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 ei->i_reserved_data_blocks = 0;
578 ei->i_reserved_meta_blocks = 0;
579 ei->i_allocated_meta_blocks = 0;
580 ei->i_delalloc_reserved_flag = 0;
581 spin_lock_init(&(ei->i_block_reservation_lock));
582 return &ei->vfs_inode;
585 static void ext4_destroy_inode(struct inode *inode)
587 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
588 printk("EXT4 Inode %p: orphan list check failed!\n",
590 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
591 EXT4_I(inode), sizeof(struct ext4_inode_info),
595 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
598 static void init_once(void *foo)
600 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
602 INIT_LIST_HEAD(&ei->i_orphan);
603 #ifdef CONFIG_EXT4DEV_FS_XATTR
604 init_rwsem(&ei->xattr_sem);
606 init_rwsem(&ei->i_data_sem);
607 inode_init_once(&ei->vfs_inode);
610 static int init_inodecache(void)
612 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
613 sizeof(struct ext4_inode_info),
614 0, (SLAB_RECLAIM_ACCOUNT|
617 if (ext4_inode_cachep == NULL)
622 static void destroy_inodecache(void)
624 kmem_cache_destroy(ext4_inode_cachep);
627 static void ext4_clear_inode(struct inode *inode)
629 struct ext4_block_alloc_info *rsv = EXT4_I(inode)->i_block_alloc_info;
630 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
631 if (EXT4_I(inode)->i_acl &&
632 EXT4_I(inode)->i_acl != EXT4_ACL_NOT_CACHED) {
633 posix_acl_release(EXT4_I(inode)->i_acl);
634 EXT4_I(inode)->i_acl = EXT4_ACL_NOT_CACHED;
636 if (EXT4_I(inode)->i_default_acl &&
637 EXT4_I(inode)->i_default_acl != EXT4_ACL_NOT_CACHED) {
638 posix_acl_release(EXT4_I(inode)->i_default_acl);
639 EXT4_I(inode)->i_default_acl = EXT4_ACL_NOT_CACHED;
642 ext4_discard_reservation(inode);
643 EXT4_I(inode)->i_block_alloc_info = NULL;
646 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
647 &EXT4_I(inode)->jinode);
650 static inline void ext4_show_quota_options(struct seq_file *seq, struct super_block *sb)
652 #if defined(CONFIG_QUOTA)
653 struct ext4_sb_info *sbi = EXT4_SB(sb);
655 if (sbi->s_jquota_fmt)
656 seq_printf(seq, ",jqfmt=%s",
657 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold": "vfsv0");
659 if (sbi->s_qf_names[USRQUOTA])
660 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
662 if (sbi->s_qf_names[GRPQUOTA])
663 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
665 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
666 seq_puts(seq, ",usrquota");
668 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
669 seq_puts(seq, ",grpquota");
675 * - it's set to a non-default value OR
676 * - if the per-sb default is different from the global default
678 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
681 unsigned long def_mount_opts;
682 struct super_block *sb = vfs->mnt_sb;
683 struct ext4_sb_info *sbi = EXT4_SB(sb);
684 struct ext4_super_block *es = sbi->s_es;
686 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
687 def_errors = le16_to_cpu(es->s_errors);
689 if (sbi->s_sb_block != 1)
690 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
691 if (test_opt(sb, MINIX_DF))
692 seq_puts(seq, ",minixdf");
693 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
694 seq_puts(seq, ",grpid");
695 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
696 seq_puts(seq, ",nogrpid");
697 if (sbi->s_resuid != EXT4_DEF_RESUID ||
698 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
699 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
701 if (sbi->s_resgid != EXT4_DEF_RESGID ||
702 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
703 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
705 if (test_opt(sb, ERRORS_RO)) {
706 if (def_errors == EXT4_ERRORS_PANIC ||
707 def_errors == EXT4_ERRORS_CONTINUE) {
708 seq_puts(seq, ",errors=remount-ro");
711 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
712 seq_puts(seq, ",errors=continue");
713 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
714 seq_puts(seq, ",errors=panic");
715 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
716 seq_puts(seq, ",nouid32");
717 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
718 seq_puts(seq, ",debug");
719 if (test_opt(sb, OLDALLOC))
720 seq_puts(seq, ",oldalloc");
721 #ifdef CONFIG_EXT4DEV_FS_XATTR
722 if (test_opt(sb, XATTR_USER) &&
723 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
724 seq_puts(seq, ",user_xattr");
725 if (!test_opt(sb, XATTR_USER) &&
726 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
727 seq_puts(seq, ",nouser_xattr");
730 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
731 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
732 seq_puts(seq, ",acl");
733 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
734 seq_puts(seq, ",noacl");
736 if (!test_opt(sb, RESERVATION))
737 seq_puts(seq, ",noreservation");
738 if (sbi->s_commit_interval) {
739 seq_printf(seq, ",commit=%u",
740 (unsigned) (sbi->s_commit_interval / HZ));
743 * We're changing the default of barrier mount option, so
744 * let's always display its mount state so it's clear what its
747 seq_puts(seq, ",barrier=");
748 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
749 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
750 seq_puts(seq, ",journal_async_commit");
751 if (test_opt(sb, NOBH))
752 seq_puts(seq, ",nobh");
753 if (!test_opt(sb, EXTENTS))
754 seq_puts(seq, ",noextents");
755 if (!test_opt(sb, MBALLOC))
756 seq_puts(seq, ",nomballoc");
757 if (test_opt(sb, I_VERSION))
758 seq_puts(seq, ",i_version");
759 if (!test_opt(sb, DELALLOC))
760 seq_puts(seq, ",nodelalloc");
764 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
766 * journal mode get enabled in different ways
767 * So just print the value even if we didn't specify it
769 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
770 seq_puts(seq, ",data=journal");
771 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
772 seq_puts(seq, ",data=ordered");
773 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
774 seq_puts(seq, ",data=writeback");
776 ext4_show_quota_options(seq, sb);
781 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
782 u64 ino, u32 generation)
786 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
787 return ERR_PTR(-ESTALE);
788 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
789 return ERR_PTR(-ESTALE);
791 /* iget isn't really right if the inode is currently unallocated!!
793 * ext4_read_inode will return a bad_inode if the inode had been
794 * deleted, so we should be safe.
796 * Currently we don't know the generation for parent directory, so
797 * a generation of 0 means "accept any"
799 inode = ext4_iget(sb, ino);
801 return ERR_CAST(inode);
802 if (generation && inode->i_generation != generation) {
804 return ERR_PTR(-ESTALE);
810 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
811 int fh_len, int fh_type)
813 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
817 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
818 int fh_len, int fh_type)
820 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
825 #define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
826 #define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
828 static int ext4_dquot_initialize(struct inode *inode, int type);
829 static int ext4_dquot_drop(struct inode *inode);
830 static int ext4_write_dquot(struct dquot *dquot);
831 static int ext4_acquire_dquot(struct dquot *dquot);
832 static int ext4_release_dquot(struct dquot *dquot);
833 static int ext4_mark_dquot_dirty(struct dquot *dquot);
834 static int ext4_write_info(struct super_block *sb, int type);
835 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
836 char *path, int remount);
837 static int ext4_quota_on_mount(struct super_block *sb, int type);
838 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
839 size_t len, loff_t off);
840 static ssize_t ext4_quota_write(struct super_block *sb, int type,
841 const char *data, size_t len, loff_t off);
843 static struct dquot_operations ext4_quota_operations = {
844 .initialize = ext4_dquot_initialize,
845 .drop = ext4_dquot_drop,
846 .alloc_space = dquot_alloc_space,
847 .alloc_inode = dquot_alloc_inode,
848 .free_space = dquot_free_space,
849 .free_inode = dquot_free_inode,
850 .transfer = dquot_transfer,
851 .write_dquot = ext4_write_dquot,
852 .acquire_dquot = ext4_acquire_dquot,
853 .release_dquot = ext4_release_dquot,
854 .mark_dirty = ext4_mark_dquot_dirty,
855 .write_info = ext4_write_info
858 static struct quotactl_ops ext4_qctl_operations = {
859 .quota_on = ext4_quota_on,
860 .quota_off = vfs_quota_off,
861 .quota_sync = vfs_quota_sync,
862 .get_info = vfs_get_dqinfo,
863 .set_info = vfs_set_dqinfo,
864 .get_dqblk = vfs_get_dqblk,
865 .set_dqblk = vfs_set_dqblk
869 static const struct super_operations ext4_sops = {
870 .alloc_inode = ext4_alloc_inode,
871 .destroy_inode = ext4_destroy_inode,
872 .write_inode = ext4_write_inode,
873 .dirty_inode = ext4_dirty_inode,
874 .delete_inode = ext4_delete_inode,
875 .put_super = ext4_put_super,
876 .write_super = ext4_write_super,
877 .sync_fs = ext4_sync_fs,
878 .write_super_lockfs = ext4_write_super_lockfs,
879 .unlockfs = ext4_unlockfs,
880 .statfs = ext4_statfs,
881 .remount_fs = ext4_remount,
882 .clear_inode = ext4_clear_inode,
883 .show_options = ext4_show_options,
885 .quota_read = ext4_quota_read,
886 .quota_write = ext4_quota_write,
890 static const struct export_operations ext4_export_ops = {
891 .fh_to_dentry = ext4_fh_to_dentry,
892 .fh_to_parent = ext4_fh_to_parent,
893 .get_parent = ext4_get_parent,
897 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
898 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
899 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
900 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
901 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
902 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
903 Opt_journal_checksum, Opt_journal_async_commit,
904 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
905 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
906 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
907 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
908 Opt_grpquota, Opt_extents, Opt_noextents, Opt_i_version,
909 Opt_mballoc, Opt_nomballoc, Opt_stripe, Opt_delalloc, Opt_nodelalloc,
912 static match_table_t tokens = {
913 {Opt_bsd_df, "bsddf"},
914 {Opt_minix_df, "minixdf"},
915 {Opt_grpid, "grpid"},
916 {Opt_grpid, "bsdgroups"},
917 {Opt_nogrpid, "nogrpid"},
918 {Opt_nogrpid, "sysvgroups"},
919 {Opt_resgid, "resgid=%u"},
920 {Opt_resuid, "resuid=%u"},
922 {Opt_err_cont, "errors=continue"},
923 {Opt_err_panic, "errors=panic"},
924 {Opt_err_ro, "errors=remount-ro"},
925 {Opt_nouid32, "nouid32"},
926 {Opt_nocheck, "nocheck"},
927 {Opt_nocheck, "check=none"},
928 {Opt_debug, "debug"},
929 {Opt_oldalloc, "oldalloc"},
930 {Opt_orlov, "orlov"},
931 {Opt_user_xattr, "user_xattr"},
932 {Opt_nouser_xattr, "nouser_xattr"},
934 {Opt_noacl, "noacl"},
935 {Opt_reservation, "reservation"},
936 {Opt_noreservation, "noreservation"},
937 {Opt_noload, "noload"},
940 {Opt_commit, "commit=%u"},
941 {Opt_journal_update, "journal=update"},
942 {Opt_journal_inum, "journal=%u"},
943 {Opt_journal_dev, "journal_dev=%u"},
944 {Opt_journal_checksum, "journal_checksum"},
945 {Opt_journal_async_commit, "journal_async_commit"},
946 {Opt_abort, "abort"},
947 {Opt_data_journal, "data=journal"},
948 {Opt_data_ordered, "data=ordered"},
949 {Opt_data_writeback, "data=writeback"},
950 {Opt_offusrjquota, "usrjquota="},
951 {Opt_usrjquota, "usrjquota=%s"},
952 {Opt_offgrpjquota, "grpjquota="},
953 {Opt_grpjquota, "grpjquota=%s"},
954 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
955 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
956 {Opt_grpquota, "grpquota"},
957 {Opt_noquota, "noquota"},
958 {Opt_quota, "quota"},
959 {Opt_usrquota, "usrquota"},
960 {Opt_barrier, "barrier=%u"},
961 {Opt_extents, "extents"},
962 {Opt_noextents, "noextents"},
963 {Opt_i_version, "i_version"},
964 {Opt_mballoc, "mballoc"},
965 {Opt_nomballoc, "nomballoc"},
966 {Opt_stripe, "stripe=%u"},
967 {Opt_resize, "resize"},
968 {Opt_delalloc, "delalloc"},
969 {Opt_nodelalloc, "nodelalloc"},
973 static ext4_fsblk_t get_sb_block(void **data)
975 ext4_fsblk_t sb_block;
976 char *options = (char *) *data;
978 if (!options || strncmp(options, "sb=", 3) != 0)
979 return 1; /* Default location */
981 /*todo: use simple_strtoll with >32bit ext4 */
982 sb_block = simple_strtoul(options, &options, 0);
983 if (*options && *options != ',') {
984 printk("EXT4-fs: Invalid sb specification: %s\n",
990 *data = (void *) options;
994 static int parse_options (char *options, struct super_block *sb,
995 unsigned int *inum, unsigned long *journal_devnum,
996 ext4_fsblk_t *n_blocks_count, int is_remount)
998 struct ext4_sb_info *sbi = EXT4_SB(sb);
1000 substring_t args[MAX_OPT_ARGS];
1007 ext4_fsblk_t last_block;
1012 while ((p = strsep (&options, ",")) != NULL) {
1017 token = match_token(p, tokens, args);
1020 clear_opt (sbi->s_mount_opt, MINIX_DF);
1023 set_opt (sbi->s_mount_opt, MINIX_DF);
1026 set_opt (sbi->s_mount_opt, GRPID);
1029 clear_opt (sbi->s_mount_opt, GRPID);
1032 if (match_int(&args[0], &option))
1034 sbi->s_resuid = option;
1037 if (match_int(&args[0], &option))
1039 sbi->s_resgid = option;
1042 /* handled by get_sb_block() instead of here */
1043 /* *sb_block = match_int(&args[0]); */
1046 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1047 clear_opt (sbi->s_mount_opt, ERRORS_RO);
1048 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
1051 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1052 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1053 set_opt (sbi->s_mount_opt, ERRORS_RO);
1056 clear_opt (sbi->s_mount_opt, ERRORS_RO);
1057 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1058 set_opt (sbi->s_mount_opt, ERRORS_CONT);
1061 set_opt (sbi->s_mount_opt, NO_UID32);
1064 clear_opt (sbi->s_mount_opt, CHECK);
1067 set_opt (sbi->s_mount_opt, DEBUG);
1070 set_opt (sbi->s_mount_opt, OLDALLOC);
1073 clear_opt (sbi->s_mount_opt, OLDALLOC);
1075 #ifdef CONFIG_EXT4DEV_FS_XATTR
1076 case Opt_user_xattr:
1077 set_opt (sbi->s_mount_opt, XATTR_USER);
1079 case Opt_nouser_xattr:
1080 clear_opt (sbi->s_mount_opt, XATTR_USER);
1083 case Opt_user_xattr:
1084 case Opt_nouser_xattr:
1085 printk("EXT4 (no)user_xattr options not supported\n");
1088 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
1090 set_opt(sbi->s_mount_opt, POSIX_ACL);
1093 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1098 printk("EXT4 (no)acl options not supported\n");
1101 case Opt_reservation:
1102 set_opt(sbi->s_mount_opt, RESERVATION);
1104 case Opt_noreservation:
1105 clear_opt(sbi->s_mount_opt, RESERVATION);
1107 case Opt_journal_update:
1109 /* Eventually we will want to be able to create
1110 a journal file here. For now, only allow the
1111 user to specify an existing inode to be the
1114 printk(KERN_ERR "EXT4-fs: cannot specify "
1115 "journal on remount\n");
1118 set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
1120 case Opt_journal_inum:
1122 printk(KERN_ERR "EXT4-fs: cannot specify "
1123 "journal on remount\n");
1126 if (match_int(&args[0], &option))
1130 case Opt_journal_dev:
1132 printk(KERN_ERR "EXT4-fs: cannot specify "
1133 "journal on remount\n");
1136 if (match_int(&args[0], &option))
1138 *journal_devnum = option;
1140 case Opt_journal_checksum:
1141 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1143 case Opt_journal_async_commit:
1144 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1145 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1148 set_opt (sbi->s_mount_opt, NOLOAD);
1151 if (match_int(&args[0], &option))
1156 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1157 sbi->s_commit_interval = HZ * option;
1159 case Opt_data_journal:
1160 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1162 case Opt_data_ordered:
1163 data_opt = EXT4_MOUNT_ORDERED_DATA;
1165 case Opt_data_writeback:
1166 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1169 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
1172 "EXT4-fs: cannot change data "
1173 "mode on remount\n");
1177 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
1178 sbi->s_mount_opt |= data_opt;
1188 if ((sb_any_quota_enabled(sb) ||
1189 sb_any_quota_suspended(sb)) &&
1190 !sbi->s_qf_names[qtype]) {
1192 "EXT4-fs: Cannot change journaled "
1193 "quota options when quota turned on.\n");
1196 qname = match_strdup(&args[0]);
1199 "EXT4-fs: not enough memory for "
1200 "storing quotafile name.\n");
1203 if (sbi->s_qf_names[qtype] &&
1204 strcmp(sbi->s_qf_names[qtype], qname)) {
1206 "EXT4-fs: %s quota file already "
1207 "specified.\n", QTYPE2NAME(qtype));
1211 sbi->s_qf_names[qtype] = qname;
1212 if (strchr(sbi->s_qf_names[qtype], '/')) {
1214 "EXT4-fs: quotafile must be on "
1215 "filesystem root.\n");
1216 kfree(sbi->s_qf_names[qtype]);
1217 sbi->s_qf_names[qtype] = NULL;
1220 set_opt(sbi->s_mount_opt, QUOTA);
1222 case Opt_offusrjquota:
1225 case Opt_offgrpjquota:
1228 if ((sb_any_quota_enabled(sb) ||
1229 sb_any_quota_suspended(sb)) &&
1230 sbi->s_qf_names[qtype]) {
1231 printk(KERN_ERR "EXT4-fs: Cannot change "
1232 "journaled quota options when "
1233 "quota turned on.\n");
1237 * The space will be released later when all options
1238 * are confirmed to be correct
1240 sbi->s_qf_names[qtype] = NULL;
1242 case Opt_jqfmt_vfsold:
1243 qfmt = QFMT_VFS_OLD;
1245 case Opt_jqfmt_vfsv0:
1248 if ((sb_any_quota_enabled(sb) ||
1249 sb_any_quota_suspended(sb)) &&
1250 sbi->s_jquota_fmt != qfmt) {
1251 printk(KERN_ERR "EXT4-fs: Cannot change "
1252 "journaled quota options when "
1253 "quota turned on.\n");
1256 sbi->s_jquota_fmt = qfmt;
1260 set_opt(sbi->s_mount_opt, QUOTA);
1261 set_opt(sbi->s_mount_opt, USRQUOTA);
1264 set_opt(sbi->s_mount_opt, QUOTA);
1265 set_opt(sbi->s_mount_opt, GRPQUOTA);
1268 if (sb_any_quota_enabled(sb)) {
1269 printk(KERN_ERR "EXT4-fs: Cannot change quota "
1270 "options when quota turned on.\n");
1273 clear_opt(sbi->s_mount_opt, QUOTA);
1274 clear_opt(sbi->s_mount_opt, USRQUOTA);
1275 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1282 "EXT4-fs: quota options not supported.\n");
1286 case Opt_offusrjquota:
1287 case Opt_offgrpjquota:
1288 case Opt_jqfmt_vfsold:
1289 case Opt_jqfmt_vfsv0:
1291 "EXT4-fs: journaled quota options not "
1298 set_opt(sbi->s_mount_opt, ABORT);
1301 if (match_int(&args[0], &option))
1304 set_opt(sbi->s_mount_opt, BARRIER);
1306 clear_opt(sbi->s_mount_opt, BARRIER);
1312 printk("EXT4-fs: resize option only available "
1316 if (match_int(&args[0], &option) != 0)
1318 *n_blocks_count = option;
1321 set_opt(sbi->s_mount_opt, NOBH);
1324 clear_opt(sbi->s_mount_opt, NOBH);
1327 if (!EXT4_HAS_INCOMPAT_FEATURE(sb,
1328 EXT4_FEATURE_INCOMPAT_EXTENTS)) {
1329 ext4_warning(sb, __func__,
1330 "extents feature not enabled "
1331 "on this filesystem, use tune2fs\n");
1334 set_opt (sbi->s_mount_opt, EXTENTS);
1338 * When e2fsprogs support resizing an already existing
1339 * ext3 file system to greater than 2**32 we need to
1340 * add support to block allocator to handle growing
1341 * already existing block mapped inode so that blocks
1342 * allocated for them fall within 2**32
1344 last_block = ext4_blocks_count(sbi->s_es) - 1;
1345 if (last_block > 0xffffffffULL) {
1346 printk(KERN_ERR "EXT4-fs: Filesystem too "
1347 "large to mount with "
1348 "-o noextents options\n");
1351 clear_opt (sbi->s_mount_opt, EXTENTS);
1354 set_opt(sbi->s_mount_opt, I_VERSION);
1355 sb->s_flags |= MS_I_VERSION;
1357 case Opt_nodelalloc:
1358 clear_opt(sbi->s_mount_opt, DELALLOC);
1361 set_opt(sbi->s_mount_opt, MBALLOC);
1364 clear_opt(sbi->s_mount_opt, MBALLOC);
1367 if (match_int(&args[0], &option))
1371 sbi->s_stripe = option;
1374 set_opt(sbi->s_mount_opt, DELALLOC);
1378 "EXT4-fs: Unrecognized mount option \"%s\" "
1379 "or missing value\n", p);
1384 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1385 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
1386 sbi->s_qf_names[USRQUOTA])
1387 clear_opt(sbi->s_mount_opt, USRQUOTA);
1389 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
1390 sbi->s_qf_names[GRPQUOTA])
1391 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1393 if ((sbi->s_qf_names[USRQUOTA] &&
1394 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
1395 (sbi->s_qf_names[GRPQUOTA] &&
1396 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1397 printk(KERN_ERR "EXT4-fs: old and new quota "
1398 "format mixing.\n");
1402 if (!sbi->s_jquota_fmt) {
1403 printk(KERN_ERR "EXT4-fs: journaled quota format "
1404 "not specified.\n");
1408 if (sbi->s_jquota_fmt) {
1409 printk(KERN_ERR "EXT4-fs: journaled quota format "
1410 "specified with no journaling "
1419 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1422 struct ext4_sb_info *sbi = EXT4_SB(sb);
1425 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1426 printk (KERN_ERR "EXT4-fs warning: revision level too high, "
1427 "forcing read-only mode\n");
1432 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1433 printk (KERN_WARNING "EXT4-fs warning: mounting unchecked fs, "
1434 "running e2fsck is recommended\n");
1435 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1436 printk (KERN_WARNING
1437 "EXT4-fs warning: mounting fs with errors, "
1438 "running e2fsck is recommended\n");
1439 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1440 le16_to_cpu(es->s_mnt_count) >=
1441 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1442 printk (KERN_WARNING
1443 "EXT4-fs warning: maximal mount count reached, "
1444 "running e2fsck is recommended\n");
1445 else if (le32_to_cpu(es->s_checkinterval) &&
1446 (le32_to_cpu(es->s_lastcheck) +
1447 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1448 printk (KERN_WARNING
1449 "EXT4-fs warning: checktime reached, "
1450 "running e2fsck is recommended\n");
1452 /* @@@ We _will_ want to clear the valid bit if we find
1453 * inconsistencies, to force a fsck at reboot. But for
1454 * a plain journaled filesystem we can keep it set as
1457 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1459 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1460 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1461 le16_add_cpu(&es->s_mnt_count, 1);
1462 es->s_mtime = cpu_to_le32(get_seconds());
1463 ext4_update_dynamic_rev(sb);
1464 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1466 ext4_commit_super(sb, es, 1);
1467 if (test_opt(sb, DEBUG))
1468 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%lu, "
1469 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1471 sbi->s_groups_count,
1472 EXT4_BLOCKS_PER_GROUP(sb),
1473 EXT4_INODES_PER_GROUP(sb),
1476 printk(KERN_INFO "EXT4 FS on %s, ", sb->s_id);
1477 if (EXT4_SB(sb)->s_journal->j_inode == NULL) {
1478 char b[BDEVNAME_SIZE];
1480 printk("external journal on %s\n",
1481 bdevname(EXT4_SB(sb)->s_journal->j_dev, b));
1483 printk("internal journal\n");
1488 static int ext4_fill_flex_info(struct super_block *sb)
1490 struct ext4_sb_info *sbi = EXT4_SB(sb);
1491 struct ext4_group_desc *gdp = NULL;
1492 struct buffer_head *bh;
1493 ext4_group_t flex_group_count;
1494 ext4_group_t flex_group;
1495 int groups_per_flex = 0;
1496 __u64 block_bitmap = 0;
1499 if (!sbi->s_es->s_log_groups_per_flex) {
1500 sbi->s_log_groups_per_flex = 0;
1504 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1505 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1507 flex_group_count = (sbi->s_groups_count + groups_per_flex - 1) /
1509 sbi->s_flex_groups = kzalloc(flex_group_count *
1510 sizeof(struct flex_groups), GFP_KERNEL);
1511 if (sbi->s_flex_groups == NULL) {
1512 printk(KERN_ERR "EXT4-fs: not enough memory for "
1513 "%lu flex groups\n", flex_group_count);
1517 gdp = ext4_get_group_desc(sb, 1, &bh);
1518 block_bitmap = ext4_block_bitmap(sb, gdp) - 1;
1520 for (i = 0; i < sbi->s_groups_count; i++) {
1521 gdp = ext4_get_group_desc(sb, i, &bh);
1523 flex_group = ext4_flex_group(sbi, i);
1524 sbi->s_flex_groups[flex_group].free_inodes +=
1525 le16_to_cpu(gdp->bg_free_inodes_count);
1526 sbi->s_flex_groups[flex_group].free_blocks +=
1527 le16_to_cpu(gdp->bg_free_blocks_count);
1535 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1536 struct ext4_group_desc *gdp)
1540 if (sbi->s_es->s_feature_ro_compat &
1541 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1542 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1543 __le32 le_group = cpu_to_le32(block_group);
1545 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1546 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1547 crc = crc16(crc, (__u8 *)gdp, offset);
1548 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1549 /* for checksum of struct ext4_group_desc do the rest...*/
1550 if ((sbi->s_es->s_feature_incompat &
1551 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1552 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1553 crc = crc16(crc, (__u8 *)gdp + offset,
1554 le16_to_cpu(sbi->s_es->s_desc_size) -
1558 return cpu_to_le16(crc);
1561 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1562 struct ext4_group_desc *gdp)
1564 if ((sbi->s_es->s_feature_ro_compat &
1565 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1566 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1572 /* Called at mount-time, super-block is locked */
1573 static int ext4_check_descriptors(struct super_block *sb)
1575 struct ext4_sb_info *sbi = EXT4_SB(sb);
1576 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1577 ext4_fsblk_t last_block;
1578 ext4_fsblk_t block_bitmap;
1579 ext4_fsblk_t inode_bitmap;
1580 ext4_fsblk_t inode_table;
1581 int flexbg_flag = 0;
1584 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1587 ext4_debug ("Checking group descriptors");
1589 for (i = 0; i < sbi->s_groups_count; i++) {
1590 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1592 if (i == sbi->s_groups_count - 1 || flexbg_flag)
1593 last_block = ext4_blocks_count(sbi->s_es) - 1;
1595 last_block = first_block +
1596 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1598 block_bitmap = ext4_block_bitmap(sb, gdp);
1599 if (block_bitmap < first_block || block_bitmap > last_block)
1601 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1602 "Block bitmap for group %lu not in group "
1603 "(block %llu)!", i, block_bitmap);
1606 inode_bitmap = ext4_inode_bitmap(sb, gdp);
1607 if (inode_bitmap < first_block || inode_bitmap > last_block)
1609 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1610 "Inode bitmap for group %lu not in group "
1611 "(block %llu)!", i, inode_bitmap);
1614 inode_table = ext4_inode_table(sb, gdp);
1615 if (inode_table < first_block ||
1616 inode_table + sbi->s_itb_per_group - 1 > last_block)
1618 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1619 "Inode table for group %lu not in group "
1620 "(block %llu)!", i, inode_table);
1623 spin_lock(sb_bgl_lock(sbi, i));
1624 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1625 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1626 "Checksum for group %lu failed (%u!=%u)\n",
1627 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1628 gdp)), le16_to_cpu(gdp->bg_checksum));
1629 if (!(sb->s_flags & MS_RDONLY))
1632 spin_unlock(sb_bgl_lock(sbi, i));
1634 first_block += EXT4_BLOCKS_PER_GROUP(sb);
1637 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1638 sbi->s_es->s_free_inodes_count=cpu_to_le32(ext4_count_free_inodes(sb));
1642 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1643 * the superblock) which were deleted from all directories, but held open by
1644 * a process at the time of a crash. We walk the list and try to delete these
1645 * inodes at recovery time (only with a read-write filesystem).
1647 * In order to keep the orphan inode chain consistent during traversal (in
1648 * case of crash during recovery), we link each inode into the superblock
1649 * orphan list_head and handle it the same way as an inode deletion during
1650 * normal operation (which journals the operations for us).
1652 * We only do an iget() and an iput() on each inode, which is very safe if we
1653 * accidentally point at an in-use or already deleted inode. The worst that
1654 * can happen in this case is that we get a "bit already cleared" message from
1655 * ext4_free_inode(). The only reason we would point at a wrong inode is if
1656 * e2fsck was run on this filesystem, and it must have already done the orphan
1657 * inode cleanup for us, so we can safely abort without any further action.
1659 static void ext4_orphan_cleanup (struct super_block * sb,
1660 struct ext4_super_block * es)
1662 unsigned int s_flags = sb->s_flags;
1663 int nr_orphans = 0, nr_truncates = 0;
1667 if (!es->s_last_orphan) {
1668 jbd_debug(4, "no orphan inodes to clean up\n");
1672 if (bdev_read_only(sb->s_bdev)) {
1673 printk(KERN_ERR "EXT4-fs: write access "
1674 "unavailable, skipping orphan cleanup.\n");
1678 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1679 if (es->s_last_orphan)
1680 jbd_debug(1, "Errors on filesystem, "
1681 "clearing orphan list.\n");
1682 es->s_last_orphan = 0;
1683 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1687 if (s_flags & MS_RDONLY) {
1688 printk(KERN_INFO "EXT4-fs: %s: orphan cleanup on readonly fs\n",
1690 sb->s_flags &= ~MS_RDONLY;
1693 /* Needed for iput() to work correctly and not trash data */
1694 sb->s_flags |= MS_ACTIVE;
1695 /* Turn on quotas so that they are updated correctly */
1696 for (i = 0; i < MAXQUOTAS; i++) {
1697 if (EXT4_SB(sb)->s_qf_names[i]) {
1698 int ret = ext4_quota_on_mount(sb, i);
1701 "EXT4-fs: Cannot turn on journaled "
1702 "quota: error %d\n", ret);
1707 while (es->s_last_orphan) {
1708 struct inode *inode;
1710 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1711 if (IS_ERR(inode)) {
1712 es->s_last_orphan = 0;
1716 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
1718 if (inode->i_nlink) {
1720 "%s: truncating inode %lu to %Ld bytes\n",
1721 __func__, inode->i_ino, inode->i_size);
1722 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1723 inode->i_ino, inode->i_size);
1724 ext4_truncate(inode);
1728 "%s: deleting unreferenced inode %lu\n",
1729 __func__, inode->i_ino);
1730 jbd_debug(2, "deleting unreferenced inode %lu\n",
1734 iput(inode); /* The delete magic happens here! */
1737 #define PLURAL(x) (x), ((x)==1) ? "" : "s"
1740 printk(KERN_INFO "EXT4-fs: %s: %d orphan inode%s deleted\n",
1741 sb->s_id, PLURAL(nr_orphans));
1743 printk(KERN_INFO "EXT4-fs: %s: %d truncate%s cleaned up\n",
1744 sb->s_id, PLURAL(nr_truncates));
1746 /* Turn quotas off */
1747 for (i = 0; i < MAXQUOTAS; i++) {
1748 if (sb_dqopt(sb)->files[i])
1749 vfs_quota_off(sb, i, 0);
1752 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1755 * Maximal extent format file size.
1756 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1757 * extent format containers, within a sector_t, and within i_blocks
1758 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1759 * so that won't be a limiting factor.
1761 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1763 static loff_t ext4_max_size(int blkbits)
1766 loff_t upper_limit = MAX_LFS_FILESIZE;
1768 /* small i_blocks in vfs inode? */
1769 if (sizeof(blkcnt_t) < sizeof(u64)) {
1771 * CONFIG_LSF is not enabled implies the inode
1772 * i_block represent total blocks in 512 bytes
1773 * 32 == size of vfs inode i_blocks * 8
1775 upper_limit = (1LL << 32) - 1;
1777 /* total blocks in file system block size */
1778 upper_limit >>= (blkbits - 9);
1779 upper_limit <<= blkbits;
1782 /* 32-bit extent-start container, ee_block */
1787 /* Sanity check against vm- & vfs- imposed limits */
1788 if (res > upper_limit)
1795 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
1796 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1797 * We need to be 1 filesystem block less than the 2^48 sector limit.
1799 static loff_t ext4_max_bitmap_size(int bits)
1801 loff_t res = EXT4_NDIR_BLOCKS;
1804 /* This is calculated to be the largest file size for a
1805 * dense, bitmapped file such that the total number of
1806 * sectors in the file, including data and all indirect blocks,
1807 * does not exceed 2^48 -1
1808 * __u32 i_blocks_lo and _u16 i_blocks_high representing the
1809 * total number of 512 bytes blocks of the file
1812 if (sizeof(blkcnt_t) < sizeof(u64)) {
1814 * CONFIG_LSF is not enabled implies the inode
1815 * i_block represent total blocks in 512 bytes
1816 * 32 == size of vfs inode i_blocks * 8
1818 upper_limit = (1LL << 32) - 1;
1820 /* total blocks in file system block size */
1821 upper_limit >>= (bits - 9);
1825 * We use 48 bit ext4_inode i_blocks
1826 * With EXT4_HUGE_FILE_FL set the i_blocks
1827 * represent total number of blocks in
1828 * file system block size
1830 upper_limit = (1LL << 48) - 1;
1834 /* indirect blocks */
1836 /* double indirect blocks */
1837 meta_blocks += 1 + (1LL << (bits-2));
1838 /* tripple indirect blocks */
1839 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1841 upper_limit -= meta_blocks;
1842 upper_limit <<= bits;
1844 res += 1LL << (bits-2);
1845 res += 1LL << (2*(bits-2));
1846 res += 1LL << (3*(bits-2));
1848 if (res > upper_limit)
1851 if (res > MAX_LFS_FILESIZE)
1852 res = MAX_LFS_FILESIZE;
1857 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
1858 ext4_fsblk_t logical_sb_block, int nr)
1860 struct ext4_sb_info *sbi = EXT4_SB(sb);
1861 ext4_group_t bg, first_meta_bg;
1864 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1866 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
1868 return logical_sb_block + nr + 1;
1869 bg = sbi->s_desc_per_block * nr;
1870 if (ext4_bg_has_super(sb, bg))
1872 return (has_super + ext4_group_first_block_no(sb, bg));
1876 * ext4_get_stripe_size: Get the stripe size.
1877 * @sbi: In memory super block info
1879 * If we have specified it via mount option, then
1880 * use the mount option value. If the value specified at mount time is
1881 * greater than the blocks per group use the super block value.
1882 * If the super block value is greater than blocks per group return 0.
1883 * Allocator needs it be less than blocks per group.
1886 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
1888 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
1889 unsigned long stripe_width =
1890 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
1892 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
1893 return sbi->s_stripe;
1895 if (stripe_width <= sbi->s_blocks_per_group)
1896 return stripe_width;
1898 if (stride <= sbi->s_blocks_per_group)
1904 static int ext4_fill_super (struct super_block *sb, void *data, int silent)
1905 __releases(kernel_lock)
1906 __acquires(kernel_lock)
1909 struct buffer_head * bh;
1910 struct ext4_super_block *es = NULL;
1911 struct ext4_sb_info *sbi;
1913 ext4_fsblk_t sb_block = get_sb_block(&data);
1914 ext4_fsblk_t logical_sb_block;
1915 unsigned long offset = 0;
1916 unsigned int journal_inum = 0;
1917 unsigned long journal_devnum = 0;
1918 unsigned long def_mount_opts;
1929 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1932 sb->s_fs_info = sbi;
1933 sbi->s_mount_opt = 0;
1934 sbi->s_resuid = EXT4_DEF_RESUID;
1935 sbi->s_resgid = EXT4_DEF_RESGID;
1936 sbi->s_sb_block = sb_block;
1940 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
1942 printk(KERN_ERR "EXT4-fs: unable to set blocksize\n");
1947 * The ext4 superblock will not be buffer aligned for other than 1kB
1948 * block sizes. We need to calculate the offset from buffer start.
1950 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
1951 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
1952 offset = do_div(logical_sb_block, blocksize);
1954 logical_sb_block = sb_block;
1957 if (!(bh = sb_bread(sb, logical_sb_block))) {
1958 printk (KERN_ERR "EXT4-fs: unable to read superblock\n");
1962 * Note: s_es must be initialized as soon as possible because
1963 * some ext4 macro-instructions depend on its value
1965 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
1967 sb->s_magic = le16_to_cpu(es->s_magic);
1968 if (sb->s_magic != EXT4_SUPER_MAGIC)
1971 /* Set defaults before we parse the mount options */
1972 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1973 if (def_mount_opts & EXT4_DEFM_DEBUG)
1974 set_opt(sbi->s_mount_opt, DEBUG);
1975 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
1976 set_opt(sbi->s_mount_opt, GRPID);
1977 if (def_mount_opts & EXT4_DEFM_UID16)
1978 set_opt(sbi->s_mount_opt, NO_UID32);
1979 #ifdef CONFIG_EXT4DEV_FS_XATTR
1980 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
1981 set_opt(sbi->s_mount_opt, XATTR_USER);
1983 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
1984 if (def_mount_opts & EXT4_DEFM_ACL)
1985 set_opt(sbi->s_mount_opt, POSIX_ACL);
1987 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
1988 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
1989 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
1990 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
1991 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
1992 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
1994 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
1995 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1996 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
1997 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1999 set_opt(sbi->s_mount_opt, ERRORS_RO);
2001 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2002 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
2004 set_opt(sbi->s_mount_opt, RESERVATION);
2005 set_opt(sbi->s_mount_opt, BARRIER);
2008 * turn on extents feature by default in ext4 filesystem
2009 * only if feature flag already set by mkfs or tune2fs.
2010 * Use -o noextents to turn it off
2012 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS))
2013 set_opt(sbi->s_mount_opt, EXTENTS);
2015 ext4_warning(sb, __func__,
2016 "extents feature not enabled on this filesystem, "
2019 * turn on mballoc code by default in ext4 filesystem
2020 * Use -o nomballoc to turn it off
2022 set_opt(sbi->s_mount_opt, MBALLOC);
2025 * enable delayed allocation by default
2026 * Use -o nodelalloc to turn it off
2028 set_opt(sbi->s_mount_opt, DELALLOC);
2031 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
2035 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2036 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2038 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2039 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2040 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2041 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
2043 "EXT4-fs warning: feature flags set on rev 0 fs, "
2044 "running e2fsck is recommended\n");
2047 * Since ext4 is still considered development code, we require
2048 * that the TEST_FILESYS flag in s->flags be set.
2050 if (!(le32_to_cpu(es->s_flags) & EXT2_FLAGS_TEST_FILESYS)) {
2051 printk(KERN_WARNING "EXT4-fs: %s: not marked "
2052 "OK to use with test code.\n", sb->s_id);
2057 * Check feature flags regardless of the revision level, since we
2058 * previously didn't change the revision level when setting the flags,
2059 * so there is a chance incompat flags are set on a rev 0 filesystem.
2061 features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
2063 printk(KERN_ERR "EXT4-fs: %s: couldn't mount because of "
2064 "unsupported optional features (%x).\n",
2065 sb->s_id, le32_to_cpu(features));
2068 features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
2069 if (!(sb->s_flags & MS_RDONLY) && features) {
2070 printk(KERN_ERR "EXT4-fs: %s: couldn't mount RDWR because of "
2071 "unsupported optional features (%x).\n",
2072 sb->s_id, le32_to_cpu(features));
2075 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2077 * Large file size enabled file system can only be
2078 * mount if kernel is build with CONFIG_LSF
2080 if (sizeof(root->i_blocks) < sizeof(u64) &&
2081 !(sb->s_flags & MS_RDONLY)) {
2082 printk(KERN_ERR "EXT4-fs: %s: Filesystem with huge "
2083 "files cannot be mounted read-write "
2084 "without CONFIG_LSF.\n", sb->s_id);
2088 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2090 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2091 blocksize > EXT4_MAX_BLOCK_SIZE) {
2093 "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
2094 blocksize, sb->s_id);
2098 if (sb->s_blocksize != blocksize) {
2100 /* Validate the filesystem blocksize */
2101 if (!sb_set_blocksize(sb, blocksize)) {
2102 printk(KERN_ERR "EXT4-fs: bad block size %d.\n",
2108 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2109 offset = do_div(logical_sb_block, blocksize);
2110 bh = sb_bread(sb, logical_sb_block);
2113 "EXT4-fs: Can't read superblock on 2nd try.\n");
2116 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
2118 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2120 "EXT4-fs: Magic mismatch, very weird !\n");
2125 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits);
2126 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits);
2128 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2129 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2130 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
2132 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2133 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
2134 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
2135 (!is_power_of_2(sbi->s_inode_size)) ||
2136 (sbi->s_inode_size > blocksize)) {
2138 "EXT4-fs: unsupported inode size: %d\n",
2142 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2143 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
2145 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2146 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
2147 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
2148 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
2149 !is_power_of_2(sbi->s_desc_size)) {
2151 "EXT4-fs: unsupported descriptor size %lu\n",
2156 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
2157 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
2158 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
2159 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
2161 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
2162 if (sbi->s_inodes_per_block == 0)
2164 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2165 sbi->s_inodes_per_block;
2166 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
2168 sbi->s_mount_state = le16_to_cpu(es->s_state);
2169 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2170 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2171 for (i=0; i < 4; i++)
2172 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2173 sbi->s_def_hash_version = es->s_def_hash_version;
2175 if (sbi->s_blocks_per_group > blocksize * 8) {
2177 "EXT4-fs: #blocks per group too big: %lu\n",
2178 sbi->s_blocks_per_group);
2181 if (sbi->s_inodes_per_group > blocksize * 8) {
2183 "EXT4-fs: #inodes per group too big: %lu\n",
2184 sbi->s_inodes_per_group);
2188 if (ext4_blocks_count(es) >
2189 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
2190 printk(KERN_ERR "EXT4-fs: filesystem on %s:"
2191 " too large to mount safely\n", sb->s_id);
2192 if (sizeof(sector_t) < 8)
2193 printk(KERN_WARNING "EXT4-fs: CONFIG_LBD not "
2198 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2201 /* ensure blocks_count calculation below doesn't sign-extend */
2202 if (ext4_blocks_count(es) + EXT4_BLOCKS_PER_GROUP(sb) <
2203 le32_to_cpu(es->s_first_data_block) + 1) {
2204 printk(KERN_WARNING "EXT4-fs: bad geometry: block count %llu, "
2205 "first data block %u, blocks per group %lu\n",
2206 ext4_blocks_count(es),
2207 le32_to_cpu(es->s_first_data_block),
2208 EXT4_BLOCKS_PER_GROUP(sb));
2211 blocks_count = (ext4_blocks_count(es) -
2212 le32_to_cpu(es->s_first_data_block) +
2213 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2214 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2215 sbi->s_groups_count = blocks_count;
2216 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2217 EXT4_DESC_PER_BLOCK(sb);
2218 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
2220 if (sbi->s_group_desc == NULL) {
2221 printk (KERN_ERR "EXT4-fs: not enough memory\n");
2225 bgl_lock_init(&sbi->s_blockgroup_lock);
2227 for (i = 0; i < db_count; i++) {
2228 block = descriptor_loc(sb, logical_sb_block, i);
2229 sbi->s_group_desc[i] = sb_bread(sb, block);
2230 if (!sbi->s_group_desc[i]) {
2231 printk (KERN_ERR "EXT4-fs: "
2232 "can't read group descriptor %d\n", i);
2237 if (!ext4_check_descriptors (sb)) {
2238 printk(KERN_ERR "EXT4-fs: group descriptors corrupted!\n");
2241 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2242 if (!ext4_fill_flex_info(sb)) {
2244 "EXT4-fs: unable to initialize "
2245 "flex_bg meta info!\n");
2249 sbi->s_gdb_count = db_count;
2250 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2251 spin_lock_init(&sbi->s_next_gen_lock);
2253 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2254 ext4_count_free_blocks(sb));
2256 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2257 ext4_count_free_inodes(sb));
2260 err = percpu_counter_init(&sbi->s_dirs_counter,
2261 ext4_count_dirs(sb));
2264 printk(KERN_ERR "EXT4-fs: insufficient memory\n");
2268 /* per fileystem reservation list head & lock */
2269 spin_lock_init(&sbi->s_rsv_window_lock);
2270 sbi->s_rsv_window_root = RB_ROOT;
2271 /* Add a single, static dummy reservation to the start of the
2272 * reservation window list --- it gives us a placeholder for
2273 * append-at-start-of-list which makes the allocation logic
2274 * _much_ simpler. */
2275 sbi->s_rsv_window_head.rsv_start = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
2276 sbi->s_rsv_window_head.rsv_end = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
2277 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
2278 sbi->s_rsv_window_head.rsv_goal_size = 0;
2279 ext4_rsv_window_add(sb, &sbi->s_rsv_window_head);
2281 sbi->s_stripe = ext4_get_stripe_size(sbi);
2284 * set up enough so that it can read an inode
2286 sb->s_op = &ext4_sops;
2287 sb->s_export_op = &ext4_export_ops;
2288 sb->s_xattr = ext4_xattr_handlers;
2290 sb->s_qcop = &ext4_qctl_operations;
2291 sb->dq_op = &ext4_quota_operations;
2293 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2297 needs_recovery = (es->s_last_orphan != 0 ||
2298 EXT4_HAS_INCOMPAT_FEATURE(sb,
2299 EXT4_FEATURE_INCOMPAT_RECOVER));
2302 * The first inode we look at is the journal inode. Don't try
2303 * root first: it may be modified in the journal!
2305 if (!test_opt(sb, NOLOAD) &&
2306 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2307 if (ext4_load_journal(sb, es, journal_devnum))
2309 if (!(sb->s_flags & MS_RDONLY) &&
2310 EXT4_SB(sb)->s_journal->j_failed_commit) {
2311 printk(KERN_CRIT "EXT4-fs error (device %s): "
2312 "ext4_fill_super: Journal transaction "
2313 "%u is corrupt\n", sb->s_id,
2314 EXT4_SB(sb)->s_journal->j_failed_commit);
2315 if (test_opt (sb, ERRORS_RO)) {
2317 "Mounting filesystem read-only\n");
2318 sb->s_flags |= MS_RDONLY;
2319 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2320 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2322 if (test_opt(sb, ERRORS_PANIC)) {
2323 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2324 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2325 ext4_commit_super(sb, es, 1);
2327 "EXT4-fs (device %s): mount failed\n",
2332 } else if (journal_inum) {
2333 if (ext4_create_journal(sb, es, journal_inum))
2338 "ext4: No journal on filesystem on %s\n",
2343 if (ext4_blocks_count(es) > 0xffffffffULL &&
2344 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2345 JBD2_FEATURE_INCOMPAT_64BIT)) {
2346 printk(KERN_ERR "ext4: Failed to set 64-bit journal feature\n");
2350 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2351 jbd2_journal_set_features(sbi->s_journal,
2352 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2353 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2354 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2355 jbd2_journal_set_features(sbi->s_journal,
2356 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2357 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2358 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2360 jbd2_journal_clear_features(sbi->s_journal,
2361 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2362 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2365 /* We have now updated the journal if required, so we can
2366 * validate the data journaling mode. */
2367 switch (test_opt(sb, DATA_FLAGS)) {
2369 /* No mode set, assume a default based on the journal
2370 * capabilities: ORDERED_DATA if the journal can
2371 * cope, else JOURNAL_DATA
2373 if (jbd2_journal_check_available_features
2374 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
2375 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2377 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2380 case EXT4_MOUNT_ORDERED_DATA:
2381 case EXT4_MOUNT_WRITEBACK_DATA:
2382 if (!jbd2_journal_check_available_features
2383 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
2384 printk(KERN_ERR "EXT4-fs: Journal does not support "
2385 "requested data journaling mode\n");
2392 if (test_opt(sb, NOBH)) {
2393 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2394 printk(KERN_WARNING "EXT4-fs: Ignoring nobh option - "
2395 "its supported only with writeback mode\n");
2396 clear_opt(sbi->s_mount_opt, NOBH);
2400 * The jbd2_journal_load will have done any necessary log recovery,
2401 * so we can safely mount the rest of the filesystem now.
2404 root = ext4_iget(sb, EXT4_ROOT_INO);
2406 printk(KERN_ERR "EXT4-fs: get root inode failed\n");
2407 ret = PTR_ERR(root);
2410 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2412 printk(KERN_ERR "EXT4-fs: corrupt root inode, run e2fsck\n");
2415 sb->s_root = d_alloc_root(root);
2417 printk(KERN_ERR "EXT4-fs: get root dentry failed\n");
2423 ext4_setup_super (sb, es, sb->s_flags & MS_RDONLY);
2425 /* determine the minimum size of new large inodes, if present */
2426 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2427 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2428 EXT4_GOOD_OLD_INODE_SIZE;
2429 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2430 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2431 if (sbi->s_want_extra_isize <
2432 le16_to_cpu(es->s_want_extra_isize))
2433 sbi->s_want_extra_isize =
2434 le16_to_cpu(es->s_want_extra_isize);
2435 if (sbi->s_want_extra_isize <
2436 le16_to_cpu(es->s_min_extra_isize))
2437 sbi->s_want_extra_isize =
2438 le16_to_cpu(es->s_min_extra_isize);
2441 /* Check if enough inode space is available */
2442 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2443 sbi->s_inode_size) {
2444 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2445 EXT4_GOOD_OLD_INODE_SIZE;
2446 printk(KERN_INFO "EXT4-fs: required extra inode space not"
2451 * akpm: core read_super() calls in here with the superblock locked.
2452 * That deadlocks, because orphan cleanup needs to lock the superblock
2453 * in numerous places. Here we just pop the lock - it's relatively
2454 * harmless, because we are now ready to accept write_super() requests,
2455 * and aviro says that's the only reason for hanging onto the
2458 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2459 ext4_orphan_cleanup(sb, es);
2460 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
2462 printk (KERN_INFO "EXT4-fs: recovery complete.\n");
2463 ext4_mark_recovery_complete(sb, es);
2464 printk (KERN_INFO "EXT4-fs: mounted filesystem with %s data mode.\n",
2465 test_opt(sb,DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA ? "journal":
2466 test_opt(sb,DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA ? "ordered":
2469 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
2470 printk(KERN_WARNING "EXT4-fs: Ignoring delalloc option - "
2471 "requested data journaling mode\n");
2472 clear_opt(sbi->s_mount_opt, DELALLOC);
2473 } else if (test_opt(sb, DELALLOC))
2474 printk(KERN_INFO "EXT4-fs: delayed allocation enabled\n");
2477 ext4_mb_init(sb, needs_recovery);
2484 printk(KERN_ERR "VFS: Can't find ext4 filesystem on dev %s.\n",
2489 jbd2_journal_destroy(sbi->s_journal);
2490 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);
2496 for (i = 0; i < db_count; i++)
2497 brelse(sbi->s_group_desc[i]);
2498 kfree(sbi->s_group_desc);
2501 for (i = 0; i < MAXQUOTAS; i++)
2502 kfree(sbi->s_qf_names[i]);
2504 ext4_blkdev_remove(sbi);
2507 sb->s_fs_info = NULL;
2514 * Setup any per-fs journal parameters now. We'll do this both on
2515 * initial mount, once the journal has been initialised but before we've
2516 * done any recovery; and again on any subsequent remount.
2518 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
2520 struct ext4_sb_info *sbi = EXT4_SB(sb);
2522 if (sbi->s_commit_interval)
2523 journal->j_commit_interval = sbi->s_commit_interval;
2524 /* We could also set up an ext4-specific default for the commit
2525 * interval here, but for now we'll just fall back to the jbd
2528 spin_lock(&journal->j_state_lock);
2529 if (test_opt(sb, BARRIER))
2530 journal->j_flags |= JBD2_BARRIER;
2532 journal->j_flags &= ~JBD2_BARRIER;
2533 spin_unlock(&journal->j_state_lock);
2536 static journal_t *ext4_get_journal(struct super_block *sb,
2537 unsigned int journal_inum)
2539 struct inode *journal_inode;
2542 /* First, test for the existence of a valid inode on disk. Bad
2543 * things happen if we iget() an unused inode, as the subsequent
2544 * iput() will try to delete it. */
2546 journal_inode = ext4_iget(sb, journal_inum);
2547 if (IS_ERR(journal_inode)) {
2548 printk(KERN_ERR "EXT4-fs: no journal found.\n");
2551 if (!journal_inode->i_nlink) {
2552 make_bad_inode(journal_inode);
2553 iput(journal_inode);
2554 printk(KERN_ERR "EXT4-fs: journal inode is deleted.\n");
2558 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
2559 journal_inode, journal_inode->i_size);
2560 if (!S_ISREG(journal_inode->i_mode)) {
2561 printk(KERN_ERR "EXT4-fs: invalid journal inode.\n");
2562 iput(journal_inode);
2566 journal = jbd2_journal_init_inode(journal_inode);
2568 printk(KERN_ERR "EXT4-fs: Could not load journal inode\n");
2569 iput(journal_inode);
2572 journal->j_private = sb;
2573 ext4_init_journal_params(sb, journal);
2577 static journal_t *ext4_get_dev_journal(struct super_block *sb,
2580 struct buffer_head * bh;
2584 int hblock, blocksize;
2585 ext4_fsblk_t sb_block;
2586 unsigned long offset;
2587 struct ext4_super_block * es;
2588 struct block_device *bdev;
2590 bdev = ext4_blkdev_get(j_dev);
2594 if (bd_claim(bdev, sb)) {
2596 "EXT4: failed to claim external journal device.\n");
2601 blocksize = sb->s_blocksize;
2602 hblock = bdev_hardsect_size(bdev);
2603 if (blocksize < hblock) {
2605 "EXT4-fs: blocksize too small for journal device.\n");
2609 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
2610 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
2611 set_blocksize(bdev, blocksize);
2612 if (!(bh = __bread(bdev, sb_block, blocksize))) {
2613 printk(KERN_ERR "EXT4-fs: couldn't read superblock of "
2614 "external journal\n");
2618 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2619 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
2620 !(le32_to_cpu(es->s_feature_incompat) &
2621 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2622 printk(KERN_ERR "EXT4-fs: external journal has "
2623 "bad superblock\n");
2628 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2629 printk(KERN_ERR "EXT4-fs: journal UUID does not match\n");
2634 len = ext4_blocks_count(es);
2635 start = sb_block + 1;
2636 brelse(bh); /* we're done with the superblock */
2638 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
2639 start, len, blocksize);
2641 printk(KERN_ERR "EXT4-fs: failed to create device journal\n");
2644 journal->j_private = sb;
2645 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2646 wait_on_buffer(journal->j_sb_buffer);
2647 if (!buffer_uptodate(journal->j_sb_buffer)) {
2648 printk(KERN_ERR "EXT4-fs: I/O error on journal device\n");
2651 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
2652 printk(KERN_ERR "EXT4-fs: External journal has more than one "
2653 "user (unsupported) - %d\n",
2654 be32_to_cpu(journal->j_superblock->s_nr_users));
2657 EXT4_SB(sb)->journal_bdev = bdev;
2658 ext4_init_journal_params(sb, journal);
2661 jbd2_journal_destroy(journal);
2663 ext4_blkdev_put(bdev);
2667 static int ext4_load_journal(struct super_block *sb,
2668 struct ext4_super_block *es,
2669 unsigned long journal_devnum)
2672 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2675 int really_read_only;
2677 if (journal_devnum &&
2678 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2679 printk(KERN_INFO "EXT4-fs: external journal device major/minor "
2680 "numbers have changed\n");
2681 journal_dev = new_decode_dev(journal_devnum);
2683 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2685 really_read_only = bdev_read_only(sb->s_bdev);
2688 * Are we loading a blank journal or performing recovery after a
2689 * crash? For recovery, we need to check in advance whether we
2690 * can get read-write access to the device.
2693 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2694 if (sb->s_flags & MS_RDONLY) {
2695 printk(KERN_INFO "EXT4-fs: INFO: recovery "
2696 "required on readonly filesystem.\n");
2697 if (really_read_only) {
2698 printk(KERN_ERR "EXT4-fs: write access "
2699 "unavailable, cannot proceed.\n");
2702 printk (KERN_INFO "EXT4-fs: write access will "
2703 "be enabled during recovery.\n");
2707 if (journal_inum && journal_dev) {
2708 printk(KERN_ERR "EXT4-fs: filesystem has both journal "
2709 "and inode journals!\n");
2714 if (!(journal = ext4_get_journal(sb, journal_inum)))
2717 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
2721 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2722 err = jbd2_journal_update_format(journal);
2724 printk(KERN_ERR "EXT4-fs: error updating journal.\n");
2725 jbd2_journal_destroy(journal);
2730 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
2731 err = jbd2_journal_wipe(journal, !really_read_only);
2733 err = jbd2_journal_load(journal);
2736 printk(KERN_ERR "EXT4-fs: error loading journal.\n");
2737 jbd2_journal_destroy(journal);
2741 EXT4_SB(sb)->s_journal = journal;
2742 ext4_clear_journal_err(sb, es);
2744 if (journal_devnum &&
2745 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2746 es->s_journal_dev = cpu_to_le32(journal_devnum);
2749 /* Make sure we flush the recovery flag to disk. */
2750 ext4_commit_super(sb, es, 1);
2756 static int ext4_create_journal(struct super_block * sb,
2757 struct ext4_super_block * es,
2758 unsigned int journal_inum)
2763 if (sb->s_flags & MS_RDONLY) {
2764 printk(KERN_ERR "EXT4-fs: readonly filesystem when trying to "
2765 "create journal.\n");
2769 journal = ext4_get_journal(sb, journal_inum);
2773 printk(KERN_INFO "EXT4-fs: creating new journal on inode %u\n",
2776 err = jbd2_journal_create(journal);
2778 printk(KERN_ERR "EXT4-fs: error creating journal.\n");
2779 jbd2_journal_destroy(journal);
2783 EXT4_SB(sb)->s_journal = journal;
2785 ext4_update_dynamic_rev(sb);
2786 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2787 EXT4_SET_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL);
2789 es->s_journal_inum = cpu_to_le32(journal_inum);
2792 /* Make sure we flush the recovery flag to disk. */
2793 ext4_commit_super(sb, es, 1);
2798 static void ext4_commit_super (struct super_block * sb,
2799 struct ext4_super_block * es,
2802 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
2806 es->s_wtime = cpu_to_le32(get_seconds());
2807 ext4_free_blocks_count_set(es, ext4_count_free_blocks(sb));
2808 es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
2809 BUFFER_TRACE(sbh, "marking dirty");
2810 mark_buffer_dirty(sbh);
2812 sync_dirty_buffer(sbh);
2817 * Have we just finished recovery? If so, and if we are mounting (or
2818 * remounting) the filesystem readonly, then we will end up with a
2819 * consistent fs on disk. Record that fact.
2821 static void ext4_mark_recovery_complete(struct super_block * sb,
2822 struct ext4_super_block * es)
2824 journal_t *journal = EXT4_SB(sb)->s_journal;
2826 jbd2_journal_lock_updates(journal);
2827 jbd2_journal_flush(journal);
2829 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
2830 sb->s_flags & MS_RDONLY) {
2831 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2833 ext4_commit_super(sb, es, 1);
2836 jbd2_journal_unlock_updates(journal);
2840 * If we are mounting (or read-write remounting) a filesystem whose journal
2841 * has recorded an error from a previous lifetime, move that error to the
2842 * main filesystem now.
2844 static void ext4_clear_journal_err(struct super_block * sb,
2845 struct ext4_super_block * es)
2851 journal = EXT4_SB(sb)->s_journal;
2854 * Now check for any error status which may have been recorded in the
2855 * journal by a prior ext4_error() or ext4_abort()
2858 j_errno = jbd2_journal_errno(journal);
2862 errstr = ext4_decode_error(sb, j_errno, nbuf);
2863 ext4_warning(sb, __func__, "Filesystem error recorded "
2864 "from previous mount: %s", errstr);
2865 ext4_warning(sb, __func__, "Marking fs in need of "
2866 "filesystem check.");
2868 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2869 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2870 ext4_commit_super (sb, es, 1);
2872 jbd2_journal_clear_err(journal);
2877 * Force the running and committing transactions to commit,
2878 * and wait on the commit.
2880 int ext4_force_commit(struct super_block *sb)
2885 if (sb->s_flags & MS_RDONLY)
2888 journal = EXT4_SB(sb)->s_journal;
2890 ret = ext4_journal_force_commit(journal);
2895 * Ext4 always journals updates to the superblock itself, so we don't
2896 * have to propagate any other updates to the superblock on disk at this
2897 * point. Just start an async writeback to get the buffers on their way
2900 * This implicitly triggers the writebehind on sync().
2903 static void ext4_write_super (struct super_block * sb)
2905 if (mutex_trylock(&sb->s_lock) != 0)
2910 static int ext4_sync_fs(struct super_block *sb, int wait)
2915 if (jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, &target)) {
2917 jbd2_log_wait_commit(EXT4_SB(sb)->s_journal, target);
2923 * LVM calls this function before a (read-only) snapshot is created. This
2924 * gives us a chance to flush the journal completely and mark the fs clean.
2926 static void ext4_write_super_lockfs(struct super_block *sb)
2930 if (!(sb->s_flags & MS_RDONLY)) {
2931 journal_t *journal = EXT4_SB(sb)->s_journal;
2933 /* Now we set up the journal barrier. */
2934 jbd2_journal_lock_updates(journal);
2935 jbd2_journal_flush(journal);
2937 /* Journal blocked and flushed, clear needs_recovery flag. */
2938 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2939 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
2944 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2945 * flag here, even though the filesystem is not technically dirty yet.
2947 static void ext4_unlockfs(struct super_block *sb)
2949 if (!(sb->s_flags & MS_RDONLY)) {
2951 /* Reser the needs_recovery flag before the fs is unlocked. */
2952 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2953 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
2955 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
2959 static int ext4_remount (struct super_block * sb, int * flags, char * data)
2961 struct ext4_super_block * es;
2962 struct ext4_sb_info *sbi = EXT4_SB(sb);
2963 ext4_fsblk_t n_blocks_count = 0;
2964 unsigned long old_sb_flags;
2965 struct ext4_mount_options old_opts;
2972 /* Store the original options */
2973 old_sb_flags = sb->s_flags;
2974 old_opts.s_mount_opt = sbi->s_mount_opt;
2975 old_opts.s_resuid = sbi->s_resuid;
2976 old_opts.s_resgid = sbi->s_resgid;
2977 old_opts.s_commit_interval = sbi->s_commit_interval;
2979 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2980 for (i = 0; i < MAXQUOTAS; i++)
2981 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2985 * Allow the "check" option to be passed as a remount option.
2987 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
2992 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT)
2993 ext4_abort(sb, __func__, "Abort forced by user");
2995 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2996 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
3000 ext4_init_journal_params(sb, sbi->s_journal);
3002 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
3003 n_blocks_count > ext4_blocks_count(es)) {
3004 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) {
3009 if (*flags & MS_RDONLY) {
3011 * First of all, the unconditional stuff we have to do
3012 * to disable replay of the journal when we next remount
3014 sb->s_flags |= MS_RDONLY;
3017 * OK, test if we are remounting a valid rw partition
3018 * readonly, and if so set the rdonly flag and then
3019 * mark the partition as valid again.
3021 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3022 (sbi->s_mount_state & EXT4_VALID_FS))
3023 es->s_state = cpu_to_le16(sbi->s_mount_state);
3026 * We have to unlock super so that we can wait for
3030 ext4_mark_recovery_complete(sb, es);
3034 if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3035 ~EXT4_FEATURE_RO_COMPAT_SUPP))) {
3036 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
3037 "remount RDWR because of unsupported "
3038 "optional features (%x).\n",
3039 sb->s_id, le32_to_cpu(ret));
3045 * Make sure the group descriptor checksums
3046 * are sane. If they aren't, refuse to
3049 for (g = 0; g < sbi->s_groups_count; g++) {
3050 struct ext4_group_desc *gdp =
3051 ext4_get_group_desc(sb, g, NULL);
3053 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3055 "EXT4-fs: ext4_remount: "
3056 "Checksum for group %lu failed (%u!=%u)\n",
3057 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3058 le16_to_cpu(gdp->bg_checksum));
3065 * If we have an unprocessed orphan list hanging
3066 * around from a previously readonly bdev mount,
3067 * require a full umount/remount for now.
3069 if (es->s_last_orphan) {
3070 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
3071 "remount RDWR because of unprocessed "
3072 "orphan inode list. Please "
3073 "umount/remount instead.\n",
3080 * Mounting a RDONLY partition read-write, so reread
3081 * and store the current valid flag. (It may have
3082 * been changed by e2fsck since we originally mounted
3085 ext4_clear_journal_err(sb, es);
3086 sbi->s_mount_state = le16_to_cpu(es->s_state);
3087 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
3089 if (!ext4_setup_super (sb, es, 0))
3090 sb->s_flags &= ~MS_RDONLY;
3094 /* Release old quota file names */
3095 for (i = 0; i < MAXQUOTAS; i++)
3096 if (old_opts.s_qf_names[i] &&
3097 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3098 kfree(old_opts.s_qf_names[i]);
3102 sb->s_flags = old_sb_flags;
3103 sbi->s_mount_opt = old_opts.s_mount_opt;
3104 sbi->s_resuid = old_opts.s_resuid;
3105 sbi->s_resgid = old_opts.s_resgid;
3106 sbi->s_commit_interval = old_opts.s_commit_interval;
3108 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3109 for (i = 0; i < MAXQUOTAS; i++) {
3110 if (sbi->s_qf_names[i] &&
3111 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3112 kfree(sbi->s_qf_names[i]);
3113 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3119 static int ext4_statfs (struct dentry * dentry, struct kstatfs * buf)
3121 struct super_block *sb = dentry->d_sb;
3122 struct ext4_sb_info *sbi = EXT4_SB(sb);
3123 struct ext4_super_block *es = sbi->s_es;
3126 if (test_opt(sb, MINIX_DF)) {
3127 sbi->s_overhead_last = 0;
3128 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
3129 ext4_group_t ngroups = sbi->s_groups_count, i;
3130 ext4_fsblk_t overhead = 0;
3134 * Compute the overhead (FS structures). This is constant
3135 * for a given filesystem unless the number of block groups
3136 * changes so we cache the previous value until it does.
3140 * All of the blocks before first_data_block are
3143 overhead = le32_to_cpu(es->s_first_data_block);
3146 * Add the overhead attributed to the superblock and
3147 * block group descriptors. If the sparse superblocks
3148 * feature is turned on, then not all groups have this.
3150 for (i = 0; i < ngroups; i++) {
3151 overhead += ext4_bg_has_super(sb, i) +
3152 ext4_bg_num_gdb(sb, i);
3157 * Every block group has an inode bitmap, a block
3158 * bitmap, and an inode table.
3160 overhead += ngroups * (2 + sbi->s_itb_per_group);
3161 sbi->s_overhead_last = overhead;
3163 sbi->s_blocks_last = ext4_blocks_count(es);
3166 buf->f_type = EXT4_SUPER_MAGIC;
3167 buf->f_bsize = sb->s_blocksize;
3168 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
3169 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter);
3170 ext4_free_blocks_count_set(es, buf->f_bfree);
3171 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3172 if (buf->f_bfree < ext4_r_blocks_count(es))
3174 buf->f_files = le32_to_cpu(es->s_inodes_count);
3175 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
3176 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
3177 buf->f_namelen = EXT4_NAME_LEN;
3178 fsid = le64_to_cpup((void *)es->s_uuid) ^
3179 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3180 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3181 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
3185 /* Helper function for writing quotas on sync - we need to start transaction before quota file
3186 * is locked for write. Otherwise the are possible deadlocks:
3187 * Process 1 Process 2
3188 * ext4_create() quota_sync()
3189 * jbd2_journal_start() write_dquot()
3190 * DQUOT_INIT() down(dqio_mutex)
3191 * down(dqio_mutex) jbd2_journal_start()
3197 static inline struct inode *dquot_to_inode(struct dquot *dquot)
3199 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3202 static int ext4_dquot_initialize(struct inode *inode, int type)
3207 /* We may create quota structure so we need to reserve enough blocks */
3208 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_INIT_BLOCKS(inode->i_sb));
3210 return PTR_ERR(handle);
3211 ret = dquot_initialize(inode, type);
3212 err = ext4_journal_stop(handle);
3218 static int ext4_dquot_drop(struct inode *inode)
3223 /* We may delete quota structure so we need to reserve enough blocks */
3224 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_DEL_BLOCKS(inode->i_sb));
3225 if (IS_ERR(handle)) {
3227 * We call dquot_drop() anyway to at least release references
3228 * to quota structures so that umount does not hang.
3231 return PTR_ERR(handle);
3233 ret = dquot_drop(inode);
3234 err = ext4_journal_stop(handle);
3240 static int ext4_write_dquot(struct dquot *dquot)
3244 struct inode *inode;
3246 inode = dquot_to_inode(dquot);
3247 handle = ext4_journal_start(inode,
3248 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
3250 return PTR_ERR(handle);
3251 ret = dquot_commit(dquot);
3252 err = ext4_journal_stop(handle);
3258 static int ext4_acquire_dquot(struct dquot *dquot)
3263 handle = ext4_journal_start(dquot_to_inode(dquot),
3264 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
3266 return PTR_ERR(handle);
3267 ret = dquot_acquire(dquot);
3268 err = ext4_journal_stop(handle);
3274 static int ext4_release_dquot(struct dquot *dquot)
3279 handle = ext4_journal_start(dquot_to_inode(dquot),
3280 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
3281 if (IS_ERR(handle)) {
3282 /* Release dquot anyway to avoid endless cycle in dqput() */
3283 dquot_release(dquot);
3284 return PTR_ERR(handle);
3286 ret = dquot_release(dquot);
3287 err = ext4_journal_stop(handle);
3293 static int ext4_mark_dquot_dirty(struct dquot *dquot)
3295 /* Are we journaling quotas? */
3296 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3297 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
3298 dquot_mark_dquot_dirty(dquot);
3299 return ext4_write_dquot(dquot);
3301 return dquot_mark_dquot_dirty(dquot);
3305 static int ext4_write_info(struct super_block *sb, int type)
3310 /* Data block + inode block */
3311 handle = ext4_journal_start(sb->s_root->d_inode, 2);
3313 return PTR_ERR(handle);
3314 ret = dquot_commit_info(sb, type);
3315 err = ext4_journal_stop(handle);
3322 * Turn on quotas during mount time - we need to find
3323 * the quota file and such...
3325 static int ext4_quota_on_mount(struct super_block *sb, int type)
3327 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3328 EXT4_SB(sb)->s_jquota_fmt, type);
3332 * Standard function to be called on quota_on
3334 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
3335 char *path, int remount)
3338 struct nameidata nd;
3340 if (!test_opt(sb, QUOTA))
3342 /* When remounting, no checks are needed and in fact, path is NULL */
3344 return vfs_quota_on(sb, type, format_id, path, remount);
3346 err = path_lookup(path, LOOKUP_FOLLOW, &nd);
3350 /* Quotafile not on the same filesystem? */
3351 if (nd.path.mnt->mnt_sb != sb) {
3355 /* Journaling quota? */
3356 if (EXT4_SB(sb)->s_qf_names[type]) {
3357 /* Quotafile not of fs root? */
3358 if (nd.path.dentry->d_parent->d_inode != sb->s_root->d_inode)
3360 "EXT4-fs: Quota file not on filesystem root. "
3361 "Journaled quota will not work.\n");
3365 * When we journal data on quota file, we have to flush journal to see
3366 * all updates to the file when we bypass pagecache...
3368 if (ext4_should_journal_data(nd.path.dentry->d_inode)) {
3370 * We don't need to lock updates but journal_flush() could
3371 * otherwise be livelocked...
3373 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
3374 jbd2_journal_flush(EXT4_SB(sb)->s_journal);
3375 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3379 return vfs_quota_on(sb, type, format_id, path, remount);
3382 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3383 * acquiring the locks... As quota files are never truncated and quota code
3384 * itself serializes the operations (and noone else should touch the files)
3385 * we don't have to be afraid of races */
3386 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
3387 size_t len, loff_t off)
3389 struct inode *inode = sb_dqopt(sb)->files[type];
3390 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3392 int offset = off & (sb->s_blocksize - 1);
3395 struct buffer_head *bh;
3396 loff_t i_size = i_size_read(inode);
3400 if (off+len > i_size)
3403 while (toread > 0) {
3404 tocopy = sb->s_blocksize - offset < toread ?
3405 sb->s_blocksize - offset : toread;
3406 bh = ext4_bread(NULL, inode, blk, 0, &err);
3409 if (!bh) /* A hole? */
3410 memset(data, 0, tocopy);
3412 memcpy(data, bh->b_data+offset, tocopy);
3422 /* Write to quotafile (we know the transaction is already started and has
3423 * enough credits) */
3424 static ssize_t ext4_quota_write(struct super_block *sb, int type,
3425 const char *data, size_t len, loff_t off)
3427 struct inode *inode = sb_dqopt(sb)->files[type];
3428 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3430 int offset = off & (sb->s_blocksize - 1);
3432 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
3433 size_t towrite = len;
3434 struct buffer_head *bh;
3435 handle_t *handle = journal_current_handle();
3438 printk(KERN_WARNING "EXT4-fs: Quota write (off=%Lu, len=%Lu)"
3439 " cancelled because transaction is not started.\n",
3440 (unsigned long long)off, (unsigned long long)len);
3443 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3444 while (towrite > 0) {
3445 tocopy = sb->s_blocksize - offset < towrite ?
3446 sb->s_blocksize - offset : towrite;
3447 bh = ext4_bread(handle, inode, blk, 1, &err);
3450 if (journal_quota) {
3451 err = ext4_journal_get_write_access(handle, bh);
3458 memcpy(bh->b_data+offset, data, tocopy);
3459 flush_dcache_page(bh->b_page);
3462 err = ext4_journal_dirty_metadata(handle, bh);
3464 /* Always do at least ordered writes for quotas */
3465 err = ext4_jbd2_file_inode(handle, inode);
3466 mark_buffer_dirty(bh);
3477 if (len == towrite) {
3478 mutex_unlock(&inode->i_mutex);
3481 if (inode->i_size < off+len-towrite) {
3482 i_size_write(inode, off+len-towrite);
3483 EXT4_I(inode)->i_disksize = inode->i_size;
3485 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3486 ext4_mark_inode_dirty(handle, inode);
3487 mutex_unlock(&inode->i_mutex);
3488 return len - towrite;
3493 static int ext4_get_sb(struct file_system_type *fs_type,
3494 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3496 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
3499 static struct file_system_type ext4dev_fs_type = {
3500 .owner = THIS_MODULE,
3502 .get_sb = ext4_get_sb,
3503 .kill_sb = kill_block_super,
3504 .fs_flags = FS_REQUIRES_DEV,
3507 static int __init init_ext4_fs(void)
3511 err = init_ext4_mballoc();
3515 err = init_ext4_xattr();
3518 err = init_inodecache();
3521 err = register_filesystem(&ext4dev_fs_type);
3526 destroy_inodecache();
3530 exit_ext4_mballoc();
3534 static void __exit exit_ext4_fs(void)
3536 unregister_filesystem(&ext4dev_fs_type);
3537 destroy_inodecache();
3539 exit_ext4_mballoc();
3542 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
3543 MODULE_DESCRIPTION("Fourth Extended Filesystem with extents");
3544 MODULE_LICENSE("GPL");
3545 module_init(init_ext4_fs)
3546 module_exit(exit_ext4_fs)