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 if (!(sb->s_flags & MS_RDONLY)) {
510 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
511 es->s_state = cpu_to_le16(sbi->s_mount_state);
512 BUFFER_TRACE(sbi->s_sbh, "marking dirty");
513 mark_buffer_dirty(sbi->s_sbh);
514 ext4_commit_super(sb, es, 1);
517 for (i = 0; i < sbi->s_gdb_count; i++)
518 brelse(sbi->s_group_desc[i]);
519 kfree(sbi->s_group_desc);
520 percpu_counter_destroy(&sbi->s_freeblocks_counter);
521 percpu_counter_destroy(&sbi->s_freeinodes_counter);
522 percpu_counter_destroy(&sbi->s_dirs_counter);
525 for (i = 0; i < MAXQUOTAS; i++)
526 kfree(sbi->s_qf_names[i]);
529 /* Debugging code just in case the in-memory inode orphan list
530 * isn't empty. The on-disk one can be non-empty if we've
531 * detected an error and taken the fs readonly, but the
532 * in-memory list had better be clean by this point. */
533 if (!list_empty(&sbi->s_orphan))
534 dump_orphan_list(sb, sbi);
535 J_ASSERT(list_empty(&sbi->s_orphan));
537 invalidate_bdev(sb->s_bdev);
538 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
540 * Invalidate the journal device's buffers. We don't want them
541 * floating about in memory - the physical journal device may
542 * hotswapped, and it breaks the `ro-after' testing code.
544 sync_blockdev(sbi->journal_bdev);
545 invalidate_bdev(sbi->journal_bdev);
546 ext4_blkdev_remove(sbi);
548 sb->s_fs_info = NULL;
553 static struct kmem_cache *ext4_inode_cachep;
556 * Called inside transaction, so use GFP_NOFS
558 static struct inode *ext4_alloc_inode(struct super_block *sb)
560 struct ext4_inode_info *ei;
562 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
565 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
566 ei->i_acl = EXT4_ACL_NOT_CACHED;
567 ei->i_default_acl = EXT4_ACL_NOT_CACHED;
569 ei->i_block_alloc_info = NULL;
570 ei->vfs_inode.i_version = 1;
571 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
572 INIT_LIST_HEAD(&ei->i_prealloc_list);
573 spin_lock_init(&ei->i_prealloc_lock);
574 return &ei->vfs_inode;
577 static void ext4_destroy_inode(struct inode *inode)
579 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
580 printk("EXT4 Inode %p: orphan list check failed!\n",
582 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
583 EXT4_I(inode), sizeof(struct ext4_inode_info),
587 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
590 static void init_once(struct kmem_cache *cachep, void *foo)
592 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
594 INIT_LIST_HEAD(&ei->i_orphan);
595 #ifdef CONFIG_EXT4DEV_FS_XATTR
596 init_rwsem(&ei->xattr_sem);
598 init_rwsem(&ei->i_data_sem);
599 inode_init_once(&ei->vfs_inode);
602 static int init_inodecache(void)
604 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
605 sizeof(struct ext4_inode_info),
606 0, (SLAB_RECLAIM_ACCOUNT|
609 if (ext4_inode_cachep == NULL)
614 static void destroy_inodecache(void)
616 kmem_cache_destroy(ext4_inode_cachep);
619 static void ext4_clear_inode(struct inode *inode)
621 struct ext4_block_alloc_info *rsv = EXT4_I(inode)->i_block_alloc_info;
622 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
623 if (EXT4_I(inode)->i_acl &&
624 EXT4_I(inode)->i_acl != EXT4_ACL_NOT_CACHED) {
625 posix_acl_release(EXT4_I(inode)->i_acl);
626 EXT4_I(inode)->i_acl = EXT4_ACL_NOT_CACHED;
628 if (EXT4_I(inode)->i_default_acl &&
629 EXT4_I(inode)->i_default_acl != EXT4_ACL_NOT_CACHED) {
630 posix_acl_release(EXT4_I(inode)->i_default_acl);
631 EXT4_I(inode)->i_default_acl = EXT4_ACL_NOT_CACHED;
634 ext4_discard_reservation(inode);
635 EXT4_I(inode)->i_block_alloc_info = NULL;
640 static inline void ext4_show_quota_options(struct seq_file *seq, struct super_block *sb)
642 #if defined(CONFIG_QUOTA)
643 struct ext4_sb_info *sbi = EXT4_SB(sb);
645 if (sbi->s_jquota_fmt)
646 seq_printf(seq, ",jqfmt=%s",
647 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold": "vfsv0");
649 if (sbi->s_qf_names[USRQUOTA])
650 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
652 if (sbi->s_qf_names[GRPQUOTA])
653 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
655 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
656 seq_puts(seq, ",usrquota");
658 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
659 seq_puts(seq, ",grpquota");
665 * - it's set to a non-default value OR
666 * - if the per-sb default is different from the global default
668 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
671 unsigned long def_mount_opts;
672 struct super_block *sb = vfs->mnt_sb;
673 struct ext4_sb_info *sbi = EXT4_SB(sb);
674 journal_t *journal = sbi->s_journal;
675 struct ext4_super_block *es = sbi->s_es;
677 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
678 def_errors = le16_to_cpu(es->s_errors);
680 if (sbi->s_sb_block != 1)
681 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
682 if (test_opt(sb, MINIX_DF))
683 seq_puts(seq, ",minixdf");
684 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
685 seq_puts(seq, ",grpid");
686 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
687 seq_puts(seq, ",nogrpid");
688 if (sbi->s_resuid != EXT4_DEF_RESUID ||
689 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
690 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
692 if (sbi->s_resgid != EXT4_DEF_RESGID ||
693 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
694 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
696 if (test_opt(sb, ERRORS_RO)) {
697 if (def_errors == EXT4_ERRORS_PANIC ||
698 def_errors == EXT4_ERRORS_CONTINUE) {
699 seq_puts(seq, ",errors=remount-ro");
702 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
703 seq_puts(seq, ",errors=continue");
704 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
705 seq_puts(seq, ",errors=panic");
706 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
707 seq_puts(seq, ",nouid32");
708 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
709 seq_puts(seq, ",debug");
710 if (test_opt(sb, OLDALLOC))
711 seq_puts(seq, ",oldalloc");
712 #ifdef CONFIG_EXT4DEV_FS_XATTR
713 if (test_opt(sb, XATTR_USER) &&
714 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
715 seq_puts(seq, ",user_xattr");
716 if (!test_opt(sb, XATTR_USER) &&
717 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
718 seq_puts(seq, ",nouser_xattr");
721 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
722 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
723 seq_puts(seq, ",acl");
724 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
725 seq_puts(seq, ",noacl");
727 if (!test_opt(sb, RESERVATION))
728 seq_puts(seq, ",noreservation");
729 if (sbi->s_commit_interval) {
730 seq_printf(seq, ",commit=%u",
731 (unsigned) (sbi->s_commit_interval / HZ));
734 * We're changing the default of barrier mount option, so
735 * let's always display its mount state so it's clear what its
738 seq_puts(seq, ",barrier=");
739 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
740 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
741 seq_puts(seq, ",journal_async_commit");
742 if (test_opt(sb, NOBH))
743 seq_puts(seq, ",nobh");
744 if (!test_opt(sb, EXTENTS))
745 seq_puts(seq, ",noextents");
746 if (!test_opt(sb, MBALLOC))
747 seq_puts(seq, ",nomballoc");
748 if (test_opt(sb, I_VERSION))
749 seq_puts(seq, ",i_version");
752 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
754 * journal mode get enabled in different ways
755 * So just print the value even if we didn't specify it
757 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
758 seq_puts(seq, ",data=journal");
759 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
760 seq_puts(seq, ",data=ordered");
761 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
762 seq_puts(seq, ",data=writeback");
764 ext4_show_quota_options(seq, sb);
769 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
770 u64 ino, u32 generation)
774 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
775 return ERR_PTR(-ESTALE);
776 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
777 return ERR_PTR(-ESTALE);
779 /* iget isn't really right if the inode is currently unallocated!!
781 * ext4_read_inode will return a bad_inode if the inode had been
782 * deleted, so we should be safe.
784 * Currently we don't know the generation for parent directory, so
785 * a generation of 0 means "accept any"
787 inode = ext4_iget(sb, ino);
789 return ERR_CAST(inode);
790 if (generation && inode->i_generation != generation) {
792 return ERR_PTR(-ESTALE);
798 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
799 int fh_len, int fh_type)
801 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
805 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
806 int fh_len, int fh_type)
808 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
813 #define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
814 #define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
816 static int ext4_dquot_initialize(struct inode *inode, int type);
817 static int ext4_dquot_drop(struct inode *inode);
818 static int ext4_write_dquot(struct dquot *dquot);
819 static int ext4_acquire_dquot(struct dquot *dquot);
820 static int ext4_release_dquot(struct dquot *dquot);
821 static int ext4_mark_dquot_dirty(struct dquot *dquot);
822 static int ext4_write_info(struct super_block *sb, int type);
823 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
824 char *path, int remount);
825 static int ext4_quota_on_mount(struct super_block *sb, int type);
826 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
827 size_t len, loff_t off);
828 static ssize_t ext4_quota_write(struct super_block *sb, int type,
829 const char *data, size_t len, loff_t off);
831 static struct dquot_operations ext4_quota_operations = {
832 .initialize = ext4_dquot_initialize,
833 .drop = ext4_dquot_drop,
834 .alloc_space = dquot_alloc_space,
835 .alloc_inode = dquot_alloc_inode,
836 .free_space = dquot_free_space,
837 .free_inode = dquot_free_inode,
838 .transfer = dquot_transfer,
839 .write_dquot = ext4_write_dquot,
840 .acquire_dquot = ext4_acquire_dquot,
841 .release_dquot = ext4_release_dquot,
842 .mark_dirty = ext4_mark_dquot_dirty,
843 .write_info = ext4_write_info
846 static struct quotactl_ops ext4_qctl_operations = {
847 .quota_on = ext4_quota_on,
848 .quota_off = vfs_quota_off,
849 .quota_sync = vfs_quota_sync,
850 .get_info = vfs_get_dqinfo,
851 .set_info = vfs_set_dqinfo,
852 .get_dqblk = vfs_get_dqblk,
853 .set_dqblk = vfs_set_dqblk
857 static const struct super_operations ext4_sops = {
858 .alloc_inode = ext4_alloc_inode,
859 .destroy_inode = ext4_destroy_inode,
860 .write_inode = ext4_write_inode,
861 .dirty_inode = ext4_dirty_inode,
862 .delete_inode = ext4_delete_inode,
863 .put_super = ext4_put_super,
864 .write_super = ext4_write_super,
865 .sync_fs = ext4_sync_fs,
866 .write_super_lockfs = ext4_write_super_lockfs,
867 .unlockfs = ext4_unlockfs,
868 .statfs = ext4_statfs,
869 .remount_fs = ext4_remount,
870 .clear_inode = ext4_clear_inode,
871 .show_options = ext4_show_options,
873 .quota_read = ext4_quota_read,
874 .quota_write = ext4_quota_write,
878 static const struct export_operations ext4_export_ops = {
879 .fh_to_dentry = ext4_fh_to_dentry,
880 .fh_to_parent = ext4_fh_to_parent,
881 .get_parent = ext4_get_parent,
885 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
886 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
887 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
888 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
889 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
890 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
891 Opt_journal_checksum, Opt_journal_async_commit,
892 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
893 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
894 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
895 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
896 Opt_grpquota, Opt_extents, Opt_noextents, Opt_i_version,
897 Opt_mballoc, Opt_nomballoc, Opt_stripe,
900 static match_table_t tokens = {
901 {Opt_bsd_df, "bsddf"},
902 {Opt_minix_df, "minixdf"},
903 {Opt_grpid, "grpid"},
904 {Opt_grpid, "bsdgroups"},
905 {Opt_nogrpid, "nogrpid"},
906 {Opt_nogrpid, "sysvgroups"},
907 {Opt_resgid, "resgid=%u"},
908 {Opt_resuid, "resuid=%u"},
910 {Opt_err_cont, "errors=continue"},
911 {Opt_err_panic, "errors=panic"},
912 {Opt_err_ro, "errors=remount-ro"},
913 {Opt_nouid32, "nouid32"},
914 {Opt_nocheck, "nocheck"},
915 {Opt_nocheck, "check=none"},
916 {Opt_debug, "debug"},
917 {Opt_oldalloc, "oldalloc"},
918 {Opt_orlov, "orlov"},
919 {Opt_user_xattr, "user_xattr"},
920 {Opt_nouser_xattr, "nouser_xattr"},
922 {Opt_noacl, "noacl"},
923 {Opt_reservation, "reservation"},
924 {Opt_noreservation, "noreservation"},
925 {Opt_noload, "noload"},
928 {Opt_commit, "commit=%u"},
929 {Opt_journal_update, "journal=update"},
930 {Opt_journal_inum, "journal=%u"},
931 {Opt_journal_dev, "journal_dev=%u"},
932 {Opt_journal_checksum, "journal_checksum"},
933 {Opt_journal_async_commit, "journal_async_commit"},
934 {Opt_abort, "abort"},
935 {Opt_data_journal, "data=journal"},
936 {Opt_data_ordered, "data=ordered"},
937 {Opt_data_writeback, "data=writeback"},
938 {Opt_offusrjquota, "usrjquota="},
939 {Opt_usrjquota, "usrjquota=%s"},
940 {Opt_offgrpjquota, "grpjquota="},
941 {Opt_grpjquota, "grpjquota=%s"},
942 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
943 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
944 {Opt_grpquota, "grpquota"},
945 {Opt_noquota, "noquota"},
946 {Opt_quota, "quota"},
947 {Opt_usrquota, "usrquota"},
948 {Opt_barrier, "barrier=%u"},
949 {Opt_extents, "extents"},
950 {Opt_noextents, "noextents"},
951 {Opt_i_version, "i_version"},
952 {Opt_mballoc, "mballoc"},
953 {Opt_nomballoc, "nomballoc"},
954 {Opt_stripe, "stripe=%u"},
955 {Opt_resize, "resize"},
959 static ext4_fsblk_t get_sb_block(void **data)
961 ext4_fsblk_t sb_block;
962 char *options = (char *) *data;
964 if (!options || strncmp(options, "sb=", 3) != 0)
965 return 1; /* Default location */
967 /*todo: use simple_strtoll with >32bit ext4 */
968 sb_block = simple_strtoul(options, &options, 0);
969 if (*options && *options != ',') {
970 printk("EXT4-fs: Invalid sb specification: %s\n",
976 *data = (void *) options;
980 static int parse_options (char *options, struct super_block *sb,
981 unsigned int *inum, unsigned long *journal_devnum,
982 ext4_fsblk_t *n_blocks_count, int is_remount)
984 struct ext4_sb_info *sbi = EXT4_SB(sb);
986 substring_t args[MAX_OPT_ARGS];
997 while ((p = strsep (&options, ",")) != NULL) {
1002 token = match_token(p, tokens, args);
1005 clear_opt (sbi->s_mount_opt, MINIX_DF);
1008 set_opt (sbi->s_mount_opt, MINIX_DF);
1011 set_opt (sbi->s_mount_opt, GRPID);
1014 clear_opt (sbi->s_mount_opt, GRPID);
1017 if (match_int(&args[0], &option))
1019 sbi->s_resuid = option;
1022 if (match_int(&args[0], &option))
1024 sbi->s_resgid = option;
1027 /* handled by get_sb_block() instead of here */
1028 /* *sb_block = match_int(&args[0]); */
1031 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1032 clear_opt (sbi->s_mount_opt, ERRORS_RO);
1033 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
1036 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1037 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1038 set_opt (sbi->s_mount_opt, ERRORS_RO);
1041 clear_opt (sbi->s_mount_opt, ERRORS_RO);
1042 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1043 set_opt (sbi->s_mount_opt, ERRORS_CONT);
1046 set_opt (sbi->s_mount_opt, NO_UID32);
1049 clear_opt (sbi->s_mount_opt, CHECK);
1052 set_opt (sbi->s_mount_opt, DEBUG);
1055 set_opt (sbi->s_mount_opt, OLDALLOC);
1058 clear_opt (sbi->s_mount_opt, OLDALLOC);
1060 #ifdef CONFIG_EXT4DEV_FS_XATTR
1061 case Opt_user_xattr:
1062 set_opt (sbi->s_mount_opt, XATTR_USER);
1064 case Opt_nouser_xattr:
1065 clear_opt (sbi->s_mount_opt, XATTR_USER);
1068 case Opt_user_xattr:
1069 case Opt_nouser_xattr:
1070 printk("EXT4 (no)user_xattr options not supported\n");
1073 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
1075 set_opt(sbi->s_mount_opt, POSIX_ACL);
1078 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1083 printk("EXT4 (no)acl options not supported\n");
1086 case Opt_reservation:
1087 set_opt(sbi->s_mount_opt, RESERVATION);
1089 case Opt_noreservation:
1090 clear_opt(sbi->s_mount_opt, RESERVATION);
1092 case Opt_journal_update:
1094 /* Eventually we will want to be able to create
1095 a journal file here. For now, only allow the
1096 user to specify an existing inode to be the
1099 printk(KERN_ERR "EXT4-fs: cannot specify "
1100 "journal on remount\n");
1103 set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
1105 case Opt_journal_inum:
1107 printk(KERN_ERR "EXT4-fs: cannot specify "
1108 "journal on remount\n");
1111 if (match_int(&args[0], &option))
1115 case Opt_journal_dev:
1117 printk(KERN_ERR "EXT4-fs: cannot specify "
1118 "journal on remount\n");
1121 if (match_int(&args[0], &option))
1123 *journal_devnum = option;
1125 case Opt_journal_checksum:
1126 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1128 case Opt_journal_async_commit:
1129 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1130 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1133 set_opt (sbi->s_mount_opt, NOLOAD);
1136 if (match_int(&args[0], &option))
1141 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1142 sbi->s_commit_interval = HZ * option;
1144 case Opt_data_journal:
1145 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1147 case Opt_data_ordered:
1148 data_opt = EXT4_MOUNT_ORDERED_DATA;
1150 case Opt_data_writeback:
1151 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1154 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
1157 "EXT4-fs: cannot change data "
1158 "mode on remount\n");
1162 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
1163 sbi->s_mount_opt |= data_opt;
1173 if ((sb_any_quota_enabled(sb) ||
1174 sb_any_quota_suspended(sb)) &&
1175 !sbi->s_qf_names[qtype]) {
1177 "EXT4-fs: Cannot change journaled "
1178 "quota options when quota turned on.\n");
1181 qname = match_strdup(&args[0]);
1184 "EXT4-fs: not enough memory for "
1185 "storing quotafile name.\n");
1188 if (sbi->s_qf_names[qtype] &&
1189 strcmp(sbi->s_qf_names[qtype], qname)) {
1191 "EXT4-fs: %s quota file already "
1192 "specified.\n", QTYPE2NAME(qtype));
1196 sbi->s_qf_names[qtype] = qname;
1197 if (strchr(sbi->s_qf_names[qtype], '/')) {
1199 "EXT4-fs: quotafile must be on "
1200 "filesystem root.\n");
1201 kfree(sbi->s_qf_names[qtype]);
1202 sbi->s_qf_names[qtype] = NULL;
1205 set_opt(sbi->s_mount_opt, QUOTA);
1207 case Opt_offusrjquota:
1210 case Opt_offgrpjquota:
1213 if ((sb_any_quota_enabled(sb) ||
1214 sb_any_quota_suspended(sb)) &&
1215 sbi->s_qf_names[qtype]) {
1216 printk(KERN_ERR "EXT4-fs: Cannot change "
1217 "journaled quota options when "
1218 "quota turned on.\n");
1222 * The space will be released later when all options
1223 * are confirmed to be correct
1225 sbi->s_qf_names[qtype] = NULL;
1227 case Opt_jqfmt_vfsold:
1228 qfmt = QFMT_VFS_OLD;
1230 case Opt_jqfmt_vfsv0:
1233 if ((sb_any_quota_enabled(sb) ||
1234 sb_any_quota_suspended(sb)) &&
1235 sbi->s_jquota_fmt != qfmt) {
1236 printk(KERN_ERR "EXT4-fs: Cannot change "
1237 "journaled quota options when "
1238 "quota turned on.\n");
1241 sbi->s_jquota_fmt = qfmt;
1245 set_opt(sbi->s_mount_opt, QUOTA);
1246 set_opt(sbi->s_mount_opt, USRQUOTA);
1249 set_opt(sbi->s_mount_opt, QUOTA);
1250 set_opt(sbi->s_mount_opt, GRPQUOTA);
1253 if (sb_any_quota_enabled(sb)) {
1254 printk(KERN_ERR "EXT4-fs: Cannot change quota "
1255 "options when quota turned on.\n");
1258 clear_opt(sbi->s_mount_opt, QUOTA);
1259 clear_opt(sbi->s_mount_opt, USRQUOTA);
1260 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1267 "EXT4-fs: quota options not supported.\n");
1271 case Opt_offusrjquota:
1272 case Opt_offgrpjquota:
1273 case Opt_jqfmt_vfsold:
1274 case Opt_jqfmt_vfsv0:
1276 "EXT4-fs: journaled quota options not "
1283 set_opt(sbi->s_mount_opt, ABORT);
1286 if (match_int(&args[0], &option))
1289 set_opt(sbi->s_mount_opt, BARRIER);
1291 clear_opt(sbi->s_mount_opt, BARRIER);
1297 printk("EXT4-fs: resize option only available "
1301 if (match_int(&args[0], &option) != 0)
1303 *n_blocks_count = option;
1306 set_opt(sbi->s_mount_opt, NOBH);
1309 clear_opt(sbi->s_mount_opt, NOBH);
1312 set_opt (sbi->s_mount_opt, EXTENTS);
1315 clear_opt (sbi->s_mount_opt, EXTENTS);
1318 set_opt(sbi->s_mount_opt, I_VERSION);
1319 sb->s_flags |= MS_I_VERSION;
1322 set_opt(sbi->s_mount_opt, MBALLOC);
1325 clear_opt(sbi->s_mount_opt, MBALLOC);
1328 if (match_int(&args[0], &option))
1332 sbi->s_stripe = option;
1336 "EXT4-fs: Unrecognized mount option \"%s\" "
1337 "or missing value\n", p);
1342 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1343 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
1344 sbi->s_qf_names[USRQUOTA])
1345 clear_opt(sbi->s_mount_opt, USRQUOTA);
1347 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
1348 sbi->s_qf_names[GRPQUOTA])
1349 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1351 if ((sbi->s_qf_names[USRQUOTA] &&
1352 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
1353 (sbi->s_qf_names[GRPQUOTA] &&
1354 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1355 printk(KERN_ERR "EXT4-fs: old and new quota "
1356 "format mixing.\n");
1360 if (!sbi->s_jquota_fmt) {
1361 printk(KERN_ERR "EXT4-fs: journaled quota format "
1362 "not specified.\n");
1366 if (sbi->s_jquota_fmt) {
1367 printk(KERN_ERR "EXT4-fs: journaled quota format "
1368 "specified with no journaling "
1377 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1380 struct ext4_sb_info *sbi = EXT4_SB(sb);
1383 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1384 printk (KERN_ERR "EXT4-fs warning: revision level too high, "
1385 "forcing read-only mode\n");
1390 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1391 printk (KERN_WARNING "EXT4-fs warning: mounting unchecked fs, "
1392 "running e2fsck is recommended\n");
1393 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1394 printk (KERN_WARNING
1395 "EXT4-fs warning: mounting fs with errors, "
1396 "running e2fsck is recommended\n");
1397 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1398 le16_to_cpu(es->s_mnt_count) >=
1399 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1400 printk (KERN_WARNING
1401 "EXT4-fs warning: maximal mount count reached, "
1402 "running e2fsck is recommended\n");
1403 else if (le32_to_cpu(es->s_checkinterval) &&
1404 (le32_to_cpu(es->s_lastcheck) +
1405 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1406 printk (KERN_WARNING
1407 "EXT4-fs warning: checktime reached, "
1408 "running e2fsck is recommended\n");
1410 /* @@@ We _will_ want to clear the valid bit if we find
1411 * inconsistencies, to force a fsck at reboot. But for
1412 * a plain journaled filesystem we can keep it set as
1415 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1417 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1418 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1419 le16_add_cpu(&es->s_mnt_count, 1);
1420 es->s_mtime = cpu_to_le32(get_seconds());
1421 ext4_update_dynamic_rev(sb);
1422 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1424 ext4_commit_super(sb, es, 1);
1425 if (test_opt(sb, DEBUG))
1426 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%lu, "
1427 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1429 sbi->s_groups_count,
1430 EXT4_BLOCKS_PER_GROUP(sb),
1431 EXT4_INODES_PER_GROUP(sb),
1434 printk(KERN_INFO "EXT4 FS on %s, ", sb->s_id);
1435 if (EXT4_SB(sb)->s_journal->j_inode == NULL) {
1436 char b[BDEVNAME_SIZE];
1438 printk("external journal on %s\n",
1439 bdevname(EXT4_SB(sb)->s_journal->j_dev, b));
1441 printk("internal journal\n");
1446 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1447 struct ext4_group_desc *gdp)
1451 if (sbi->s_es->s_feature_ro_compat &
1452 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1453 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1454 __le32 le_group = cpu_to_le32(block_group);
1456 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1457 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1458 crc = crc16(crc, (__u8 *)gdp, offset);
1459 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1460 /* for checksum of struct ext4_group_desc do the rest...*/
1461 if ((sbi->s_es->s_feature_incompat &
1462 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1463 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1464 crc = crc16(crc, (__u8 *)gdp + offset,
1465 le16_to_cpu(sbi->s_es->s_desc_size) -
1469 return cpu_to_le16(crc);
1472 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1473 struct ext4_group_desc *gdp)
1475 if ((sbi->s_es->s_feature_ro_compat &
1476 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1477 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1483 /* Called at mount-time, super-block is locked */
1484 static int ext4_check_descriptors(struct super_block *sb)
1486 struct ext4_sb_info *sbi = EXT4_SB(sb);
1487 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1488 ext4_fsblk_t last_block;
1489 ext4_fsblk_t block_bitmap;
1490 ext4_fsblk_t inode_bitmap;
1491 ext4_fsblk_t inode_table;
1492 int flexbg_flag = 0;
1495 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1498 ext4_debug ("Checking group descriptors");
1500 for (i = 0; i < sbi->s_groups_count; i++) {
1501 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1503 if (i == sbi->s_groups_count - 1 || flexbg_flag)
1504 last_block = ext4_blocks_count(sbi->s_es) - 1;
1506 last_block = first_block +
1507 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1509 block_bitmap = ext4_block_bitmap(sb, gdp);
1510 if (block_bitmap < first_block || block_bitmap > last_block)
1512 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1513 "Block bitmap for group %lu not in group "
1514 "(block %llu)!", i, block_bitmap);
1517 inode_bitmap = ext4_inode_bitmap(sb, gdp);
1518 if (inode_bitmap < first_block || inode_bitmap > last_block)
1520 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1521 "Inode bitmap for group %lu not in group "
1522 "(block %llu)!", i, inode_bitmap);
1525 inode_table = ext4_inode_table(sb, gdp);
1526 if (inode_table < first_block ||
1527 inode_table + sbi->s_itb_per_group - 1 > last_block)
1529 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1530 "Inode table for group %lu not in group "
1531 "(block %llu)!", i, inode_table);
1534 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1535 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1536 "Checksum for group %lu failed (%u!=%u)\n",
1537 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1538 gdp)), le16_to_cpu(gdp->bg_checksum));
1542 first_block += EXT4_BLOCKS_PER_GROUP(sb);
1545 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1546 sbi->s_es->s_free_inodes_count=cpu_to_le32(ext4_count_free_inodes(sb));
1550 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1551 * the superblock) which were deleted from all directories, but held open by
1552 * a process at the time of a crash. We walk the list and try to delete these
1553 * inodes at recovery time (only with a read-write filesystem).
1555 * In order to keep the orphan inode chain consistent during traversal (in
1556 * case of crash during recovery), we link each inode into the superblock
1557 * orphan list_head and handle it the same way as an inode deletion during
1558 * normal operation (which journals the operations for us).
1560 * We only do an iget() and an iput() on each inode, which is very safe if we
1561 * accidentally point at an in-use or already deleted inode. The worst that
1562 * can happen in this case is that we get a "bit already cleared" message from
1563 * ext4_free_inode(). The only reason we would point at a wrong inode is if
1564 * e2fsck was run on this filesystem, and it must have already done the orphan
1565 * inode cleanup for us, so we can safely abort without any further action.
1567 static void ext4_orphan_cleanup (struct super_block * sb,
1568 struct ext4_super_block * es)
1570 unsigned int s_flags = sb->s_flags;
1571 int nr_orphans = 0, nr_truncates = 0;
1575 if (!es->s_last_orphan) {
1576 jbd_debug(4, "no orphan inodes to clean up\n");
1580 if (bdev_read_only(sb->s_bdev)) {
1581 printk(KERN_ERR "EXT4-fs: write access "
1582 "unavailable, skipping orphan cleanup.\n");
1586 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1587 if (es->s_last_orphan)
1588 jbd_debug(1, "Errors on filesystem, "
1589 "clearing orphan list.\n");
1590 es->s_last_orphan = 0;
1591 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1595 if (s_flags & MS_RDONLY) {
1596 printk(KERN_INFO "EXT4-fs: %s: orphan cleanup on readonly fs\n",
1598 sb->s_flags &= ~MS_RDONLY;
1601 /* Needed for iput() to work correctly and not trash data */
1602 sb->s_flags |= MS_ACTIVE;
1603 /* Turn on quotas so that they are updated correctly */
1604 for (i = 0; i < MAXQUOTAS; i++) {
1605 if (EXT4_SB(sb)->s_qf_names[i]) {
1606 int ret = ext4_quota_on_mount(sb, i);
1609 "EXT4-fs: Cannot turn on journaled "
1610 "quota: error %d\n", ret);
1615 while (es->s_last_orphan) {
1616 struct inode *inode;
1618 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1619 if (IS_ERR(inode)) {
1620 es->s_last_orphan = 0;
1624 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
1626 if (inode->i_nlink) {
1628 "%s: truncating inode %lu to %Ld bytes\n",
1629 __func__, inode->i_ino, inode->i_size);
1630 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1631 inode->i_ino, inode->i_size);
1632 ext4_truncate(inode);
1636 "%s: deleting unreferenced inode %lu\n",
1637 __func__, inode->i_ino);
1638 jbd_debug(2, "deleting unreferenced inode %lu\n",
1642 iput(inode); /* The delete magic happens here! */
1645 #define PLURAL(x) (x), ((x)==1) ? "" : "s"
1648 printk(KERN_INFO "EXT4-fs: %s: %d orphan inode%s deleted\n",
1649 sb->s_id, PLURAL(nr_orphans));
1651 printk(KERN_INFO "EXT4-fs: %s: %d truncate%s cleaned up\n",
1652 sb->s_id, PLURAL(nr_truncates));
1654 /* Turn quotas off */
1655 for (i = 0; i < MAXQUOTAS; i++) {
1656 if (sb_dqopt(sb)->files[i])
1657 vfs_quota_off(sb, i, 0);
1660 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1663 * Maximal extent format file size.
1664 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1665 * extent format containers, within a sector_t, and within i_blocks
1666 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1667 * so that won't be a limiting factor.
1669 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1671 static loff_t ext4_max_size(int blkbits)
1674 loff_t upper_limit = MAX_LFS_FILESIZE;
1676 /* small i_blocks in vfs inode? */
1677 if (sizeof(blkcnt_t) < sizeof(u64)) {
1679 * CONFIG_LSF is not enabled implies the inode
1680 * i_block represent total blocks in 512 bytes
1681 * 32 == size of vfs inode i_blocks * 8
1683 upper_limit = (1LL << 32) - 1;
1685 /* total blocks in file system block size */
1686 upper_limit >>= (blkbits - 9);
1687 upper_limit <<= blkbits;
1690 /* 32-bit extent-start container, ee_block */
1695 /* Sanity check against vm- & vfs- imposed limits */
1696 if (res > upper_limit)
1703 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
1704 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1705 * We need to be 1 filesystem block less than the 2^48 sector limit.
1707 static loff_t ext4_max_bitmap_size(int bits)
1709 loff_t res = EXT4_NDIR_BLOCKS;
1712 /* This is calculated to be the largest file size for a
1713 * dense, bitmapped file such that the total number of
1714 * sectors in the file, including data and all indirect blocks,
1715 * does not exceed 2^48 -1
1716 * __u32 i_blocks_lo and _u16 i_blocks_high representing the
1717 * total number of 512 bytes blocks of the file
1720 if (sizeof(blkcnt_t) < sizeof(u64)) {
1722 * CONFIG_LSF is not enabled implies the inode
1723 * i_block represent total blocks in 512 bytes
1724 * 32 == size of vfs inode i_blocks * 8
1726 upper_limit = (1LL << 32) - 1;
1728 /* total blocks in file system block size */
1729 upper_limit >>= (bits - 9);
1733 * We use 48 bit ext4_inode i_blocks
1734 * With EXT4_HUGE_FILE_FL set the i_blocks
1735 * represent total number of blocks in
1736 * file system block size
1738 upper_limit = (1LL << 48) - 1;
1742 /* indirect blocks */
1744 /* double indirect blocks */
1745 meta_blocks += 1 + (1LL << (bits-2));
1746 /* tripple indirect blocks */
1747 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1749 upper_limit -= meta_blocks;
1750 upper_limit <<= bits;
1752 res += 1LL << (bits-2);
1753 res += 1LL << (2*(bits-2));
1754 res += 1LL << (3*(bits-2));
1756 if (res > upper_limit)
1759 if (res > MAX_LFS_FILESIZE)
1760 res = MAX_LFS_FILESIZE;
1765 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
1766 ext4_fsblk_t logical_sb_block, int nr)
1768 struct ext4_sb_info *sbi = EXT4_SB(sb);
1769 ext4_group_t bg, first_meta_bg;
1772 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1774 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
1776 return logical_sb_block + nr + 1;
1777 bg = sbi->s_desc_per_block * nr;
1778 if (ext4_bg_has_super(sb, bg))
1780 return (has_super + ext4_group_first_block_no(sb, bg));
1784 * ext4_get_stripe_size: Get the stripe size.
1785 * @sbi: In memory super block info
1787 * If we have specified it via mount option, then
1788 * use the mount option value. If the value specified at mount time is
1789 * greater than the blocks per group use the super block value.
1790 * If the super block value is greater than blocks per group return 0.
1791 * Allocator needs it be less than blocks per group.
1794 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
1796 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
1797 unsigned long stripe_width =
1798 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
1800 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
1801 return sbi->s_stripe;
1803 if (stripe_width <= sbi->s_blocks_per_group)
1804 return stripe_width;
1806 if (stride <= sbi->s_blocks_per_group)
1812 static int ext4_fill_super (struct super_block *sb, void *data, int silent)
1813 __releases(kernel_sem)
1814 __acquires(kernel_sem)
1817 struct buffer_head * bh;
1818 struct ext4_super_block *es = NULL;
1819 struct ext4_sb_info *sbi;
1821 ext4_fsblk_t sb_block = get_sb_block(&data);
1822 ext4_fsblk_t logical_sb_block;
1823 unsigned long offset = 0;
1824 unsigned int journal_inum = 0;
1825 unsigned long journal_devnum = 0;
1826 unsigned long def_mount_opts;
1837 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1840 sb->s_fs_info = sbi;
1841 sbi->s_mount_opt = 0;
1842 sbi->s_resuid = EXT4_DEF_RESUID;
1843 sbi->s_resgid = EXT4_DEF_RESGID;
1844 sbi->s_sb_block = sb_block;
1848 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
1850 printk(KERN_ERR "EXT4-fs: unable to set blocksize\n");
1854 if (!sb_set_blocksize(sb, blocksize)) {
1855 printk(KERN_ERR "EXT4-fs: bad blocksize %d.\n", blocksize);
1860 * The ext4 superblock will not be buffer aligned for other than 1kB
1861 * block sizes. We need to calculate the offset from buffer start.
1863 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
1864 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
1865 offset = do_div(logical_sb_block, blocksize);
1867 logical_sb_block = sb_block;
1870 if (!(bh = sb_bread(sb, logical_sb_block))) {
1871 printk (KERN_ERR "EXT4-fs: unable to read superblock\n");
1875 * Note: s_es must be initialized as soon as possible because
1876 * some ext4 macro-instructions depend on its value
1878 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
1880 sb->s_magic = le16_to_cpu(es->s_magic);
1881 if (sb->s_magic != EXT4_SUPER_MAGIC)
1884 /* Set defaults before we parse the mount options */
1885 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1886 if (def_mount_opts & EXT4_DEFM_DEBUG)
1887 set_opt(sbi->s_mount_opt, DEBUG);
1888 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
1889 set_opt(sbi->s_mount_opt, GRPID);
1890 if (def_mount_opts & EXT4_DEFM_UID16)
1891 set_opt(sbi->s_mount_opt, NO_UID32);
1892 #ifdef CONFIG_EXT4DEV_FS_XATTR
1893 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
1894 set_opt(sbi->s_mount_opt, XATTR_USER);
1896 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
1897 if (def_mount_opts & EXT4_DEFM_ACL)
1898 set_opt(sbi->s_mount_opt, POSIX_ACL);
1900 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
1901 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
1902 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
1903 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
1904 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
1905 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
1907 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
1908 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1909 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
1910 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1912 set_opt(sbi->s_mount_opt, ERRORS_RO);
1914 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1915 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1917 set_opt(sbi->s_mount_opt, RESERVATION);
1918 set_opt(sbi->s_mount_opt, BARRIER);
1921 * turn on extents feature by default in ext4 filesystem
1922 * User -o noextents to turn it off
1924 set_opt(sbi->s_mount_opt, EXTENTS);
1926 * turn on mballoc feature by default in ext4 filesystem
1927 * User -o nomballoc to turn it off
1929 set_opt(sbi->s_mount_opt, MBALLOC);
1931 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1935 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1936 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1938 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
1939 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
1940 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1941 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
1943 "EXT4-fs warning: feature flags set on rev 0 fs, "
1944 "running e2fsck is recommended\n");
1947 * Since ext4 is still considered development code, we require
1948 * that the TEST_FILESYS flag in s->flags be set.
1950 if (!(le32_to_cpu(es->s_flags) & EXT2_FLAGS_TEST_FILESYS)) {
1951 printk(KERN_WARNING "EXT4-fs: %s: not marked "
1952 "OK to use with test code.\n", sb->s_id);
1957 * Check feature flags regardless of the revision level, since we
1958 * previously didn't change the revision level when setting the flags,
1959 * so there is a chance incompat flags are set on a rev 0 filesystem.
1961 features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
1963 printk(KERN_ERR "EXT4-fs: %s: couldn't mount because of "
1964 "unsupported optional features (%x).\n",
1965 sb->s_id, le32_to_cpu(features));
1968 features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
1969 if (!(sb->s_flags & MS_RDONLY) && features) {
1970 printk(KERN_ERR "EXT4-fs: %s: couldn't mount RDWR because of "
1971 "unsupported optional features (%x).\n",
1972 sb->s_id, le32_to_cpu(features));
1975 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
1977 * Large file size enabled file system can only be
1978 * mount if kernel is build with CONFIG_LSF
1980 if (sizeof(root->i_blocks) < sizeof(u64) &&
1981 !(sb->s_flags & MS_RDONLY)) {
1982 printk(KERN_ERR "EXT4-fs: %s: Filesystem with huge "
1983 "files cannot be mounted read-write "
1984 "without CONFIG_LSF.\n", sb->s_id);
1988 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
1990 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
1991 blocksize > EXT4_MAX_BLOCK_SIZE) {
1993 "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
1994 blocksize, sb->s_id);
1998 if (sb->s_blocksize != blocksize) {
2000 /* Validate the filesystem blocksize */
2001 if (!sb_set_blocksize(sb, blocksize)) {
2002 printk(KERN_ERR "EXT4-fs: bad block size %d.\n",
2008 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2009 offset = do_div(logical_sb_block, blocksize);
2010 bh = sb_bread(sb, logical_sb_block);
2013 "EXT4-fs: Can't read superblock on 2nd try.\n");
2016 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
2018 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2020 "EXT4-fs: Magic mismatch, very weird !\n");
2025 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits);
2026 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits);
2028 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2029 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2030 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
2032 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2033 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
2034 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
2035 (!is_power_of_2(sbi->s_inode_size)) ||
2036 (sbi->s_inode_size > blocksize)) {
2038 "EXT4-fs: unsupported inode size: %d\n",
2042 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2043 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
2045 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2046 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
2047 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
2048 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
2049 !is_power_of_2(sbi->s_desc_size)) {
2051 "EXT4-fs: unsupported descriptor size %lu\n",
2056 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
2057 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
2058 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
2059 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
2061 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
2062 if (sbi->s_inodes_per_block == 0)
2064 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2065 sbi->s_inodes_per_block;
2066 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
2068 sbi->s_mount_state = le16_to_cpu(es->s_state);
2069 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2070 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2071 for (i=0; i < 4; i++)
2072 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2073 sbi->s_def_hash_version = es->s_def_hash_version;
2075 if (sbi->s_blocks_per_group > blocksize * 8) {
2077 "EXT4-fs: #blocks per group too big: %lu\n",
2078 sbi->s_blocks_per_group);
2081 if (sbi->s_inodes_per_group > blocksize * 8) {
2083 "EXT4-fs: #inodes per group too big: %lu\n",
2084 sbi->s_inodes_per_group);
2088 if (ext4_blocks_count(es) >
2089 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
2090 printk(KERN_ERR "EXT4-fs: filesystem on %s:"
2091 " too large to mount safely\n", sb->s_id);
2092 if (sizeof(sector_t) < 8)
2093 printk(KERN_WARNING "EXT4-fs: CONFIG_LBD not "
2098 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2101 /* ensure blocks_count calculation below doesn't sign-extend */
2102 if (ext4_blocks_count(es) + EXT4_BLOCKS_PER_GROUP(sb) <
2103 le32_to_cpu(es->s_first_data_block) + 1) {
2104 printk(KERN_WARNING "EXT4-fs: bad geometry: block count %llu, "
2105 "first data block %u, blocks per group %lu\n",
2106 ext4_blocks_count(es),
2107 le32_to_cpu(es->s_first_data_block),
2108 EXT4_BLOCKS_PER_GROUP(sb));
2111 blocks_count = (ext4_blocks_count(es) -
2112 le32_to_cpu(es->s_first_data_block) +
2113 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2114 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2115 sbi->s_groups_count = blocks_count;
2116 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2117 EXT4_DESC_PER_BLOCK(sb);
2118 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
2120 if (sbi->s_group_desc == NULL) {
2121 printk (KERN_ERR "EXT4-fs: not enough memory\n");
2125 bgl_lock_init(&sbi->s_blockgroup_lock);
2127 for (i = 0; i < db_count; i++) {
2128 block = descriptor_loc(sb, logical_sb_block, i);
2129 sbi->s_group_desc[i] = sb_bread(sb, block);
2130 if (!sbi->s_group_desc[i]) {
2131 printk (KERN_ERR "EXT4-fs: "
2132 "can't read group descriptor %d\n", i);
2137 if (!ext4_check_descriptors (sb)) {
2138 printk(KERN_ERR "EXT4-fs: group descriptors corrupted!\n");
2141 sbi->s_gdb_count = db_count;
2142 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2143 spin_lock_init(&sbi->s_next_gen_lock);
2145 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2146 ext4_count_free_blocks(sb));
2148 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2149 ext4_count_free_inodes(sb));
2152 err = percpu_counter_init(&sbi->s_dirs_counter,
2153 ext4_count_dirs(sb));
2156 printk(KERN_ERR "EXT4-fs: insufficient memory\n");
2160 /* per fileystem reservation list head & lock */
2161 spin_lock_init(&sbi->s_rsv_window_lock);
2162 sbi->s_rsv_window_root = RB_ROOT;
2163 /* Add a single, static dummy reservation to the start of the
2164 * reservation window list --- it gives us a placeholder for
2165 * append-at-start-of-list which makes the allocation logic
2166 * _much_ simpler. */
2167 sbi->s_rsv_window_head.rsv_start = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
2168 sbi->s_rsv_window_head.rsv_end = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
2169 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
2170 sbi->s_rsv_window_head.rsv_goal_size = 0;
2171 ext4_rsv_window_add(sb, &sbi->s_rsv_window_head);
2173 sbi->s_stripe = ext4_get_stripe_size(sbi);
2176 * set up enough so that it can read an inode
2178 sb->s_op = &ext4_sops;
2179 sb->s_export_op = &ext4_export_ops;
2180 sb->s_xattr = ext4_xattr_handlers;
2182 sb->s_qcop = &ext4_qctl_operations;
2183 sb->dq_op = &ext4_quota_operations;
2185 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2189 needs_recovery = (es->s_last_orphan != 0 ||
2190 EXT4_HAS_INCOMPAT_FEATURE(sb,
2191 EXT4_FEATURE_INCOMPAT_RECOVER));
2194 * The first inode we look at is the journal inode. Don't try
2195 * root first: it may be modified in the journal!
2197 if (!test_opt(sb, NOLOAD) &&
2198 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2199 if (ext4_load_journal(sb, es, journal_devnum))
2201 if (!(sb->s_flags & MS_RDONLY) &&
2202 EXT4_SB(sb)->s_journal->j_failed_commit) {
2203 printk(KERN_CRIT "EXT4-fs error (device %s): "
2204 "ext4_fill_super: Journal transaction "
2205 "%u is corrupt\n", sb->s_id,
2206 EXT4_SB(sb)->s_journal->j_failed_commit);
2207 if (test_opt (sb, ERRORS_RO)) {
2209 "Mounting filesystem read-only\n");
2210 sb->s_flags |= MS_RDONLY;
2211 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2212 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2214 if (test_opt(sb, ERRORS_PANIC)) {
2215 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2216 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2217 ext4_commit_super(sb, es, 1);
2219 "EXT4-fs (device %s): mount failed\n",
2224 } else if (journal_inum) {
2225 if (ext4_create_journal(sb, es, journal_inum))
2230 "ext4: No journal on filesystem on %s\n",
2235 if (ext4_blocks_count(es) > 0xffffffffULL &&
2236 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2237 JBD2_FEATURE_INCOMPAT_64BIT)) {
2238 printk(KERN_ERR "ext4: Failed to set 64-bit journal feature\n");
2242 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2243 jbd2_journal_set_features(sbi->s_journal,
2244 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2245 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2246 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2247 jbd2_journal_set_features(sbi->s_journal,
2248 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2249 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2250 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2252 jbd2_journal_clear_features(sbi->s_journal,
2253 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2254 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2257 /* We have now updated the journal if required, so we can
2258 * validate the data journaling mode. */
2259 switch (test_opt(sb, DATA_FLAGS)) {
2261 /* No mode set, assume a default based on the journal
2262 * capabilities: ORDERED_DATA if the journal can
2263 * cope, else JOURNAL_DATA
2265 if (jbd2_journal_check_available_features
2266 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
2267 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2269 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2272 case EXT4_MOUNT_ORDERED_DATA:
2273 case EXT4_MOUNT_WRITEBACK_DATA:
2274 if (!jbd2_journal_check_available_features
2275 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
2276 printk(KERN_ERR "EXT4-fs: Journal does not support "
2277 "requested data journaling mode\n");
2284 if (test_opt(sb, NOBH)) {
2285 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2286 printk(KERN_WARNING "EXT4-fs: Ignoring nobh option - "
2287 "its supported only with writeback mode\n");
2288 clear_opt(sbi->s_mount_opt, NOBH);
2292 * The jbd2_journal_load will have done any necessary log recovery,
2293 * so we can safely mount the rest of the filesystem now.
2296 root = ext4_iget(sb, EXT4_ROOT_INO);
2298 printk(KERN_ERR "EXT4-fs: get root inode failed\n");
2299 ret = PTR_ERR(root);
2302 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2304 printk(KERN_ERR "EXT4-fs: corrupt root inode, run e2fsck\n");
2307 sb->s_root = d_alloc_root(root);
2309 printk(KERN_ERR "EXT4-fs: get root dentry failed\n");
2315 ext4_setup_super (sb, es, sb->s_flags & MS_RDONLY);
2317 /* determine the minimum size of new large inodes, if present */
2318 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2319 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2320 EXT4_GOOD_OLD_INODE_SIZE;
2321 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2322 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2323 if (sbi->s_want_extra_isize <
2324 le16_to_cpu(es->s_want_extra_isize))
2325 sbi->s_want_extra_isize =
2326 le16_to_cpu(es->s_want_extra_isize);
2327 if (sbi->s_want_extra_isize <
2328 le16_to_cpu(es->s_min_extra_isize))
2329 sbi->s_want_extra_isize =
2330 le16_to_cpu(es->s_min_extra_isize);
2333 /* Check if enough inode space is available */
2334 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2335 sbi->s_inode_size) {
2336 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2337 EXT4_GOOD_OLD_INODE_SIZE;
2338 printk(KERN_INFO "EXT4-fs: required extra inode space not"
2343 * akpm: core read_super() calls in here with the superblock locked.
2344 * That deadlocks, because orphan cleanup needs to lock the superblock
2345 * in numerous places. Here we just pop the lock - it's relatively
2346 * harmless, because we are now ready to accept write_super() requests,
2347 * and aviro says that's the only reason for hanging onto the
2350 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2351 ext4_orphan_cleanup(sb, es);
2352 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
2354 printk (KERN_INFO "EXT4-fs: recovery complete.\n");
2355 ext4_mark_recovery_complete(sb, es);
2356 printk (KERN_INFO "EXT4-fs: mounted filesystem with %s data mode.\n",
2357 test_opt(sb,DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA ? "journal":
2358 test_opt(sb,DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA ? "ordered":
2362 ext4_mb_init(sb, needs_recovery);
2369 printk(KERN_ERR "VFS: Can't find ext4 filesystem on dev %s.\n",
2374 jbd2_journal_destroy(sbi->s_journal);
2376 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2377 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2378 percpu_counter_destroy(&sbi->s_dirs_counter);
2380 for (i = 0; i < db_count; i++)
2381 brelse(sbi->s_group_desc[i]);
2382 kfree(sbi->s_group_desc);
2385 for (i = 0; i < MAXQUOTAS; i++)
2386 kfree(sbi->s_qf_names[i]);
2388 ext4_blkdev_remove(sbi);
2391 sb->s_fs_info = NULL;
2398 * Setup any per-fs journal parameters now. We'll do this both on
2399 * initial mount, once the journal has been initialised but before we've
2400 * done any recovery; and again on any subsequent remount.
2402 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
2404 struct ext4_sb_info *sbi = EXT4_SB(sb);
2406 if (sbi->s_commit_interval)
2407 journal->j_commit_interval = sbi->s_commit_interval;
2408 /* We could also set up an ext4-specific default for the commit
2409 * interval here, but for now we'll just fall back to the jbd
2412 spin_lock(&journal->j_state_lock);
2413 if (test_opt(sb, BARRIER))
2414 journal->j_flags |= JBD2_BARRIER;
2416 journal->j_flags &= ~JBD2_BARRIER;
2417 spin_unlock(&journal->j_state_lock);
2420 static journal_t *ext4_get_journal(struct super_block *sb,
2421 unsigned int journal_inum)
2423 struct inode *journal_inode;
2426 /* First, test for the existence of a valid inode on disk. Bad
2427 * things happen if we iget() an unused inode, as the subsequent
2428 * iput() will try to delete it. */
2430 journal_inode = ext4_iget(sb, journal_inum);
2431 if (IS_ERR(journal_inode)) {
2432 printk(KERN_ERR "EXT4-fs: no journal found.\n");
2435 if (!journal_inode->i_nlink) {
2436 make_bad_inode(journal_inode);
2437 iput(journal_inode);
2438 printk(KERN_ERR "EXT4-fs: journal inode is deleted.\n");
2442 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
2443 journal_inode, journal_inode->i_size);
2444 if (!S_ISREG(journal_inode->i_mode)) {
2445 printk(KERN_ERR "EXT4-fs: invalid journal inode.\n");
2446 iput(journal_inode);
2450 journal = jbd2_journal_init_inode(journal_inode);
2452 printk(KERN_ERR "EXT4-fs: Could not load journal inode\n");
2453 iput(journal_inode);
2456 journal->j_private = sb;
2457 ext4_init_journal_params(sb, journal);
2461 static journal_t *ext4_get_dev_journal(struct super_block *sb,
2464 struct buffer_head * bh;
2468 int hblock, blocksize;
2469 ext4_fsblk_t sb_block;
2470 unsigned long offset;
2471 struct ext4_super_block * es;
2472 struct block_device *bdev;
2474 bdev = ext4_blkdev_get(j_dev);
2478 if (bd_claim(bdev, sb)) {
2480 "EXT4: failed to claim external journal device.\n");
2485 blocksize = sb->s_blocksize;
2486 hblock = bdev_hardsect_size(bdev);
2487 if (blocksize < hblock) {
2489 "EXT4-fs: blocksize too small for journal device.\n");
2493 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
2494 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
2495 set_blocksize(bdev, blocksize);
2496 if (!(bh = __bread(bdev, sb_block, blocksize))) {
2497 printk(KERN_ERR "EXT4-fs: couldn't read superblock of "
2498 "external journal\n");
2502 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2503 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
2504 !(le32_to_cpu(es->s_feature_incompat) &
2505 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2506 printk(KERN_ERR "EXT4-fs: external journal has "
2507 "bad superblock\n");
2512 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2513 printk(KERN_ERR "EXT4-fs: journal UUID does not match\n");
2518 len = ext4_blocks_count(es);
2519 start = sb_block + 1;
2520 brelse(bh); /* we're done with the superblock */
2522 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
2523 start, len, blocksize);
2525 printk(KERN_ERR "EXT4-fs: failed to create device journal\n");
2528 journal->j_private = sb;
2529 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2530 wait_on_buffer(journal->j_sb_buffer);
2531 if (!buffer_uptodate(journal->j_sb_buffer)) {
2532 printk(KERN_ERR "EXT4-fs: I/O error on journal device\n");
2535 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
2536 printk(KERN_ERR "EXT4-fs: External journal has more than one "
2537 "user (unsupported) - %d\n",
2538 be32_to_cpu(journal->j_superblock->s_nr_users));
2541 EXT4_SB(sb)->journal_bdev = bdev;
2542 ext4_init_journal_params(sb, journal);
2545 jbd2_journal_destroy(journal);
2547 ext4_blkdev_put(bdev);
2551 static int ext4_load_journal(struct super_block *sb,
2552 struct ext4_super_block *es,
2553 unsigned long journal_devnum)
2556 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2559 int really_read_only;
2561 if (journal_devnum &&
2562 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2563 printk(KERN_INFO "EXT4-fs: external journal device major/minor "
2564 "numbers have changed\n");
2565 journal_dev = new_decode_dev(journal_devnum);
2567 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2569 really_read_only = bdev_read_only(sb->s_bdev);
2572 * Are we loading a blank journal or performing recovery after a
2573 * crash? For recovery, we need to check in advance whether we
2574 * can get read-write access to the device.
2577 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2578 if (sb->s_flags & MS_RDONLY) {
2579 printk(KERN_INFO "EXT4-fs: INFO: recovery "
2580 "required on readonly filesystem.\n");
2581 if (really_read_only) {
2582 printk(KERN_ERR "EXT4-fs: write access "
2583 "unavailable, cannot proceed.\n");
2586 printk (KERN_INFO "EXT4-fs: write access will "
2587 "be enabled during recovery.\n");
2591 if (journal_inum && journal_dev) {
2592 printk(KERN_ERR "EXT4-fs: filesystem has both journal "
2593 "and inode journals!\n");
2598 if (!(journal = ext4_get_journal(sb, journal_inum)))
2601 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
2605 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2606 err = jbd2_journal_update_format(journal);
2608 printk(KERN_ERR "EXT4-fs: error updating journal.\n");
2609 jbd2_journal_destroy(journal);
2614 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
2615 err = jbd2_journal_wipe(journal, !really_read_only);
2617 err = jbd2_journal_load(journal);
2620 printk(KERN_ERR "EXT4-fs: error loading journal.\n");
2621 jbd2_journal_destroy(journal);
2625 EXT4_SB(sb)->s_journal = journal;
2626 ext4_clear_journal_err(sb, es);
2628 if (journal_devnum &&
2629 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2630 es->s_journal_dev = cpu_to_le32(journal_devnum);
2633 /* Make sure we flush the recovery flag to disk. */
2634 ext4_commit_super(sb, es, 1);
2640 static int ext4_create_journal(struct super_block * sb,
2641 struct ext4_super_block * es,
2642 unsigned int journal_inum)
2647 if (sb->s_flags & MS_RDONLY) {
2648 printk(KERN_ERR "EXT4-fs: readonly filesystem when trying to "
2649 "create journal.\n");
2653 journal = ext4_get_journal(sb, journal_inum);
2657 printk(KERN_INFO "EXT4-fs: creating new journal on inode %u\n",
2660 err = jbd2_journal_create(journal);
2662 printk(KERN_ERR "EXT4-fs: error creating journal.\n");
2663 jbd2_journal_destroy(journal);
2667 EXT4_SB(sb)->s_journal = journal;
2669 ext4_update_dynamic_rev(sb);
2670 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2671 EXT4_SET_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL);
2673 es->s_journal_inum = cpu_to_le32(journal_inum);
2676 /* Make sure we flush the recovery flag to disk. */
2677 ext4_commit_super(sb, es, 1);
2682 static void ext4_commit_super (struct super_block * sb,
2683 struct ext4_super_block * es,
2686 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
2690 es->s_wtime = cpu_to_le32(get_seconds());
2691 ext4_free_blocks_count_set(es, ext4_count_free_blocks(sb));
2692 es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
2693 BUFFER_TRACE(sbh, "marking dirty");
2694 mark_buffer_dirty(sbh);
2696 sync_dirty_buffer(sbh);
2701 * Have we just finished recovery? If so, and if we are mounting (or
2702 * remounting) the filesystem readonly, then we will end up with a
2703 * consistent fs on disk. Record that fact.
2705 static void ext4_mark_recovery_complete(struct super_block * sb,
2706 struct ext4_super_block * es)
2708 journal_t *journal = EXT4_SB(sb)->s_journal;
2710 jbd2_journal_lock_updates(journal);
2711 jbd2_journal_flush(journal);
2713 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
2714 sb->s_flags & MS_RDONLY) {
2715 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2717 ext4_commit_super(sb, es, 1);
2720 jbd2_journal_unlock_updates(journal);
2724 * If we are mounting (or read-write remounting) a filesystem whose journal
2725 * has recorded an error from a previous lifetime, move that error to the
2726 * main filesystem now.
2728 static void ext4_clear_journal_err(struct super_block * sb,
2729 struct ext4_super_block * es)
2735 journal = EXT4_SB(sb)->s_journal;
2738 * Now check for any error status which may have been recorded in the
2739 * journal by a prior ext4_error() or ext4_abort()
2742 j_errno = jbd2_journal_errno(journal);
2746 errstr = ext4_decode_error(sb, j_errno, nbuf);
2747 ext4_warning(sb, __func__, "Filesystem error recorded "
2748 "from previous mount: %s", errstr);
2749 ext4_warning(sb, __func__, "Marking fs in need of "
2750 "filesystem check.");
2752 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2753 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2754 ext4_commit_super (sb, es, 1);
2756 jbd2_journal_clear_err(journal);
2761 * Force the running and committing transactions to commit,
2762 * and wait on the commit.
2764 int ext4_force_commit(struct super_block *sb)
2769 if (sb->s_flags & MS_RDONLY)
2772 journal = EXT4_SB(sb)->s_journal;
2774 ret = ext4_journal_force_commit(journal);
2779 * Ext4 always journals updates to the superblock itself, so we don't
2780 * have to propagate any other updates to the superblock on disk at this
2781 * point. Just start an async writeback to get the buffers on their way
2784 * This implicitly triggers the writebehind on sync().
2787 static void ext4_write_super (struct super_block * sb)
2789 if (mutex_trylock(&sb->s_lock) != 0)
2794 static int ext4_sync_fs(struct super_block *sb, int wait)
2799 if (jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, &target)) {
2801 jbd2_log_wait_commit(EXT4_SB(sb)->s_journal, target);
2807 * LVM calls this function before a (read-only) snapshot is created. This
2808 * gives us a chance to flush the journal completely and mark the fs clean.
2810 static void ext4_write_super_lockfs(struct super_block *sb)
2814 if (!(sb->s_flags & MS_RDONLY)) {
2815 journal_t *journal = EXT4_SB(sb)->s_journal;
2817 /* Now we set up the journal barrier. */
2818 jbd2_journal_lock_updates(journal);
2819 jbd2_journal_flush(journal);
2821 /* Journal blocked and flushed, clear needs_recovery flag. */
2822 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2823 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
2828 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2829 * flag here, even though the filesystem is not technically dirty yet.
2831 static void ext4_unlockfs(struct super_block *sb)
2833 if (!(sb->s_flags & MS_RDONLY)) {
2835 /* Reser the needs_recovery flag before the fs is unlocked. */
2836 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2837 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
2839 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
2843 static int ext4_remount (struct super_block * sb, int * flags, char * data)
2845 struct ext4_super_block * es;
2846 struct ext4_sb_info *sbi = EXT4_SB(sb);
2847 ext4_fsblk_t n_blocks_count = 0;
2848 unsigned long old_sb_flags;
2849 struct ext4_mount_options old_opts;
2855 /* Store the original options */
2856 old_sb_flags = sb->s_flags;
2857 old_opts.s_mount_opt = sbi->s_mount_opt;
2858 old_opts.s_resuid = sbi->s_resuid;
2859 old_opts.s_resgid = sbi->s_resgid;
2860 old_opts.s_commit_interval = sbi->s_commit_interval;
2862 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2863 for (i = 0; i < MAXQUOTAS; i++)
2864 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2868 * Allow the "check" option to be passed as a remount option.
2870 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
2875 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT)
2876 ext4_abort(sb, __func__, "Abort forced by user");
2878 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2879 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2883 ext4_init_journal_params(sb, sbi->s_journal);
2885 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
2886 n_blocks_count > ext4_blocks_count(es)) {
2887 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) {
2892 if (*flags & MS_RDONLY) {
2894 * First of all, the unconditional stuff we have to do
2895 * to disable replay of the journal when we next remount
2897 sb->s_flags |= MS_RDONLY;
2900 * OK, test if we are remounting a valid rw partition
2901 * readonly, and if so set the rdonly flag and then
2902 * mark the partition as valid again.
2904 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
2905 (sbi->s_mount_state & EXT4_VALID_FS))
2906 es->s_state = cpu_to_le16(sbi->s_mount_state);
2909 * We have to unlock super so that we can wait for
2913 ext4_mark_recovery_complete(sb, es);
2917 if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2918 ~EXT4_FEATURE_RO_COMPAT_SUPP))) {
2919 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
2920 "remount RDWR because of unsupported "
2921 "optional features (%x).\n",
2922 sb->s_id, le32_to_cpu(ret));
2928 * If we have an unprocessed orphan list hanging
2929 * around from a previously readonly bdev mount,
2930 * require a full umount/remount for now.
2932 if (es->s_last_orphan) {
2933 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
2934 "remount RDWR because of unprocessed "
2935 "orphan inode list. Please "
2936 "umount/remount instead.\n",
2943 * Mounting a RDONLY partition read-write, so reread
2944 * and store the current valid flag. (It may have
2945 * been changed by e2fsck since we originally mounted
2948 ext4_clear_journal_err(sb, es);
2949 sbi->s_mount_state = le16_to_cpu(es->s_state);
2950 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
2952 if (!ext4_setup_super (sb, es, 0))
2953 sb->s_flags &= ~MS_RDONLY;
2957 /* Release old quota file names */
2958 for (i = 0; i < MAXQUOTAS; i++)
2959 if (old_opts.s_qf_names[i] &&
2960 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2961 kfree(old_opts.s_qf_names[i]);
2965 sb->s_flags = old_sb_flags;
2966 sbi->s_mount_opt = old_opts.s_mount_opt;
2967 sbi->s_resuid = old_opts.s_resuid;
2968 sbi->s_resgid = old_opts.s_resgid;
2969 sbi->s_commit_interval = old_opts.s_commit_interval;
2971 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
2972 for (i = 0; i < MAXQUOTAS; i++) {
2973 if (sbi->s_qf_names[i] &&
2974 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2975 kfree(sbi->s_qf_names[i]);
2976 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
2982 static int ext4_statfs (struct dentry * dentry, struct kstatfs * buf)
2984 struct super_block *sb = dentry->d_sb;
2985 struct ext4_sb_info *sbi = EXT4_SB(sb);
2986 struct ext4_super_block *es = sbi->s_es;
2989 if (test_opt(sb, MINIX_DF)) {
2990 sbi->s_overhead_last = 0;
2991 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
2992 ext4_group_t ngroups = sbi->s_groups_count, i;
2993 ext4_fsblk_t overhead = 0;
2997 * Compute the overhead (FS structures). This is constant
2998 * for a given filesystem unless the number of block groups
2999 * changes so we cache the previous value until it does.
3003 * All of the blocks before first_data_block are
3006 overhead = le32_to_cpu(es->s_first_data_block);
3009 * Add the overhead attributed to the superblock and
3010 * block group descriptors. If the sparse superblocks
3011 * feature is turned on, then not all groups have this.
3013 for (i = 0; i < ngroups; i++) {
3014 overhead += ext4_bg_has_super(sb, i) +
3015 ext4_bg_num_gdb(sb, i);
3020 * Every block group has an inode bitmap, a block
3021 * bitmap, and an inode table.
3023 overhead += ngroups * (2 + sbi->s_itb_per_group);
3024 sbi->s_overhead_last = overhead;
3026 sbi->s_blocks_last = ext4_blocks_count(es);
3029 buf->f_type = EXT4_SUPER_MAGIC;
3030 buf->f_bsize = sb->s_blocksize;
3031 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
3032 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter);
3033 ext4_free_blocks_count_set(es, buf->f_bfree);
3034 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3035 if (buf->f_bfree < ext4_r_blocks_count(es))
3037 buf->f_files = le32_to_cpu(es->s_inodes_count);
3038 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
3039 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
3040 buf->f_namelen = EXT4_NAME_LEN;
3041 fsid = le64_to_cpup((void *)es->s_uuid) ^
3042 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3043 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3044 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
3048 /* Helper function for writing quotas on sync - we need to start transaction before quota file
3049 * is locked for write. Otherwise the are possible deadlocks:
3050 * Process 1 Process 2
3051 * ext4_create() quota_sync()
3052 * jbd2_journal_start() write_dquot()
3053 * DQUOT_INIT() down(dqio_mutex)
3054 * down(dqio_mutex) jbd2_journal_start()
3060 static inline struct inode *dquot_to_inode(struct dquot *dquot)
3062 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3065 static int ext4_dquot_initialize(struct inode *inode, int type)
3070 /* We may create quota structure so we need to reserve enough blocks */
3071 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_INIT_BLOCKS(inode->i_sb));
3073 return PTR_ERR(handle);
3074 ret = dquot_initialize(inode, type);
3075 err = ext4_journal_stop(handle);
3081 static int ext4_dquot_drop(struct inode *inode)
3086 /* We may delete quota structure so we need to reserve enough blocks */
3087 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_DEL_BLOCKS(inode->i_sb));
3088 if (IS_ERR(handle)) {
3090 * We call dquot_drop() anyway to at least release references
3091 * to quota structures so that umount does not hang.
3094 return PTR_ERR(handle);
3096 ret = dquot_drop(inode);
3097 err = ext4_journal_stop(handle);
3103 static int ext4_write_dquot(struct dquot *dquot)
3107 struct inode *inode;
3109 inode = dquot_to_inode(dquot);
3110 handle = ext4_journal_start(inode,
3111 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
3113 return PTR_ERR(handle);
3114 ret = dquot_commit(dquot);
3115 err = ext4_journal_stop(handle);
3121 static int ext4_acquire_dquot(struct dquot *dquot)
3126 handle = ext4_journal_start(dquot_to_inode(dquot),
3127 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
3129 return PTR_ERR(handle);
3130 ret = dquot_acquire(dquot);
3131 err = ext4_journal_stop(handle);
3137 static int ext4_release_dquot(struct dquot *dquot)
3142 handle = ext4_journal_start(dquot_to_inode(dquot),
3143 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
3144 if (IS_ERR(handle)) {
3145 /* Release dquot anyway to avoid endless cycle in dqput() */
3146 dquot_release(dquot);
3147 return PTR_ERR(handle);
3149 ret = dquot_release(dquot);
3150 err = ext4_journal_stop(handle);
3156 static int ext4_mark_dquot_dirty(struct dquot *dquot)
3158 /* Are we journaling quotas? */
3159 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3160 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
3161 dquot_mark_dquot_dirty(dquot);
3162 return ext4_write_dquot(dquot);
3164 return dquot_mark_dquot_dirty(dquot);
3168 static int ext4_write_info(struct super_block *sb, int type)
3173 /* Data block + inode block */
3174 handle = ext4_journal_start(sb->s_root->d_inode, 2);
3176 return PTR_ERR(handle);
3177 ret = dquot_commit_info(sb, type);
3178 err = ext4_journal_stop(handle);
3185 * Turn on quotas during mount time - we need to find
3186 * the quota file and such...
3188 static int ext4_quota_on_mount(struct super_block *sb, int type)
3190 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3191 EXT4_SB(sb)->s_jquota_fmt, type);
3195 * Standard function to be called on quota_on
3197 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
3198 char *path, int remount)
3201 struct nameidata nd;
3203 if (!test_opt(sb, QUOTA))
3205 /* When remounting, no checks are needed and in fact, path is NULL */
3207 return vfs_quota_on(sb, type, format_id, path, remount);
3209 err = path_lookup(path, LOOKUP_FOLLOW, &nd);
3213 /* Quotafile not on the same filesystem? */
3214 if (nd.path.mnt->mnt_sb != sb) {
3218 /* Journaling quota? */
3219 if (EXT4_SB(sb)->s_qf_names[type]) {
3220 /* Quotafile not of fs root? */
3221 if (nd.path.dentry->d_parent->d_inode != sb->s_root->d_inode)
3223 "EXT4-fs: Quota file not on filesystem root. "
3224 "Journaled quota will not work.\n");
3228 * When we journal data on quota file, we have to flush journal to see
3229 * all updates to the file when we bypass pagecache...
3231 if (ext4_should_journal_data(nd.path.dentry->d_inode)) {
3233 * We don't need to lock updates but journal_flush() could
3234 * otherwise be livelocked...
3236 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
3237 jbd2_journal_flush(EXT4_SB(sb)->s_journal);
3238 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3242 return vfs_quota_on(sb, type, format_id, path, remount);
3245 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3246 * acquiring the locks... As quota files are never truncated and quota code
3247 * itself serializes the operations (and noone else should touch the files)
3248 * we don't have to be afraid of races */
3249 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
3250 size_t len, loff_t off)
3252 struct inode *inode = sb_dqopt(sb)->files[type];
3253 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3255 int offset = off & (sb->s_blocksize - 1);
3258 struct buffer_head *bh;
3259 loff_t i_size = i_size_read(inode);
3263 if (off+len > i_size)
3266 while (toread > 0) {
3267 tocopy = sb->s_blocksize - offset < toread ?
3268 sb->s_blocksize - offset : toread;
3269 bh = ext4_bread(NULL, inode, blk, 0, &err);
3272 if (!bh) /* A hole? */
3273 memset(data, 0, tocopy);
3275 memcpy(data, bh->b_data+offset, tocopy);
3285 /* Write to quotafile (we know the transaction is already started and has
3286 * enough credits) */
3287 static ssize_t ext4_quota_write(struct super_block *sb, int type,
3288 const char *data, size_t len, loff_t off)
3290 struct inode *inode = sb_dqopt(sb)->files[type];
3291 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3293 int offset = off & (sb->s_blocksize - 1);
3295 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
3296 size_t towrite = len;
3297 struct buffer_head *bh;
3298 handle_t *handle = journal_current_handle();
3301 printk(KERN_WARNING "EXT4-fs: Quota write (off=%Lu, len=%Lu)"
3302 " cancelled because transaction is not started.\n",
3303 (unsigned long long)off, (unsigned long long)len);
3306 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3307 while (towrite > 0) {
3308 tocopy = sb->s_blocksize - offset < towrite ?
3309 sb->s_blocksize - offset : towrite;
3310 bh = ext4_bread(handle, inode, blk, 1, &err);
3313 if (journal_quota) {
3314 err = ext4_journal_get_write_access(handle, bh);
3321 memcpy(bh->b_data+offset, data, tocopy);
3322 flush_dcache_page(bh->b_page);
3325 err = ext4_journal_dirty_metadata(handle, bh);
3327 /* Always do at least ordered writes for quotas */
3328 err = ext4_journal_dirty_data(handle, bh);
3329 mark_buffer_dirty(bh);
3340 if (len == towrite) {
3341 mutex_unlock(&inode->i_mutex);
3344 if (inode->i_size < off+len-towrite) {
3345 i_size_write(inode, off+len-towrite);
3346 EXT4_I(inode)->i_disksize = inode->i_size;
3348 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3349 ext4_mark_inode_dirty(handle, inode);
3350 mutex_unlock(&inode->i_mutex);
3351 return len - towrite;
3356 static int ext4_get_sb(struct file_system_type *fs_type,
3357 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3359 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
3362 static struct file_system_type ext4dev_fs_type = {
3363 .owner = THIS_MODULE,
3365 .get_sb = ext4_get_sb,
3366 .kill_sb = kill_block_super,
3367 .fs_flags = FS_REQUIRES_DEV,
3370 static int __init init_ext4_fs(void)
3374 err = init_ext4_mballoc();
3378 err = init_ext4_xattr();
3381 err = init_inodecache();
3384 err = register_filesystem(&ext4dev_fs_type);
3389 destroy_inodecache();
3393 exit_ext4_mballoc();
3397 static void __exit exit_ext4_fs(void)
3399 unregister_filesystem(&ext4dev_fs_type);
3400 destroy_inodecache();
3402 exit_ext4_mballoc();
3405 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
3406 MODULE_DESCRIPTION("Fourth Extended Filesystem with extents");
3407 MODULE_LICENSE("GPL");
3408 module_init(init_ext4_fs)
3409 module_exit(exit_ext4_fs)