initial commit with v2.6.32.60
[linux-2.6.32.60-moxart.git] / fs / ext3 / super.c
blob9cfa28daf823de2a78487ec36d715ddbd7172142
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/smp_lock.h>
31 #include <linux/buffer_head.h>
32 #include <linux/exportfs.h>
33 #include <linux/vfs.h>
34 #include <linux/random.h>
35 #include <linux/mount.h>
36 #include <linux/namei.h>
37 #include <linux/quotaops.h>
38 #include <linux/seq_file.h>
39 #include <linux/log2.h>
41 #include <asm/uaccess.h>
43 #include "xattr.h"
44 #include "acl.h"
45 #include "namei.h"
47 #ifdef CONFIG_EXT3_DEFAULTS_TO_ORDERED
48 #define EXT3_MOUNT_DEFAULT_DATA_MODE EXT3_MOUNT_ORDERED_DATA
49 #else
50 #define EXT3_MOUNT_DEFAULT_DATA_MODE EXT3_MOUNT_WRITEBACK_DATA
51 #endif
53 static int ext3_load_journal(struct super_block *, struct ext3_super_block *,
54 unsigned long journal_devnum);
55 static int ext3_create_journal(struct super_block *, struct ext3_super_block *,
56 unsigned int);
57 static int ext3_commit_super(struct super_block *sb,
58 struct ext3_super_block *es,
59 int sync);
60 static void ext3_mark_recovery_complete(struct super_block * sb,
61 struct ext3_super_block * es);
62 static void ext3_clear_journal_err(struct super_block * sb,
63 struct ext3_super_block * es);
64 static int ext3_sync_fs(struct super_block *sb, int wait);
65 static const char *ext3_decode_error(struct super_block * sb, int errno,
66 char nbuf[16]);
67 static int ext3_remount (struct super_block * sb, int * flags, char * data);
68 static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf);
69 static int ext3_unfreeze(struct super_block *sb);
70 static int ext3_freeze(struct super_block *sb);
73 * Wrappers for journal_start/end.
75 * The only special thing we need to do here is to make sure that all
76 * journal_end calls result in the superblock being marked dirty, so
77 * that sync() will call the filesystem's write_super callback if
78 * appropriate.
80 handle_t *ext3_journal_start_sb(struct super_block *sb, int nblocks)
82 journal_t *journal;
84 if (sb->s_flags & MS_RDONLY)
85 return ERR_PTR(-EROFS);
87 /* Special case here: if the journal has aborted behind our
88 * backs (eg. EIO in the commit thread), then we still need to
89 * take the FS itself readonly cleanly. */
90 journal = EXT3_SB(sb)->s_journal;
91 if (is_journal_aborted(journal)) {
92 ext3_abort(sb, __func__,
93 "Detected aborted journal");
94 return ERR_PTR(-EROFS);
97 return journal_start(journal, nblocks);
101 * The only special thing we need to do here is to make sure that all
102 * journal_stop calls result in the superblock being marked dirty, so
103 * that sync() will call the filesystem's write_super callback if
104 * appropriate.
106 int __ext3_journal_stop(const char *where, handle_t *handle)
108 struct super_block *sb;
109 int err;
110 int rc;
112 sb = handle->h_transaction->t_journal->j_private;
113 err = handle->h_err;
114 rc = journal_stop(handle);
116 if (!err)
117 err = rc;
118 if (err)
119 __ext3_std_error(sb, where, err);
120 return err;
123 void ext3_journal_abort_handle(const char *caller, const char *err_fn,
124 struct buffer_head *bh, handle_t *handle, int err)
126 char nbuf[16];
127 const char *errstr = ext3_decode_error(NULL, err, nbuf);
129 if (bh)
130 BUFFER_TRACE(bh, "abort");
132 if (!handle->h_err)
133 handle->h_err = err;
135 if (is_handle_aborted(handle))
136 return;
138 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
139 caller, errstr, err_fn);
141 journal_abort_handle(handle);
144 /* Deal with the reporting of failure conditions on a filesystem such as
145 * inconsistencies detected or read IO failures.
147 * On ext2, we can store the error state of the filesystem in the
148 * superblock. That is not possible on ext3, because we may have other
149 * write ordering constraints on the superblock which prevent us from
150 * writing it out straight away; and given that the journal is about to
151 * be aborted, we can't rely on the current, or future, transactions to
152 * write out the superblock safely.
154 * We'll just use the journal_abort() error code to record an error in
155 * the journal instead. On recovery, the journal will compain about
156 * that error until we've noted it down and cleared it.
159 static void ext3_handle_error(struct super_block *sb)
161 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
163 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
164 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
166 if (sb->s_flags & MS_RDONLY)
167 return;
169 if (!test_opt (sb, ERRORS_CONT)) {
170 journal_t *journal = EXT3_SB(sb)->s_journal;
172 EXT3_SB(sb)->s_mount_opt |= EXT3_MOUNT_ABORT;
173 if (journal)
174 journal_abort(journal, -EIO);
176 if (test_opt (sb, ERRORS_RO)) {
177 printk (KERN_CRIT "Remounting filesystem read-only\n");
178 sb->s_flags |= MS_RDONLY;
180 ext3_commit_super(sb, es, 1);
181 if (test_opt(sb, ERRORS_PANIC))
182 panic("EXT3-fs (device %s): panic forced after error\n",
183 sb->s_id);
186 void ext3_error (struct super_block * sb, const char * function,
187 const char * fmt, ...)
189 va_list args;
191 va_start(args, fmt);
192 printk(KERN_CRIT "EXT3-fs error (device %s): %s: ",sb->s_id, function);
193 vprintk(fmt, args);
194 printk("\n");
195 va_end(args);
197 ext3_handle_error(sb);
200 static const char *ext3_decode_error(struct super_block * sb, int errno,
201 char nbuf[16])
203 char *errstr = NULL;
205 switch (errno) {
206 case -EIO:
207 errstr = "IO failure";
208 break;
209 case -ENOMEM:
210 errstr = "Out of memory";
211 break;
212 case -EROFS:
213 if (!sb || EXT3_SB(sb)->s_journal->j_flags & JFS_ABORT)
214 errstr = "Journal has aborted";
215 else
216 errstr = "Readonly filesystem";
217 break;
218 default:
219 /* If the caller passed in an extra buffer for unknown
220 * errors, textualise them now. Else we just return
221 * NULL. */
222 if (nbuf) {
223 /* Check for truncated error codes... */
224 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
225 errstr = nbuf;
227 break;
230 return errstr;
233 /* __ext3_std_error decodes expected errors from journaling functions
234 * automatically and invokes the appropriate error response. */
236 void __ext3_std_error (struct super_block * sb, const char * function,
237 int errno)
239 char nbuf[16];
240 const char *errstr;
242 /* Special case: if the error is EROFS, and we're not already
243 * inside a transaction, then there's really no point in logging
244 * an error. */
245 if (errno == -EROFS && journal_current_handle() == NULL &&
246 (sb->s_flags & MS_RDONLY))
247 return;
249 errstr = ext3_decode_error(sb, errno, nbuf);
250 printk (KERN_CRIT "EXT3-fs error (device %s) in %s: %s\n",
251 sb->s_id, function, errstr);
253 ext3_handle_error(sb);
257 * ext3_abort is a much stronger failure handler than ext3_error. The
258 * abort function may be used to deal with unrecoverable failures such
259 * as journal IO errors or ENOMEM at a critical moment in log management.
261 * We unconditionally force the filesystem into an ABORT|READONLY state,
262 * unless the error response on the fs has been set to panic in which
263 * case we take the easy way out and panic immediately.
266 void ext3_abort (struct super_block * sb, const char * function,
267 const char * fmt, ...)
269 va_list args;
271 printk (KERN_CRIT "ext3_abort called.\n");
273 va_start(args, fmt);
274 printk(KERN_CRIT "EXT3-fs error (device %s): %s: ",sb->s_id, function);
275 vprintk(fmt, args);
276 printk("\n");
277 va_end(args);
279 if (test_opt(sb, ERRORS_PANIC))
280 panic("EXT3-fs panic from previous error\n");
282 if (sb->s_flags & MS_RDONLY)
283 return;
285 printk(KERN_CRIT "Remounting filesystem read-only\n");
286 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
287 sb->s_flags |= MS_RDONLY;
288 EXT3_SB(sb)->s_mount_opt |= EXT3_MOUNT_ABORT;
289 if (EXT3_SB(sb)->s_journal)
290 journal_abort(EXT3_SB(sb)->s_journal, -EIO);
293 void ext3_warning (struct super_block * sb, const char * function,
294 const char * fmt, ...)
296 va_list args;
298 va_start(args, fmt);
299 printk(KERN_WARNING "EXT3-fs warning (device %s): %s: ",
300 sb->s_id, function);
301 vprintk(fmt, args);
302 printk("\n");
303 va_end(args);
306 void ext3_update_dynamic_rev(struct super_block *sb)
308 struct ext3_super_block *es = EXT3_SB(sb)->s_es;
310 if (le32_to_cpu(es->s_rev_level) > EXT3_GOOD_OLD_REV)
311 return;
313 ext3_warning(sb, __func__,
314 "updating to rev %d because of new feature flag, "
315 "running e2fsck is recommended",
316 EXT3_DYNAMIC_REV);
318 es->s_first_ino = cpu_to_le32(EXT3_GOOD_OLD_FIRST_INO);
319 es->s_inode_size = cpu_to_le16(EXT3_GOOD_OLD_INODE_SIZE);
320 es->s_rev_level = cpu_to_le32(EXT3_DYNAMIC_REV);
321 /* leave es->s_feature_*compat flags alone */
322 /* es->s_uuid will be set by e2fsck if empty */
325 * The rest of the superblock fields should be zero, and if not it
326 * means they are likely already in use, so leave them alone. We
327 * can leave it up to e2fsck to clean up any inconsistencies there.
332 * Open the external journal device
334 static struct block_device *ext3_blkdev_get(dev_t dev)
336 struct block_device *bdev;
337 char b[BDEVNAME_SIZE];
339 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
340 if (IS_ERR(bdev))
341 goto fail;
342 return bdev;
344 fail:
345 printk(KERN_ERR "EXT3: failed to open journal device %s: %ld\n",
346 __bdevname(dev, b), PTR_ERR(bdev));
347 return NULL;
351 * Release the journal device
353 static int ext3_blkdev_put(struct block_device *bdev)
355 bd_release(bdev);
356 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
359 static int ext3_blkdev_remove(struct ext3_sb_info *sbi)
361 struct block_device *bdev;
362 int ret = -ENODEV;
364 bdev = sbi->journal_bdev;
365 if (bdev) {
366 ret = ext3_blkdev_put(bdev);
367 sbi->journal_bdev = NULL;
369 return ret;
372 static inline struct inode *orphan_list_entry(struct list_head *l)
374 return &list_entry(l, struct ext3_inode_info, i_orphan)->vfs_inode;
377 static void dump_orphan_list(struct super_block *sb, struct ext3_sb_info *sbi)
379 struct list_head *l;
381 printk(KERN_ERR "sb orphan head is %d\n",
382 le32_to_cpu(sbi->s_es->s_last_orphan));
384 printk(KERN_ERR "sb_info orphan list:\n");
385 list_for_each(l, &sbi->s_orphan) {
386 struct inode *inode = orphan_list_entry(l);
387 printk(KERN_ERR " "
388 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
389 inode->i_sb->s_id, inode->i_ino, inode,
390 inode->i_mode, inode->i_nlink,
391 NEXT_ORPHAN(inode));
395 static void ext3_put_super (struct super_block * sb)
397 struct ext3_sb_info *sbi = EXT3_SB(sb);
398 struct ext3_super_block *es = sbi->s_es;
399 int i, err;
401 lock_kernel();
403 ext3_xattr_put_super(sb);
404 err = journal_destroy(sbi->s_journal);
405 sbi->s_journal = NULL;
406 if (err < 0)
407 ext3_abort(sb, __func__, "Couldn't clean up the journal");
409 if (!(sb->s_flags & MS_RDONLY)) {
410 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
411 es->s_state = cpu_to_le16(sbi->s_mount_state);
412 BUFFER_TRACE(sbi->s_sbh, "marking dirty");
413 mark_buffer_dirty(sbi->s_sbh);
414 ext3_commit_super(sb, es, 1);
417 for (i = 0; i < sbi->s_gdb_count; i++)
418 brelse(sbi->s_group_desc[i]);
419 kfree(sbi->s_group_desc);
420 percpu_counter_destroy(&sbi->s_freeblocks_counter);
421 percpu_counter_destroy(&sbi->s_freeinodes_counter);
422 percpu_counter_destroy(&sbi->s_dirs_counter);
423 brelse(sbi->s_sbh);
424 #ifdef CONFIG_QUOTA
425 for (i = 0; i < MAXQUOTAS; i++)
426 kfree(sbi->s_qf_names[i]);
427 #endif
429 /* Debugging code just in case the in-memory inode orphan list
430 * isn't empty. The on-disk one can be non-empty if we've
431 * detected an error and taken the fs readonly, but the
432 * in-memory list had better be clean by this point. */
433 if (!list_empty(&sbi->s_orphan))
434 dump_orphan_list(sb, sbi);
435 J_ASSERT(list_empty(&sbi->s_orphan));
437 invalidate_bdev(sb->s_bdev);
438 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
440 * Invalidate the journal device's buffers. We don't want them
441 * floating about in memory - the physical journal device may
442 * hotswapped, and it breaks the `ro-after' testing code.
444 sync_blockdev(sbi->journal_bdev);
445 invalidate_bdev(sbi->journal_bdev);
446 ext3_blkdev_remove(sbi);
448 sb->s_fs_info = NULL;
449 kfree(sbi->s_blockgroup_lock);
450 kfree(sbi);
452 unlock_kernel();
455 static struct kmem_cache *ext3_inode_cachep;
458 * Called inside transaction, so use GFP_NOFS
460 static struct inode *ext3_alloc_inode(struct super_block *sb)
462 struct ext3_inode_info *ei;
464 ei = kmem_cache_alloc(ext3_inode_cachep, GFP_NOFS);
465 if (!ei)
466 return NULL;
467 ei->i_block_alloc_info = NULL;
468 ei->vfs_inode.i_version = 1;
469 atomic_set(&ei->i_datasync_tid, 0);
470 atomic_set(&ei->i_sync_tid, 0);
471 return &ei->vfs_inode;
474 static void ext3_destroy_inode(struct inode *inode)
476 if (!list_empty(&(EXT3_I(inode)->i_orphan))) {
477 printk("EXT3 Inode %p: orphan list check failed!\n",
478 EXT3_I(inode));
479 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
480 EXT3_I(inode), sizeof(struct ext3_inode_info),
481 false);
482 dump_stack();
484 kmem_cache_free(ext3_inode_cachep, EXT3_I(inode));
487 static void init_once(void *foo)
489 struct ext3_inode_info *ei = (struct ext3_inode_info *) foo;
491 INIT_LIST_HEAD(&ei->i_orphan);
492 #ifdef CONFIG_EXT3_FS_XATTR
493 init_rwsem(&ei->xattr_sem);
494 #endif
495 mutex_init(&ei->truncate_mutex);
496 inode_init_once(&ei->vfs_inode);
499 static int init_inodecache(void)
501 ext3_inode_cachep = kmem_cache_create("ext3_inode_cache",
502 sizeof(struct ext3_inode_info),
503 0, (SLAB_RECLAIM_ACCOUNT|
504 SLAB_MEM_SPREAD),
505 init_once);
506 if (ext3_inode_cachep == NULL)
507 return -ENOMEM;
508 return 0;
511 static void destroy_inodecache(void)
513 kmem_cache_destroy(ext3_inode_cachep);
516 static void ext3_clear_inode(struct inode *inode)
518 struct ext3_block_alloc_info *rsv = EXT3_I(inode)->i_block_alloc_info;
519 ext3_discard_reservation(inode);
520 EXT3_I(inode)->i_block_alloc_info = NULL;
521 if (unlikely(rsv))
522 kfree(rsv);
525 static inline void ext3_show_quota_options(struct seq_file *seq, struct super_block *sb)
527 #if defined(CONFIG_QUOTA)
528 struct ext3_sb_info *sbi = EXT3_SB(sb);
530 if (sbi->s_jquota_fmt)
531 seq_printf(seq, ",jqfmt=%s",
532 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold": "vfsv0");
534 if (sbi->s_qf_names[USRQUOTA])
535 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
537 if (sbi->s_qf_names[GRPQUOTA])
538 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
540 if (sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA)
541 seq_puts(seq, ",usrquota");
543 if (sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA)
544 seq_puts(seq, ",grpquota");
545 #endif
548 static char *data_mode_string(unsigned long mode)
550 switch (mode) {
551 case EXT3_MOUNT_JOURNAL_DATA:
552 return "journal";
553 case EXT3_MOUNT_ORDERED_DATA:
554 return "ordered";
555 case EXT3_MOUNT_WRITEBACK_DATA:
556 return "writeback";
558 return "unknown";
562 * Show an option if
563 * - it's set to a non-default value OR
564 * - if the per-sb default is different from the global default
566 static int ext3_show_options(struct seq_file *seq, struct vfsmount *vfs)
568 struct super_block *sb = vfs->mnt_sb;
569 struct ext3_sb_info *sbi = EXT3_SB(sb);
570 struct ext3_super_block *es = sbi->s_es;
571 unsigned long def_mount_opts;
573 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
575 if (sbi->s_sb_block != 1)
576 seq_printf(seq, ",sb=%lu", sbi->s_sb_block);
577 if (test_opt(sb, MINIX_DF))
578 seq_puts(seq, ",minixdf");
579 if (test_opt(sb, GRPID))
580 seq_puts(seq, ",grpid");
581 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT3_DEFM_BSDGROUPS))
582 seq_puts(seq, ",nogrpid");
583 if (sbi->s_resuid != EXT3_DEF_RESUID ||
584 le16_to_cpu(es->s_def_resuid) != EXT3_DEF_RESUID) {
585 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
587 if (sbi->s_resgid != EXT3_DEF_RESGID ||
588 le16_to_cpu(es->s_def_resgid) != EXT3_DEF_RESGID) {
589 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
591 if (test_opt(sb, ERRORS_RO)) {
592 int def_errors = le16_to_cpu(es->s_errors);
594 if (def_errors == EXT3_ERRORS_PANIC ||
595 def_errors == EXT3_ERRORS_CONTINUE) {
596 seq_puts(seq, ",errors=remount-ro");
599 if (test_opt(sb, ERRORS_CONT))
600 seq_puts(seq, ",errors=continue");
601 if (test_opt(sb, ERRORS_PANIC))
602 seq_puts(seq, ",errors=panic");
603 if (test_opt(sb, NO_UID32))
604 seq_puts(seq, ",nouid32");
605 if (test_opt(sb, DEBUG))
606 seq_puts(seq, ",debug");
607 if (test_opt(sb, OLDALLOC))
608 seq_puts(seq, ",oldalloc");
609 #ifdef CONFIG_EXT3_FS_XATTR
610 if (test_opt(sb, XATTR_USER))
611 seq_puts(seq, ",user_xattr");
612 if (!test_opt(sb, XATTR_USER) &&
613 (def_mount_opts & EXT3_DEFM_XATTR_USER)) {
614 seq_puts(seq, ",nouser_xattr");
616 #endif
617 #ifdef CONFIG_EXT3_FS_POSIX_ACL
618 if (test_opt(sb, POSIX_ACL))
619 seq_puts(seq, ",acl");
620 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT3_DEFM_ACL))
621 seq_puts(seq, ",noacl");
622 #endif
623 if (!test_opt(sb, RESERVATION))
624 seq_puts(seq, ",noreservation");
625 if (sbi->s_commit_interval) {
626 seq_printf(seq, ",commit=%u",
627 (unsigned) (sbi->s_commit_interval / HZ));
629 if (test_opt(sb, BARRIER))
630 seq_puts(seq, ",barrier=1");
631 if (test_opt(sb, NOBH))
632 seq_puts(seq, ",nobh");
634 seq_printf(seq, ",data=%s", data_mode_string(sbi->s_mount_opt &
635 EXT3_MOUNT_DATA_FLAGS));
636 if (test_opt(sb, DATA_ERR_ABORT))
637 seq_puts(seq, ",data_err=abort");
639 ext3_show_quota_options(seq, sb);
641 return 0;
645 static struct inode *ext3_nfs_get_inode(struct super_block *sb,
646 u64 ino, u32 generation)
648 struct inode *inode;
650 if (ino < EXT3_FIRST_INO(sb) && ino != EXT3_ROOT_INO)
651 return ERR_PTR(-ESTALE);
652 if (ino > le32_to_cpu(EXT3_SB(sb)->s_es->s_inodes_count))
653 return ERR_PTR(-ESTALE);
655 /* iget isn't really right if the inode is currently unallocated!!
657 * ext3_read_inode will return a bad_inode if the inode had been
658 * deleted, so we should be safe.
660 * Currently we don't know the generation for parent directory, so
661 * a generation of 0 means "accept any"
663 inode = ext3_iget(sb, ino);
664 if (IS_ERR(inode))
665 return ERR_CAST(inode);
666 if (generation && inode->i_generation != generation) {
667 iput(inode);
668 return ERR_PTR(-ESTALE);
671 return inode;
674 static struct dentry *ext3_fh_to_dentry(struct super_block *sb, struct fid *fid,
675 int fh_len, int fh_type)
677 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
678 ext3_nfs_get_inode);
681 static struct dentry *ext3_fh_to_parent(struct super_block *sb, struct fid *fid,
682 int fh_len, int fh_type)
684 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
685 ext3_nfs_get_inode);
689 * Try to release metadata pages (indirect blocks, directories) which are
690 * mapped via the block device. Since these pages could have journal heads
691 * which would prevent try_to_free_buffers() from freeing them, we must use
692 * jbd layer's try_to_free_buffers() function to release them.
694 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
695 gfp_t wait)
697 journal_t *journal = EXT3_SB(sb)->s_journal;
699 WARN_ON(PageChecked(page));
700 if (!page_has_buffers(page))
701 return 0;
702 if (journal)
703 return journal_try_to_free_buffers(journal, page,
704 wait & ~__GFP_WAIT);
705 return try_to_free_buffers(page);
708 #ifdef CONFIG_QUOTA
709 #define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
710 #define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
712 static int ext3_write_dquot(struct dquot *dquot);
713 static int ext3_acquire_dquot(struct dquot *dquot);
714 static int ext3_release_dquot(struct dquot *dquot);
715 static int ext3_mark_dquot_dirty(struct dquot *dquot);
716 static int ext3_write_info(struct super_block *sb, int type);
717 static int ext3_quota_on(struct super_block *sb, int type, int format_id,
718 char *path, int remount);
719 static int ext3_quota_on_mount(struct super_block *sb, int type);
720 static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
721 size_t len, loff_t off);
722 static ssize_t ext3_quota_write(struct super_block *sb, int type,
723 const char *data, size_t len, loff_t off);
725 static const struct dquot_operations ext3_quota_operations = {
726 .initialize = dquot_initialize,
727 .drop = dquot_drop,
728 .alloc_space = dquot_alloc_space,
729 .alloc_inode = dquot_alloc_inode,
730 .free_space = dquot_free_space,
731 .free_inode = dquot_free_inode,
732 .transfer = dquot_transfer,
733 .write_dquot = ext3_write_dquot,
734 .acquire_dquot = ext3_acquire_dquot,
735 .release_dquot = ext3_release_dquot,
736 .mark_dirty = ext3_mark_dquot_dirty,
737 .write_info = ext3_write_info,
738 .alloc_dquot = dquot_alloc,
739 .destroy_dquot = dquot_destroy,
742 static const struct quotactl_ops ext3_qctl_operations = {
743 .quota_on = ext3_quota_on,
744 .quota_off = vfs_quota_off,
745 .quota_sync = vfs_quota_sync,
746 .get_info = vfs_get_dqinfo,
747 .set_info = vfs_set_dqinfo,
748 .get_dqblk = vfs_get_dqblk,
749 .set_dqblk = vfs_set_dqblk
751 #endif
753 static const struct super_operations ext3_sops = {
754 .alloc_inode = ext3_alloc_inode,
755 .destroy_inode = ext3_destroy_inode,
756 .write_inode = ext3_write_inode,
757 .dirty_inode = ext3_dirty_inode,
758 .delete_inode = ext3_delete_inode,
759 .put_super = ext3_put_super,
760 .sync_fs = ext3_sync_fs,
761 .freeze_fs = ext3_freeze,
762 .unfreeze_fs = ext3_unfreeze,
763 .statfs = ext3_statfs,
764 .remount_fs = ext3_remount,
765 .clear_inode = ext3_clear_inode,
766 .show_options = ext3_show_options,
767 #ifdef CONFIG_QUOTA
768 .quota_read = ext3_quota_read,
769 .quota_write = ext3_quota_write,
770 #endif
771 .bdev_try_to_free_page = bdev_try_to_free_page,
774 static const struct export_operations ext3_export_ops = {
775 .fh_to_dentry = ext3_fh_to_dentry,
776 .fh_to_parent = ext3_fh_to_parent,
777 .get_parent = ext3_get_parent,
780 enum {
781 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
782 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
783 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
784 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
785 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
786 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
787 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
788 Opt_data_err_abort, Opt_data_err_ignore,
789 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
790 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
791 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
792 Opt_grpquota
795 static const match_table_t tokens = {
796 {Opt_bsd_df, "bsddf"},
797 {Opt_minix_df, "minixdf"},
798 {Opt_grpid, "grpid"},
799 {Opt_grpid, "bsdgroups"},
800 {Opt_nogrpid, "nogrpid"},
801 {Opt_nogrpid, "sysvgroups"},
802 {Opt_resgid, "resgid=%u"},
803 {Opt_resuid, "resuid=%u"},
804 {Opt_sb, "sb=%u"},
805 {Opt_err_cont, "errors=continue"},
806 {Opt_err_panic, "errors=panic"},
807 {Opt_err_ro, "errors=remount-ro"},
808 {Opt_nouid32, "nouid32"},
809 {Opt_nocheck, "nocheck"},
810 {Opt_nocheck, "check=none"},
811 {Opt_debug, "debug"},
812 {Opt_oldalloc, "oldalloc"},
813 {Opt_orlov, "orlov"},
814 {Opt_user_xattr, "user_xattr"},
815 {Opt_nouser_xattr, "nouser_xattr"},
816 {Opt_acl, "acl"},
817 {Opt_noacl, "noacl"},
818 {Opt_reservation, "reservation"},
819 {Opt_noreservation, "noreservation"},
820 {Opt_noload, "noload"},
821 {Opt_nobh, "nobh"},
822 {Opt_bh, "bh"},
823 {Opt_commit, "commit=%u"},
824 {Opt_journal_update, "journal=update"},
825 {Opt_journal_inum, "journal=%u"},
826 {Opt_journal_dev, "journal_dev=%u"},
827 {Opt_abort, "abort"},
828 {Opt_data_journal, "data=journal"},
829 {Opt_data_ordered, "data=ordered"},
830 {Opt_data_writeback, "data=writeback"},
831 {Opt_data_err_abort, "data_err=abort"},
832 {Opt_data_err_ignore, "data_err=ignore"},
833 {Opt_offusrjquota, "usrjquota="},
834 {Opt_usrjquota, "usrjquota=%s"},
835 {Opt_offgrpjquota, "grpjquota="},
836 {Opt_grpjquota, "grpjquota=%s"},
837 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
838 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
839 {Opt_grpquota, "grpquota"},
840 {Opt_noquota, "noquota"},
841 {Opt_quota, "quota"},
842 {Opt_usrquota, "usrquota"},
843 {Opt_barrier, "barrier=%u"},
844 {Opt_resize, "resize"},
845 {Opt_err, NULL},
848 static ext3_fsblk_t get_sb_block(void **data)
850 ext3_fsblk_t sb_block;
851 char *options = (char *) *data;
853 if (!options || strncmp(options, "sb=", 3) != 0)
854 return 1; /* Default location */
855 options += 3;
856 /*todo: use simple_strtoll with >32bit ext3 */
857 sb_block = simple_strtoul(options, &options, 0);
858 if (*options && *options != ',') {
859 printk("EXT3-fs: Invalid sb specification: %s\n",
860 (char *) *data);
861 return 1;
863 if (*options == ',')
864 options++;
865 *data = (void *) options;
866 return sb_block;
869 static int parse_options (char *options, struct super_block *sb,
870 unsigned int *inum, unsigned long *journal_devnum,
871 ext3_fsblk_t *n_blocks_count, int is_remount)
873 struct ext3_sb_info *sbi = EXT3_SB(sb);
874 char * p;
875 substring_t args[MAX_OPT_ARGS];
876 int data_opt = 0;
877 int option;
878 #ifdef CONFIG_QUOTA
879 int qtype, qfmt;
880 char *qname;
881 #endif
883 if (!options)
884 return 1;
886 while ((p = strsep (&options, ",")) != NULL) {
887 int token;
888 if (!*p)
889 continue;
891 token = match_token(p, tokens, args);
892 switch (token) {
893 case Opt_bsd_df:
894 clear_opt (sbi->s_mount_opt, MINIX_DF);
895 break;
896 case Opt_minix_df:
897 set_opt (sbi->s_mount_opt, MINIX_DF);
898 break;
899 case Opt_grpid:
900 set_opt (sbi->s_mount_opt, GRPID);
901 break;
902 case Opt_nogrpid:
903 clear_opt (sbi->s_mount_opt, GRPID);
904 break;
905 case Opt_resuid:
906 if (match_int(&args[0], &option))
907 return 0;
908 sbi->s_resuid = option;
909 break;
910 case Opt_resgid:
911 if (match_int(&args[0], &option))
912 return 0;
913 sbi->s_resgid = option;
914 break;
915 case Opt_sb:
916 /* handled by get_sb_block() instead of here */
917 /* *sb_block = match_int(&args[0]); */
918 break;
919 case Opt_err_panic:
920 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
921 clear_opt (sbi->s_mount_opt, ERRORS_RO);
922 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
923 break;
924 case Opt_err_ro:
925 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
926 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
927 set_opt (sbi->s_mount_opt, ERRORS_RO);
928 break;
929 case Opt_err_cont:
930 clear_opt (sbi->s_mount_opt, ERRORS_RO);
931 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
932 set_opt (sbi->s_mount_opt, ERRORS_CONT);
933 break;
934 case Opt_nouid32:
935 set_opt (sbi->s_mount_opt, NO_UID32);
936 break;
937 case Opt_nocheck:
938 clear_opt (sbi->s_mount_opt, CHECK);
939 break;
940 case Opt_debug:
941 set_opt (sbi->s_mount_opt, DEBUG);
942 break;
943 case Opt_oldalloc:
944 set_opt (sbi->s_mount_opt, OLDALLOC);
945 break;
946 case Opt_orlov:
947 clear_opt (sbi->s_mount_opt, OLDALLOC);
948 break;
949 #ifdef CONFIG_EXT3_FS_XATTR
950 case Opt_user_xattr:
951 set_opt (sbi->s_mount_opt, XATTR_USER);
952 break;
953 case Opt_nouser_xattr:
954 clear_opt (sbi->s_mount_opt, XATTR_USER);
955 break;
956 #else
957 case Opt_user_xattr:
958 case Opt_nouser_xattr:
959 printk("EXT3 (no)user_xattr options not supported\n");
960 break;
961 #endif
962 #ifdef CONFIG_EXT3_FS_POSIX_ACL
963 case Opt_acl:
964 set_opt(sbi->s_mount_opt, POSIX_ACL);
965 break;
966 case Opt_noacl:
967 clear_opt(sbi->s_mount_opt, POSIX_ACL);
968 break;
969 #else
970 case Opt_acl:
971 case Opt_noacl:
972 printk("EXT3 (no)acl options not supported\n");
973 break;
974 #endif
975 case Opt_reservation:
976 set_opt(sbi->s_mount_opt, RESERVATION);
977 break;
978 case Opt_noreservation:
979 clear_opt(sbi->s_mount_opt, RESERVATION);
980 break;
981 case Opt_journal_update:
982 /* @@@ FIXME */
983 /* Eventually we will want to be able to create
984 a journal file here. For now, only allow the
985 user to specify an existing inode to be the
986 journal file. */
987 if (is_remount) {
988 printk(KERN_ERR "EXT3-fs: cannot specify "
989 "journal on remount\n");
990 return 0;
992 set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
993 break;
994 case Opt_journal_inum:
995 if (is_remount) {
996 printk(KERN_ERR "EXT3-fs: cannot specify "
997 "journal on remount\n");
998 return 0;
1000 if (match_int(&args[0], &option))
1001 return 0;
1002 *inum = option;
1003 break;
1004 case Opt_journal_dev:
1005 if (is_remount) {
1006 printk(KERN_ERR "EXT3-fs: cannot specify "
1007 "journal on remount\n");
1008 return 0;
1010 if (match_int(&args[0], &option))
1011 return 0;
1012 *journal_devnum = option;
1013 break;
1014 case Opt_noload:
1015 set_opt (sbi->s_mount_opt, NOLOAD);
1016 break;
1017 case Opt_commit:
1018 if (match_int(&args[0], &option))
1019 return 0;
1020 if (option < 0)
1021 return 0;
1022 if (option == 0)
1023 option = JBD_DEFAULT_MAX_COMMIT_AGE;
1024 sbi->s_commit_interval = HZ * option;
1025 break;
1026 case Opt_data_journal:
1027 data_opt = EXT3_MOUNT_JOURNAL_DATA;
1028 goto datacheck;
1029 case Opt_data_ordered:
1030 data_opt = EXT3_MOUNT_ORDERED_DATA;
1031 goto datacheck;
1032 case Opt_data_writeback:
1033 data_opt = EXT3_MOUNT_WRITEBACK_DATA;
1034 datacheck:
1035 if (is_remount) {
1036 if ((sbi->s_mount_opt & EXT3_MOUNT_DATA_FLAGS)
1037 == data_opt)
1038 break;
1039 printk(KERN_ERR
1040 "EXT3-fs (device %s): Cannot change "
1041 "data mode on remount. The filesystem "
1042 "is mounted in data=%s mode and you "
1043 "try to remount it in data=%s mode.\n",
1044 sb->s_id,
1045 data_mode_string(sbi->s_mount_opt &
1046 EXT3_MOUNT_DATA_FLAGS),
1047 data_mode_string(data_opt));
1048 return 0;
1049 } else {
1050 sbi->s_mount_opt &= ~EXT3_MOUNT_DATA_FLAGS;
1051 sbi->s_mount_opt |= data_opt;
1053 break;
1054 case Opt_data_err_abort:
1055 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1056 break;
1057 case Opt_data_err_ignore:
1058 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1059 break;
1060 #ifdef CONFIG_QUOTA
1061 case Opt_usrjquota:
1062 qtype = USRQUOTA;
1063 goto set_qf_name;
1064 case Opt_grpjquota:
1065 qtype = GRPQUOTA;
1066 set_qf_name:
1067 if (sb_any_quota_loaded(sb) &&
1068 !sbi->s_qf_names[qtype]) {
1069 printk(KERN_ERR
1070 "EXT3-fs: Cannot change journaled "
1071 "quota options when quota turned on.\n");
1072 return 0;
1074 qname = match_strdup(&args[0]);
1075 if (!qname) {
1076 printk(KERN_ERR
1077 "EXT3-fs: not enough memory for "
1078 "storing quotafile name.\n");
1079 return 0;
1081 if (sbi->s_qf_names[qtype] &&
1082 strcmp(sbi->s_qf_names[qtype], qname)) {
1083 printk(KERN_ERR
1084 "EXT3-fs: %s quota file already "
1085 "specified.\n", QTYPE2NAME(qtype));
1086 kfree(qname);
1087 return 0;
1089 sbi->s_qf_names[qtype] = qname;
1090 if (strchr(sbi->s_qf_names[qtype], '/')) {
1091 printk(KERN_ERR
1092 "EXT3-fs: quotafile must be on "
1093 "filesystem root.\n");
1094 kfree(sbi->s_qf_names[qtype]);
1095 sbi->s_qf_names[qtype] = NULL;
1096 return 0;
1098 set_opt(sbi->s_mount_opt, QUOTA);
1099 break;
1100 case Opt_offusrjquota:
1101 qtype = USRQUOTA;
1102 goto clear_qf_name;
1103 case Opt_offgrpjquota:
1104 qtype = GRPQUOTA;
1105 clear_qf_name:
1106 if (sb_any_quota_loaded(sb) &&
1107 sbi->s_qf_names[qtype]) {
1108 printk(KERN_ERR "EXT3-fs: Cannot change "
1109 "journaled quota options when "
1110 "quota turned on.\n");
1111 return 0;
1114 * The space will be released later when all options
1115 * are confirmed to be correct
1117 sbi->s_qf_names[qtype] = NULL;
1118 break;
1119 case Opt_jqfmt_vfsold:
1120 qfmt = QFMT_VFS_OLD;
1121 goto set_qf_format;
1122 case Opt_jqfmt_vfsv0:
1123 qfmt = QFMT_VFS_V0;
1124 set_qf_format:
1125 if (sb_any_quota_loaded(sb) &&
1126 sbi->s_jquota_fmt != qfmt) {
1127 printk(KERN_ERR "EXT3-fs: Cannot change "
1128 "journaled quota options when "
1129 "quota turned on.\n");
1130 return 0;
1132 sbi->s_jquota_fmt = qfmt;
1133 break;
1134 case Opt_quota:
1135 case Opt_usrquota:
1136 set_opt(sbi->s_mount_opt, QUOTA);
1137 set_opt(sbi->s_mount_opt, USRQUOTA);
1138 break;
1139 case Opt_grpquota:
1140 set_opt(sbi->s_mount_opt, QUOTA);
1141 set_opt(sbi->s_mount_opt, GRPQUOTA);
1142 break;
1143 case Opt_noquota:
1144 if (sb_any_quota_loaded(sb)) {
1145 printk(KERN_ERR "EXT3-fs: Cannot change quota "
1146 "options when quota turned on.\n");
1147 return 0;
1149 clear_opt(sbi->s_mount_opt, QUOTA);
1150 clear_opt(sbi->s_mount_opt, USRQUOTA);
1151 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1152 break;
1153 #else
1154 case Opt_quota:
1155 case Opt_usrquota:
1156 case Opt_grpquota:
1157 printk(KERN_ERR
1158 "EXT3-fs: quota options not supported.\n");
1159 break;
1160 case Opt_usrjquota:
1161 case Opt_grpjquota:
1162 case Opt_offusrjquota:
1163 case Opt_offgrpjquota:
1164 case Opt_jqfmt_vfsold:
1165 case Opt_jqfmt_vfsv0:
1166 printk(KERN_ERR
1167 "EXT3-fs: journaled quota options not "
1168 "supported.\n");
1169 break;
1170 case Opt_noquota:
1171 break;
1172 #endif
1173 case Opt_abort:
1174 set_opt(sbi->s_mount_opt, ABORT);
1175 break;
1176 case Opt_barrier:
1177 if (match_int(&args[0], &option))
1178 return 0;
1179 if (option)
1180 set_opt(sbi->s_mount_opt, BARRIER);
1181 else
1182 clear_opt(sbi->s_mount_opt, BARRIER);
1183 break;
1184 case Opt_ignore:
1185 break;
1186 case Opt_resize:
1187 if (!is_remount) {
1188 printk("EXT3-fs: resize option only available "
1189 "for remount\n");
1190 return 0;
1192 if (match_int(&args[0], &option) != 0)
1193 return 0;
1194 *n_blocks_count = option;
1195 break;
1196 case Opt_nobh:
1197 set_opt(sbi->s_mount_opt, NOBH);
1198 break;
1199 case Opt_bh:
1200 clear_opt(sbi->s_mount_opt, NOBH);
1201 break;
1202 default:
1203 printk (KERN_ERR
1204 "EXT3-fs: Unrecognized mount option \"%s\" "
1205 "or missing value\n", p);
1206 return 0;
1209 #ifdef CONFIG_QUOTA
1210 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1211 if ((sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA) &&
1212 sbi->s_qf_names[USRQUOTA])
1213 clear_opt(sbi->s_mount_opt, USRQUOTA);
1215 if ((sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA) &&
1216 sbi->s_qf_names[GRPQUOTA])
1217 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1219 if ((sbi->s_qf_names[USRQUOTA] &&
1220 (sbi->s_mount_opt & EXT3_MOUNT_GRPQUOTA)) ||
1221 (sbi->s_qf_names[GRPQUOTA] &&
1222 (sbi->s_mount_opt & EXT3_MOUNT_USRQUOTA))) {
1223 printk(KERN_ERR "EXT3-fs: old and new quota "
1224 "format mixing.\n");
1225 return 0;
1228 if (!sbi->s_jquota_fmt) {
1229 printk(KERN_ERR "EXT3-fs: journaled quota format "
1230 "not specified.\n");
1231 return 0;
1233 } else {
1234 if (sbi->s_jquota_fmt) {
1235 printk(KERN_ERR "EXT3-fs: journaled quota format "
1236 "specified with no journaling "
1237 "enabled.\n");
1238 return 0;
1241 #endif
1242 return 1;
1245 static int ext3_setup_super(struct super_block *sb, struct ext3_super_block *es,
1246 int read_only)
1248 struct ext3_sb_info *sbi = EXT3_SB(sb);
1249 int res = 0;
1251 if (le32_to_cpu(es->s_rev_level) > EXT3_MAX_SUPP_REV) {
1252 printk (KERN_ERR "EXT3-fs warning: revision level too high, "
1253 "forcing read-only mode\n");
1254 res = MS_RDONLY;
1256 if (read_only)
1257 return res;
1258 if (!(sbi->s_mount_state & EXT3_VALID_FS))
1259 printk (KERN_WARNING "EXT3-fs warning: mounting unchecked fs, "
1260 "running e2fsck is recommended\n");
1261 else if ((sbi->s_mount_state & EXT3_ERROR_FS))
1262 printk (KERN_WARNING
1263 "EXT3-fs warning: mounting fs with errors, "
1264 "running e2fsck is recommended\n");
1265 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1266 le16_to_cpu(es->s_mnt_count) >=
1267 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1268 printk (KERN_WARNING
1269 "EXT3-fs warning: maximal mount count reached, "
1270 "running e2fsck is recommended\n");
1271 else if (le32_to_cpu(es->s_checkinterval) &&
1272 (le32_to_cpu(es->s_lastcheck) +
1273 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1274 printk (KERN_WARNING
1275 "EXT3-fs warning: checktime reached, "
1276 "running e2fsck is recommended\n");
1277 #if 0
1278 /* @@@ We _will_ want to clear the valid bit if we find
1279 inconsistencies, to force a fsck at reboot. But for
1280 a plain journaled filesystem we can keep it set as
1281 valid forever! :) */
1282 es->s_state &= cpu_to_le16(~EXT3_VALID_FS);
1283 #endif
1284 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1285 es->s_max_mnt_count = cpu_to_le16(EXT3_DFL_MAX_MNT_COUNT);
1286 le16_add_cpu(&es->s_mnt_count, 1);
1287 es->s_mtime = cpu_to_le32(get_seconds());
1288 ext3_update_dynamic_rev(sb);
1289 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
1291 ext3_commit_super(sb, es, 1);
1292 if (test_opt(sb, DEBUG))
1293 printk(KERN_INFO "[EXT3 FS bs=%lu, gc=%lu, "
1294 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1295 sb->s_blocksize,
1296 sbi->s_groups_count,
1297 EXT3_BLOCKS_PER_GROUP(sb),
1298 EXT3_INODES_PER_GROUP(sb),
1299 sbi->s_mount_opt);
1301 printk(KERN_INFO "EXT3 FS on %s, ", sb->s_id);
1302 if (EXT3_SB(sb)->s_journal->j_inode == NULL) {
1303 char b[BDEVNAME_SIZE];
1305 printk("external journal on %s\n",
1306 bdevname(EXT3_SB(sb)->s_journal->j_dev, b));
1307 } else {
1308 printk("internal journal\n");
1310 return res;
1313 /* Called at mount-time, super-block is locked */
1314 static int ext3_check_descriptors(struct super_block *sb)
1316 struct ext3_sb_info *sbi = EXT3_SB(sb);
1317 int i;
1319 ext3_debug ("Checking group descriptors");
1321 for (i = 0; i < sbi->s_groups_count; i++) {
1322 struct ext3_group_desc *gdp = ext3_get_group_desc(sb, i, NULL);
1323 ext3_fsblk_t first_block = ext3_group_first_block_no(sb, i);
1324 ext3_fsblk_t last_block;
1326 if (i == sbi->s_groups_count - 1)
1327 last_block = le32_to_cpu(sbi->s_es->s_blocks_count) - 1;
1328 else
1329 last_block = first_block +
1330 (EXT3_BLOCKS_PER_GROUP(sb) - 1);
1332 if (le32_to_cpu(gdp->bg_block_bitmap) < first_block ||
1333 le32_to_cpu(gdp->bg_block_bitmap) > last_block)
1335 ext3_error (sb, "ext3_check_descriptors",
1336 "Block bitmap for group %d"
1337 " not in group (block %lu)!",
1338 i, (unsigned long)
1339 le32_to_cpu(gdp->bg_block_bitmap));
1340 return 0;
1342 if (le32_to_cpu(gdp->bg_inode_bitmap) < first_block ||
1343 le32_to_cpu(gdp->bg_inode_bitmap) > last_block)
1345 ext3_error (sb, "ext3_check_descriptors",
1346 "Inode bitmap for group %d"
1347 " not in group (block %lu)!",
1348 i, (unsigned long)
1349 le32_to_cpu(gdp->bg_inode_bitmap));
1350 return 0;
1352 if (le32_to_cpu(gdp->bg_inode_table) < first_block ||
1353 le32_to_cpu(gdp->bg_inode_table) + sbi->s_itb_per_group - 1 >
1354 last_block)
1356 ext3_error (sb, "ext3_check_descriptors",
1357 "Inode table for group %d"
1358 " not in group (block %lu)!",
1359 i, (unsigned long)
1360 le32_to_cpu(gdp->bg_inode_table));
1361 return 0;
1365 sbi->s_es->s_free_blocks_count=cpu_to_le32(ext3_count_free_blocks(sb));
1366 sbi->s_es->s_free_inodes_count=cpu_to_le32(ext3_count_free_inodes(sb));
1367 return 1;
1371 /* ext3_orphan_cleanup() walks a singly-linked list of inodes (starting at
1372 * the superblock) which were deleted from all directories, but held open by
1373 * a process at the time of a crash. We walk the list and try to delete these
1374 * inodes at recovery time (only with a read-write filesystem).
1376 * In order to keep the orphan inode chain consistent during traversal (in
1377 * case of crash during recovery), we link each inode into the superblock
1378 * orphan list_head and handle it the same way as an inode deletion during
1379 * normal operation (which journals the operations for us).
1381 * We only do an iget() and an iput() on each inode, which is very safe if we
1382 * accidentally point at an in-use or already deleted inode. The worst that
1383 * can happen in this case is that we get a "bit already cleared" message from
1384 * ext3_free_inode(). The only reason we would point at a wrong inode is if
1385 * e2fsck was run on this filesystem, and it must have already done the orphan
1386 * inode cleanup for us, so we can safely abort without any further action.
1388 static void ext3_orphan_cleanup (struct super_block * sb,
1389 struct ext3_super_block * es)
1391 unsigned int s_flags = sb->s_flags;
1392 int nr_orphans = 0, nr_truncates = 0;
1393 #ifdef CONFIG_QUOTA
1394 int i;
1395 #endif
1396 if (!es->s_last_orphan) {
1397 jbd_debug(4, "no orphan inodes to clean up\n");
1398 return;
1401 if (bdev_read_only(sb->s_bdev)) {
1402 printk(KERN_ERR "EXT3-fs: write access "
1403 "unavailable, skipping orphan cleanup.\n");
1404 return;
1407 /* Check if feature set allows readwrite operations */
1408 if (EXT3_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP)) {
1409 printk(KERN_INFO "EXT3-fs: %s: Skipping orphan cleanup due to "
1410 "unknown ROCOMPAT features\n", sb->s_id);
1411 return;
1414 if (EXT3_SB(sb)->s_mount_state & EXT3_ERROR_FS) {
1415 if (es->s_last_orphan)
1416 jbd_debug(1, "Errors on filesystem, "
1417 "clearing orphan list.\n");
1418 es->s_last_orphan = 0;
1419 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1420 return;
1423 if (s_flags & MS_RDONLY) {
1424 printk(KERN_INFO "EXT3-fs: %s: orphan cleanup on readonly fs\n",
1425 sb->s_id);
1426 sb->s_flags &= ~MS_RDONLY;
1428 #ifdef CONFIG_QUOTA
1429 /* Needed for iput() to work correctly and not trash data */
1430 sb->s_flags |= MS_ACTIVE;
1431 /* Turn on quotas so that they are updated correctly */
1432 for (i = 0; i < MAXQUOTAS; i++) {
1433 if (EXT3_SB(sb)->s_qf_names[i]) {
1434 int ret = ext3_quota_on_mount(sb, i);
1435 if (ret < 0)
1436 printk(KERN_ERR
1437 "EXT3-fs: Cannot turn on journaled "
1438 "quota: error %d\n", ret);
1441 #endif
1443 while (es->s_last_orphan) {
1444 struct inode *inode;
1446 inode = ext3_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1447 if (IS_ERR(inode)) {
1448 es->s_last_orphan = 0;
1449 break;
1452 list_add(&EXT3_I(inode)->i_orphan, &EXT3_SB(sb)->s_orphan);
1453 vfs_dq_init(inode);
1454 if (inode->i_nlink) {
1455 printk(KERN_DEBUG
1456 "%s: truncating inode %lu to %Ld bytes\n",
1457 __func__, inode->i_ino, inode->i_size);
1458 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1459 inode->i_ino, inode->i_size);
1460 ext3_truncate(inode);
1461 nr_truncates++;
1462 } else {
1463 printk(KERN_DEBUG
1464 "%s: deleting unreferenced inode %lu\n",
1465 __func__, inode->i_ino);
1466 jbd_debug(2, "deleting unreferenced inode %lu\n",
1467 inode->i_ino);
1468 nr_orphans++;
1470 iput(inode); /* The delete magic happens here! */
1473 #define PLURAL(x) (x), ((x)==1) ? "" : "s"
1475 if (nr_orphans)
1476 printk(KERN_INFO "EXT3-fs: %s: %d orphan inode%s deleted\n",
1477 sb->s_id, PLURAL(nr_orphans));
1478 if (nr_truncates)
1479 printk(KERN_INFO "EXT3-fs: %s: %d truncate%s cleaned up\n",
1480 sb->s_id, PLURAL(nr_truncates));
1481 #ifdef CONFIG_QUOTA
1482 /* Turn quotas off */
1483 for (i = 0; i < MAXQUOTAS; i++) {
1484 if (sb_dqopt(sb)->files[i])
1485 vfs_quota_off(sb, i, 0);
1487 #endif
1488 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1492 * Maximal file size. There is a direct, and {,double-,triple-}indirect
1493 * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
1494 * We need to be 1 filesystem block less than the 2^32 sector limit.
1496 static loff_t ext3_max_size(int bits)
1498 loff_t res = EXT3_NDIR_BLOCKS;
1499 int meta_blocks;
1500 loff_t upper_limit;
1502 /* This is calculated to be the largest file size for a
1503 * dense, file such that the total number of
1504 * sectors in the file, including data and all indirect blocks,
1505 * does not exceed 2^32 -1
1506 * __u32 i_blocks representing the total number of
1507 * 512 bytes blocks of the file
1509 upper_limit = (1LL << 32) - 1;
1511 /* total blocks in file system block size */
1512 upper_limit >>= (bits - 9);
1515 /* indirect blocks */
1516 meta_blocks = 1;
1517 /* double indirect blocks */
1518 meta_blocks += 1 + (1LL << (bits-2));
1519 /* tripple indirect blocks */
1520 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1522 upper_limit -= meta_blocks;
1523 upper_limit <<= bits;
1525 res += 1LL << (bits-2);
1526 res += 1LL << (2*(bits-2));
1527 res += 1LL << (3*(bits-2));
1528 res <<= bits;
1529 if (res > upper_limit)
1530 res = upper_limit;
1532 if (res > MAX_LFS_FILESIZE)
1533 res = MAX_LFS_FILESIZE;
1535 return res;
1538 static ext3_fsblk_t descriptor_loc(struct super_block *sb,
1539 ext3_fsblk_t logic_sb_block,
1540 int nr)
1542 struct ext3_sb_info *sbi = EXT3_SB(sb);
1543 unsigned long bg, first_meta_bg;
1544 int has_super = 0;
1546 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1548 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_META_BG) ||
1549 nr < first_meta_bg)
1550 return (logic_sb_block + nr + 1);
1551 bg = sbi->s_desc_per_block * nr;
1552 if (ext3_bg_has_super(sb, bg))
1553 has_super = 1;
1554 return (has_super + ext3_group_first_block_no(sb, bg));
1558 static int ext3_fill_super (struct super_block *sb, void *data, int silent)
1560 struct buffer_head * bh;
1561 struct ext3_super_block *es = NULL;
1562 struct ext3_sb_info *sbi;
1563 ext3_fsblk_t block;
1564 ext3_fsblk_t sb_block = get_sb_block(&data);
1565 ext3_fsblk_t logic_sb_block;
1566 unsigned long offset = 0;
1567 unsigned int journal_inum = 0;
1568 unsigned long journal_devnum = 0;
1569 unsigned long def_mount_opts;
1570 struct inode *root;
1571 int blocksize;
1572 int hblock;
1573 int db_count;
1574 int i;
1575 int needs_recovery;
1576 int ret = -EINVAL;
1577 __le32 features;
1578 int err;
1580 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1581 if (!sbi)
1582 return -ENOMEM;
1584 sbi->s_blockgroup_lock =
1585 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
1586 if (!sbi->s_blockgroup_lock) {
1587 kfree(sbi);
1588 return -ENOMEM;
1590 sb->s_fs_info = sbi;
1591 sbi->s_mount_opt = 0;
1592 sbi->s_resuid = EXT3_DEF_RESUID;
1593 sbi->s_resgid = EXT3_DEF_RESGID;
1594 sbi->s_sb_block = sb_block;
1596 unlock_kernel();
1598 blocksize = sb_min_blocksize(sb, EXT3_MIN_BLOCK_SIZE);
1599 if (!blocksize) {
1600 printk(KERN_ERR "EXT3-fs: unable to set blocksize\n");
1601 goto out_fail;
1605 * The ext3 superblock will not be buffer aligned for other than 1kB
1606 * block sizes. We need to calculate the offset from buffer start.
1608 if (blocksize != EXT3_MIN_BLOCK_SIZE) {
1609 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1610 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1611 } else {
1612 logic_sb_block = sb_block;
1615 if (!(bh = sb_bread(sb, logic_sb_block))) {
1616 printk (KERN_ERR "EXT3-fs: unable to read superblock\n");
1617 goto out_fail;
1620 * Note: s_es must be initialized as soon as possible because
1621 * some ext3 macro-instructions depend on its value
1623 es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
1624 sbi->s_es = es;
1625 sb->s_magic = le16_to_cpu(es->s_magic);
1626 if (sb->s_magic != EXT3_SUPER_MAGIC)
1627 goto cantfind_ext3;
1629 /* Set defaults before we parse the mount options */
1630 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1631 if (def_mount_opts & EXT3_DEFM_DEBUG)
1632 set_opt(sbi->s_mount_opt, DEBUG);
1633 if (def_mount_opts & EXT3_DEFM_BSDGROUPS)
1634 set_opt(sbi->s_mount_opt, GRPID);
1635 if (def_mount_opts & EXT3_DEFM_UID16)
1636 set_opt(sbi->s_mount_opt, NO_UID32);
1637 #ifdef CONFIG_EXT3_FS_XATTR
1638 if (def_mount_opts & EXT3_DEFM_XATTR_USER)
1639 set_opt(sbi->s_mount_opt, XATTR_USER);
1640 #endif
1641 #ifdef CONFIG_EXT3_FS_POSIX_ACL
1642 if (def_mount_opts & EXT3_DEFM_ACL)
1643 set_opt(sbi->s_mount_opt, POSIX_ACL);
1644 #endif
1645 if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_DATA)
1646 sbi->s_mount_opt |= EXT3_MOUNT_JOURNAL_DATA;
1647 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_ORDERED)
1648 sbi->s_mount_opt |= EXT3_MOUNT_ORDERED_DATA;
1649 else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_WBACK)
1650 sbi->s_mount_opt |= EXT3_MOUNT_WRITEBACK_DATA;
1652 if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_PANIC)
1653 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1654 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_CONTINUE)
1655 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1656 else
1657 set_opt(sbi->s_mount_opt, ERRORS_RO);
1659 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1660 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1662 set_opt(sbi->s_mount_opt, RESERVATION);
1664 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1665 NULL, 0))
1666 goto failed_mount;
1668 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1669 ((sbi->s_mount_opt & EXT3_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1671 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV &&
1672 (EXT3_HAS_COMPAT_FEATURE(sb, ~0U) ||
1673 EXT3_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1674 EXT3_HAS_INCOMPAT_FEATURE(sb, ~0U)))
1675 printk(KERN_WARNING
1676 "EXT3-fs warning: feature flags set on rev 0 fs, "
1677 "running e2fsck is recommended\n");
1679 * Check feature flags regardless of the revision level, since we
1680 * previously didn't change the revision level when setting the flags,
1681 * so there is a chance incompat flags are set on a rev 0 filesystem.
1683 features = EXT3_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP);
1684 if (features) {
1685 printk(KERN_ERR "EXT3-fs: %s: couldn't mount because of "
1686 "unsupported optional features (%x).\n",
1687 sb->s_id, le32_to_cpu(features));
1688 goto failed_mount;
1690 features = EXT3_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP);
1691 if (!(sb->s_flags & MS_RDONLY) && features) {
1692 printk(KERN_ERR "EXT3-fs: %s: couldn't mount RDWR because of "
1693 "unsupported optional features (%x).\n",
1694 sb->s_id, le32_to_cpu(features));
1695 goto failed_mount;
1697 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
1699 if (blocksize < EXT3_MIN_BLOCK_SIZE ||
1700 blocksize > EXT3_MAX_BLOCK_SIZE) {
1701 printk(KERN_ERR
1702 "EXT3-fs: Unsupported filesystem blocksize %d on %s.\n",
1703 blocksize, sb->s_id);
1704 goto failed_mount;
1707 hblock = bdev_logical_block_size(sb->s_bdev);
1708 if (sb->s_blocksize != blocksize) {
1710 * Make sure the blocksize for the filesystem is larger
1711 * than the hardware sectorsize for the machine.
1713 if (blocksize < hblock) {
1714 printk(KERN_ERR "EXT3-fs: blocksize %d too small for "
1715 "device blocksize %d.\n", blocksize, hblock);
1716 goto failed_mount;
1719 brelse (bh);
1720 if (!sb_set_blocksize(sb, blocksize)) {
1721 printk(KERN_ERR "EXT3-fs: bad blocksize %d.\n",
1722 blocksize);
1723 goto out_fail;
1725 logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize;
1726 offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize;
1727 bh = sb_bread(sb, logic_sb_block);
1728 if (!bh) {
1729 printk(KERN_ERR
1730 "EXT3-fs: Can't read superblock on 2nd try.\n");
1731 goto failed_mount;
1733 es = (struct ext3_super_block *)(((char *)bh->b_data) + offset);
1734 sbi->s_es = es;
1735 if (es->s_magic != cpu_to_le16(EXT3_SUPER_MAGIC)) {
1736 printk (KERN_ERR
1737 "EXT3-fs: Magic mismatch, very weird !\n");
1738 goto failed_mount;
1742 sb->s_maxbytes = ext3_max_size(sb->s_blocksize_bits);
1744 if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV) {
1745 sbi->s_inode_size = EXT3_GOOD_OLD_INODE_SIZE;
1746 sbi->s_first_ino = EXT3_GOOD_OLD_FIRST_INO;
1747 } else {
1748 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1749 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
1750 if ((sbi->s_inode_size < EXT3_GOOD_OLD_INODE_SIZE) ||
1751 (!is_power_of_2(sbi->s_inode_size)) ||
1752 (sbi->s_inode_size > blocksize)) {
1753 printk (KERN_ERR
1754 "EXT3-fs: unsupported inode size: %d\n",
1755 sbi->s_inode_size);
1756 goto failed_mount;
1759 sbi->s_frag_size = EXT3_MIN_FRAG_SIZE <<
1760 le32_to_cpu(es->s_log_frag_size);
1761 if (blocksize != sbi->s_frag_size) {
1762 printk(KERN_ERR
1763 "EXT3-fs: fragsize %lu != blocksize %u (unsupported)\n",
1764 sbi->s_frag_size, blocksize);
1765 goto failed_mount;
1767 sbi->s_frags_per_block = 1;
1768 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1769 sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
1770 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
1771 if (EXT3_INODE_SIZE(sb) == 0 || EXT3_INODES_PER_GROUP(sb) == 0)
1772 goto cantfind_ext3;
1773 sbi->s_inodes_per_block = blocksize / EXT3_INODE_SIZE(sb);
1774 if (sbi->s_inodes_per_block == 0)
1775 goto cantfind_ext3;
1776 sbi->s_itb_per_group = sbi->s_inodes_per_group /
1777 sbi->s_inodes_per_block;
1778 sbi->s_desc_per_block = blocksize / sizeof(struct ext3_group_desc);
1779 sbi->s_sbh = bh;
1780 sbi->s_mount_state = le16_to_cpu(es->s_state);
1781 sbi->s_addr_per_block_bits = ilog2(EXT3_ADDR_PER_BLOCK(sb));
1782 sbi->s_desc_per_block_bits = ilog2(EXT3_DESC_PER_BLOCK(sb));
1783 for (i=0; i < 4; i++)
1784 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
1785 sbi->s_def_hash_version = es->s_def_hash_version;
1786 i = le32_to_cpu(es->s_flags);
1787 if (i & EXT2_FLAGS_UNSIGNED_HASH)
1788 sbi->s_hash_unsigned = 3;
1789 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
1790 #ifdef __CHAR_UNSIGNED__
1791 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
1792 sbi->s_hash_unsigned = 3;
1793 #else
1794 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
1795 #endif
1798 if (sbi->s_blocks_per_group > blocksize * 8) {
1799 printk (KERN_ERR
1800 "EXT3-fs: #blocks per group too big: %lu\n",
1801 sbi->s_blocks_per_group);
1802 goto failed_mount;
1804 if (sbi->s_frags_per_group > blocksize * 8) {
1805 printk (KERN_ERR
1806 "EXT3-fs: #fragments per group too big: %lu\n",
1807 sbi->s_frags_per_group);
1808 goto failed_mount;
1810 if (sbi->s_inodes_per_group > blocksize * 8) {
1811 printk (KERN_ERR
1812 "EXT3-fs: #inodes per group too big: %lu\n",
1813 sbi->s_inodes_per_group);
1814 goto failed_mount;
1817 if (le32_to_cpu(es->s_blocks_count) >
1818 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
1819 printk(KERN_ERR "EXT3-fs: filesystem on %s:"
1820 " too large to mount safely\n", sb->s_id);
1821 if (sizeof(sector_t) < 8)
1822 printk(KERN_WARNING "EXT3-fs: CONFIG_LBDAF not "
1823 "enabled\n");
1824 goto failed_mount;
1827 if (EXT3_BLOCKS_PER_GROUP(sb) == 0)
1828 goto cantfind_ext3;
1829 sbi->s_groups_count = ((le32_to_cpu(es->s_blocks_count) -
1830 le32_to_cpu(es->s_first_data_block) - 1)
1831 / EXT3_BLOCKS_PER_GROUP(sb)) + 1;
1832 db_count = (sbi->s_groups_count + EXT3_DESC_PER_BLOCK(sb) - 1) /
1833 EXT3_DESC_PER_BLOCK(sb);
1834 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
1835 GFP_KERNEL);
1836 if (sbi->s_group_desc == NULL) {
1837 printk (KERN_ERR "EXT3-fs: not enough memory\n");
1838 goto failed_mount;
1841 bgl_lock_init(sbi->s_blockgroup_lock);
1843 for (i = 0; i < db_count; i++) {
1844 block = descriptor_loc(sb, logic_sb_block, i);
1845 sbi->s_group_desc[i] = sb_bread(sb, block);
1846 if (!sbi->s_group_desc[i]) {
1847 printk (KERN_ERR "EXT3-fs: "
1848 "can't read group descriptor %d\n", i);
1849 db_count = i;
1850 goto failed_mount2;
1853 if (!ext3_check_descriptors (sb)) {
1854 printk(KERN_ERR "EXT3-fs: group descriptors corrupted!\n");
1855 goto failed_mount2;
1857 sbi->s_gdb_count = db_count;
1858 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1859 spin_lock_init(&sbi->s_next_gen_lock);
1861 err = percpu_counter_init(&sbi->s_freeblocks_counter,
1862 ext3_count_free_blocks(sb));
1863 if (!err) {
1864 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1865 ext3_count_free_inodes(sb));
1867 if (!err) {
1868 err = percpu_counter_init(&sbi->s_dirs_counter,
1869 ext3_count_dirs(sb));
1871 if (err) {
1872 printk(KERN_ERR "EXT3-fs: insufficient memory\n");
1873 goto failed_mount3;
1876 /* per fileystem reservation list head & lock */
1877 spin_lock_init(&sbi->s_rsv_window_lock);
1878 sbi->s_rsv_window_root = RB_ROOT;
1879 /* Add a single, static dummy reservation to the start of the
1880 * reservation window list --- it gives us a placeholder for
1881 * append-at-start-of-list which makes the allocation logic
1882 * _much_ simpler. */
1883 sbi->s_rsv_window_head.rsv_start = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1884 sbi->s_rsv_window_head.rsv_end = EXT3_RESERVE_WINDOW_NOT_ALLOCATED;
1885 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1886 sbi->s_rsv_window_head.rsv_goal_size = 0;
1887 ext3_rsv_window_add(sb, &sbi->s_rsv_window_head);
1890 * set up enough so that it can read an inode
1892 sb->s_op = &ext3_sops;
1893 sb->s_export_op = &ext3_export_ops;
1894 sb->s_xattr = ext3_xattr_handlers;
1895 #ifdef CONFIG_QUOTA
1896 sb->s_qcop = &ext3_qctl_operations;
1897 sb->dq_op = &ext3_quota_operations;
1898 #endif
1899 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
1901 sb->s_root = NULL;
1903 needs_recovery = (es->s_last_orphan != 0 ||
1904 EXT3_HAS_INCOMPAT_FEATURE(sb,
1905 EXT3_FEATURE_INCOMPAT_RECOVER));
1908 * The first inode we look at is the journal inode. Don't try
1909 * root first: it may be modified in the journal!
1911 if (!test_opt(sb, NOLOAD) &&
1912 EXT3_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL)) {
1913 if (ext3_load_journal(sb, es, journal_devnum))
1914 goto failed_mount3;
1915 } else if (journal_inum) {
1916 if (ext3_create_journal(sb, es, journal_inum))
1917 goto failed_mount3;
1918 } else {
1919 if (!silent)
1920 printk (KERN_ERR
1921 "ext3: No journal on filesystem on %s\n",
1922 sb->s_id);
1923 goto failed_mount3;
1926 /* We have now updated the journal if required, so we can
1927 * validate the data journaling mode. */
1928 switch (test_opt(sb, DATA_FLAGS)) {
1929 case 0:
1930 /* No mode set, assume a default based on the journal
1931 capabilities: ORDERED_DATA if the journal can
1932 cope, else JOURNAL_DATA */
1933 if (journal_check_available_features
1934 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE))
1935 set_opt(sbi->s_mount_opt, DEFAULT_DATA_MODE);
1936 else
1937 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
1938 break;
1940 case EXT3_MOUNT_ORDERED_DATA:
1941 case EXT3_MOUNT_WRITEBACK_DATA:
1942 if (!journal_check_available_features
1943 (sbi->s_journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE)) {
1944 printk(KERN_ERR "EXT3-fs: Journal does not support "
1945 "requested data journaling mode\n");
1946 goto failed_mount4;
1948 default:
1949 break;
1952 if (test_opt(sb, NOBH)) {
1953 if (!(test_opt(sb, DATA_FLAGS) == EXT3_MOUNT_WRITEBACK_DATA)) {
1954 printk(KERN_WARNING "EXT3-fs: Ignoring nobh option - "
1955 "its supported only with writeback mode\n");
1956 clear_opt(sbi->s_mount_opt, NOBH);
1960 * The journal_load will have done any necessary log recovery,
1961 * so we can safely mount the rest of the filesystem now.
1964 root = ext3_iget(sb, EXT3_ROOT_INO);
1965 if (IS_ERR(root)) {
1966 printk(KERN_ERR "EXT3-fs: get root inode failed\n");
1967 ret = PTR_ERR(root);
1968 goto failed_mount4;
1970 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1971 iput(root);
1972 printk(KERN_ERR "EXT3-fs: corrupt root inode, run e2fsck\n");
1973 goto failed_mount4;
1975 sb->s_root = d_alloc_root(root);
1976 if (!sb->s_root) {
1977 printk(KERN_ERR "EXT3-fs: get root dentry failed\n");
1978 iput(root);
1979 ret = -ENOMEM;
1980 goto failed_mount4;
1983 ext3_setup_super (sb, es, sb->s_flags & MS_RDONLY);
1985 * akpm: core read_super() calls in here with the superblock locked.
1986 * That deadlocks, because orphan cleanup needs to lock the superblock
1987 * in numerous places. Here we just pop the lock - it's relatively
1988 * harmless, because we are now ready to accept write_super() requests,
1989 * and aviro says that's the only reason for hanging onto the
1990 * superblock lock.
1992 EXT3_SB(sb)->s_mount_state |= EXT3_ORPHAN_FS;
1993 ext3_orphan_cleanup(sb, es);
1994 EXT3_SB(sb)->s_mount_state &= ~EXT3_ORPHAN_FS;
1995 if (needs_recovery)
1996 printk (KERN_INFO "EXT3-fs: recovery complete.\n");
1997 ext3_mark_recovery_complete(sb, es);
1998 printk (KERN_INFO "EXT3-fs: mounted filesystem with %s data mode.\n",
1999 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA ? "journal":
2000 test_opt(sb,DATA_FLAGS) == EXT3_MOUNT_ORDERED_DATA ? "ordered":
2001 "writeback");
2003 lock_kernel();
2004 return 0;
2006 cantfind_ext3:
2007 if (!silent)
2008 printk(KERN_ERR "VFS: Can't find ext3 filesystem on dev %s.\n",
2009 sb->s_id);
2010 goto failed_mount;
2012 failed_mount4:
2013 journal_destroy(sbi->s_journal);
2014 failed_mount3:
2015 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2016 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2017 percpu_counter_destroy(&sbi->s_dirs_counter);
2018 failed_mount2:
2019 for (i = 0; i < db_count; i++)
2020 brelse(sbi->s_group_desc[i]);
2021 kfree(sbi->s_group_desc);
2022 failed_mount:
2023 #ifdef CONFIG_QUOTA
2024 for (i = 0; i < MAXQUOTAS; i++)
2025 kfree(sbi->s_qf_names[i]);
2026 #endif
2027 ext3_blkdev_remove(sbi);
2028 brelse(bh);
2029 out_fail:
2030 sb->s_fs_info = NULL;
2031 kfree(sbi->s_blockgroup_lock);
2032 kfree(sbi);
2033 lock_kernel();
2034 return ret;
2038 * Setup any per-fs journal parameters now. We'll do this both on
2039 * initial mount, once the journal has been initialised but before we've
2040 * done any recovery; and again on any subsequent remount.
2042 static void ext3_init_journal_params(struct super_block *sb, journal_t *journal)
2044 struct ext3_sb_info *sbi = EXT3_SB(sb);
2046 if (sbi->s_commit_interval)
2047 journal->j_commit_interval = sbi->s_commit_interval;
2048 /* We could also set up an ext3-specific default for the commit
2049 * interval here, but for now we'll just fall back to the jbd
2050 * default. */
2052 spin_lock(&journal->j_state_lock);
2053 if (test_opt(sb, BARRIER))
2054 journal->j_flags |= JFS_BARRIER;
2055 else
2056 journal->j_flags &= ~JFS_BARRIER;
2057 if (test_opt(sb, DATA_ERR_ABORT))
2058 journal->j_flags |= JFS_ABORT_ON_SYNCDATA_ERR;
2059 else
2060 journal->j_flags &= ~JFS_ABORT_ON_SYNCDATA_ERR;
2061 spin_unlock(&journal->j_state_lock);
2064 static journal_t *ext3_get_journal(struct super_block *sb,
2065 unsigned int journal_inum)
2067 struct inode *journal_inode;
2068 journal_t *journal;
2070 /* First, test for the existence of a valid inode on disk. Bad
2071 * things happen if we iget() an unused inode, as the subsequent
2072 * iput() will try to delete it. */
2074 journal_inode = ext3_iget(sb, journal_inum);
2075 if (IS_ERR(journal_inode)) {
2076 printk(KERN_ERR "EXT3-fs: no journal found.\n");
2077 return NULL;
2079 if (!journal_inode->i_nlink) {
2080 make_bad_inode(journal_inode);
2081 iput(journal_inode);
2082 printk(KERN_ERR "EXT3-fs: journal inode is deleted.\n");
2083 return NULL;
2086 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
2087 journal_inode, journal_inode->i_size);
2088 if (!S_ISREG(journal_inode->i_mode)) {
2089 printk(KERN_ERR "EXT3-fs: invalid journal inode.\n");
2090 iput(journal_inode);
2091 return NULL;
2094 journal = journal_init_inode(journal_inode);
2095 if (!journal) {
2096 printk(KERN_ERR "EXT3-fs: Could not load journal inode\n");
2097 iput(journal_inode);
2098 return NULL;
2100 journal->j_private = sb;
2101 ext3_init_journal_params(sb, journal);
2102 return journal;
2105 static journal_t *ext3_get_dev_journal(struct super_block *sb,
2106 dev_t j_dev)
2108 struct buffer_head * bh;
2109 journal_t *journal;
2110 ext3_fsblk_t start;
2111 ext3_fsblk_t len;
2112 int hblock, blocksize;
2113 ext3_fsblk_t sb_block;
2114 unsigned long offset;
2115 struct ext3_super_block * es;
2116 struct block_device *bdev;
2118 bdev = ext3_blkdev_get(j_dev);
2119 if (bdev == NULL)
2120 return NULL;
2122 if (bd_claim(bdev, sb)) {
2123 printk(KERN_ERR
2124 "EXT3: failed to claim external journal device.\n");
2125 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
2126 return NULL;
2129 blocksize = sb->s_blocksize;
2130 hblock = bdev_logical_block_size(bdev);
2131 if (blocksize < hblock) {
2132 printk(KERN_ERR
2133 "EXT3-fs: blocksize too small for journal device.\n");
2134 goto out_bdev;
2137 sb_block = EXT3_MIN_BLOCK_SIZE / blocksize;
2138 offset = EXT3_MIN_BLOCK_SIZE % blocksize;
2139 set_blocksize(bdev, blocksize);
2140 if (!(bh = __bread(bdev, sb_block, blocksize))) {
2141 printk(KERN_ERR "EXT3-fs: couldn't read superblock of "
2142 "external journal\n");
2143 goto out_bdev;
2146 es = (struct ext3_super_block *) (((char *)bh->b_data) + offset);
2147 if ((le16_to_cpu(es->s_magic) != EXT3_SUPER_MAGIC) ||
2148 !(le32_to_cpu(es->s_feature_incompat) &
2149 EXT3_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2150 printk(KERN_ERR "EXT3-fs: external journal has "
2151 "bad superblock\n");
2152 brelse(bh);
2153 goto out_bdev;
2156 if (memcmp(EXT3_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2157 printk(KERN_ERR "EXT3-fs: journal UUID does not match\n");
2158 brelse(bh);
2159 goto out_bdev;
2162 len = le32_to_cpu(es->s_blocks_count);
2163 start = sb_block + 1;
2164 brelse(bh); /* we're done with the superblock */
2166 journal = journal_init_dev(bdev, sb->s_bdev,
2167 start, len, blocksize);
2168 if (!journal) {
2169 printk(KERN_ERR "EXT3-fs: failed to create device journal\n");
2170 goto out_bdev;
2172 journal->j_private = sb;
2173 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2174 wait_on_buffer(journal->j_sb_buffer);
2175 if (!buffer_uptodate(journal->j_sb_buffer)) {
2176 printk(KERN_ERR "EXT3-fs: I/O error on journal device\n");
2177 goto out_journal;
2179 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
2180 printk(KERN_ERR "EXT3-fs: External journal has more than one "
2181 "user (unsupported) - %d\n",
2182 be32_to_cpu(journal->j_superblock->s_nr_users));
2183 goto out_journal;
2185 EXT3_SB(sb)->journal_bdev = bdev;
2186 ext3_init_journal_params(sb, journal);
2187 return journal;
2188 out_journal:
2189 journal_destroy(journal);
2190 out_bdev:
2191 ext3_blkdev_put(bdev);
2192 return NULL;
2195 static int ext3_load_journal(struct super_block *sb,
2196 struct ext3_super_block *es,
2197 unsigned long journal_devnum)
2199 journal_t *journal;
2200 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2201 dev_t journal_dev;
2202 int err = 0;
2203 int really_read_only;
2205 if (journal_devnum &&
2206 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2207 printk(KERN_INFO "EXT3-fs: external journal device major/minor "
2208 "numbers have changed\n");
2209 journal_dev = new_decode_dev(journal_devnum);
2210 } else
2211 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2213 really_read_only = bdev_read_only(sb->s_bdev);
2216 * Are we loading a blank journal or performing recovery after a
2217 * crash? For recovery, we need to check in advance whether we
2218 * can get read-write access to the device.
2221 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER)) {
2222 if (sb->s_flags & MS_RDONLY) {
2223 printk(KERN_INFO "EXT3-fs: INFO: recovery "
2224 "required on readonly filesystem.\n");
2225 if (really_read_only) {
2226 printk(KERN_ERR "EXT3-fs: write access "
2227 "unavailable, cannot proceed.\n");
2228 return -EROFS;
2230 printk (KERN_INFO "EXT3-fs: write access will "
2231 "be enabled during recovery.\n");
2235 if (journal_inum && journal_dev) {
2236 printk(KERN_ERR "EXT3-fs: filesystem has both journal "
2237 "and inode journals!\n");
2238 return -EINVAL;
2241 if (journal_inum) {
2242 if (!(journal = ext3_get_journal(sb, journal_inum)))
2243 return -EINVAL;
2244 } else {
2245 if (!(journal = ext3_get_dev_journal(sb, journal_dev)))
2246 return -EINVAL;
2249 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2250 err = journal_update_format(journal);
2251 if (err) {
2252 printk(KERN_ERR "EXT3-fs: error updating journal.\n");
2253 journal_destroy(journal);
2254 return err;
2258 if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER))
2259 err = journal_wipe(journal, !really_read_only);
2260 if (!err)
2261 err = journal_load(journal);
2263 if (err) {
2264 printk(KERN_ERR "EXT3-fs: error loading journal.\n");
2265 journal_destroy(journal);
2266 return err;
2269 EXT3_SB(sb)->s_journal = journal;
2270 ext3_clear_journal_err(sb, es);
2272 if (journal_devnum &&
2273 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2274 es->s_journal_dev = cpu_to_le32(journal_devnum);
2276 /* Make sure we flush the recovery flag to disk. */
2277 ext3_commit_super(sb, es, 1);
2280 return 0;
2283 static int ext3_create_journal(struct super_block * sb,
2284 struct ext3_super_block * es,
2285 unsigned int journal_inum)
2287 journal_t *journal;
2288 int err;
2290 if (sb->s_flags & MS_RDONLY) {
2291 printk(KERN_ERR "EXT3-fs: readonly filesystem when trying to "
2292 "create journal.\n");
2293 return -EROFS;
2296 journal = ext3_get_journal(sb, journal_inum);
2297 if (!journal)
2298 return -EINVAL;
2300 printk(KERN_INFO "EXT3-fs: creating new journal on inode %u\n",
2301 journal_inum);
2303 err = journal_create(journal);
2304 if (err) {
2305 printk(KERN_ERR "EXT3-fs: error creating journal.\n");
2306 journal_destroy(journal);
2307 return -EIO;
2310 EXT3_SB(sb)->s_journal = journal;
2312 ext3_update_dynamic_rev(sb);
2313 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2314 EXT3_SET_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_HAS_JOURNAL);
2316 es->s_journal_inum = cpu_to_le32(journal_inum);
2318 /* Make sure we flush the recovery flag to disk. */
2319 ext3_commit_super(sb, es, 1);
2321 return 0;
2324 static int ext3_commit_super(struct super_block *sb,
2325 struct ext3_super_block *es,
2326 int sync)
2328 struct buffer_head *sbh = EXT3_SB(sb)->s_sbh;
2329 int error = 0;
2331 if (!sbh)
2332 return error;
2334 * If the file system is mounted read-only, don't update the
2335 * superblock write time. This avoids updating the superblock
2336 * write time when we are mounting the root file system
2337 * read/only but we need to replay the journal; at that point,
2338 * for people who are east of GMT and who make their clock
2339 * tick in localtime for Windows bug-for-bug compatibility,
2340 * the clock is set in the future, and this will cause e2fsck
2341 * to complain and force a full file system check.
2343 if (!(sb->s_flags & MS_RDONLY))
2344 es->s_wtime = cpu_to_le32(get_seconds());
2345 es->s_free_blocks_count = cpu_to_le32(ext3_count_free_blocks(sb));
2346 es->s_free_inodes_count = cpu_to_le32(ext3_count_free_inodes(sb));
2347 BUFFER_TRACE(sbh, "marking dirty");
2348 mark_buffer_dirty(sbh);
2349 if (sync)
2350 error = sync_dirty_buffer(sbh);
2351 return error;
2356 * Have we just finished recovery? If so, and if we are mounting (or
2357 * remounting) the filesystem readonly, then we will end up with a
2358 * consistent fs on disk. Record that fact.
2360 static void ext3_mark_recovery_complete(struct super_block * sb,
2361 struct ext3_super_block * es)
2363 journal_t *journal = EXT3_SB(sb)->s_journal;
2365 journal_lock_updates(journal);
2366 if (journal_flush(journal) < 0)
2367 goto out;
2369 lock_super(sb);
2370 if (EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER) &&
2371 sb->s_flags & MS_RDONLY) {
2372 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2373 ext3_commit_super(sb, es, 1);
2375 unlock_super(sb);
2377 out:
2378 journal_unlock_updates(journal);
2382 * If we are mounting (or read-write remounting) a filesystem whose journal
2383 * has recorded an error from a previous lifetime, move that error to the
2384 * main filesystem now.
2386 static void ext3_clear_journal_err(struct super_block * sb,
2387 struct ext3_super_block * es)
2389 journal_t *journal;
2390 int j_errno;
2391 const char *errstr;
2393 journal = EXT3_SB(sb)->s_journal;
2396 * Now check for any error status which may have been recorded in the
2397 * journal by a prior ext3_error() or ext3_abort()
2400 j_errno = journal_errno(journal);
2401 if (j_errno) {
2402 char nbuf[16];
2404 errstr = ext3_decode_error(sb, j_errno, nbuf);
2405 ext3_warning(sb, __func__, "Filesystem error recorded "
2406 "from previous mount: %s", errstr);
2407 ext3_warning(sb, __func__, "Marking fs in need of "
2408 "filesystem check.");
2410 EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS;
2411 es->s_state |= cpu_to_le16(EXT3_ERROR_FS);
2412 ext3_commit_super (sb, es, 1);
2414 journal_clear_err(journal);
2419 * Force the running and committing transactions to commit,
2420 * and wait on the commit.
2422 int ext3_force_commit(struct super_block *sb)
2424 journal_t *journal;
2425 int ret;
2427 if (sb->s_flags & MS_RDONLY)
2428 return 0;
2430 journal = EXT3_SB(sb)->s_journal;
2431 ret = ext3_journal_force_commit(journal);
2432 return ret;
2435 static int ext3_sync_fs(struct super_block *sb, int wait)
2437 tid_t target;
2439 if (journal_start_commit(EXT3_SB(sb)->s_journal, &target)) {
2440 if (wait)
2441 log_wait_commit(EXT3_SB(sb)->s_journal, target);
2443 return 0;
2447 * LVM calls this function before a (read-only) snapshot is created. This
2448 * gives us a chance to flush the journal completely and mark the fs clean.
2450 static int ext3_freeze(struct super_block *sb)
2452 int error = 0;
2453 journal_t *journal;
2455 if (!(sb->s_flags & MS_RDONLY)) {
2456 journal = EXT3_SB(sb)->s_journal;
2458 /* Now we set up the journal barrier. */
2459 journal_lock_updates(journal);
2462 * We don't want to clear needs_recovery flag when we failed
2463 * to flush the journal.
2465 error = journal_flush(journal);
2466 if (error < 0)
2467 goto out;
2469 /* Journal blocked and flushed, clear needs_recovery flag. */
2470 EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2471 error = ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2472 if (error)
2473 goto out;
2475 return 0;
2477 out:
2478 journal_unlock_updates(journal);
2479 return error;
2483 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2484 * flag here, even though the filesystem is not technically dirty yet.
2486 static int ext3_unfreeze(struct super_block *sb)
2488 if (!(sb->s_flags & MS_RDONLY)) {
2489 lock_super(sb);
2490 /* Reser the needs_recovery flag before the fs is unlocked. */
2491 EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER);
2492 ext3_commit_super(sb, EXT3_SB(sb)->s_es, 1);
2493 unlock_super(sb);
2494 journal_unlock_updates(EXT3_SB(sb)->s_journal);
2496 return 0;
2499 static int ext3_remount (struct super_block * sb, int * flags, char * data)
2501 struct ext3_super_block * es;
2502 struct ext3_sb_info *sbi = EXT3_SB(sb);
2503 ext3_fsblk_t n_blocks_count = 0;
2504 unsigned long old_sb_flags;
2505 struct ext3_mount_options old_opts;
2506 int err;
2507 #ifdef CONFIG_QUOTA
2508 int i;
2509 #endif
2511 lock_kernel();
2513 /* Store the original options */
2514 lock_super(sb);
2515 old_sb_flags = sb->s_flags;
2516 old_opts.s_mount_opt = sbi->s_mount_opt;
2517 old_opts.s_resuid = sbi->s_resuid;
2518 old_opts.s_resgid = sbi->s_resgid;
2519 old_opts.s_commit_interval = sbi->s_commit_interval;
2520 #ifdef CONFIG_QUOTA
2521 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2522 for (i = 0; i < MAXQUOTAS; i++)
2523 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2524 #endif
2527 * Allow the "check" option to be passed as a remount option.
2529 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
2530 err = -EINVAL;
2531 goto restore_opts;
2534 if (sbi->s_mount_opt & EXT3_MOUNT_ABORT)
2535 ext3_abort(sb, __func__, "Abort forced by user");
2537 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2538 ((sbi->s_mount_opt & EXT3_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2540 es = sbi->s_es;
2542 ext3_init_journal_params(sb, sbi->s_journal);
2544 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
2545 n_blocks_count > le32_to_cpu(es->s_blocks_count)) {
2546 if (sbi->s_mount_opt & EXT3_MOUNT_ABORT) {
2547 err = -EROFS;
2548 goto restore_opts;
2551 if (*flags & MS_RDONLY) {
2553 * First of all, the unconditional stuff we have to do
2554 * to disable replay of the journal when we next remount
2556 sb->s_flags |= MS_RDONLY;
2559 * OK, test if we are remounting a valid rw partition
2560 * readonly, and if so set the rdonly flag and then
2561 * mark the partition as valid again.
2563 if (!(es->s_state & cpu_to_le16(EXT3_VALID_FS)) &&
2564 (sbi->s_mount_state & EXT3_VALID_FS))
2565 es->s_state = cpu_to_le16(sbi->s_mount_state);
2568 * We have to unlock super so that we can wait for
2569 * transactions.
2571 unlock_super(sb);
2572 ext3_mark_recovery_complete(sb, es);
2573 lock_super(sb);
2574 } else {
2575 __le32 ret;
2576 if ((ret = EXT3_HAS_RO_COMPAT_FEATURE(sb,
2577 ~EXT3_FEATURE_RO_COMPAT_SUPP))) {
2578 printk(KERN_WARNING "EXT3-fs: %s: couldn't "
2579 "remount RDWR because of unsupported "
2580 "optional features (%x).\n",
2581 sb->s_id, le32_to_cpu(ret));
2582 err = -EROFS;
2583 goto restore_opts;
2587 * If we have an unprocessed orphan list hanging
2588 * around from a previously readonly bdev mount,
2589 * require a full umount/remount for now.
2591 if (es->s_last_orphan) {
2592 printk(KERN_WARNING "EXT3-fs: %s: couldn't "
2593 "remount RDWR because of unprocessed "
2594 "orphan inode list. Please "
2595 "umount/remount instead.\n",
2596 sb->s_id);
2597 err = -EINVAL;
2598 goto restore_opts;
2602 * Mounting a RDONLY partition read-write, so reread
2603 * and store the current valid flag. (It may have
2604 * been changed by e2fsck since we originally mounted
2605 * the partition.)
2607 ext3_clear_journal_err(sb, es);
2608 sbi->s_mount_state = le16_to_cpu(es->s_state);
2609 if ((err = ext3_group_extend(sb, es, n_blocks_count)))
2610 goto restore_opts;
2611 if (!ext3_setup_super (sb, es, 0))
2612 sb->s_flags &= ~MS_RDONLY;
2615 #ifdef CONFIG_QUOTA
2616 /* Release old quota file names */
2617 for (i = 0; i < MAXQUOTAS; i++)
2618 if (old_opts.s_qf_names[i] &&
2619 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2620 kfree(old_opts.s_qf_names[i]);
2621 #endif
2622 unlock_super(sb);
2623 unlock_kernel();
2624 return 0;
2625 restore_opts:
2626 sb->s_flags = old_sb_flags;
2627 sbi->s_mount_opt = old_opts.s_mount_opt;
2628 sbi->s_resuid = old_opts.s_resuid;
2629 sbi->s_resgid = old_opts.s_resgid;
2630 sbi->s_commit_interval = old_opts.s_commit_interval;
2631 #ifdef CONFIG_QUOTA
2632 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
2633 for (i = 0; i < MAXQUOTAS; i++) {
2634 if (sbi->s_qf_names[i] &&
2635 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2636 kfree(sbi->s_qf_names[i]);
2637 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
2639 #endif
2640 unlock_super(sb);
2641 unlock_kernel();
2642 return err;
2645 static int ext3_statfs (struct dentry * dentry, struct kstatfs * buf)
2647 struct super_block *sb = dentry->d_sb;
2648 struct ext3_sb_info *sbi = EXT3_SB(sb);
2649 struct ext3_super_block *es = sbi->s_es;
2650 u64 fsid;
2652 if (test_opt(sb, MINIX_DF)) {
2653 sbi->s_overhead_last = 0;
2654 } else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
2655 unsigned long ngroups = sbi->s_groups_count, i;
2656 ext3_fsblk_t overhead = 0;
2657 smp_rmb();
2660 * Compute the overhead (FS structures). This is constant
2661 * for a given filesystem unless the number of block groups
2662 * changes so we cache the previous value until it does.
2666 * All of the blocks before first_data_block are
2667 * overhead
2669 overhead = le32_to_cpu(es->s_first_data_block);
2672 * Add the overhead attributed to the superblock and
2673 * block group descriptors. If the sparse superblocks
2674 * feature is turned on, then not all groups have this.
2676 for (i = 0; i < ngroups; i++) {
2677 overhead += ext3_bg_has_super(sb, i) +
2678 ext3_bg_num_gdb(sb, i);
2679 cond_resched();
2683 * Every block group has an inode bitmap, a block
2684 * bitmap, and an inode table.
2686 overhead += ngroups * (2 + sbi->s_itb_per_group);
2687 sbi->s_overhead_last = overhead;
2688 smp_wmb();
2689 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
2692 buf->f_type = EXT3_SUPER_MAGIC;
2693 buf->f_bsize = sb->s_blocksize;
2694 buf->f_blocks = le32_to_cpu(es->s_blocks_count) - sbi->s_overhead_last;
2695 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter);
2696 buf->f_bavail = buf->f_bfree - le32_to_cpu(es->s_r_blocks_count);
2697 if (buf->f_bfree < le32_to_cpu(es->s_r_blocks_count))
2698 buf->f_bavail = 0;
2699 buf->f_files = le32_to_cpu(es->s_inodes_count);
2700 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
2701 buf->f_namelen = EXT3_NAME_LEN;
2702 fsid = le64_to_cpup((void *)es->s_uuid) ^
2703 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
2704 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
2705 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
2706 return 0;
2709 /* Helper function for writing quotas on sync - we need to start transaction before quota file
2710 * is locked for write. Otherwise the are possible deadlocks:
2711 * Process 1 Process 2
2712 * ext3_create() quota_sync()
2713 * journal_start() write_dquot()
2714 * vfs_dq_init() down(dqio_mutex)
2715 * down(dqio_mutex) journal_start()
2719 #ifdef CONFIG_QUOTA
2721 static inline struct inode *dquot_to_inode(struct dquot *dquot)
2723 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
2726 static int ext3_write_dquot(struct dquot *dquot)
2728 int ret, err;
2729 handle_t *handle;
2730 struct inode *inode;
2732 inode = dquot_to_inode(dquot);
2733 handle = ext3_journal_start(inode,
2734 EXT3_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
2735 if (IS_ERR(handle))
2736 return PTR_ERR(handle);
2737 ret = dquot_commit(dquot);
2738 err = ext3_journal_stop(handle);
2739 if (!ret)
2740 ret = err;
2741 return ret;
2744 static int ext3_acquire_dquot(struct dquot *dquot)
2746 int ret, err;
2747 handle_t *handle;
2749 handle = ext3_journal_start(dquot_to_inode(dquot),
2750 EXT3_QUOTA_INIT_BLOCKS(dquot->dq_sb));
2751 if (IS_ERR(handle))
2752 return PTR_ERR(handle);
2753 ret = dquot_acquire(dquot);
2754 err = ext3_journal_stop(handle);
2755 if (!ret)
2756 ret = err;
2757 return ret;
2760 static int ext3_release_dquot(struct dquot *dquot)
2762 int ret, err;
2763 handle_t *handle;
2765 handle = ext3_journal_start(dquot_to_inode(dquot),
2766 EXT3_QUOTA_DEL_BLOCKS(dquot->dq_sb));
2767 if (IS_ERR(handle)) {
2768 /* Release dquot anyway to avoid endless cycle in dqput() */
2769 dquot_release(dquot);
2770 return PTR_ERR(handle);
2772 ret = dquot_release(dquot);
2773 err = ext3_journal_stop(handle);
2774 if (!ret)
2775 ret = err;
2776 return ret;
2779 static int ext3_mark_dquot_dirty(struct dquot *dquot)
2781 /* Are we journaling quotas? */
2782 if (EXT3_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
2783 EXT3_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
2784 dquot_mark_dquot_dirty(dquot);
2785 return ext3_write_dquot(dquot);
2786 } else {
2787 return dquot_mark_dquot_dirty(dquot);
2791 static int ext3_write_info(struct super_block *sb, int type)
2793 int ret, err;
2794 handle_t *handle;
2796 /* Data block + inode block */
2797 handle = ext3_journal_start(sb->s_root->d_inode, 2);
2798 if (IS_ERR(handle))
2799 return PTR_ERR(handle);
2800 ret = dquot_commit_info(sb, type);
2801 err = ext3_journal_stop(handle);
2802 if (!ret)
2803 ret = err;
2804 return ret;
2808 * Turn on quotas during mount time - we need to find
2809 * the quota file and such...
2811 static int ext3_quota_on_mount(struct super_block *sb, int type)
2813 return vfs_quota_on_mount(sb, EXT3_SB(sb)->s_qf_names[type],
2814 EXT3_SB(sb)->s_jquota_fmt, type);
2818 * Standard function to be called on quota_on
2820 static int ext3_quota_on(struct super_block *sb, int type, int format_id,
2821 char *name, int remount)
2823 int err;
2824 struct path path;
2826 if (!test_opt(sb, QUOTA))
2827 return -EINVAL;
2828 /* When remounting, no checks are needed and in fact, name is NULL */
2829 if (remount)
2830 return vfs_quota_on(sb, type, format_id, name, remount);
2832 err = kern_path(name, LOOKUP_FOLLOW, &path);
2833 if (err)
2834 return err;
2836 /* Quotafile not on the same filesystem? */
2837 if (path.mnt->mnt_sb != sb) {
2838 path_put(&path);
2839 return -EXDEV;
2841 /* Journaling quota? */
2842 if (EXT3_SB(sb)->s_qf_names[type]) {
2843 /* Quotafile not of fs root? */
2844 if (path.dentry->d_parent != sb->s_root)
2845 printk(KERN_WARNING
2846 "EXT3-fs: Quota file not on filesystem root. "
2847 "Journaled quota will not work.\n");
2851 * When we journal data on quota file, we have to flush journal to see
2852 * all updates to the file when we bypass pagecache...
2854 if (ext3_should_journal_data(path.dentry->d_inode)) {
2856 * We don't need to lock updates but journal_flush() could
2857 * otherwise be livelocked...
2859 journal_lock_updates(EXT3_SB(sb)->s_journal);
2860 err = journal_flush(EXT3_SB(sb)->s_journal);
2861 journal_unlock_updates(EXT3_SB(sb)->s_journal);
2862 if (err) {
2863 path_put(&path);
2864 return err;
2868 err = vfs_quota_on_path(sb, type, format_id, &path);
2869 path_put(&path);
2870 return err;
2873 /* Read data from quotafile - avoid pagecache and such because we cannot afford
2874 * acquiring the locks... As quota files are never truncated and quota code
2875 * itself serializes the operations (and noone else should touch the files)
2876 * we don't have to be afraid of races */
2877 static ssize_t ext3_quota_read(struct super_block *sb, int type, char *data,
2878 size_t len, loff_t off)
2880 struct inode *inode = sb_dqopt(sb)->files[type];
2881 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2882 int err = 0;
2883 int offset = off & (sb->s_blocksize - 1);
2884 int tocopy;
2885 size_t toread;
2886 struct buffer_head *bh;
2887 loff_t i_size = i_size_read(inode);
2889 if (off > i_size)
2890 return 0;
2891 if (off+len > i_size)
2892 len = i_size-off;
2893 toread = len;
2894 while (toread > 0) {
2895 tocopy = sb->s_blocksize - offset < toread ?
2896 sb->s_blocksize - offset : toread;
2897 bh = ext3_bread(NULL, inode, blk, 0, &err);
2898 if (err)
2899 return err;
2900 if (!bh) /* A hole? */
2901 memset(data, 0, tocopy);
2902 else
2903 memcpy(data, bh->b_data+offset, tocopy);
2904 brelse(bh);
2905 offset = 0;
2906 toread -= tocopy;
2907 data += tocopy;
2908 blk++;
2910 return len;
2913 /* Write to quotafile (we know the transaction is already started and has
2914 * enough credits) */
2915 static ssize_t ext3_quota_write(struct super_block *sb, int type,
2916 const char *data, size_t len, loff_t off)
2918 struct inode *inode = sb_dqopt(sb)->files[type];
2919 sector_t blk = off >> EXT3_BLOCK_SIZE_BITS(sb);
2920 int err = 0;
2921 int offset = off & (sb->s_blocksize - 1);
2922 int tocopy;
2923 int journal_quota = EXT3_SB(sb)->s_qf_names[type] != NULL;
2924 size_t towrite = len;
2925 struct buffer_head *bh;
2926 handle_t *handle = journal_current_handle();
2928 if (!handle) {
2929 printk(KERN_WARNING "EXT3-fs: Quota write (off=%Lu, len=%Lu)"
2930 " cancelled because transaction is not started.\n",
2931 (unsigned long long)off, (unsigned long long)len);
2932 return -EIO;
2934 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
2935 while (towrite > 0) {
2936 tocopy = sb->s_blocksize - offset < towrite ?
2937 sb->s_blocksize - offset : towrite;
2938 bh = ext3_bread(handle, inode, blk, 1, &err);
2939 if (!bh)
2940 goto out;
2941 if (journal_quota) {
2942 err = ext3_journal_get_write_access(handle, bh);
2943 if (err) {
2944 brelse(bh);
2945 goto out;
2948 lock_buffer(bh);
2949 memcpy(bh->b_data+offset, data, tocopy);
2950 flush_dcache_page(bh->b_page);
2951 unlock_buffer(bh);
2952 if (journal_quota)
2953 err = ext3_journal_dirty_metadata(handle, bh);
2954 else {
2955 /* Always do at least ordered writes for quotas */
2956 err = ext3_journal_dirty_data(handle, bh);
2957 mark_buffer_dirty(bh);
2959 brelse(bh);
2960 if (err)
2961 goto out;
2962 offset = 0;
2963 towrite -= tocopy;
2964 data += tocopy;
2965 blk++;
2967 out:
2968 if (len == towrite) {
2969 mutex_unlock(&inode->i_mutex);
2970 return err;
2972 if (inode->i_size < off+len-towrite) {
2973 i_size_write(inode, off+len-towrite);
2974 EXT3_I(inode)->i_disksize = inode->i_size;
2976 inode->i_version++;
2977 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
2978 ext3_mark_inode_dirty(handle, inode);
2979 mutex_unlock(&inode->i_mutex);
2980 return len - towrite;
2983 #endif
2985 static int ext3_get_sb(struct file_system_type *fs_type,
2986 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
2988 return get_sb_bdev(fs_type, flags, dev_name, data, ext3_fill_super, mnt);
2991 static struct file_system_type ext3_fs_type = {
2992 .owner = THIS_MODULE,
2993 .name = "ext3",
2994 .get_sb = ext3_get_sb,
2995 .kill_sb = kill_block_super,
2996 .fs_flags = FS_REQUIRES_DEV,
2999 static int __init init_ext3_fs(void)
3001 int err = init_ext3_xattr();
3002 if (err)
3003 return err;
3004 err = init_inodecache();
3005 if (err)
3006 goto out1;
3007 err = register_filesystem(&ext3_fs_type);
3008 if (err)
3009 goto out;
3010 return 0;
3011 out:
3012 destroy_inodecache();
3013 out1:
3014 exit_ext3_xattr();
3015 return err;
3018 static void __exit exit_ext3_fs(void)
3020 unregister_filesystem(&ext3_fs_type);
3021 destroy_inodecache();
3022 exit_ext3_xattr();
3025 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
3026 MODULE_DESCRIPTION("Second Extended Filesystem with journaling extensions");
3027 MODULE_LICENSE("GPL");
3028 module_init(init_ext3_fs)
3029 module_exit(exit_ext3_fs)