2 * linux/fs/ext4/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)
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>
22 #include <linux/time.h>
23 #include <linux/jbd2.h>
24 #include <linux/slab.h>
25 #include <linux/init.h>
26 #include <linux/blkdev.h>
27 #include <linux/parser.h>
28 #include <linux/smp_lock.h>
29 #include <linux/buffer_head.h>
30 #include <linux/exportfs.h>
31 #include <linux/vfs.h>
32 #include <linux/random.h>
33 #include <linux/mount.h>
34 #include <linux/namei.h>
35 #include <linux/quotaops.h>
36 #include <linux/seq_file.h>
37 #include <linux/proc_fs.h>
38 #include <linux/marker.h>
39 #include <linux/log2.h>
40 #include <linux/crc16.h>
41 #include <asm/uaccess.h>
44 #include "ext4_jbd2.h"
50 struct proc_dir_entry *ext4_proc_root;
52 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
53 unsigned long journal_devnum);
54 static int ext4_create_journal(struct super_block *, struct ext4_super_block *,
56 static void ext4_commit_super(struct super_block *sb,
57 struct ext4_super_block *es, int sync);
58 static void ext4_mark_recovery_complete(struct super_block *sb,
59 struct ext4_super_block *es);
60 static void ext4_clear_journal_err(struct super_block *sb,
61 struct ext4_super_block *es);
62 static int ext4_sync_fs(struct super_block *sb, int wait);
63 static const char *ext4_decode_error(struct super_block *sb, int errno,
65 static int ext4_remount(struct super_block *sb, int *flags, char *data);
66 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
67 static void ext4_unlockfs(struct super_block *sb);
68 static void ext4_write_super(struct super_block *sb);
69 static void ext4_write_super_lockfs(struct super_block *sb);
72 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
73 struct ext4_group_desc *bg)
75 return le32_to_cpu(bg->bg_block_bitmap_lo) |
76 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
77 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
80 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
81 struct ext4_group_desc *bg)
83 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
84 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
85 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
88 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
89 struct ext4_group_desc *bg)
91 return le32_to_cpu(bg->bg_inode_table_lo) |
92 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
93 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
96 void ext4_block_bitmap_set(struct super_block *sb,
97 struct ext4_group_desc *bg, ext4_fsblk_t blk)
99 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
100 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
101 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
104 void ext4_inode_bitmap_set(struct super_block *sb,
105 struct ext4_group_desc *bg, ext4_fsblk_t blk)
107 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
108 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
109 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
112 void ext4_inode_table_set(struct super_block *sb,
113 struct ext4_group_desc *bg, ext4_fsblk_t blk)
115 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
116 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
117 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
121 * Wrappers for jbd2_journal_start/end.
123 * The only special thing we need to do here is to make sure that all
124 * journal_end calls result in the superblock being marked dirty, so
125 * that sync() will call the filesystem's write_super callback if
128 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
132 if (sb->s_flags & MS_RDONLY)
133 return ERR_PTR(-EROFS);
135 /* Special case here: if the journal has aborted behind our
136 * backs (eg. EIO in the commit thread), then we still need to
137 * take the FS itself readonly cleanly. */
138 journal = EXT4_SB(sb)->s_journal;
140 if (is_journal_aborted(journal)) {
141 ext4_abort(sb, __func__,
142 "Detected aborted journal");
143 return ERR_PTR(-EROFS);
145 return jbd2_journal_start(journal, nblocks);
148 * We're not journaling, return the appropriate indication.
150 current->journal_info = EXT4_NOJOURNAL_HANDLE;
151 return current->journal_info;
155 * The only special thing we need to do here is to make sure that all
156 * jbd2_journal_stop calls result in the superblock being marked dirty, so
157 * that sync() will call the filesystem's write_super callback if
160 int __ext4_journal_stop(const char *where, handle_t *handle)
162 struct super_block *sb;
166 if (!ext4_handle_valid(handle)) {
168 * Do this here since we don't call jbd2_journal_stop() in
171 current->journal_info = NULL;
174 sb = handle->h_transaction->t_journal->j_private;
176 rc = jbd2_journal_stop(handle);
181 __ext4_std_error(sb, where, err);
185 void ext4_journal_abort_handle(const char *caller, const char *err_fn,
186 struct buffer_head *bh, handle_t *handle, int err)
189 const char *errstr = ext4_decode_error(NULL, err, nbuf);
191 BUG_ON(!ext4_handle_valid(handle));
194 BUFFER_TRACE(bh, "abort");
199 if (is_handle_aborted(handle))
202 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
203 caller, errstr, err_fn);
205 jbd2_journal_abort_handle(handle);
208 /* Deal with the reporting of failure conditions on a filesystem such as
209 * inconsistencies detected or read IO failures.
211 * On ext2, we can store the error state of the filesystem in the
212 * superblock. That is not possible on ext4, because we may have other
213 * write ordering constraints on the superblock which prevent us from
214 * writing it out straight away; and given that the journal is about to
215 * be aborted, we can't rely on the current, or future, transactions to
216 * write out the superblock safely.
218 * We'll just use the jbd2_journal_abort() error code to record an error in
219 * the journal instead. On recovery, the journal will compain about
220 * that error until we've noted it down and cleared it.
223 static void ext4_handle_error(struct super_block *sb)
225 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
227 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
228 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
230 if (sb->s_flags & MS_RDONLY)
233 if (!test_opt(sb, ERRORS_CONT)) {
234 journal_t *journal = EXT4_SB(sb)->s_journal;
236 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
238 jbd2_journal_abort(journal, -EIO);
240 if (test_opt(sb, ERRORS_RO)) {
241 printk(KERN_CRIT "Remounting filesystem read-only\n");
242 sb->s_flags |= MS_RDONLY;
244 ext4_commit_super(sb, es, 1);
245 if (test_opt(sb, ERRORS_PANIC))
246 panic("EXT4-fs (device %s): panic forced after error\n",
250 void ext4_error(struct super_block *sb, const char *function,
251 const char *fmt, ...)
256 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
261 ext4_handle_error(sb);
264 static const char *ext4_decode_error(struct super_block *sb, int errno,
271 errstr = "IO failure";
274 errstr = "Out of memory";
277 if (!sb || EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)
278 errstr = "Journal has aborted";
280 errstr = "Readonly filesystem";
283 /* If the caller passed in an extra buffer for unknown
284 * errors, textualise them now. Else we just return
287 /* Check for truncated error codes... */
288 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
297 /* __ext4_std_error decodes expected errors from journaling functions
298 * automatically and invokes the appropriate error response. */
300 void __ext4_std_error(struct super_block *sb, const char *function, int errno)
305 /* Special case: if the error is EROFS, and we're not already
306 * inside a transaction, then there's really no point in logging
308 if (errno == -EROFS && journal_current_handle() == NULL &&
309 (sb->s_flags & MS_RDONLY))
312 errstr = ext4_decode_error(sb, errno, nbuf);
313 printk(KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
314 sb->s_id, function, errstr);
316 ext4_handle_error(sb);
320 * ext4_abort is a much stronger failure handler than ext4_error. The
321 * abort function may be used to deal with unrecoverable failures such
322 * as journal IO errors or ENOMEM at a critical moment in log management.
324 * We unconditionally force the filesystem into an ABORT|READONLY state,
325 * unless the error response on the fs has been set to panic in which
326 * case we take the easy way out and panic immediately.
329 void ext4_abort(struct super_block *sb, const char *function,
330 const char *fmt, ...)
334 printk(KERN_CRIT "ext4_abort called.\n");
337 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
342 if (test_opt(sb, ERRORS_PANIC))
343 panic("EXT4-fs panic from previous error\n");
345 if (sb->s_flags & MS_RDONLY)
348 printk(KERN_CRIT "Remounting filesystem read-only\n");
349 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
350 sb->s_flags |= MS_RDONLY;
351 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
352 if (EXT4_SB(sb)->s_journal)
353 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
356 void ext4_warning(struct super_block *sb, const char *function,
357 const char *fmt, ...)
362 printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
369 void ext4_update_dynamic_rev(struct super_block *sb)
371 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
373 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
376 ext4_warning(sb, __func__,
377 "updating to rev %d because of new feature flag, "
378 "running e2fsck is recommended",
381 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
382 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
383 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
384 /* leave es->s_feature_*compat flags alone */
385 /* es->s_uuid will be set by e2fsck if empty */
388 * The rest of the superblock fields should be zero, and if not it
389 * means they are likely already in use, so leave them alone. We
390 * can leave it up to e2fsck to clean up any inconsistencies there.
395 * Open the external journal device
397 static struct block_device *ext4_blkdev_get(dev_t dev)
399 struct block_device *bdev;
400 char b[BDEVNAME_SIZE];
402 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
408 printk(KERN_ERR "EXT4: failed to open journal device %s: %ld\n",
409 __bdevname(dev, b), PTR_ERR(bdev));
414 * Release the journal device
416 static int ext4_blkdev_put(struct block_device *bdev)
419 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
422 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
424 struct block_device *bdev;
427 bdev = sbi->journal_bdev;
429 ret = ext4_blkdev_put(bdev);
430 sbi->journal_bdev = NULL;
435 static inline struct inode *orphan_list_entry(struct list_head *l)
437 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
440 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
444 printk(KERN_ERR "sb orphan head is %d\n",
445 le32_to_cpu(sbi->s_es->s_last_orphan));
447 printk(KERN_ERR "sb_info orphan list:\n");
448 list_for_each(l, &sbi->s_orphan) {
449 struct inode *inode = orphan_list_entry(l);
451 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
452 inode->i_sb->s_id, inode->i_ino, inode,
453 inode->i_mode, inode->i_nlink,
458 static void ext4_put_super(struct super_block *sb)
460 struct ext4_sb_info *sbi = EXT4_SB(sb);
461 struct ext4_super_block *es = sbi->s_es;
465 ext4_ext_release(sb);
466 ext4_xattr_put_super(sb);
467 if (sbi->s_journal) {
468 err = jbd2_journal_destroy(sbi->s_journal);
469 sbi->s_journal = NULL;
471 ext4_abort(sb, __func__,
472 "Couldn't clean up the journal");
474 if (!(sb->s_flags & MS_RDONLY)) {
475 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
476 es->s_state = cpu_to_le16(sbi->s_mount_state);
477 ext4_commit_super(sb, es, 1);
480 remove_proc_entry("inode_readahead_blks", sbi->s_proc);
481 remove_proc_entry(sb->s_id, ext4_proc_root);
484 for (i = 0; i < sbi->s_gdb_count; i++)
485 brelse(sbi->s_group_desc[i]);
486 kfree(sbi->s_group_desc);
487 kfree(sbi->s_flex_groups);
488 percpu_counter_destroy(&sbi->s_freeblocks_counter);
489 percpu_counter_destroy(&sbi->s_freeinodes_counter);
490 percpu_counter_destroy(&sbi->s_dirs_counter);
491 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
494 for (i = 0; i < MAXQUOTAS; i++)
495 kfree(sbi->s_qf_names[i]);
498 /* Debugging code just in case the in-memory inode orphan list
499 * isn't empty. The on-disk one can be non-empty if we've
500 * detected an error and taken the fs readonly, but the
501 * in-memory list had better be clean by this point. */
502 if (!list_empty(&sbi->s_orphan))
503 dump_orphan_list(sb, sbi);
504 J_ASSERT(list_empty(&sbi->s_orphan));
506 invalidate_bdev(sb->s_bdev);
507 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
509 * Invalidate the journal device's buffers. We don't want them
510 * floating about in memory - the physical journal device may
511 * hotswapped, and it breaks the `ro-after' testing code.
513 sync_blockdev(sbi->journal_bdev);
514 invalidate_bdev(sbi->journal_bdev);
515 ext4_blkdev_remove(sbi);
517 sb->s_fs_info = NULL;
522 static struct kmem_cache *ext4_inode_cachep;
525 * Called inside transaction, so use GFP_NOFS
527 static struct inode *ext4_alloc_inode(struct super_block *sb)
529 struct ext4_inode_info *ei;
531 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
534 #ifdef CONFIG_EXT4_FS_POSIX_ACL
535 ei->i_acl = EXT4_ACL_NOT_CACHED;
536 ei->i_default_acl = EXT4_ACL_NOT_CACHED;
538 ei->vfs_inode.i_version = 1;
539 ei->vfs_inode.i_data.writeback_index = 0;
540 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
541 INIT_LIST_HEAD(&ei->i_prealloc_list);
542 spin_lock_init(&ei->i_prealloc_lock);
544 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
545 * therefore it can be null here. Don't check it, just initialize
548 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
549 ei->i_reserved_data_blocks = 0;
550 ei->i_reserved_meta_blocks = 0;
551 ei->i_allocated_meta_blocks = 0;
552 ei->i_delalloc_reserved_flag = 0;
553 spin_lock_init(&(ei->i_block_reservation_lock));
554 return &ei->vfs_inode;
557 static void ext4_destroy_inode(struct inode *inode)
559 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
560 printk("EXT4 Inode %p: orphan list check failed!\n",
562 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
563 EXT4_I(inode), sizeof(struct ext4_inode_info),
567 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
570 static void init_once(void *foo)
572 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
574 INIT_LIST_HEAD(&ei->i_orphan);
575 #ifdef CONFIG_EXT4_FS_XATTR
576 init_rwsem(&ei->xattr_sem);
578 init_rwsem(&ei->i_data_sem);
579 inode_init_once(&ei->vfs_inode);
582 static int init_inodecache(void)
584 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
585 sizeof(struct ext4_inode_info),
586 0, (SLAB_RECLAIM_ACCOUNT|
589 if (ext4_inode_cachep == NULL)
594 static void destroy_inodecache(void)
596 kmem_cache_destroy(ext4_inode_cachep);
599 static void ext4_clear_inode(struct inode *inode)
601 #ifdef CONFIG_EXT4_FS_POSIX_ACL
602 if (EXT4_I(inode)->i_acl &&
603 EXT4_I(inode)->i_acl != EXT4_ACL_NOT_CACHED) {
604 posix_acl_release(EXT4_I(inode)->i_acl);
605 EXT4_I(inode)->i_acl = EXT4_ACL_NOT_CACHED;
607 if (EXT4_I(inode)->i_default_acl &&
608 EXT4_I(inode)->i_default_acl != EXT4_ACL_NOT_CACHED) {
609 posix_acl_release(EXT4_I(inode)->i_default_acl);
610 EXT4_I(inode)->i_default_acl = EXT4_ACL_NOT_CACHED;
613 ext4_discard_preallocations(inode);
614 if (EXT4_JOURNAL(inode))
615 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
616 &EXT4_I(inode)->jinode);
619 static inline void ext4_show_quota_options(struct seq_file *seq,
620 struct super_block *sb)
622 #if defined(CONFIG_QUOTA)
623 struct ext4_sb_info *sbi = EXT4_SB(sb);
625 if (sbi->s_jquota_fmt)
626 seq_printf(seq, ",jqfmt=%s",
627 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold" : "vfsv0");
629 if (sbi->s_qf_names[USRQUOTA])
630 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
632 if (sbi->s_qf_names[GRPQUOTA])
633 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
635 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
636 seq_puts(seq, ",usrquota");
638 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
639 seq_puts(seq, ",grpquota");
645 * - it's set to a non-default value OR
646 * - if the per-sb default is different from the global default
648 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
651 unsigned long def_mount_opts;
652 struct super_block *sb = vfs->mnt_sb;
653 struct ext4_sb_info *sbi = EXT4_SB(sb);
654 struct ext4_super_block *es = sbi->s_es;
656 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
657 def_errors = le16_to_cpu(es->s_errors);
659 if (sbi->s_sb_block != 1)
660 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
661 if (test_opt(sb, MINIX_DF))
662 seq_puts(seq, ",minixdf");
663 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
664 seq_puts(seq, ",grpid");
665 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
666 seq_puts(seq, ",nogrpid");
667 if (sbi->s_resuid != EXT4_DEF_RESUID ||
668 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
669 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
671 if (sbi->s_resgid != EXT4_DEF_RESGID ||
672 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
673 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
675 if (test_opt(sb, ERRORS_RO)) {
676 if (def_errors == EXT4_ERRORS_PANIC ||
677 def_errors == EXT4_ERRORS_CONTINUE) {
678 seq_puts(seq, ",errors=remount-ro");
681 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
682 seq_puts(seq, ",errors=continue");
683 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
684 seq_puts(seq, ",errors=panic");
685 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
686 seq_puts(seq, ",nouid32");
687 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
688 seq_puts(seq, ",debug");
689 if (test_opt(sb, OLDALLOC))
690 seq_puts(seq, ",oldalloc");
691 #ifdef CONFIG_EXT4_FS_XATTR
692 if (test_opt(sb, XATTR_USER) &&
693 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
694 seq_puts(seq, ",user_xattr");
695 if (!test_opt(sb, XATTR_USER) &&
696 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
697 seq_puts(seq, ",nouser_xattr");
700 #ifdef CONFIG_EXT4_FS_POSIX_ACL
701 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
702 seq_puts(seq, ",acl");
703 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
704 seq_puts(seq, ",noacl");
706 if (!test_opt(sb, RESERVATION))
707 seq_puts(seq, ",noreservation");
708 if (sbi->s_commit_interval) {
709 seq_printf(seq, ",commit=%u",
710 (unsigned) (sbi->s_commit_interval / HZ));
713 * We're changing the default of barrier mount option, so
714 * let's always display its mount state so it's clear what its
717 seq_puts(seq, ",barrier=");
718 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
719 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
720 seq_puts(seq, ",journal_async_commit");
721 if (test_opt(sb, NOBH))
722 seq_puts(seq, ",nobh");
723 if (!test_opt(sb, EXTENTS))
724 seq_puts(seq, ",noextents");
725 if (test_opt(sb, I_VERSION))
726 seq_puts(seq, ",i_version");
727 if (!test_opt(sb, DELALLOC))
728 seq_puts(seq, ",nodelalloc");
732 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
734 * journal mode get enabled in different ways
735 * So just print the value even if we didn't specify it
737 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
738 seq_puts(seq, ",data=journal");
739 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
740 seq_puts(seq, ",data=ordered");
741 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
742 seq_puts(seq, ",data=writeback");
744 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
745 seq_printf(seq, ",inode_readahead_blks=%u",
746 sbi->s_inode_readahead_blks);
748 if (test_opt(sb, DATA_ERR_ABORT))
749 seq_puts(seq, ",data_err=abort");
751 ext4_show_quota_options(seq, sb);
756 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
757 u64 ino, u32 generation)
761 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
762 return ERR_PTR(-ESTALE);
763 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
764 return ERR_PTR(-ESTALE);
766 /* iget isn't really right if the inode is currently unallocated!!
768 * ext4_read_inode will return a bad_inode if the inode had been
769 * deleted, so we should be safe.
771 * Currently we don't know the generation for parent directory, so
772 * a generation of 0 means "accept any"
774 inode = ext4_iget(sb, ino);
776 return ERR_CAST(inode);
777 if (generation && inode->i_generation != generation) {
779 return ERR_PTR(-ESTALE);
785 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
786 int fh_len, int fh_type)
788 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
792 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
793 int fh_len, int fh_type)
795 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
800 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
801 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
803 static int ext4_dquot_initialize(struct inode *inode, int type);
804 static int ext4_dquot_drop(struct inode *inode);
805 static int ext4_write_dquot(struct dquot *dquot);
806 static int ext4_acquire_dquot(struct dquot *dquot);
807 static int ext4_release_dquot(struct dquot *dquot);
808 static int ext4_mark_dquot_dirty(struct dquot *dquot);
809 static int ext4_write_info(struct super_block *sb, int type);
810 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
811 char *path, int remount);
812 static int ext4_quota_on_mount(struct super_block *sb, int type);
813 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
814 size_t len, loff_t off);
815 static ssize_t ext4_quota_write(struct super_block *sb, int type,
816 const char *data, size_t len, loff_t off);
818 static struct dquot_operations ext4_quota_operations = {
819 .initialize = ext4_dquot_initialize,
820 .drop = ext4_dquot_drop,
821 .alloc_space = dquot_alloc_space,
822 .alloc_inode = dquot_alloc_inode,
823 .free_space = dquot_free_space,
824 .free_inode = dquot_free_inode,
825 .transfer = dquot_transfer,
826 .write_dquot = ext4_write_dquot,
827 .acquire_dquot = ext4_acquire_dquot,
828 .release_dquot = ext4_release_dquot,
829 .mark_dirty = ext4_mark_dquot_dirty,
830 .write_info = ext4_write_info
833 static struct quotactl_ops ext4_qctl_operations = {
834 .quota_on = ext4_quota_on,
835 .quota_off = vfs_quota_off,
836 .quota_sync = vfs_quota_sync,
837 .get_info = vfs_get_dqinfo,
838 .set_info = vfs_set_dqinfo,
839 .get_dqblk = vfs_get_dqblk,
840 .set_dqblk = vfs_set_dqblk
844 static const struct super_operations ext4_sops = {
845 .alloc_inode = ext4_alloc_inode,
846 .destroy_inode = ext4_destroy_inode,
847 .write_inode = ext4_write_inode,
848 .dirty_inode = ext4_dirty_inode,
849 .delete_inode = ext4_delete_inode,
850 .put_super = ext4_put_super,
851 .write_super = ext4_write_super,
852 .sync_fs = ext4_sync_fs,
853 .write_super_lockfs = ext4_write_super_lockfs,
854 .unlockfs = ext4_unlockfs,
855 .statfs = ext4_statfs,
856 .remount_fs = ext4_remount,
857 .clear_inode = ext4_clear_inode,
858 .show_options = ext4_show_options,
860 .quota_read = ext4_quota_read,
861 .quota_write = ext4_quota_write,
865 static const struct export_operations ext4_export_ops = {
866 .fh_to_dentry = ext4_fh_to_dentry,
867 .fh_to_parent = ext4_fh_to_parent,
868 .get_parent = ext4_get_parent,
872 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
873 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
874 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
875 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
876 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
877 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
878 Opt_journal_checksum, Opt_journal_async_commit,
879 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
880 Opt_data_err_abort, Opt_data_err_ignore,
881 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
882 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
883 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
884 Opt_grpquota, Opt_extents, Opt_noextents, Opt_i_version,
885 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
886 Opt_inode_readahead_blks
889 static const match_table_t tokens = {
890 {Opt_bsd_df, "bsddf"},
891 {Opt_minix_df, "minixdf"},
892 {Opt_grpid, "grpid"},
893 {Opt_grpid, "bsdgroups"},
894 {Opt_nogrpid, "nogrpid"},
895 {Opt_nogrpid, "sysvgroups"},
896 {Opt_resgid, "resgid=%u"},
897 {Opt_resuid, "resuid=%u"},
899 {Opt_err_cont, "errors=continue"},
900 {Opt_err_panic, "errors=panic"},
901 {Opt_err_ro, "errors=remount-ro"},
902 {Opt_nouid32, "nouid32"},
903 {Opt_debug, "debug"},
904 {Opt_oldalloc, "oldalloc"},
905 {Opt_orlov, "orlov"},
906 {Opt_user_xattr, "user_xattr"},
907 {Opt_nouser_xattr, "nouser_xattr"},
909 {Opt_noacl, "noacl"},
910 {Opt_reservation, "reservation"},
911 {Opt_noreservation, "noreservation"},
912 {Opt_noload, "noload"},
915 {Opt_commit, "commit=%u"},
916 {Opt_journal_update, "journal=update"},
917 {Opt_journal_inum, "journal=%u"},
918 {Opt_journal_dev, "journal_dev=%u"},
919 {Opt_journal_checksum, "journal_checksum"},
920 {Opt_journal_async_commit, "journal_async_commit"},
921 {Opt_abort, "abort"},
922 {Opt_data_journal, "data=journal"},
923 {Opt_data_ordered, "data=ordered"},
924 {Opt_data_writeback, "data=writeback"},
925 {Opt_data_err_abort, "data_err=abort"},
926 {Opt_data_err_ignore, "data_err=ignore"},
927 {Opt_offusrjquota, "usrjquota="},
928 {Opt_usrjquota, "usrjquota=%s"},
929 {Opt_offgrpjquota, "grpjquota="},
930 {Opt_grpjquota, "grpjquota=%s"},
931 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
932 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
933 {Opt_grpquota, "grpquota"},
934 {Opt_noquota, "noquota"},
935 {Opt_quota, "quota"},
936 {Opt_usrquota, "usrquota"},
937 {Opt_barrier, "barrier=%u"},
938 {Opt_extents, "extents"},
939 {Opt_noextents, "noextents"},
940 {Opt_i_version, "i_version"},
941 {Opt_stripe, "stripe=%u"},
942 {Opt_resize, "resize"},
943 {Opt_delalloc, "delalloc"},
944 {Opt_nodelalloc, "nodelalloc"},
945 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
949 static ext4_fsblk_t get_sb_block(void **data)
951 ext4_fsblk_t sb_block;
952 char *options = (char *) *data;
954 if (!options || strncmp(options, "sb=", 3) != 0)
955 return 1; /* Default location */
957 /*todo: use simple_strtoll with >32bit ext4 */
958 sb_block = simple_strtoul(options, &options, 0);
959 if (*options && *options != ',') {
960 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
966 *data = (void *) options;
970 static int parse_options(char *options, struct super_block *sb,
971 unsigned int *inum, unsigned long *journal_devnum,
972 ext4_fsblk_t *n_blocks_count, int is_remount)
974 struct ext4_sb_info *sbi = EXT4_SB(sb);
976 substring_t args[MAX_OPT_ARGS];
983 ext4_fsblk_t last_block;
988 while ((p = strsep(&options, ",")) != NULL) {
993 token = match_token(p, tokens, args);
996 clear_opt(sbi->s_mount_opt, MINIX_DF);
999 set_opt(sbi->s_mount_opt, MINIX_DF);
1002 set_opt(sbi->s_mount_opt, GRPID);
1005 clear_opt(sbi->s_mount_opt, GRPID);
1008 if (match_int(&args[0], &option))
1010 sbi->s_resuid = option;
1013 if (match_int(&args[0], &option))
1015 sbi->s_resgid = option;
1018 /* handled by get_sb_block() instead of here */
1019 /* *sb_block = match_int(&args[0]); */
1022 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1023 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1024 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1027 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1028 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1029 set_opt(sbi->s_mount_opt, ERRORS_RO);
1032 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1033 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1034 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1037 set_opt(sbi->s_mount_opt, NO_UID32);
1040 set_opt(sbi->s_mount_opt, DEBUG);
1043 set_opt(sbi->s_mount_opt, OLDALLOC);
1046 clear_opt(sbi->s_mount_opt, OLDALLOC);
1048 #ifdef CONFIG_EXT4_FS_XATTR
1049 case Opt_user_xattr:
1050 set_opt(sbi->s_mount_opt, XATTR_USER);
1052 case Opt_nouser_xattr:
1053 clear_opt(sbi->s_mount_opt, XATTR_USER);
1056 case Opt_user_xattr:
1057 case Opt_nouser_xattr:
1058 printk(KERN_ERR "EXT4 (no)user_xattr options "
1062 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1064 set_opt(sbi->s_mount_opt, POSIX_ACL);
1067 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1072 printk(KERN_ERR "EXT4 (no)acl options "
1076 case Opt_reservation:
1077 set_opt(sbi->s_mount_opt, RESERVATION);
1079 case Opt_noreservation:
1080 clear_opt(sbi->s_mount_opt, RESERVATION);
1082 case Opt_journal_update:
1084 /* Eventually we will want to be able to create
1085 a journal file here. For now, only allow the
1086 user to specify an existing inode to be the
1089 printk(KERN_ERR "EXT4-fs: cannot specify "
1090 "journal on remount\n");
1093 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
1095 case Opt_journal_inum:
1097 printk(KERN_ERR "EXT4-fs: cannot specify "
1098 "journal on remount\n");
1101 if (match_int(&args[0], &option))
1105 case Opt_journal_dev:
1107 printk(KERN_ERR "EXT4-fs: cannot specify "
1108 "journal on remount\n");
1111 if (match_int(&args[0], &option))
1113 *journal_devnum = option;
1115 case Opt_journal_checksum:
1116 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1118 case Opt_journal_async_commit:
1119 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1120 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1123 set_opt(sbi->s_mount_opt, NOLOAD);
1126 if (match_int(&args[0], &option))
1131 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1132 sbi->s_commit_interval = HZ * option;
1134 case Opt_data_journal:
1135 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1137 case Opt_data_ordered:
1138 data_opt = EXT4_MOUNT_ORDERED_DATA;
1140 case Opt_data_writeback:
1141 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1144 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
1147 "EXT4-fs: cannot change data "
1148 "mode on remount\n");
1152 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
1153 sbi->s_mount_opt |= data_opt;
1156 case Opt_data_err_abort:
1157 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1159 case Opt_data_err_ignore:
1160 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1169 if ((sb_any_quota_enabled(sb) ||
1170 sb_any_quota_suspended(sb)) &&
1171 !sbi->s_qf_names[qtype]) {
1173 "EXT4-fs: Cannot change journaled "
1174 "quota options when quota turned on.\n");
1177 qname = match_strdup(&args[0]);
1180 "EXT4-fs: not enough memory for "
1181 "storing quotafile name.\n");
1184 if (sbi->s_qf_names[qtype] &&
1185 strcmp(sbi->s_qf_names[qtype], qname)) {
1187 "EXT4-fs: %s quota file already "
1188 "specified.\n", QTYPE2NAME(qtype));
1192 sbi->s_qf_names[qtype] = qname;
1193 if (strchr(sbi->s_qf_names[qtype], '/')) {
1195 "EXT4-fs: quotafile must be on "
1196 "filesystem root.\n");
1197 kfree(sbi->s_qf_names[qtype]);
1198 sbi->s_qf_names[qtype] = NULL;
1201 set_opt(sbi->s_mount_opt, QUOTA);
1203 case Opt_offusrjquota:
1206 case Opt_offgrpjquota:
1209 if ((sb_any_quota_enabled(sb) ||
1210 sb_any_quota_suspended(sb)) &&
1211 sbi->s_qf_names[qtype]) {
1212 printk(KERN_ERR "EXT4-fs: Cannot change "
1213 "journaled quota options when "
1214 "quota turned on.\n");
1218 * The space will be released later when all options
1219 * are confirmed to be correct
1221 sbi->s_qf_names[qtype] = NULL;
1223 case Opt_jqfmt_vfsold:
1224 qfmt = QFMT_VFS_OLD;
1226 case Opt_jqfmt_vfsv0:
1229 if ((sb_any_quota_enabled(sb) ||
1230 sb_any_quota_suspended(sb)) &&
1231 sbi->s_jquota_fmt != qfmt) {
1232 printk(KERN_ERR "EXT4-fs: Cannot change "
1233 "journaled quota options when "
1234 "quota turned on.\n");
1237 sbi->s_jquota_fmt = qfmt;
1241 set_opt(sbi->s_mount_opt, QUOTA);
1242 set_opt(sbi->s_mount_opt, USRQUOTA);
1245 set_opt(sbi->s_mount_opt, QUOTA);
1246 set_opt(sbi->s_mount_opt, GRPQUOTA);
1249 if (sb_any_quota_enabled(sb)) {
1250 printk(KERN_ERR "EXT4-fs: Cannot change quota "
1251 "options when quota turned on.\n");
1254 clear_opt(sbi->s_mount_opt, QUOTA);
1255 clear_opt(sbi->s_mount_opt, USRQUOTA);
1256 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1263 "EXT4-fs: quota options not supported.\n");
1267 case Opt_offusrjquota:
1268 case Opt_offgrpjquota:
1269 case Opt_jqfmt_vfsold:
1270 case Opt_jqfmt_vfsv0:
1272 "EXT4-fs: journaled quota options not "
1279 set_opt(sbi->s_mount_opt, ABORT);
1282 if (match_int(&args[0], &option))
1285 set_opt(sbi->s_mount_opt, BARRIER);
1287 clear_opt(sbi->s_mount_opt, BARRIER);
1293 printk("EXT4-fs: resize option only available "
1297 if (match_int(&args[0], &option) != 0)
1299 *n_blocks_count = option;
1302 set_opt(sbi->s_mount_opt, NOBH);
1305 clear_opt(sbi->s_mount_opt, NOBH);
1308 if (!EXT4_HAS_INCOMPAT_FEATURE(sb,
1309 EXT4_FEATURE_INCOMPAT_EXTENTS)) {
1310 ext4_warning(sb, __func__,
1311 "extents feature not enabled "
1312 "on this filesystem, use tune2fs\n");
1315 set_opt(sbi->s_mount_opt, EXTENTS);
1319 * When e2fsprogs support resizing an already existing
1320 * ext3 file system to greater than 2**32 we need to
1321 * add support to block allocator to handle growing
1322 * already existing block mapped inode so that blocks
1323 * allocated for them fall within 2**32
1325 last_block = ext4_blocks_count(sbi->s_es) - 1;
1326 if (last_block > 0xffffffffULL) {
1327 printk(KERN_ERR "EXT4-fs: Filesystem too "
1328 "large to mount with "
1329 "-o noextents options\n");
1332 clear_opt(sbi->s_mount_opt, EXTENTS);
1335 set_opt(sbi->s_mount_opt, I_VERSION);
1336 sb->s_flags |= MS_I_VERSION;
1338 case Opt_nodelalloc:
1339 clear_opt(sbi->s_mount_opt, DELALLOC);
1342 if (match_int(&args[0], &option))
1346 sbi->s_stripe = option;
1349 set_opt(sbi->s_mount_opt, DELALLOC);
1351 case Opt_inode_readahead_blks:
1352 if (match_int(&args[0], &option))
1354 if (option < 0 || option > (1 << 30))
1356 sbi->s_inode_readahead_blks = option;
1360 "EXT4-fs: Unrecognized mount option \"%s\" "
1361 "or missing value\n", p);
1366 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1367 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
1368 sbi->s_qf_names[USRQUOTA])
1369 clear_opt(sbi->s_mount_opt, USRQUOTA);
1371 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
1372 sbi->s_qf_names[GRPQUOTA])
1373 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1375 if ((sbi->s_qf_names[USRQUOTA] &&
1376 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
1377 (sbi->s_qf_names[GRPQUOTA] &&
1378 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1379 printk(KERN_ERR "EXT4-fs: old and new quota "
1380 "format mixing.\n");
1384 if (!sbi->s_jquota_fmt) {
1385 printk(KERN_ERR "EXT4-fs: journaled quota format "
1386 "not specified.\n");
1390 if (sbi->s_jquota_fmt) {
1391 printk(KERN_ERR "EXT4-fs: journaled quota format "
1392 "specified with no journaling "
1401 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1404 struct ext4_sb_info *sbi = EXT4_SB(sb);
1407 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1408 printk(KERN_ERR "EXT4-fs warning: revision level too high, "
1409 "forcing read-only mode\n");
1414 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1415 printk(KERN_WARNING "EXT4-fs warning: mounting unchecked fs, "
1416 "running e2fsck is recommended\n");
1417 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1419 "EXT4-fs warning: mounting fs with errors, "
1420 "running e2fsck is recommended\n");
1421 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1422 le16_to_cpu(es->s_mnt_count) >=
1423 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1425 "EXT4-fs warning: maximal mount count reached, "
1426 "running e2fsck is recommended\n");
1427 else if (le32_to_cpu(es->s_checkinterval) &&
1428 (le32_to_cpu(es->s_lastcheck) +
1429 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1431 "EXT4-fs warning: checktime reached, "
1432 "running e2fsck is recommended\n");
1433 if (!sbi->s_journal)
1434 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1435 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1436 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1437 le16_add_cpu(&es->s_mnt_count, 1);
1438 es->s_mtime = cpu_to_le32(get_seconds());
1439 ext4_update_dynamic_rev(sb);
1441 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1443 ext4_commit_super(sb, es, 1);
1444 if (test_opt(sb, DEBUG))
1445 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%lu, "
1446 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1448 sbi->s_groups_count,
1449 EXT4_BLOCKS_PER_GROUP(sb),
1450 EXT4_INODES_PER_GROUP(sb),
1453 if (EXT4_SB(sb)->s_journal) {
1454 printk(KERN_INFO "EXT4 FS on %s, %s journal on %s\n",
1455 sb->s_id, EXT4_SB(sb)->s_journal->j_inode ? "internal" :
1456 "external", EXT4_SB(sb)->s_journal->j_devname);
1458 printk(KERN_INFO "EXT4 FS on %s, no journal\n", sb->s_id);
1463 static int ext4_fill_flex_info(struct super_block *sb)
1465 struct ext4_sb_info *sbi = EXT4_SB(sb);
1466 struct ext4_group_desc *gdp = NULL;
1467 struct buffer_head *bh;
1468 ext4_group_t flex_group_count;
1469 ext4_group_t flex_group;
1470 int groups_per_flex = 0;
1473 if (!sbi->s_es->s_log_groups_per_flex) {
1474 sbi->s_log_groups_per_flex = 0;
1478 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1479 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1481 /* We allocate both existing and potentially added groups */
1482 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1483 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1484 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1485 sbi->s_flex_groups = kzalloc(flex_group_count *
1486 sizeof(struct flex_groups), GFP_KERNEL);
1487 if (sbi->s_flex_groups == NULL) {
1488 printk(KERN_ERR "EXT4-fs: not enough memory for "
1489 "%lu flex groups\n", flex_group_count);
1493 for (i = 0; i < sbi->s_groups_count; i++) {
1494 gdp = ext4_get_group_desc(sb, i, &bh);
1496 flex_group = ext4_flex_group(sbi, i);
1497 sbi->s_flex_groups[flex_group].free_inodes +=
1498 le16_to_cpu(gdp->bg_free_inodes_count);
1499 sbi->s_flex_groups[flex_group].free_blocks +=
1500 le16_to_cpu(gdp->bg_free_blocks_count);
1508 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1509 struct ext4_group_desc *gdp)
1513 if (sbi->s_es->s_feature_ro_compat &
1514 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1515 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1516 __le32 le_group = cpu_to_le32(block_group);
1518 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1519 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1520 crc = crc16(crc, (__u8 *)gdp, offset);
1521 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1522 /* for checksum of struct ext4_group_desc do the rest...*/
1523 if ((sbi->s_es->s_feature_incompat &
1524 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1525 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1526 crc = crc16(crc, (__u8 *)gdp + offset,
1527 le16_to_cpu(sbi->s_es->s_desc_size) -
1531 return cpu_to_le16(crc);
1534 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1535 struct ext4_group_desc *gdp)
1537 if ((sbi->s_es->s_feature_ro_compat &
1538 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1539 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1545 /* Called at mount-time, super-block is locked */
1546 static int ext4_check_descriptors(struct super_block *sb)
1548 struct ext4_sb_info *sbi = EXT4_SB(sb);
1549 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1550 ext4_fsblk_t last_block;
1551 ext4_fsblk_t block_bitmap;
1552 ext4_fsblk_t inode_bitmap;
1553 ext4_fsblk_t inode_table;
1554 int flexbg_flag = 0;
1557 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1560 ext4_debug("Checking group descriptors");
1562 for (i = 0; i < sbi->s_groups_count; i++) {
1563 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1565 if (i == sbi->s_groups_count - 1 || flexbg_flag)
1566 last_block = ext4_blocks_count(sbi->s_es) - 1;
1568 last_block = first_block +
1569 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1571 block_bitmap = ext4_block_bitmap(sb, gdp);
1572 if (block_bitmap < first_block || block_bitmap > last_block) {
1573 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1574 "Block bitmap for group %lu not in group "
1575 "(block %llu)!\n", i, block_bitmap);
1578 inode_bitmap = ext4_inode_bitmap(sb, gdp);
1579 if (inode_bitmap < first_block || inode_bitmap > last_block) {
1580 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1581 "Inode bitmap for group %lu not in group "
1582 "(block %llu)!\n", i, inode_bitmap);
1585 inode_table = ext4_inode_table(sb, gdp);
1586 if (inode_table < first_block ||
1587 inode_table + sbi->s_itb_per_group - 1 > last_block) {
1588 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1589 "Inode table for group %lu not in group "
1590 "(block %llu)!\n", i, inode_table);
1593 spin_lock(sb_bgl_lock(sbi, i));
1594 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1595 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1596 "Checksum for group %lu failed (%u!=%u)\n",
1597 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1598 gdp)), le16_to_cpu(gdp->bg_checksum));
1599 if (!(sb->s_flags & MS_RDONLY)) {
1600 spin_unlock(sb_bgl_lock(sbi, i));
1604 spin_unlock(sb_bgl_lock(sbi, i));
1606 first_block += EXT4_BLOCKS_PER_GROUP(sb);
1609 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1610 sbi->s_es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
1614 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1615 * the superblock) which were deleted from all directories, but held open by
1616 * a process at the time of a crash. We walk the list and try to delete these
1617 * inodes at recovery time (only with a read-write filesystem).
1619 * In order to keep the orphan inode chain consistent during traversal (in
1620 * case of crash during recovery), we link each inode into the superblock
1621 * orphan list_head and handle it the same way as an inode deletion during
1622 * normal operation (which journals the operations for us).
1624 * We only do an iget() and an iput() on each inode, which is very safe if we
1625 * accidentally point at an in-use or already deleted inode. The worst that
1626 * can happen in this case is that we get a "bit already cleared" message from
1627 * ext4_free_inode(). The only reason we would point at a wrong inode is if
1628 * e2fsck was run on this filesystem, and it must have already done the orphan
1629 * inode cleanup for us, so we can safely abort without any further action.
1631 static void ext4_orphan_cleanup(struct super_block *sb,
1632 struct ext4_super_block *es)
1634 unsigned int s_flags = sb->s_flags;
1635 int nr_orphans = 0, nr_truncates = 0;
1639 if (!es->s_last_orphan) {
1640 jbd_debug(4, "no orphan inodes to clean up\n");
1644 if (bdev_read_only(sb->s_bdev)) {
1645 printk(KERN_ERR "EXT4-fs: write access "
1646 "unavailable, skipping orphan cleanup.\n");
1650 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1651 if (es->s_last_orphan)
1652 jbd_debug(1, "Errors on filesystem, "
1653 "clearing orphan list.\n");
1654 es->s_last_orphan = 0;
1655 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1659 if (s_flags & MS_RDONLY) {
1660 printk(KERN_INFO "EXT4-fs: %s: orphan cleanup on readonly fs\n",
1662 sb->s_flags &= ~MS_RDONLY;
1665 /* Needed for iput() to work correctly and not trash data */
1666 sb->s_flags |= MS_ACTIVE;
1667 /* Turn on quotas so that they are updated correctly */
1668 for (i = 0; i < MAXQUOTAS; i++) {
1669 if (EXT4_SB(sb)->s_qf_names[i]) {
1670 int ret = ext4_quota_on_mount(sb, i);
1673 "EXT4-fs: Cannot turn on journaled "
1674 "quota: error %d\n", ret);
1679 while (es->s_last_orphan) {
1680 struct inode *inode;
1682 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1683 if (IS_ERR(inode)) {
1684 es->s_last_orphan = 0;
1688 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
1690 if (inode->i_nlink) {
1692 "%s: truncating inode %lu to %lld bytes\n",
1693 __func__, inode->i_ino, inode->i_size);
1694 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
1695 inode->i_ino, inode->i_size);
1696 ext4_truncate(inode);
1700 "%s: deleting unreferenced inode %lu\n",
1701 __func__, inode->i_ino);
1702 jbd_debug(2, "deleting unreferenced inode %lu\n",
1706 iput(inode); /* The delete magic happens here! */
1709 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
1712 printk(KERN_INFO "EXT4-fs: %s: %d orphan inode%s deleted\n",
1713 sb->s_id, PLURAL(nr_orphans));
1715 printk(KERN_INFO "EXT4-fs: %s: %d truncate%s cleaned up\n",
1716 sb->s_id, PLURAL(nr_truncates));
1718 /* Turn quotas off */
1719 for (i = 0; i < MAXQUOTAS; i++) {
1720 if (sb_dqopt(sb)->files[i])
1721 vfs_quota_off(sb, i, 0);
1724 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1727 * Maximal extent format file size.
1728 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1729 * extent format containers, within a sector_t, and within i_blocks
1730 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1731 * so that won't be a limiting factor.
1733 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1735 static loff_t ext4_max_size(int blkbits, int has_huge_files)
1738 loff_t upper_limit = MAX_LFS_FILESIZE;
1740 /* small i_blocks in vfs inode? */
1741 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1743 * CONFIG_LBD is not enabled implies the inode
1744 * i_block represent total blocks in 512 bytes
1745 * 32 == size of vfs inode i_blocks * 8
1747 upper_limit = (1LL << 32) - 1;
1749 /* total blocks in file system block size */
1750 upper_limit >>= (blkbits - 9);
1751 upper_limit <<= blkbits;
1754 /* 32-bit extent-start container, ee_block */
1759 /* Sanity check against vm- & vfs- imposed limits */
1760 if (res > upper_limit)
1767 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
1768 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1769 * We need to be 1 filesystem block less than the 2^48 sector limit.
1771 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
1773 loff_t res = EXT4_NDIR_BLOCKS;
1776 /* This is calculated to be the largest file size for a
1777 * dense, bitmapped file such that the total number of
1778 * sectors in the file, including data and all indirect blocks,
1779 * does not exceed 2^48 -1
1780 * __u32 i_blocks_lo and _u16 i_blocks_high representing the
1781 * total number of 512 bytes blocks of the file
1784 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
1786 * !has_huge_files or CONFIG_LBD is not enabled
1787 * implies the inode i_block represent total blocks in
1788 * 512 bytes 32 == size of vfs inode i_blocks * 8
1790 upper_limit = (1LL << 32) - 1;
1792 /* total blocks in file system block size */
1793 upper_limit >>= (bits - 9);
1797 * We use 48 bit ext4_inode i_blocks
1798 * With EXT4_HUGE_FILE_FL set the i_blocks
1799 * represent total number of blocks in
1800 * file system block size
1802 upper_limit = (1LL << 48) - 1;
1806 /* indirect blocks */
1808 /* double indirect blocks */
1809 meta_blocks += 1 + (1LL << (bits-2));
1810 /* tripple indirect blocks */
1811 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1813 upper_limit -= meta_blocks;
1814 upper_limit <<= bits;
1816 res += 1LL << (bits-2);
1817 res += 1LL << (2*(bits-2));
1818 res += 1LL << (3*(bits-2));
1820 if (res > upper_limit)
1823 if (res > MAX_LFS_FILESIZE)
1824 res = MAX_LFS_FILESIZE;
1829 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
1830 ext4_fsblk_t logical_sb_block, int nr)
1832 struct ext4_sb_info *sbi = EXT4_SB(sb);
1833 ext4_group_t bg, first_meta_bg;
1836 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1838 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
1840 return logical_sb_block + nr + 1;
1841 bg = sbi->s_desc_per_block * nr;
1842 if (ext4_bg_has_super(sb, bg))
1844 return (has_super + ext4_group_first_block_no(sb, bg));
1848 * ext4_get_stripe_size: Get the stripe size.
1849 * @sbi: In memory super block info
1851 * If we have specified it via mount option, then
1852 * use the mount option value. If the value specified at mount time is
1853 * greater than the blocks per group use the super block value.
1854 * If the super block value is greater than blocks per group return 0.
1855 * Allocator needs it be less than blocks per group.
1858 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
1860 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
1861 unsigned long stripe_width =
1862 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
1864 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
1865 return sbi->s_stripe;
1867 if (stripe_width <= sbi->s_blocks_per_group)
1868 return stripe_width;
1870 if (stride <= sbi->s_blocks_per_group)
1876 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
1877 __releases(kernel_lock)
1878 __acquires(kernel_lock)
1881 struct buffer_head *bh;
1882 struct ext4_super_block *es = NULL;
1883 struct ext4_sb_info *sbi;
1885 ext4_fsblk_t sb_block = get_sb_block(&data);
1886 ext4_fsblk_t logical_sb_block;
1887 unsigned long offset = 0;
1888 unsigned int journal_inum = 0;
1889 unsigned long journal_devnum = 0;
1890 unsigned long def_mount_opts;
1898 int needs_recovery, has_huge_files;
1903 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1906 sb->s_fs_info = sbi;
1907 sbi->s_mount_opt = 0;
1908 sbi->s_resuid = EXT4_DEF_RESUID;
1909 sbi->s_resgid = EXT4_DEF_RESGID;
1910 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
1911 sbi->s_sb_block = sb_block;
1915 /* Cleanup superblock name */
1916 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
1919 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
1921 printk(KERN_ERR "EXT4-fs: unable to set blocksize\n");
1926 * The ext4 superblock will not be buffer aligned for other than 1kB
1927 * block sizes. We need to calculate the offset from buffer start.
1929 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
1930 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
1931 offset = do_div(logical_sb_block, blocksize);
1933 logical_sb_block = sb_block;
1936 if (!(bh = sb_bread(sb, logical_sb_block))) {
1937 printk(KERN_ERR "EXT4-fs: unable to read superblock\n");
1941 * Note: s_es must be initialized as soon as possible because
1942 * some ext4 macro-instructions depend on its value
1944 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
1946 sb->s_magic = le16_to_cpu(es->s_magic);
1947 if (sb->s_magic != EXT4_SUPER_MAGIC)
1950 /* Set defaults before we parse the mount options */
1951 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1952 if (def_mount_opts & EXT4_DEFM_DEBUG)
1953 set_opt(sbi->s_mount_opt, DEBUG);
1954 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
1955 set_opt(sbi->s_mount_opt, GRPID);
1956 if (def_mount_opts & EXT4_DEFM_UID16)
1957 set_opt(sbi->s_mount_opt, NO_UID32);
1958 #ifdef CONFIG_EXT4_FS_XATTR
1959 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
1960 set_opt(sbi->s_mount_opt, XATTR_USER);
1962 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1963 if (def_mount_opts & EXT4_DEFM_ACL)
1964 set_opt(sbi->s_mount_opt, POSIX_ACL);
1966 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
1967 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
1968 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
1969 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
1970 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
1971 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
1973 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
1974 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1975 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
1976 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1978 set_opt(sbi->s_mount_opt, ERRORS_RO);
1980 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1981 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1983 set_opt(sbi->s_mount_opt, RESERVATION);
1984 set_opt(sbi->s_mount_opt, BARRIER);
1987 * turn on extents feature by default in ext4 filesystem
1988 * only if feature flag already set by mkfs or tune2fs.
1989 * Use -o noextents to turn it off
1991 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS))
1992 set_opt(sbi->s_mount_opt, EXTENTS);
1994 ext4_warning(sb, __func__,
1995 "extents feature not enabled on this filesystem, "
1999 * enable delayed allocation by default
2000 * Use -o nodelalloc to turn it off
2002 set_opt(sbi->s_mount_opt, DELALLOC);
2005 if (!parse_options((char *) data, sb, &journal_inum, &journal_devnum,
2009 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2010 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2012 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2013 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2014 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2015 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
2017 "EXT4-fs warning: feature flags set on rev 0 fs, "
2018 "running e2fsck is recommended\n");
2021 * Check feature flags regardless of the revision level, since we
2022 * previously didn't change the revision level when setting the flags,
2023 * so there is a chance incompat flags are set on a rev 0 filesystem.
2025 features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
2027 printk(KERN_ERR "EXT4-fs: %s: couldn't mount because of "
2028 "unsupported optional features (%x).\n",
2029 sb->s_id, le32_to_cpu(features));
2032 features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
2033 if (!(sb->s_flags & MS_RDONLY) && features) {
2034 printk(KERN_ERR "EXT4-fs: %s: couldn't mount RDWR because of "
2035 "unsupported optional features (%x).\n",
2036 sb->s_id, le32_to_cpu(features));
2039 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2040 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
2041 if (has_huge_files) {
2043 * Large file size enabled file system can only be
2044 * mount if kernel is build with CONFIG_LBD
2046 if (sizeof(root->i_blocks) < sizeof(u64) &&
2047 !(sb->s_flags & MS_RDONLY)) {
2048 printk(KERN_ERR "EXT4-fs: %s: Filesystem with huge "
2049 "files cannot be mounted read-write "
2050 "without CONFIG_LBD.\n", sb->s_id);
2054 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2056 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2057 blocksize > EXT4_MAX_BLOCK_SIZE) {
2059 "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
2060 blocksize, sb->s_id);
2064 if (sb->s_blocksize != blocksize) {
2066 /* Validate the filesystem blocksize */
2067 if (!sb_set_blocksize(sb, blocksize)) {
2068 printk(KERN_ERR "EXT4-fs: bad block size %d.\n",
2074 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2075 offset = do_div(logical_sb_block, blocksize);
2076 bh = sb_bread(sb, logical_sb_block);
2079 "EXT4-fs: Can't read superblock on 2nd try.\n");
2082 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
2084 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2086 "EXT4-fs: Magic mismatch, very weird !\n");
2091 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2093 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
2095 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2096 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2097 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
2099 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2100 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
2101 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
2102 (!is_power_of_2(sbi->s_inode_size)) ||
2103 (sbi->s_inode_size > blocksize)) {
2105 "EXT4-fs: unsupported inode size: %d\n",
2109 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2110 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
2112 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2113 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
2114 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
2115 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
2116 !is_power_of_2(sbi->s_desc_size)) {
2118 "EXT4-fs: unsupported descriptor size %lu\n",
2123 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
2124 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
2125 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
2126 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
2128 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
2129 if (sbi->s_inodes_per_block == 0)
2131 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2132 sbi->s_inodes_per_block;
2133 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
2135 sbi->s_mount_state = le16_to_cpu(es->s_state);
2136 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2137 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2138 for (i = 0; i < 4; i++)
2139 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2140 sbi->s_def_hash_version = es->s_def_hash_version;
2141 i = le32_to_cpu(es->s_flags);
2142 if (i & EXT2_FLAGS_UNSIGNED_HASH)
2143 sbi->s_hash_unsigned = 3;
2144 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2145 #ifdef __CHAR_UNSIGNED__
2146 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2147 sbi->s_hash_unsigned = 3;
2149 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2154 if (sbi->s_blocks_per_group > blocksize * 8) {
2156 "EXT4-fs: #blocks per group too big: %lu\n",
2157 sbi->s_blocks_per_group);
2160 if (sbi->s_inodes_per_group > blocksize * 8) {
2162 "EXT4-fs: #inodes per group too big: %lu\n",
2163 sbi->s_inodes_per_group);
2167 if (ext4_blocks_count(es) >
2168 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
2169 printk(KERN_ERR "EXT4-fs: filesystem on %s:"
2170 " too large to mount safely\n", sb->s_id);
2171 if (sizeof(sector_t) < 8)
2172 printk(KERN_WARNING "EXT4-fs: CONFIG_LBD not "
2177 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2180 /* ensure blocks_count calculation below doesn't sign-extend */
2181 if (ext4_blocks_count(es) + EXT4_BLOCKS_PER_GROUP(sb) <
2182 le32_to_cpu(es->s_first_data_block) + 1) {
2183 printk(KERN_WARNING "EXT4-fs: bad geometry: block count %llu, "
2184 "first data block %u, blocks per group %lu\n",
2185 ext4_blocks_count(es),
2186 le32_to_cpu(es->s_first_data_block),
2187 EXT4_BLOCKS_PER_GROUP(sb));
2190 blocks_count = (ext4_blocks_count(es) -
2191 le32_to_cpu(es->s_first_data_block) +
2192 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2193 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2194 sbi->s_groups_count = blocks_count;
2195 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2196 EXT4_DESC_PER_BLOCK(sb);
2197 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
2199 if (sbi->s_group_desc == NULL) {
2200 printk(KERN_ERR "EXT4-fs: not enough memory\n");
2204 #ifdef CONFIG_PROC_FS
2206 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
2209 proc_create_data("inode_readahead_blks", 0644, sbi->s_proc,
2211 &sbi->s_inode_readahead_blks);
2214 bgl_lock_init(&sbi->s_blockgroup_lock);
2216 for (i = 0; i < db_count; i++) {
2217 block = descriptor_loc(sb, logical_sb_block, i);
2218 sbi->s_group_desc[i] = sb_bread(sb, block);
2219 if (!sbi->s_group_desc[i]) {
2220 printk(KERN_ERR "EXT4-fs: "
2221 "can't read group descriptor %d\n", i);
2226 if (!ext4_check_descriptors(sb)) {
2227 printk(KERN_ERR "EXT4-fs: group descriptors corrupted!\n");
2230 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2231 if (!ext4_fill_flex_info(sb)) {
2233 "EXT4-fs: unable to initialize "
2234 "flex_bg meta info!\n");
2238 sbi->s_gdb_count = db_count;
2239 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2240 spin_lock_init(&sbi->s_next_gen_lock);
2242 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2243 ext4_count_free_blocks(sb));
2245 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2246 ext4_count_free_inodes(sb));
2249 err = percpu_counter_init(&sbi->s_dirs_counter,
2250 ext4_count_dirs(sb));
2253 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2256 printk(KERN_ERR "EXT4-fs: insufficient memory\n");
2260 sbi->s_stripe = ext4_get_stripe_size(sbi);
2263 * set up enough so that it can read an inode
2265 sb->s_op = &ext4_sops;
2266 sb->s_export_op = &ext4_export_ops;
2267 sb->s_xattr = ext4_xattr_handlers;
2269 sb->s_qcop = &ext4_qctl_operations;
2270 sb->dq_op = &ext4_quota_operations;
2272 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2276 needs_recovery = (es->s_last_orphan != 0 ||
2277 EXT4_HAS_INCOMPAT_FEATURE(sb,
2278 EXT4_FEATURE_INCOMPAT_RECOVER));
2281 * The first inode we look at is the journal inode. Don't try
2282 * root first: it may be modified in the journal!
2284 if (!test_opt(sb, NOLOAD) &&
2285 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2286 if (ext4_load_journal(sb, es, journal_devnum))
2288 if (!(sb->s_flags & MS_RDONLY) &&
2289 EXT4_SB(sb)->s_journal->j_failed_commit) {
2290 printk(KERN_CRIT "EXT4-fs error (device %s): "
2291 "ext4_fill_super: Journal transaction "
2292 "%u is corrupt\n", sb->s_id,
2293 EXT4_SB(sb)->s_journal->j_failed_commit);
2294 if (test_opt(sb, ERRORS_RO)) {
2296 "Mounting filesystem read-only\n");
2297 sb->s_flags |= MS_RDONLY;
2298 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2299 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2301 if (test_opt(sb, ERRORS_PANIC)) {
2302 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2303 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2304 ext4_commit_super(sb, es, 1);
2308 } else if (journal_inum) {
2309 if (ext4_create_journal(sb, es, journal_inum))
2311 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2312 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2313 printk(KERN_ERR "EXT4-fs: required journal recovery "
2314 "suppressed and not mounted read-only\n");
2317 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2318 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2319 sbi->s_journal = NULL;
2324 if (ext4_blocks_count(es) > 0xffffffffULL &&
2325 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2326 JBD2_FEATURE_INCOMPAT_64BIT)) {
2327 printk(KERN_ERR "ext4: Failed to set 64-bit journal feature\n");
2331 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2332 jbd2_journal_set_features(sbi->s_journal,
2333 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2334 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2335 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2336 jbd2_journal_set_features(sbi->s_journal,
2337 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2338 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2339 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2341 jbd2_journal_clear_features(sbi->s_journal,
2342 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2343 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2346 /* We have now updated the journal if required, so we can
2347 * validate the data journaling mode. */
2348 switch (test_opt(sb, DATA_FLAGS)) {
2350 /* No mode set, assume a default based on the journal
2351 * capabilities: ORDERED_DATA if the journal can
2352 * cope, else JOURNAL_DATA
2354 if (jbd2_journal_check_available_features
2355 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
2356 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2358 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2361 case EXT4_MOUNT_ORDERED_DATA:
2362 case EXT4_MOUNT_WRITEBACK_DATA:
2363 if (!jbd2_journal_check_available_features
2364 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
2365 printk(KERN_ERR "EXT4-fs: Journal does not support "
2366 "requested data journaling mode\n");
2375 if (test_opt(sb, NOBH)) {
2376 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2377 printk(KERN_WARNING "EXT4-fs: Ignoring nobh option - "
2378 "its supported only with writeback mode\n");
2379 clear_opt(sbi->s_mount_opt, NOBH);
2383 * The jbd2_journal_load will have done any necessary log recovery,
2384 * so we can safely mount the rest of the filesystem now.
2387 root = ext4_iget(sb, EXT4_ROOT_INO);
2389 printk(KERN_ERR "EXT4-fs: get root inode failed\n");
2390 ret = PTR_ERR(root);
2393 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2395 printk(KERN_ERR "EXT4-fs: corrupt root inode, run e2fsck\n");
2398 sb->s_root = d_alloc_root(root);
2400 printk(KERN_ERR "EXT4-fs: get root dentry failed\n");
2406 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
2408 /* determine the minimum size of new large inodes, if present */
2409 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2410 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2411 EXT4_GOOD_OLD_INODE_SIZE;
2412 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2413 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2414 if (sbi->s_want_extra_isize <
2415 le16_to_cpu(es->s_want_extra_isize))
2416 sbi->s_want_extra_isize =
2417 le16_to_cpu(es->s_want_extra_isize);
2418 if (sbi->s_want_extra_isize <
2419 le16_to_cpu(es->s_min_extra_isize))
2420 sbi->s_want_extra_isize =
2421 le16_to_cpu(es->s_min_extra_isize);
2424 /* Check if enough inode space is available */
2425 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2426 sbi->s_inode_size) {
2427 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2428 EXT4_GOOD_OLD_INODE_SIZE;
2429 printk(KERN_INFO "EXT4-fs: required extra inode space not"
2433 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
2434 printk(KERN_WARNING "EXT4-fs: Ignoring delalloc option - "
2435 "requested data journaling mode\n");
2436 clear_opt(sbi->s_mount_opt, DELALLOC);
2437 } else if (test_opt(sb, DELALLOC))
2438 printk(KERN_INFO "EXT4-fs: delayed allocation enabled\n");
2441 err = ext4_mb_init(sb, needs_recovery);
2443 printk(KERN_ERR "EXT4-fs: failed to initalize mballoc (%d)\n",
2449 * akpm: core read_super() calls in here with the superblock locked.
2450 * That deadlocks, because orphan cleanup needs to lock the superblock
2451 * in numerous places. Here we just pop the lock - it's relatively
2452 * harmless, because we are now ready to accept write_super() requests,
2453 * and aviro says that's the only reason for hanging onto the
2456 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2457 ext4_orphan_cleanup(sb, es);
2458 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
2459 if (needs_recovery) {
2460 printk(KERN_INFO "EXT4-fs: recovery complete.\n");
2461 ext4_mark_recovery_complete(sb, es);
2463 if (EXT4_SB(sb)->s_journal) {
2464 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2465 descr = " journalled data mode";
2466 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2467 descr = " ordered data mode";
2469 descr = " writeback data mode";
2471 descr = "out journal";
2473 printk(KERN_INFO "EXT4-fs: mounted filesystem %s with%s\n",
2481 printk(KERN_ERR "VFS: Can't find ext4 filesystem on dev %s.\n",
2486 printk(KERN_ERR "EXT4-fs (device %s): mount failed\n", sb->s_id);
2487 if (sbi->s_journal) {
2488 jbd2_journal_destroy(sbi->s_journal);
2489 sbi->s_journal = NULL;
2492 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2493 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2494 percpu_counter_destroy(&sbi->s_dirs_counter);
2495 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
2497 for (i = 0; i < db_count; i++)
2498 brelse(sbi->s_group_desc[i]);
2499 kfree(sbi->s_group_desc);
2502 remove_proc_entry("inode_readahead_blks", sbi->s_proc);
2503 remove_proc_entry(sb->s_id, ext4_proc_root);
2506 for (i = 0; i < MAXQUOTAS; i++)
2507 kfree(sbi->s_qf_names[i]);
2509 ext4_blkdev_remove(sbi);
2512 sb->s_fs_info = NULL;
2519 * Setup any per-fs journal parameters now. We'll do this both on
2520 * initial mount, once the journal has been initialised but before we've
2521 * done any recovery; and again on any subsequent remount.
2523 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
2525 struct ext4_sb_info *sbi = EXT4_SB(sb);
2527 if (sbi->s_commit_interval)
2528 journal->j_commit_interval = sbi->s_commit_interval;
2529 /* We could also set up an ext4-specific default for the commit
2530 * interval here, but for now we'll just fall back to the jbd
2533 spin_lock(&journal->j_state_lock);
2534 if (test_opt(sb, BARRIER))
2535 journal->j_flags |= JBD2_BARRIER;
2537 journal->j_flags &= ~JBD2_BARRIER;
2538 if (test_opt(sb, DATA_ERR_ABORT))
2539 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
2541 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
2542 spin_unlock(&journal->j_state_lock);
2545 static journal_t *ext4_get_journal(struct super_block *sb,
2546 unsigned int journal_inum)
2548 struct inode *journal_inode;
2551 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2553 /* First, test for the existence of a valid inode on disk. Bad
2554 * things happen if we iget() an unused inode, as the subsequent
2555 * iput() will try to delete it. */
2557 journal_inode = ext4_iget(sb, journal_inum);
2558 if (IS_ERR(journal_inode)) {
2559 printk(KERN_ERR "EXT4-fs: no journal found.\n");
2562 if (!journal_inode->i_nlink) {
2563 make_bad_inode(journal_inode);
2564 iput(journal_inode);
2565 printk(KERN_ERR "EXT4-fs: journal inode is deleted.\n");
2569 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
2570 journal_inode, journal_inode->i_size);
2571 if (!S_ISREG(journal_inode->i_mode)) {
2572 printk(KERN_ERR "EXT4-fs: invalid journal inode.\n");
2573 iput(journal_inode);
2577 journal = jbd2_journal_init_inode(journal_inode);
2579 printk(KERN_ERR "EXT4-fs: Could not load journal inode\n");
2580 iput(journal_inode);
2583 journal->j_private = sb;
2584 ext4_init_journal_params(sb, journal);
2588 static journal_t *ext4_get_dev_journal(struct super_block *sb,
2591 struct buffer_head *bh;
2595 int hblock, blocksize;
2596 ext4_fsblk_t sb_block;
2597 unsigned long offset;
2598 struct ext4_super_block *es;
2599 struct block_device *bdev;
2601 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2603 bdev = ext4_blkdev_get(j_dev);
2607 if (bd_claim(bdev, sb)) {
2609 "EXT4: failed to claim external journal device.\n");
2610 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
2614 blocksize = sb->s_blocksize;
2615 hblock = bdev_hardsect_size(bdev);
2616 if (blocksize < hblock) {
2618 "EXT4-fs: blocksize too small for journal device.\n");
2622 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
2623 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
2624 set_blocksize(bdev, blocksize);
2625 if (!(bh = __bread(bdev, sb_block, blocksize))) {
2626 printk(KERN_ERR "EXT4-fs: couldn't read superblock of "
2627 "external journal\n");
2631 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2632 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
2633 !(le32_to_cpu(es->s_feature_incompat) &
2634 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2635 printk(KERN_ERR "EXT4-fs: external journal has "
2636 "bad superblock\n");
2641 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2642 printk(KERN_ERR "EXT4-fs: journal UUID does not match\n");
2647 len = ext4_blocks_count(es);
2648 start = sb_block + 1;
2649 brelse(bh); /* we're done with the superblock */
2651 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
2652 start, len, blocksize);
2654 printk(KERN_ERR "EXT4-fs: failed to create device journal\n");
2657 journal->j_private = sb;
2658 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2659 wait_on_buffer(journal->j_sb_buffer);
2660 if (!buffer_uptodate(journal->j_sb_buffer)) {
2661 printk(KERN_ERR "EXT4-fs: I/O error on journal device\n");
2664 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
2665 printk(KERN_ERR "EXT4-fs: External journal has more than one "
2666 "user (unsupported) - %d\n",
2667 be32_to_cpu(journal->j_superblock->s_nr_users));
2670 EXT4_SB(sb)->journal_bdev = bdev;
2671 ext4_init_journal_params(sb, journal);
2674 jbd2_journal_destroy(journal);
2676 ext4_blkdev_put(bdev);
2680 static int ext4_load_journal(struct super_block *sb,
2681 struct ext4_super_block *es,
2682 unsigned long journal_devnum)
2685 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2688 int really_read_only;
2690 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2692 if (journal_devnum &&
2693 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2694 printk(KERN_INFO "EXT4-fs: external journal device major/minor "
2695 "numbers have changed\n");
2696 journal_dev = new_decode_dev(journal_devnum);
2698 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2700 really_read_only = bdev_read_only(sb->s_bdev);
2703 * Are we loading a blank journal or performing recovery after a
2704 * crash? For recovery, we need to check in advance whether we
2705 * can get read-write access to the device.
2708 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2709 if (sb->s_flags & MS_RDONLY) {
2710 printk(KERN_INFO "EXT4-fs: INFO: recovery "
2711 "required on readonly filesystem.\n");
2712 if (really_read_only) {
2713 printk(KERN_ERR "EXT4-fs: write access "
2714 "unavailable, cannot proceed.\n");
2717 printk(KERN_INFO "EXT4-fs: write access will "
2718 "be enabled during recovery.\n");
2722 if (journal_inum && journal_dev) {
2723 printk(KERN_ERR "EXT4-fs: filesystem has both journal "
2724 "and inode journals!\n");
2729 if (!(journal = ext4_get_journal(sb, journal_inum)))
2732 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
2736 if (journal->j_flags & JBD2_BARRIER)
2737 printk(KERN_INFO "EXT4-fs: barriers enabled\n");
2739 printk(KERN_INFO "EXT4-fs: barriers disabled\n");
2741 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2742 err = jbd2_journal_update_format(journal);
2744 printk(KERN_ERR "EXT4-fs: error updating journal.\n");
2745 jbd2_journal_destroy(journal);
2750 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
2751 err = jbd2_journal_wipe(journal, !really_read_only);
2753 err = jbd2_journal_load(journal);
2756 printk(KERN_ERR "EXT4-fs: error loading journal.\n");
2757 jbd2_journal_destroy(journal);
2761 EXT4_SB(sb)->s_journal = journal;
2762 ext4_clear_journal_err(sb, es);
2764 if (journal_devnum &&
2765 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2766 es->s_journal_dev = cpu_to_le32(journal_devnum);
2769 /* Make sure we flush the recovery flag to disk. */
2770 ext4_commit_super(sb, es, 1);
2776 static int ext4_create_journal(struct super_block *sb,
2777 struct ext4_super_block *es,
2778 unsigned int journal_inum)
2783 if (sb->s_flags & MS_RDONLY) {
2784 printk(KERN_ERR "EXT4-fs: readonly filesystem when trying to "
2785 "create journal.\n");
2789 journal = ext4_get_journal(sb, journal_inum);
2793 printk(KERN_INFO "EXT4-fs: creating new journal on inode %u\n",
2796 err = jbd2_journal_create(journal);
2798 printk(KERN_ERR "EXT4-fs: error creating journal.\n");
2799 jbd2_journal_destroy(journal);
2803 EXT4_SB(sb)->s_journal = journal;
2805 ext4_update_dynamic_rev(sb);
2806 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2807 EXT4_SET_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL);
2809 es->s_journal_inum = cpu_to_le32(journal_inum);
2812 /* Make sure we flush the recovery flag to disk. */
2813 ext4_commit_super(sb, es, 1);
2818 static void ext4_commit_super(struct super_block *sb,
2819 struct ext4_super_block *es, int sync)
2821 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
2825 if (buffer_write_io_error(sbh)) {
2827 * Oh, dear. A previous attempt to write the
2828 * superblock failed. This could happen because the
2829 * USB device was yanked out. Or it could happen to
2830 * be a transient write error and maybe the block will
2831 * be remapped. Nothing we can do but to retry the
2832 * write and hope for the best.
2834 printk(KERN_ERR "ext4: previous I/O error to "
2835 "superblock detected for %s.\n", sb->s_id);
2836 clear_buffer_write_io_error(sbh);
2837 set_buffer_uptodate(sbh);
2839 es->s_wtime = cpu_to_le32(get_seconds());
2840 ext4_free_blocks_count_set(es, ext4_count_free_blocks(sb));
2841 es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
2842 BUFFER_TRACE(sbh, "marking dirty");
2843 mark_buffer_dirty(sbh);
2845 sync_dirty_buffer(sbh);
2846 if (buffer_write_io_error(sbh)) {
2847 printk(KERN_ERR "ext4: I/O error while writing "
2848 "superblock for %s.\n", sb->s_id);
2849 clear_buffer_write_io_error(sbh);
2850 set_buffer_uptodate(sbh);
2857 * Have we just finished recovery? If so, and if we are mounting (or
2858 * remounting) the filesystem readonly, then we will end up with a
2859 * consistent fs on disk. Record that fact.
2861 static void ext4_mark_recovery_complete(struct super_block *sb,
2862 struct ext4_super_block *es)
2864 journal_t *journal = EXT4_SB(sb)->s_journal;
2866 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2867 BUG_ON(journal != NULL);
2870 jbd2_journal_lock_updates(journal);
2871 if (jbd2_journal_flush(journal) < 0)
2875 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
2876 sb->s_flags & MS_RDONLY) {
2877 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2879 ext4_commit_super(sb, es, 1);
2884 jbd2_journal_unlock_updates(journal);
2888 * If we are mounting (or read-write remounting) a filesystem whose journal
2889 * has recorded an error from a previous lifetime, move that error to the
2890 * main filesystem now.
2892 static void ext4_clear_journal_err(struct super_block *sb,
2893 struct ext4_super_block *es)
2899 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2901 journal = EXT4_SB(sb)->s_journal;
2904 * Now check for any error status which may have been recorded in the
2905 * journal by a prior ext4_error() or ext4_abort()
2908 j_errno = jbd2_journal_errno(journal);
2912 errstr = ext4_decode_error(sb, j_errno, nbuf);
2913 ext4_warning(sb, __func__, "Filesystem error recorded "
2914 "from previous mount: %s", errstr);
2915 ext4_warning(sb, __func__, "Marking fs in need of "
2916 "filesystem check.");
2918 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2919 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2920 ext4_commit_super(sb, es, 1);
2922 jbd2_journal_clear_err(journal);
2927 * Force the running and committing transactions to commit,
2928 * and wait on the commit.
2930 int ext4_force_commit(struct super_block *sb)
2935 if (sb->s_flags & MS_RDONLY)
2938 journal = EXT4_SB(sb)->s_journal;
2941 ret = ext4_journal_force_commit(journal);
2948 * Ext4 always journals updates to the superblock itself, so we don't
2949 * have to propagate any other updates to the superblock on disk at this
2950 * point. (We can probably nuke this function altogether, and remove
2951 * any mention to sb->s_dirt in all of fs/ext4; eventual cleanup...)
2953 static void ext4_write_super(struct super_block *sb)
2955 if (EXT4_SB(sb)->s_journal) {
2956 if (mutex_trylock(&sb->s_lock) != 0)
2960 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
2964 static int ext4_sync_fs(struct super_block *sb, int wait)
2968 trace_mark(ext4_sync_fs, "dev %s wait %d", sb->s_id, wait);
2970 if (EXT4_SB(sb)->s_journal) {
2972 ret = ext4_force_commit(sb);
2974 jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, NULL);
2976 ext4_commit_super(sb, EXT4_SB(sb)->s_es, wait);
2982 * LVM calls this function before a (read-only) snapshot is created. This
2983 * gives us a chance to flush the journal completely and mark the fs clean.
2985 static void ext4_write_super_lockfs(struct super_block *sb)
2989 if (!(sb->s_flags & MS_RDONLY)) {
2990 journal_t *journal = EXT4_SB(sb)->s_journal;
2993 /* Now we set up the journal barrier. */
2994 jbd2_journal_lock_updates(journal);
2997 * We don't want to clear needs_recovery flag when we
2998 * failed to flush the journal.
3000 if (jbd2_journal_flush(journal) < 0)
3004 /* Journal blocked and flushed, clear needs_recovery flag. */
3005 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3006 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
3011 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3012 * flag here, even though the filesystem is not technically dirty yet.
3014 static void ext4_unlockfs(struct super_block *sb)
3016 if (EXT4_SB(sb)->s_journal && !(sb->s_flags & MS_RDONLY)) {
3018 /* Reser the needs_recovery flag before the fs is unlocked. */
3019 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3020 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
3022 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3026 static int ext4_remount(struct super_block *sb, int *flags, char *data)
3028 struct ext4_super_block *es;
3029 struct ext4_sb_info *sbi = EXT4_SB(sb);
3030 ext4_fsblk_t n_blocks_count = 0;
3031 unsigned long old_sb_flags;
3032 struct ext4_mount_options old_opts;
3039 /* Store the original options */
3040 old_sb_flags = sb->s_flags;
3041 old_opts.s_mount_opt = sbi->s_mount_opt;
3042 old_opts.s_resuid = sbi->s_resuid;
3043 old_opts.s_resgid = sbi->s_resgid;
3044 old_opts.s_commit_interval = sbi->s_commit_interval;
3046 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3047 for (i = 0; i < MAXQUOTAS; i++)
3048 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3052 * Allow the "check" option to be passed as a remount option.
3054 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
3059 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT)
3060 ext4_abort(sb, __func__, "Abort forced by user");
3062 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3063 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
3068 ext4_init_journal_params(sb, sbi->s_journal);
3070 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
3071 n_blocks_count > ext4_blocks_count(es)) {
3072 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) {
3077 if (*flags & MS_RDONLY) {
3079 * First of all, the unconditional stuff we have to do
3080 * to disable replay of the journal when we next remount
3082 sb->s_flags |= MS_RDONLY;
3085 * OK, test if we are remounting a valid rw partition
3086 * readonly, and if so set the rdonly flag and then
3087 * mark the partition as valid again.
3089 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3090 (sbi->s_mount_state & EXT4_VALID_FS))
3091 es->s_state = cpu_to_le16(sbi->s_mount_state);
3094 * We have to unlock super so that we can wait for
3097 if (sbi->s_journal) {
3099 ext4_mark_recovery_complete(sb, es);
3104 if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3105 ~EXT4_FEATURE_RO_COMPAT_SUPP))) {
3106 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
3107 "remount RDWR because of unsupported "
3108 "optional features (%x).\n",
3109 sb->s_id, le32_to_cpu(ret));
3115 * Make sure the group descriptor checksums
3116 * are sane. If they aren't, refuse to
3119 for (g = 0; g < sbi->s_groups_count; g++) {
3120 struct ext4_group_desc *gdp =
3121 ext4_get_group_desc(sb, g, NULL);
3123 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3125 "EXT4-fs: ext4_remount: "
3126 "Checksum for group %lu failed (%u!=%u)\n",
3127 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3128 le16_to_cpu(gdp->bg_checksum));
3135 * If we have an unprocessed orphan list hanging
3136 * around from a previously readonly bdev mount,
3137 * require a full umount/remount for now.
3139 if (es->s_last_orphan) {
3140 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
3141 "remount RDWR because of unprocessed "
3142 "orphan inode list. Please "
3143 "umount/remount instead.\n",
3150 * Mounting a RDONLY partition read-write, so reread
3151 * and store the current valid flag. (It may have
3152 * been changed by e2fsck since we originally mounted
3156 ext4_clear_journal_err(sb, es);
3157 sbi->s_mount_state = le16_to_cpu(es->s_state);
3158 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
3160 if (!ext4_setup_super(sb, es, 0))
3161 sb->s_flags &= ~MS_RDONLY;
3164 if (sbi->s_journal == NULL)
3165 ext4_commit_super(sb, es, 1);
3168 /* Release old quota file names */
3169 for (i = 0; i < MAXQUOTAS; i++)
3170 if (old_opts.s_qf_names[i] &&
3171 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3172 kfree(old_opts.s_qf_names[i]);
3176 sb->s_flags = old_sb_flags;
3177 sbi->s_mount_opt = old_opts.s_mount_opt;
3178 sbi->s_resuid = old_opts.s_resuid;
3179 sbi->s_resgid = old_opts.s_resgid;
3180 sbi->s_commit_interval = old_opts.s_commit_interval;
3182 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3183 for (i = 0; i < MAXQUOTAS; i++) {
3184 if (sbi->s_qf_names[i] &&
3185 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3186 kfree(sbi->s_qf_names[i]);
3187 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3193 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
3195 struct super_block *sb = dentry->d_sb;
3196 struct ext4_sb_info *sbi = EXT4_SB(sb);
3197 struct ext4_super_block *es = sbi->s_es;
3200 if (test_opt(sb, MINIX_DF)) {
3201 sbi->s_overhead_last = 0;
3202 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
3203 ext4_group_t ngroups = sbi->s_groups_count, i;
3204 ext4_fsblk_t overhead = 0;
3208 * Compute the overhead (FS structures). This is constant
3209 * for a given filesystem unless the number of block groups
3210 * changes so we cache the previous value until it does.
3214 * All of the blocks before first_data_block are
3217 overhead = le32_to_cpu(es->s_first_data_block);
3220 * Add the overhead attributed to the superblock and
3221 * block group descriptors. If the sparse superblocks
3222 * feature is turned on, then not all groups have this.
3224 for (i = 0; i < ngroups; i++) {
3225 overhead += ext4_bg_has_super(sb, i) +
3226 ext4_bg_num_gdb(sb, i);
3231 * Every block group has an inode bitmap, a block
3232 * bitmap, and an inode table.
3234 overhead += ngroups * (2 + sbi->s_itb_per_group);
3235 sbi->s_overhead_last = overhead;
3237 sbi->s_blocks_last = ext4_blocks_count(es);
3240 buf->f_type = EXT4_SUPER_MAGIC;
3241 buf->f_bsize = sb->s_blocksize;
3242 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
3243 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3244 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
3245 ext4_free_blocks_count_set(es, buf->f_bfree);
3246 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3247 if (buf->f_bfree < ext4_r_blocks_count(es))
3249 buf->f_files = le32_to_cpu(es->s_inodes_count);
3250 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
3251 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
3252 buf->f_namelen = EXT4_NAME_LEN;
3253 fsid = le64_to_cpup((void *)es->s_uuid) ^
3254 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3255 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3256 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
3260 /* Helper function for writing quotas on sync - we need to start transaction before quota file
3261 * is locked for write. Otherwise the are possible deadlocks:
3262 * Process 1 Process 2
3263 * ext4_create() quota_sync()
3264 * jbd2_journal_start() write_dquot()
3265 * DQUOT_INIT() down(dqio_mutex)
3266 * down(dqio_mutex) jbd2_journal_start()
3272 static inline struct inode *dquot_to_inode(struct dquot *dquot)
3274 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3277 static int ext4_dquot_initialize(struct inode *inode, int type)
3282 /* We may create quota structure so we need to reserve enough blocks */
3283 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_INIT_BLOCKS(inode->i_sb));
3285 return PTR_ERR(handle);
3286 ret = dquot_initialize(inode, type);
3287 err = ext4_journal_stop(handle);
3293 static int ext4_dquot_drop(struct inode *inode)
3298 /* We may delete quota structure so we need to reserve enough blocks */
3299 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_DEL_BLOCKS(inode->i_sb));
3300 if (IS_ERR(handle)) {
3302 * We call dquot_drop() anyway to at least release references
3303 * to quota structures so that umount does not hang.
3306 return PTR_ERR(handle);
3308 ret = dquot_drop(inode);
3309 err = ext4_journal_stop(handle);
3315 static int ext4_write_dquot(struct dquot *dquot)
3319 struct inode *inode;
3321 inode = dquot_to_inode(dquot);
3322 handle = ext4_journal_start(inode,
3323 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
3325 return PTR_ERR(handle);
3326 ret = dquot_commit(dquot);
3327 err = ext4_journal_stop(handle);
3333 static int ext4_acquire_dquot(struct dquot *dquot)
3338 handle = ext4_journal_start(dquot_to_inode(dquot),
3339 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
3341 return PTR_ERR(handle);
3342 ret = dquot_acquire(dquot);
3343 err = ext4_journal_stop(handle);
3349 static int ext4_release_dquot(struct dquot *dquot)
3354 handle = ext4_journal_start(dquot_to_inode(dquot),
3355 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
3356 if (IS_ERR(handle)) {
3357 /* Release dquot anyway to avoid endless cycle in dqput() */
3358 dquot_release(dquot);
3359 return PTR_ERR(handle);
3361 ret = dquot_release(dquot);
3362 err = ext4_journal_stop(handle);
3368 static int ext4_mark_dquot_dirty(struct dquot *dquot)
3370 /* Are we journaling quotas? */
3371 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3372 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
3373 dquot_mark_dquot_dirty(dquot);
3374 return ext4_write_dquot(dquot);
3376 return dquot_mark_dquot_dirty(dquot);
3380 static int ext4_write_info(struct super_block *sb, int type)
3385 /* Data block + inode block */
3386 handle = ext4_journal_start(sb->s_root->d_inode, 2);
3388 return PTR_ERR(handle);
3389 ret = dquot_commit_info(sb, type);
3390 err = ext4_journal_stop(handle);
3397 * Turn on quotas during mount time - we need to find
3398 * the quota file and such...
3400 static int ext4_quota_on_mount(struct super_block *sb, int type)
3402 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3403 EXT4_SB(sb)->s_jquota_fmt, type);
3407 * Standard function to be called on quota_on
3409 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
3410 char *name, int remount)
3415 if (!test_opt(sb, QUOTA))
3417 /* When remounting, no checks are needed and in fact, name is NULL */
3419 return vfs_quota_on(sb, type, format_id, name, remount);
3421 err = kern_path(name, LOOKUP_FOLLOW, &path);
3425 /* Quotafile not on the same filesystem? */
3426 if (path.mnt->mnt_sb != sb) {
3430 /* Journaling quota? */
3431 if (EXT4_SB(sb)->s_qf_names[type]) {
3432 /* Quotafile not in fs root? */
3433 if (path.dentry->d_parent != sb->s_root)
3435 "EXT4-fs: Quota file not on filesystem root. "
3436 "Journaled quota will not work.\n");
3440 * When we journal data on quota file, we have to flush journal to see
3441 * all updates to the file when we bypass pagecache...
3443 if (EXT4_SB(sb)->s_journal &&
3444 ext4_should_journal_data(path.dentry->d_inode)) {
3446 * We don't need to lock updates but journal_flush() could
3447 * otherwise be livelocked...
3449 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
3450 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
3451 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3458 err = vfs_quota_on_path(sb, type, format_id, &path);
3463 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3464 * acquiring the locks... As quota files are never truncated and quota code
3465 * itself serializes the operations (and noone else should touch the files)
3466 * we don't have to be afraid of races */
3467 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
3468 size_t len, loff_t off)
3470 struct inode *inode = sb_dqopt(sb)->files[type];
3471 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3473 int offset = off & (sb->s_blocksize - 1);
3476 struct buffer_head *bh;
3477 loff_t i_size = i_size_read(inode);
3481 if (off+len > i_size)
3484 while (toread > 0) {
3485 tocopy = sb->s_blocksize - offset < toread ?
3486 sb->s_blocksize - offset : toread;
3487 bh = ext4_bread(NULL, inode, blk, 0, &err);
3490 if (!bh) /* A hole? */
3491 memset(data, 0, tocopy);
3493 memcpy(data, bh->b_data+offset, tocopy);
3503 /* Write to quotafile (we know the transaction is already started and has
3504 * enough credits) */
3505 static ssize_t ext4_quota_write(struct super_block *sb, int type,
3506 const char *data, size_t len, loff_t off)
3508 struct inode *inode = sb_dqopt(sb)->files[type];
3509 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3511 int offset = off & (sb->s_blocksize - 1);
3513 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
3514 size_t towrite = len;
3515 struct buffer_head *bh;
3516 handle_t *handle = journal_current_handle();
3518 if (EXT4_SB(sb)->s_journal && !handle) {
3519 printk(KERN_WARNING "EXT4-fs: Quota write (off=%llu, len=%llu)"
3520 " cancelled because transaction is not started.\n",
3521 (unsigned long long)off, (unsigned long long)len);
3524 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3525 while (towrite > 0) {
3526 tocopy = sb->s_blocksize - offset < towrite ?
3527 sb->s_blocksize - offset : towrite;
3528 bh = ext4_bread(handle, inode, blk, 1, &err);
3531 if (journal_quota) {
3532 err = ext4_journal_get_write_access(handle, bh);
3539 memcpy(bh->b_data+offset, data, tocopy);
3540 flush_dcache_page(bh->b_page);
3543 err = ext4_handle_dirty_metadata(handle, NULL, bh);
3545 /* Always do at least ordered writes for quotas */
3546 err = ext4_jbd2_file_inode(handle, inode);
3547 mark_buffer_dirty(bh);
3558 if (len == towrite) {
3559 mutex_unlock(&inode->i_mutex);
3562 if (inode->i_size < off+len-towrite) {
3563 i_size_write(inode, off+len-towrite);
3564 EXT4_I(inode)->i_disksize = inode->i_size;
3566 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3567 ext4_mark_inode_dirty(handle, inode);
3568 mutex_unlock(&inode->i_mutex);
3569 return len - towrite;
3574 static int ext4_get_sb(struct file_system_type *fs_type,
3575 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3577 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
3580 #ifdef CONFIG_PROC_FS
3581 static int ext4_ui_proc_show(struct seq_file *m, void *v)
3583 unsigned int *p = m->private;
3585 seq_printf(m, "%u\n", *p);
3589 static int ext4_ui_proc_open(struct inode *inode, struct file *file)
3591 return single_open(file, ext4_ui_proc_show, PDE(inode)->data);
3594 static ssize_t ext4_ui_proc_write(struct file *file, const char __user *buf,
3595 size_t cnt, loff_t *ppos)
3597 unsigned long *p = PDE(file->f_path.dentry->d_inode)->data;
3600 if (cnt >= sizeof(str))
3602 if (copy_from_user(str, buf, cnt))
3605 *p = simple_strtoul(str, NULL, 0);
3609 const struct file_operations ext4_ui_proc_fops = {
3610 .owner = THIS_MODULE,
3611 .open = ext4_ui_proc_open,
3613 .llseek = seq_lseek,
3614 .release = single_release,
3615 .write = ext4_ui_proc_write,
3619 static struct file_system_type ext4_fs_type = {
3620 .owner = THIS_MODULE,
3622 .get_sb = ext4_get_sb,
3623 .kill_sb = kill_block_super,
3624 .fs_flags = FS_REQUIRES_DEV,
3627 #ifdef CONFIG_EXT4DEV_COMPAT
3628 static int ext4dev_get_sb(struct file_system_type *fs_type,
3629 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3631 printk(KERN_WARNING "EXT4-fs: Update your userspace programs "
3632 "to mount using ext4\n");
3633 printk(KERN_WARNING "EXT4-fs: ext4dev backwards compatibility "
3634 "will go away by 2.6.31\n");
3635 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
3638 static struct file_system_type ext4dev_fs_type = {
3639 .owner = THIS_MODULE,
3641 .get_sb = ext4dev_get_sb,
3642 .kill_sb = kill_block_super,
3643 .fs_flags = FS_REQUIRES_DEV,
3645 MODULE_ALIAS("ext4dev");
3648 static int __init init_ext4_fs(void)
3652 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
3653 err = init_ext4_mballoc();
3657 err = init_ext4_xattr();
3660 err = init_inodecache();
3663 err = register_filesystem(&ext4_fs_type);
3666 #ifdef CONFIG_EXT4DEV_COMPAT
3667 err = register_filesystem(&ext4dev_fs_type);
3669 unregister_filesystem(&ext4_fs_type);
3675 destroy_inodecache();
3679 exit_ext4_mballoc();
3683 static void __exit exit_ext4_fs(void)
3685 unregister_filesystem(&ext4_fs_type);
3686 #ifdef CONFIG_EXT4DEV_COMPAT
3687 unregister_filesystem(&ext4dev_fs_type);
3689 destroy_inodecache();
3691 exit_ext4_mballoc();
3692 remove_proc_entry("fs/ext4", NULL);
3695 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
3696 MODULE_DESCRIPTION("Fourth Extended Filesystem with extents");
3697 MODULE_LICENSE("GPL");
3698 module_init(init_ext4_fs)
3699 module_exit(exit_ext4_fs)