic("EXT3-fs (%s): panic forced after error\n",
199 sb->s_id);
202 void ext3_error(struct super_block *sb, const char *function,
203 const char *fmt, ...)
205 struct va_format vaf;
206 va_list args;
208 va_start(args, fmt);
210 vaf.fmt = fmt;
211 vaf.va = &args;
213 printk(KERN_CRIT "EXT3-fs error (device %s): %s: %pV\n",
214 sb->s_id, function, &vaf);
216 va_end(args);
218 ext3_handle_error(sb);
221 static const char *ext3_decode_error(struct super_block * sb, int errno,
222 char nbuf[16])
224 char *errstr = NULL;
226 switch (errno) {
227 case -EIO:
228 errstr = "IO failure";
229 break;
230 case -ENOMEM:
231 errstr = "Out of memory";
232 break;
233 case -EROFS:
234 if (!sb || EXT3_SB(sb)->s_journal->j_flags & JFS_ABORT)
235 errstr = "Journal has aborted";
236 else
237 errstr = "Readonly filesystem";
238 break;
239 default:
240 /* If the caller passed in an extra buffer for unknown
241 * errors, textualise them now. Else we just return
242 * NULL. */
243 if (nbuf) {
244 /* Check for truncated error codes... */
245 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
246 errstr = nbuf;
248 break;
251 return errstr;
254 /* __ext3_std_error decodes expected errors from journaling functions
255 * automatically and invokes the appropriate error response. */
257 void __ext3_std_error (struct super_block * sb, const char * function,
258 int errno)
260 char nbuf[16];
261 const char *errstr;
263 /* Special case: if the error is EROFS, and we're not already
264 * inside a transaction, then there's really no point in logging
265 * an error. */
266 if (errno == -EROFS && journal_current_handle() == NULL &&
267 (sb->s_flags & MS_RDONLY))
268 return;
270 errstr = ext3_decode_error(sb, errno, nbuf);
271 ext3_msg(sb, KERN_CRIT, "error in %s: %s", function, errstr);
273 ext3_handle_error(sb);
277 * ext3_abort is a much stronger failure handler than ext3_error. The
278 * abort function may be used to deal with unrecoverable failures such
279 * as journal IO errors or ENOMEM at a critical moment in log management.
281 * We unconditionally force the filesystem into an ABORT|READONLY state,
282 * unless the error response on the fs has been set to panic in which
283 * case we take the easy way out and panic immediately.
286 void ext3_abort(struct super_block *sb, const char *function,
287 const char *fmt, ...)
289 struct va_format vaf;
290 va_list args;
292 va_start(args, fmt);
294 vaf.fmt = fmt;
295 vaf.va = &args;
297 printk(KERN_CRIT "EXT3-fs (%s): error: %s: %pV\n",
298 sb->s_id, function, &vaf);
300 va_end(args);
302 if (test_opt(sb, ERRORS_PANIC))
303 panic("EXT3-fs: panic from previous error\n");
305 if (sb->s_flags & MS_RDONLY)
306 return;
308 ext3_msg(sb, KERN_CRIT,
309 "error: remounting filesystem read-only");
310 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
311 sb->s_flags |= MS_RDONLY;
312 set_opt(EXT3_SB(sb)->s_mount_opt, ABORT);
313 if (EXT3_SB(sb)->s_journal)
314 journal_abort(EXT3_SB(sb)->s_journal, -EIO);
317 void ext3_warning(struct super_block *sb, const char *function,
318 const char *fmt, ...)
320 struct va_format vaf;
321 va_list args;
323 va_start(args, fmt);
325 vaf.fmt = fmt;
326 vaf.va = &args;
328 printk(KERN_WARNING "EXT3-fs (%s): warning: %s: %pV\n",
329 sb->s_id, function, &vaf);
331 va_end(args);
334 void ext3_update_dynamic_rev(struct super_block *sb)
336 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
338 if (le32_to_cpu(es->s_rev_level) > EXT3_GOOD_OLD_REV)
339 return;
341 ext3_msg(sb, KERN_WARNING,
342 "warning: updating to rev %d because of "
343 "new feature flag, running e2fsck is recommended",
344 EXT3_DYNAMIC_REV);
346 es->s_first_ino = cpu_to_le32(EXT3_GOOD_OLD_FIRST_INO);
347 es->s_inode_size = cpu_to_le16(EXT3_GOOD_OLD_INODE_SIZE);
348 es->s_rev_level = cpu_to_le32(EXT3_DYNAMIC_REV);
349 /* leave es->s_feature_*compat flags alone */
350 /* es->s_uuid will be set by e2fsck if empty */
353 * The rest of the superblock fields should be zero, and if not it
354 * means they are likely already in use, so leave them alone. We
355 * can leave it up to e2fsck to clean up any inconsistencies there.
360 * Open the external journal device
362 static struct block_device *ext3_blkdev_get(dev_t dev, struct super_block *sb)
364 struct block_device *bdev;
365 char b[BDEVNAME_SIZE];
367 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
368 if (IS_ERR(bdev))
369 goto fail;
370 return bdev;
372 fail:
373 ext3_msg(sb, "error: failed to open journal device %s: %ld",
374 __bdevname(dev, b), PTR_ERR(bdev));
376 return NULL;
380 * Release the journal device
382 static int ext3_blkdev_put(struct block_device *bdev)
384 bd_release(bdev);
385 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
388 static int ext3_blkdev_remove(struct ext3_sb_info *sbi)
390 struct block_device *bdev;
391 int ret = -ENODEV;
393 bdev = sbi->journal_bdev;
394 if (bdev) {
395 ret = ext3_blkdev_put(bdev);
396 sbi->journal_bdev = NULL;
398 return ret;
401 static inline struct inode *orphan_list_entry(struct list_head *l)
403 return &list_entry(l, struct ext3_inode_info, i_orphan)->vfs_inode;
406 static void dump_orphan_list(struct super_block *sb, struct ext3_sb_info *sbi)
408 struct list_head *l;
410 ext3_msg(sb, KERN_ERR, "error: sb orphan head is %d",
411 le32_to_cpu(sbi->s_es->s_last_orphan));
413 ext3_msg(sb, KERN_ERR, "sb_info orphan list:");
414 list_for_each(l, &sbi->s_orphan) {
415 struct inode *inode = orphan_list_entry(l);
416 ext3_msg(sb, KERN_ERR, " "
417 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
418 inode->i_sb->s_id, inode->i_ino, inode,
419 inode->i_mode, inode->i_nlink,
420 NEXT_ORPHAN(inode));
424 static void ext3_put_super (struct super_block * sb)
426 struct ext3_sb_info *sbi = EXT3_SB(sb);
427 struct ext3_super_block *es = sbi->s_es;
428 int i, err;
430 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
431 ext3_xattr_put_super(sb);
432 err = journal_destroy(sbi->s_journal);
433 sbi->s_journal = NULL;
434 if (err < 0)
435 ext3_abort(sb, __func__, "Couldn't clean up the journal");
437 if (!(sb->s_flags & MS_RDONLY)) {
438 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
439 es->s_state = cpu_to_le16(sbi->s_mount_state);
440 BUFFER_TRACE(sbi->s_sbh, "marking dirty");
441 mark_buffer_dirty(sbi->s_sbh);
442 ext3_commit_super(sb, es, 1);
445 for (i = 0; i < sbi->s_gdb_count; i++)
446 brelse(sbi->s_group_desc[i]);
447 kfree(sbi->s_group_desc);
448 percpu_counter_destroy(&sbi->s_freeblocks_counter);
449 percpu_counter_destroy(&sbi->s_freeinodes_counter);
450 percpu_counter_destroy(&sbi->s_dirs_counter);
451 brelse(sbi->s_sbh);
452 #ifdef CONFIG_QUOTA
453 for (i = 0; i < MAXQUOTAS; i++)
454 kfree(sbi->s_qf_names[i]);
455 #endif
457 /* Debugging code just in case the in-memory inode orphan list
458 * isn't empty. The on-disk one can be non-empty if we've
459 * detected an error and taken the fs readonly, but the
460 * in-memory list had better be clean by this point. */
461 if (!list_empty(&sbi->s_orphan))
462 dump_orphan_list(sb, sbi);
463 J_ASSERT(list_empty(&sbi->s_orphan));
465 invalidate_bdev(sb->s_bdev);
466 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
468 * Invalidate the journal device's buffers. We don't want them
469 * floating about in memory - the physical journal device may
470 * hotswapped, and it breaks the `ro-after' testing code.
472 sync_blockdev(sbi->journal_bdev);
473 invalidate_bdev(sbi->journal_bdev);
474 ext3_blkdev_remove(sbi);
476 sb->s_fs_info = NULL;
477 kfree(sbi->s_blockgroup_lock);
478 kfree(sbi);
481 static struct kmem_cache *ext3_inode_cachep;
484 * Called inside transaction, so use GFP_NOFS
486 static struct inode *ext3_alloc_inode(struct super_block *sb)
488 struct ext3_inode_info *ei;
490 ei = kmem_cache_alloc(ext3_inode_cachep, GFP_NOFS);
491 if (!ei)
492 return NULL;
493 ei->i_block_alloc_info = NULL;
494 ei->vfs_inode.i_version = 1;
495 atomic_set(&ei->i_datasync_tid, 0);
496 atomic_set(&ei->i_sync_tid, 0);
497 return &ei->vfs_inode;
500 static void ext3_i_callback(struct rcu_head *head)
502 struct inode *inode = container_of(head, struct inode, i_rcu);
503 INIT_LIST_HEAD(&inode->i_dentry);
504 kmem_cache_free(ext3_inode_cachep, EXT3_I(inode));
507 static void ext3_destroy_inode(struct inode *inode)
509 if (!list_empty(&(EXT3_I(inode)->i_orphan))) {
510 printk("EXT3 Inode %p: orphan list check failed!\n",
511 EXT3_I(inode));
512 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
513 EXT3_I(inode), sizeof(struct ext3_inode_info),
514 false);
515 dump_stack();
517 call_rcu(&inode->i_rcu, ext3_i_callback);
520 static void init_once(void *foo)
522 struct ext3_inode_info *ei = (struct ext3_inode_info *) foo;
524 INIT_LIST_HEAD(&ei->i_orphan);
525 #ifdef CONFIG_EXT3_FS_XATTR
526 init_rwsem(&ei->xattr_sem);
527 #endif
528 mutex_init(&ei->truncate_mutex);
529 inode_init_once(&ei->vfs_inode);
532 static int init_inodecache(void)
534 ext3_inode_cachep = kmem_cache_create("ext3_inode_cache",
535 sizeof(struct ext3_inode_info),
536 0, (SLAB_RECLAIM_ACCOUNT|
537 SLAB_MEM_SPREAD),
538 init_once);
539 if (ext3_inode_cachep == NULL)
540 return -ENOMEM;
541 return 0;
544 static void destroy_inodecache(void)
546 kmem_cache_destroy(ext3_inode_cachep);
549 static inline void ext3_show_quota_options(struct seq_file *seq, struct super_block *sb)
551 #if defined(CONFIG_QUOTA)
552 struct ext3_sb_info *sbi = EXT3_SB(sb);
554 if (sbi->s_jquota_fmt) {
555 char *fmtname = "";
557 switch (sbi->s_jquota_fmt) {
558 case QFMT_VFS_OLD:
559 fmtname = "vfsold";
560 break;
561 case QFMT_VFS_V0:
562 fmtname = "vfsv0";
563 break;
564 case QFMT_VFS_V1:
565 fmtname = "vfsv1";
566 break;
568 seq_printf(seq, ",jqfmt=%s", fmtname);
571 if (sbi->s_qf_names[USRQUOTA])
572 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
574 if (sbi->s_qf_names[GRPQUOTA])
575 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
577 if (test_opt(sb, USRQUOTA))
578 seq_puts(seq, ",usrquota");
580 if (test_opt(sb, GRPQUOTA))
581 seq_puts(seq, ",grpquota");
582 #endif
585 static char *data_mode_string(unsigned long mode)
587 switch (mode) {
588 case EXT3_MOUNT_JOURNAL_DATA:
589 return "journal";
590 case EXT3_MOUNT_ORDERED_DATA:
591 return "ordered";
592 case EXT3_MOUNT_WRITEBACK_DATA:
593 return "writeback";
595 return "unknown";
599 * Show an option if
600 * - it's set to a non-default value OR
601 * - if the per-sb default is different from the global default
603 static int ext3_show_options(struct seq_file *seq, struct vfsmount *vfs)
605 struct super_block *sb = vfs->mnt_sb;
606 struct ext3_sb_info *sbi = EXT3_SB(sb);
607 struct ext3_super_block *es = sbi->s_es;
608 unsigned long def_mount_opts;
610 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
612 if (sbi->s_sb_block != 1)
613 seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
614 if (test_opt(sb, MINIX_DF))
615 seq_puts(seq, ",minixdf");
616 if (test_opt(sb, GRPID))
617 seq_puts(seq, ",grpid");
618 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT3_DEFM_BSDGROUPS))
619 seq_puts(seq, ",nogrpid");
620 if (sbi->s_resuid != EXT3_DEF_RESUID ||
621 le16_to_cpu(es->s_def_resuid) != EXT3_DEF_RESUID) {
622 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
624 if (sbi->s_resgid != EXT3_DEF_RESGID ||
625 le16_to_cpu(es->s_def_resgid) != EXT3_DEF_RESGID) {
626 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
628 if (test_opt(sb, ERRORS_RO)) {
629 int def_errors = le16_to_cpu(es->s_errors);
631 if (def_errors == EXT3_ERRORS_PANIC ||
632 def_errors == EXT3_ERRORS_CONTINUE) {
633 seq_puts(seq, ",errors=remount-ro");
636 if (test_opt(sb, ERRORS_CONT))
637 seq_puts(seq, ",errors=continue");
638 if (test_opt(sb, ERRORS_PANIC))
639 seq_puts(seq, ",errors=panic");
640 if (test_opt(sb, NO_UID32))
641 seq_puts(seq, ",nouid32");
642 if (test_opt(sb, DEBUG))
643 seq_puts(seq, ",debug");
644 if (test_opt(sb, OLDALLOC))
645 seq_puts(seq, ",oldalloc");
646 #ifdef CONFIG_EXT3_FS_XATTR
647 if (test_opt(sb, XATTR_USER))
648 seq_puts(seq, ",user_xattr");
649 if (!test_opt(sb, XATTR_USER) &&
650 (def_mount_opts & EXT3_DEFM_XATTR_USER)) {
651 seq_puts(seq, ",nouser_xattr");
653 #endif
654 #ifdef CONFIG_EXT3_FS_POSIX_ACL
655 if (test_opt(sb, POSIX_ACL))
656 seq_puts(seq, ",acl");
657 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT3_DEFM_ACL))
658 seq_puts(seq, ",noacl");
659 #endif
660 if (!test_opt(sb, RESERVATION))
661 seq_puts(seq, ",noreservation");
662 if (sbi->s_commit_interval) {
663 seq_printf(seq, ",commit=%u",
664 (unsigned) (sbi->s_commit_interval / HZ));
668 * Always display barrier state so it's clear what the status is.
670 seq_puts(seq, ",barrier=");
671 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
672 seq_printf(seq, ",data=%s", data_mode_string(test_opt(sb, DATA_FLAGS)));
673 if (test_opt(sb, DATA_ERR_ABORT))
674 seq_puts(seq, ",data_err=abort");
676 if (test_opt(sb, NOLOAD))
677 seq_puts(seq, ",norecovery");
679 ext3_show_quota_options(seq, sb);
681 return 0;
685 static struct inode *ext3_nfs_get_inode(struct super_block *sb,
686 u64 ino, u32 generation)
688 struct inode *inode;
690 if (ino < EXT3_FIRST_INO(sb) && ino != EXT3_ROOT_INO)
691 return ERR_PTR(-ESTALE);
692 if (ino > le32_to_cpu(EXT3_SB(sb)->s_es->s_inodes_count))
693 return ERR_PTR(-ESTALE);
695 /* iget isn't really right if the inode is currently unallocated!!
697 * ext3_read_inode will return a bad_inode if the inode had been
698 * deleted, so we should be safe.
700 * Currently we don't know the generation for parent directory, so
701 * a generation of 0 means "accept any"
703 inode = ext3_iget(sb, ino);
704 if (IS_ERR(inode))
705 return ERR_CAST(inode);
706 if (generation && inode->i_generation != generation) {
707 iput(inode);
708 return ERR_PTR(-ESTALE);
711 return inode;
714 static struct dentry *ext3_fh_to_dentry(struct super_block *sb, struct fid *fid,
715 int fh_len, int fh_type)
717 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
718 ext3_nfs_get_inode);
721 static struct dentry *ext3_fh_to_parent(struct super_block *sb, struct fid *fid,
722 int fh_len, int fh_type)
724 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
725 ext3_nfs_get_inode);
729 * Try to release metadata pages (indirect blocks, directories) which are
730 * mapped via the block device. Since these pages could have journal heads
731 * which would prevent try_to_free_buffers() from freeing them, we must use
732 * jbd layer's try_to_free_buffers() function to release them.
734 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
735 gfp_t wait)
737 journal_t *journal = EXT3_SB(sb)->s_journal;
739 WARN_ON(PageChecked(page));
740 if (!page_has_buffers(page))
741 return 0;
742 if (journal)
743 return journal_try_to_free_buffers(journal, page,
744 wait & ~__GFP_WAIT);
745 return try_to_free_buffers(page);
748 #ifdef CONFIG_QUOTA
749 #define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
750 #define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
752 static int ext3_write_dquot(struct dquot *dquot);
753 static int ext3_acquire_dquot(struct dquot *dquot);
754 static int ext3_release_dquot(struct dquot *dquot);
755 static int ext3_mark_dquot_dirty(struct dquot *dquot);
756 static int ext3_write_info(struct super_block *sb, int type);
757 static int ext3_quota_on(struct super_block *sb, int type, int format_id,
758 struct path *path);
759 static int ext3_quota_on_mount(struct super_block *sb, int type);
760 static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
761 size_t len, loff_t off);
762 static ssize_t ext3_quota_write(struct super_block *sb, int type,
763 const char *data, size_t len, loff_t off);
765 static const struct dquot_operations ext3_quota_operations = {
766 .write_dquot = ext3_write_dquot,
767 .acquire_dquot = ext3_acquire_dquot,
768 .release_dquot = ext3_release_dquot,
769 .mark_dirty = ext3_mark_dquot_dirty,
770 .write_info = ext3_write_info,
771 .alloc_dquot = dquot_alloc,
772 .destroy_dquot = dquot_destroy,
775 static const struct quotactl_ops ext3_qctl_operations = {
776 .quota_on = ext3_quota_on,
777 .quota_off = dquot_quota_off,
778 .quota_sync = dquot_quota_sync,
779 .get_info = dquot_get_dqinfo,
780 .set_info = dquot_set_dqinfo,
781 .get_dqblk = dquot_get_dqblk,
782 .set_dqblk = dquot_set_dqblk
784 #endif
786 static const struct super_operations ext3_sops = {
787 .alloc_inode = ext3_alloc_inode,
788 .destroy_inode = ext3_destroy_inode,
789 .write_inode = ext3_write_inode,
790 .dirty_inode = ext3_dirty_inode,
791 .evict_inode = ext3_evict_inode,
792 .put_super = ext3_put_super,
793 .sync_fs = ext3_sync_fs,
794 .freeze_fs = ext3_freeze,
795 .unfreeze_fs = ext3_unfreeze,
796 .statfs = ext3_statfs,
797 .remount_fs = ext3_remount,
798 .show_options = ext3_show_options,
799 #ifdef CONFIG_QUOTA
800 .quota_read = ext3_quota_read,
801 .quota_write = ext3_quota_write,
802 #endif
803 .bdev_try_to_free_page = bdev_try_to_free_page,
806 static const struct export_operations ext3_export_ops = {
807 .fh_to_dentry = ext3_fh_to_dentry,
808 .fh_to_parent = ext3_fh_to_parent,
809 .get_parent = ext3_get_parent,
812 enum {
813 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
814 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
815 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
816 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
817 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
818 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
819 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
820 Opt_data_err_abort, Opt_data_err_ignore,
821 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
822 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
823 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
824 Opt_resize, Opt_usrquota, Opt_grpquota
827 static const match_table_t tokens = {
828 {Opt_bsd_df, "bsddf"},
829 {Opt_minix_df, "minixdf"},
830 {Opt_grpid, "grpid"},
831 {Opt_grpid, "bsdgroups"},
832 {Opt_nogrpid, "nogrpid"},
833 {Opt_nogrpid, "sysvgroups"},
834 {Opt_resgid, "resgid=%u"},
835 {Opt_resuid, "resuid=%u"},
836 {Opt_sb, "sb=%u"},
837 {Opt_err_cont, "errors=continue"},
838 {Opt_err_panic, "errors=panic"},
839 {Opt_err_ro, "errors=remount-ro"},
840 {Opt_nouid32, "nouid32"},
841 {Opt_nocheck, "nocheck"},
842 {Opt_nocheck, "check=none"},
843 {Opt_debug, "debug"},
844 {Opt_oldalloc, "oldalloc"},
845 {Opt_orlov, "orlov"},
846 {Opt_user_xattr, "user_xattr"},
847 {Opt_nouser_xattr, "nouser_xattr"},
848 {Opt_acl, "acl"},
849 {Opt_noacl, "noacl"},
850 {Opt_reservation, "reservation"},
851 {Opt_noreservation, "noreservation"},
852 {Opt_noload, "noload"},
853 {Opt_noload, "norecovery"},
854 {Opt_nobh, "nobh"},
855 {Opt_bh, "bh"},
856 {Opt_commit, "commit=%u"},
857 {Opt_journal_update, "journal=update"},
858 {Opt_journal_inum, "journal=%u"},
859 {Opt_journal_dev, "journal_dev=%u"},
860 {Opt_abort, "abort"},
861 {Opt_data_journal, "data=journal"},
862 {Opt_data_ordered, "data=ordered"},
863 {Opt_data_writeback, "data=writeback"},
864 {Opt_data_err_abort, "data_err=abort"},
865 {Opt_data_err_ignore, "data_err=ignore"},
866 {Opt_offusrjquota, "usrjquota="},
867 {Opt_usrjquota, "usrjquota=%s"},
868 {Opt_offgrpjquota, "grpjquota="},
869 {Opt_grpjquota, "grpjquota=%s"},
870 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
871 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
872 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
873 {Opt_grpquota, "grpquota"},
874 {Opt_noquota, "noquota"},
875 {Opt_quota, "quota"},
876 {Opt_usrquota, "usrquota"},
877 {Opt_barrier, "barrier=%u"},
878 {Opt_barrier, "barrier"},
879 {Opt_nobarrier, "nobarrier"},
880 {Opt_resize, "resize"},
881 {Opt_err, NULL},
884 static ext3_fsblk_t get_sb_block(void **data, struct super_block *sb)
886 ext3_fsblk_t sb_block;
887 char *options = (char *) *data;
889 if (!options || strncmp(options, "sb=", 3) != 0)
890 return 1; /* Default location */
891 options += 3;
892 /*todo: use simple_strtoll with >32bit ext3 */
893 sb_block = simple_strtoul(options, &options, 0);
894 if (*options && *options != ',') {
895 ext3_msg(sb, "error: invalid sb specification: %s",
896 (char *) *data);
897 return 1;
899 if (*options == ',')
900 options++;
901 *data = (void *) options;
902 return sb_block;
905 #ifdef CONFIG_QUOTA
906 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
908 struct ext3_sb_info *sbi = EXT3_SB(sb);
909 char *qname;
911 if (sb_any_quota_loaded(sb) &&
912 !sbi->s_qf_names[qtype]) {
913 ext3_msg(sb, KERN_ERR,
914 "Cannot change journaled "
915 "quota options when quota turned on");
916 return 0;
918 qname = match_strdup(args);
919 if (!qname) {
920 ext3_msg(sb, KERN_ERR,
921 "Not enough memory for storing quotafile name");
922 return 0;
924 if (sbi->s_qf_names[qtype] &&
925 strcmp(sbi->s_qf_names[qtype], qname)) {
926 ext3_msg(sb, KERN_ERR,
927 "%s quota file already specified", QTYPE2NAME(qtype));
928 kfree(qname);
929 return 0;
931 sbi->s_qf_names[qtype] = qname;
932 if (strchr(sbi->s_qf_names[qtype], '/')) {
933 ext3_msg(sb, KERN_ERR,
934 "quotafile must be on filesystem root");
935 kfree(sbi->s_qf_names[qtype]);
936 sbi->s_qf_names[qtype] = NULL;
937 return 0;
939 set_opt(sbi->s_mount_opt, QUOTA);
940 return 1;
943 static int clear_qf_name(struct super_block *sb, int qtype) {
945 struct ext3_sb_info *sbi = EXT3_SB(sb);
947 if (sb_any_quota_loaded(sb) &&
948 sbi->s_qf_names[qtype]) {
949 ext3_msg(sb, KERN_ERR, "Cannot change journaled quota options"
950 " when quota turned on");
951 return 0;
954 * The space will be released later when all options are confirmed
955 * to be correct
957 sbi->s_qf_names[qtype] = NULL;
958 return 1;
960 #endif
962 static int parse_options (char *options, struct super_block *sb,
963 unsigned int *inum, unsigned long *journal_devnum,
964 ext3_fsblk_t *n_blocks_count, int is_remount)
966 struct ext3_sb_info *sbi = EXT3_SB(sb);
967 char * p;
968 substring_t args[MAX_OPT_ARGS];
969 int data_opt = 0;
970 int option;
971 #ifdef CONFIG_QUOTA
972 int qfmt;
973 #endif
975 if (!options)
976 return 1;
978 while ((p = strsep (&options, ",")) != NULL) {
979 int token;
980 if (!*p)
981 continue;
983 * Initialize args struct so we know whether arg was
984 * found; some options take optional arguments.
986 args[0].to = args[0].from = 0;
987 token = match_token(p, tokens, args);
988 switch (token) {
989 case Opt_bsd_df:
990 clear_opt (sbi->s_mount_opt, MINIX_DF);
991 break;
992 case Opt_minix_df:
993 set_opt (sbi->s_mount_opt, MINIX_DF);
994 break;
995 case Opt_grpid:
996 set_opt (sbi->s_mount_opt, GRPID);
997 break;
998 case Opt_nogrpid:
999 clear_opt (sbi->s_mount_opt, GRPID);
1000 break;
1001 case Opt_resuid:
1002 if (match_int(&args[0], &option))
1003 return 0;
1004 sbi->s_resuid = option;
1005 break;
1006 case Opt_resgid:
1007 if (match_int(&args[0], &option))
1008 return 0;
1009 sbi->s_resgid = option;
1010 break;
1011 case Opt_sb:
1012 /* handled by get_sb_block() instead of here */
1013 /* *sb_block = match_int(&args[0]); */
1014 break;
1015 case Opt_err_panic:
1016 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1017 clear_opt (sbi->s_mount_opt, ERRORS_RO);
1018 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
1019 break;
1020 case Opt_err_ro:
1021 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1022 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1023 set_opt (sbi->s_mount_opt, ERRORS_RO);
1024 break;
1025 case Opt_err_cont:
1026 clear_opt (sbi->s_mount_opt, ERRORS_RO);
1027 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1028 set_opt (sbi->s_mount_opt, ERRORS_CONT);
1029 break;
1030 case Opt_nouid32:
1031 set_opt (sbi->s_mount_opt, NO_UID32);
1032 break;
1033 case Opt_nocheck:
1034 clear_opt (sbi->s_mount_opt, CHECK);
1035 break;
1036 case Opt_debug:
1037 set_opt (sbi->s_mount_opt, DEBUG);
1038 break;
1039 case Opt_oldalloc:
1040 set_opt (sbi->s_mount_opt, OLDALLOC);
1041 break;
1042 case Opt_orlov:
1043 clear_opt (sbi->s_mount_opt, OLDALLOC);
1044 break;
1045 #ifdef CONFIG_EXT3_FS_XATTR
1046 case Opt_user_xattr:
1047 set_opt (sbi->s_mount_opt, XATTR_USER);
1048 break;
1049 case Opt_nouser_xattr:
1050 clear_opt (sbi->s_mount_opt, XATTR_USER);
1051 break;
1052 #else
1053 case Opt_user_xattr:
1054 case Opt_nouser_xattr:
1055 ext3_msg(sb, KERN_INFO,
1056 "(no)user_xattr options not supported");
1057 break;
1058 #endif
1059 #ifdef CONFIG_EXT3_FS_POSIX_ACL
1060 case Opt_acl:
1061 set_opt(sbi->s_mount_opt, POSIX_ACL);
1062 break;
1063 case Opt_noacl:
1064 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1065 break;
1066 #else
1067 case Opt_acl:
1068 case Opt_noacl:
1069 ext3_msg(sb, KERN_INFO,
1070 "(no)acl options not supported");
1071 break;
1072 #endif
1073 case Opt_reservation:
1074 set_opt(sbi->s_mount_opt, RESERVATION);
1075 break;
1076 case Opt_noreservation:
1077 clear_opt(sbi->s_mount_opt, RESERVATION);
1078 break;
1079 case Opt_journal_update:
1080 /* @@@ FIXME */
1081 /* Eventually we will want to be able to create
1082 a journal file here. For now, only allow the
1083 user to specify an existing inode to be the
1084 journal file. */
1085 if (is_remount) {
1086 ext3_msg(sb, KERN_ERR, "error: cannot specify "
1087 "journal on remount");
1088 return 0;
1090 set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
1091 break;
1092 case Opt_journal_inum:
1093 if (is_remount) {
1094 ext3_msg(sb, KERN_ERR, "error: cannot specify "
1095 "journal on remount");
1096 return 0;
1098 if (match_int(&args[0], &option))
1099 return 0;
1100 *inum = option;
1101 break;
1102 case Opt_journal_dev:
1103 if (is_remount) {
1104 ext3_msg(sb, KERN_ERR, "error: cannot specify "
1105 "journal on remount");
1106 return 0;
1108 if (match_int(&args[0], &option))
1109 return 0;
1110 *journal_devnum = option;
1111 break;
1112 case Opt_noload:
1113 set_opt (sbi->s_mount_opt, NOLOAD);
1114 break;
1115 case Opt_commit:
1116 if (match_int(&args[0], &option))
1117 return 0;
1118 if (option < 0)
1119 return 0;
1120 if (option == 0)
1121 option = JBD_DEFAULT_MAX_COMMIT_AGE;
1122 sbi->s_commit_interval = HZ * option;
1123 break;
1124 case Opt_data_journal:
1125 data_opt = EXT3_MOUNT_JOURNAL_DATA;
1126 goto datacheck;
1127 case Opt_data_ordered:
1128 data_opt = EXT3_MOUNT_ORDERED_DATA;
1129 goto datacheck;
1130 case Opt_data_writeback:
1131 data_opt = EXT3_MOUNT_WRITEBACK_DATA;
1132 datacheck:
1133 if (is_remount) {
1134 if (test_opt(sb, DATA_FLAGS) == data_opt)
1135 break;
1136 ext3_msg(sb, KERN_ERR,
1137 "error: cannot change "
1138 "data mode on remount. The filesystem "
1139 "is mounted in data=%s mode and you "
1140 "try to remount it in data=%s mode.",
1141 data_mode_string(test_opt(sb,
1142 DATA_FLAGS)),
1143 data_mode_string(data_opt));
1144 return 0;
1145 } else {
1146 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
1147 sbi->s_mount_opt |= data_opt;
1149 break;
1150 case Opt_data_err_abort:
1151 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1152 break;
1153 case Opt_data_err_ignore:
1154 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1155 break;
1156 #ifdef CONFIG_QUOTA
1157 case Opt_usrjquota:
1158 if (!set_qf_name(sb, USRQUOTA, &args[0]))
1159 return 0;
1160 break;
1161 case Opt_grpjquota:
1162 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
1163 return 0;
1164 break;
1165 case Opt_offusrjquota:
1166 if (!clear_qf_name(sb, USRQUOTA))
1167 return 0;
1168 break;
1169 case Opt_offgrpjquota:
1170 if (!clear_qf_name(sb, GRPQUOTA))
1171 return 0;
1172 break;
1173 case Opt_jqfmt_vfsold:
1174 qfmt = QFMT_VFS_OLD;
1175 goto set_qf_format;
1176 case Opt_jqfmt_vfsv0:
1177 qfmt = QFMT_VFS_V0;
1178 goto set_qf_format;
1179 case Opt_jqfmt_vfsv1:
1180 qfmt = QFMT_VFS_V1;
1181 set_qf_format:
1182 if (sb_any_quota_loaded(sb) &&
1183 sbi->s_jquota_fmt != qfmt) {
1184 ext3_msg(sb, KERN_ERR, "error: cannot change "
1185 "journaled quota options when "
1186 "quota turned on.");
1187 return 0;
1189 sbi->s_jquota_fmt = qfmt;
1190 break;
1191 case Opt_quota:
1192 case Opt_usrquota:
1193 set_opt(sbi->s_mount_opt, QUOTA);
1194 set_opt(sbi->s_mount_opt, USRQUOTA);
1195 break;
1196 case Opt_grpquota:
1197 set_opt(sbi->s_mount_opt, QUOTA);
1198 set_opt(sbi->s_mount_opt, GRPQUOTA);
1199 break;
1200 case Opt_noquota:
1201 if (sb_any_quota_loaded(sb)) {
1202 ext3_msg(sb, KERN_ERR, "error: cannot change "
1203 "quota options when quota turned on.");
1204 return 0;
1206 clear_opt(sbi->s_mount_opt, QUOTA);
1207 clear_opt(sbi->s_mount_opt, USRQUOTA);
1208 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1209 break;
1210 #else
1211 case Opt_quota:
1212 case Opt_usrquota:
1213 case Opt_grpquota:
1214 ext3_msg(sb, KERN_ERR,
1215 "error: quota options not supported.");
1216 break;
1217 case Opt_usrjquota:
1218 case Opt_grpjquota:
1219 case Opt_offusrjquota:
1220 case Opt_offgrpjquota:
1221 case Opt_jqfmt_vfsold:
1222 case Opt_jqfmt_vfsv0:
1223 case Opt_jqfmt_vfsv1:
1224 ext3_msg(sb, KERN_ERR,
1225 "error: journaled quota options not "
1226 "supported.");
1227 break;
1228 case Opt_noquota:
1229 break;
1230 #endif
1231 case Opt_abort:
1232 set_opt(sbi->s_mount_opt, ABORT);
1233 break;
1234 case Opt_nobarrier:
1235 clear_opt(sbi->s_mount_opt, BARRIER);
1236 break;
1237 case Opt_barrier:
1238 if (args[0].from) {
1239 if (match_int(&args[0], &option))
1240 return 0;
1241 } else
1242 option = 1; /* No argument, default to 1 */
1243 if (option)
1244 set_opt(sbi->s_mount_opt, BARRIER);
1245 else
1246 clear_opt(sbi->s_mount_opt, BARRIER);
1247 break;
1248 case Opt_ignore:
1249 break;
1250 case Opt_resize:
1251 if (!is_remount) {
1252 ext3_msg(sb, KERN_ERR,
1253 "error: resize option only available "
1254 "for remount");
1255 return 0;
1257 if (match_int(&args[0], &option) != 0)
1258 return 0;
1259 *n_blocks_count = option;
1260 break;
1261 case Opt_nobh:
1262 ext3_msg(sb, KERN_WARNING,
1263 "warning: ignoring deprecated nobh option");
1264 break;
1265 case Opt_bh:
1266 ext3_msg(sb, KERN_WARNING,
1267 "warning: ignoring deprecated bh option");
1268 break;
1269 default:
1270 ext3_msg(sb, KERN_ERR,
1271 "error: unrecognized mount option \"%s\" "
1272 "or missing value", p);
1273 return 0;
1276 #ifdef CONFIG_QUOTA
1277 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1278 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1279 clear_opt(sbi->s_mount_opt, USRQUOTA);
1280 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1281 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1283 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1284 ext3_msg(sb, KERN_ERR, "error: old and new quota "
1285 "format mixing.");
1286 return 0;
1289 if (!sbi->s_jquota_fmt) {
1290 ext3_msg(sb, KERN_ERR, "error: journaled quota format "
1291 "not specified.");
1292 return 0;
1294 } else {
1295 if (sbi->s_jquota_fmt) {
1296 ext3_msg(sb, KERN_ERR, "error: journaled quota format "
1297 "specified with no journaling "
1298 "enabled.");
1299 return 0;
1302 #endif
1303 return 1;
1306 static int ext3_setup_super(struct super_block *sb, struct ext3_super_block *es,
1307 int read_only)
1309 struct ext3_sb_info *sbi = EXT3_SB(sb);
1310 int res = 0;
1312 if (le32_to_cpu(es->s_rev_level) > EXT3_MAX_SUPP_REV) {
1313 ext3_msg(sb, KERN_ERR,
1314 "error: revision level too high, "
1315 "forcing read-only mode");
1316 res = MS_RDONLY;
1318 if (read_only)
1319 return res;
1320 if (!(sbi->s_mount_state & EXT3_VALID_FS))
1321 ext3_msg(sb, KERN_WARNING,
1322 "warning: mounting unchecked fs, "
1323 "running e2fsck is recommended");
1324 else if ((sbi->s_mount_state & EXT3_ERROR_FS))
1325 ext3_msg(sb, KERN_WARNING,
1326 "warning: mounting fs with errors, "
1327 "running e2fsck is recommended");
1328 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1329 le16_to_cpu(es->s_mnt_count) >=
1330 le16_to_cpu(es->s_max_mnt_count))
1331 ext3_msg(sb, KERN_WARNING,
1332 "warning: maximal mount count reached, "
1333 "running e2fsck is recommended");
1334 else if (le32_to_cpu(es->s_checkinterval) &&
1335 (le32_to_cpu(es->s_lastcheck) +
1336 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1337 ext3_msg(sb, KERN_WARNING,
1338 "warning: checktime reached, "
1339 "running e2fsck is recommended");
1340 #if 0
1341 /* @@@ We _will_ want to clear the valid bit if we find
1342 inconsistencies, to force a fsck at reboot. But for
1343 a plain journaled filesystem we can keep it set as
1344 valid forever! :) */
1345 es->s_state &= cpu_to_le16(~EXT3_VALID_FS);
1346 #endif
1347 if (!le16_to_cpu(es->s_max_mnt_count))
1348 es->s_max_mnt_count = cpu_to_le16(EXT3_DFL_MAX_MNT_COUNT);
1349 le16_add_cpu(&es->s_mnt_count, 1);
1350 es->s_mtime = cpu_to_le32(get_seconds());
1351 ext3_update_dynamic_rev(sb);
1352 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
1354 ext3_commit_super(sb, es, 1);
1355 if (test_opt(sb, DEBUG))
1356 ext3_msg(sb, KERN_INFO, "[bs=%lu, gc=%lu, "
1357 "bpg=%lu, ipg=%lu, mo=%04lx]",
1358 sb->s_blocksize,
1359 sbi->s_groups_count,
1360 EXT3_BLOCKS_PER_GROUP(sb),
1361 EXT3_INODES_PER_GROUP(sb),
1362 sbi->s_mount_opt);
1364 if (EXT3_SB(sb)->s_journal->j_inode == NULL) {
1365 char b[BDEVNAME_SIZE];
1366 ext3_msg(sb, KERN_INFO, "using external journal on %s",
1367 bdevname(EXT3_SB(sb)->s_journal->j_dev, b));
1368 } else {
1369 ext3_msg(sb, KERN_INFO, "using internal journal");
1371 return res;
1374 /* Called at mount-time, super-block is locked */
1375 static int ext3_check_descriptors(struct super_block *sb)
1377 struct ext3_sb_info *sbi = EXT3_SB(sb);
1378 int i;
1380 ext3_debug ("Checking group descriptors");
1382 for (i = 0; i < sbi->s_groups_count; i++) {
1383 struct ext3_group_desc *gdp = ext3_get_group_desc(sb, i, NULL);
1384 ext3_fsblk_t first_block = ext3_group_first_block_no(sb, i);
1385 ext3_fsblk_t last_block;
1387 if (i == sbi->s_groups_count - 1)
1388 last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
1389 else
1390 last_block = first_block +
1391 (EXT3_BLOCKS_PER_GROUP(sb) - 1);
1393 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
1394 le32_to_cpu(gdp->bg_block_bitmap) > last_block)
1396 ext3_error (sb, "ext3_check_descriptors",
1397 "Block bitmap for group %d"
1398 " not in group (block %lu)!",
1399 i, (unsigned long)
1400 le32_to_cpu(gdp->bg_block_bitmap));
1401 return 0;
1403 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
1404 le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
1406 ext3_error (sb, "ext3_check_descriptors",
1407 "Inode bitmap for group %d"
1408 " not in group (block %lu)!",
1409 i, (unsigned long)
1410 le32_to_cpu(gdp->bg_inode_bitmap));
1411 return 0;
1413 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
1414 le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
1415 last_block)
1417 ext3_error (sb, "ext3_check_descriptors",
1418 "Inode table for group %d"
1419 " not in group (block %lu)!",
1420 i, (unsigned long)
1421 le32_to_cpu(gdp->bg_inode_table));
1422 return 0;
1426 sbi->s_es->s_free_blocks_count=cpu_to_le32(ext3_count_free_blocks(sb));
1427 sbi->s_es->s_free_inodes_count=cpu_to_le32(ext3_count_free_inodes(sb));
1428 return 1;
1432 /* ext3_orphan_cleanup() walks a singly-linked list of inodes (starting at
1433 * the superblock) which were deleted from all directories, but held open by
1434 * a process at the time of a crash. We walk the list and try to delete these
1435 * inodes at recovery time (only with a read-write filesystem).
1437 * In order to keep the orphan inode chain consistent during traversal (in
1438 * case of crash during recovery), we link each inode into the superblock
1439 * orphan list_head and handle it the same way as an inode deletion during
1440 * normal operation (which journals the operations for us).
1442 * We only do an iget() and an iput() on each inode, which is very safe if we
1443 * accidentally point at an in-use or already deleted inode. The worst that
1444 * can happen in this case is that we get a "bit already cleared" message from
1445 * ext3_free_inode(). The only reason we would point at a wrong inode is if
1446 * e2fsck was run on this filesystem, and it must have already done the orphan
1447 * inode cleanup for us, so we can safely abort without any further action.
1449 static void ext3_orphan_cleanup (struct super_block * sb,
1450 struct ext3_super_block * es)
1452 unsigned int s_flags = sb->s_flags;
1453 int nr_orphans = 0, nr_truncates = 0;
1454 #ifdef CONFIG_QUOTA
1455 int i;
1456 #endif
1457 if (!es->s_last_orphan) {
1458 jbd_debug(4, "no orphan inodes to clean up\n");
1459 return;
1462 if (bdev_read_only(sb->s_bdev)) {
1463 ext3_msg(sb, KERN_ERR, "error: write access "
1464 "unavailable, skipping orphan cleanup.");
1465 return;
1468 if (EXT3_SB(sb)->s_mount_state & EXT3_ERROR_FS) {
1469 if (es->s_last_orphan)
1470 jbd_debug(1, "Errors on filesystem, "
1471 "clearing orphan list.\n");
1472 es->s_last_orphan = 0;
1473 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1474 return;
1477 if (s_flags & MS_RDONLY) {
1478 ext3_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
1479 sb->s_flags &= ~MS_RDONLY;
1481 #ifdef CONFIG_QUOTA
1482 /* Needed for iput() to work correctly and not trash data */
1483 sb->s_flags |= MS_ACTIVE;
1484 /* Turn on quotas so that they are updated correctly */
1485 for (i = 0; i < MAXQUOTAS; i++) {
1486 if (EXT3_SB(sb)->s_qf_names[i]) {
1487 int ret = ext3_quota_on_mount(sb, i);
1488 if (ret < 0)
1489 ext3_msg(sb, KERN_ERR,
1490 "error: cannot turn on journaled "
1491 "quota: %d", ret);
1494 #endif
1496 while (es->s_last_orphan) {
1497 struct inode *inode;
1499 inode = ext3_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1500 if (IS_ERR(inode)) {
1501 es->s_last_orphan = 0;
1502 break;
1505 list_add(&EXT3_I(inode)->i_orphan, &EXT3_SB(sb)->s_orphan);
1506 dquot_initialize(inode);
1507 if (inode->i_nlink) {
1508 printk(KERN_DEBUG
1509 "%s: truncating inode %lu to %Ld bytes\n",
1510 __func__, inode->i_ino, inode->i_size);
1511 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1512 inode->i_ino, inode->i_size);
1513 ext3_truncate(inode);
1514 nr_truncates++;
1515 } else {
1516 printk(KERN_DEBUG
1517 "%s: deleting unreferenced inode %lu\n",
1518 __func__, inode->i_ino);
1519 jbd_debug(2, "deleting unreferenced inode %lu\n",
1520 inode->i_ino);
1521 nr_orphans++;
1523 iput(inode); /* The delete magic happens here! */
1526 #define PLURAL(x) (x), ((x)==1) ? "" : "s"
1528 if (nr_orphans)
1529 ext3_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
1530 PLURAL(nr_orphans));
1531 if (nr_truncates)
1532 ext3_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
1533 PLURAL(nr_truncates));
1534 #ifdef CONFIG_QUOTA
1535 /* Turn quotas off */
1536 for (i = 0; i < MAXQUOTAS; i++) {
1537 if (sb_dqopt(sb)->files[i])
1538 dquot_quota_off(sb, i);
1540 #endif
1541 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1545 * Maximal file size. There is a direct, and {,double-,triple-}indirect
1546 * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
1547 * We need to be 1 filesystem block less than the 2^32 sector limit.
1549 static loff_t ext3_max_size(int bits)
1551 loff_t res = EXT3_NDIR_BLOCKS;
1552 int meta_blocks;
1553 loff_t upper_limit;
1555 /* This is calculated to be the largest file size for a
1556 * dense, file such that the total number of
1557 * sectors in the file, including data and all indirect blocks,
1558 * does not exceed 2^32 -1
1559 * __u32 i_blocks representing the total number of
1560 * 512 bytes blocks of the file
1562 upper_limit = (1LL << 32) - 1;
1564 /* total blocks in file system block size */
1565 upper_limit >>= (bits - 9);
1568 /* indirect blocks */
1569 meta_blocks = 1;
1570 /* double indirect blocks */
1571 meta_blocks += 1 + (1LL << (bits-2));
1572 /* tripple indirect blocks */
1573 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1575 upper_limit -= meta_blocks;
1576 upper_limit <<= bits;
1578 res += 1LL << (bits-2);
1579 res += 1LL << (2*(bits-2));
1580 res += 1LL << (3*(bits-2));
1581 res <<= bits;
1582 if (res > upper_limit)
1583 res = upper_limit;
1585 if (res > MAX_LFS_FILESIZE)
1586 res = MAX_LFS_FILESIZE;
1588 return res;
1591 static ext3_fsblk_t descriptor_loc(struct super_block *sb,
1592 ext3_fsblk_t logic_sb_block,
1593 int nr)
1595 struct ext3_sb_info *sbi = EXT3_SB(sb);
1596 unsigned long bg, first_meta_bg;
1597 int has_super = 0;
1599 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1601 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_META_BG) ||
1602 nr < first_meta_bg)
1603 return (logic_sb_block + nr + 1);
1604 bg = sbi->s_desc_per_block * nr;
1605 if (ext3_bg_has_super(sb, bg))
1606 has_super = 1;
1607 return (has_super + ext3_group_first_block_no(sb, bg));
1611 static int ext3_fill_super (struct super_block *sb, void *data, int silent)
1613 struct buffer_head * bh;
1614 struct ext3_super_block *es = NULL;
1615 struct ext3_sb_info *sbi;
1616 ext3_fsblk_t block;
1617 ext3_fsblk_t sb_block = get_sb_block(&data, sb);
1618 ext3_fsblk_t logic_sb_block;
1619 unsigned long offset = 0;
1620 unsigned int journal_inum = 0;
1621 unsigned long journal_devnum = 0;
1622 unsigned long def_mount_opts;
1623 struct inode *root;
1624 int blocksize;
1625 int hblock;
1626 int db_count;
1627 int i;
1628 int needs_recovery;
1629 int ret = -EINVAL;
1630 __le32 features;
1631 int err;
1633 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1634 if (!sbi)
1635 return -ENOMEM;
1637 sbi->s_blockgroup_lock =
1638 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
1639 if (!sbi->s_blockgroup_lock) {
1640 kfree(sbi);
1641 return -ENOMEM;
1643 sb->s_fs_info = sbi;
1644 sbi->s_mount_opt = 0;
1645 sbi->s_resuid = EXT3_DEF_RESUID;
1646 sbi->s_resgid = EXT3_DEF_RESGID;
1647 sbi->s_sb_block = sb_block;
1649 blocksize = sb_min_blocksize(sb, EXT3_MIN_BLOCK_SIZE);
1650 if (!blocksize) {
1651 ext3_msg(sb, KERN_ERR, "error: unable to set blocksize");
1652 goto out_fail;
1656 * The ext3 superblock will not be buffer aligned for other than 1kB
1657 * block sizes. We need to calculate the offset from buffer start.
1659 if (blocksize != EXT3_MIN_BLOCK_SIZE) {
1660 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1661 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1662 } else {
1663 logic_sb_block = sb_block;
1666 if (!(bh = sb_bread(sb, logic_sb_block))) {
1667 ext3_msg(sb, KERN_ERR, "error: unable to read superblock");
1668 goto out_fail;
1671 * Note: s_es must be initialized as soon as possible because
1672 * some ext3 macro-instructions depend on its value
1674 es = (struct ext3_super_block *) (bh->b_data + offset);
1675 sbi->s_es = es;
1676 sb->s_magic = le16_to_cpu(es->s_magic);
1677 if (sb->s_magic != EXT3_SUPER_MAGIC)
1678 goto cantfind_ext3;
1680 /* Set defaults before we parse the mount options */
1681 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1682 if (def_mount_opts & EXT3_DEFM_DEBUG)
1683 set_opt(sbi->s_mount_opt, DEBUG);
1684 if (def_mount_opts & EXT3_DEFM_BSDGROUPS)
1685 set_opt(sbi->s_mount_opt, GRPID);
1686 if (def_mount_opts & EXT3_DEFM_UID16)
1687 set_opt(sbi->s_mount_opt, NO_UID32);
1688 #ifdef CONFIG_EXT3_FS_XATTR
1689 if (def_mount_opts & EXT3_DEFM_XATTR_USER)
1690 set_opt(sbi->s_mount_opt, XATTR_USER);
1691 #endif
1692 #ifdef CONFIG_EXT3_FS_POSIX_ACL
1693 if (def_mount_opts & EXT3_DEFM_ACL)
1694 set_opt(sbi->s_mount_opt, POSIX_ACL);
1695 #endif
1696 if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_DATA)
1697 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
1698 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_ORDERED)
1699 set_opt(sbi->s_mount_opt, ORDERED_DATA);
1700 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_WBACK)
1701 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
1703 if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_PANIC)
1704 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1705 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_CONTINUE)
1706 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1707 else
1708 set_opt(sbi->s_mount_opt, ERRORS_RO);
1710 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1711 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1713 set_opt(sbi->s_mount_opt, RESERVATION);
1715 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1716 NULL, 0))
1717 goto failed_mount;
1719 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1720 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
1722 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV &&
1723 (EXT3_HAS_COMPAT_FEATURE(sb, ~0U) ||
1724 EXT3_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1725 EXT3_HAS_INCOMPAT_FEATURE(sb, ~0U)))
1726 ext3_msg(sb, KERN_WARNING,
1727 "warning: feature flags set on rev 0 fs, "
1728 "running e2fsck is recommended");
1730 * Check feature flags regardless of the revision level, since we
1731 * previously didn't change the revision level when setting the flags,
1732 * so there is a chance incompat flags are set on a rev 0 filesystem.
1734 features = EXT3_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP);
1735 if (features) {
1736 ext3_msg(sb, KERN_ERR,
1737 "error: couldn't mount because of unsupported "
1738 "optional features (%x)", le32_to_cpu(features));
1739 goto failed_mount;
1741 features = EXT3_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP);
1742 if (!(sb->s_flags & MS_RDONLY) && features) {
1743 ext3_msg(sb, KERN_ERR,
1744 "error: couldn't mount RDWR because of unsupported "
1745 "optional features (%x)", le32_to_cpu(features));
1746 goto failed_mount;
1748 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
1750 if (blocksize < EXT3_MIN_BLOCK_SIZE ||
1751 blocksize > EXT3_MAX_BLOCK_SIZE) {
1752 ext3_msg(sb, KERN_ERR,
1753 "error: couldn't mount because of unsupported "
1754 "filesystem blocksize %d", blocksize);
1755 goto failed_mount;
1758 hblock = bdev_logical_block_size(sb->s_bdev);
1759 if (sb->s_blocksize != blocksize) {
1761 * Make sure the blocksize for the filesystem is larger
1762 * than the hardware sectorsize for the machine.
1764 if (blocksize < hblock) {
1765 ext3_msg(sb, KERN_ERR,
1766 "error: fsblocksize %d too small for "
1767 "hardware sectorsize %d", blocksize, hblock);
1768 goto failed_mount;
1771 brelse (bh);
1772 if (!sb_set_blocksize(sb, blocksize)) {
1773 ext3_msg(sb, KERN_ERR,
1774 "error: bad blocksize %d", blocksize);
1775 goto out_fail;
1777 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1778 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1779 bh = sb_bread(sb, logic_sb_block);
1780 if (!bh) {
1781 ext3_msg(sb, KERN_ERR,
1782 "error: can't read superblock on 2nd try");
1783 goto failed_mount;
1785 es = (struct ext3_super_block *)(bh->b_data + offset);
1786 sbi->s_es = es;
1787 if (es->s_magic != cpu_to_le16(EXT3_SUPER_MAGIC)) {
1788 ext3_msg(sb, KERN_ERR,
1789 "error: magic mismatch");
1790 goto failed_mount;
1794 sb->s_maxbytes = ext3_max_size(sb->s_blocksize_bits);
1796 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV) {
1797 sbi->s_inode_size = EXT3_GOOD_OLD_INODE_SIZE;
1798 sbi->s_first_ino = EXT3_GOOD_OLD_FIRST_INO;
1799 } else {
1800 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1801 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1802 if ((sbi->s_inode_size < EXT3_GOOD_OLD_INODE_SIZE) ||
1803 (!is_power_of_2(sbi->s_inode_size)) ||
1804 (sbi->s_inode_size > blocksize)) {
1805 ext3_msg(sb, KERN_ERR,
1806 "error: unsupported inode size: %d",
1807 sbi->s_inode_size);
1808 goto failed_mount;
1811 sbi->s_frag_size = EXT3_MIN_FRAG_SIZE <<
1812 le32_to_cpu(es->s_log_frag_size);
1813 if (blocksize != sbi->s_frag_size) {
1814 ext3_msg(sb, KERN_ERR,
1815 "error: fragsize %lu != blocksize %u (unsupported)",
1816 sbi->s_frag_size, blocksize);
1817 goto failed_mount;
1819 sbi->s_frags_per_block = 1;
1820 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1821 sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
1822 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
1823 if (EXT3_INODE_SIZE(sb) == 0 || EXT3_INODES_PER_GROUP(sb) == 0)
1824 goto cantfind_ext3;
1825 sbi->s_inodes_per_block = blocksize / EXT3_INODE_SIZE(sb);
1826 if (sbi->s_inodes_per_block == 0)
1827 goto cantfind_ext3;
1828 sbi->s_itb_per_group = sbi->s_inodes_per_group /
1829 sbi->s_inodes_per_block;
1830 sbi->s_desc_per_block = blocksize / sizeof(struct ext3_group_desc);
1831 sbi->s_sbh = bh;
1832 sbi->s_mount_state = le16_to_cpu(es->s_state);
1833 sbi->s_addr_per_block_bits = ilog2(EXT3_ADDR_PER_BLOCK(sb));
1834 sbi->s_desc_per_block_bits = ilog2(EXT3_DESC_PER_BLOCK(sb));
1835 for (i=0; i < 4; i++)
1836 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
1837 sbi->s_def_hash_version = es->s_def_hash_version;
1838 i = le32_to_cpu(es->s_flags);
1839 if (i & EXT2_FLAGS_UNSIGNED_HASH)
1840 sbi->s_hash_unsigned = 3;
1841 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
1842 #ifdef __CHAR_UNSIGNED__
1843 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
1844 sbi->s_hash_unsigned = 3;
1845 #else
1846 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
1847 #endif
1850 if (sbi->s_blocks_per_group > blocksize * 8) {
1851 ext3_msg(sb, KERN_ERR,
1852 "#blocks per group too big: %lu",
1853 sbi->s_blocks_per_group);
1854 goto failed_mount;
1856 if (sbi->s_frags_per_group > blocksize * 8) {
1857 ext3_msg(sb, KERN_ERR,
1858 "error: #fragments per group too big: %lu",
1859 sbi->s_frags_per_group);
1860 goto failed_mount;
1862 if (sbi->s_inodes_per_group > blocksize * 8) {
1863 ext3_msg(sb, KERN_ERR,
1864 "error: #inodes per group too big: %lu",
1865 sbi->s_inodes_per_group);
1866 goto failed_mount;
1869 err = generic_check_addressable(sb->s_blocksize_bits,
1870 le32_to_cpu(es->s_blocks_count));
1871 if (err) {
1872 ext3_msg(sb, KERN_ERR,
1873 "error: filesystem is too large to mount safely");
1874 if (sizeof(sector_t) < 8)
1875 ext3_msg(sb, KERN_ERR,
1876 "error: CONFIG_LBDAF not enabled");
1877 ret = err;
1878 goto failed_mount;
1881 if (EXT3_BLOCKS_PER_GROUP(sb) == 0)
1882 goto cantfind_ext3;
1883 sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1884 le32_to_cpu(es->s_first_data_block) - 1)
1885 / EXT3_BLOCKS_PER_GROUP(sb)) + 1;
1886 db_count = DIV_ROUND_UP(sbi->s_groups_count, EXT3_DESC_PER_BLOCK(sb));
1887 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
1888 GFP_KERNEL);
1889 if (sbi->s_group_desc == NULL) {
1890 ext3_msg(sb, KERN_ERR,
1891 "error: not enough memory");
1892 ret = -ENOMEM;
1893 goto failed_mount;
1896 bgl_lock_init(sbi->s_blockgroup_lock);
1898 for (i = 0; i < db_count; i++) {
1899 block = descriptor_loc(sb, logic_sb_block, i);
1900 sbi->s_group_desc[i] = sb_bread(sb, block);
1901 if (!sbi->s_group_desc[i]) {
1902 ext3_msg(sb, KERN_ERR,
1903 "error: can't read group descriptor %d", i);
1904 db_count = i;
1905 goto failed_mount2;
1908 if (!ext3_check_descriptors (sb)) {
1909 ext3_msg(sb, KERN_ERR,
1910 "error: group descriptors corrupted");
1911 goto failed_mount2;
1913 sbi->s_gdb_count = db_count;
1914 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1915 spin_lock_init(&sbi->s_next_gen_lock);
1917 /* per fileystem reservation list head & lock */
1918 spin_lock_init(&sbi->s_rsv_window_lock);
1919 sbi->s_rsv_window_root = RB_ROOT;
1920 /* Add a single, static dummy reservation to the start of the
1921 * reservation window list --- it gives us a placeholder for
1922 * append-at-start-of-list which makes the allocation logic
1923 * _much_ simpler. */
1924 sbi->s_rsv_window_head.rsv_start = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1925 sbi->s_rsv_window_head.rsv_end = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1926 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1927 sbi->s_rsv_window_head.rsv_goal_size = 0;
1928 ext3_rsv_window_add(sb, &sbi->s_rsv_window_head);
1931 * set up enough so that it can read an inode
1933 sb->s_op = &ext3_sops;
1934 sb->s_export_op = &ext3_export_ops;
1935 sb->s_xattr = ext3_xattr_handlers;
1936 #ifdef CONFIG_QUOTA
1937 sb->s_qcop = &ext3_qctl_operations;
1938 sb->dq_op = &ext3_quota_operations;
1939 #endif
1940 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
1941 mutex_init(&sbi->s_orphan_lock);
1942 mutex_init(&sbi->s_resize_lock);
1944 sb->s_root = NULL;
1946 needs_recovery = (es->s_last_orphan != 0 ||
1947 EXT3_HAS_INCOMPAT_FEATURE(sb,
1948 EXT3_FEATURE_INCOMPAT_RECOVER));
1951 * The first inode we look at is the journal inode. Don't try
1952 * root first: it may be modified in the journal!
1954 if (!test_opt(sb, NOLOAD) &&
1955 EXT3_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL)) {
1956 if (ext3_load_journal(sb, es, journal_devnum))
1957 goto failed_mount2;
1958 } else if (journal_inum) {
1959 if (ext3_create_journal(sb, es, journal_inum))
1960 goto failed_mount2;
1961 } else {
1962 if (!silent)
1963 ext3_msg(sb, KERN_ERR,
1964 "error: no journal found. "
1965 "mounting ext3 over ext2?");
1966 goto failed_mount2;
1968 err = percpu_counter_init(&sbi->s_freeblocks_counter,
1969 ext3_count_free_blocks(sb));
1970 if (!err) {
1971 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1972 ext3_count_free_inodes(sb));
1974 if (!err) {
1975 err = percpu_counter_init(&sbi->s_dirs_counter,
1976 ext3_count_dirs(sb));
1978 if (err) {
1979 ext3_msg(sb, KERN_ERR, "error: insufficient memory");
1980 ret = err;
1981 goto failed_mount3;
1984 /* We have now updated the journal if required, so we can
1985 * validate the data journaling mode. */
1986 switch (test_opt(sb, DATA_FLAGS)) {
1987 case 0:
1988 /* No mode set, assume a default based on the journal
1989 capabilities: ORDERED_DATA if the journal can
1990 cope, else JOURNAL_DATA */
1991 if (journal_check_available_features
1992 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE))
1993 set_opt(sbi->s_mount_opt, DEFAULT_DATA_MODE);
1994 else
1995 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
1996 break;
1998 case EXT3_MOUNT_ORDERED_DATA:
1999 case EXT3_MOUNT_WRITEBACK_DATA:
2000 if (!journal_check_available_features
2001 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE)) {
2002 ext3_msg(sb, KERN_ERR,
2003 "error: journal does not support "
2004 "requested data journaling mode");
2005 goto failed_mount3;
2007 default:
2008 break;
2012 * The journal_load will have done any necessary log recovery,
2013 * so we can safely mount the rest of the filesystem now.
2016 root = ext3_iget(sb, EXT3_ROOT_INO);
2017 if (IS_ERR(root)) {
2018 ext3_msg(sb, KERN_ERR, "error: get root inode failed");
2019 ret = PTR_ERR(root);
2020 goto failed_mount3;
2022 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2023 iput(root);
2024 ext3_msg(sb, KERN_ERR, "error: corrupt root inode, run e2fsck");
2025 goto failed_mount3;
2027 sb->s_root = d_alloc_root(root);
2028 if (!sb->s_root) {
2029 ext3_msg(sb, KERN_ERR, "error: get root dentry failed");
2030 iput(root);
2031 ret = -ENOMEM;
2032 goto failed_mount3;
2035 ext3_setup_super (sb, es, sb->s_flags & MS_RDONLY);
2037 EXT3_SB(sb)->s_mount_state |= EXT3_ORPHAN_FS;
2038 ext3_orphan_cleanup(sb, es);
2039 EXT3_SB(sb)->s_mount_state &= ~EXT3_ORPHAN_FS;
2040 if (needs_recovery)
2041 ext3_msg(sb, KERN_INFO, "recovery complete");
2042 ext3_mark_recovery_complete(sb, es);
2043 ext3_msg(sb, KERN_INFO, "mounted filesystem with %s data mode",
2044 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA ? "journal":
2045 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA ? "ordered":
2046 "writeback");
2048 return 0;
2050 cantfind_ext3:
2051 if (!silent)
2052 ext3_msg(sb, KERN_INFO,
2053 "error: can't find ext3 filesystem on dev %s.",
2054 sb->s_id);
2055 goto failed_mount;
2057 failed_mount3:
2058 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2059 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2060 percpu_counter_destroy(&sbi->s_dirs_counter);
2061 journal_destroy(sbi->s_journal);
2062 failed_mount2:
2063 for (i = 0; i < db_count; i++)
2064 brelse(sbi->s_group_desc[i]);
2065 kfree(sbi->s_group_desc);
2066 failed_mount:
2067 #ifdef CONFIG_QUOTA
2068 for (i = 0; i < MAXQUOTAS; i++)
2069 kfree(sbi->s_qf_names[i]);
2070 #endif
2071 ext3_blkdev_remove(sbi);
2072 brelse(bh);
2073 out_fail:
2074 sb->s_fs_info = NULL;
2075 kfree(sbi->s_blockgroup_lock);
2076 kfree(sbi);
2077 return ret;
2081 * Setup any per-fs journal parameters now. We'll do this both on
2082 * initial mount, once the journal has been initialised but before we've
2083 * done any recovery; and again on any subsequent remount.
2085 static void ext3_init_journal_params(struct super_block *sb, journal_t *journal)
2087 struct ext3_sb_info *sbi = EXT3_SB(sb);
2089 if (sbi->s_commit_interval)
2090 journal->j_commit_interval = sbi->s_commit_interval;
2091 /* We could also set up an ext3-specific default for the commit
2092 * interval here, but for now we'll just fall back to the jbd
2093 * default. */
2095 spin_lock(&journal->j_state_lock);
2096 if (test_opt(sb, BARRIER))
2097 journal->j_flags |= JFS_BARRIER;
2098 else
2099 journal->j_flags &= ~JFS_BARRIER;
2100 if (test_opt(sb, DATA_ERR_ABORT))
2101 journal->j_flags |= JFS_ABORT_ON_SYNCDATA_ERR;
2102 else
2103 journal->j_flags &= ~JFS_ABORT_ON_SYNCDATA_ERR;
2104 spin_unlock(&journal->j_state_lock);
2107 static journal_t *ext3_get_journal(struct super_block *sb,
2108 unsigned int journal_inum)
2110 struct inode *journal_inode;
2111 journal_t *journal;
2113 /* First, test for the existence of a valid inode on disk. Bad
2114 * things happen if we iget() an unused inode, as the subsequent
2115 * iput() will try to delete it. */
2117 journal_inode = ext3_iget(sb, journal_inum);
2118 if (IS_ERR(journal_inode)) {
2119 ext3_msg(sb, KERN_ERR, "error: no journal found");
2120 return NULL;
2122 if (!journal_inode->i_nlink) {
2123 make_bad_inode(journal_inode);
2124 iput(journal_inode);
2125 ext3_msg(sb, KERN_ERR, "error: journal inode is deleted");
2126 return NULL;
2129 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
2130 journal_inode, journal_inode->i_size);
2131 if (!S_ISREG(journal_inode->i_mode)) {
2132 ext3_msg(sb, KERN_ERR, "error: invalid journal inode");
2133 iput(journal_inode);
2134 return NULL;
2137 journal = journal_init_inode(journal_inode);
2138 if (!journal) {
2139 ext3_msg(sb, KERN_ERR, "error: could not load journal inode");
2140 iput(journal_inode);
2141 return NULL;
2143 journal->j_private = sb;
2144 ext3_init_journal_params(sb, journal);
2145 return journal;
2148 static journal_t *ext3_get_dev_journal(struct super_block *sb,
2149 dev_t j_dev)
2151 struct buffer_head * bh;
2152 journal_t *journal;
2153 ext3_fsblk_t start;
2154 ext3_fsblk_t len;
2155 int hblock, blocksize;
2156 ext3_fsblk_t sb_block;
2157 unsigned long offset;
2158 struct ext3_super_block * es;
2159 struct block_device *bdev;
2161 bdev = ext3_blkdev_get(j_dev, sb);
2162 if (bdev == NULL)
2163 return NULL;
2165 if (bd_claim(bdev, sb)) {
2166 ext3_msg(sb, KERN_ERR,
2167 "error: failed to claim external journal device");
2168 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
2169 return NULL;
2172 blocksize = sb->s_blocksize;
2173 hblock = bdev_logical_block_size(bdev);
2174 if (blocksize < hblock) {
2175 ext3_msg(sb, KERN_ERR,
2176 "error: blocksize too small for journal device");
2177 goto out_bdev;
2180 sb_block = EXT3_MIN_BLOCK_SIZE / blocksize;
2181 offset = EXT3_MIN_BLOCK_SIZE % blocksize;
2182 set_blocksize(bdev, blocksize);
2183 if (!(bh = __bread(bdev, sb_block, blocksize))) {
2184 ext3_msg(sb, KERN_ERR, "error: couldn't read superblock of "
2185 "external journal");
2186 goto out_bdev;
2189 es = (struct ext3_super_block *) (bh->b_data + offset);
2190 if ((le16_to_cpu(es->s_magic) != EXT3_SUPER_MAGIC) ||
2191 !(le32_to_cpu(es->s_feature_incompat) &
2192 EXT3_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2193 ext3_msg(sb, KERN_ERR, "error: external journal has "
2194 "bad superblock");
2195 brelse(bh);
2196 goto out_bdev;
2199 if (memcmp(EXT3_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2200 ext3_msg(sb, KERN_ERR, "error: journal UUID does not match");
2201 brelse(bh);
2202 goto out_bdev;
2205 len = le32_to_cpu(es->s_blocks_count);
2206 start = sb_block + 1;
2207 brelse(bh); /* we're done with the superblock */
2209 journal = journal_init_dev(bdev, sb->s_bdev,
2210 start, len, blocksize);
2211 if (!journal) {
2212 ext3_msg(sb, KERN_ERR,
2213 "error: failed to create device journal");
2214 goto out_bdev;
2216 journal->j_private = sb;
2217 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2218 wait_on_buffer(journal->j_sb_buffer);
2219 if (!buffer_uptodate(journal->j_sb_buffer)) {
2220 ext3_msg(sb, KERN_ERR, "I/O error on journal device");
2221 goto out_journal;
2223 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
2224 ext3_msg(sb, KERN_ERR,
2225 "error: external journal has more than one "
2226 "user (unsupported) - %d",
2227 be32_to_cpu(journal->j_superblock->s_nr_users));
2228 goto out_journal;
2230 EXT3_SB(sb)->journal_bdev = bdev;
2231 ext3_init_journal_params(sb, journal);
2232 return journal;
2233 out_journal:
2234 journal_destroy(journal);
2235 out_bdev:
2236 ext3_blkdev_put(bdev);
2237 return NULL;
2240 static int ext3_load_journal(struct super_block *sb,
2241 struct ext3_super_block *es,
2242 unsigned long journal_devnum)
2244 journal_t *journal;
2245 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2246 dev_t journal_dev;
2247 int err = 0;
2248 int really_read_only;
2250 if (journal_devnum &&
2251 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2252 ext3_msg(sb, KERN_INFO, "external journal device major/minor "
2253 "numbers have changed");
2254 journal_dev = new_decode_dev(journal_devnum);
2255 } else
2256 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2258 really_read_only = bdev_read_only(sb->s_bdev);
2261 * Are we loading a blank journal or performing recovery after a
2262 * crash? For recovery, we need to check in advance whether we
2263 * can get read-write access to the device.
2266 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER)) {
2267 if (sb->s_flags & MS_RDONLY) {
2268 ext3_msg(sb, KERN_INFO,
2269 "recovery required on readonly filesystem");
2270 if (really_read_only) {
2271 ext3_msg(sb, KERN_ERR, "error: write access "
2272 "unavailable, cannot proceed");
2273 return -EROFS;
2275 ext3_msg(sb, KERN_INFO,
2276 "write access will be enabled during recovery");
2280 if (journal_inum && journal_dev) {
2281 ext3_msg(sb, KERN_ERR, "error: filesystem has both journal "
2282 "and inode journals");
2283 return -EINVAL;
2286 if (journal_inum) {
2287 if (!(journal = ext3_get_journal(sb, journal_inum)))
2288 return -EINVAL;
2289 } else {
2290 if (!(journal = ext3_get_dev_journal(sb, journal_dev)))
2291 return -EINVAL;
2294 if (!(journal->j_flags & JFS_BARRIER))
2295 printk(KERN_INFO "EXT3-fs: barriers not enabled\n");
2297 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2298 err = journal_update_format(journal);
2299 if (err) {
2300 ext3_msg(sb, KERN_ERR, "error updating journal");
2301 journal_destroy(journal);
2302 return err;
2306 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER))
2307 err = journal_wipe(journal, !really_read_only);
2308 if (!err)
2309 err = journal_load(journal);
2311 if (err) {
2312 ext3_msg(sb, KERN_ERR, "error loading journal");
2313 journal_destroy(journal);
2314 return err;
2317 EXT3_SB(sb)->s_journal = journal;
2318 ext3_clear_journal_err(sb, es);
2320 if (!really_read_only && journal_devnum &&
2321 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2322 es->s_journal_dev = cpu_to_le32(journal_devnum);
2324 /* Make sure we flush the recovery flag to disk. */
2325 ext3_commit_super(sb, es, 1);
2328 return 0;
2331 static int ext3_create_journal(struct super_block *sb,
2332 struct ext3_super_block *es,
2333 unsigned int journal_inum)
2335 journal_t *journal;
2336 int err;
2338 if (sb->s_flags & MS_RDONLY) {
2339 ext3_msg(sb, KERN_ERR,
2340 "error: readonly filesystem when trying to "
2341 "create journal");
2342 return -EROFS;
2345 journal = ext3_get_journal(sb, journal_inum);
2346 if (!journal)
2347 return -EINVAL;
2349 ext3_msg(sb, KERN_INFO, "creating new journal on inode %u",
2350 journal_inum);
2352 err = journal_create(journal);
2353 if (err) {
2354 ext3_msg(sb, KERN_ERR, "error creating journal");
2355 journal_destroy(journal);
2356 return -EIO;
2359 EXT3_SB(sb)->s_journal = journal;
2361 ext3_update_dynamic_rev(sb);
2362 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2363 EXT3_SET_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL);
2365 es->s_journal_inum = cpu_to_le32(journal_inum);
2367 /* Make sure we flush the recovery flag to disk. */
2368 ext3_commit_super(sb, es, 1);
2370 return 0;
2373 static int ext3_commit_super(struct super_block *sb,
2374 struct ext3_super_block *es,
2375 int sync)
2377 struct buffer_head *sbh = EXT3_SB(sb)->s_sbh;
2378 int error = 0;
2380 if (!sbh)
2381 return error;
2383 if (buffer_write_io_error(sbh)) {
2385 * Oh, dear. A previous attempt to write the
2386 * superblock failed. This could happen because the
2387 * USB device was yanked out. Or it could happen to
2388 * be a transient write error and maybe the block will
2389 * be remapped. Nothing we can do but to retry the
2390 * write and hope for the best.
2392 ext3_msg(sb, KERN_ERR, "previous I/O error to "
2393 "superblock detected");
2394 clear_buffer_write_io_error(sbh);
2395 set_buffer_uptodate(sbh);
2398 * If the file system is mounted read-only, don't update the
2399 * superblock write time. This avoids updating the superblock
2400 * write time when we are mounting the root file system
2401 * read/only but we need to replay the journal; at that point,
2402 * for people who are east of GMT and who make their clock
2403 * tick in localtime for Windows bug-for-bug compatibility,
2404 * the clock is set in the future, and this will cause e2fsck
2405 * to complain and force a full file system check.
2407 if (!(sb->s_flags & MS_RDONLY))
2408 es->s_wtime = cpu_to_le32(get_seconds());
2409 es->s_free_blocks_count = cpu_to_le32(ext3_count_free_blocks(sb));
2410 es->s_free_inodes_count = cpu_to_le32(ext3_count_free_inodes(sb));
2411 BUFFER_TRACE(sbh, "marking dirty");
2412 mark_buffer_dirty(sbh);
2413 if (sync) {
2414 error = sync_dirty_buffer(sbh);
2415 if (buffer_write_io_error(sbh)) {
2416 ext3_msg(sb, KERN_ERR, "I/O error while writing "
2417 "superblock");
2418 clear_buffer_write_io_error(sbh);
2419 set_buffer_uptodate(sbh);
2422 return error;
2427 * Have we just finished recovery? If so, and if we are mounting (or
2428 * remounting) the filesystem readonly, then we will end up with a
2429 * consistent fs on disk. Record that fact.
2431 static void ext3_mark_recovery_complete(struct super_block * sb,
2432 struct ext3_super_block * es)
2434 journal_t *journal = EXT3_SB(sb)->s_journal;
2436 journal_lock_updates(journal);
2437 if (journal_flush(journal) < 0)
2438 goto out;
2440 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER) &&
2441 sb->s_flags & MS_RDONLY) {
2442 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2443 ext3_commit_super(sb, es, 1);
2446 out:
2447 journal_unlock_updates(journal);
2451 * If we are mounting (or read-write remounting) a filesystem whose journal
2452 * has recorded an error from a previous lifetime, move that error to the
2453 * main filesystem now.
2455 static void ext3_clear_journal_err(struct super_block *sb,
2456 struct ext3_super_block *es)
2458 journal_t *journal;
2459 int j_errno;
2460 const char *errstr;
2462 journal = EXT3_SB(sb)->s_journal;
2465 * Now check for any error status which may have been recorded in the
2466 * journal by a prior ext3_error() or ext3_abort()
2469 j_errno = journal_errno(journal);
2470 if (j_errno) {
2471 char nbuf[16];
2473 errstr = ext3_decode_error(sb, j_errno, nbuf);
2474 ext3_warning(sb, __func__, "Filesystem error recorded "
2475 "from previous mount: %s", errstr);
2476 ext3_warning(sb, __func__, "Marking fs in need of "
2477 "filesystem check.");
2479 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
2480 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
2481 ext3_commit_super (sb, es, 1);
2483 journal_clear_err(journal);
2488 * Force the running and committing transactions to commit,
2489 * and wait on the commit.
2491 int ext3_force_commit(struct super_block *sb)
2493 journal_t *journal;
2494 int ret;
2496 if (sb->s_flags & MS_RDONLY)
2497 return 0;
2499 journal = EXT3_SB(sb)->s_journal;
2500 ret = ext3_journal_force_commit(journal);
2501 return ret;
2504 static int ext3_sync_fs(struct super_block *sb, int wait)
2506 tid_t target;
2508 if (journal_start_commit(EXT3_SB(sb)->s_journal, &target)) {
2509 if (wait)
2510 log_wait_commit(EXT3_SB(sb)->s_journal, target);
2512 return 0;
2516 * LVM calls this function before a (read-only) snapshot is created. This
2517 * gives us a chance to flush the journal completely and mark the fs clean.
2519 static int ext3_freeze(struct super_block *sb)
2521 int error = 0;
2522 journal_t *journal;
2524 if (!(sb->s_flags & MS_RDONLY)) {
2525 journal = EXT3_SB(sb)->s_journal;
2527 /* Now we set up the journal barrier. */
2528 journal_lock_updates(journal);
2531 * We don't want to clear needs_recovery flag when we failed
2532 * to flush the journal.
2534 error = journal_flush(journal);
2535 if (error < 0)
2536 goto out;
2538 /* Journal blocked and flushed, clear needs_recovery flag. */
2539 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2540 error = ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2541 if (error)
2542 goto out;
2544 return 0;
2546 out:
2547 journal_unlock_updates(journal);
2548 return error;
2552 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2553 * flag here, even though the filesystem is not technically dirty yet.
2555 static int ext3_unfreeze(struct super_block *sb)
2557 if (!(sb->s_flags & MS_RDONLY)) {
2558 lock_super(sb);
2559 /* Reser the needs_recovery flag before the fs is unlocked. */
2560 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2561 ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2562 unlock_super(sb);
2563 journal_unlock_updates(EXT3_SB(sb)->s_journal);
2565 return 0;
2568 static int ext3_remount (struct super_block * sb, int * flags, char * data)
2570 struct ext3_super_block * es;
2571 struct ext3_sb_info *sbi = EXT3_SB(sb);
2572 ext3_fsblk_t n_blocks_count = 0;
2573 unsigned long old_sb_flags;
2574 struct ext3_mount_options old_opts;
2575 int enable_quota = 0;
2576 int err;
2577 #ifdef CONFIG_QUOTA
2578 int i;
2579 #endif
2581 /* Store the original options */
2582 lock_super(sb);
2583 old_sb_flags = sb->s_flags;
2584 old_opts.s_mount_opt = sbi->s_mount_opt;
2585 old_opts.s_resuid = sbi->s_resuid;
2586 old_opts.s_resgid = sbi->s_resgid;
2587 old_opts.s_commit_interval = sbi->s_commit_interval;
2588 #ifdef CONFIG_QUOTA
2589 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2590 for (i = 0; i < MAXQUOTAS; i++)
2591 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2592 #endif
2595 * Allow the "check" option to be passed as a remount option.
2597 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
2598 err = -EINVAL;
2599 goto restore_opts;
2602 if (test_opt(sb, ABORT))
2603 ext3_abort(sb, __func__, "Abort forced by user");
2605 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2606 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
2608 es = sbi->s_es;
2610 ext3_init_journal_params(sb, sbi->s_journal);
2612 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
2613 n_blocks_count > le32_to_cpu(es->s_blocks_count)) {
2614 if (test_opt(sb, ABORT)) {
2615 err = -EROFS;
2616 goto restore_opts;
2619 if (*flags & MS_RDONLY) {
2620 err = dquot_suspend(sb, -1);
2621 if (err < 0)
2622 goto restore_opts;
2625 * First of all, the unconditional stuff we have to do
2626 * to disable replay of the journal when we next remount
2628 sb->s_flags |= MS_RDONLY;
2631 * OK, test if we are remounting a valid rw partition
2632 * readonly, and if so set the rdonly flag and then
2633 * mark the partition as valid again.
2635 if (!(es->s_state & cpu_to_le16(EXT3_VALID_FS)) &&
2636 (sbi->s_mount_state & EXT3_VALID_FS))
2637 es->s_state = cpu_to_le16(sbi->s_mount_state);
2639 ext3_mark_recovery_complete(sb, es);
2640 } else {
2641 __le32 ret;
2642 if ((ret = EXT3_HAS_RO_COMPAT_FEATURE(sb,
2643 ~EXT3_FEATURE_RO_COMPAT_SUPP))) {
2644 ext3_msg(sb, KERN_WARNING,
2645 "warning: couldn't remount RDWR "
2646 "because of unsupported optional "
2647 "features (%x)", le32_to_cpu(ret));
2648 err = -EROFS;
2649 goto restore_opts;
2653 * If we have an unprocessed orphan list hanging
2654 * around from a previously readonly bdev mount,
2655 * require a full umount/remount for now.
2657 if (es->s_last_orphan) {
2658 ext3_msg(sb, KERN_WARNING, "warning: couldn't "
2659 "remount RDWR because of unprocessed "
2660 "orphan inode list. Please "
2661 "umount/remount instead.");
2662 err = -EINVAL;
2663 goto restore_opts;
2667 * Mounting a RDONLY partition read-write, so reread
2668 * and store the current valid flag. (It may have
2669 * been changed by e2fsck since we originally mounted
2670 * the partition.)
2672 ext3_clear_journal_err(sb, es);
2673 sbi->s_mount_state = le16_to_cpu(es->s_state);
2674 if ((err = ext3_group_extend(sb, es, n_blocks_count)))
2675 goto restore_opts;
2676 if (!ext3_setup_super (sb, es, 0))
2677 sb->s_flags &= ~MS_RDONLY;
2678 enable_quota = 1;
2681 #ifdef CONFIG_QUOTA
2682 /* Release old quota file names */
2683 for (i = 0; i < MAXQUOTAS; i++)
2684 if (old_opts.s_qf_names[i] &&
2685 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2686 kfree(old_opts.s_qf_names[i]);
2687 #endif
2688 unlock_super(sb);
2690 if (enable_quota)
2691 dquot_resume(sb, -1);
2692 return 0;
2693 restore_opts:
2694 sb->s_flags = old_sb_flags;
2695 sbi->s_mount_opt = old_opts.s_mount_opt;
2696 sbi->s_resuid = old_opts.s_resuid;
2697 sbi->s_resgid = old_opts.s_resgid;
2698 sbi->s_commit_interval = old_opts.s_commit_interval;
2699 #ifdef CONFIG_QUOTA
2700 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
2701 for (i = 0; i < MAXQUOTAS; i++) {
2702 if (sbi->s_qf_names[i] &&
2703 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2704 kfree(sbi->s_qf_names[i]);
2705 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
2707 #endif
2708 unlock_super(sb);
2709 return err;
2712 static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf)
2714 struct super_block *sb = dentry->d_sb;
2715 struct ext3_sb_info *sbi = EXT3_SB(sb);
2716 struct ext3_super_block *es = sbi->s_es;
2717 u64 fsid;
2719 if (test_opt(sb, MINIX_DF)) {
2720 sbi->s_overhead_last = 0;
2721 } else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
2722 unsigned long ngroups = sbi->s_groups_count, i;
2723 ext3_fsblk_t overhead = 0;
2724 smp_rmb();
2727 * Compute the overhead (FS structures). This is constant
2728 * for a given filesystem unless the number of block groups
2729 * changes so we cache the previous value until it does.
2733 * All of the blocks before first_data_block are
2734 * overhead
2736 overhead = le32_to_cpu(es->s_first_data_block);
2739 * Add the overhead attributed to the superblock and
2740 * block group descriptors. If the sparse superblocks
2741 * feature is turned on, then not all groups have this.
2743 for (i = 0; i < ngroups; i++) {
2744 overhead += ext3_bg_has_super(sb, i) +
2745 ext3_bg_num_gdb(sb, i);
2746 cond_resched();
2750 * Every block group has an inode bitmap, a block
2751 * bitmap, and an inode table.
2753 overhead += ngroups * (2 + sbi->s_itb_per_group);
2754 sbi->s_overhead_last = overhead;
2755 smp_wmb();
2756 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
2759 buf->f_type = EXT3_SUPER_MAGIC;
2760 buf->f_bsize = sb->s_blocksize;
2761 buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
2762 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter);
2763 buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
2764 if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
2765 buf->f_bavail = 0;
2766 buf->f_files = le32_to_cpu(es->s_inodes_count);
2767 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
2768 buf->f_namelen = EXT3_NAME_LEN;
2769 fsid = le64_to_cpup((void *)es->s_uuid) ^
2770 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
2771 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
2772 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
2773 return 0;
2776 /* Helper function for writing quotas on sync - we need to start transaction before quota file
2777 * is locked for write. Otherwise the are possible deadlocks:
2778 * Process 1 Process 2
2779 * ext3_create() quota_sync()
2780 * journal_start() write_dquot()
2781 * dquot_initialize() down(dqio_mutex)
2782 * down(dqio_mutex) journal_start()
2786 #ifdef CONFIG_QUOTA
2788 static inline struct inode *dquot_to_inode(struct dquot *dquot)
2790 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
2793 static int ext3_write_dquot(struct dquot *dquot)
2795 int ret, err;
2796 handle_t *handle;
2797 struct inode *inode;
2799 inode = dquot_to_inode(dquot);
2800 handle = ext3_journal_start(inode,
2801 EXT3_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
2802 if (IS_ERR(handle))
2803 return PTR_ERR(handle);
2804 ret = dquot_commit(dquot);
2805 err = ext3_journal_stop(handle);
2806 if (!ret)
2807 ret = err;
2808 return ret;
2811 static int ext3_acquire_dquot(struct dquot *dquot)
2813 int ret, err;
2814 handle_t *handle;
2816 handle = ext3_journal_start(dquot_to_inode(dquot),
2817 EXT3_QUOTA_INIT_BLOCKS(dquot->dq_sb));
2818 if (IS_ERR(handle))
2819 return PTR_ERR(handle);
2820 ret = dquot_acquire(dquot);
2821 err = ext3_journal_stop(handle);
2822 if (!ret)
2823 ret = err;
2824 return ret;
2827 static int ext3_release_dquot(struct dquot *dquot)
2829 int ret, err;
2830 handle_t *handle;
2832 handle = ext3_journal_start(dquot_to_inode(dquot),
2833 EXT3_QUOTA_DEL_BLOCKS(dquot->dq_sb));
2834 if (IS_ERR(handle)) {
2835 /* Release dquot anyway to avoid endless cycle in dqput() */
2836 dquot_release(dquot);
2837 return PTR_ERR(handle);
2839 ret = dquot_release(dquot);
2840 err = ext3_journal_stop(handle);
2841 if (!ret)
2842 ret = err;
2843 return ret;
2846 static int ext3_mark_dquot_dirty(struct dquot *dquot)
2848 /* Are we journaling quotas? */
2849 if (EXT3_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
2850 EXT3_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
2851 dquot_mark_dquot_dirty(dquot);
2852 return ext3_write_dquot(dquot);
2853 } else {
2854 return dquot_mark_dquot_dirty(dquot);
2858 static int ext3_write_info(struct super_block *sb, int type)
2860 int ret, err;
2861 handle_t *handle;
2863 /* Data block + inode block */
2864 handle = ext3_journal_start(sb->s_root->d_inode, 2);
2865 if (IS_ERR(handle))
2866 return PTR_ERR(handle);
2867 ret = dquot_commit_info(sb, type);
2868 err = ext3_journal_stop(handle);
2869 if (!ret)
2870 ret = err;
2871 return ret;
2875 * Turn on quotas during mount time - we need to find
2876 * the quota file and such...
2878 static int ext3_quota_on_mount(struct super_block *sb, int type)
2880 return dquot_quota_on_mount(sb, EXT3_SB(sb)->s_qf_names[type],
2881 EXT3_SB(sb)->s_jquota_fmt, type);
2885 * Standard function to be called on quota_on
2887 static int ext3_quota_on(struct super_block *sb, int type, int format_id,
2888 struct path *path)
2890 int err;
2892 if (!test_opt(sb, QUOTA))
2893 return -EINVAL;
2895 /* Quotafile not on the same filesystem? */
2896 if (path->mnt->mnt_sb != sb)
2897 return -EXDEV;
2898 /* Journaling quota? */
2899 if (EXT3_SB(sb)->s_qf_names[type]) {
2900 /* Quotafile not of fs root? */
2901 if (path->dentry->d_parent != sb->s_root)
2902 ext3_msg(sb, KERN_WARNING,
2903 "warning: Quota file not on filesystem root. "
2904 "Journaled quota will not work.");
2908 * When we journal data on quota file, we have to flush journal to see
2909 * all updates to the file when we bypass pagecache...
2911 if (ext3_should_journal_data(path->dentry->d_inode)) {
2913 * We don't need to lock updates but journal_flush() could
2914 * otherwise be livelocked...
2916 journal_lock_updates(EXT3_SB(sb)->s_journal);
2917 err = journal_flush(EXT3_SB(sb)->s_journal);
2918 journal_unlock_updates(EXT3_SB(sb)->s_journal);
2919 if (err)
2920 return err;
2923 return dquot_quota_on(sb, type, format_id, path);
2926 /* Read data from quotafile - avoid pagecache and such because we cannot afford
2927 * acquiring the locks... As quota files are never truncated and quota code
2928 * itself serializes the operations (and noone else should touch the files)
2929 * we don't have to be afraid of races */
2930 static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
2931 size_t len, loff_t off)
2933 struct inode *inode = sb_dqopt(sb)->files[type];
2934 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2935 int err = 0;
2936 int offset = off & (sb->s_blocksize - 1);
2937 int tocopy;
2938 size_t toread;
2939 struct buffer_head *bh;
2940 loff_t i_size = i_size_read(inode);
2942 if (off > i_size)
2943 return 0;
2944 if (off+len > i_size)
2945 len = i_size-off;
2946 toread = len;
2947 while (toread > 0) {
2948 tocopy = sb->s_blocksize - offset < toread ?
2949 sb->s_blocksize - offset : toread;
2950 bh = ext3_bread(NULL, inode, blk, 0, &err);
2951 if (err)
2952 return err;
2953 if (!bh) /* A hole? */
2954 memset(data, 0, tocopy);
2955 else
2956 memcpy(data, bh->b_data+offset, tocopy);
2957 brelse(bh);
2958 offset = 0;
2959 toread -= tocopy;
2960 data += tocopy;
2961 blk++;
2963 return len;
2966 /* Write to quotafile (we know the transaction is already started and has
2967 * enough credits) */
2968 static ssize_t ext3_quota_write(struct super_block *sb, int type,
2969 const char *data, size_t len, loff_t off)
2971 struct inode *inode = sb_dqopt(sb)->files[type];
2972 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2973 int err = 0;
2974 int offset = off & (sb->s_blocksize - 1);
2975 int journal_quota = EXT3_SB(sb)->s_qf_names[type] != NULL;
2976 struct buffer_head *bh;
2977 handle_t *handle = journal_current_handle();
2979 if (!handle) {
2980 ext3_msg(sb, KERN_WARNING,
2981 "warning: quota write (off=%llu, len=%llu)"
2982 " cancelled because transaction is not started.",
2983 (unsigned long long)off, (unsigned long long)len);
2984 return -EIO;
2988 * Since we account only one data block in transaction credits,
2989 * then it is impossible to cross a block boundary.
2991 if (sb->s_blocksize - offset < len) {
2992 ext3_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
2993 " cancelled because not block aligned",
2994 (unsigned long long)off, (unsigned long long)len);
2995 return -EIO;
2997 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
2998 bh = ext3_bread(handle, inode, blk, 1, &err);
2999 if (!bh)
3000 goto out;
3001 if (journal_quota) {
3002 err = ext3_journal_get_write_access(handle, bh);
3003 if (err) {
3004 brelse(bh);
3005 goto out;
3008 lock_buffer(bh);
3009 memcpy(bh->b_data+offset, data, len);
3010 flush_dcache_page(bh->b_page);
3011 unlock_buffer(bh);
3012 if (journal_quota)
3013 err = ext3_journal_dirty_metadata(handle, bh);
3014 else {
3015 /* Always do at least ordered writes for quotas */
3016 err = ext3_journal_dirty_data(handle, bh);
3017 mark_buffer_dirty(bh);
3019 brelse(bh);
3020 out:
3021 if (err) {
3022 mutex_unlock(&inode->i_mutex);
3023 return err;
3025 if (inode->i_size < off + len) {
3026 i_size_write(inode, off + len);
3027 EXT3_I(inode)->i_disksize = inode->i_size;
3029 inode->i_version++;
3030 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3031 ext3_mark_inode_dirty(handle, inode);
3032 mutex_unlock(&inode->i_mutex);
3033 return len;
3036 #endif
3038 static struct dentry *ext3_mount(struct file_system_type *fs_type,
3039 int flags, const char *dev_name, void *data)
3041 return mount_bdev(fs_type, flags, dev_name, data, ext3_fill_super);
3044 static struct file_system_type ext3_fs_type = {
3045 .owner = THIS_MODULE,
3046 .name = "ext3",
3047 .mount = ext3_mount,
3048 .kill_sb = kill_block_super,
3049 .fs_flags = FS_REQUIRES_DEV,
3052 static int __init init_ext3_fs(void)
3054 int err = init_ext3_xattr();
3055 if (err)
3056 return err;
3057 err = init_inodecache();
3058 if (err)
3059 goto out1;
3060 err = register_filesystem(&ext3_fs_type);
3061 if (err)
3062 goto out;
3063 return 0;
3064 out:
3065 destroy_inodecache();
3066 out1:
3067 exit_ext3_xattr();
3068 return err;
3071 static void __exit exit_ext3_fs(void)
3073 unregister_filesystem(&ext3_fs_type);
3074 destroy_inodecache();
3075 exit_ext3_xattr();
3078 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
3079 MODULE_DESCRIPTION("Second Extended Filesystem with journaling extensions");
3080 MODULE_LICENSE("GPL");
3081 module_init(init_ext3_fs)
3082 module_exit(exit_ext3_fs)