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, int sync);
54 static void ext4_mark_recovery_complete(struct super_block *sb,
55 struct ext4_super_block *es);
56 static void ext4_clear_journal_err(struct super_block *sb,
57 struct ext4_super_block *es);
58 static int ext4_sync_fs(struct super_block *sb, int wait);
59 static const char *ext4_decode_error(struct super_block *sb, int errno,
61 static int ext4_remount(struct super_block *sb, int *flags, char *data);
62 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
63 static void ext4_unlockfs(struct super_block *sb);
64 static void ext4_write_super(struct super_block *sb);
65 static void ext4_write_super_lockfs(struct super_block *sb);
68 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
69 struct ext4_group_desc *bg)
71 return le32_to_cpu(bg->bg_block_bitmap_lo) |
72 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
73 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
76 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
77 struct ext4_group_desc *bg)
79 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
80 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
81 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
84 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
85 struct ext4_group_desc *bg)
87 return le32_to_cpu(bg->bg_inode_table_lo) |
88 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
89 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
92 void ext4_block_bitmap_set(struct super_block *sb,
93 struct ext4_group_desc *bg, ext4_fsblk_t blk)
95 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
96 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
97 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
100 void ext4_inode_bitmap_set(struct super_block *sb,
101 struct ext4_group_desc *bg, ext4_fsblk_t blk)
103 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
104 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
105 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
108 void ext4_inode_table_set(struct super_block *sb,
109 struct ext4_group_desc *bg, ext4_fsblk_t blk)
111 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
112 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
113 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
117 * Wrappers for jbd2_journal_start/end.
119 * The only special thing we need to do here is to make sure that all
120 * journal_end calls result in the superblock being marked dirty, so
121 * that sync() will call the filesystem's write_super callback if
124 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
128 if (sb->s_flags & MS_RDONLY)
129 return ERR_PTR(-EROFS);
131 /* Special case here: if the journal has aborted behind our
132 * backs (eg. EIO in the commit thread), then we still need to
133 * take the FS itself readonly cleanly. */
134 journal = EXT4_SB(sb)->s_journal;
135 if (is_journal_aborted(journal)) {
136 ext4_abort(sb, __func__,
137 "Detected aborted journal");
138 return ERR_PTR(-EROFS);
141 return jbd2_journal_start(journal, nblocks);
145 * The only special thing we need to do here is to make sure that all
146 * jbd2_journal_stop calls result in the superblock being marked dirty, so
147 * that sync() will call the filesystem's write_super callback if
150 int __ext4_journal_stop(const char *where, handle_t *handle)
152 struct super_block *sb;
156 sb = handle->h_transaction->t_journal->j_private;
158 rc = jbd2_journal_stop(handle);
163 __ext4_std_error(sb, where, err);
167 void ext4_journal_abort_handle(const char *caller, const char *err_fn,
168 struct buffer_head *bh, handle_t *handle, int err)
171 const char *errstr = ext4_decode_error(NULL, err, nbuf);
174 BUFFER_TRACE(bh, "abort");
179 if (is_handle_aborted(handle))
182 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
183 caller, errstr, err_fn);
185 jbd2_journal_abort_handle(handle);
188 /* Deal with the reporting of failure conditions on a filesystem such as
189 * inconsistencies detected or read IO failures.
191 * On ext2, we can store the error state of the filesystem in the
192 * superblock. That is not possible on ext4, because we may have other
193 * write ordering constraints on the superblock which prevent us from
194 * writing it out straight away; and given that the journal is about to
195 * be aborted, we can't rely on the current, or future, transactions to
196 * write out the superblock safely.
198 * We'll just use the jbd2_journal_abort() error code to record an error in
199 * the journal instead. On recovery, the journal will compain about
200 * that error until we've noted it down and cleared it.
203 static void ext4_handle_error(struct super_block *sb)
205 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
207 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
208 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
210 if (sb->s_flags & MS_RDONLY)
213 if (!test_opt(sb, ERRORS_CONT)) {
214 journal_t *journal = EXT4_SB(sb)->s_journal;
216 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
218 jbd2_journal_abort(journal, -EIO);
220 if (test_opt(sb, ERRORS_RO)) {
221 printk(KERN_CRIT "Remounting filesystem read-only\n");
222 sb->s_flags |= MS_RDONLY;
224 ext4_commit_super(sb, es, 1);
225 if (test_opt(sb, ERRORS_PANIC))
226 panic("EXT4-fs (device %s): panic forced after error\n",
230 void ext4_error(struct super_block *sb, const char *function,
231 const char *fmt, ...)
236 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
241 ext4_handle_error(sb);
244 static const char *ext4_decode_error(struct super_block *sb, int errno,
251 errstr = "IO failure";
254 errstr = "Out of memory";
257 if (!sb || EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)
258 errstr = "Journal has aborted";
260 errstr = "Readonly filesystem";
263 /* If the caller passed in an extra buffer for unknown
264 * errors, textualise them now. Else we just return
267 /* Check for truncated error codes... */
268 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
277 /* __ext4_std_error decodes expected errors from journaling functions
278 * automatically and invokes the appropriate error response. */
280 void __ext4_std_error(struct super_block *sb, const char *function, int errno)
285 /* Special case: if the error is EROFS, and we're not already
286 * inside a transaction, then there's really no point in logging
288 if (errno == -EROFS && journal_current_handle() == NULL &&
289 (sb->s_flags & MS_RDONLY))
292 errstr = ext4_decode_error(sb, errno, nbuf);
293 printk(KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
294 sb->s_id, function, errstr);
296 ext4_handle_error(sb);
300 * ext4_abort is a much stronger failure handler than ext4_error. The
301 * abort function may be used to deal with unrecoverable failures such
302 * as journal IO errors or ENOMEM at a critical moment in log management.
304 * We unconditionally force the filesystem into an ABORT|READONLY state,
305 * unless the error response on the fs has been set to panic in which
306 * case we take the easy way out and panic immediately.
309 void ext4_abort(struct super_block *sb, const char *function,
310 const char *fmt, ...)
314 printk(KERN_CRIT "ext4_abort called.\n");
317 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
322 if (test_opt(sb, ERRORS_PANIC))
323 panic("EXT4-fs panic from previous error\n");
325 if (sb->s_flags & MS_RDONLY)
328 printk(KERN_CRIT "Remounting filesystem read-only\n");
329 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
330 sb->s_flags |= MS_RDONLY;
331 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
332 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
335 void ext4_warning(struct super_block *sb, const char *function,
336 const char *fmt, ...)
341 printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
348 void ext4_update_dynamic_rev(struct super_block *sb)
350 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
352 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
355 ext4_warning(sb, __func__,
356 "updating to rev %d because of new feature flag, "
357 "running e2fsck is recommended",
360 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
361 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
362 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
363 /* leave es->s_feature_*compat flags alone */
364 /* es->s_uuid will be set by e2fsck if empty */
367 * The rest of the superblock fields should be zero, and if not it
368 * means they are likely already in use, so leave them alone. We
369 * can leave it up to e2fsck to clean up any inconsistencies there.
373 int ext4_update_compat_feature(handle_t *handle,
374 struct super_block *sb, __u32 compat)
377 if (!EXT4_HAS_COMPAT_FEATURE(sb, compat)) {
378 err = ext4_journal_get_write_access(handle,
382 EXT4_SET_COMPAT_FEATURE(sb, compat);
385 BUFFER_TRACE(EXT4_SB(sb)->s_sbh,
386 "call ext4_journal_dirty_met adata");
387 err = ext4_journal_dirty_metadata(handle,
393 int ext4_update_rocompat_feature(handle_t *handle,
394 struct super_block *sb, __u32 rocompat)
397 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, rocompat)) {
398 err = ext4_journal_get_write_access(handle,
402 EXT4_SET_RO_COMPAT_FEATURE(sb, rocompat);
405 BUFFER_TRACE(EXT4_SB(sb)->s_sbh,
406 "call ext4_journal_dirty_met adata");
407 err = ext4_journal_dirty_metadata(handle,
413 int ext4_update_incompat_feature(handle_t *handle,
414 struct super_block *sb, __u32 incompat)
417 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, incompat)) {
418 err = ext4_journal_get_write_access(handle,
422 EXT4_SET_INCOMPAT_FEATURE(sb, incompat);
425 BUFFER_TRACE(EXT4_SB(sb)->s_sbh,
426 "call ext4_journal_dirty_met adata");
427 err = ext4_journal_dirty_metadata(handle,
434 * Open the external journal device
436 static struct block_device *ext4_blkdev_get(dev_t dev)
438 struct block_device *bdev;
439 char b[BDEVNAME_SIZE];
441 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
447 printk(KERN_ERR "EXT4: failed to open journal device %s: %ld\n",
448 __bdevname(dev, b), PTR_ERR(bdev));
453 * Release the journal device
455 static int ext4_blkdev_put(struct block_device *bdev)
458 return blkdev_put(bdev);
461 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
463 struct block_device *bdev;
466 bdev = sbi->journal_bdev;
468 ret = ext4_blkdev_put(bdev);
469 sbi->journal_bdev = NULL;
474 static inline struct inode *orphan_list_entry(struct list_head *l)
476 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
479 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
483 printk(KERN_ERR "sb orphan head is %d\n",
484 le32_to_cpu(sbi->s_es->s_last_orphan));
486 printk(KERN_ERR "sb_info orphan list:\n");
487 list_for_each(l, &sbi->s_orphan) {
488 struct inode *inode = orphan_list_entry(l);
490 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
491 inode->i_sb->s_id, inode->i_ino, inode,
492 inode->i_mode, inode->i_nlink,
497 static void ext4_put_super(struct super_block *sb)
499 struct ext4_sb_info *sbi = EXT4_SB(sb);
500 struct ext4_super_block *es = sbi->s_es;
504 ext4_ext_release(sb);
505 ext4_xattr_put_super(sb);
506 jbd2_journal_destroy(sbi->s_journal);
507 sbi->s_journal = NULL;
508 if (!(sb->s_flags & MS_RDONLY)) {
509 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
510 es->s_state = cpu_to_le16(sbi->s_mount_state);
511 BUFFER_TRACE(sbi->s_sbh, "marking dirty");
512 mark_buffer_dirty(sbi->s_sbh);
513 ext4_commit_super(sb, es, 1);
516 for (i = 0; i < sbi->s_gdb_count; i++)
517 brelse(sbi->s_group_desc[i]);
518 kfree(sbi->s_group_desc);
519 kfree(sbi->s_flex_groups);
520 percpu_counter_destroy(&sbi->s_freeblocks_counter);
521 percpu_counter_destroy(&sbi->s_freeinodes_counter);
522 percpu_counter_destroy(&sbi->s_dirs_counter);
523 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
526 for (i = 0; i < MAXQUOTAS; i++)
527 kfree(sbi->s_qf_names[i]);
530 /* Debugging code just in case the in-memory inode orphan list
531 * isn't empty. The on-disk one can be non-empty if we've
532 * detected an error and taken the fs readonly, but the
533 * in-memory list had better be clean by this point. */
534 if (!list_empty(&sbi->s_orphan))
535 dump_orphan_list(sb, sbi);
536 J_ASSERT(list_empty(&sbi->s_orphan));
538 invalidate_bdev(sb->s_bdev);
539 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
541 * Invalidate the journal device's buffers. We don't want them
542 * floating about in memory - the physical journal device may
543 * hotswapped, and it breaks the `ro-after' testing code.
545 sync_blockdev(sbi->journal_bdev);
546 invalidate_bdev(sbi->journal_bdev);
547 ext4_blkdev_remove(sbi);
549 sb->s_fs_info = NULL;
554 static struct kmem_cache *ext4_inode_cachep;
557 * Called inside transaction, so use GFP_NOFS
559 static struct inode *ext4_alloc_inode(struct super_block *sb)
561 struct ext4_inode_info *ei;
563 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
566 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
567 ei->i_acl = EXT4_ACL_NOT_CACHED;
568 ei->i_default_acl = EXT4_ACL_NOT_CACHED;
570 ei->i_block_alloc_info = NULL;
571 ei->vfs_inode.i_version = 1;
572 ei->vfs_inode.i_data.writeback_index = 0;
573 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
574 INIT_LIST_HEAD(&ei->i_prealloc_list);
575 spin_lock_init(&ei->i_prealloc_lock);
576 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
577 ei->i_reserved_data_blocks = 0;
578 ei->i_reserved_meta_blocks = 0;
579 ei->i_allocated_meta_blocks = 0;
580 ei->i_delalloc_reserved_flag = 0;
581 spin_lock_init(&(ei->i_block_reservation_lock));
582 return &ei->vfs_inode;
585 static void ext4_destroy_inode(struct inode *inode)
587 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
588 printk("EXT4 Inode %p: orphan list check failed!\n",
590 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
591 EXT4_I(inode), sizeof(struct ext4_inode_info),
595 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
598 static void init_once(void *foo)
600 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
602 INIT_LIST_HEAD(&ei->i_orphan);
603 #ifdef CONFIG_EXT4DEV_FS_XATTR
604 init_rwsem(&ei->xattr_sem);
606 init_rwsem(&ei->i_data_sem);
607 inode_init_once(&ei->vfs_inode);
610 static int init_inodecache(void)
612 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
613 sizeof(struct ext4_inode_info),
614 0, (SLAB_RECLAIM_ACCOUNT|
617 if (ext4_inode_cachep == NULL)
622 static void destroy_inodecache(void)
624 kmem_cache_destroy(ext4_inode_cachep);
627 static void ext4_clear_inode(struct inode *inode)
629 struct ext4_block_alloc_info *rsv = EXT4_I(inode)->i_block_alloc_info;
630 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
631 if (EXT4_I(inode)->i_acl &&
632 EXT4_I(inode)->i_acl != EXT4_ACL_NOT_CACHED) {
633 posix_acl_release(EXT4_I(inode)->i_acl);
634 EXT4_I(inode)->i_acl = EXT4_ACL_NOT_CACHED;
636 if (EXT4_I(inode)->i_default_acl &&
637 EXT4_I(inode)->i_default_acl != EXT4_ACL_NOT_CACHED) {
638 posix_acl_release(EXT4_I(inode)->i_default_acl);
639 EXT4_I(inode)->i_default_acl = EXT4_ACL_NOT_CACHED;
642 ext4_discard_reservation(inode);
643 EXT4_I(inode)->i_block_alloc_info = NULL;
646 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
647 &EXT4_I(inode)->jinode);
650 static inline void ext4_show_quota_options(struct seq_file *seq,
651 struct super_block *sb)
653 #if defined(CONFIG_QUOTA)
654 struct ext4_sb_info *sbi = EXT4_SB(sb);
656 if (sbi->s_jquota_fmt)
657 seq_printf(seq, ",jqfmt=%s",
658 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold" : "vfsv0");
660 if (sbi->s_qf_names[USRQUOTA])
661 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
663 if (sbi->s_qf_names[GRPQUOTA])
664 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
666 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
667 seq_puts(seq, ",usrquota");
669 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
670 seq_puts(seq, ",grpquota");
676 * - it's set to a non-default value OR
677 * - if the per-sb default is different from the global default
679 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
682 unsigned long def_mount_opts;
683 struct super_block *sb = vfs->mnt_sb;
684 struct ext4_sb_info *sbi = EXT4_SB(sb);
685 struct ext4_super_block *es = sbi->s_es;
687 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
688 def_errors = le16_to_cpu(es->s_errors);
690 if (sbi->s_sb_block != 1)
691 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
692 if (test_opt(sb, MINIX_DF))
693 seq_puts(seq, ",minixdf");
694 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
695 seq_puts(seq, ",grpid");
696 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
697 seq_puts(seq, ",nogrpid");
698 if (sbi->s_resuid != EXT4_DEF_RESUID ||
699 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
700 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
702 if (sbi->s_resgid != EXT4_DEF_RESGID ||
703 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
704 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
706 if (test_opt(sb, ERRORS_RO)) {
707 if (def_errors == EXT4_ERRORS_PANIC ||
708 def_errors == EXT4_ERRORS_CONTINUE) {
709 seq_puts(seq, ",errors=remount-ro");
712 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
713 seq_puts(seq, ",errors=continue");
714 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
715 seq_puts(seq, ",errors=panic");
716 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
717 seq_puts(seq, ",nouid32");
718 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
719 seq_puts(seq, ",debug");
720 if (test_opt(sb, OLDALLOC))
721 seq_puts(seq, ",oldalloc");
722 #ifdef CONFIG_EXT4DEV_FS_XATTR
723 if (test_opt(sb, XATTR_USER) &&
724 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
725 seq_puts(seq, ",user_xattr");
726 if (!test_opt(sb, XATTR_USER) &&
727 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
728 seq_puts(seq, ",nouser_xattr");
731 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
732 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
733 seq_puts(seq, ",acl");
734 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
735 seq_puts(seq, ",noacl");
737 if (!test_opt(sb, RESERVATION))
738 seq_puts(seq, ",noreservation");
739 if (sbi->s_commit_interval) {
740 seq_printf(seq, ",commit=%u",
741 (unsigned) (sbi->s_commit_interval / HZ));
744 * We're changing the default of barrier mount option, so
745 * let's always display its mount state so it's clear what its
748 seq_puts(seq, ",barrier=");
749 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
750 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
751 seq_puts(seq, ",journal_async_commit");
752 if (test_opt(sb, NOBH))
753 seq_puts(seq, ",nobh");
754 if (!test_opt(sb, EXTENTS))
755 seq_puts(seq, ",noextents");
756 if (!test_opt(sb, MBALLOC))
757 seq_puts(seq, ",nomballoc");
758 if (test_opt(sb, I_VERSION))
759 seq_puts(seq, ",i_version");
760 if (!test_opt(sb, DELALLOC))
761 seq_puts(seq, ",nodelalloc");
765 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
767 * journal mode get enabled in different ways
768 * So just print the value even if we didn't specify it
770 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
771 seq_puts(seq, ",data=journal");
772 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
773 seq_puts(seq, ",data=ordered");
774 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
775 seq_puts(seq, ",data=writeback");
777 ext4_show_quota_options(seq, sb);
782 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
783 u64 ino, u32 generation)
787 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
788 return ERR_PTR(-ESTALE);
789 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
790 return ERR_PTR(-ESTALE);
792 /* iget isn't really right if the inode is currently unallocated!!
794 * ext4_read_inode will return a bad_inode if the inode had been
795 * deleted, so we should be safe.
797 * Currently we don't know the generation for parent directory, so
798 * a generation of 0 means "accept any"
800 inode = ext4_iget(sb, ino);
802 return ERR_CAST(inode);
803 if (generation && inode->i_generation != generation) {
805 return ERR_PTR(-ESTALE);
811 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
812 int fh_len, int fh_type)
814 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
818 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
819 int fh_len, int fh_type)
821 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
826 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
827 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
829 static int ext4_dquot_initialize(struct inode *inode, int type);
830 static int ext4_dquot_drop(struct inode *inode);
831 static int ext4_write_dquot(struct dquot *dquot);
832 static int ext4_acquire_dquot(struct dquot *dquot);
833 static int ext4_release_dquot(struct dquot *dquot);
834 static int ext4_mark_dquot_dirty(struct dquot *dquot);
835 static int ext4_write_info(struct super_block *sb, int type);
836 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
837 char *path, int remount);
838 static int ext4_quota_on_mount(struct super_block *sb, int type);
839 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
840 size_t len, loff_t off);
841 static ssize_t ext4_quota_write(struct super_block *sb, int type,
842 const char *data, size_t len, loff_t off);
844 static struct dquot_operations ext4_quota_operations = {
845 .initialize = ext4_dquot_initialize,
846 .drop = ext4_dquot_drop,
847 .alloc_space = dquot_alloc_space,
848 .alloc_inode = dquot_alloc_inode,
849 .free_space = dquot_free_space,
850 .free_inode = dquot_free_inode,
851 .transfer = dquot_transfer,
852 .write_dquot = ext4_write_dquot,
853 .acquire_dquot = ext4_acquire_dquot,
854 .release_dquot = ext4_release_dquot,
855 .mark_dirty = ext4_mark_dquot_dirty,
856 .write_info = ext4_write_info
859 static struct quotactl_ops ext4_qctl_operations = {
860 .quota_on = ext4_quota_on,
861 .quota_off = vfs_quota_off,
862 .quota_sync = vfs_quota_sync,
863 .get_info = vfs_get_dqinfo,
864 .set_info = vfs_set_dqinfo,
865 .get_dqblk = vfs_get_dqblk,
866 .set_dqblk = vfs_set_dqblk
870 static const struct super_operations ext4_sops = {
871 .alloc_inode = ext4_alloc_inode,
872 .destroy_inode = ext4_destroy_inode,
873 .write_inode = ext4_write_inode,
874 .dirty_inode = ext4_dirty_inode,
875 .delete_inode = ext4_delete_inode,
876 .put_super = ext4_put_super,
877 .write_super = ext4_write_super,
878 .sync_fs = ext4_sync_fs,
879 .write_super_lockfs = ext4_write_super_lockfs,
880 .unlockfs = ext4_unlockfs,
881 .statfs = ext4_statfs,
882 .remount_fs = ext4_remount,
883 .clear_inode = ext4_clear_inode,
884 .show_options = ext4_show_options,
886 .quota_read = ext4_quota_read,
887 .quota_write = ext4_quota_write,
891 static const struct export_operations ext4_export_ops = {
892 .fh_to_dentry = ext4_fh_to_dentry,
893 .fh_to_parent = ext4_fh_to_parent,
894 .get_parent = ext4_get_parent,
898 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
899 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
900 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
901 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
902 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
903 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
904 Opt_journal_checksum, Opt_journal_async_commit,
905 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
906 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
907 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
908 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
909 Opt_grpquota, Opt_extents, Opt_noextents, Opt_i_version,
910 Opt_mballoc, Opt_nomballoc, Opt_stripe, Opt_delalloc, Opt_nodelalloc,
913 static match_table_t tokens = {
914 {Opt_bsd_df, "bsddf"},
915 {Opt_minix_df, "minixdf"},
916 {Opt_grpid, "grpid"},
917 {Opt_grpid, "bsdgroups"},
918 {Opt_nogrpid, "nogrpid"},
919 {Opt_nogrpid, "sysvgroups"},
920 {Opt_resgid, "resgid=%u"},
921 {Opt_resuid, "resuid=%u"},
923 {Opt_err_cont, "errors=continue"},
924 {Opt_err_panic, "errors=panic"},
925 {Opt_err_ro, "errors=remount-ro"},
926 {Opt_nouid32, "nouid32"},
927 {Opt_nocheck, "nocheck"},
928 {Opt_nocheck, "check=none"},
929 {Opt_debug, "debug"},
930 {Opt_oldalloc, "oldalloc"},
931 {Opt_orlov, "orlov"},
932 {Opt_user_xattr, "user_xattr"},
933 {Opt_nouser_xattr, "nouser_xattr"},
935 {Opt_noacl, "noacl"},
936 {Opt_reservation, "reservation"},
937 {Opt_noreservation, "noreservation"},
938 {Opt_noload, "noload"},
941 {Opt_commit, "commit=%u"},
942 {Opt_journal_update, "journal=update"},
943 {Opt_journal_inum, "journal=%u"},
944 {Opt_journal_dev, "journal_dev=%u"},
945 {Opt_journal_checksum, "journal_checksum"},
946 {Opt_journal_async_commit, "journal_async_commit"},
947 {Opt_abort, "abort"},
948 {Opt_data_journal, "data=journal"},
949 {Opt_data_ordered, "data=ordered"},
950 {Opt_data_writeback, "data=writeback"},
951 {Opt_offusrjquota, "usrjquota="},
952 {Opt_usrjquota, "usrjquota=%s"},
953 {Opt_offgrpjquota, "grpjquota="},
954 {Opt_grpjquota, "grpjquota=%s"},
955 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
956 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
957 {Opt_grpquota, "grpquota"},
958 {Opt_noquota, "noquota"},
959 {Opt_quota, "quota"},
960 {Opt_usrquota, "usrquota"},
961 {Opt_barrier, "barrier=%u"},
962 {Opt_extents, "extents"},
963 {Opt_noextents, "noextents"},
964 {Opt_i_version, "i_version"},
965 {Opt_mballoc, "mballoc"},
966 {Opt_nomballoc, "nomballoc"},
967 {Opt_stripe, "stripe=%u"},
968 {Opt_resize, "resize"},
969 {Opt_delalloc, "delalloc"},
970 {Opt_nodelalloc, "nodelalloc"},
974 static ext4_fsblk_t get_sb_block(void **data)
976 ext4_fsblk_t sb_block;
977 char *options = (char *) *data;
979 if (!options || strncmp(options, "sb=", 3) != 0)
980 return 1; /* Default location */
982 /*todo: use simple_strtoll with >32bit ext4 */
983 sb_block = simple_strtoul(options, &options, 0);
984 if (*options && *options != ',') {
985 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
991 *data = (void *) options;
995 static int parse_options(char *options, struct super_block *sb,
996 unsigned int *inum, unsigned long *journal_devnum,
997 ext4_fsblk_t *n_blocks_count, int is_remount)
999 struct ext4_sb_info *sbi = EXT4_SB(sb);
1001 substring_t args[MAX_OPT_ARGS];
1008 ext4_fsblk_t last_block;
1013 while ((p = strsep(&options, ",")) != NULL) {
1018 token = match_token(p, tokens, args);
1021 clear_opt(sbi->s_mount_opt, MINIX_DF);
1024 set_opt(sbi->s_mount_opt, MINIX_DF);
1027 set_opt(sbi->s_mount_opt, GRPID);
1030 clear_opt(sbi->s_mount_opt, GRPID);
1033 if (match_int(&args[0], &option))
1035 sbi->s_resuid = option;
1038 if (match_int(&args[0], &option))
1040 sbi->s_resgid = option;
1043 /* handled by get_sb_block() instead of here */
1044 /* *sb_block = match_int(&args[0]); */
1047 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1048 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1049 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1052 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1053 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1054 set_opt(sbi->s_mount_opt, ERRORS_RO);
1057 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1058 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1059 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1062 set_opt(sbi->s_mount_opt, NO_UID32);
1065 clear_opt(sbi->s_mount_opt, CHECK);
1068 set_opt(sbi->s_mount_opt, DEBUG);
1071 set_opt(sbi->s_mount_opt, OLDALLOC);
1074 clear_opt(sbi->s_mount_opt, OLDALLOC);
1076 #ifdef CONFIG_EXT4DEV_FS_XATTR
1077 case Opt_user_xattr:
1078 set_opt(sbi->s_mount_opt, XATTR_USER);
1080 case Opt_nouser_xattr:
1081 clear_opt(sbi->s_mount_opt, XATTR_USER);
1084 case Opt_user_xattr:
1085 case Opt_nouser_xattr:
1086 printk(KERN_ERR "EXT4 (no)user_xattr options "
1090 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
1092 set_opt(sbi->s_mount_opt, POSIX_ACL);
1095 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1100 printk(KERN_ERR "EXT4 (no)acl options "
1104 case Opt_reservation:
1105 set_opt(sbi->s_mount_opt, RESERVATION);
1107 case Opt_noreservation:
1108 clear_opt(sbi->s_mount_opt, RESERVATION);
1110 case Opt_journal_update:
1112 /* Eventually we will want to be able to create
1113 a journal file here. For now, only allow the
1114 user to specify an existing inode to be the
1117 printk(KERN_ERR "EXT4-fs: cannot specify "
1118 "journal on remount\n");
1121 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
1123 case Opt_journal_inum:
1125 printk(KERN_ERR "EXT4-fs: cannot specify "
1126 "journal on remount\n");
1129 if (match_int(&args[0], &option))
1133 case Opt_journal_dev:
1135 printk(KERN_ERR "EXT4-fs: cannot specify "
1136 "journal on remount\n");
1139 if (match_int(&args[0], &option))
1141 *journal_devnum = option;
1143 case Opt_journal_checksum:
1144 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1146 case Opt_journal_async_commit:
1147 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1148 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1151 set_opt(sbi->s_mount_opt, NOLOAD);
1154 if (match_int(&args[0], &option))
1159 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1160 sbi->s_commit_interval = HZ * option;
1162 case Opt_data_journal:
1163 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1165 case Opt_data_ordered:
1166 data_opt = EXT4_MOUNT_ORDERED_DATA;
1168 case Opt_data_writeback:
1169 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1172 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
1175 "EXT4-fs: cannot change data "
1176 "mode on remount\n");
1180 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
1181 sbi->s_mount_opt |= data_opt;
1191 if ((sb_any_quota_enabled(sb) ||
1192 sb_any_quota_suspended(sb)) &&
1193 !sbi->s_qf_names[qtype]) {
1195 "EXT4-fs: Cannot change journaled "
1196 "quota options when quota turned on.\n");
1199 qname = match_strdup(&args[0]);
1202 "EXT4-fs: not enough memory for "
1203 "storing quotafile name.\n");
1206 if (sbi->s_qf_names[qtype] &&
1207 strcmp(sbi->s_qf_names[qtype], qname)) {
1209 "EXT4-fs: %s quota file already "
1210 "specified.\n", QTYPE2NAME(qtype));
1214 sbi->s_qf_names[qtype] = qname;
1215 if (strchr(sbi->s_qf_names[qtype], '/')) {
1217 "EXT4-fs: quotafile must be on "
1218 "filesystem root.\n");
1219 kfree(sbi->s_qf_names[qtype]);
1220 sbi->s_qf_names[qtype] = NULL;
1223 set_opt(sbi->s_mount_opt, QUOTA);
1225 case Opt_offusrjquota:
1228 case Opt_offgrpjquota:
1231 if ((sb_any_quota_enabled(sb) ||
1232 sb_any_quota_suspended(sb)) &&
1233 sbi->s_qf_names[qtype]) {
1234 printk(KERN_ERR "EXT4-fs: Cannot change "
1235 "journaled quota options when "
1236 "quota turned on.\n");
1240 * The space will be released later when all options
1241 * are confirmed to be correct
1243 sbi->s_qf_names[qtype] = NULL;
1245 case Opt_jqfmt_vfsold:
1246 qfmt = QFMT_VFS_OLD;
1248 case Opt_jqfmt_vfsv0:
1251 if ((sb_any_quota_enabled(sb) ||
1252 sb_any_quota_suspended(sb)) &&
1253 sbi->s_jquota_fmt != qfmt) {
1254 printk(KERN_ERR "EXT4-fs: Cannot change "
1255 "journaled quota options when "
1256 "quota turned on.\n");
1259 sbi->s_jquota_fmt = qfmt;
1263 set_opt(sbi->s_mount_opt, QUOTA);
1264 set_opt(sbi->s_mount_opt, USRQUOTA);
1267 set_opt(sbi->s_mount_opt, QUOTA);
1268 set_opt(sbi->s_mount_opt, GRPQUOTA);
1271 if (sb_any_quota_enabled(sb)) {
1272 printk(KERN_ERR "EXT4-fs: Cannot change quota "
1273 "options when quota turned on.\n");
1276 clear_opt(sbi->s_mount_opt, QUOTA);
1277 clear_opt(sbi->s_mount_opt, USRQUOTA);
1278 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1285 "EXT4-fs: quota options not supported.\n");
1289 case Opt_offusrjquota:
1290 case Opt_offgrpjquota:
1291 case Opt_jqfmt_vfsold:
1292 case Opt_jqfmt_vfsv0:
1294 "EXT4-fs: journaled quota options not "
1301 set_opt(sbi->s_mount_opt, ABORT);
1304 if (match_int(&args[0], &option))
1307 set_opt(sbi->s_mount_opt, BARRIER);
1309 clear_opt(sbi->s_mount_opt, BARRIER);
1315 printk("EXT4-fs: resize option only available "
1319 if (match_int(&args[0], &option) != 0)
1321 *n_blocks_count = option;
1324 set_opt(sbi->s_mount_opt, NOBH);
1327 clear_opt(sbi->s_mount_opt, NOBH);
1330 if (!EXT4_HAS_INCOMPAT_FEATURE(sb,
1331 EXT4_FEATURE_INCOMPAT_EXTENTS)) {
1332 ext4_warning(sb, __func__,
1333 "extents feature not enabled "
1334 "on this filesystem, use tune2fs\n");
1337 set_opt(sbi->s_mount_opt, EXTENTS);
1341 * When e2fsprogs support resizing an already existing
1342 * ext3 file system to greater than 2**32 we need to
1343 * add support to block allocator to handle growing
1344 * already existing block mapped inode so that blocks
1345 * allocated for them fall within 2**32
1347 last_block = ext4_blocks_count(sbi->s_es) - 1;
1348 if (last_block > 0xffffffffULL) {
1349 printk(KERN_ERR "EXT4-fs: Filesystem too "
1350 "large to mount with "
1351 "-o noextents options\n");
1354 clear_opt(sbi->s_mount_opt, EXTENTS);
1357 set_opt(sbi->s_mount_opt, I_VERSION);
1358 sb->s_flags |= MS_I_VERSION;
1360 case Opt_nodelalloc:
1361 clear_opt(sbi->s_mount_opt, DELALLOC);
1364 set_opt(sbi->s_mount_opt, MBALLOC);
1367 clear_opt(sbi->s_mount_opt, MBALLOC);
1370 if (match_int(&args[0], &option))
1374 sbi->s_stripe = option;
1377 set_opt(sbi->s_mount_opt, DELALLOC);
1381 "EXT4-fs: Unrecognized mount option \"%s\" "
1382 "or missing value\n", p);
1387 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1388 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
1389 sbi->s_qf_names[USRQUOTA])
1390 clear_opt(sbi->s_mount_opt, USRQUOTA);
1392 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
1393 sbi->s_qf_names[GRPQUOTA])
1394 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1396 if ((sbi->s_qf_names[USRQUOTA] &&
1397 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
1398 (sbi->s_qf_names[GRPQUOTA] &&
1399 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1400 printk(KERN_ERR "EXT4-fs: old and new quota "
1401 "format mixing.\n");
1405 if (!sbi->s_jquota_fmt) {
1406 printk(KERN_ERR "EXT4-fs: journaled quota format "
1407 "not specified.\n");
1411 if (sbi->s_jquota_fmt) {
1412 printk(KERN_ERR "EXT4-fs: journaled quota format "
1413 "specified with no journaling "
1422 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1425 struct ext4_sb_info *sbi = EXT4_SB(sb);
1428 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1429 printk(KERN_ERR "EXT4-fs warning: revision level too high, "
1430 "forcing read-only mode\n");
1435 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1436 printk(KERN_WARNING "EXT4-fs warning: mounting unchecked fs, "
1437 "running e2fsck is recommended\n");
1438 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1440 "EXT4-fs warning: mounting fs with errors, "
1441 "running e2fsck is recommended\n");
1442 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1443 le16_to_cpu(es->s_mnt_count) >=
1444 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1446 "EXT4-fs warning: maximal mount count reached, "
1447 "running e2fsck is recommended\n");
1448 else if (le32_to_cpu(es->s_checkinterval) &&
1449 (le32_to_cpu(es->s_lastcheck) +
1450 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1452 "EXT4-fs warning: checktime reached, "
1453 "running e2fsck is recommended\n");
1455 /* @@@ We _will_ want to clear the valid bit if we find
1456 * inconsistencies, to force a fsck at reboot. But for
1457 * a plain journaled filesystem we can keep it set as
1460 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1462 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1463 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1464 le16_add_cpu(&es->s_mnt_count, 1);
1465 es->s_mtime = cpu_to_le32(get_seconds());
1466 ext4_update_dynamic_rev(sb);
1467 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1469 ext4_commit_super(sb, es, 1);
1470 if (test_opt(sb, DEBUG))
1471 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%lu, "
1472 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1474 sbi->s_groups_count,
1475 EXT4_BLOCKS_PER_GROUP(sb),
1476 EXT4_INODES_PER_GROUP(sb),
1479 if (EXT4_SB(sb)->s_journal->j_inode == NULL) {
1480 char b[BDEVNAME_SIZE];
1482 printk(KERN_INFO "EXT4 FS on %s, external journal on %s\n",
1483 sb->s_id, bdevname(EXT4_SB(sb)->s_journal->j_dev, b));
1485 printk(KERN_INFO "EXT4 FS on %s, internal journal\n",
1491 static int ext4_fill_flex_info(struct super_block *sb)
1493 struct ext4_sb_info *sbi = EXT4_SB(sb);
1494 struct ext4_group_desc *gdp = NULL;
1495 struct buffer_head *bh;
1496 ext4_group_t flex_group_count;
1497 ext4_group_t flex_group;
1498 int groups_per_flex = 0;
1499 __u64 block_bitmap = 0;
1502 if (!sbi->s_es->s_log_groups_per_flex) {
1503 sbi->s_log_groups_per_flex = 0;
1507 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1508 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1510 flex_group_count = (sbi->s_groups_count + groups_per_flex - 1) /
1512 sbi->s_flex_groups = kzalloc(flex_group_count *
1513 sizeof(struct flex_groups), GFP_KERNEL);
1514 if (sbi->s_flex_groups == NULL) {
1515 printk(KERN_ERR "EXT4-fs: not enough memory for "
1516 "%lu flex groups\n", flex_group_count);
1520 gdp = ext4_get_group_desc(sb, 1, &bh);
1521 block_bitmap = ext4_block_bitmap(sb, gdp) - 1;
1523 for (i = 0; i < sbi->s_groups_count; i++) {
1524 gdp = ext4_get_group_desc(sb, i, &bh);
1526 flex_group = ext4_flex_group(sbi, i);
1527 sbi->s_flex_groups[flex_group].free_inodes +=
1528 le16_to_cpu(gdp->bg_free_inodes_count);
1529 sbi->s_flex_groups[flex_group].free_blocks +=
1530 le16_to_cpu(gdp->bg_free_blocks_count);
1538 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1539 struct ext4_group_desc *gdp)
1543 if (sbi->s_es->s_feature_ro_compat &
1544 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1545 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1546 __le32 le_group = cpu_to_le32(block_group);
1548 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1549 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1550 crc = crc16(crc, (__u8 *)gdp, offset);
1551 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1552 /* for checksum of struct ext4_group_desc do the rest...*/
1553 if ((sbi->s_es->s_feature_incompat &
1554 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1555 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1556 crc = crc16(crc, (__u8 *)gdp + offset,
1557 le16_to_cpu(sbi->s_es->s_desc_size) -
1561 return cpu_to_le16(crc);
1564 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1565 struct ext4_group_desc *gdp)
1567 if ((sbi->s_es->s_feature_ro_compat &
1568 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1569 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1575 /* Called at mount-time, super-block is locked */
1576 static int ext4_check_descriptors(struct super_block *sb)
1578 struct ext4_sb_info *sbi = EXT4_SB(sb);
1579 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1580 ext4_fsblk_t last_block;
1581 ext4_fsblk_t block_bitmap;
1582 ext4_fsblk_t inode_bitmap;
1583 ext4_fsblk_t inode_table;
1584 int flexbg_flag = 0;
1587 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1590 ext4_debug("Checking group descriptors");
1592 for (i = 0; i < sbi->s_groups_count; i++) {
1593 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1595 if (i == sbi->s_groups_count - 1 || flexbg_flag)
1596 last_block = ext4_blocks_count(sbi->s_es) - 1;
1598 last_block = first_block +
1599 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1601 block_bitmap = ext4_block_bitmap(sb, gdp);
1602 if (block_bitmap < first_block || block_bitmap > last_block) {
1603 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1604 "Block bitmap for group %lu not in group "
1605 "(block %llu)!", i, block_bitmap);
1608 inode_bitmap = ext4_inode_bitmap(sb, gdp);
1609 if (inode_bitmap < first_block || inode_bitmap > last_block) {
1610 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1611 "Inode bitmap for group %lu not in group "
1612 "(block %llu)!", i, inode_bitmap);
1615 inode_table = ext4_inode_table(sb, gdp);
1616 if (inode_table < first_block ||
1617 inode_table + sbi->s_itb_per_group - 1 > last_block) {
1618 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1619 "Inode table for group %lu not in group "
1620 "(block %llu)!", i, inode_table);
1623 spin_lock(sb_bgl_lock(sbi, i));
1624 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1625 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1626 "Checksum for group %lu failed (%u!=%u)\n",
1627 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1628 gdp)), le16_to_cpu(gdp->bg_checksum));
1629 if (!(sb->s_flags & MS_RDONLY))
1632 spin_unlock(sb_bgl_lock(sbi, i));
1634 first_block += EXT4_BLOCKS_PER_GROUP(sb);
1637 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1638 sbi->s_es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
1642 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1643 * the superblock) which were deleted from all directories, but held open by
1644 * a process at the time of a crash. We walk the list and try to delete these
1645 * inodes at recovery time (only with a read-write filesystem).
1647 * In order to keep the orphan inode chain consistent during traversal (in
1648 * case of crash during recovery), we link each inode into the superblock
1649 * orphan list_head and handle it the same way as an inode deletion during
1650 * normal operation (which journals the operations for us).
1652 * We only do an iget() and an iput() on each inode, which is very safe if we
1653 * accidentally point at an in-use or already deleted inode. The worst that
1654 * can happen in this case is that we get a "bit already cleared" message from
1655 * ext4_free_inode(). The only reason we would point at a wrong inode is if
1656 * e2fsck was run on this filesystem, and it must have already done the orphan
1657 * inode cleanup for us, so we can safely abort without any further action.
1659 static void ext4_orphan_cleanup(struct super_block *sb,
1660 struct ext4_super_block *es)
1662 unsigned int s_flags = sb->s_flags;
1663 int nr_orphans = 0, nr_truncates = 0;
1667 if (!es->s_last_orphan) {
1668 jbd_debug(4, "no orphan inodes to clean up\n");
1672 if (bdev_read_only(sb->s_bdev)) {
1673 printk(KERN_ERR "EXT4-fs: write access "
1674 "unavailable, skipping orphan cleanup.\n");
1678 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1679 if (es->s_last_orphan)
1680 jbd_debug(1, "Errors on filesystem, "
1681 "clearing orphan list.\n");
1682 es->s_last_orphan = 0;
1683 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1687 if (s_flags & MS_RDONLY) {
1688 printk(KERN_INFO "EXT4-fs: %s: orphan cleanup on readonly fs\n",
1690 sb->s_flags &= ~MS_RDONLY;
1693 /* Needed for iput() to work correctly and not trash data */
1694 sb->s_flags |= MS_ACTIVE;
1695 /* Turn on quotas so that they are updated correctly */
1696 for (i = 0; i < MAXQUOTAS; i++) {
1697 if (EXT4_SB(sb)->s_qf_names[i]) {
1698 int ret = ext4_quota_on_mount(sb, i);
1701 "EXT4-fs: Cannot turn on journaled "
1702 "quota: error %d\n", ret);
1707 while (es->s_last_orphan) {
1708 struct inode *inode;
1710 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1711 if (IS_ERR(inode)) {
1712 es->s_last_orphan = 0;
1716 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
1718 if (inode->i_nlink) {
1720 "%s: truncating inode %lu to %lld bytes\n",
1721 __func__, inode->i_ino, inode->i_size);
1722 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
1723 inode->i_ino, inode->i_size);
1724 ext4_truncate(inode);
1728 "%s: deleting unreferenced inode %lu\n",
1729 __func__, inode->i_ino);
1730 jbd_debug(2, "deleting unreferenced inode %lu\n",
1734 iput(inode); /* The delete magic happens here! */
1737 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
1740 printk(KERN_INFO "EXT4-fs: %s: %d orphan inode%s deleted\n",
1741 sb->s_id, PLURAL(nr_orphans));
1743 printk(KERN_INFO "EXT4-fs: %s: %d truncate%s cleaned up\n",
1744 sb->s_id, PLURAL(nr_truncates));
1746 /* Turn quotas off */
1747 for (i = 0; i < MAXQUOTAS; i++) {
1748 if (sb_dqopt(sb)->files[i])
1749 vfs_quota_off(sb, i, 0);
1752 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1755 * Maximal extent format file size.
1756 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1757 * extent format containers, within a sector_t, and within i_blocks
1758 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1759 * so that won't be a limiting factor.
1761 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1763 static loff_t ext4_max_size(int blkbits)
1766 loff_t upper_limit = MAX_LFS_FILESIZE;
1768 /* small i_blocks in vfs inode? */
1769 if (sizeof(blkcnt_t) < sizeof(u64)) {
1771 * CONFIG_LSF is not enabled implies the inode
1772 * i_block represent total blocks in 512 bytes
1773 * 32 == size of vfs inode i_blocks * 8
1775 upper_limit = (1LL << 32) - 1;
1777 /* total blocks in file system block size */
1778 upper_limit >>= (blkbits - 9);
1779 upper_limit <<= blkbits;
1782 /* 32-bit extent-start container, ee_block */
1787 /* Sanity check against vm- & vfs- imposed limits */
1788 if (res > upper_limit)
1795 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
1796 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1797 * We need to be 1 filesystem block less than the 2^48 sector limit.
1799 static loff_t ext4_max_bitmap_size(int bits)
1801 loff_t res = EXT4_NDIR_BLOCKS;
1804 /* This is calculated to be the largest file size for a
1805 * dense, bitmapped file such that the total number of
1806 * sectors in the file, including data and all indirect blocks,
1807 * does not exceed 2^48 -1
1808 * __u32 i_blocks_lo and _u16 i_blocks_high representing the
1809 * total number of 512 bytes blocks of the file
1812 if (sizeof(blkcnt_t) < sizeof(u64)) {
1814 * CONFIG_LSF is not enabled implies the inode
1815 * i_block represent total blocks in 512 bytes
1816 * 32 == size of vfs inode i_blocks * 8
1818 upper_limit = (1LL << 32) - 1;
1820 /* total blocks in file system block size */
1821 upper_limit >>= (bits - 9);
1825 * We use 48 bit ext4_inode i_blocks
1826 * With EXT4_HUGE_FILE_FL set the i_blocks
1827 * represent total number of blocks in
1828 * file system block size
1830 upper_limit = (1LL << 48) - 1;
1834 /* indirect blocks */
1836 /* double indirect blocks */
1837 meta_blocks += 1 + (1LL << (bits-2));
1838 /* tripple indirect blocks */
1839 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1841 upper_limit -= meta_blocks;
1842 upper_limit <<= bits;
1844 res += 1LL << (bits-2);
1845 res += 1LL << (2*(bits-2));
1846 res += 1LL << (3*(bits-2));
1848 if (res > upper_limit)
1851 if (res > MAX_LFS_FILESIZE)
1852 res = MAX_LFS_FILESIZE;
1857 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
1858 ext4_fsblk_t logical_sb_block, int nr)
1860 struct ext4_sb_info *sbi = EXT4_SB(sb);
1861 ext4_group_t bg, first_meta_bg;
1864 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1866 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
1868 return logical_sb_block + nr + 1;
1869 bg = sbi->s_desc_per_block * nr;
1870 if (ext4_bg_has_super(sb, bg))
1872 return (has_super + ext4_group_first_block_no(sb, bg));
1876 * ext4_get_stripe_size: Get the stripe size.
1877 * @sbi: In memory super block info
1879 * If we have specified it via mount option, then
1880 * use the mount option value. If the value specified at mount time is
1881 * greater than the blocks per group use the super block value.
1882 * If the super block value is greater than blocks per group return 0.
1883 * Allocator needs it be less than blocks per group.
1886 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
1888 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
1889 unsigned long stripe_width =
1890 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
1892 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
1893 return sbi->s_stripe;
1895 if (stripe_width <= sbi->s_blocks_per_group)
1896 return stripe_width;
1898 if (stride <= sbi->s_blocks_per_group)
1904 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
1905 __releases(kernel_lock)
1906 __acquires(kernel_lock)
1909 struct buffer_head *bh;
1910 struct ext4_super_block *es = NULL;
1911 struct ext4_sb_info *sbi;
1913 ext4_fsblk_t sb_block = get_sb_block(&data);
1914 ext4_fsblk_t logical_sb_block;
1915 unsigned long offset = 0;
1916 unsigned int journal_inum = 0;
1917 unsigned long journal_devnum = 0;
1918 unsigned long def_mount_opts;
1929 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1932 sb->s_fs_info = sbi;
1933 sbi->s_mount_opt = 0;
1934 sbi->s_resuid = EXT4_DEF_RESUID;
1935 sbi->s_resgid = EXT4_DEF_RESGID;
1936 sbi->s_sb_block = sb_block;
1940 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
1942 printk(KERN_ERR "EXT4-fs: unable to set blocksize\n");
1947 * The ext4 superblock will not be buffer aligned for other than 1kB
1948 * block sizes. We need to calculate the offset from buffer start.
1950 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
1951 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
1952 offset = do_div(logical_sb_block, blocksize);
1954 logical_sb_block = sb_block;
1957 if (!(bh = sb_bread(sb, logical_sb_block))) {
1958 printk(KERN_ERR "EXT4-fs: unable to read superblock\n");
1962 * Note: s_es must be initialized as soon as possible because
1963 * some ext4 macro-instructions depend on its value
1965 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
1967 sb->s_magic = le16_to_cpu(es->s_magic);
1968 if (sb->s_magic != EXT4_SUPER_MAGIC)
1971 /* Set defaults before we parse the mount options */
1972 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1973 if (def_mount_opts & EXT4_DEFM_DEBUG)
1974 set_opt(sbi->s_mount_opt, DEBUG);
1975 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
1976 set_opt(sbi->s_mount_opt, GRPID);
1977 if (def_mount_opts & EXT4_DEFM_UID16)
1978 set_opt(sbi->s_mount_opt, NO_UID32);
1979 #ifdef CONFIG_EXT4DEV_FS_XATTR
1980 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
1981 set_opt(sbi->s_mount_opt, XATTR_USER);
1983 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
1984 if (def_mount_opts & EXT4_DEFM_ACL)
1985 set_opt(sbi->s_mount_opt, POSIX_ACL);
1987 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
1988 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
1989 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
1990 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
1991 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
1992 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
1994 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
1995 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1996 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
1997 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1999 set_opt(sbi->s_mount_opt, ERRORS_RO);
2001 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2002 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
2004 set_opt(sbi->s_mount_opt, RESERVATION);
2005 set_opt(sbi->s_mount_opt, BARRIER);
2008 * turn on extents feature by default in ext4 filesystem
2009 * only if feature flag already set by mkfs or tune2fs.
2010 * Use -o noextents to turn it off
2012 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS))
2013 set_opt(sbi->s_mount_opt, EXTENTS);
2015 ext4_warning(sb, __func__,
2016 "extents feature not enabled on this filesystem, "
2019 * turn on mballoc code by default in ext4 filesystem
2020 * Use -o nomballoc to turn it off
2022 set_opt(sbi->s_mount_opt, MBALLOC);
2025 * enable delayed allocation by default
2026 * Use -o nodelalloc to turn it off
2028 set_opt(sbi->s_mount_opt, DELALLOC);
2031 if (!parse_options((char *) data, sb, &journal_inum, &journal_devnum,
2035 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2036 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2038 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2039 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2040 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2041 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
2043 "EXT4-fs warning: feature flags set on rev 0 fs, "
2044 "running e2fsck is recommended\n");
2047 * Since ext4 is still considered development code, we require
2048 * that the TEST_FILESYS flag in s->flags be set.
2050 if (!(le32_to_cpu(es->s_flags) & EXT2_FLAGS_TEST_FILESYS)) {
2051 printk(KERN_WARNING "EXT4-fs: %s: not marked "
2052 "OK to use with test code.\n", sb->s_id);
2057 * Check feature flags regardless of the revision level, since we
2058 * previously didn't change the revision level when setting the flags,
2059 * so there is a chance incompat flags are set on a rev 0 filesystem.
2061 features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
2063 printk(KERN_ERR "EXT4-fs: %s: couldn't mount because of "
2064 "unsupported optional features (%x).\n",
2065 sb->s_id, le32_to_cpu(features));
2068 features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
2069 if (!(sb->s_flags & MS_RDONLY) && features) {
2070 printk(KERN_ERR "EXT4-fs: %s: couldn't mount RDWR because of "
2071 "unsupported optional features (%x).\n",
2072 sb->s_id, le32_to_cpu(features));
2075 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2077 * Large file size enabled file system can only be
2078 * mount if kernel is build with CONFIG_LSF
2080 if (sizeof(root->i_blocks) < sizeof(u64) &&
2081 !(sb->s_flags & MS_RDONLY)) {
2082 printk(KERN_ERR "EXT4-fs: %s: Filesystem with huge "
2083 "files cannot be mounted read-write "
2084 "without CONFIG_LSF.\n", sb->s_id);
2088 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2090 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2091 blocksize > EXT4_MAX_BLOCK_SIZE) {
2093 "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
2094 blocksize, sb->s_id);
2098 if (sb->s_blocksize != blocksize) {
2100 /* Validate the filesystem blocksize */
2101 if (!sb_set_blocksize(sb, blocksize)) {
2102 printk(KERN_ERR "EXT4-fs: bad block size %d.\n",
2108 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2109 offset = do_div(logical_sb_block, blocksize);
2110 bh = sb_bread(sb, logical_sb_block);
2113 "EXT4-fs: Can't read superblock on 2nd try.\n");
2116 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
2118 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2120 "EXT4-fs: Magic mismatch, very weird !\n");
2125 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits);
2126 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits);
2128 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2129 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2130 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
2132 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2133 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
2134 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
2135 (!is_power_of_2(sbi->s_inode_size)) ||
2136 (sbi->s_inode_size > blocksize)) {
2138 "EXT4-fs: unsupported inode size: %d\n",
2142 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2143 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
2145 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2146 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
2147 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
2148 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
2149 !is_power_of_2(sbi->s_desc_size)) {
2151 "EXT4-fs: unsupported descriptor size %lu\n",
2156 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
2157 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
2158 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
2159 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
2161 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
2162 if (sbi->s_inodes_per_block == 0)
2164 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2165 sbi->s_inodes_per_block;
2166 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
2168 sbi->s_mount_state = le16_to_cpu(es->s_state);
2169 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2170 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2171 for (i = 0; i < 4; i++)
2172 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2173 sbi->s_def_hash_version = es->s_def_hash_version;
2175 if (sbi->s_blocks_per_group > blocksize * 8) {
2177 "EXT4-fs: #blocks per group too big: %lu\n",
2178 sbi->s_blocks_per_group);
2181 if (sbi->s_inodes_per_group > blocksize * 8) {
2183 "EXT4-fs: #inodes per group too big: %lu\n",
2184 sbi->s_inodes_per_group);
2188 if (ext4_blocks_count(es) >
2189 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
2190 printk(KERN_ERR "EXT4-fs: filesystem on %s:"
2191 " too large to mount safely\n", sb->s_id);
2192 if (sizeof(sector_t) < 8)
2193 printk(KERN_WARNING "EXT4-fs: CONFIG_LBD not "
2198 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2201 /* ensure blocks_count calculation below doesn't sign-extend */
2202 if (ext4_blocks_count(es) + EXT4_BLOCKS_PER_GROUP(sb) <
2203 le32_to_cpu(es->s_first_data_block) + 1) {
2204 printk(KERN_WARNING "EXT4-fs: bad geometry: block count %llu, "
2205 "first data block %u, blocks per group %lu\n",
2206 ext4_blocks_count(es),
2207 le32_to_cpu(es->s_first_data_block),
2208 EXT4_BLOCKS_PER_GROUP(sb));
2211 blocks_count = (ext4_blocks_count(es) -
2212 le32_to_cpu(es->s_first_data_block) +
2213 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2214 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2215 sbi->s_groups_count = blocks_count;
2216 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2217 EXT4_DESC_PER_BLOCK(sb);
2218 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
2220 if (sbi->s_group_desc == NULL) {
2221 printk(KERN_ERR "EXT4-fs: not enough memory\n");
2225 bgl_lock_init(&sbi->s_blockgroup_lock);
2227 for (i = 0; i < db_count; i++) {
2228 block = descriptor_loc(sb, logical_sb_block, i);
2229 sbi->s_group_desc[i] = sb_bread(sb, block);
2230 if (!sbi->s_group_desc[i]) {
2231 printk(KERN_ERR "EXT4-fs: "
2232 "can't read group descriptor %d\n", i);
2237 if (!ext4_check_descriptors(sb)) {
2238 printk(KERN_ERR "EXT4-fs: group descriptors corrupted!\n");
2241 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2242 if (!ext4_fill_flex_info(sb)) {
2244 "EXT4-fs: unable to initialize "
2245 "flex_bg meta info!\n");
2249 sbi->s_gdb_count = db_count;
2250 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2251 spin_lock_init(&sbi->s_next_gen_lock);
2253 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2254 ext4_count_free_blocks(sb));
2256 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2257 ext4_count_free_inodes(sb));
2260 err = percpu_counter_init(&sbi->s_dirs_counter,
2261 ext4_count_dirs(sb));
2264 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2267 printk(KERN_ERR "EXT4-fs: insufficient memory\n");
2271 /* per fileystem reservation list head & lock */
2272 spin_lock_init(&sbi->s_rsv_window_lock);
2273 sbi->s_rsv_window_root = RB_ROOT;
2274 /* Add a single, static dummy reservation to the start of the
2275 * reservation window list --- it gives us a placeholder for
2276 * append-at-start-of-list which makes the allocation logic
2277 * _much_ simpler. */
2278 sbi->s_rsv_window_head.rsv_start = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
2279 sbi->s_rsv_window_head.rsv_end = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
2280 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
2281 sbi->s_rsv_window_head.rsv_goal_size = 0;
2282 ext4_rsv_window_add(sb, &sbi->s_rsv_window_head);
2284 sbi->s_stripe = ext4_get_stripe_size(sbi);
2287 * set up enough so that it can read an inode
2289 sb->s_op = &ext4_sops;
2290 sb->s_export_op = &ext4_export_ops;
2291 sb->s_xattr = ext4_xattr_handlers;
2293 sb->s_qcop = &ext4_qctl_operations;
2294 sb->dq_op = &ext4_quota_operations;
2296 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2300 needs_recovery = (es->s_last_orphan != 0 ||
2301 EXT4_HAS_INCOMPAT_FEATURE(sb,
2302 EXT4_FEATURE_INCOMPAT_RECOVER));
2305 * The first inode we look at is the journal inode. Don't try
2306 * root first: it may be modified in the journal!
2308 if (!test_opt(sb, NOLOAD) &&
2309 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2310 if (ext4_load_journal(sb, es, journal_devnum))
2312 if (!(sb->s_flags & MS_RDONLY) &&
2313 EXT4_SB(sb)->s_journal->j_failed_commit) {
2314 printk(KERN_CRIT "EXT4-fs error (device %s): "
2315 "ext4_fill_super: Journal transaction "
2316 "%u is corrupt\n", sb->s_id,
2317 EXT4_SB(sb)->s_journal->j_failed_commit);
2318 if (test_opt(sb, ERRORS_RO)) {
2320 "Mounting filesystem read-only\n");
2321 sb->s_flags |= MS_RDONLY;
2322 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2323 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2325 if (test_opt(sb, ERRORS_PANIC)) {
2326 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2327 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2328 ext4_commit_super(sb, es, 1);
2330 "EXT4-fs (device %s): mount failed\n",
2335 } else if (journal_inum) {
2336 if (ext4_create_journal(sb, es, journal_inum))
2341 "ext4: No journal on filesystem on %s\n",
2346 if (ext4_blocks_count(es) > 0xffffffffULL &&
2347 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2348 JBD2_FEATURE_INCOMPAT_64BIT)) {
2349 printk(KERN_ERR "ext4: Failed to set 64-bit journal feature\n");
2353 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2354 jbd2_journal_set_features(sbi->s_journal,
2355 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2356 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2357 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2358 jbd2_journal_set_features(sbi->s_journal,
2359 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2360 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2361 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2363 jbd2_journal_clear_features(sbi->s_journal,
2364 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2365 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2368 /* We have now updated the journal if required, so we can
2369 * validate the data journaling mode. */
2370 switch (test_opt(sb, DATA_FLAGS)) {
2372 /* No mode set, assume a default based on the journal
2373 * capabilities: ORDERED_DATA if the journal can
2374 * cope, else JOURNAL_DATA
2376 if (jbd2_journal_check_available_features
2377 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
2378 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2380 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2383 case EXT4_MOUNT_ORDERED_DATA:
2384 case EXT4_MOUNT_WRITEBACK_DATA:
2385 if (!jbd2_journal_check_available_features
2386 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
2387 printk(KERN_ERR "EXT4-fs: Journal does not support "
2388 "requested data journaling mode\n");
2395 if (test_opt(sb, NOBH)) {
2396 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2397 printk(KERN_WARNING "EXT4-fs: Ignoring nobh option - "
2398 "its supported only with writeback mode\n");
2399 clear_opt(sbi->s_mount_opt, NOBH);
2403 * The jbd2_journal_load will have done any necessary log recovery,
2404 * so we can safely mount the rest of the filesystem now.
2407 root = ext4_iget(sb, EXT4_ROOT_INO);
2409 printk(KERN_ERR "EXT4-fs: get root inode failed\n");
2410 ret = PTR_ERR(root);
2413 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2415 printk(KERN_ERR "EXT4-fs: corrupt root inode, run e2fsck\n");
2418 sb->s_root = d_alloc_root(root);
2420 printk(KERN_ERR "EXT4-fs: get root dentry failed\n");
2426 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
2428 /* determine the minimum size of new large inodes, if present */
2429 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2430 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2431 EXT4_GOOD_OLD_INODE_SIZE;
2432 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2433 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2434 if (sbi->s_want_extra_isize <
2435 le16_to_cpu(es->s_want_extra_isize))
2436 sbi->s_want_extra_isize =
2437 le16_to_cpu(es->s_want_extra_isize);
2438 if (sbi->s_want_extra_isize <
2439 le16_to_cpu(es->s_min_extra_isize))
2440 sbi->s_want_extra_isize =
2441 le16_to_cpu(es->s_min_extra_isize);
2444 /* Check if enough inode space is available */
2445 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2446 sbi->s_inode_size) {
2447 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2448 EXT4_GOOD_OLD_INODE_SIZE;
2449 printk(KERN_INFO "EXT4-fs: required extra inode space not"
2454 * akpm: core read_super() calls in here with the superblock locked.
2455 * That deadlocks, because orphan cleanup needs to lock the superblock
2456 * in numerous places. Here we just pop the lock - it's relatively
2457 * harmless, because we are now ready to accept write_super() requests,
2458 * and aviro says that's the only reason for hanging onto the
2461 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2462 ext4_orphan_cleanup(sb, es);
2463 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
2465 printk(KERN_INFO "EXT4-fs: recovery complete.\n");
2466 ext4_mark_recovery_complete(sb, es);
2467 printk(KERN_INFO "EXT4-fs: mounted filesystem with %s data mode.\n",
2468 test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA ? "journal":
2469 test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA ? "ordered":
2472 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
2473 printk(KERN_WARNING "EXT4-fs: Ignoring delalloc option - "
2474 "requested data journaling mode\n");
2475 clear_opt(sbi->s_mount_opt, DELALLOC);
2476 } else if (test_opt(sb, DELALLOC))
2477 printk(KERN_INFO "EXT4-fs: delayed allocation enabled\n");
2480 ext4_mb_init(sb, needs_recovery);
2487 printk(KERN_ERR "VFS: Can't find ext4 filesystem on dev %s.\n",
2492 jbd2_journal_destroy(sbi->s_journal);
2493 sbi->s_journal = NULL;
2495 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2496 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2497 percpu_counter_destroy(&sbi->s_dirs_counter);
2498 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
2500 for (i = 0; i < db_count; i++)
2501 brelse(sbi->s_group_desc[i]);
2502 kfree(sbi->s_group_desc);
2505 for (i = 0; i < MAXQUOTAS; i++)
2506 kfree(sbi->s_qf_names[i]);
2508 ext4_blkdev_remove(sbi);
2511 sb->s_fs_info = NULL;
2518 * Setup any per-fs journal parameters now. We'll do this both on
2519 * initial mount, once the journal has been initialised but before we've
2520 * done any recovery; and again on any subsequent remount.
2522 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
2524 struct ext4_sb_info *sbi = EXT4_SB(sb);
2526 if (sbi->s_commit_interval)
2527 journal->j_commit_interval = sbi->s_commit_interval;
2528 /* We could also set up an ext4-specific default for the commit
2529 * interval here, but for now we'll just fall back to the jbd
2532 spin_lock(&journal->j_state_lock);
2533 if (test_opt(sb, BARRIER))
2534 journal->j_flags |= JBD2_BARRIER;
2536 journal->j_flags &= ~JBD2_BARRIER;
2537 spin_unlock(&journal->j_state_lock);
2540 static journal_t *ext4_get_journal(struct super_block *sb,
2541 unsigned int journal_inum)
2543 struct inode *journal_inode;
2546 /* First, test for the existence of a valid inode on disk. Bad
2547 * things happen if we iget() an unused inode, as the subsequent
2548 * iput() will try to delete it. */
2550 journal_inode = ext4_iget(sb, journal_inum);
2551 if (IS_ERR(journal_inode)) {
2552 printk(KERN_ERR "EXT4-fs: no journal found.\n");
2555 if (!journal_inode->i_nlink) {
2556 make_bad_inode(journal_inode);
2557 iput(journal_inode);
2558 printk(KERN_ERR "EXT4-fs: journal inode is deleted.\n");
2562 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
2563 journal_inode, journal_inode->i_size);
2564 if (!S_ISREG(journal_inode->i_mode)) {
2565 printk(KERN_ERR "EXT4-fs: invalid journal inode.\n");
2566 iput(journal_inode);
2570 journal = jbd2_journal_init_inode(journal_inode);
2572 printk(KERN_ERR "EXT4-fs: Could not load journal inode\n");
2573 iput(journal_inode);
2576 journal->j_private = sb;
2577 ext4_init_journal_params(sb, journal);
2581 static journal_t *ext4_get_dev_journal(struct super_block *sb,
2584 struct buffer_head *bh;
2588 int hblock, blocksize;
2589 ext4_fsblk_t sb_block;
2590 unsigned long offset;
2591 struct ext4_super_block *es;
2592 struct block_device *bdev;
2594 bdev = ext4_blkdev_get(j_dev);
2598 if (bd_claim(bdev, sb)) {
2600 "EXT4: failed to claim external journal device.\n");
2605 blocksize = sb->s_blocksize;
2606 hblock = bdev_hardsect_size(bdev);
2607 if (blocksize < hblock) {
2609 "EXT4-fs: blocksize too small for journal device.\n");
2613 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
2614 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
2615 set_blocksize(bdev, blocksize);
2616 if (!(bh = __bread(bdev, sb_block, blocksize))) {
2617 printk(KERN_ERR "EXT4-fs: couldn't read superblock of "
2618 "external journal\n");
2622 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2623 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
2624 !(le32_to_cpu(es->s_feature_incompat) &
2625 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2626 printk(KERN_ERR "EXT4-fs: external journal has "
2627 "bad superblock\n");
2632 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2633 printk(KERN_ERR "EXT4-fs: journal UUID does not match\n");
2638 len = ext4_blocks_count(es);
2639 start = sb_block + 1;
2640 brelse(bh); /* we're done with the superblock */
2642 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
2643 start, len, blocksize);
2645 printk(KERN_ERR "EXT4-fs: failed to create device journal\n");
2648 journal->j_private = sb;
2649 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2650 wait_on_buffer(journal->j_sb_buffer);
2651 if (!buffer_uptodate(journal->j_sb_buffer)) {
2652 printk(KERN_ERR "EXT4-fs: I/O error on journal device\n");
2655 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
2656 printk(KERN_ERR "EXT4-fs: External journal has more than one "
2657 "user (unsupported) - %d\n",
2658 be32_to_cpu(journal->j_superblock->s_nr_users));
2661 EXT4_SB(sb)->journal_bdev = bdev;
2662 ext4_init_journal_params(sb, journal);
2665 jbd2_journal_destroy(journal);
2667 ext4_blkdev_put(bdev);
2671 static int ext4_load_journal(struct super_block *sb,
2672 struct ext4_super_block *es,
2673 unsigned long journal_devnum)
2676 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2679 int really_read_only;
2681 if (journal_devnum &&
2682 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2683 printk(KERN_INFO "EXT4-fs: external journal device major/minor "
2684 "numbers have changed\n");
2685 journal_dev = new_decode_dev(journal_devnum);
2687 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2689 really_read_only = bdev_read_only(sb->s_bdev);
2692 * Are we loading a blank journal or performing recovery after a
2693 * crash? For recovery, we need to check in advance whether we
2694 * can get read-write access to the device.
2697 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2698 if (sb->s_flags & MS_RDONLY) {
2699 printk(KERN_INFO "EXT4-fs: INFO: recovery "
2700 "required on readonly filesystem.\n");
2701 if (really_read_only) {
2702 printk(KERN_ERR "EXT4-fs: write access "
2703 "unavailable, cannot proceed.\n");
2706 printk(KERN_INFO "EXT4-fs: write access will "
2707 "be enabled during recovery.\n");
2711 if (journal_inum && journal_dev) {
2712 printk(KERN_ERR "EXT4-fs: filesystem has both journal "
2713 "and inode journals!\n");
2718 if (!(journal = ext4_get_journal(sb, journal_inum)))
2721 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
2725 if (journal->j_flags & JBD2_BARRIER)
2726 printk(KERN_INFO "EXT4-fs: barriers enabled\n");
2728 printk(KERN_INFO "EXT4-fs: barriers disabled\n");
2730 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2731 err = jbd2_journal_update_format(journal);
2733 printk(KERN_ERR "EXT4-fs: error updating journal.\n");
2734 jbd2_journal_destroy(journal);
2739 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
2740 err = jbd2_journal_wipe(journal, !really_read_only);
2742 err = jbd2_journal_load(journal);
2745 printk(KERN_ERR "EXT4-fs: error loading journal.\n");
2746 jbd2_journal_destroy(journal);
2750 EXT4_SB(sb)->s_journal = journal;
2751 ext4_clear_journal_err(sb, es);
2753 if (journal_devnum &&
2754 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2755 es->s_journal_dev = cpu_to_le32(journal_devnum);
2758 /* Make sure we flush the recovery flag to disk. */
2759 ext4_commit_super(sb, es, 1);
2765 static int ext4_create_journal(struct super_block *sb,
2766 struct ext4_super_block *es,
2767 unsigned int journal_inum)
2772 if (sb->s_flags & MS_RDONLY) {
2773 printk(KERN_ERR "EXT4-fs: readonly filesystem when trying to "
2774 "create journal.\n");
2778 journal = ext4_get_journal(sb, journal_inum);
2782 printk(KERN_INFO "EXT4-fs: creating new journal on inode %u\n",
2785 err = jbd2_journal_create(journal);
2787 printk(KERN_ERR "EXT4-fs: error creating journal.\n");
2788 jbd2_journal_destroy(journal);
2792 EXT4_SB(sb)->s_journal = journal;
2794 ext4_update_dynamic_rev(sb);
2795 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2796 EXT4_SET_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL);
2798 es->s_journal_inum = cpu_to_le32(journal_inum);
2801 /* Make sure we flush the recovery flag to disk. */
2802 ext4_commit_super(sb, es, 1);
2807 static void ext4_commit_super(struct super_block *sb,
2808 struct ext4_super_block *es, int sync)
2810 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
2814 es->s_wtime = cpu_to_le32(get_seconds());
2815 ext4_free_blocks_count_set(es, ext4_count_free_blocks(sb));
2816 es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
2817 BUFFER_TRACE(sbh, "marking dirty");
2818 mark_buffer_dirty(sbh);
2820 sync_dirty_buffer(sbh);
2825 * Have we just finished recovery? If so, and if we are mounting (or
2826 * remounting) the filesystem readonly, then we will end up with a
2827 * consistent fs on disk. Record that fact.
2829 static void ext4_mark_recovery_complete(struct super_block *sb,
2830 struct ext4_super_block *es)
2832 journal_t *journal = EXT4_SB(sb)->s_journal;
2834 jbd2_journal_lock_updates(journal);
2835 jbd2_journal_flush(journal);
2837 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
2838 sb->s_flags & MS_RDONLY) {
2839 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2841 ext4_commit_super(sb, es, 1);
2844 jbd2_journal_unlock_updates(journal);
2848 * If we are mounting (or read-write remounting) a filesystem whose journal
2849 * has recorded an error from a previous lifetime, move that error to the
2850 * main filesystem now.
2852 static void ext4_clear_journal_err(struct super_block *sb,
2853 struct ext4_super_block *es)
2859 journal = EXT4_SB(sb)->s_journal;
2862 * Now check for any error status which may have been recorded in the
2863 * journal by a prior ext4_error() or ext4_abort()
2866 j_errno = jbd2_journal_errno(journal);
2870 errstr = ext4_decode_error(sb, j_errno, nbuf);
2871 ext4_warning(sb, __func__, "Filesystem error recorded "
2872 "from previous mount: %s", errstr);
2873 ext4_warning(sb, __func__, "Marking fs in need of "
2874 "filesystem check.");
2876 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2877 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2878 ext4_commit_super(sb, es, 1);
2880 jbd2_journal_clear_err(journal);
2885 * Force the running and committing transactions to commit,
2886 * and wait on the commit.
2888 int ext4_force_commit(struct super_block *sb)
2893 if (sb->s_flags & MS_RDONLY)
2896 journal = EXT4_SB(sb)->s_journal;
2898 ret = ext4_journal_force_commit(journal);
2903 * Ext4 always journals updates to the superblock itself, so we don't
2904 * have to propagate any other updates to the superblock on disk at this
2905 * point. Just start an async writeback to get the buffers on their way
2908 * This implicitly triggers the writebehind on sync().
2911 static void ext4_write_super(struct super_block *sb)
2913 if (mutex_trylock(&sb->s_lock) != 0)
2918 static int ext4_sync_fs(struct super_block *sb, int wait)
2923 if (jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, &target)) {
2925 jbd2_log_wait_commit(EXT4_SB(sb)->s_journal, target);
2931 * LVM calls this function before a (read-only) snapshot is created. This
2932 * gives us a chance to flush the journal completely and mark the fs clean.
2934 static void ext4_write_super_lockfs(struct super_block *sb)
2938 if (!(sb->s_flags & MS_RDONLY)) {
2939 journal_t *journal = EXT4_SB(sb)->s_journal;
2941 /* Now we set up the journal barrier. */
2942 jbd2_journal_lock_updates(journal);
2943 jbd2_journal_flush(journal);
2945 /* Journal blocked and flushed, clear needs_recovery flag. */
2946 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2947 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
2952 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2953 * flag here, even though the filesystem is not technically dirty yet.
2955 static void ext4_unlockfs(struct super_block *sb)
2957 if (!(sb->s_flags & MS_RDONLY)) {
2959 /* Reser the needs_recovery flag before the fs is unlocked. */
2960 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2961 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
2963 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
2967 static int ext4_remount(struct super_block *sb, int *flags, char *data)
2969 struct ext4_super_block *es;
2970 struct ext4_sb_info *sbi = EXT4_SB(sb);
2971 ext4_fsblk_t n_blocks_count = 0;
2972 unsigned long old_sb_flags;
2973 struct ext4_mount_options old_opts;
2980 /* Store the original options */
2981 old_sb_flags = sb->s_flags;
2982 old_opts.s_mount_opt = sbi->s_mount_opt;
2983 old_opts.s_resuid = sbi->s_resuid;
2984 old_opts.s_resgid = sbi->s_resgid;
2985 old_opts.s_commit_interval = sbi->s_commit_interval;
2987 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2988 for (i = 0; i < MAXQUOTAS; i++)
2989 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2993 * Allow the "check" option to be passed as a remount option.
2995 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
3000 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT)
3001 ext4_abort(sb, __func__, "Abort forced by user");
3003 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3004 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
3008 ext4_init_journal_params(sb, sbi->s_journal);
3010 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
3011 n_blocks_count > ext4_blocks_count(es)) {
3012 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) {
3017 if (*flags & MS_RDONLY) {
3019 * First of all, the unconditional stuff we have to do
3020 * to disable replay of the journal when we next remount
3022 sb->s_flags |= MS_RDONLY;
3025 * OK, test if we are remounting a valid rw partition
3026 * readonly, and if so set the rdonly flag and then
3027 * mark the partition as valid again.
3029 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3030 (sbi->s_mount_state & EXT4_VALID_FS))
3031 es->s_state = cpu_to_le16(sbi->s_mount_state);
3034 * We have to unlock super so that we can wait for
3038 ext4_mark_recovery_complete(sb, es);
3042 if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3043 ~EXT4_FEATURE_RO_COMPAT_SUPP))) {
3044 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
3045 "remount RDWR because of unsupported "
3046 "optional features (%x).\n",
3047 sb->s_id, le32_to_cpu(ret));
3053 * Make sure the group descriptor checksums
3054 * are sane. If they aren't, refuse to
3057 for (g = 0; g < sbi->s_groups_count; g++) {
3058 struct ext4_group_desc *gdp =
3059 ext4_get_group_desc(sb, g, NULL);
3061 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3063 "EXT4-fs: ext4_remount: "
3064 "Checksum for group %lu failed (%u!=%u)\n",
3065 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3066 le16_to_cpu(gdp->bg_checksum));
3073 * If we have an unprocessed orphan list hanging
3074 * around from a previously readonly bdev mount,
3075 * require a full umount/remount for now.
3077 if (es->s_last_orphan) {
3078 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
3079 "remount RDWR because of unprocessed "
3080 "orphan inode list. Please "
3081 "umount/remount instead.\n",
3088 * Mounting a RDONLY partition read-write, so reread
3089 * and store the current valid flag. (It may have
3090 * been changed by e2fsck since we originally mounted
3093 ext4_clear_journal_err(sb, es);
3094 sbi->s_mount_state = le16_to_cpu(es->s_state);
3095 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
3097 if (!ext4_setup_super(sb, es, 0))
3098 sb->s_flags &= ~MS_RDONLY;
3102 /* Release old quota file names */
3103 for (i = 0; i < MAXQUOTAS; i++)
3104 if (old_opts.s_qf_names[i] &&
3105 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3106 kfree(old_opts.s_qf_names[i]);
3110 sb->s_flags = old_sb_flags;
3111 sbi->s_mount_opt = old_opts.s_mount_opt;
3112 sbi->s_resuid = old_opts.s_resuid;
3113 sbi->s_resgid = old_opts.s_resgid;
3114 sbi->s_commit_interval = old_opts.s_commit_interval;
3116 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3117 for (i = 0; i < MAXQUOTAS; i++) {
3118 if (sbi->s_qf_names[i] &&
3119 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3120 kfree(sbi->s_qf_names[i]);
3121 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3127 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
3129 struct super_block *sb = dentry->d_sb;
3130 struct ext4_sb_info *sbi = EXT4_SB(sb);
3131 struct ext4_super_block *es = sbi->s_es;
3134 if (test_opt(sb, MINIX_DF)) {
3135 sbi->s_overhead_last = 0;
3136 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
3137 ext4_group_t ngroups = sbi->s_groups_count, i;
3138 ext4_fsblk_t overhead = 0;
3142 * Compute the overhead (FS structures). This is constant
3143 * for a given filesystem unless the number of block groups
3144 * changes so we cache the previous value until it does.
3148 * All of the blocks before first_data_block are
3151 overhead = le32_to_cpu(es->s_first_data_block);
3154 * Add the overhead attributed to the superblock and
3155 * block group descriptors. If the sparse superblocks
3156 * feature is turned on, then not all groups have this.
3158 for (i = 0; i < ngroups; i++) {
3159 overhead += ext4_bg_has_super(sb, i) +
3160 ext4_bg_num_gdb(sb, i);
3165 * Every block group has an inode bitmap, a block
3166 * bitmap, and an inode table.
3168 overhead += ngroups * (2 + sbi->s_itb_per_group);
3169 sbi->s_overhead_last = overhead;
3171 sbi->s_blocks_last = ext4_blocks_count(es);
3174 buf->f_type = EXT4_SUPER_MAGIC;
3175 buf->f_bsize = sb->s_blocksize;
3176 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
3177 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3178 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
3179 ext4_free_blocks_count_set(es, buf->f_bfree);
3180 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3181 if (buf->f_bfree < ext4_r_blocks_count(es))
3183 buf->f_files = le32_to_cpu(es->s_inodes_count);
3184 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
3185 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
3186 buf->f_namelen = EXT4_NAME_LEN;
3187 fsid = le64_to_cpup((void *)es->s_uuid) ^
3188 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3189 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3190 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
3194 /* Helper function for writing quotas on sync - we need to start transaction before quota file
3195 * is locked for write. Otherwise the are possible deadlocks:
3196 * Process 1 Process 2
3197 * ext4_create() quota_sync()
3198 * jbd2_journal_start() write_dquot()
3199 * DQUOT_INIT() down(dqio_mutex)
3200 * down(dqio_mutex) jbd2_journal_start()
3206 static inline struct inode *dquot_to_inode(struct dquot *dquot)
3208 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3211 static int ext4_dquot_initialize(struct inode *inode, int type)
3216 /* We may create quota structure so we need to reserve enough blocks */
3217 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_INIT_BLOCKS(inode->i_sb));
3219 return PTR_ERR(handle);
3220 ret = dquot_initialize(inode, type);
3221 err = ext4_journal_stop(handle);
3227 static int ext4_dquot_drop(struct inode *inode)
3232 /* We may delete quota structure so we need to reserve enough blocks */
3233 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_DEL_BLOCKS(inode->i_sb));
3234 if (IS_ERR(handle)) {
3236 * We call dquot_drop() anyway to at least release references
3237 * to quota structures so that umount does not hang.
3240 return PTR_ERR(handle);
3242 ret = dquot_drop(inode);
3243 err = ext4_journal_stop(handle);
3249 static int ext4_write_dquot(struct dquot *dquot)
3253 struct inode *inode;
3255 inode = dquot_to_inode(dquot);
3256 handle = ext4_journal_start(inode,
3257 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
3259 return PTR_ERR(handle);
3260 ret = dquot_commit(dquot);
3261 err = ext4_journal_stop(handle);
3267 static int ext4_acquire_dquot(struct dquot *dquot)
3272 handle = ext4_journal_start(dquot_to_inode(dquot),
3273 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
3275 return PTR_ERR(handle);
3276 ret = dquot_acquire(dquot);
3277 err = ext4_journal_stop(handle);
3283 static int ext4_release_dquot(struct dquot *dquot)
3288 handle = ext4_journal_start(dquot_to_inode(dquot),
3289 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
3290 if (IS_ERR(handle)) {
3291 /* Release dquot anyway to avoid endless cycle in dqput() */
3292 dquot_release(dquot);
3293 return PTR_ERR(handle);
3295 ret = dquot_release(dquot);
3296 err = ext4_journal_stop(handle);
3302 static int ext4_mark_dquot_dirty(struct dquot *dquot)
3304 /* Are we journaling quotas? */
3305 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3306 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
3307 dquot_mark_dquot_dirty(dquot);
3308 return ext4_write_dquot(dquot);
3310 return dquot_mark_dquot_dirty(dquot);
3314 static int ext4_write_info(struct super_block *sb, int type)
3319 /* Data block + inode block */
3320 handle = ext4_journal_start(sb->s_root->d_inode, 2);
3322 return PTR_ERR(handle);
3323 ret = dquot_commit_info(sb, type);
3324 err = ext4_journal_stop(handle);
3331 * Turn on quotas during mount time - we need to find
3332 * the quota file and such...
3334 static int ext4_quota_on_mount(struct super_block *sb, int type)
3336 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3337 EXT4_SB(sb)->s_jquota_fmt, type);
3341 * Standard function to be called on quota_on
3343 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
3344 char *path, int remount)
3347 struct nameidata nd;
3349 if (!test_opt(sb, QUOTA))
3351 /* When remounting, no checks are needed and in fact, path is NULL */
3353 return vfs_quota_on(sb, type, format_id, path, remount);
3355 err = path_lookup(path, LOOKUP_FOLLOW, &nd);
3359 /* Quotafile not on the same filesystem? */
3360 if (nd.path.mnt->mnt_sb != sb) {
3364 /* Journaling quota? */
3365 if (EXT4_SB(sb)->s_qf_names[type]) {
3366 /* Quotafile not in fs root? */
3367 if (nd.path.dentry->d_parent->d_inode != sb->s_root->d_inode)
3369 "EXT4-fs: Quota file not on filesystem root. "
3370 "Journaled quota will not work.\n");
3374 * When we journal data on quota file, we have to flush journal to see
3375 * all updates to the file when we bypass pagecache...
3377 if (ext4_should_journal_data(nd.path.dentry->d_inode)) {
3379 * We don't need to lock updates but journal_flush() could
3380 * otherwise be livelocked...
3382 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
3383 jbd2_journal_flush(EXT4_SB(sb)->s_journal);
3384 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3387 err = vfs_quota_on_path(sb, type, format_id, &nd.path);
3392 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3393 * acquiring the locks... As quota files are never truncated and quota code
3394 * itself serializes the operations (and noone else should touch the files)
3395 * we don't have to be afraid of races */
3396 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
3397 size_t len, loff_t off)
3399 struct inode *inode = sb_dqopt(sb)->files[type];
3400 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3402 int offset = off & (sb->s_blocksize - 1);
3405 struct buffer_head *bh;
3406 loff_t i_size = i_size_read(inode);
3410 if (off+len > i_size)
3413 while (toread > 0) {
3414 tocopy = sb->s_blocksize - offset < toread ?
3415 sb->s_blocksize - offset : toread;
3416 bh = ext4_bread(NULL, inode, blk, 0, &err);
3419 if (!bh) /* A hole? */
3420 memset(data, 0, tocopy);
3422 memcpy(data, bh->b_data+offset, tocopy);
3432 /* Write to quotafile (we know the transaction is already started and has
3433 * enough credits) */
3434 static ssize_t ext4_quota_write(struct super_block *sb, int type,
3435 const char *data, size_t len, loff_t off)
3437 struct inode *inode = sb_dqopt(sb)->files[type];
3438 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3440 int offset = off & (sb->s_blocksize - 1);
3442 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
3443 size_t towrite = len;
3444 struct buffer_head *bh;
3445 handle_t *handle = journal_current_handle();
3448 printk(KERN_WARNING "EXT4-fs: Quota write (off=%llu, len=%llu)"
3449 " cancelled because transaction is not started.\n",
3450 (unsigned long long)off, (unsigned long long)len);
3453 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3454 while (towrite > 0) {
3455 tocopy = sb->s_blocksize - offset < towrite ?
3456 sb->s_blocksize - offset : towrite;
3457 bh = ext4_bread(handle, inode, blk, 1, &err);
3460 if (journal_quota) {
3461 err = ext4_journal_get_write_access(handle, bh);
3468 memcpy(bh->b_data+offset, data, tocopy);
3469 flush_dcache_page(bh->b_page);
3472 err = ext4_journal_dirty_metadata(handle, bh);
3474 /* Always do at least ordered writes for quotas */
3475 err = ext4_jbd2_file_inode(handle, inode);
3476 mark_buffer_dirty(bh);
3487 if (len == towrite) {
3488 mutex_unlock(&inode->i_mutex);
3491 if (inode->i_size < off+len-towrite) {
3492 i_size_write(inode, off+len-towrite);
3493 EXT4_I(inode)->i_disksize = inode->i_size;
3495 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3496 ext4_mark_inode_dirty(handle, inode);
3497 mutex_unlock(&inode->i_mutex);
3498 return len - towrite;
3503 static int ext4_get_sb(struct file_system_type *fs_type,
3504 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3506 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
3509 static struct file_system_type ext4dev_fs_type = {
3510 .owner = THIS_MODULE,
3512 .get_sb = ext4_get_sb,
3513 .kill_sb = kill_block_super,
3514 .fs_flags = FS_REQUIRES_DEV,
3517 static int __init init_ext4_fs(void)
3521 err = init_ext4_mballoc();
3525 err = init_ext4_xattr();
3528 err = init_inodecache();
3531 err = register_filesystem(&ext4dev_fs_type);
3536 destroy_inodecache();
3540 exit_ext4_mballoc();
3544 static void __exit exit_ext4_fs(void)
3546 unregister_filesystem(&ext4dev_fs_type);
3547 destroy_inodecache();
3549 exit_ext4_mballoc();
3552 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
3553 MODULE_DESCRIPTION("Fourth Extended Filesystem with extents");
3554 MODULE_LICENSE("GPL");
3555 module_init(init_ext4_fs)
3556 module_exit(exit_ext4_fs)