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 struct ext4_super_block *es = sbi->s_es;
676 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
677 def_errors = le16_to_cpu(es->s_errors);
679 if (sbi->s_sb_block != 1)
680 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
681 if (test_opt(sb, MINIX_DF))
682 seq_puts(seq, ",minixdf");
683 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
684 seq_puts(seq, ",grpid");
685 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
686 seq_puts(seq, ",nogrpid");
687 if (sbi->s_resuid != EXT4_DEF_RESUID ||
688 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
689 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
691 if (sbi->s_resgid != EXT4_DEF_RESGID ||
692 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
693 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
695 if (test_opt(sb, ERRORS_RO)) {
696 if (def_errors == EXT4_ERRORS_PANIC ||
697 def_errors == EXT4_ERRORS_CONTINUE) {
698 seq_puts(seq, ",errors=remount-ro");
701 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
702 seq_puts(seq, ",errors=continue");
703 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
704 seq_puts(seq, ",errors=panic");
705 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
706 seq_puts(seq, ",nouid32");
707 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
708 seq_puts(seq, ",debug");
709 if (test_opt(sb, OLDALLOC))
710 seq_puts(seq, ",oldalloc");
711 #ifdef CONFIG_EXT4DEV_FS_XATTR
712 if (test_opt(sb, XATTR_USER) &&
713 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
714 seq_puts(seq, ",user_xattr");
715 if (!test_opt(sb, XATTR_USER) &&
716 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
717 seq_puts(seq, ",nouser_xattr");
720 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
721 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
722 seq_puts(seq, ",acl");
723 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
724 seq_puts(seq, ",noacl");
726 if (!test_opt(sb, RESERVATION))
727 seq_puts(seq, ",noreservation");
728 if (sbi->s_commit_interval) {
729 seq_printf(seq, ",commit=%u",
730 (unsigned) (sbi->s_commit_interval / HZ));
733 * We're changing the default of barrier mount option, so
734 * let's always display its mount state so it's clear what its
737 seq_puts(seq, ",barrier=");
738 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
739 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
740 seq_puts(seq, ",journal_async_commit");
741 if (test_opt(sb, NOBH))
742 seq_puts(seq, ",nobh");
743 if (!test_opt(sb, EXTENTS))
744 seq_puts(seq, ",noextents");
745 if (!test_opt(sb, MBALLOC))
746 seq_puts(seq, ",nomballoc");
747 if (test_opt(sb, I_VERSION))
748 seq_puts(seq, ",i_version");
751 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
753 * journal mode get enabled in different ways
754 * So just print the value even if we didn't specify it
756 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
757 seq_puts(seq, ",data=journal");
758 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
759 seq_puts(seq, ",data=ordered");
760 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
761 seq_puts(seq, ",data=writeback");
763 ext4_show_quota_options(seq, sb);
768 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
769 u64 ino, u32 generation)
773 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
774 return ERR_PTR(-ESTALE);
775 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
776 return ERR_PTR(-ESTALE);
778 /* iget isn't really right if the inode is currently unallocated!!
780 * ext4_read_inode will return a bad_inode if the inode had been
781 * deleted, so we should be safe.
783 * Currently we don't know the generation for parent directory, so
784 * a generation of 0 means "accept any"
786 inode = ext4_iget(sb, ino);
788 return ERR_CAST(inode);
789 if (generation && inode->i_generation != generation) {
791 return ERR_PTR(-ESTALE);
797 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
798 int fh_len, int fh_type)
800 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
804 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
805 int fh_len, int fh_type)
807 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
812 #define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
813 #define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
815 static int ext4_dquot_initialize(struct inode *inode, int type);
816 static int ext4_dquot_drop(struct inode *inode);
817 static int ext4_write_dquot(struct dquot *dquot);
818 static int ext4_acquire_dquot(struct dquot *dquot);
819 static int ext4_release_dquot(struct dquot *dquot);
820 static int ext4_mark_dquot_dirty(struct dquot *dquot);
821 static int ext4_write_info(struct super_block *sb, int type);
822 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
823 char *path, int remount);
824 static int ext4_quota_on_mount(struct super_block *sb, int type);
825 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
826 size_t len, loff_t off);
827 static ssize_t ext4_quota_write(struct super_block *sb, int type,
828 const char *data, size_t len, loff_t off);
830 static struct dquot_operations ext4_quota_operations = {
831 .initialize = ext4_dquot_initialize,
832 .drop = ext4_dquot_drop,
833 .alloc_space = dquot_alloc_space,
834 .alloc_inode = dquot_alloc_inode,
835 .free_space = dquot_free_space,
836 .free_inode = dquot_free_inode,
837 .transfer = dquot_transfer,
838 .write_dquot = ext4_write_dquot,
839 .acquire_dquot = ext4_acquire_dquot,
840 .release_dquot = ext4_release_dquot,
841 .mark_dirty = ext4_mark_dquot_dirty,
842 .write_info = ext4_write_info
845 static struct quotactl_ops ext4_qctl_operations = {
846 .quota_on = ext4_quota_on,
847 .quota_off = vfs_quota_off,
848 .quota_sync = vfs_quota_sync,
849 .get_info = vfs_get_dqinfo,
850 .set_info = vfs_set_dqinfo,
851 .get_dqblk = vfs_get_dqblk,
852 .set_dqblk = vfs_set_dqblk
856 static const struct super_operations ext4_sops = {
857 .alloc_inode = ext4_alloc_inode,
858 .destroy_inode = ext4_destroy_inode,
859 .write_inode = ext4_write_inode,
860 .dirty_inode = ext4_dirty_inode,
861 .delete_inode = ext4_delete_inode,
862 .put_super = ext4_put_super,
863 .write_super = ext4_write_super,
864 .sync_fs = ext4_sync_fs,
865 .write_super_lockfs = ext4_write_super_lockfs,
866 .unlockfs = ext4_unlockfs,
867 .statfs = ext4_statfs,
868 .remount_fs = ext4_remount,
869 .clear_inode = ext4_clear_inode,
870 .show_options = ext4_show_options,
872 .quota_read = ext4_quota_read,
873 .quota_write = ext4_quota_write,
877 static const struct export_operations ext4_export_ops = {
878 .fh_to_dentry = ext4_fh_to_dentry,
879 .fh_to_parent = ext4_fh_to_parent,
880 .get_parent = ext4_get_parent,
884 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
885 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
886 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
887 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
888 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
889 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
890 Opt_journal_checksum, Opt_journal_async_commit,
891 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
892 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
893 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
894 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
895 Opt_grpquota, Opt_extents, Opt_noextents, Opt_i_version,
896 Opt_mballoc, Opt_nomballoc, Opt_stripe,
899 static match_table_t tokens = {
900 {Opt_bsd_df, "bsddf"},
901 {Opt_minix_df, "minixdf"},
902 {Opt_grpid, "grpid"},
903 {Opt_grpid, "bsdgroups"},
904 {Opt_nogrpid, "nogrpid"},
905 {Opt_nogrpid, "sysvgroups"},
906 {Opt_resgid, "resgid=%u"},
907 {Opt_resuid, "resuid=%u"},
909 {Opt_err_cont, "errors=continue"},
910 {Opt_err_panic, "errors=panic"},
911 {Opt_err_ro, "errors=remount-ro"},
912 {Opt_nouid32, "nouid32"},
913 {Opt_nocheck, "nocheck"},
914 {Opt_nocheck, "check=none"},
915 {Opt_debug, "debug"},
916 {Opt_oldalloc, "oldalloc"},
917 {Opt_orlov, "orlov"},
918 {Opt_user_xattr, "user_xattr"},
919 {Opt_nouser_xattr, "nouser_xattr"},
921 {Opt_noacl, "noacl"},
922 {Opt_reservation, "reservation"},
923 {Opt_noreservation, "noreservation"},
924 {Opt_noload, "noload"},
927 {Opt_commit, "commit=%u"},
928 {Opt_journal_update, "journal=update"},
929 {Opt_journal_inum, "journal=%u"},
930 {Opt_journal_dev, "journal_dev=%u"},
931 {Opt_journal_checksum, "journal_checksum"},
932 {Opt_journal_async_commit, "journal_async_commit"},
933 {Opt_abort, "abort"},
934 {Opt_data_journal, "data=journal"},
935 {Opt_data_ordered, "data=ordered"},
936 {Opt_data_writeback, "data=writeback"},
937 {Opt_offusrjquota, "usrjquota="},
938 {Opt_usrjquota, "usrjquota=%s"},
939 {Opt_offgrpjquota, "grpjquota="},
940 {Opt_grpjquota, "grpjquota=%s"},
941 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
942 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
943 {Opt_grpquota, "grpquota"},
944 {Opt_noquota, "noquota"},
945 {Opt_quota, "quota"},
946 {Opt_usrquota, "usrquota"},
947 {Opt_barrier, "barrier=%u"},
948 {Opt_extents, "extents"},
949 {Opt_noextents, "noextents"},
950 {Opt_i_version, "i_version"},
951 {Opt_mballoc, "mballoc"},
952 {Opt_nomballoc, "nomballoc"},
953 {Opt_stripe, "stripe=%u"},
954 {Opt_resize, "resize"},
958 static ext4_fsblk_t get_sb_block(void **data)
960 ext4_fsblk_t sb_block;
961 char *options = (char *) *data;
963 if (!options || strncmp(options, "sb=", 3) != 0)
964 return 1; /* Default location */
966 /*todo: use simple_strtoll with >32bit ext4 */
967 sb_block = simple_strtoul(options, &options, 0);
968 if (*options && *options != ',') {
969 printk("EXT4-fs: Invalid sb specification: %s\n",
975 *data = (void *) options;
979 static int parse_options (char *options, struct super_block *sb,
980 unsigned int *inum, unsigned long *journal_devnum,
981 ext4_fsblk_t *n_blocks_count, int is_remount)
983 struct ext4_sb_info *sbi = EXT4_SB(sb);
985 substring_t args[MAX_OPT_ARGS];
996 while ((p = strsep (&options, ",")) != NULL) {
1001 token = match_token(p, tokens, args);
1004 clear_opt (sbi->s_mount_opt, MINIX_DF);
1007 set_opt (sbi->s_mount_opt, MINIX_DF);
1010 set_opt (sbi->s_mount_opt, GRPID);
1013 clear_opt (sbi->s_mount_opt, GRPID);
1016 if (match_int(&args[0], &option))
1018 sbi->s_resuid = option;
1021 if (match_int(&args[0], &option))
1023 sbi->s_resgid = option;
1026 /* handled by get_sb_block() instead of here */
1027 /* *sb_block = match_int(&args[0]); */
1030 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1031 clear_opt (sbi->s_mount_opt, ERRORS_RO);
1032 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
1035 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1036 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1037 set_opt (sbi->s_mount_opt, ERRORS_RO);
1040 clear_opt (sbi->s_mount_opt, ERRORS_RO);
1041 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1042 set_opt (sbi->s_mount_opt, ERRORS_CONT);
1045 set_opt (sbi->s_mount_opt, NO_UID32);
1048 clear_opt (sbi->s_mount_opt, CHECK);
1051 set_opt (sbi->s_mount_opt, DEBUG);
1054 set_opt (sbi->s_mount_opt, OLDALLOC);
1057 clear_opt (sbi->s_mount_opt, OLDALLOC);
1059 #ifdef CONFIG_EXT4DEV_FS_XATTR
1060 case Opt_user_xattr:
1061 set_opt (sbi->s_mount_opt, XATTR_USER);
1063 case Opt_nouser_xattr:
1064 clear_opt (sbi->s_mount_opt, XATTR_USER);
1067 case Opt_user_xattr:
1068 case Opt_nouser_xattr:
1069 printk("EXT4 (no)user_xattr options not supported\n");
1072 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
1074 set_opt(sbi->s_mount_opt, POSIX_ACL);
1077 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1082 printk("EXT4 (no)acl options not supported\n");
1085 case Opt_reservation:
1086 set_opt(sbi->s_mount_opt, RESERVATION);
1088 case Opt_noreservation:
1089 clear_opt(sbi->s_mount_opt, RESERVATION);
1091 case Opt_journal_update:
1093 /* Eventually we will want to be able to create
1094 a journal file here. For now, only allow the
1095 user to specify an existing inode to be the
1098 printk(KERN_ERR "EXT4-fs: cannot specify "
1099 "journal on remount\n");
1102 set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
1104 case Opt_journal_inum:
1106 printk(KERN_ERR "EXT4-fs: cannot specify "
1107 "journal on remount\n");
1110 if (match_int(&args[0], &option))
1114 case Opt_journal_dev:
1116 printk(KERN_ERR "EXT4-fs: cannot specify "
1117 "journal on remount\n");
1120 if (match_int(&args[0], &option))
1122 *journal_devnum = option;
1124 case Opt_journal_checksum:
1125 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1127 case Opt_journal_async_commit:
1128 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1129 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1132 set_opt (sbi->s_mount_opt, NOLOAD);
1135 if (match_int(&args[0], &option))
1140 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1141 sbi->s_commit_interval = HZ * option;
1143 case Opt_data_journal:
1144 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1146 case Opt_data_ordered:
1147 data_opt = EXT4_MOUNT_ORDERED_DATA;
1149 case Opt_data_writeback:
1150 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1153 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
1156 "EXT4-fs: cannot change data "
1157 "mode on remount\n");
1161 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
1162 sbi->s_mount_opt |= data_opt;
1172 if ((sb_any_quota_enabled(sb) ||
1173 sb_any_quota_suspended(sb)) &&
1174 !sbi->s_qf_names[qtype]) {
1176 "EXT4-fs: Cannot change journaled "
1177 "quota options when quota turned on.\n");
1180 qname = match_strdup(&args[0]);
1183 "EXT4-fs: not enough memory for "
1184 "storing quotafile name.\n");
1187 if (sbi->s_qf_names[qtype] &&
1188 strcmp(sbi->s_qf_names[qtype], qname)) {
1190 "EXT4-fs: %s quota file already "
1191 "specified.\n", QTYPE2NAME(qtype));
1195 sbi->s_qf_names[qtype] = qname;
1196 if (strchr(sbi->s_qf_names[qtype], '/')) {
1198 "EXT4-fs: quotafile must be on "
1199 "filesystem root.\n");
1200 kfree(sbi->s_qf_names[qtype]);
1201 sbi->s_qf_names[qtype] = NULL;
1204 set_opt(sbi->s_mount_opt, QUOTA);
1206 case Opt_offusrjquota:
1209 case Opt_offgrpjquota:
1212 if ((sb_any_quota_enabled(sb) ||
1213 sb_any_quota_suspended(sb)) &&
1214 sbi->s_qf_names[qtype]) {
1215 printk(KERN_ERR "EXT4-fs: Cannot change "
1216 "journaled quota options when "
1217 "quota turned on.\n");
1221 * The space will be released later when all options
1222 * are confirmed to be correct
1224 sbi->s_qf_names[qtype] = NULL;
1226 case Opt_jqfmt_vfsold:
1227 qfmt = QFMT_VFS_OLD;
1229 case Opt_jqfmt_vfsv0:
1232 if ((sb_any_quota_enabled(sb) ||
1233 sb_any_quota_suspended(sb)) &&
1234 sbi->s_jquota_fmt != qfmt) {
1235 printk(KERN_ERR "EXT4-fs: Cannot change "
1236 "journaled quota options when "
1237 "quota turned on.\n");
1240 sbi->s_jquota_fmt = qfmt;
1244 set_opt(sbi->s_mount_opt, QUOTA);
1245 set_opt(sbi->s_mount_opt, USRQUOTA);
1248 set_opt(sbi->s_mount_opt, QUOTA);
1249 set_opt(sbi->s_mount_opt, GRPQUOTA);
1252 if (sb_any_quota_enabled(sb)) {
1253 printk(KERN_ERR "EXT4-fs: Cannot change quota "
1254 "options when quota turned on.\n");
1257 clear_opt(sbi->s_mount_opt, QUOTA);
1258 clear_opt(sbi->s_mount_opt, USRQUOTA);
1259 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1266 "EXT4-fs: quota options not supported.\n");
1270 case Opt_offusrjquota:
1271 case Opt_offgrpjquota:
1272 case Opt_jqfmt_vfsold:
1273 case Opt_jqfmt_vfsv0:
1275 "EXT4-fs: journaled quota options not "
1282 set_opt(sbi->s_mount_opt, ABORT);
1285 if (match_int(&args[0], &option))
1288 set_opt(sbi->s_mount_opt, BARRIER);
1290 clear_opt(sbi->s_mount_opt, BARRIER);
1296 printk("EXT4-fs: resize option only available "
1300 if (match_int(&args[0], &option) != 0)
1302 *n_blocks_count = option;
1305 set_opt(sbi->s_mount_opt, NOBH);
1308 clear_opt(sbi->s_mount_opt, NOBH);
1311 set_opt (sbi->s_mount_opt, EXTENTS);
1314 clear_opt (sbi->s_mount_opt, EXTENTS);
1317 set_opt(sbi->s_mount_opt, I_VERSION);
1318 sb->s_flags |= MS_I_VERSION;
1321 set_opt(sbi->s_mount_opt, MBALLOC);
1324 clear_opt(sbi->s_mount_opt, MBALLOC);
1327 if (match_int(&args[0], &option))
1331 sbi->s_stripe = option;
1335 "EXT4-fs: Unrecognized mount option \"%s\" "
1336 "or missing value\n", p);
1341 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1342 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
1343 sbi->s_qf_names[USRQUOTA])
1344 clear_opt(sbi->s_mount_opt, USRQUOTA);
1346 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
1347 sbi->s_qf_names[GRPQUOTA])
1348 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1350 if ((sbi->s_qf_names[USRQUOTA] &&
1351 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
1352 (sbi->s_qf_names[GRPQUOTA] &&
1353 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1354 printk(KERN_ERR "EXT4-fs: old and new quota "
1355 "format mixing.\n");
1359 if (!sbi->s_jquota_fmt) {
1360 printk(KERN_ERR "EXT4-fs: journaled quota format "
1361 "not specified.\n");
1365 if (sbi->s_jquota_fmt) {
1366 printk(KERN_ERR "EXT4-fs: journaled quota format "
1367 "specified with no journaling "
1376 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1379 struct ext4_sb_info *sbi = EXT4_SB(sb);
1382 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1383 printk (KERN_ERR "EXT4-fs warning: revision level too high, "
1384 "forcing read-only mode\n");
1389 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1390 printk (KERN_WARNING "EXT4-fs warning: mounting unchecked fs, "
1391 "running e2fsck is recommended\n");
1392 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1393 printk (KERN_WARNING
1394 "EXT4-fs warning: mounting fs with errors, "
1395 "running e2fsck is recommended\n");
1396 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1397 le16_to_cpu(es->s_mnt_count) >=
1398 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1399 printk (KERN_WARNING
1400 "EXT4-fs warning: maximal mount count reached, "
1401 "running e2fsck is recommended\n");
1402 else if (le32_to_cpu(es->s_checkinterval) &&
1403 (le32_to_cpu(es->s_lastcheck) +
1404 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1405 printk (KERN_WARNING
1406 "EXT4-fs warning: checktime reached, "
1407 "running e2fsck is recommended\n");
1409 /* @@@ We _will_ want to clear the valid bit if we find
1410 * inconsistencies, to force a fsck at reboot. But for
1411 * a plain journaled filesystem we can keep it set as
1414 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1416 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1417 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1418 le16_add_cpu(&es->s_mnt_count, 1);
1419 es->s_mtime = cpu_to_le32(get_seconds());
1420 ext4_update_dynamic_rev(sb);
1421 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1423 ext4_commit_super(sb, es, 1);
1424 if (test_opt(sb, DEBUG))
1425 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%lu, "
1426 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1428 sbi->s_groups_count,
1429 EXT4_BLOCKS_PER_GROUP(sb),
1430 EXT4_INODES_PER_GROUP(sb),
1433 printk(KERN_INFO "EXT4 FS on %s, ", sb->s_id);
1434 if (EXT4_SB(sb)->s_journal->j_inode == NULL) {
1435 char b[BDEVNAME_SIZE];
1437 printk("external journal on %s\n",
1438 bdevname(EXT4_SB(sb)->s_journal->j_dev, b));
1440 printk("internal journal\n");
1445 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1446 struct ext4_group_desc *gdp)
1450 if (sbi->s_es->s_feature_ro_compat &
1451 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1452 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1453 __le32 le_group = cpu_to_le32(block_group);
1455 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1456 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1457 crc = crc16(crc, (__u8 *)gdp, offset);
1458 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1459 /* for checksum of struct ext4_group_desc do the rest...*/
1460 if ((sbi->s_es->s_feature_incompat &
1461 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1462 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1463 crc = crc16(crc, (__u8 *)gdp + offset,
1464 le16_to_cpu(sbi->s_es->s_desc_size) -
1468 return cpu_to_le16(crc);
1471 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1472 struct ext4_group_desc *gdp)
1474 if ((sbi->s_es->s_feature_ro_compat &
1475 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1476 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1482 /* Called at mount-time, super-block is locked */
1483 static int ext4_check_descriptors(struct super_block *sb)
1485 struct ext4_sb_info *sbi = EXT4_SB(sb);
1486 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1487 ext4_fsblk_t last_block;
1488 ext4_fsblk_t block_bitmap;
1489 ext4_fsblk_t inode_bitmap;
1490 ext4_fsblk_t inode_table;
1491 int flexbg_flag = 0;
1494 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1497 ext4_debug ("Checking group descriptors");
1499 for (i = 0; i < sbi->s_groups_count; i++) {
1500 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1502 if (i == sbi->s_groups_count - 1 || flexbg_flag)
1503 last_block = ext4_blocks_count(sbi->s_es) - 1;
1505 last_block = first_block +
1506 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1508 block_bitmap = ext4_block_bitmap(sb, gdp);
1509 if (block_bitmap < first_block || block_bitmap > last_block)
1511 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1512 "Block bitmap for group %lu not in group "
1513 "(block %llu)!", i, block_bitmap);
1516 inode_bitmap = ext4_inode_bitmap(sb, gdp);
1517 if (inode_bitmap < first_block || inode_bitmap > last_block)
1519 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1520 "Inode bitmap for group %lu not in group "
1521 "(block %llu)!", i, inode_bitmap);
1524 inode_table = ext4_inode_table(sb, gdp);
1525 if (inode_table < first_block ||
1526 inode_table + sbi->s_itb_per_group - 1 > last_block)
1528 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1529 "Inode table for group %lu not in group "
1530 "(block %llu)!", i, inode_table);
1533 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1534 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1535 "Checksum for group %lu failed (%u!=%u)\n",
1536 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1537 gdp)), le16_to_cpu(gdp->bg_checksum));
1541 first_block += EXT4_BLOCKS_PER_GROUP(sb);
1544 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1545 sbi->s_es->s_free_inodes_count=cpu_to_le32(ext4_count_free_inodes(sb));
1549 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1550 * the superblock) which were deleted from all directories, but held open by
1551 * a process at the time of a crash. We walk the list and try to delete these
1552 * inodes at recovery time (only with a read-write filesystem).
1554 * In order to keep the orphan inode chain consistent during traversal (in
1555 * case of crash during recovery), we link each inode into the superblock
1556 * orphan list_head and handle it the same way as an inode deletion during
1557 * normal operation (which journals the operations for us).
1559 * We only do an iget() and an iput() on each inode, which is very safe if we
1560 * accidentally point at an in-use or already deleted inode. The worst that
1561 * can happen in this case is that we get a "bit already cleared" message from
1562 * ext4_free_inode(). The only reason we would point at a wrong inode is if
1563 * e2fsck was run on this filesystem, and it must have already done the orphan
1564 * inode cleanup for us, so we can safely abort without any further action.
1566 static void ext4_orphan_cleanup (struct super_block * sb,
1567 struct ext4_super_block * es)
1569 unsigned int s_flags = sb->s_flags;
1570 int nr_orphans = 0, nr_truncates = 0;
1574 if (!es->s_last_orphan) {
1575 jbd_debug(4, "no orphan inodes to clean up\n");
1579 if (bdev_read_only(sb->s_bdev)) {
1580 printk(KERN_ERR "EXT4-fs: write access "
1581 "unavailable, skipping orphan cleanup.\n");
1585 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1586 if (es->s_last_orphan)
1587 jbd_debug(1, "Errors on filesystem, "
1588 "clearing orphan list.\n");
1589 es->s_last_orphan = 0;
1590 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1594 if (s_flags & MS_RDONLY) {
1595 printk(KERN_INFO "EXT4-fs: %s: orphan cleanup on readonly fs\n",
1597 sb->s_flags &= ~MS_RDONLY;
1600 /* Needed for iput() to work correctly and not trash data */
1601 sb->s_flags |= MS_ACTIVE;
1602 /* Turn on quotas so that they are updated correctly */
1603 for (i = 0; i < MAXQUOTAS; i++) {
1604 if (EXT4_SB(sb)->s_qf_names[i]) {
1605 int ret = ext4_quota_on_mount(sb, i);
1608 "EXT4-fs: Cannot turn on journaled "
1609 "quota: error %d\n", ret);
1614 while (es->s_last_orphan) {
1615 struct inode *inode;
1617 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1618 if (IS_ERR(inode)) {
1619 es->s_last_orphan = 0;
1623 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
1625 if (inode->i_nlink) {
1627 "%s: truncating inode %lu to %Ld bytes\n",
1628 __func__, inode->i_ino, inode->i_size);
1629 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1630 inode->i_ino, inode->i_size);
1631 ext4_truncate(inode);
1635 "%s: deleting unreferenced inode %lu\n",
1636 __func__, inode->i_ino);
1637 jbd_debug(2, "deleting unreferenced inode %lu\n",
1641 iput(inode); /* The delete magic happens here! */
1644 #define PLURAL(x) (x), ((x)==1) ? "" : "s"
1647 printk(KERN_INFO "EXT4-fs: %s: %d orphan inode%s deleted\n",
1648 sb->s_id, PLURAL(nr_orphans));
1650 printk(KERN_INFO "EXT4-fs: %s: %d truncate%s cleaned up\n",
1651 sb->s_id, PLURAL(nr_truncates));
1653 /* Turn quotas off */
1654 for (i = 0; i < MAXQUOTAS; i++) {
1655 if (sb_dqopt(sb)->files[i])
1656 vfs_quota_off(sb, i, 0);
1659 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1662 * Maximal extent format file size.
1663 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1664 * extent format containers, within a sector_t, and within i_blocks
1665 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1666 * so that won't be a limiting factor.
1668 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1670 static loff_t ext4_max_size(int blkbits)
1673 loff_t upper_limit = MAX_LFS_FILESIZE;
1675 /* small i_blocks in vfs inode? */
1676 if (sizeof(blkcnt_t) < sizeof(u64)) {
1678 * CONFIG_LSF is not enabled implies the inode
1679 * i_block represent total blocks in 512 bytes
1680 * 32 == size of vfs inode i_blocks * 8
1682 upper_limit = (1LL << 32) - 1;
1684 /* total blocks in file system block size */
1685 upper_limit >>= (blkbits - 9);
1686 upper_limit <<= blkbits;
1689 /* 32-bit extent-start container, ee_block */
1694 /* Sanity check against vm- & vfs- imposed limits */
1695 if (res > upper_limit)
1702 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
1703 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1704 * We need to be 1 filesystem block less than the 2^48 sector limit.
1706 static loff_t ext4_max_bitmap_size(int bits)
1708 loff_t res = EXT4_NDIR_BLOCKS;
1711 /* This is calculated to be the largest file size for a
1712 * dense, bitmapped file such that the total number of
1713 * sectors in the file, including data and all indirect blocks,
1714 * does not exceed 2^48 -1
1715 * __u32 i_blocks_lo and _u16 i_blocks_high representing the
1716 * total number of 512 bytes blocks of the file
1719 if (sizeof(blkcnt_t) < sizeof(u64)) {
1721 * CONFIG_LSF is not enabled implies the inode
1722 * i_block represent total blocks in 512 bytes
1723 * 32 == size of vfs inode i_blocks * 8
1725 upper_limit = (1LL << 32) - 1;
1727 /* total blocks in file system block size */
1728 upper_limit >>= (bits - 9);
1732 * We use 48 bit ext4_inode i_blocks
1733 * With EXT4_HUGE_FILE_FL set the i_blocks
1734 * represent total number of blocks in
1735 * file system block size
1737 upper_limit = (1LL << 48) - 1;
1741 /* indirect blocks */
1743 /* double indirect blocks */
1744 meta_blocks += 1 + (1LL << (bits-2));
1745 /* tripple indirect blocks */
1746 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1748 upper_limit -= meta_blocks;
1749 upper_limit <<= bits;
1751 res += 1LL << (bits-2);
1752 res += 1LL << (2*(bits-2));
1753 res += 1LL << (3*(bits-2));
1755 if (res > upper_limit)
1758 if (res > MAX_LFS_FILESIZE)
1759 res = MAX_LFS_FILESIZE;
1764 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
1765 ext4_fsblk_t logical_sb_block, int nr)
1767 struct ext4_sb_info *sbi = EXT4_SB(sb);
1768 ext4_group_t bg, first_meta_bg;
1771 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1773 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
1775 return logical_sb_block + nr + 1;
1776 bg = sbi->s_desc_per_block * nr;
1777 if (ext4_bg_has_super(sb, bg))
1779 return (has_super + ext4_group_first_block_no(sb, bg));
1783 * ext4_get_stripe_size: Get the stripe size.
1784 * @sbi: In memory super block info
1786 * If we have specified it via mount option, then
1787 * use the mount option value. If the value specified at mount time is
1788 * greater than the blocks per group use the super block value.
1789 * If the super block value is greater than blocks per group return 0.
1790 * Allocator needs it be less than blocks per group.
1793 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
1795 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
1796 unsigned long stripe_width =
1797 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
1799 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
1800 return sbi->s_stripe;
1802 if (stripe_width <= sbi->s_blocks_per_group)
1803 return stripe_width;
1805 if (stride <= sbi->s_blocks_per_group)
1811 static int ext4_fill_super (struct super_block *sb, void *data, int silent)
1812 __releases(kernel_sem)
1813 __acquires(kernel_sem)
1816 struct buffer_head * bh;
1817 struct ext4_super_block *es = NULL;
1818 struct ext4_sb_info *sbi;
1820 ext4_fsblk_t sb_block = get_sb_block(&data);
1821 ext4_fsblk_t logical_sb_block;
1822 unsigned long offset = 0;
1823 unsigned int journal_inum = 0;
1824 unsigned long journal_devnum = 0;
1825 unsigned long def_mount_opts;
1836 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1839 sb->s_fs_info = sbi;
1840 sbi->s_mount_opt = 0;
1841 sbi->s_resuid = EXT4_DEF_RESUID;
1842 sbi->s_resgid = EXT4_DEF_RESGID;
1843 sbi->s_sb_block = sb_block;
1847 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
1849 printk(KERN_ERR "EXT4-fs: unable to set blocksize\n");
1853 if (!sb_set_blocksize(sb, blocksize)) {
1854 printk(KERN_ERR "EXT4-fs: bad blocksize %d.\n", blocksize);
1859 * The ext4 superblock will not be buffer aligned for other than 1kB
1860 * block sizes. We need to calculate the offset from buffer start.
1862 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
1863 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
1864 offset = do_div(logical_sb_block, blocksize);
1866 logical_sb_block = sb_block;
1869 if (!(bh = sb_bread(sb, logical_sb_block))) {
1870 printk (KERN_ERR "EXT4-fs: unable to read superblock\n");
1874 * Note: s_es must be initialized as soon as possible because
1875 * some ext4 macro-instructions depend on its value
1877 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
1879 sb->s_magic = le16_to_cpu(es->s_magic);
1880 if (sb->s_magic != EXT4_SUPER_MAGIC)
1883 /* Set defaults before we parse the mount options */
1884 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1885 if (def_mount_opts & EXT4_DEFM_DEBUG)
1886 set_opt(sbi->s_mount_opt, DEBUG);
1887 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
1888 set_opt(sbi->s_mount_opt, GRPID);
1889 if (def_mount_opts & EXT4_DEFM_UID16)
1890 set_opt(sbi->s_mount_opt, NO_UID32);
1891 #ifdef CONFIG_EXT4DEV_FS_XATTR
1892 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
1893 set_opt(sbi->s_mount_opt, XATTR_USER);
1895 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
1896 if (def_mount_opts & EXT4_DEFM_ACL)
1897 set_opt(sbi->s_mount_opt, POSIX_ACL);
1899 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
1900 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
1901 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
1902 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
1903 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
1904 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
1906 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
1907 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1908 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
1909 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1911 set_opt(sbi->s_mount_opt, ERRORS_RO);
1913 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1914 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1916 set_opt(sbi->s_mount_opt, RESERVATION);
1917 set_opt(sbi->s_mount_opt, BARRIER);
1920 * turn on extents feature by default in ext4 filesystem
1921 * User -o noextents to turn it off
1923 set_opt(sbi->s_mount_opt, EXTENTS);
1925 * turn on mballoc feature by default in ext4 filesystem
1926 * User -o nomballoc to turn it off
1928 set_opt(sbi->s_mount_opt, MBALLOC);
1930 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1934 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1935 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1937 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
1938 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
1939 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1940 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
1942 "EXT4-fs warning: feature flags set on rev 0 fs, "
1943 "running e2fsck is recommended\n");
1946 * Since ext4 is still considered development code, we require
1947 * that the TEST_FILESYS flag in s->flags be set.
1949 if (!(le32_to_cpu(es->s_flags) & EXT2_FLAGS_TEST_FILESYS)) {
1950 printk(KERN_WARNING "EXT4-fs: %s: not marked "
1951 "OK to use with test code.\n", sb->s_id);
1956 * Check feature flags regardless of the revision level, since we
1957 * previously didn't change the revision level when setting the flags,
1958 * so there is a chance incompat flags are set on a rev 0 filesystem.
1960 features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
1962 printk(KERN_ERR "EXT4-fs: %s: couldn't mount because of "
1963 "unsupported optional features (%x).\n",
1964 sb->s_id, le32_to_cpu(features));
1967 features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
1968 if (!(sb->s_flags & MS_RDONLY) && features) {
1969 printk(KERN_ERR "EXT4-fs: %s: couldn't mount RDWR because of "
1970 "unsupported optional features (%x).\n",
1971 sb->s_id, le32_to_cpu(features));
1974 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
1976 * Large file size enabled file system can only be
1977 * mount if kernel is build with CONFIG_LSF
1979 if (sizeof(root->i_blocks) < sizeof(u64) &&
1980 !(sb->s_flags & MS_RDONLY)) {
1981 printk(KERN_ERR "EXT4-fs: %s: Filesystem with huge "
1982 "files cannot be mounted read-write "
1983 "without CONFIG_LSF.\n", sb->s_id);
1987 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
1989 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
1990 blocksize > EXT4_MAX_BLOCK_SIZE) {
1992 "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
1993 blocksize, sb->s_id);
1997 if (sb->s_blocksize != blocksize) {
1999 /* Validate the filesystem blocksize */
2000 if (!sb_set_blocksize(sb, blocksize)) {
2001 printk(KERN_ERR "EXT4-fs: bad block size %d.\n",
2007 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2008 offset = do_div(logical_sb_block, blocksize);
2009 bh = sb_bread(sb, logical_sb_block);
2012 "EXT4-fs: Can't read superblock on 2nd try.\n");
2015 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
2017 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2019 "EXT4-fs: Magic mismatch, very weird !\n");
2024 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits);
2025 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits);
2027 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2028 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2029 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
2031 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2032 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
2033 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
2034 (!is_power_of_2(sbi->s_inode_size)) ||
2035 (sbi->s_inode_size > blocksize)) {
2037 "EXT4-fs: unsupported inode size: %d\n",
2041 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2042 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
2044 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2045 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
2046 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
2047 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
2048 !is_power_of_2(sbi->s_desc_size)) {
2050 "EXT4-fs: unsupported descriptor size %lu\n",
2055 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
2056 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
2057 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
2058 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
2060 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
2061 if (sbi->s_inodes_per_block == 0)
2063 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2064 sbi->s_inodes_per_block;
2065 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
2067 sbi->s_mount_state = le16_to_cpu(es->s_state);
2068 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2069 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2070 for (i=0; i < 4; i++)
2071 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2072 sbi->s_def_hash_version = es->s_def_hash_version;
2074 if (sbi->s_blocks_per_group > blocksize * 8) {
2076 "EXT4-fs: #blocks per group too big: %lu\n",
2077 sbi->s_blocks_per_group);
2080 if (sbi->s_inodes_per_group > blocksize * 8) {
2082 "EXT4-fs: #inodes per group too big: %lu\n",
2083 sbi->s_inodes_per_group);
2087 if (ext4_blocks_count(es) >
2088 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
2089 printk(KERN_ERR "EXT4-fs: filesystem on %s:"
2090 " too large to mount safely\n", sb->s_id);
2091 if (sizeof(sector_t) < 8)
2092 printk(KERN_WARNING "EXT4-fs: CONFIG_LBD not "
2097 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2100 /* ensure blocks_count calculation below doesn't sign-extend */
2101 if (ext4_blocks_count(es) + EXT4_BLOCKS_PER_GROUP(sb) <
2102 le32_to_cpu(es->s_first_data_block) + 1) {
2103 printk(KERN_WARNING "EXT4-fs: bad geometry: block count %llu, "
2104 "first data block %u, blocks per group %lu\n",
2105 ext4_blocks_count(es),
2106 le32_to_cpu(es->s_first_data_block),
2107 EXT4_BLOCKS_PER_GROUP(sb));
2110 blocks_count = (ext4_blocks_count(es) -
2111 le32_to_cpu(es->s_first_data_block) +
2112 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2113 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2114 sbi->s_groups_count = blocks_count;
2115 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2116 EXT4_DESC_PER_BLOCK(sb);
2117 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
2119 if (sbi->s_group_desc == NULL) {
2120 printk (KERN_ERR "EXT4-fs: not enough memory\n");
2124 bgl_lock_init(&sbi->s_blockgroup_lock);
2126 for (i = 0; i < db_count; i++) {
2127 block = descriptor_loc(sb, logical_sb_block, i);
2128 sbi->s_group_desc[i] = sb_bread(sb, block);
2129 if (!sbi->s_group_desc[i]) {
2130 printk (KERN_ERR "EXT4-fs: "
2131 "can't read group descriptor %d\n", i);
2136 if (!ext4_check_descriptors (sb)) {
2137 printk(KERN_ERR "EXT4-fs: group descriptors corrupted!\n");
2140 sbi->s_gdb_count = db_count;
2141 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2142 spin_lock_init(&sbi->s_next_gen_lock);
2144 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2145 ext4_count_free_blocks(sb));
2147 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2148 ext4_count_free_inodes(sb));
2151 err = percpu_counter_init(&sbi->s_dirs_counter,
2152 ext4_count_dirs(sb));
2155 printk(KERN_ERR "EXT4-fs: insufficient memory\n");
2159 /* per fileystem reservation list head & lock */
2160 spin_lock_init(&sbi->s_rsv_window_lock);
2161 sbi->s_rsv_window_root = RB_ROOT;
2162 /* Add a single, static dummy reservation to the start of the
2163 * reservation window list --- it gives us a placeholder for
2164 * append-at-start-of-list which makes the allocation logic
2165 * _much_ simpler. */
2166 sbi->s_rsv_window_head.rsv_start = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
2167 sbi->s_rsv_window_head.rsv_end = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
2168 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
2169 sbi->s_rsv_window_head.rsv_goal_size = 0;
2170 ext4_rsv_window_add(sb, &sbi->s_rsv_window_head);
2172 sbi->s_stripe = ext4_get_stripe_size(sbi);
2175 * set up enough so that it can read an inode
2177 sb->s_op = &ext4_sops;
2178 sb->s_export_op = &ext4_export_ops;
2179 sb->s_xattr = ext4_xattr_handlers;
2181 sb->s_qcop = &ext4_qctl_operations;
2182 sb->dq_op = &ext4_quota_operations;
2184 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2188 needs_recovery = (es->s_last_orphan != 0 ||
2189 EXT4_HAS_INCOMPAT_FEATURE(sb,
2190 EXT4_FEATURE_INCOMPAT_RECOVER));
2193 * The first inode we look at is the journal inode. Don't try
2194 * root first: it may be modified in the journal!
2196 if (!test_opt(sb, NOLOAD) &&
2197 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2198 if (ext4_load_journal(sb, es, journal_devnum))
2200 if (!(sb->s_flags & MS_RDONLY) &&
2201 EXT4_SB(sb)->s_journal->j_failed_commit) {
2202 printk(KERN_CRIT "EXT4-fs error (device %s): "
2203 "ext4_fill_super: Journal transaction "
2204 "%u is corrupt\n", sb->s_id,
2205 EXT4_SB(sb)->s_journal->j_failed_commit);
2206 if (test_opt (sb, ERRORS_RO)) {
2208 "Mounting filesystem read-only\n");
2209 sb->s_flags |= MS_RDONLY;
2210 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2211 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2213 if (test_opt(sb, ERRORS_PANIC)) {
2214 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2215 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2216 ext4_commit_super(sb, es, 1);
2218 "EXT4-fs (device %s): mount failed\n",
2223 } else if (journal_inum) {
2224 if (ext4_create_journal(sb, es, journal_inum))
2229 "ext4: No journal on filesystem on %s\n",
2234 if (ext4_blocks_count(es) > 0xffffffffULL &&
2235 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2236 JBD2_FEATURE_INCOMPAT_64BIT)) {
2237 printk(KERN_ERR "ext4: Failed to set 64-bit journal feature\n");
2241 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2242 jbd2_journal_set_features(sbi->s_journal,
2243 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2244 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2245 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2246 jbd2_journal_set_features(sbi->s_journal,
2247 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2248 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2249 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2251 jbd2_journal_clear_features(sbi->s_journal,
2252 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2253 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2256 /* We have now updated the journal if required, so we can
2257 * validate the data journaling mode. */
2258 switch (test_opt(sb, DATA_FLAGS)) {
2260 /* No mode set, assume a default based on the journal
2261 * capabilities: ORDERED_DATA if the journal can
2262 * cope, else JOURNAL_DATA
2264 if (jbd2_journal_check_available_features
2265 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
2266 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2268 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2271 case EXT4_MOUNT_ORDERED_DATA:
2272 case EXT4_MOUNT_WRITEBACK_DATA:
2273 if (!jbd2_journal_check_available_features
2274 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
2275 printk(KERN_ERR "EXT4-fs: Journal does not support "
2276 "requested data journaling mode\n");
2283 if (test_opt(sb, NOBH)) {
2284 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2285 printk(KERN_WARNING "EXT4-fs: Ignoring nobh option - "
2286 "its supported only with writeback mode\n");
2287 clear_opt(sbi->s_mount_opt, NOBH);
2291 * The jbd2_journal_load will have done any necessary log recovery,
2292 * so we can safely mount the rest of the filesystem now.
2295 root = ext4_iget(sb, EXT4_ROOT_INO);
2297 printk(KERN_ERR "EXT4-fs: get root inode failed\n");
2298 ret = PTR_ERR(root);
2301 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2303 printk(KERN_ERR "EXT4-fs: corrupt root inode, run e2fsck\n");
2306 sb->s_root = d_alloc_root(root);
2308 printk(KERN_ERR "EXT4-fs: get root dentry failed\n");
2314 ext4_setup_super (sb, es, sb->s_flags & MS_RDONLY);
2316 /* determine the minimum size of new large inodes, if present */
2317 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2318 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2319 EXT4_GOOD_OLD_INODE_SIZE;
2320 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2321 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2322 if (sbi->s_want_extra_isize <
2323 le16_to_cpu(es->s_want_extra_isize))
2324 sbi->s_want_extra_isize =
2325 le16_to_cpu(es->s_want_extra_isize);
2326 if (sbi->s_want_extra_isize <
2327 le16_to_cpu(es->s_min_extra_isize))
2328 sbi->s_want_extra_isize =
2329 le16_to_cpu(es->s_min_extra_isize);
2332 /* Check if enough inode space is available */
2333 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2334 sbi->s_inode_size) {
2335 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2336 EXT4_GOOD_OLD_INODE_SIZE;
2337 printk(KERN_INFO "EXT4-fs: required extra inode space not"
2342 * akpm: core read_super() calls in here with the superblock locked.
2343 * That deadlocks, because orphan cleanup needs to lock the superblock
2344 * in numerous places. Here we just pop the lock - it's relatively
2345 * harmless, because we are now ready to accept write_super() requests,
2346 * and aviro says that's the only reason for hanging onto the
2349 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2350 ext4_orphan_cleanup(sb, es);
2351 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
2353 printk (KERN_INFO "EXT4-fs: recovery complete.\n");
2354 ext4_mark_recovery_complete(sb, es);
2355 printk (KERN_INFO "EXT4-fs: mounted filesystem with %s data mode.\n",
2356 test_opt(sb,DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA ? "journal":
2357 test_opt(sb,DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA ? "ordered":
2361 ext4_mb_init(sb, needs_recovery);
2368 printk(KERN_ERR "VFS: Can't find ext4 filesystem on dev %s.\n",
2373 jbd2_journal_destroy(sbi->s_journal);
2375 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2376 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2377 percpu_counter_destroy(&sbi->s_dirs_counter);
2379 for (i = 0; i < db_count; i++)
2380 brelse(sbi->s_group_desc[i]);
2381 kfree(sbi->s_group_desc);
2384 for (i = 0; i < MAXQUOTAS; i++)
2385 kfree(sbi->s_qf_names[i]);
2387 ext4_blkdev_remove(sbi);
2390 sb->s_fs_info = NULL;
2397 * Setup any per-fs journal parameters now. We'll do this both on
2398 * initial mount, once the journal has been initialised but before we've
2399 * done any recovery; and again on any subsequent remount.
2401 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
2403 struct ext4_sb_info *sbi = EXT4_SB(sb);
2405 if (sbi->s_commit_interval)
2406 journal->j_commit_interval = sbi->s_commit_interval;
2407 /* We could also set up an ext4-specific default for the commit
2408 * interval here, but for now we'll just fall back to the jbd
2411 spin_lock(&journal->j_state_lock);
2412 if (test_opt(sb, BARRIER))
2413 journal->j_flags |= JBD2_BARRIER;
2415 journal->j_flags &= ~JBD2_BARRIER;
2416 spin_unlock(&journal->j_state_lock);
2419 static journal_t *ext4_get_journal(struct super_block *sb,
2420 unsigned int journal_inum)
2422 struct inode *journal_inode;
2425 /* First, test for the existence of a valid inode on disk. Bad
2426 * things happen if we iget() an unused inode, as the subsequent
2427 * iput() will try to delete it. */
2429 journal_inode = ext4_iget(sb, journal_inum);
2430 if (IS_ERR(journal_inode)) {
2431 printk(KERN_ERR "EXT4-fs: no journal found.\n");
2434 if (!journal_inode->i_nlink) {
2435 make_bad_inode(journal_inode);
2436 iput(journal_inode);
2437 printk(KERN_ERR "EXT4-fs: journal inode is deleted.\n");
2441 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
2442 journal_inode, journal_inode->i_size);
2443 if (!S_ISREG(journal_inode->i_mode)) {
2444 printk(KERN_ERR "EXT4-fs: invalid journal inode.\n");
2445 iput(journal_inode);
2449 journal = jbd2_journal_init_inode(journal_inode);
2451 printk(KERN_ERR "EXT4-fs: Could not load journal inode\n");
2452 iput(journal_inode);
2455 journal->j_private = sb;
2456 ext4_init_journal_params(sb, journal);
2460 static journal_t *ext4_get_dev_journal(struct super_block *sb,
2463 struct buffer_head * bh;
2467 int hblock, blocksize;
2468 ext4_fsblk_t sb_block;
2469 unsigned long offset;
2470 struct ext4_super_block * es;
2471 struct block_device *bdev;
2473 bdev = ext4_blkdev_get(j_dev);
2477 if (bd_claim(bdev, sb)) {
2479 "EXT4: failed to claim external journal device.\n");
2484 blocksize = sb->s_blocksize;
2485 hblock = bdev_hardsect_size(bdev);
2486 if (blocksize < hblock) {
2488 "EXT4-fs: blocksize too small for journal device.\n");
2492 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
2493 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
2494 set_blocksize(bdev, blocksize);
2495 if (!(bh = __bread(bdev, sb_block, blocksize))) {
2496 printk(KERN_ERR "EXT4-fs: couldn't read superblock of "
2497 "external journal\n");
2501 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2502 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
2503 !(le32_to_cpu(es->s_feature_incompat) &
2504 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2505 printk(KERN_ERR "EXT4-fs: external journal has "
2506 "bad superblock\n");
2511 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2512 printk(KERN_ERR "EXT4-fs: journal UUID does not match\n");
2517 len = ext4_blocks_count(es);
2518 start = sb_block + 1;
2519 brelse(bh); /* we're done with the superblock */
2521 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
2522 start, len, blocksize);
2524 printk(KERN_ERR "EXT4-fs: failed to create device journal\n");
2527 journal->j_private = sb;
2528 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2529 wait_on_buffer(journal->j_sb_buffer);
2530 if (!buffer_uptodate(journal->j_sb_buffer)) {
2531 printk(KERN_ERR "EXT4-fs: I/O error on journal device\n");
2534 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
2535 printk(KERN_ERR "EXT4-fs: External journal has more than one "
2536 "user (unsupported) - %d\n",
2537 be32_to_cpu(journal->j_superblock->s_nr_users));
2540 EXT4_SB(sb)->journal_bdev = bdev;
2541 ext4_init_journal_params(sb, journal);
2544 jbd2_journal_destroy(journal);
2546 ext4_blkdev_put(bdev);
2550 static int ext4_load_journal(struct super_block *sb,
2551 struct ext4_super_block *es,
2552 unsigned long journal_devnum)
2555 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2558 int really_read_only;
2560 if (journal_devnum &&
2561 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2562 printk(KERN_INFO "EXT4-fs: external journal device major/minor "
2563 "numbers have changed\n");
2564 journal_dev = new_decode_dev(journal_devnum);
2566 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2568 really_read_only = bdev_read_only(sb->s_bdev);
2571 * Are we loading a blank journal or performing recovery after a
2572 * crash? For recovery, we need to check in advance whether we
2573 * can get read-write access to the device.
2576 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2577 if (sb->s_flags & MS_RDONLY) {
2578 printk(KERN_INFO "EXT4-fs: INFO: recovery "
2579 "required on readonly filesystem.\n");
2580 if (really_read_only) {
2581 printk(KERN_ERR "EXT4-fs: write access "
2582 "unavailable, cannot proceed.\n");
2585 printk (KERN_INFO "EXT4-fs: write access will "
2586 "be enabled during recovery.\n");
2590 if (journal_inum && journal_dev) {
2591 printk(KERN_ERR "EXT4-fs: filesystem has both journal "
2592 "and inode journals!\n");
2597 if (!(journal = ext4_get_journal(sb, journal_inum)))
2600 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
2604 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2605 err = jbd2_journal_update_format(journal);
2607 printk(KERN_ERR "EXT4-fs: error updating journal.\n");
2608 jbd2_journal_destroy(journal);
2613 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
2614 err = jbd2_journal_wipe(journal, !really_read_only);
2616 err = jbd2_journal_load(journal);
2619 printk(KERN_ERR "EXT4-fs: error loading journal.\n");
2620 jbd2_journal_destroy(journal);
2624 EXT4_SB(sb)->s_journal = journal;
2625 ext4_clear_journal_err(sb, es);
2627 if (journal_devnum &&
2628 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2629 es->s_journal_dev = cpu_to_le32(journal_devnum);
2632 /* Make sure we flush the recovery flag to disk. */
2633 ext4_commit_super(sb, es, 1);
2639 static int ext4_create_journal(struct super_block * sb,
2640 struct ext4_super_block * es,
2641 unsigned int journal_inum)
2646 if (sb->s_flags & MS_RDONLY) {
2647 printk(KERN_ERR "EXT4-fs: readonly filesystem when trying to "
2648 "create journal.\n");
2652 journal = ext4_get_journal(sb, journal_inum);
2656 printk(KERN_INFO "EXT4-fs: creating new journal on inode %u\n",
2659 err = jbd2_journal_create(journal);
2661 printk(KERN_ERR "EXT4-fs: error creating journal.\n");
2662 jbd2_journal_destroy(journal);
2666 EXT4_SB(sb)->s_journal = journal;
2668 ext4_update_dynamic_rev(sb);
2669 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2670 EXT4_SET_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL);
2672 es->s_journal_inum = cpu_to_le32(journal_inum);
2675 /* Make sure we flush the recovery flag to disk. */
2676 ext4_commit_super(sb, es, 1);
2681 static void ext4_commit_super (struct super_block * sb,
2682 struct ext4_super_block * es,
2685 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
2689 es->s_wtime = cpu_to_le32(get_seconds());
2690 ext4_free_blocks_count_set(es, ext4_count_free_blocks(sb));
2691 es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
2692 BUFFER_TRACE(sbh, "marking dirty");
2693 mark_buffer_dirty(sbh);
2695 sync_dirty_buffer(sbh);
2700 * Have we just finished recovery? If so, and if we are mounting (or
2701 * remounting) the filesystem readonly, then we will end up with a
2702 * consistent fs on disk. Record that fact.
2704 static void ext4_mark_recovery_complete(struct super_block * sb,
2705 struct ext4_super_block * es)
2707 journal_t *journal = EXT4_SB(sb)->s_journal;
2709 jbd2_journal_lock_updates(journal);
2710 jbd2_journal_flush(journal);
2712 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
2713 sb->s_flags & MS_RDONLY) {
2714 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2716 ext4_commit_super(sb, es, 1);
2719 jbd2_journal_unlock_updates(journal);
2723 * If we are mounting (or read-write remounting) a filesystem whose journal
2724 * has recorded an error from a previous lifetime, move that error to the
2725 * main filesystem now.
2727 static void ext4_clear_journal_err(struct super_block * sb,
2728 struct ext4_super_block * es)
2734 journal = EXT4_SB(sb)->s_journal;
2737 * Now check for any error status which may have been recorded in the
2738 * journal by a prior ext4_error() or ext4_abort()
2741 j_errno = jbd2_journal_errno(journal);
2745 errstr = ext4_decode_error(sb, j_errno, nbuf);
2746 ext4_warning(sb, __func__, "Filesystem error recorded "
2747 "from previous mount: %s", errstr);
2748 ext4_warning(sb, __func__, "Marking fs in need of "
2749 "filesystem check.");
2751 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2752 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2753 ext4_commit_super (sb, es, 1);
2755 jbd2_journal_clear_err(journal);
2760 * Force the running and committing transactions to commit,
2761 * and wait on the commit.
2763 int ext4_force_commit(struct super_block *sb)
2768 if (sb->s_flags & MS_RDONLY)
2771 journal = EXT4_SB(sb)->s_journal;
2773 ret = ext4_journal_force_commit(journal);
2778 * Ext4 always journals updates to the superblock itself, so we don't
2779 * have to propagate any other updates to the superblock on disk at this
2780 * point. Just start an async writeback to get the buffers on their way
2783 * This implicitly triggers the writebehind on sync().
2786 static void ext4_write_super (struct super_block * sb)
2788 if (mutex_trylock(&sb->s_lock) != 0)
2793 static int ext4_sync_fs(struct super_block *sb, int wait)
2798 if (jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, &target)) {
2800 jbd2_log_wait_commit(EXT4_SB(sb)->s_journal, target);
2806 * LVM calls this function before a (read-only) snapshot is created. This
2807 * gives us a chance to flush the journal completely and mark the fs clean.
2809 static void ext4_write_super_lockfs(struct super_block *sb)
2813 if (!(sb->s_flags & MS_RDONLY)) {
2814 journal_t *journal = EXT4_SB(sb)->s_journal;
2816 /* Now we set up the journal barrier. */
2817 jbd2_journal_lock_updates(journal);
2818 jbd2_journal_flush(journal);
2820 /* Journal blocked and flushed, clear needs_recovery flag. */
2821 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2822 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
2827 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2828 * flag here, even though the filesystem is not technically dirty yet.
2830 static void ext4_unlockfs(struct super_block *sb)
2832 if (!(sb->s_flags & MS_RDONLY)) {
2834 /* Reser the needs_recovery flag before the fs is unlocked. */
2835 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2836 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
2838 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
2842 static int ext4_remount (struct super_block * sb, int * flags, char * data)
2844 struct ext4_super_block * es;
2845 struct ext4_sb_info *sbi = EXT4_SB(sb);
2846 ext4_fsblk_t n_blocks_count = 0;
2847 unsigned long old_sb_flags;
2848 struct ext4_mount_options old_opts;
2854 /* Store the original options */
2855 old_sb_flags = sb->s_flags;
2856 old_opts.s_mount_opt = sbi->s_mount_opt;
2857 old_opts.s_resuid = sbi->s_resuid;
2858 old_opts.s_resgid = sbi->s_resgid;
2859 old_opts.s_commit_interval = sbi->s_commit_interval;
2861 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2862 for (i = 0; i < MAXQUOTAS; i++)
2863 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2867 * Allow the "check" option to be passed as a remount option.
2869 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
2874 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT)
2875 ext4_abort(sb, __func__, "Abort forced by user");
2877 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2878 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2882 ext4_init_journal_params(sb, sbi->s_journal);
2884 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
2885 n_blocks_count > ext4_blocks_count(es)) {
2886 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) {
2891 if (*flags & MS_RDONLY) {
2893 * First of all, the unconditional stuff we have to do
2894 * to disable replay of the journal when we next remount
2896 sb->s_flags |= MS_RDONLY;
2899 * OK, test if we are remounting a valid rw partition
2900 * readonly, and if so set the rdonly flag and then
2901 * mark the partition as valid again.
2903 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
2904 (sbi->s_mount_state & EXT4_VALID_FS))
2905 es->s_state = cpu_to_le16(sbi->s_mount_state);
2908 * We have to unlock super so that we can wait for
2912 ext4_mark_recovery_complete(sb, es);
2916 if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2917 ~EXT4_FEATURE_RO_COMPAT_SUPP))) {
2918 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
2919 "remount RDWR because of unsupported "
2920 "optional features (%x).\n",
2921 sb->s_id, le32_to_cpu(ret));
2927 * If we have an unprocessed orphan list hanging
2928 * around from a previously readonly bdev mount,
2929 * require a full umount/remount for now.
2931 if (es->s_last_orphan) {
2932 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
2933 "remount RDWR because of unprocessed "
2934 "orphan inode list. Please "
2935 "umount/remount instead.\n",
2942 * Mounting a RDONLY partition read-write, so reread
2943 * and store the current valid flag. (It may have
2944 * been changed by e2fsck since we originally mounted
2947 ext4_clear_journal_err(sb, es);
2948 sbi->s_mount_state = le16_to_cpu(es->s_state);
2949 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
2951 if (!ext4_setup_super (sb, es, 0))
2952 sb->s_flags &= ~MS_RDONLY;
2956 /* Release old quota file names */
2957 for (i = 0; i < MAXQUOTAS; i++)
2958 if (old_opts.s_qf_names[i] &&
2959 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2960 kfree(old_opts.s_qf_names[i]);
2964 sb->s_flags = old_sb_flags;
2965 sbi->s_mount_opt = old_opts.s_mount_opt;
2966 sbi->s_resuid = old_opts.s_resuid;
2967 sbi->s_resgid = old_opts.s_resgid;
2968 sbi->s_commit_interval = old_opts.s_commit_interval;
2970 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
2971 for (i = 0; i < MAXQUOTAS; i++) {
2972 if (sbi->s_qf_names[i] &&
2973 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2974 kfree(sbi->s_qf_names[i]);
2975 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
2981 static int ext4_statfs (struct dentry * dentry, struct kstatfs * buf)
2983 struct super_block *sb = dentry->d_sb;
2984 struct ext4_sb_info *sbi = EXT4_SB(sb);
2985 struct ext4_super_block *es = sbi->s_es;
2988 if (test_opt(sb, MINIX_DF)) {
2989 sbi->s_overhead_last = 0;
2990 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
2991 ext4_group_t ngroups = sbi->s_groups_count, i;
2992 ext4_fsblk_t overhead = 0;
2996 * Compute the overhead (FS structures). This is constant
2997 * for a given filesystem unless the number of block groups
2998 * changes so we cache the previous value until it does.
3002 * All of the blocks before first_data_block are
3005 overhead = le32_to_cpu(es->s_first_data_block);
3008 * Add the overhead attributed to the superblock and
3009 * block group descriptors. If the sparse superblocks
3010 * feature is turned on, then not all groups have this.
3012 for (i = 0; i < ngroups; i++) {
3013 overhead += ext4_bg_has_super(sb, i) +
3014 ext4_bg_num_gdb(sb, i);
3019 * Every block group has an inode bitmap, a block
3020 * bitmap, and an inode table.
3022 overhead += ngroups * (2 + sbi->s_itb_per_group);
3023 sbi->s_overhead_last = overhead;
3025 sbi->s_blocks_last = ext4_blocks_count(es);
3028 buf->f_type = EXT4_SUPER_MAGIC;
3029 buf->f_bsize = sb->s_blocksize;
3030 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
3031 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter);
3032 ext4_free_blocks_count_set(es, buf->f_bfree);
3033 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3034 if (buf->f_bfree < ext4_r_blocks_count(es))
3036 buf->f_files = le32_to_cpu(es->s_inodes_count);
3037 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
3038 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
3039 buf->f_namelen = EXT4_NAME_LEN;
3040 fsid = le64_to_cpup((void *)es->s_uuid) ^
3041 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3042 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3043 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
3047 /* Helper function for writing quotas on sync - we need to start transaction before quota file
3048 * is locked for write. Otherwise the are possible deadlocks:
3049 * Process 1 Process 2
3050 * ext4_create() quota_sync()
3051 * jbd2_journal_start() write_dquot()
3052 * DQUOT_INIT() down(dqio_mutex)
3053 * down(dqio_mutex) jbd2_journal_start()
3059 static inline struct inode *dquot_to_inode(struct dquot *dquot)
3061 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3064 static int ext4_dquot_initialize(struct inode *inode, int type)
3069 /* We may create quota structure so we need to reserve enough blocks */
3070 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_INIT_BLOCKS(inode->i_sb));
3072 return PTR_ERR(handle);
3073 ret = dquot_initialize(inode, type);
3074 err = ext4_journal_stop(handle);
3080 static int ext4_dquot_drop(struct inode *inode)
3085 /* We may delete quota structure so we need to reserve enough blocks */
3086 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_DEL_BLOCKS(inode->i_sb));
3087 if (IS_ERR(handle)) {
3089 * We call dquot_drop() anyway to at least release references
3090 * to quota structures so that umount does not hang.
3093 return PTR_ERR(handle);
3095 ret = dquot_drop(inode);
3096 err = ext4_journal_stop(handle);
3102 static int ext4_write_dquot(struct dquot *dquot)
3106 struct inode *inode;
3108 inode = dquot_to_inode(dquot);
3109 handle = ext4_journal_start(inode,
3110 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
3112 return PTR_ERR(handle);
3113 ret = dquot_commit(dquot);
3114 err = ext4_journal_stop(handle);
3120 static int ext4_acquire_dquot(struct dquot *dquot)
3125 handle = ext4_journal_start(dquot_to_inode(dquot),
3126 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
3128 return PTR_ERR(handle);
3129 ret = dquot_acquire(dquot);
3130 err = ext4_journal_stop(handle);
3136 static int ext4_release_dquot(struct dquot *dquot)
3141 handle = ext4_journal_start(dquot_to_inode(dquot),
3142 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
3143 if (IS_ERR(handle)) {
3144 /* Release dquot anyway to avoid endless cycle in dqput() */
3145 dquot_release(dquot);
3146 return PTR_ERR(handle);
3148 ret = dquot_release(dquot);
3149 err = ext4_journal_stop(handle);
3155 static int ext4_mark_dquot_dirty(struct dquot *dquot)
3157 /* Are we journaling quotas? */
3158 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3159 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
3160 dquot_mark_dquot_dirty(dquot);
3161 return ext4_write_dquot(dquot);
3163 return dquot_mark_dquot_dirty(dquot);
3167 static int ext4_write_info(struct super_block *sb, int type)
3172 /* Data block + inode block */
3173 handle = ext4_journal_start(sb->s_root->d_inode, 2);
3175 return PTR_ERR(handle);
3176 ret = dquot_commit_info(sb, type);
3177 err = ext4_journal_stop(handle);
3184 * Turn on quotas during mount time - we need to find
3185 * the quota file and such...
3187 static int ext4_quota_on_mount(struct super_block *sb, int type)
3189 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3190 EXT4_SB(sb)->s_jquota_fmt, type);
3194 * Standard function to be called on quota_on
3196 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
3197 char *path, int remount)
3200 struct nameidata nd;
3202 if (!test_opt(sb, QUOTA))
3204 /* When remounting, no checks are needed and in fact, path is NULL */
3206 return vfs_quota_on(sb, type, format_id, path, remount);
3208 err = path_lookup(path, LOOKUP_FOLLOW, &nd);
3212 /* Quotafile not on the same filesystem? */
3213 if (nd.path.mnt->mnt_sb != sb) {
3217 /* Journaling quota? */
3218 if (EXT4_SB(sb)->s_qf_names[type]) {
3219 /* Quotafile not of fs root? */
3220 if (nd.path.dentry->d_parent->d_inode != sb->s_root->d_inode)
3222 "EXT4-fs: Quota file not on filesystem root. "
3223 "Journaled quota will not work.\n");
3227 * When we journal data on quota file, we have to flush journal to see
3228 * all updates to the file when we bypass pagecache...
3230 if (ext4_should_journal_data(nd.path.dentry->d_inode)) {
3232 * We don't need to lock updates but journal_flush() could
3233 * otherwise be livelocked...
3235 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
3236 jbd2_journal_flush(EXT4_SB(sb)->s_journal);
3237 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3241 return vfs_quota_on(sb, type, format_id, path, remount);
3244 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3245 * acquiring the locks... As quota files are never truncated and quota code
3246 * itself serializes the operations (and noone else should touch the files)
3247 * we don't have to be afraid of races */
3248 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
3249 size_t len, loff_t off)
3251 struct inode *inode = sb_dqopt(sb)->files[type];
3252 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3254 int offset = off & (sb->s_blocksize - 1);
3257 struct buffer_head *bh;
3258 loff_t i_size = i_size_read(inode);
3262 if (off+len > i_size)
3265 while (toread > 0) {
3266 tocopy = sb->s_blocksize - offset < toread ?
3267 sb->s_blocksize - offset : toread;
3268 bh = ext4_bread(NULL, inode, blk, 0, &err);
3271 if (!bh) /* A hole? */
3272 memset(data, 0, tocopy);
3274 memcpy(data, bh->b_data+offset, tocopy);
3284 /* Write to quotafile (we know the transaction is already started and has
3285 * enough credits) */
3286 static ssize_t ext4_quota_write(struct super_block *sb, int type,
3287 const char *data, size_t len, loff_t off)
3289 struct inode *inode = sb_dqopt(sb)->files[type];
3290 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3292 int offset = off & (sb->s_blocksize - 1);
3294 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
3295 size_t towrite = len;
3296 struct buffer_head *bh;
3297 handle_t *handle = journal_current_handle();
3300 printk(KERN_WARNING "EXT4-fs: Quota write (off=%Lu, len=%Lu)"
3301 " cancelled because transaction is not started.\n",
3302 (unsigned long long)off, (unsigned long long)len);
3305 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3306 while (towrite > 0) {
3307 tocopy = sb->s_blocksize - offset < towrite ?
3308 sb->s_blocksize - offset : towrite;
3309 bh = ext4_bread(handle, inode, blk, 1, &err);
3312 if (journal_quota) {
3313 err = ext4_journal_get_write_access(handle, bh);
3320 memcpy(bh->b_data+offset, data, tocopy);
3321 flush_dcache_page(bh->b_page);
3324 err = ext4_journal_dirty_metadata(handle, bh);
3326 /* Always do at least ordered writes for quotas */
3327 err = ext4_journal_dirty_data(handle, bh);
3328 mark_buffer_dirty(bh);
3339 if (len == towrite) {
3340 mutex_unlock(&inode->i_mutex);
3343 if (inode->i_size < off+len-towrite) {
3344 i_size_write(inode, off+len-towrite);
3345 EXT4_I(inode)->i_disksize = inode->i_size;
3347 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3348 ext4_mark_inode_dirty(handle, inode);
3349 mutex_unlock(&inode->i_mutex);
3350 return len - towrite;
3355 static int ext4_get_sb(struct file_system_type *fs_type,
3356 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3358 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
3361 static struct file_system_type ext4dev_fs_type = {
3362 .owner = THIS_MODULE,
3364 .get_sb = ext4_get_sb,
3365 .kill_sb = kill_block_super,
3366 .fs_flags = FS_REQUIRES_DEV,
3369 static int __init init_ext4_fs(void)
3373 err = init_ext4_mballoc();
3377 err = init_ext4_xattr();
3380 err = init_inodecache();
3383 err = register_filesystem(&ext4dev_fs_type);
3388 destroy_inodecache();
3392 exit_ext4_mballoc();
3396 static void __exit exit_ext4_fs(void)
3398 unregister_filesystem(&ext4dev_fs_type);
3399 destroy_inodecache();
3401 exit_ext4_mballoc();
3404 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
3405 MODULE_DESCRIPTION("Fourth Extended Filesystem with extents");
3406 MODULE_LICENSE("GPL");
3407 module_init(init_ext4_fs)
3408 module_exit(exit_ext4_fs)