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/log2.h>
38 #include <linux/crc16.h>
39 #include <asm/uaccess.h>
42 #include "ext4_jbd2.h"
48 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
49 unsigned long journal_devnum);
50 static int ext4_create_journal(struct super_block *, struct ext4_super_block *,
52 static void ext4_commit_super (struct super_block * sb,
53 struct ext4_super_block * es,
55 static void ext4_mark_recovery_complete(struct super_block * sb,
56 struct ext4_super_block * es);
57 static void ext4_clear_journal_err(struct super_block * sb,
58 struct ext4_super_block * es);
59 static int ext4_sync_fs(struct super_block *sb, int wait);
60 static const char *ext4_decode_error(struct super_block * sb, int errno,
62 static int ext4_remount (struct super_block * sb, int * flags, char * data);
63 static int ext4_statfs (struct dentry * dentry, struct kstatfs * buf);
64 static void ext4_unlockfs(struct super_block *sb);
65 static void ext4_write_super (struct super_block * sb);
66 static void ext4_write_super_lockfs(struct super_block *sb);
69 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
70 struct ext4_group_desc *bg)
72 return le32_to_cpu(bg->bg_block_bitmap_lo) |
73 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
74 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
77 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
78 struct ext4_group_desc *bg)
80 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
81 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
82 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
85 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
86 struct ext4_group_desc *bg)
88 return le32_to_cpu(bg->bg_inode_table_lo) |
89 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
90 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
93 void ext4_block_bitmap_set(struct super_block *sb,
94 struct ext4_group_desc *bg, ext4_fsblk_t blk)
96 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
97 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
98 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
101 void ext4_inode_bitmap_set(struct super_block *sb,
102 struct ext4_group_desc *bg, ext4_fsblk_t blk)
104 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
105 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
106 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
109 void ext4_inode_table_set(struct super_block *sb,
110 struct ext4_group_desc *bg, ext4_fsblk_t blk)
112 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
113 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
114 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
118 * Wrappers for jbd2_journal_start/end.
120 * The only special thing we need to do here is to make sure that all
121 * journal_end calls result in the superblock being marked dirty, so
122 * that sync() will call the filesystem's write_super callback if
125 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
129 if (sb->s_flags & MS_RDONLY)
130 return ERR_PTR(-EROFS);
132 /* Special case here: if the journal has aborted behind our
133 * backs (eg. EIO in the commit thread), then we still need to
134 * take the FS itself readonly cleanly. */
135 journal = EXT4_SB(sb)->s_journal;
136 if (is_journal_aborted(journal)) {
137 ext4_abort(sb, __func__,
138 "Detected aborted journal");
139 return ERR_PTR(-EROFS);
142 return jbd2_journal_start(journal, nblocks);
146 * The only special thing we need to do here is to make sure that all
147 * jbd2_journal_stop calls result in the superblock being marked dirty, so
148 * that sync() will call the filesystem's write_super callback if
151 int __ext4_journal_stop(const char *where, handle_t *handle)
153 struct super_block *sb;
157 sb = handle->h_transaction->t_journal->j_private;
159 rc = jbd2_journal_stop(handle);
164 __ext4_std_error(sb, where, err);
168 void ext4_journal_abort_handle(const char *caller, const char *err_fn,
169 struct buffer_head *bh, handle_t *handle, int err)
172 const char *errstr = ext4_decode_error(NULL, err, nbuf);
175 BUFFER_TRACE(bh, "abort");
180 if (is_handle_aborted(handle))
183 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
184 caller, errstr, err_fn);
186 jbd2_journal_abort_handle(handle);
189 /* Deal with the reporting of failure conditions on a filesystem such as
190 * inconsistencies detected or read IO failures.
192 * On ext2, we can store the error state of the filesystem in the
193 * superblock. That is not possible on ext4, because we may have other
194 * write ordering constraints on the superblock which prevent us from
195 * writing it out straight away; and given that the journal is about to
196 * be aborted, we can't rely on the current, or future, transactions to
197 * write out the superblock safely.
199 * We'll just use the jbd2_journal_abort() error code to record an error in
200 * the journal instead. On recovery, the journal will compain about
201 * that error until we've noted it down and cleared it.
204 static void ext4_handle_error(struct super_block *sb)
206 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
208 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
209 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
211 if (sb->s_flags & MS_RDONLY)
214 if (!test_opt (sb, ERRORS_CONT)) {
215 journal_t *journal = EXT4_SB(sb)->s_journal;
217 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
219 jbd2_journal_abort(journal, -EIO);
221 if (test_opt (sb, ERRORS_RO)) {
222 printk (KERN_CRIT "Remounting filesystem read-only\n");
223 sb->s_flags |= MS_RDONLY;
225 ext4_commit_super(sb, es, 1);
226 if (test_opt(sb, ERRORS_PANIC))
227 panic("EXT4-fs (device %s): panic forced after error\n",
231 void ext4_error (struct super_block * sb, const char * function,
232 const char * fmt, ...)
237 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ",sb->s_id, function);
242 ext4_handle_error(sb);
245 static const char *ext4_decode_error(struct super_block * sb, int errno,
252 errstr = "IO failure";
255 errstr = "Out of memory";
258 if (!sb || EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)
259 errstr = "Journal has aborted";
261 errstr = "Readonly filesystem";
264 /* If the caller passed in an extra buffer for unknown
265 * errors, textualise them now. Else we just return
268 /* Check for truncated error codes... */
269 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
278 /* __ext4_std_error decodes expected errors from journaling functions
279 * automatically and invokes the appropriate error response. */
281 void __ext4_std_error (struct super_block * sb, const char * function,
287 /* Special case: if the error is EROFS, and we're not already
288 * inside a transaction, then there's really no point in logging
290 if (errno == -EROFS && journal_current_handle() == NULL &&
291 (sb->s_flags & MS_RDONLY))
294 errstr = ext4_decode_error(sb, errno, nbuf);
295 printk (KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
296 sb->s_id, function, errstr);
298 ext4_handle_error(sb);
302 * ext4_abort is a much stronger failure handler than ext4_error. The
303 * abort function may be used to deal with unrecoverable failures such
304 * as journal IO errors or ENOMEM at a critical moment in log management.
306 * We unconditionally force the filesystem into an ABORT|READONLY state,
307 * unless the error response on the fs has been set to panic in which
308 * case we take the easy way out and panic immediately.
311 void ext4_abort (struct super_block * sb, const char * function,
312 const char * fmt, ...)
316 printk (KERN_CRIT "ext4_abort called.\n");
319 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ",sb->s_id, function);
324 if (test_opt(sb, ERRORS_PANIC))
325 panic("EXT4-fs panic from previous error\n");
327 if (sb->s_flags & MS_RDONLY)
330 printk(KERN_CRIT "Remounting filesystem read-only\n");
331 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
332 sb->s_flags |= MS_RDONLY;
333 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
334 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
337 void ext4_warning (struct super_block * sb, const char * function,
338 const char * fmt, ...)
343 printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
350 void ext4_update_dynamic_rev(struct super_block *sb)
352 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
354 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
357 ext4_warning(sb, __func__,
358 "updating to rev %d because of new feature flag, "
359 "running e2fsck is recommended",
362 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
363 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
364 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
365 /* leave es->s_feature_*compat flags alone */
366 /* es->s_uuid will be set by e2fsck if empty */
369 * The rest of the superblock fields should be zero, and if not it
370 * means they are likely already in use, so leave them alone. We
371 * can leave it up to e2fsck to clean up any inconsistencies there.
375 int ext4_update_compat_feature(handle_t *handle,
376 struct super_block *sb, __u32 compat)
379 if (!EXT4_HAS_COMPAT_FEATURE(sb, compat)) {
380 err = ext4_journal_get_write_access(handle,
384 EXT4_SET_COMPAT_FEATURE(sb, compat);
387 BUFFER_TRACE(EXT4_SB(sb)->s_sbh,
388 "call ext4_journal_dirty_met adata");
389 err = ext4_journal_dirty_metadata(handle,
395 int ext4_update_rocompat_feature(handle_t *handle,
396 struct super_block *sb, __u32 rocompat)
399 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, rocompat)) {
400 err = ext4_journal_get_write_access(handle,
404 EXT4_SET_RO_COMPAT_FEATURE(sb, rocompat);
407 BUFFER_TRACE(EXT4_SB(sb)->s_sbh,
408 "call ext4_journal_dirty_met adata");
409 err = ext4_journal_dirty_metadata(handle,
415 int ext4_update_incompat_feature(handle_t *handle,
416 struct super_block *sb, __u32 incompat)
419 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, incompat)) {
420 err = ext4_journal_get_write_access(handle,
424 EXT4_SET_INCOMPAT_FEATURE(sb, incompat);
427 BUFFER_TRACE(EXT4_SB(sb)->s_sbh,
428 "call ext4_journal_dirty_met adata");
429 err = ext4_journal_dirty_metadata(handle,
436 * Open the external journal device
438 static struct block_device *ext4_blkdev_get(dev_t dev)
440 struct block_device *bdev;
441 char b[BDEVNAME_SIZE];
443 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
449 printk(KERN_ERR "EXT4: failed to open journal device %s: %ld\n",
450 __bdevname(dev, b), PTR_ERR(bdev));
455 * Release the journal device
457 static int ext4_blkdev_put(struct block_device *bdev)
460 return blkdev_put(bdev);
463 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
465 struct block_device *bdev;
468 bdev = sbi->journal_bdev;
470 ret = ext4_blkdev_put(bdev);
471 sbi->journal_bdev = NULL;
476 static inline struct inode *orphan_list_entry(struct list_head *l)
478 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
481 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
485 printk(KERN_ERR "sb orphan head is %d\n",
486 le32_to_cpu(sbi->s_es->s_last_orphan));
488 printk(KERN_ERR "sb_info orphan list:\n");
489 list_for_each(l, &sbi->s_orphan) {
490 struct inode *inode = orphan_list_entry(l);
492 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
493 inode->i_sb->s_id, inode->i_ino, inode,
494 inode->i_mode, inode->i_nlink,
499 static void ext4_put_super (struct super_block * sb)
501 struct ext4_sb_info *sbi = EXT4_SB(sb);
502 struct ext4_super_block *es = sbi->s_es;
506 ext4_ext_release(sb);
507 ext4_xattr_put_super(sb);
508 jbd2_journal_destroy(sbi->s_journal);
509 sbi->s_journal = NULL;
510 if (!(sb->s_flags & MS_RDONLY)) {
511 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
512 es->s_state = cpu_to_le16(sbi->s_mount_state);
513 BUFFER_TRACE(sbi->s_sbh, "marking dirty");
514 mark_buffer_dirty(sbi->s_sbh);
515 ext4_commit_super(sb, es, 1);
518 for (i = 0; i < sbi->s_gdb_count; i++)
519 brelse(sbi->s_group_desc[i]);
520 kfree(sbi->s_group_desc);
521 kfree(sbi->s_flex_groups);
522 percpu_counter_destroy(&sbi->s_freeblocks_counter);
523 percpu_counter_destroy(&sbi->s_freeinodes_counter);
524 percpu_counter_destroy(&sbi->s_dirs_counter);
527 for (i = 0; i < MAXQUOTAS; i++)
528 kfree(sbi->s_qf_names[i]);
531 /* Debugging code just in case the in-memory inode orphan list
532 * isn't empty. The on-disk one can be non-empty if we've
533 * detected an error and taken the fs readonly, but the
534 * in-memory list had better be clean by this point. */
535 if (!list_empty(&sbi->s_orphan))
536 dump_orphan_list(sb, sbi);
537 J_ASSERT(list_empty(&sbi->s_orphan));
539 invalidate_bdev(sb->s_bdev);
540 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
542 * Invalidate the journal device's buffers. We don't want them
543 * floating about in memory - the physical journal device may
544 * hotswapped, and it breaks the `ro-after' testing code.
546 sync_blockdev(sbi->journal_bdev);
547 invalidate_bdev(sbi->journal_bdev);
548 ext4_blkdev_remove(sbi);
550 sb->s_fs_info = NULL;
555 static struct kmem_cache *ext4_inode_cachep;
558 * Called inside transaction, so use GFP_NOFS
560 static struct inode *ext4_alloc_inode(struct super_block *sb)
562 struct ext4_inode_info *ei;
564 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
567 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
568 ei->i_acl = EXT4_ACL_NOT_CACHED;
569 ei->i_default_acl = EXT4_ACL_NOT_CACHED;
571 ei->i_block_alloc_info = NULL;
572 ei->vfs_inode.i_version = 1;
573 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
574 INIT_LIST_HEAD(&ei->i_prealloc_list);
575 spin_lock_init(&ei->i_prealloc_lock);
576 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
577 return &ei->vfs_inode;
580 static void ext4_destroy_inode(struct inode *inode)
582 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
583 printk("EXT4 Inode %p: orphan list check failed!\n",
585 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
586 EXT4_I(inode), sizeof(struct ext4_inode_info),
590 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
593 static void init_once(struct kmem_cache *cachep, void *foo)
595 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
597 INIT_LIST_HEAD(&ei->i_orphan);
598 #ifdef CONFIG_EXT4DEV_FS_XATTR
599 init_rwsem(&ei->xattr_sem);
601 init_rwsem(&ei->i_data_sem);
602 inode_init_once(&ei->vfs_inode);
605 static int init_inodecache(void)
607 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
608 sizeof(struct ext4_inode_info),
609 0, (SLAB_RECLAIM_ACCOUNT|
612 if (ext4_inode_cachep == NULL)
617 static void destroy_inodecache(void)
619 kmem_cache_destroy(ext4_inode_cachep);
622 static void ext4_clear_inode(struct inode *inode)
624 struct ext4_block_alloc_info *rsv = EXT4_I(inode)->i_block_alloc_info;
625 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
626 if (EXT4_I(inode)->i_acl &&
627 EXT4_I(inode)->i_acl != EXT4_ACL_NOT_CACHED) {
628 posix_acl_release(EXT4_I(inode)->i_acl);
629 EXT4_I(inode)->i_acl = EXT4_ACL_NOT_CACHED;
631 if (EXT4_I(inode)->i_default_acl &&
632 EXT4_I(inode)->i_default_acl != EXT4_ACL_NOT_CACHED) {
633 posix_acl_release(EXT4_I(inode)->i_default_acl);
634 EXT4_I(inode)->i_default_acl = EXT4_ACL_NOT_CACHED;
637 ext4_discard_reservation(inode);
638 EXT4_I(inode)->i_block_alloc_info = NULL;
641 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
642 &EXT4_I(inode)->jinode);
645 static inline void ext4_show_quota_options(struct seq_file *seq, struct super_block *sb)
647 #if defined(CONFIG_QUOTA)
648 struct ext4_sb_info *sbi = EXT4_SB(sb);
650 if (sbi->s_jquota_fmt)
651 seq_printf(seq, ",jqfmt=%s",
652 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold": "vfsv0");
654 if (sbi->s_qf_names[USRQUOTA])
655 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
657 if (sbi->s_qf_names[GRPQUOTA])
658 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
660 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
661 seq_puts(seq, ",usrquota");
663 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
664 seq_puts(seq, ",grpquota");
670 * - it's set to a non-default value OR
671 * - if the per-sb default is different from the global default
673 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
676 unsigned long def_mount_opts;
677 struct super_block *sb = vfs->mnt_sb;
678 struct ext4_sb_info *sbi = EXT4_SB(sb);
679 struct ext4_super_block *es = sbi->s_es;
681 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
682 def_errors = le16_to_cpu(es->s_errors);
684 if (sbi->s_sb_block != 1)
685 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
686 if (test_opt(sb, MINIX_DF))
687 seq_puts(seq, ",minixdf");
688 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
689 seq_puts(seq, ",grpid");
690 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
691 seq_puts(seq, ",nogrpid");
692 if (sbi->s_resuid != EXT4_DEF_RESUID ||
693 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
694 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
696 if (sbi->s_resgid != EXT4_DEF_RESGID ||
697 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
698 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
700 if (test_opt(sb, ERRORS_RO)) {
701 if (def_errors == EXT4_ERRORS_PANIC ||
702 def_errors == EXT4_ERRORS_CONTINUE) {
703 seq_puts(seq, ",errors=remount-ro");
706 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
707 seq_puts(seq, ",errors=continue");
708 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
709 seq_puts(seq, ",errors=panic");
710 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
711 seq_puts(seq, ",nouid32");
712 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
713 seq_puts(seq, ",debug");
714 if (test_opt(sb, OLDALLOC))
715 seq_puts(seq, ",oldalloc");
716 #ifdef CONFIG_EXT4DEV_FS_XATTR
717 if (test_opt(sb, XATTR_USER) &&
718 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
719 seq_puts(seq, ",user_xattr");
720 if (!test_opt(sb, XATTR_USER) &&
721 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
722 seq_puts(seq, ",nouser_xattr");
725 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
726 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
727 seq_puts(seq, ",acl");
728 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
729 seq_puts(seq, ",noacl");
731 if (!test_opt(sb, RESERVATION))
732 seq_puts(seq, ",noreservation");
733 if (sbi->s_commit_interval) {
734 seq_printf(seq, ",commit=%u",
735 (unsigned) (sbi->s_commit_interval / HZ));
738 * We're changing the default of barrier mount option, so
739 * let's always display its mount state so it's clear what its
742 seq_puts(seq, ",barrier=");
743 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
744 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
745 seq_puts(seq, ",journal_async_commit");
746 if (test_opt(sb, NOBH))
747 seq_puts(seq, ",nobh");
748 if (!test_opt(sb, EXTENTS))
749 seq_puts(seq, ",noextents");
750 if (!test_opt(sb, MBALLOC))
751 seq_puts(seq, ",nomballoc");
752 if (test_opt(sb, I_VERSION))
753 seq_puts(seq, ",i_version");
756 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
758 * journal mode get enabled in different ways
759 * So just print the value even if we didn't specify it
761 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
762 seq_puts(seq, ",data=journal");
763 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
764 seq_puts(seq, ",data=ordered");
765 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
766 seq_puts(seq, ",data=writeback");
768 ext4_show_quota_options(seq, sb);
773 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
774 u64 ino, u32 generation)
778 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
779 return ERR_PTR(-ESTALE);
780 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
781 return ERR_PTR(-ESTALE);
783 /* iget isn't really right if the inode is currently unallocated!!
785 * ext4_read_inode will return a bad_inode if the inode had been
786 * deleted, so we should be safe.
788 * Currently we don't know the generation for parent directory, so
789 * a generation of 0 means "accept any"
791 inode = ext4_iget(sb, ino);
793 return ERR_CAST(inode);
794 if (generation && inode->i_generation != generation) {
796 return ERR_PTR(-ESTALE);
802 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
803 int fh_len, int fh_type)
805 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
809 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
810 int fh_len, int fh_type)
812 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
817 #define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
818 #define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
820 static int ext4_dquot_initialize(struct inode *inode, int type);
821 static int ext4_dquot_drop(struct inode *inode);
822 static int ext4_write_dquot(struct dquot *dquot);
823 static int ext4_acquire_dquot(struct dquot *dquot);
824 static int ext4_release_dquot(struct dquot *dquot);
825 static int ext4_mark_dquot_dirty(struct dquot *dquot);
826 static int ext4_write_info(struct super_block *sb, int type);
827 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
828 char *path, int remount);
829 static int ext4_quota_on_mount(struct super_block *sb, int type);
830 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
831 size_t len, loff_t off);
832 static ssize_t ext4_quota_write(struct super_block *sb, int type,
833 const char *data, size_t len, loff_t off);
835 static struct dquot_operations ext4_quota_operations = {
836 .initialize = ext4_dquot_initialize,
837 .drop = ext4_dquot_drop,
838 .alloc_space = dquot_alloc_space,
839 .alloc_inode = dquot_alloc_inode,
840 .free_space = dquot_free_space,
841 .free_inode = dquot_free_inode,
842 .transfer = dquot_transfer,
843 .write_dquot = ext4_write_dquot,
844 .acquire_dquot = ext4_acquire_dquot,
845 .release_dquot = ext4_release_dquot,
846 .mark_dirty = ext4_mark_dquot_dirty,
847 .write_info = ext4_write_info
850 static struct quotactl_ops ext4_qctl_operations = {
851 .quota_on = ext4_quota_on,
852 .quota_off = vfs_quota_off,
853 .quota_sync = vfs_quota_sync,
854 .get_info = vfs_get_dqinfo,
855 .set_info = vfs_set_dqinfo,
856 .get_dqblk = vfs_get_dqblk,
857 .set_dqblk = vfs_set_dqblk
861 static const struct super_operations ext4_sops = {
862 .alloc_inode = ext4_alloc_inode,
863 .destroy_inode = ext4_destroy_inode,
864 .write_inode = ext4_write_inode,
865 .dirty_inode = ext4_dirty_inode,
866 .delete_inode = ext4_delete_inode,
867 .put_super = ext4_put_super,
868 .write_super = ext4_write_super,
869 .sync_fs = ext4_sync_fs,
870 .write_super_lockfs = ext4_write_super_lockfs,
871 .unlockfs = ext4_unlockfs,
872 .statfs = ext4_statfs,
873 .remount_fs = ext4_remount,
874 .clear_inode = ext4_clear_inode,
875 .show_options = ext4_show_options,
877 .quota_read = ext4_quota_read,
878 .quota_write = ext4_quota_write,
882 static const struct export_operations ext4_export_ops = {
883 .fh_to_dentry = ext4_fh_to_dentry,
884 .fh_to_parent = ext4_fh_to_parent,
885 .get_parent = ext4_get_parent,
889 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
890 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
891 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
892 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
893 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
894 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
895 Opt_journal_checksum, Opt_journal_async_commit,
896 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
897 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
898 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
899 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
900 Opt_grpquota, Opt_extents, Opt_noextents, Opt_i_version,
901 Opt_mballoc, Opt_nomballoc, Opt_stripe, Opt_delalloc,
904 static match_table_t tokens = {
905 {Opt_bsd_df, "bsddf"},
906 {Opt_minix_df, "minixdf"},
907 {Opt_grpid, "grpid"},
908 {Opt_grpid, "bsdgroups"},
909 {Opt_nogrpid, "nogrpid"},
910 {Opt_nogrpid, "sysvgroups"},
911 {Opt_resgid, "resgid=%u"},
912 {Opt_resuid, "resuid=%u"},
914 {Opt_err_cont, "errors=continue"},
915 {Opt_err_panic, "errors=panic"},
916 {Opt_err_ro, "errors=remount-ro"},
917 {Opt_nouid32, "nouid32"},
918 {Opt_nocheck, "nocheck"},
919 {Opt_nocheck, "check=none"},
920 {Opt_debug, "debug"},
921 {Opt_oldalloc, "oldalloc"},
922 {Opt_orlov, "orlov"},
923 {Opt_user_xattr, "user_xattr"},
924 {Opt_nouser_xattr, "nouser_xattr"},
926 {Opt_noacl, "noacl"},
927 {Opt_reservation, "reservation"},
928 {Opt_noreservation, "noreservation"},
929 {Opt_noload, "noload"},
932 {Opt_commit, "commit=%u"},
933 {Opt_journal_update, "journal=update"},
934 {Opt_journal_inum, "journal=%u"},
935 {Opt_journal_dev, "journal_dev=%u"},
936 {Opt_journal_checksum, "journal_checksum"},
937 {Opt_journal_async_commit, "journal_async_commit"},
938 {Opt_abort, "abort"},
939 {Opt_data_journal, "data=journal"},
940 {Opt_data_ordered, "data=ordered"},
941 {Opt_data_writeback, "data=writeback"},
942 {Opt_offusrjquota, "usrjquota="},
943 {Opt_usrjquota, "usrjquota=%s"},
944 {Opt_offgrpjquota, "grpjquota="},
945 {Opt_grpjquota, "grpjquota=%s"},
946 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
947 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
948 {Opt_grpquota, "grpquota"},
949 {Opt_noquota, "noquota"},
950 {Opt_quota, "quota"},
951 {Opt_usrquota, "usrquota"},
952 {Opt_barrier, "barrier=%u"},
953 {Opt_extents, "extents"},
954 {Opt_noextents, "noextents"},
955 {Opt_i_version, "i_version"},
956 {Opt_mballoc, "mballoc"},
957 {Opt_nomballoc, "nomballoc"},
958 {Opt_stripe, "stripe=%u"},
959 {Opt_resize, "resize"},
960 {Opt_delalloc, "delalloc"},
964 static ext4_fsblk_t get_sb_block(void **data)
966 ext4_fsblk_t sb_block;
967 char *options = (char *) *data;
969 if (!options || strncmp(options, "sb=", 3) != 0)
970 return 1; /* Default location */
972 /*todo: use simple_strtoll with >32bit ext4 */
973 sb_block = simple_strtoul(options, &options, 0);
974 if (*options && *options != ',') {
975 printk("EXT4-fs: Invalid sb specification: %s\n",
981 *data = (void *) options;
985 static int parse_options (char *options, struct super_block *sb,
986 unsigned int *inum, unsigned long *journal_devnum,
987 ext4_fsblk_t *n_blocks_count, int is_remount)
989 struct ext4_sb_info *sbi = EXT4_SB(sb);
991 substring_t args[MAX_OPT_ARGS];
1002 while ((p = strsep (&options, ",")) != NULL) {
1007 token = match_token(p, tokens, args);
1010 clear_opt (sbi->s_mount_opt, MINIX_DF);
1013 set_opt (sbi->s_mount_opt, MINIX_DF);
1016 set_opt (sbi->s_mount_opt, GRPID);
1019 clear_opt (sbi->s_mount_opt, GRPID);
1022 if (match_int(&args[0], &option))
1024 sbi->s_resuid = option;
1027 if (match_int(&args[0], &option))
1029 sbi->s_resgid = option;
1032 /* handled by get_sb_block() instead of here */
1033 /* *sb_block = match_int(&args[0]); */
1036 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1037 clear_opt (sbi->s_mount_opt, ERRORS_RO);
1038 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
1041 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1042 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1043 set_opt (sbi->s_mount_opt, ERRORS_RO);
1046 clear_opt (sbi->s_mount_opt, ERRORS_RO);
1047 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1048 set_opt (sbi->s_mount_opt, ERRORS_CONT);
1051 set_opt (sbi->s_mount_opt, NO_UID32);
1054 clear_opt (sbi->s_mount_opt, CHECK);
1057 set_opt (sbi->s_mount_opt, DEBUG);
1060 set_opt (sbi->s_mount_opt, OLDALLOC);
1063 clear_opt (sbi->s_mount_opt, OLDALLOC);
1065 #ifdef CONFIG_EXT4DEV_FS_XATTR
1066 case Opt_user_xattr:
1067 set_opt (sbi->s_mount_opt, XATTR_USER);
1069 case Opt_nouser_xattr:
1070 clear_opt (sbi->s_mount_opt, XATTR_USER);
1073 case Opt_user_xattr:
1074 case Opt_nouser_xattr:
1075 printk("EXT4 (no)user_xattr options not supported\n");
1078 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
1080 set_opt(sbi->s_mount_opt, POSIX_ACL);
1083 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1088 printk("EXT4 (no)acl options not supported\n");
1091 case Opt_reservation:
1092 set_opt(sbi->s_mount_opt, RESERVATION);
1094 case Opt_noreservation:
1095 clear_opt(sbi->s_mount_opt, RESERVATION);
1097 case Opt_journal_update:
1099 /* Eventually we will want to be able to create
1100 a journal file here. For now, only allow the
1101 user to specify an existing inode to be the
1104 printk(KERN_ERR "EXT4-fs: cannot specify "
1105 "journal on remount\n");
1108 set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
1110 case Opt_journal_inum:
1112 printk(KERN_ERR "EXT4-fs: cannot specify "
1113 "journal on remount\n");
1116 if (match_int(&args[0], &option))
1120 case Opt_journal_dev:
1122 printk(KERN_ERR "EXT4-fs: cannot specify "
1123 "journal on remount\n");
1126 if (match_int(&args[0], &option))
1128 *journal_devnum = option;
1130 case Opt_journal_checksum:
1131 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1133 case Opt_journal_async_commit:
1134 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1135 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1138 set_opt (sbi->s_mount_opt, NOLOAD);
1141 if (match_int(&args[0], &option))
1146 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1147 sbi->s_commit_interval = HZ * option;
1149 case Opt_data_journal:
1150 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1152 case Opt_data_ordered:
1153 data_opt = EXT4_MOUNT_ORDERED_DATA;
1155 case Opt_data_writeback:
1156 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1159 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
1162 "EXT4-fs: cannot change data "
1163 "mode on remount\n");
1167 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
1168 sbi->s_mount_opt |= data_opt;
1178 if ((sb_any_quota_enabled(sb) ||
1179 sb_any_quota_suspended(sb)) &&
1180 !sbi->s_qf_names[qtype]) {
1182 "EXT4-fs: Cannot change journaled "
1183 "quota options when quota turned on.\n");
1186 qname = match_strdup(&args[0]);
1189 "EXT4-fs: not enough memory for "
1190 "storing quotafile name.\n");
1193 if (sbi->s_qf_names[qtype] &&
1194 strcmp(sbi->s_qf_names[qtype], qname)) {
1196 "EXT4-fs: %s quota file already "
1197 "specified.\n", QTYPE2NAME(qtype));
1201 sbi->s_qf_names[qtype] = qname;
1202 if (strchr(sbi->s_qf_names[qtype], '/')) {
1204 "EXT4-fs: quotafile must be on "
1205 "filesystem root.\n");
1206 kfree(sbi->s_qf_names[qtype]);
1207 sbi->s_qf_names[qtype] = NULL;
1210 set_opt(sbi->s_mount_opt, QUOTA);
1212 case Opt_offusrjquota:
1215 case Opt_offgrpjquota:
1218 if ((sb_any_quota_enabled(sb) ||
1219 sb_any_quota_suspended(sb)) &&
1220 sbi->s_qf_names[qtype]) {
1221 printk(KERN_ERR "EXT4-fs: Cannot change "
1222 "journaled quota options when "
1223 "quota turned on.\n");
1227 * The space will be released later when all options
1228 * are confirmed to be correct
1230 sbi->s_qf_names[qtype] = NULL;
1232 case Opt_jqfmt_vfsold:
1233 qfmt = QFMT_VFS_OLD;
1235 case Opt_jqfmt_vfsv0:
1238 if ((sb_any_quota_enabled(sb) ||
1239 sb_any_quota_suspended(sb)) &&
1240 sbi->s_jquota_fmt != qfmt) {
1241 printk(KERN_ERR "EXT4-fs: Cannot change "
1242 "journaled quota options when "
1243 "quota turned on.\n");
1246 sbi->s_jquota_fmt = qfmt;
1250 set_opt(sbi->s_mount_opt, QUOTA);
1251 set_opt(sbi->s_mount_opt, USRQUOTA);
1254 set_opt(sbi->s_mount_opt, QUOTA);
1255 set_opt(sbi->s_mount_opt, GRPQUOTA);
1258 if (sb_any_quota_enabled(sb)) {
1259 printk(KERN_ERR "EXT4-fs: Cannot change quota "
1260 "options when quota turned on.\n");
1263 clear_opt(sbi->s_mount_opt, QUOTA);
1264 clear_opt(sbi->s_mount_opt, USRQUOTA);
1265 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1272 "EXT4-fs: quota options not supported.\n");
1276 case Opt_offusrjquota:
1277 case Opt_offgrpjquota:
1278 case Opt_jqfmt_vfsold:
1279 case Opt_jqfmt_vfsv0:
1281 "EXT4-fs: journaled quota options not "
1288 set_opt(sbi->s_mount_opt, ABORT);
1291 if (match_int(&args[0], &option))
1294 set_opt(sbi->s_mount_opt, BARRIER);
1296 clear_opt(sbi->s_mount_opt, BARRIER);
1302 printk("EXT4-fs: resize option only available "
1306 if (match_int(&args[0], &option) != 0)
1308 *n_blocks_count = option;
1311 set_opt(sbi->s_mount_opt, NOBH);
1314 clear_opt(sbi->s_mount_opt, NOBH);
1317 set_opt (sbi->s_mount_opt, EXTENTS);
1320 clear_opt (sbi->s_mount_opt, EXTENTS);
1323 set_opt(sbi->s_mount_opt, I_VERSION);
1324 sb->s_flags |= MS_I_VERSION;
1327 set_opt(sbi->s_mount_opt, MBALLOC);
1330 clear_opt(sbi->s_mount_opt, MBALLOC);
1333 if (match_int(&args[0], &option))
1337 sbi->s_stripe = option;
1340 set_opt(sbi->s_mount_opt, DELALLOC);
1344 "EXT4-fs: Unrecognized mount option \"%s\" "
1345 "or missing value\n", p);
1350 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1351 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
1352 sbi->s_qf_names[USRQUOTA])
1353 clear_opt(sbi->s_mount_opt, USRQUOTA);
1355 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
1356 sbi->s_qf_names[GRPQUOTA])
1357 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1359 if ((sbi->s_qf_names[USRQUOTA] &&
1360 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
1361 (sbi->s_qf_names[GRPQUOTA] &&
1362 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1363 printk(KERN_ERR "EXT4-fs: old and new quota "
1364 "format mixing.\n");
1368 if (!sbi->s_jquota_fmt) {
1369 printk(KERN_ERR "EXT4-fs: journaled quota format "
1370 "not specified.\n");
1374 if (sbi->s_jquota_fmt) {
1375 printk(KERN_ERR "EXT4-fs: journaled quota format "
1376 "specified with no journaling "
1385 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1388 struct ext4_sb_info *sbi = EXT4_SB(sb);
1391 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1392 printk (KERN_ERR "EXT4-fs warning: revision level too high, "
1393 "forcing read-only mode\n");
1398 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1399 printk (KERN_WARNING "EXT4-fs warning: mounting unchecked fs, "
1400 "running e2fsck is recommended\n");
1401 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1402 printk (KERN_WARNING
1403 "EXT4-fs warning: mounting fs with errors, "
1404 "running e2fsck is recommended\n");
1405 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1406 le16_to_cpu(es->s_mnt_count) >=
1407 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1408 printk (KERN_WARNING
1409 "EXT4-fs warning: maximal mount count reached, "
1410 "running e2fsck is recommended\n");
1411 else if (le32_to_cpu(es->s_checkinterval) &&
1412 (le32_to_cpu(es->s_lastcheck) +
1413 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1414 printk (KERN_WARNING
1415 "EXT4-fs warning: checktime reached, "
1416 "running e2fsck is recommended\n");
1418 /* @@@ We _will_ want to clear the valid bit if we find
1419 * inconsistencies, to force a fsck at reboot. But for
1420 * a plain journaled filesystem we can keep it set as
1423 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1425 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1426 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1427 le16_add_cpu(&es->s_mnt_count, 1);
1428 es->s_mtime = cpu_to_le32(get_seconds());
1429 ext4_update_dynamic_rev(sb);
1430 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1432 ext4_commit_super(sb, es, 1);
1433 if (test_opt(sb, DEBUG))
1434 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%lu, "
1435 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1437 sbi->s_groups_count,
1438 EXT4_BLOCKS_PER_GROUP(sb),
1439 EXT4_INODES_PER_GROUP(sb),
1442 printk(KERN_INFO "EXT4 FS on %s, ", sb->s_id);
1443 if (EXT4_SB(sb)->s_journal->j_inode == NULL) {
1444 char b[BDEVNAME_SIZE];
1446 printk("external journal on %s\n",
1447 bdevname(EXT4_SB(sb)->s_journal->j_dev, b));
1449 printk("internal journal\n");
1454 static int ext4_fill_flex_info(struct super_block *sb)
1456 struct ext4_sb_info *sbi = EXT4_SB(sb);
1457 struct ext4_group_desc *gdp = NULL;
1458 struct buffer_head *bh;
1459 ext4_group_t flex_group_count;
1460 ext4_group_t flex_group;
1461 int groups_per_flex = 0;
1462 __u64 block_bitmap = 0;
1465 if (!sbi->s_es->s_log_groups_per_flex) {
1466 sbi->s_log_groups_per_flex = 0;
1470 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1471 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1473 flex_group_count = (sbi->s_groups_count + groups_per_flex - 1) /
1475 sbi->s_flex_groups = kmalloc(flex_group_count *
1476 sizeof(struct flex_groups), GFP_KERNEL);
1477 if (sbi->s_flex_groups == NULL) {
1478 printk(KERN_ERR "EXT4-fs: not enough memory\n");
1481 memset(sbi->s_flex_groups, 0, flex_group_count *
1482 sizeof(struct flex_groups));
1484 gdp = ext4_get_group_desc(sb, 1, &bh);
1485 block_bitmap = ext4_block_bitmap(sb, gdp) - 1;
1487 for (i = 0; i < sbi->s_groups_count; i++) {
1488 gdp = ext4_get_group_desc(sb, i, &bh);
1490 flex_group = ext4_flex_group(sbi, i);
1491 sbi->s_flex_groups[flex_group].free_inodes +=
1492 le16_to_cpu(gdp->bg_free_inodes_count);
1493 sbi->s_flex_groups[flex_group].free_blocks +=
1494 le16_to_cpu(gdp->bg_free_blocks_count);
1502 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1503 struct ext4_group_desc *gdp)
1507 if (sbi->s_es->s_feature_ro_compat &
1508 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1509 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1510 __le32 le_group = cpu_to_le32(block_group);
1512 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1513 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1514 crc = crc16(crc, (__u8 *)gdp, offset);
1515 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1516 /* for checksum of struct ext4_group_desc do the rest...*/
1517 if ((sbi->s_es->s_feature_incompat &
1518 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1519 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1520 crc = crc16(crc, (__u8 *)gdp + offset,
1521 le16_to_cpu(sbi->s_es->s_desc_size) -
1525 return cpu_to_le16(crc);
1528 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1529 struct ext4_group_desc *gdp)
1531 if ((sbi->s_es->s_feature_ro_compat &
1532 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1533 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1539 /* Called at mount-time, super-block is locked */
1540 static int ext4_check_descriptors(struct super_block *sb)
1542 struct ext4_sb_info *sbi = EXT4_SB(sb);
1543 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1544 ext4_fsblk_t last_block;
1545 ext4_fsblk_t block_bitmap;
1546 ext4_fsblk_t inode_bitmap;
1547 ext4_fsblk_t inode_table;
1548 int flexbg_flag = 0;
1551 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1554 ext4_debug ("Checking group descriptors");
1556 for (i = 0; i < sbi->s_groups_count; i++) {
1557 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1559 if (i == sbi->s_groups_count - 1 || flexbg_flag)
1560 last_block = ext4_blocks_count(sbi->s_es) - 1;
1562 last_block = first_block +
1563 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1565 block_bitmap = ext4_block_bitmap(sb, gdp);
1566 if (block_bitmap < first_block || block_bitmap > last_block)
1568 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1569 "Block bitmap for group %lu not in group "
1570 "(block %llu)!", i, block_bitmap);
1573 inode_bitmap = ext4_inode_bitmap(sb, gdp);
1574 if (inode_bitmap < first_block || inode_bitmap > last_block)
1576 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1577 "Inode bitmap for group %lu not in group "
1578 "(block %llu)!", i, inode_bitmap);
1581 inode_table = ext4_inode_table(sb, gdp);
1582 if (inode_table < first_block ||
1583 inode_table + sbi->s_itb_per_group - 1 > last_block)
1585 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1586 "Inode table for group %lu not in group "
1587 "(block %llu)!", i, inode_table);
1590 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1591 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1592 "Checksum for group %lu failed (%u!=%u)\n",
1593 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1594 gdp)), le16_to_cpu(gdp->bg_checksum));
1598 first_block += EXT4_BLOCKS_PER_GROUP(sb);
1601 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1602 sbi->s_es->s_free_inodes_count=cpu_to_le32(ext4_count_free_inodes(sb));
1606 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1607 * the superblock) which were deleted from all directories, but held open by
1608 * a process at the time of a crash. We walk the list and try to delete these
1609 * inodes at recovery time (only with a read-write filesystem).
1611 * In order to keep the orphan inode chain consistent during traversal (in
1612 * case of crash during recovery), we link each inode into the superblock
1613 * orphan list_head and handle it the same way as an inode deletion during
1614 * normal operation (which journals the operations for us).
1616 * We only do an iget() and an iput() on each inode, which is very safe if we
1617 * accidentally point at an in-use or already deleted inode. The worst that
1618 * can happen in this case is that we get a "bit already cleared" message from
1619 * ext4_free_inode(). The only reason we would point at a wrong inode is if
1620 * e2fsck was run on this filesystem, and it must have already done the orphan
1621 * inode cleanup for us, so we can safely abort without any further action.
1623 static void ext4_orphan_cleanup (struct super_block * sb,
1624 struct ext4_super_block * es)
1626 unsigned int s_flags = sb->s_flags;
1627 int nr_orphans = 0, nr_truncates = 0;
1631 if (!es->s_last_orphan) {
1632 jbd_debug(4, "no orphan inodes to clean up\n");
1636 if (bdev_read_only(sb->s_bdev)) {
1637 printk(KERN_ERR "EXT4-fs: write access "
1638 "unavailable, skipping orphan cleanup.\n");
1642 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1643 if (es->s_last_orphan)
1644 jbd_debug(1, "Errors on filesystem, "
1645 "clearing orphan list.\n");
1646 es->s_last_orphan = 0;
1647 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1651 if (s_flags & MS_RDONLY) {
1652 printk(KERN_INFO "EXT4-fs: %s: orphan cleanup on readonly fs\n",
1654 sb->s_flags &= ~MS_RDONLY;
1657 /* Needed for iput() to work correctly and not trash data */
1658 sb->s_flags |= MS_ACTIVE;
1659 /* Turn on quotas so that they are updated correctly */
1660 for (i = 0; i < MAXQUOTAS; i++) {
1661 if (EXT4_SB(sb)->s_qf_names[i]) {
1662 int ret = ext4_quota_on_mount(sb, i);
1665 "EXT4-fs: Cannot turn on journaled "
1666 "quota: error %d\n", ret);
1671 while (es->s_last_orphan) {
1672 struct inode *inode;
1674 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1675 if (IS_ERR(inode)) {
1676 es->s_last_orphan = 0;
1680 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
1682 if (inode->i_nlink) {
1684 "%s: truncating inode %lu to %Ld bytes\n",
1685 __func__, inode->i_ino, inode->i_size);
1686 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1687 inode->i_ino, inode->i_size);
1688 ext4_truncate(inode);
1692 "%s: deleting unreferenced inode %lu\n",
1693 __func__, inode->i_ino);
1694 jbd_debug(2, "deleting unreferenced inode %lu\n",
1698 iput(inode); /* The delete magic happens here! */
1701 #define PLURAL(x) (x), ((x)==1) ? "" : "s"
1704 printk(KERN_INFO "EXT4-fs: %s: %d orphan inode%s deleted\n",
1705 sb->s_id, PLURAL(nr_orphans));
1707 printk(KERN_INFO "EXT4-fs: %s: %d truncate%s cleaned up\n",
1708 sb->s_id, PLURAL(nr_truncates));
1710 /* Turn quotas off */
1711 for (i = 0; i < MAXQUOTAS; i++) {
1712 if (sb_dqopt(sb)->files[i])
1713 vfs_quota_off(sb, i, 0);
1716 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1719 * Maximal extent format file size.
1720 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1721 * extent format containers, within a sector_t, and within i_blocks
1722 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1723 * so that won't be a limiting factor.
1725 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1727 static loff_t ext4_max_size(int blkbits)
1730 loff_t upper_limit = MAX_LFS_FILESIZE;
1732 /* small i_blocks in vfs inode? */
1733 if (sizeof(blkcnt_t) < sizeof(u64)) {
1735 * CONFIG_LSF is not enabled implies the inode
1736 * i_block represent total blocks in 512 bytes
1737 * 32 == size of vfs inode i_blocks * 8
1739 upper_limit = (1LL << 32) - 1;
1741 /* total blocks in file system block size */
1742 upper_limit >>= (blkbits - 9);
1743 upper_limit <<= blkbits;
1746 /* 32-bit extent-start container, ee_block */
1751 /* Sanity check against vm- & vfs- imposed limits */
1752 if (res > upper_limit)
1759 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
1760 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1761 * We need to be 1 filesystem block less than the 2^48 sector limit.
1763 static loff_t ext4_max_bitmap_size(int bits)
1765 loff_t res = EXT4_NDIR_BLOCKS;
1768 /* This is calculated to be the largest file size for a
1769 * dense, bitmapped file such that the total number of
1770 * sectors in the file, including data and all indirect blocks,
1771 * does not exceed 2^48 -1
1772 * __u32 i_blocks_lo and _u16 i_blocks_high representing the
1773 * total number of 512 bytes blocks of the file
1776 if (sizeof(blkcnt_t) < sizeof(u64)) {
1778 * CONFIG_LSF is not enabled implies the inode
1779 * i_block represent total blocks in 512 bytes
1780 * 32 == size of vfs inode i_blocks * 8
1782 upper_limit = (1LL << 32) - 1;
1784 /* total blocks in file system block size */
1785 upper_limit >>= (bits - 9);
1789 * We use 48 bit ext4_inode i_blocks
1790 * With EXT4_HUGE_FILE_FL set the i_blocks
1791 * represent total number of blocks in
1792 * file system block size
1794 upper_limit = (1LL << 48) - 1;
1798 /* indirect blocks */
1800 /* double indirect blocks */
1801 meta_blocks += 1 + (1LL << (bits-2));
1802 /* tripple indirect blocks */
1803 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1805 upper_limit -= meta_blocks;
1806 upper_limit <<= bits;
1808 res += 1LL << (bits-2);
1809 res += 1LL << (2*(bits-2));
1810 res += 1LL << (3*(bits-2));
1812 if (res > upper_limit)
1815 if (res > MAX_LFS_FILESIZE)
1816 res = MAX_LFS_FILESIZE;
1821 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
1822 ext4_fsblk_t logical_sb_block, int nr)
1824 struct ext4_sb_info *sbi = EXT4_SB(sb);
1825 ext4_group_t bg, first_meta_bg;
1828 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1830 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
1832 return logical_sb_block + nr + 1;
1833 bg = sbi->s_desc_per_block * nr;
1834 if (ext4_bg_has_super(sb, bg))
1836 return (has_super + ext4_group_first_block_no(sb, bg));
1840 * ext4_get_stripe_size: Get the stripe size.
1841 * @sbi: In memory super block info
1843 * If we have specified it via mount option, then
1844 * use the mount option value. If the value specified at mount time is
1845 * greater than the blocks per group use the super block value.
1846 * If the super block value is greater than blocks per group return 0.
1847 * Allocator needs it be less than blocks per group.
1850 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
1852 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
1853 unsigned long stripe_width =
1854 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
1856 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
1857 return sbi->s_stripe;
1859 if (stripe_width <= sbi->s_blocks_per_group)
1860 return stripe_width;
1862 if (stride <= sbi->s_blocks_per_group)
1868 static int ext4_fill_super (struct super_block *sb, void *data, int silent)
1869 __releases(kernel_lock)
1870 __acquires(kernel_lock)
1873 struct buffer_head * bh;
1874 struct ext4_super_block *es = NULL;
1875 struct ext4_sb_info *sbi;
1877 ext4_fsblk_t sb_block = get_sb_block(&data);
1878 ext4_fsblk_t logical_sb_block;
1879 unsigned long offset = 0;
1880 unsigned int journal_inum = 0;
1881 unsigned long journal_devnum = 0;
1882 unsigned long def_mount_opts;
1893 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1896 sb->s_fs_info = sbi;
1897 sbi->s_mount_opt = 0;
1898 sbi->s_resuid = EXT4_DEF_RESUID;
1899 sbi->s_resgid = EXT4_DEF_RESGID;
1900 sbi->s_sb_block = sb_block;
1904 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
1906 printk(KERN_ERR "EXT4-fs: unable to set blocksize\n");
1911 * The ext4 superblock will not be buffer aligned for other than 1kB
1912 * block sizes. We need to calculate the offset from buffer start.
1914 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
1915 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
1916 offset = do_div(logical_sb_block, blocksize);
1918 logical_sb_block = sb_block;
1921 if (!(bh = sb_bread(sb, logical_sb_block))) {
1922 printk (KERN_ERR "EXT4-fs: unable to read superblock\n");
1926 * Note: s_es must be initialized as soon as possible because
1927 * some ext4 macro-instructions depend on its value
1929 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
1931 sb->s_magic = le16_to_cpu(es->s_magic);
1932 if (sb->s_magic != EXT4_SUPER_MAGIC)
1935 /* Set defaults before we parse the mount options */
1936 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1937 if (def_mount_opts & EXT4_DEFM_DEBUG)
1938 set_opt(sbi->s_mount_opt, DEBUG);
1939 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
1940 set_opt(sbi->s_mount_opt, GRPID);
1941 if (def_mount_opts & EXT4_DEFM_UID16)
1942 set_opt(sbi->s_mount_opt, NO_UID32);
1943 #ifdef CONFIG_EXT4DEV_FS_XATTR
1944 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
1945 set_opt(sbi->s_mount_opt, XATTR_USER);
1947 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
1948 if (def_mount_opts & EXT4_DEFM_ACL)
1949 set_opt(sbi->s_mount_opt, POSIX_ACL);
1951 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
1952 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
1953 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
1954 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
1955 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
1956 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
1958 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
1959 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1960 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
1961 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1963 set_opt(sbi->s_mount_opt, ERRORS_RO);
1965 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1966 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1968 set_opt(sbi->s_mount_opt, RESERVATION);
1969 set_opt(sbi->s_mount_opt, BARRIER);
1972 * turn on extents feature by default in ext4 filesystem
1973 * User -o noextents to turn it off
1975 set_opt(sbi->s_mount_opt, EXTENTS);
1977 * turn on mballoc feature by default in ext4 filesystem
1978 * User -o nomballoc to turn it off
1980 set_opt(sbi->s_mount_opt, MBALLOC);
1982 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1986 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1987 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1989 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
1990 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
1991 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1992 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
1994 "EXT4-fs warning: feature flags set on rev 0 fs, "
1995 "running e2fsck is recommended\n");
1998 * Since ext4 is still considered development code, we require
1999 * that the TEST_FILESYS flag in s->flags be set.
2001 if (!(le32_to_cpu(es->s_flags) & EXT2_FLAGS_TEST_FILESYS)) {
2002 printk(KERN_WARNING "EXT4-fs: %s: not marked "
2003 "OK to use with test code.\n", sb->s_id);
2008 * Check feature flags regardless of the revision level, since we
2009 * previously didn't change the revision level when setting the flags,
2010 * so there is a chance incompat flags are set on a rev 0 filesystem.
2012 features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
2014 printk(KERN_ERR "EXT4-fs: %s: couldn't mount because of "
2015 "unsupported optional features (%x).\n",
2016 sb->s_id, le32_to_cpu(features));
2019 features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
2020 if (!(sb->s_flags & MS_RDONLY) && features) {
2021 printk(KERN_ERR "EXT4-fs: %s: couldn't mount RDWR because of "
2022 "unsupported optional features (%x).\n",
2023 sb->s_id, le32_to_cpu(features));
2026 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2028 * Large file size enabled file system can only be
2029 * mount if kernel is build with CONFIG_LSF
2031 if (sizeof(root->i_blocks) < sizeof(u64) &&
2032 !(sb->s_flags & MS_RDONLY)) {
2033 printk(KERN_ERR "EXT4-fs: %s: Filesystem with huge "
2034 "files cannot be mounted read-write "
2035 "without CONFIG_LSF.\n", sb->s_id);
2039 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2041 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2042 blocksize > EXT4_MAX_BLOCK_SIZE) {
2044 "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
2045 blocksize, sb->s_id);
2049 if (sb->s_blocksize != blocksize) {
2051 /* Validate the filesystem blocksize */
2052 if (!sb_set_blocksize(sb, blocksize)) {
2053 printk(KERN_ERR "EXT4-fs: bad block size %d.\n",
2059 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2060 offset = do_div(logical_sb_block, blocksize);
2061 bh = sb_bread(sb, logical_sb_block);
2064 "EXT4-fs: Can't read superblock on 2nd try.\n");
2067 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
2069 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2071 "EXT4-fs: Magic mismatch, very weird !\n");
2076 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits);
2077 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits);
2079 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2080 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2081 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
2083 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2084 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
2085 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
2086 (!is_power_of_2(sbi->s_inode_size)) ||
2087 (sbi->s_inode_size > blocksize)) {
2089 "EXT4-fs: unsupported inode size: %d\n",
2093 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2094 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
2096 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2097 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
2098 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
2099 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
2100 !is_power_of_2(sbi->s_desc_size)) {
2102 "EXT4-fs: unsupported descriptor size %lu\n",
2107 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
2108 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
2109 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
2110 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
2112 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
2113 if (sbi->s_inodes_per_block == 0)
2115 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2116 sbi->s_inodes_per_block;
2117 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
2119 sbi->s_mount_state = le16_to_cpu(es->s_state);
2120 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2121 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2122 for (i=0; i < 4; i++)
2123 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2124 sbi->s_def_hash_version = es->s_def_hash_version;
2126 if (sbi->s_blocks_per_group > blocksize * 8) {
2128 "EXT4-fs: #blocks per group too big: %lu\n",
2129 sbi->s_blocks_per_group);
2132 if (sbi->s_inodes_per_group > blocksize * 8) {
2134 "EXT4-fs: #inodes per group too big: %lu\n",
2135 sbi->s_inodes_per_group);
2139 if (ext4_blocks_count(es) >
2140 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
2141 printk(KERN_ERR "EXT4-fs: filesystem on %s:"
2142 " too large to mount safely\n", sb->s_id);
2143 if (sizeof(sector_t) < 8)
2144 printk(KERN_WARNING "EXT4-fs: CONFIG_LBD not "
2149 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2152 /* ensure blocks_count calculation below doesn't sign-extend */
2153 if (ext4_blocks_count(es) + EXT4_BLOCKS_PER_GROUP(sb) <
2154 le32_to_cpu(es->s_first_data_block) + 1) {
2155 printk(KERN_WARNING "EXT4-fs: bad geometry: block count %llu, "
2156 "first data block %u, blocks per group %lu\n",
2157 ext4_blocks_count(es),
2158 le32_to_cpu(es->s_first_data_block),
2159 EXT4_BLOCKS_PER_GROUP(sb));
2162 blocks_count = (ext4_blocks_count(es) -
2163 le32_to_cpu(es->s_first_data_block) +
2164 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2165 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2166 sbi->s_groups_count = blocks_count;
2167 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2168 EXT4_DESC_PER_BLOCK(sb);
2169 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
2171 if (sbi->s_group_desc == NULL) {
2172 printk (KERN_ERR "EXT4-fs: not enough memory\n");
2176 bgl_lock_init(&sbi->s_blockgroup_lock);
2178 for (i = 0; i < db_count; i++) {
2179 block = descriptor_loc(sb, logical_sb_block, i);
2180 sbi->s_group_desc[i] = sb_bread(sb, block);
2181 if (!sbi->s_group_desc[i]) {
2182 printk (KERN_ERR "EXT4-fs: "
2183 "can't read group descriptor %d\n", i);
2188 if (!ext4_check_descriptors (sb)) {
2189 printk(KERN_ERR "EXT4-fs: group descriptors corrupted!\n");
2192 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2193 if (!ext4_fill_flex_info(sb)) {
2195 "EXT4-fs: unable to initialize "
2196 "flex_bg meta info!\n");
2200 sbi->s_gdb_count = db_count;
2201 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2202 spin_lock_init(&sbi->s_next_gen_lock);
2204 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2205 ext4_count_free_blocks(sb));
2207 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2208 ext4_count_free_inodes(sb));
2211 err = percpu_counter_init(&sbi->s_dirs_counter,
2212 ext4_count_dirs(sb));
2215 printk(KERN_ERR "EXT4-fs: insufficient memory\n");
2219 /* per fileystem reservation list head & lock */
2220 spin_lock_init(&sbi->s_rsv_window_lock);
2221 sbi->s_rsv_window_root = RB_ROOT;
2222 /* Add a single, static dummy reservation to the start of the
2223 * reservation window list --- it gives us a placeholder for
2224 * append-at-start-of-list which makes the allocation logic
2225 * _much_ simpler. */
2226 sbi->s_rsv_window_head.rsv_start = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
2227 sbi->s_rsv_window_head.rsv_end = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
2228 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
2229 sbi->s_rsv_window_head.rsv_goal_size = 0;
2230 ext4_rsv_window_add(sb, &sbi->s_rsv_window_head);
2232 sbi->s_stripe = ext4_get_stripe_size(sbi);
2235 * set up enough so that it can read an inode
2237 sb->s_op = &ext4_sops;
2238 sb->s_export_op = &ext4_export_ops;
2239 sb->s_xattr = ext4_xattr_handlers;
2241 sb->s_qcop = &ext4_qctl_operations;
2242 sb->dq_op = &ext4_quota_operations;
2244 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2248 needs_recovery = (es->s_last_orphan != 0 ||
2249 EXT4_HAS_INCOMPAT_FEATURE(sb,
2250 EXT4_FEATURE_INCOMPAT_RECOVER));
2253 * The first inode we look at is the journal inode. Don't try
2254 * root first: it may be modified in the journal!
2256 if (!test_opt(sb, NOLOAD) &&
2257 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2258 if (ext4_load_journal(sb, es, journal_devnum))
2260 if (!(sb->s_flags & MS_RDONLY) &&
2261 EXT4_SB(sb)->s_journal->j_failed_commit) {
2262 printk(KERN_CRIT "EXT4-fs error (device %s): "
2263 "ext4_fill_super: Journal transaction "
2264 "%u is corrupt\n", sb->s_id,
2265 EXT4_SB(sb)->s_journal->j_failed_commit);
2266 if (test_opt (sb, ERRORS_RO)) {
2268 "Mounting filesystem read-only\n");
2269 sb->s_flags |= MS_RDONLY;
2270 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2271 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2273 if (test_opt(sb, ERRORS_PANIC)) {
2274 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2275 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2276 ext4_commit_super(sb, es, 1);
2278 "EXT4-fs (device %s): mount failed\n",
2283 } else if (journal_inum) {
2284 if (ext4_create_journal(sb, es, journal_inum))
2289 "ext4: No journal on filesystem on %s\n",
2294 if (ext4_blocks_count(es) > 0xffffffffULL &&
2295 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2296 JBD2_FEATURE_INCOMPAT_64BIT)) {
2297 printk(KERN_ERR "ext4: Failed to set 64-bit journal feature\n");
2301 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2302 jbd2_journal_set_features(sbi->s_journal,
2303 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2304 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2305 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2306 jbd2_journal_set_features(sbi->s_journal,
2307 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2308 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2309 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2311 jbd2_journal_clear_features(sbi->s_journal,
2312 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2313 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2316 /* We have now updated the journal if required, so we can
2317 * validate the data journaling mode. */
2318 switch (test_opt(sb, DATA_FLAGS)) {
2320 /* No mode set, assume a default based on the journal
2321 * capabilities: ORDERED_DATA if the journal can
2322 * cope, else JOURNAL_DATA
2324 if (jbd2_journal_check_available_features
2325 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
2326 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2328 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2331 case EXT4_MOUNT_ORDERED_DATA:
2332 case EXT4_MOUNT_WRITEBACK_DATA:
2333 if (!jbd2_journal_check_available_features
2334 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
2335 printk(KERN_ERR "EXT4-fs: Journal does not support "
2336 "requested data journaling mode\n");
2343 if (test_opt(sb, NOBH)) {
2344 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2345 printk(KERN_WARNING "EXT4-fs: Ignoring nobh option - "
2346 "its supported only with writeback mode\n");
2347 clear_opt(sbi->s_mount_opt, NOBH);
2351 * The jbd2_journal_load will have done any necessary log recovery,
2352 * so we can safely mount the rest of the filesystem now.
2355 root = ext4_iget(sb, EXT4_ROOT_INO);
2357 printk(KERN_ERR "EXT4-fs: get root inode failed\n");
2358 ret = PTR_ERR(root);
2361 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2363 printk(KERN_ERR "EXT4-fs: corrupt root inode, run e2fsck\n");
2366 sb->s_root = d_alloc_root(root);
2368 printk(KERN_ERR "EXT4-fs: get root dentry failed\n");
2374 ext4_setup_super (sb, es, sb->s_flags & MS_RDONLY);
2376 /* determine the minimum size of new large inodes, if present */
2377 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2378 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2379 EXT4_GOOD_OLD_INODE_SIZE;
2380 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2381 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2382 if (sbi->s_want_extra_isize <
2383 le16_to_cpu(es->s_want_extra_isize))
2384 sbi->s_want_extra_isize =
2385 le16_to_cpu(es->s_want_extra_isize);
2386 if (sbi->s_want_extra_isize <
2387 le16_to_cpu(es->s_min_extra_isize))
2388 sbi->s_want_extra_isize =
2389 le16_to_cpu(es->s_min_extra_isize);
2392 /* Check if enough inode space is available */
2393 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2394 sbi->s_inode_size) {
2395 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2396 EXT4_GOOD_OLD_INODE_SIZE;
2397 printk(KERN_INFO "EXT4-fs: required extra inode space not"
2402 * akpm: core read_super() calls in here with the superblock locked.
2403 * That deadlocks, because orphan cleanup needs to lock the superblock
2404 * in numerous places. Here we just pop the lock - it's relatively
2405 * harmless, because we are now ready to accept write_super() requests,
2406 * and aviro says that's the only reason for hanging onto the
2409 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2410 ext4_orphan_cleanup(sb, es);
2411 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
2413 printk (KERN_INFO "EXT4-fs: recovery complete.\n");
2414 ext4_mark_recovery_complete(sb, es);
2415 printk (KERN_INFO "EXT4-fs: mounted filesystem with %s data mode.\n",
2416 test_opt(sb,DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA ? "journal":
2417 test_opt(sb,DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA ? "ordered":
2421 ext4_mb_init(sb, needs_recovery);
2428 printk(KERN_ERR "VFS: Can't find ext4 filesystem on dev %s.\n",
2433 jbd2_journal_destroy(sbi->s_journal);
2434 sbi->s_journal = NULL;
2436 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2437 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2438 percpu_counter_destroy(&sbi->s_dirs_counter);
2440 for (i = 0; i < db_count; i++)
2441 brelse(sbi->s_group_desc[i]);
2442 kfree(sbi->s_group_desc);
2445 for (i = 0; i < MAXQUOTAS; i++)
2446 kfree(sbi->s_qf_names[i]);
2448 ext4_blkdev_remove(sbi);
2451 sb->s_fs_info = NULL;
2458 * Setup any per-fs journal parameters now. We'll do this both on
2459 * initial mount, once the journal has been initialised but before we've
2460 * done any recovery; and again on any subsequent remount.
2462 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
2464 struct ext4_sb_info *sbi = EXT4_SB(sb);
2466 if (sbi->s_commit_interval)
2467 journal->j_commit_interval = sbi->s_commit_interval;
2468 /* We could also set up an ext4-specific default for the commit
2469 * interval here, but for now we'll just fall back to the jbd
2472 spin_lock(&journal->j_state_lock);
2473 if (test_opt(sb, BARRIER))
2474 journal->j_flags |= JBD2_BARRIER;
2476 journal->j_flags &= ~JBD2_BARRIER;
2477 spin_unlock(&journal->j_state_lock);
2480 static journal_t *ext4_get_journal(struct super_block *sb,
2481 unsigned int journal_inum)
2483 struct inode *journal_inode;
2486 /* First, test for the existence of a valid inode on disk. Bad
2487 * things happen if we iget() an unused inode, as the subsequent
2488 * iput() will try to delete it. */
2490 journal_inode = ext4_iget(sb, journal_inum);
2491 if (IS_ERR(journal_inode)) {
2492 printk(KERN_ERR "EXT4-fs: no journal found.\n");
2495 if (!journal_inode->i_nlink) {
2496 make_bad_inode(journal_inode);
2497 iput(journal_inode);
2498 printk(KERN_ERR "EXT4-fs: journal inode is deleted.\n");
2502 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
2503 journal_inode, journal_inode->i_size);
2504 if (!S_ISREG(journal_inode->i_mode)) {
2505 printk(KERN_ERR "EXT4-fs: invalid journal inode.\n");
2506 iput(journal_inode);
2510 journal = jbd2_journal_init_inode(journal_inode);
2512 printk(KERN_ERR "EXT4-fs: Could not load journal inode\n");
2513 iput(journal_inode);
2516 journal->j_private = sb;
2517 ext4_init_journal_params(sb, journal);
2521 static journal_t *ext4_get_dev_journal(struct super_block *sb,
2524 struct buffer_head * bh;
2528 int hblock, blocksize;
2529 ext4_fsblk_t sb_block;
2530 unsigned long offset;
2531 struct ext4_super_block * es;
2532 struct block_device *bdev;
2534 bdev = ext4_blkdev_get(j_dev);
2538 if (bd_claim(bdev, sb)) {
2540 "EXT4: failed to claim external journal device.\n");
2545 blocksize = sb->s_blocksize;
2546 hblock = bdev_hardsect_size(bdev);
2547 if (blocksize < hblock) {
2549 "EXT4-fs: blocksize too small for journal device.\n");
2553 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
2554 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
2555 set_blocksize(bdev, blocksize);
2556 if (!(bh = __bread(bdev, sb_block, blocksize))) {
2557 printk(KERN_ERR "EXT4-fs: couldn't read superblock of "
2558 "external journal\n");
2562 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2563 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
2564 !(le32_to_cpu(es->s_feature_incompat) &
2565 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2566 printk(KERN_ERR "EXT4-fs: external journal has "
2567 "bad superblock\n");
2572 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2573 printk(KERN_ERR "EXT4-fs: journal UUID does not match\n");
2578 len = ext4_blocks_count(es);
2579 start = sb_block + 1;
2580 brelse(bh); /* we're done with the superblock */
2582 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
2583 start, len, blocksize);
2585 printk(KERN_ERR "EXT4-fs: failed to create device journal\n");
2588 journal->j_private = sb;
2589 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2590 wait_on_buffer(journal->j_sb_buffer);
2591 if (!buffer_uptodate(journal->j_sb_buffer)) {
2592 printk(KERN_ERR "EXT4-fs: I/O error on journal device\n");
2595 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
2596 printk(KERN_ERR "EXT4-fs: External journal has more than one "
2597 "user (unsupported) - %d\n",
2598 be32_to_cpu(journal->j_superblock->s_nr_users));
2601 EXT4_SB(sb)->journal_bdev = bdev;
2602 ext4_init_journal_params(sb, journal);
2605 jbd2_journal_destroy(journal);
2607 ext4_blkdev_put(bdev);
2611 static int ext4_load_journal(struct super_block *sb,
2612 struct ext4_super_block *es,
2613 unsigned long journal_devnum)
2616 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2619 int really_read_only;
2621 if (journal_devnum &&
2622 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2623 printk(KERN_INFO "EXT4-fs: external journal device major/minor "
2624 "numbers have changed\n");
2625 journal_dev = new_decode_dev(journal_devnum);
2627 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2629 really_read_only = bdev_read_only(sb->s_bdev);
2632 * Are we loading a blank journal or performing recovery after a
2633 * crash? For recovery, we need to check in advance whether we
2634 * can get read-write access to the device.
2637 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2638 if (sb->s_flags & MS_RDONLY) {
2639 printk(KERN_INFO "EXT4-fs: INFO: recovery "
2640 "required on readonly filesystem.\n");
2641 if (really_read_only) {
2642 printk(KERN_ERR "EXT4-fs: write access "
2643 "unavailable, cannot proceed.\n");
2646 printk (KERN_INFO "EXT4-fs: write access will "
2647 "be enabled during recovery.\n");
2651 if (journal_inum && journal_dev) {
2652 printk(KERN_ERR "EXT4-fs: filesystem has both journal "
2653 "and inode journals!\n");
2658 if (!(journal = ext4_get_journal(sb, journal_inum)))
2661 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
2665 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2666 err = jbd2_journal_update_format(journal);
2668 printk(KERN_ERR "EXT4-fs: error updating journal.\n");
2669 jbd2_journal_destroy(journal);
2674 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
2675 err = jbd2_journal_wipe(journal, !really_read_only);
2677 err = jbd2_journal_load(journal);
2680 printk(KERN_ERR "EXT4-fs: error loading journal.\n");
2681 jbd2_journal_destroy(journal);
2685 EXT4_SB(sb)->s_journal = journal;
2686 ext4_clear_journal_err(sb, es);
2688 if (journal_devnum &&
2689 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2690 es->s_journal_dev = cpu_to_le32(journal_devnum);
2693 /* Make sure we flush the recovery flag to disk. */
2694 ext4_commit_super(sb, es, 1);
2700 static int ext4_create_journal(struct super_block * sb,
2701 struct ext4_super_block * es,
2702 unsigned int journal_inum)
2707 if (sb->s_flags & MS_RDONLY) {
2708 printk(KERN_ERR "EXT4-fs: readonly filesystem when trying to "
2709 "create journal.\n");
2713 journal = ext4_get_journal(sb, journal_inum);
2717 printk(KERN_INFO "EXT4-fs: creating new journal on inode %u\n",
2720 err = jbd2_journal_create(journal);
2722 printk(KERN_ERR "EXT4-fs: error creating journal.\n");
2723 jbd2_journal_destroy(journal);
2727 EXT4_SB(sb)->s_journal = journal;
2729 ext4_update_dynamic_rev(sb);
2730 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2731 EXT4_SET_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL);
2733 es->s_journal_inum = cpu_to_le32(journal_inum);
2736 /* Make sure we flush the recovery flag to disk. */
2737 ext4_commit_super(sb, es, 1);
2742 static void ext4_commit_super (struct super_block * sb,
2743 struct ext4_super_block * es,
2746 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
2750 es->s_wtime = cpu_to_le32(get_seconds());
2751 ext4_free_blocks_count_set(es, ext4_count_free_blocks(sb));
2752 es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
2753 BUFFER_TRACE(sbh, "marking dirty");
2754 mark_buffer_dirty(sbh);
2756 sync_dirty_buffer(sbh);
2761 * Have we just finished recovery? If so, and if we are mounting (or
2762 * remounting) the filesystem readonly, then we will end up with a
2763 * consistent fs on disk. Record that fact.
2765 static void ext4_mark_recovery_complete(struct super_block * sb,
2766 struct ext4_super_block * es)
2768 journal_t *journal = EXT4_SB(sb)->s_journal;
2770 jbd2_journal_lock_updates(journal);
2771 jbd2_journal_flush(journal);
2773 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
2774 sb->s_flags & MS_RDONLY) {
2775 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2777 ext4_commit_super(sb, es, 1);
2780 jbd2_journal_unlock_updates(journal);
2784 * If we are mounting (or read-write remounting) a filesystem whose journal
2785 * has recorded an error from a previous lifetime, move that error to the
2786 * main filesystem now.
2788 static void ext4_clear_journal_err(struct super_block * sb,
2789 struct ext4_super_block * es)
2795 journal = EXT4_SB(sb)->s_journal;
2798 * Now check for any error status which may have been recorded in the
2799 * journal by a prior ext4_error() or ext4_abort()
2802 j_errno = jbd2_journal_errno(journal);
2806 errstr = ext4_decode_error(sb, j_errno, nbuf);
2807 ext4_warning(sb, __func__, "Filesystem error recorded "
2808 "from previous mount: %s", errstr);
2809 ext4_warning(sb, __func__, "Marking fs in need of "
2810 "filesystem check.");
2812 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2813 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2814 ext4_commit_super (sb, es, 1);
2816 jbd2_journal_clear_err(journal);
2821 * Force the running and committing transactions to commit,
2822 * and wait on the commit.
2824 int ext4_force_commit(struct super_block *sb)
2829 if (sb->s_flags & MS_RDONLY)
2832 journal = EXT4_SB(sb)->s_journal;
2834 ret = ext4_journal_force_commit(journal);
2839 * Ext4 always journals updates to the superblock itself, so we don't
2840 * have to propagate any other updates to the superblock on disk at this
2841 * point. Just start an async writeback to get the buffers on their way
2844 * This implicitly triggers the writebehind on sync().
2847 static void ext4_write_super (struct super_block * sb)
2849 if (mutex_trylock(&sb->s_lock) != 0)
2854 static int ext4_sync_fs(struct super_block *sb, int wait)
2859 if (jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, &target)) {
2861 jbd2_log_wait_commit(EXT4_SB(sb)->s_journal, target);
2867 * LVM calls this function before a (read-only) snapshot is created. This
2868 * gives us a chance to flush the journal completely and mark the fs clean.
2870 static void ext4_write_super_lockfs(struct super_block *sb)
2874 if (!(sb->s_flags & MS_RDONLY)) {
2875 journal_t *journal = EXT4_SB(sb)->s_journal;
2877 /* Now we set up the journal barrier. */
2878 jbd2_journal_lock_updates(journal);
2879 jbd2_journal_flush(journal);
2881 /* Journal blocked and flushed, clear needs_recovery flag. */
2882 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2883 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
2888 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2889 * flag here, even though the filesystem is not technically dirty yet.
2891 static void ext4_unlockfs(struct super_block *sb)
2893 if (!(sb->s_flags & MS_RDONLY)) {
2895 /* Reser the needs_recovery flag before the fs is unlocked. */
2896 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2897 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
2899 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
2903 static int ext4_remount (struct super_block * sb, int * flags, char * data)
2905 struct ext4_super_block * es;
2906 struct ext4_sb_info *sbi = EXT4_SB(sb);
2907 ext4_fsblk_t n_blocks_count = 0;
2908 unsigned long old_sb_flags;
2909 struct ext4_mount_options old_opts;
2915 /* Store the original options */
2916 old_sb_flags = sb->s_flags;
2917 old_opts.s_mount_opt = sbi->s_mount_opt;
2918 old_opts.s_resuid = sbi->s_resuid;
2919 old_opts.s_resgid = sbi->s_resgid;
2920 old_opts.s_commit_interval = sbi->s_commit_interval;
2922 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2923 for (i = 0; i < MAXQUOTAS; i++)
2924 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2928 * Allow the "check" option to be passed as a remount option.
2930 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
2935 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT)
2936 ext4_abort(sb, __func__, "Abort forced by user");
2938 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2939 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2943 ext4_init_journal_params(sb, sbi->s_journal);
2945 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
2946 n_blocks_count > ext4_blocks_count(es)) {
2947 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) {
2952 if (*flags & MS_RDONLY) {
2954 * First of all, the unconditional stuff we have to do
2955 * to disable replay of the journal when we next remount
2957 sb->s_flags |= MS_RDONLY;
2960 * OK, test if we are remounting a valid rw partition
2961 * readonly, and if so set the rdonly flag and then
2962 * mark the partition as valid again.
2964 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
2965 (sbi->s_mount_state & EXT4_VALID_FS))
2966 es->s_state = cpu_to_le16(sbi->s_mount_state);
2969 * We have to unlock super so that we can wait for
2973 ext4_mark_recovery_complete(sb, es);
2977 if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2978 ~EXT4_FEATURE_RO_COMPAT_SUPP))) {
2979 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
2980 "remount RDWR because of unsupported "
2981 "optional features (%x).\n",
2982 sb->s_id, le32_to_cpu(ret));
2988 * If we have an unprocessed orphan list hanging
2989 * around from a previously readonly bdev mount,
2990 * require a full umount/remount for now.
2992 if (es->s_last_orphan) {
2993 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
2994 "remount RDWR because of unprocessed "
2995 "orphan inode list. Please "
2996 "umount/remount instead.\n",
3003 * Mounting a RDONLY partition read-write, so reread
3004 * and store the current valid flag. (It may have
3005 * been changed by e2fsck since we originally mounted
3008 ext4_clear_journal_err(sb, es);
3009 sbi->s_mount_state = le16_to_cpu(es->s_state);
3010 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
3012 if (!ext4_setup_super (sb, es, 0))
3013 sb->s_flags &= ~MS_RDONLY;
3017 /* Release old quota file names */
3018 for (i = 0; i < MAXQUOTAS; i++)
3019 if (old_opts.s_qf_names[i] &&
3020 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3021 kfree(old_opts.s_qf_names[i]);
3025 sb->s_flags = old_sb_flags;
3026 sbi->s_mount_opt = old_opts.s_mount_opt;
3027 sbi->s_resuid = old_opts.s_resuid;
3028 sbi->s_resgid = old_opts.s_resgid;
3029 sbi->s_commit_interval = old_opts.s_commit_interval;
3031 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3032 for (i = 0; i < MAXQUOTAS; i++) {
3033 if (sbi->s_qf_names[i] &&
3034 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3035 kfree(sbi->s_qf_names[i]);
3036 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3042 static int ext4_statfs (struct dentry * dentry, struct kstatfs * buf)
3044 struct super_block *sb = dentry->d_sb;
3045 struct ext4_sb_info *sbi = EXT4_SB(sb);
3046 struct ext4_super_block *es = sbi->s_es;
3049 if (test_opt(sb, MINIX_DF)) {
3050 sbi->s_overhead_last = 0;
3051 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
3052 ext4_group_t ngroups = sbi->s_groups_count, i;
3053 ext4_fsblk_t overhead = 0;
3057 * Compute the overhead (FS structures). This is constant
3058 * for a given filesystem unless the number of block groups
3059 * changes so we cache the previous value until it does.
3063 * All of the blocks before first_data_block are
3066 overhead = le32_to_cpu(es->s_first_data_block);
3069 * Add the overhead attributed to the superblock and
3070 * block group descriptors. If the sparse superblocks
3071 * feature is turned on, then not all groups have this.
3073 for (i = 0; i < ngroups; i++) {
3074 overhead += ext4_bg_has_super(sb, i) +
3075 ext4_bg_num_gdb(sb, i);
3080 * Every block group has an inode bitmap, a block
3081 * bitmap, and an inode table.
3083 overhead += ngroups * (2 + sbi->s_itb_per_group);
3084 sbi->s_overhead_last = overhead;
3086 sbi->s_blocks_last = ext4_blocks_count(es);
3089 buf->f_type = EXT4_SUPER_MAGIC;
3090 buf->f_bsize = sb->s_blocksize;
3091 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
3092 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter);
3093 ext4_free_blocks_count_set(es, buf->f_bfree);
3094 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3095 if (buf->f_bfree < ext4_r_blocks_count(es))
3097 buf->f_files = le32_to_cpu(es->s_inodes_count);
3098 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
3099 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
3100 buf->f_namelen = EXT4_NAME_LEN;
3101 fsid = le64_to_cpup((void *)es->s_uuid) ^
3102 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3103 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3104 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
3108 /* Helper function for writing quotas on sync - we need to start transaction before quota file
3109 * is locked for write. Otherwise the are possible deadlocks:
3110 * Process 1 Process 2
3111 * ext4_create() quota_sync()
3112 * jbd2_journal_start() write_dquot()
3113 * DQUOT_INIT() down(dqio_mutex)
3114 * down(dqio_mutex) jbd2_journal_start()
3120 static inline struct inode *dquot_to_inode(struct dquot *dquot)
3122 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3125 static int ext4_dquot_initialize(struct inode *inode, int type)
3130 /* We may create quota structure so we need to reserve enough blocks */
3131 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_INIT_BLOCKS(inode->i_sb));
3133 return PTR_ERR(handle);
3134 ret = dquot_initialize(inode, type);
3135 err = ext4_journal_stop(handle);
3141 static int ext4_dquot_drop(struct inode *inode)
3146 /* We may delete quota structure so we need to reserve enough blocks */
3147 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_DEL_BLOCKS(inode->i_sb));
3148 if (IS_ERR(handle)) {
3150 * We call dquot_drop() anyway to at least release references
3151 * to quota structures so that umount does not hang.
3154 return PTR_ERR(handle);
3156 ret = dquot_drop(inode);
3157 err = ext4_journal_stop(handle);
3163 static int ext4_write_dquot(struct dquot *dquot)
3167 struct inode *inode;
3169 inode = dquot_to_inode(dquot);
3170 handle = ext4_journal_start(inode,
3171 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
3173 return PTR_ERR(handle);
3174 ret = dquot_commit(dquot);
3175 err = ext4_journal_stop(handle);
3181 static int ext4_acquire_dquot(struct dquot *dquot)
3186 handle = ext4_journal_start(dquot_to_inode(dquot),
3187 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
3189 return PTR_ERR(handle);
3190 ret = dquot_acquire(dquot);
3191 err = ext4_journal_stop(handle);
3197 static int ext4_release_dquot(struct dquot *dquot)
3202 handle = ext4_journal_start(dquot_to_inode(dquot),
3203 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
3204 if (IS_ERR(handle)) {
3205 /* Release dquot anyway to avoid endless cycle in dqput() */
3206 dquot_release(dquot);
3207 return PTR_ERR(handle);
3209 ret = dquot_release(dquot);
3210 err = ext4_journal_stop(handle);
3216 static int ext4_mark_dquot_dirty(struct dquot *dquot)
3218 /* Are we journaling quotas? */
3219 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3220 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
3221 dquot_mark_dquot_dirty(dquot);
3222 return ext4_write_dquot(dquot);
3224 return dquot_mark_dquot_dirty(dquot);
3228 static int ext4_write_info(struct super_block *sb, int type)
3233 /* Data block + inode block */
3234 handle = ext4_journal_start(sb->s_root->d_inode, 2);
3236 return PTR_ERR(handle);
3237 ret = dquot_commit_info(sb, type);
3238 err = ext4_journal_stop(handle);
3245 * Turn on quotas during mount time - we need to find
3246 * the quota file and such...
3248 static int ext4_quota_on_mount(struct super_block *sb, int type)
3250 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3251 EXT4_SB(sb)->s_jquota_fmt, type);
3255 * Standard function to be called on quota_on
3257 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
3258 char *path, int remount)
3261 struct nameidata nd;
3263 if (!test_opt(sb, QUOTA))
3265 /* When remounting, no checks are needed and in fact, path is NULL */
3267 return vfs_quota_on(sb, type, format_id, path, remount);
3269 err = path_lookup(path, LOOKUP_FOLLOW, &nd);
3273 /* Quotafile not on the same filesystem? */
3274 if (nd.path.mnt->mnt_sb != sb) {
3278 /* Journaling quota? */
3279 if (EXT4_SB(sb)->s_qf_names[type]) {
3280 /* Quotafile not of fs root? */
3281 if (nd.path.dentry->d_parent->d_inode != sb->s_root->d_inode)
3283 "EXT4-fs: Quota file not on filesystem root. "
3284 "Journaled quota will not work.\n");
3288 * When we journal data on quota file, we have to flush journal to see
3289 * all updates to the file when we bypass pagecache...
3291 if (ext4_should_journal_data(nd.path.dentry->d_inode)) {
3293 * We don't need to lock updates but journal_flush() could
3294 * otherwise be livelocked...
3296 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
3297 jbd2_journal_flush(EXT4_SB(sb)->s_journal);
3298 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3302 return vfs_quota_on(sb, type, format_id, path, remount);
3305 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3306 * acquiring the locks... As quota files are never truncated and quota code
3307 * itself serializes the operations (and noone else should touch the files)
3308 * we don't have to be afraid of races */
3309 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
3310 size_t len, loff_t off)
3312 struct inode *inode = sb_dqopt(sb)->files[type];
3313 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3315 int offset = off & (sb->s_blocksize - 1);
3318 struct buffer_head *bh;
3319 loff_t i_size = i_size_read(inode);
3323 if (off+len > i_size)
3326 while (toread > 0) {
3327 tocopy = sb->s_blocksize - offset < toread ?
3328 sb->s_blocksize - offset : toread;
3329 bh = ext4_bread(NULL, inode, blk, 0, &err);
3332 if (!bh) /* A hole? */
3333 memset(data, 0, tocopy);
3335 memcpy(data, bh->b_data+offset, tocopy);
3345 /* Write to quotafile (we know the transaction is already started and has
3346 * enough credits) */
3347 static ssize_t ext4_quota_write(struct super_block *sb, int type,
3348 const char *data, size_t len, loff_t off)
3350 struct inode *inode = sb_dqopt(sb)->files[type];
3351 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3353 int offset = off & (sb->s_blocksize - 1);
3355 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
3356 size_t towrite = len;
3357 struct buffer_head *bh;
3358 handle_t *handle = journal_current_handle();
3361 printk(KERN_WARNING "EXT4-fs: Quota write (off=%Lu, len=%Lu)"
3362 " cancelled because transaction is not started.\n",
3363 (unsigned long long)off, (unsigned long long)len);
3366 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3367 while (towrite > 0) {
3368 tocopy = sb->s_blocksize - offset < towrite ?
3369 sb->s_blocksize - offset : towrite;
3370 bh = ext4_bread(handle, inode, blk, 1, &err);
3373 if (journal_quota) {
3374 err = ext4_journal_get_write_access(handle, bh);
3381 memcpy(bh->b_data+offset, data, tocopy);
3382 flush_dcache_page(bh->b_page);
3385 err = ext4_journal_dirty_metadata(handle, bh);
3387 /* Always do at least ordered writes for quotas */
3388 err = ext4_jbd2_file_inode(handle, inode);
3389 mark_buffer_dirty(bh);
3400 if (len == towrite) {
3401 mutex_unlock(&inode->i_mutex);
3404 if (inode->i_size < off+len-towrite) {
3405 i_size_write(inode, off+len-towrite);
3406 EXT4_I(inode)->i_disksize = inode->i_size;
3408 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3409 ext4_mark_inode_dirty(handle, inode);
3410 mutex_unlock(&inode->i_mutex);
3411 return len - towrite;
3416 static int ext4_get_sb(struct file_system_type *fs_type,
3417 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3419 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
3422 static struct file_system_type ext4dev_fs_type = {
3423 .owner = THIS_MODULE,
3425 .get_sb = ext4_get_sb,
3426 .kill_sb = kill_block_super,
3427 .fs_flags = FS_REQUIRES_DEV,
3430 static int __init init_ext4_fs(void)
3434 err = init_ext4_mballoc();
3438 err = init_ext4_xattr();
3441 err = init_inodecache();
3444 err = register_filesystem(&ext4dev_fs_type);
3449 destroy_inodecache();
3453 exit_ext4_mballoc();
3457 static void __exit exit_ext4_fs(void)
3459 unregister_filesystem(&ext4dev_fs_type);
3460 destroy_inodecache();
3462 exit_ext4_mballoc();
3465 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
3466 MODULE_DESCRIPTION("Fourth Extended Filesystem with extents");
3467 MODULE_LICENSE("GPL");
3468 module_init(init_ext4_fs)
3469 module_exit(exit_ext4_fs)