2 * linux/fs/ext4/super.c
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
11 * linux/fs/minix/inode.c
13 * Copyright (C) 1991, 1992 Linus Torvalds
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
19 #include <linux/module.h>
20 #include <linux/string.h>
22 #include <linux/time.h>
23 #include <linux/jbd2.h>
24 #include <linux/slab.h>
25 #include <linux/init.h>
26 #include <linux/blkdev.h>
27 #include <linux/parser.h>
28 #include <linux/smp_lock.h>
29 #include <linux/buffer_head.h>
30 #include <linux/exportfs.h>
31 #include <linux/vfs.h>
32 #include <linux/random.h>
33 #include <linux/mount.h>
34 #include <linux/namei.h>
35 #include <linux/quotaops.h>
36 #include <linux/seq_file.h>
37 #include <linux/log2.h>
38 #include <linux/crc16.h>
39 #include <asm/uaccess.h>
42 #include "ext4_jbd2.h"
48 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
49 unsigned long journal_devnum);
50 static int ext4_create_journal(struct super_block *, struct ext4_super_block *,
52 static void ext4_commit_super (struct super_block * sb,
53 struct ext4_super_block * es,
55 static void ext4_mark_recovery_complete(struct super_block * sb,
56 struct ext4_super_block * es);
57 static void ext4_clear_journal_err(struct super_block * sb,
58 struct ext4_super_block * es);
59 static int ext4_sync_fs(struct super_block *sb, int wait);
60 static const char *ext4_decode_error(struct super_block * sb, int errno,
62 static int ext4_remount (struct super_block * sb, int * flags, char * data);
63 static int ext4_statfs (struct dentry * dentry, struct kstatfs * buf);
64 static void ext4_unlockfs(struct super_block *sb);
65 static void ext4_write_super (struct super_block * sb);
66 static void ext4_write_super_lockfs(struct super_block *sb);
69 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
70 struct ext4_group_desc *bg)
72 return le32_to_cpu(bg->bg_block_bitmap_lo) |
73 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
74 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
77 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
78 struct ext4_group_desc *bg)
80 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
81 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
82 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
85 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
86 struct ext4_group_desc *bg)
88 return le32_to_cpu(bg->bg_inode_table_lo) |
89 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
90 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
93 void ext4_block_bitmap_set(struct super_block *sb,
94 struct ext4_group_desc *bg, ext4_fsblk_t blk)
96 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
97 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
98 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
101 void ext4_inode_bitmap_set(struct super_block *sb,
102 struct ext4_group_desc *bg, ext4_fsblk_t blk)
104 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
105 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
106 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
109 void ext4_inode_table_set(struct super_block *sb,
110 struct ext4_group_desc *bg, ext4_fsblk_t blk)
112 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
113 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
114 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
118 * Wrappers for jbd2_journal_start/end.
120 * The only special thing we need to do here is to make sure that all
121 * journal_end calls result in the superblock being marked dirty, so
122 * that sync() will call the filesystem's write_super callback if
125 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
129 if (sb->s_flags & MS_RDONLY)
130 return ERR_PTR(-EROFS);
132 /* Special case here: if the journal has aborted behind our
133 * backs (eg. EIO in the commit thread), then we still need to
134 * take the FS itself readonly cleanly. */
135 journal = EXT4_SB(sb)->s_journal;
136 if (is_journal_aborted(journal)) {
137 ext4_abort(sb, __func__,
138 "Detected aborted journal");
139 return ERR_PTR(-EROFS);
142 return jbd2_journal_start(journal, nblocks);
146 * The only special thing we need to do here is to make sure that all
147 * jbd2_journal_stop calls result in the superblock being marked dirty, so
148 * that sync() will call the filesystem's write_super callback if
151 int __ext4_journal_stop(const char *where, handle_t *handle)
153 struct super_block *sb;
157 sb = handle->h_transaction->t_journal->j_private;
159 rc = jbd2_journal_stop(handle);
164 __ext4_std_error(sb, where, err);
168 void ext4_journal_abort_handle(const char *caller, const char *err_fn,
169 struct buffer_head *bh, handle_t *handle, int err)
172 const char *errstr = ext4_decode_error(NULL, err, nbuf);
175 BUFFER_TRACE(bh, "abort");
180 if (is_handle_aborted(handle))
183 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
184 caller, errstr, err_fn);
186 jbd2_journal_abort_handle(handle);
189 /* Deal with the reporting of failure conditions on a filesystem such as
190 * inconsistencies detected or read IO failures.
192 * On ext2, we can store the error state of the filesystem in the
193 * superblock. That is not possible on ext4, because we may have other
194 * write ordering constraints on the superblock which prevent us from
195 * writing it out straight away; and given that the journal is about to
196 * be aborted, we can't rely on the current, or future, transactions to
197 * write out the superblock safely.
199 * We'll just use the jbd2_journal_abort() error code to record an error in
200 * the journal instead. On recovery, the journal will compain about
201 * that error until we've noted it down and cleared it.
204 static void ext4_handle_error(struct super_block *sb)
206 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
208 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
209 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
211 if (sb->s_flags & MS_RDONLY)
214 if (!test_opt (sb, ERRORS_CONT)) {
215 journal_t *journal = EXT4_SB(sb)->s_journal;
217 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
219 jbd2_journal_abort(journal, -EIO);
221 if (test_opt (sb, ERRORS_RO)) {
222 printk (KERN_CRIT "Remounting filesystem read-only\n");
223 sb->s_flags |= MS_RDONLY;
225 ext4_commit_super(sb, es, 1);
226 if (test_opt(sb, ERRORS_PANIC))
227 panic("EXT4-fs (device %s): panic forced after error\n",
231 void ext4_error (struct super_block * sb, const char * function,
232 const char * fmt, ...)
237 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ",sb->s_id, function);
242 ext4_handle_error(sb);
245 static const char *ext4_decode_error(struct super_block * sb, int errno,
252 errstr = "IO failure";
255 errstr = "Out of memory";
258 if (!sb || EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)
259 errstr = "Journal has aborted";
261 errstr = "Readonly filesystem";
264 /* If the caller passed in an extra buffer for unknown
265 * errors, textualise them now. Else we just return
268 /* Check for truncated error codes... */
269 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
278 /* __ext4_std_error decodes expected errors from journaling functions
279 * automatically and invokes the appropriate error response. */
281 void __ext4_std_error (struct super_block * sb, const char * function,
287 /* Special case: if the error is EROFS, and we're not already
288 * inside a transaction, then there's really no point in logging
290 if (errno == -EROFS && journal_current_handle() == NULL &&
291 (sb->s_flags & MS_RDONLY))
294 errstr = ext4_decode_error(sb, errno, nbuf);
295 printk (KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
296 sb->s_id, function, errstr);
298 ext4_handle_error(sb);
302 * ext4_abort is a much stronger failure handler than ext4_error. The
303 * abort function may be used to deal with unrecoverable failures such
304 * as journal IO errors or ENOMEM at a critical moment in log management.
306 * We unconditionally force the filesystem into an ABORT|READONLY state,
307 * unless the error response on the fs has been set to panic in which
308 * case we take the easy way out and panic immediately.
311 void ext4_abort (struct super_block * sb, const char * function,
312 const char * fmt, ...)
316 printk (KERN_CRIT "ext4_abort called.\n");
319 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ",sb->s_id, function);
324 if (test_opt(sb, ERRORS_PANIC))
325 panic("EXT4-fs panic from previous error\n");
327 if (sb->s_flags & MS_RDONLY)
330 printk(KERN_CRIT "Remounting filesystem read-only\n");
331 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
332 sb->s_flags |= MS_RDONLY;
333 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
334 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
337 void ext4_warning (struct super_block * sb, const char * function,
338 const char * fmt, ...)
343 printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
350 void ext4_update_dynamic_rev(struct super_block *sb)
352 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
354 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
357 ext4_warning(sb, __func__,
358 "updating to rev %d because of new feature flag, "
359 "running e2fsck is recommended",
362 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
363 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
364 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
365 /* leave es->s_feature_*compat flags alone */
366 /* es->s_uuid will be set by e2fsck if empty */
369 * The rest of the superblock fields should be zero, and if not it
370 * means they are likely already in use, so leave them alone. We
371 * can leave it up to e2fsck to clean up any inconsistencies there.
375 int ext4_update_compat_feature(handle_t *handle,
376 struct super_block *sb, __u32 compat)
379 if (!EXT4_HAS_COMPAT_FEATURE(sb, compat)) {
380 err = ext4_journal_get_write_access(handle,
384 EXT4_SET_COMPAT_FEATURE(sb, compat);
387 BUFFER_TRACE(EXT4_SB(sb)->s_sbh,
388 "call ext4_journal_dirty_met adata");
389 err = ext4_journal_dirty_metadata(handle,
395 int ext4_update_rocompat_feature(handle_t *handle,
396 struct super_block *sb, __u32 rocompat)
399 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, rocompat)) {
400 err = ext4_journal_get_write_access(handle,
404 EXT4_SET_RO_COMPAT_FEATURE(sb, rocompat);
407 BUFFER_TRACE(EXT4_SB(sb)->s_sbh,
408 "call ext4_journal_dirty_met adata");
409 err = ext4_journal_dirty_metadata(handle,
415 int ext4_update_incompat_feature(handle_t *handle,
416 struct super_block *sb, __u32 incompat)
419 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, incompat)) {
420 err = ext4_journal_get_write_access(handle,
424 EXT4_SET_INCOMPAT_FEATURE(sb, incompat);
427 BUFFER_TRACE(EXT4_SB(sb)->s_sbh,
428 "call ext4_journal_dirty_met adata");
429 err = ext4_journal_dirty_metadata(handle,
436 * Open the external journal device
438 static struct block_device *ext4_blkdev_get(dev_t dev)
440 struct block_device *bdev;
441 char b[BDEVNAME_SIZE];
443 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
449 printk(KERN_ERR "EXT4: failed to open journal device %s: %ld\n",
450 __bdevname(dev, b), PTR_ERR(bdev));
455 * Release the journal device
457 static int ext4_blkdev_put(struct block_device *bdev)
460 return blkdev_put(bdev);
463 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
465 struct block_device *bdev;
468 bdev = sbi->journal_bdev;
470 ret = ext4_blkdev_put(bdev);
471 sbi->journal_bdev = NULL;
476 static inline struct inode *orphan_list_entry(struct list_head *l)
478 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
481 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
485 printk(KERN_ERR "sb orphan head is %d\n",
486 le32_to_cpu(sbi->s_es->s_last_orphan));
488 printk(KERN_ERR "sb_info orphan list:\n");
489 list_for_each(l, &sbi->s_orphan) {
490 struct inode *inode = orphan_list_entry(l);
492 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
493 inode->i_sb->s_id, inode->i_ino, inode,
494 inode->i_mode, inode->i_nlink,
499 static void ext4_put_super (struct super_block * sb)
501 struct ext4_sb_info *sbi = EXT4_SB(sb);
502 struct ext4_super_block *es = sbi->s_es;
506 ext4_ext_release(sb);
507 ext4_xattr_put_super(sb);
508 jbd2_journal_destroy(sbi->s_journal);
509 sbi->s_journal = NULL;
510 if (!(sb->s_flags & MS_RDONLY)) {
511 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
512 es->s_state = cpu_to_le16(sbi->s_mount_state);
513 BUFFER_TRACE(sbi->s_sbh, "marking dirty");
514 mark_buffer_dirty(sbi->s_sbh);
515 ext4_commit_super(sb, es, 1);
518 for (i = 0; i < sbi->s_gdb_count; i++)
519 brelse(sbi->s_group_desc[i]);
520 kfree(sbi->s_group_desc);
521 kfree(sbi->s_flex_groups);
522 percpu_counter_destroy(&sbi->s_freeblocks_counter);
523 percpu_counter_destroy(&sbi->s_freeinodes_counter);
524 percpu_counter_destroy(&sbi->s_dirs_counter);
527 for (i = 0; i < MAXQUOTAS; i++)
528 kfree(sbi->s_qf_names[i]);
531 /* Debugging code just in case the in-memory inode orphan list
532 * isn't empty. The on-disk one can be non-empty if we've
533 * detected an error and taken the fs readonly, but the
534 * in-memory list had better be clean by this point. */
535 if (!list_empty(&sbi->s_orphan))
536 dump_orphan_list(sb, sbi);
537 J_ASSERT(list_empty(&sbi->s_orphan));
539 invalidate_bdev(sb->s_bdev);
540 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
542 * Invalidate the journal device's buffers. We don't want them
543 * floating about in memory - the physical journal device may
544 * hotswapped, and it breaks the `ro-after' testing code.
546 sync_blockdev(sbi->journal_bdev);
547 invalidate_bdev(sbi->journal_bdev);
548 ext4_blkdev_remove(sbi);
550 sb->s_fs_info = NULL;
555 static struct kmem_cache *ext4_inode_cachep;
558 * Called inside transaction, so use GFP_NOFS
560 static struct inode *ext4_alloc_inode(struct super_block *sb)
562 struct ext4_inode_info *ei;
564 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
567 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
568 ei->i_acl = EXT4_ACL_NOT_CACHED;
569 ei->i_default_acl = EXT4_ACL_NOT_CACHED;
571 ei->i_block_alloc_info = NULL;
572 ei->vfs_inode.i_version = 1;
573 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
574 INIT_LIST_HEAD(&ei->i_prealloc_list);
575 spin_lock_init(&ei->i_prealloc_lock);
576 return &ei->vfs_inode;
579 static void ext4_destroy_inode(struct inode *inode)
581 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
582 printk("EXT4 Inode %p: orphan list check failed!\n",
584 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
585 EXT4_I(inode), sizeof(struct ext4_inode_info),
589 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
592 static void init_once(struct kmem_cache *cachep, void *foo)
594 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
596 INIT_LIST_HEAD(&ei->i_orphan);
597 #ifdef CONFIG_EXT4DEV_FS_XATTR
598 init_rwsem(&ei->xattr_sem);
600 init_rwsem(&ei->i_data_sem);
601 inode_init_once(&ei->vfs_inode);
604 static int init_inodecache(void)
606 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
607 sizeof(struct ext4_inode_info),
608 0, (SLAB_RECLAIM_ACCOUNT|
611 if (ext4_inode_cachep == NULL)
616 static void destroy_inodecache(void)
618 kmem_cache_destroy(ext4_inode_cachep);
621 static void ext4_clear_inode(struct inode *inode)
623 struct ext4_block_alloc_info *rsv = EXT4_I(inode)->i_block_alloc_info;
624 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
625 if (EXT4_I(inode)->i_acl &&
626 EXT4_I(inode)->i_acl != EXT4_ACL_NOT_CACHED) {
627 posix_acl_release(EXT4_I(inode)->i_acl);
628 EXT4_I(inode)->i_acl = EXT4_ACL_NOT_CACHED;
630 if (EXT4_I(inode)->i_default_acl &&
631 EXT4_I(inode)->i_default_acl != EXT4_ACL_NOT_CACHED) {
632 posix_acl_release(EXT4_I(inode)->i_default_acl);
633 EXT4_I(inode)->i_default_acl = EXT4_ACL_NOT_CACHED;
636 ext4_discard_reservation(inode);
637 EXT4_I(inode)->i_block_alloc_info = NULL;
642 static inline void ext4_show_quota_options(struct seq_file *seq, struct super_block *sb)
644 #if defined(CONFIG_QUOTA)
645 struct ext4_sb_info *sbi = EXT4_SB(sb);
647 if (sbi->s_jquota_fmt)
648 seq_printf(seq, ",jqfmt=%s",
649 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold": "vfsv0");
651 if (sbi->s_qf_names[USRQUOTA])
652 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
654 if (sbi->s_qf_names[GRPQUOTA])
655 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
657 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
658 seq_puts(seq, ",usrquota");
660 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
661 seq_puts(seq, ",grpquota");
667 * - it's set to a non-default value OR
668 * - if the per-sb default is different from the global default
670 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
673 unsigned long def_mount_opts;
674 struct super_block *sb = vfs->mnt_sb;
675 struct ext4_sb_info *sbi = EXT4_SB(sb);
676 struct ext4_super_block *es = sbi->s_es;
678 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
679 def_errors = le16_to_cpu(es->s_errors);
681 if (sbi->s_sb_block != 1)
682 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
683 if (test_opt(sb, MINIX_DF))
684 seq_puts(seq, ",minixdf");
685 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
686 seq_puts(seq, ",grpid");
687 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
688 seq_puts(seq, ",nogrpid");
689 if (sbi->s_resuid != EXT4_DEF_RESUID ||
690 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
691 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
693 if (sbi->s_resgid != EXT4_DEF_RESGID ||
694 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
695 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
697 if (test_opt(sb, ERRORS_RO)) {
698 if (def_errors == EXT4_ERRORS_PANIC ||
699 def_errors == EXT4_ERRORS_CONTINUE) {
700 seq_puts(seq, ",errors=remount-ro");
703 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
704 seq_puts(seq, ",errors=continue");
705 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
706 seq_puts(seq, ",errors=panic");
707 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
708 seq_puts(seq, ",nouid32");
709 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
710 seq_puts(seq, ",debug");
711 if (test_opt(sb, OLDALLOC))
712 seq_puts(seq, ",oldalloc");
713 #ifdef CONFIG_EXT4DEV_FS_XATTR
714 if (test_opt(sb, XATTR_USER) &&
715 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
716 seq_puts(seq, ",user_xattr");
717 if (!test_opt(sb, XATTR_USER) &&
718 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
719 seq_puts(seq, ",nouser_xattr");
722 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
723 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
724 seq_puts(seq, ",acl");
725 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
726 seq_puts(seq, ",noacl");
728 if (!test_opt(sb, RESERVATION))
729 seq_puts(seq, ",noreservation");
730 if (sbi->s_commit_interval) {
731 seq_printf(seq, ",commit=%u",
732 (unsigned) (sbi->s_commit_interval / HZ));
735 * We're changing the default of barrier mount option, so
736 * let's always display its mount state so it's clear what its
739 seq_puts(seq, ",barrier=");
740 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
741 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
742 seq_puts(seq, ",journal_async_commit");
743 if (test_opt(sb, NOBH))
744 seq_puts(seq, ",nobh");
745 if (!test_opt(sb, EXTENTS))
746 seq_puts(seq, ",noextents");
747 if (!test_opt(sb, MBALLOC))
748 seq_puts(seq, ",nomballoc");
749 if (test_opt(sb, I_VERSION))
750 seq_puts(seq, ",i_version");
753 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
755 * journal mode get enabled in different ways
756 * So just print the value even if we didn't specify it
758 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
759 seq_puts(seq, ",data=journal");
760 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
761 seq_puts(seq, ",data=ordered");
762 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
763 seq_puts(seq, ",data=writeback");
765 ext4_show_quota_options(seq, sb);
770 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
771 u64 ino, u32 generation)
775 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
776 return ERR_PTR(-ESTALE);
777 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
778 return ERR_PTR(-ESTALE);
780 /* iget isn't really right if the inode is currently unallocated!!
782 * ext4_read_inode will return a bad_inode if the inode had been
783 * deleted, so we should be safe.
785 * Currently we don't know the generation for parent directory, so
786 * a generation of 0 means "accept any"
788 inode = ext4_iget(sb, ino);
790 return ERR_CAST(inode);
791 if (generation && inode->i_generation != generation) {
793 return ERR_PTR(-ESTALE);
799 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
800 int fh_len, int fh_type)
802 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
806 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
807 int fh_len, int fh_type)
809 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
814 #define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
815 #define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
817 static int ext4_dquot_initialize(struct inode *inode, int type);
818 static int ext4_dquot_drop(struct inode *inode);
819 static int ext4_write_dquot(struct dquot *dquot);
820 static int ext4_acquire_dquot(struct dquot *dquot);
821 static int ext4_release_dquot(struct dquot *dquot);
822 static int ext4_mark_dquot_dirty(struct dquot *dquot);
823 static int ext4_write_info(struct super_block *sb, int type);
824 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
825 char *path, int remount);
826 static int ext4_quota_on_mount(struct super_block *sb, int type);
827 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
828 size_t len, loff_t off);
829 static ssize_t ext4_quota_write(struct super_block *sb, int type,
830 const char *data, size_t len, loff_t off);
832 static struct dquot_operations ext4_quota_operations = {
833 .initialize = ext4_dquot_initialize,
834 .drop = ext4_dquot_drop,
835 .alloc_space = dquot_alloc_space,
836 .alloc_inode = dquot_alloc_inode,
837 .free_space = dquot_free_space,
838 .free_inode = dquot_free_inode,
839 .transfer = dquot_transfer,
840 .write_dquot = ext4_write_dquot,
841 .acquire_dquot = ext4_acquire_dquot,
842 .release_dquot = ext4_release_dquot,
843 .mark_dirty = ext4_mark_dquot_dirty,
844 .write_info = ext4_write_info
847 static struct quotactl_ops ext4_qctl_operations = {
848 .quota_on = ext4_quota_on,
849 .quota_off = vfs_quota_off,
850 .quota_sync = vfs_quota_sync,
851 .get_info = vfs_get_dqinfo,
852 .set_info = vfs_set_dqinfo,
853 .get_dqblk = vfs_get_dqblk,
854 .set_dqblk = vfs_set_dqblk
858 static const struct super_operations ext4_sops = {
859 .alloc_inode = ext4_alloc_inode,
860 .destroy_inode = ext4_destroy_inode,
861 .write_inode = ext4_write_inode,
862 .dirty_inode = ext4_dirty_inode,
863 .delete_inode = ext4_delete_inode,
864 .put_super = ext4_put_super,
865 .write_super = ext4_write_super,
866 .sync_fs = ext4_sync_fs,
867 .write_super_lockfs = ext4_write_super_lockfs,
868 .unlockfs = ext4_unlockfs,
869 .statfs = ext4_statfs,
870 .remount_fs = ext4_remount,
871 .clear_inode = ext4_clear_inode,
872 .show_options = ext4_show_options,
874 .quota_read = ext4_quota_read,
875 .quota_write = ext4_quota_write,
879 static const struct export_operations ext4_export_ops = {
880 .fh_to_dentry = ext4_fh_to_dentry,
881 .fh_to_parent = ext4_fh_to_parent,
882 .get_parent = ext4_get_parent,
886 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
887 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
888 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
889 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
890 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
891 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
892 Opt_journal_checksum, Opt_journal_async_commit,
893 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
894 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
895 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
896 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
897 Opt_grpquota, Opt_extents, Opt_noextents, Opt_i_version,
898 Opt_mballoc, Opt_nomballoc, Opt_stripe,
901 static match_table_t tokens = {
902 {Opt_bsd_df, "bsddf"},
903 {Opt_minix_df, "minixdf"},
904 {Opt_grpid, "grpid"},
905 {Opt_grpid, "bsdgroups"},
906 {Opt_nogrpid, "nogrpid"},
907 {Opt_nogrpid, "sysvgroups"},
908 {Opt_resgid, "resgid=%u"},
909 {Opt_resuid, "resuid=%u"},
911 {Opt_err_cont, "errors=continue"},
912 {Opt_err_panic, "errors=panic"},
913 {Opt_err_ro, "errors=remount-ro"},
914 {Opt_nouid32, "nouid32"},
915 {Opt_nocheck, "nocheck"},
916 {Opt_nocheck, "check=none"},
917 {Opt_debug, "debug"},
918 {Opt_oldalloc, "oldalloc"},
919 {Opt_orlov, "orlov"},
920 {Opt_user_xattr, "user_xattr"},
921 {Opt_nouser_xattr, "nouser_xattr"},
923 {Opt_noacl, "noacl"},
924 {Opt_reservation, "reservation"},
925 {Opt_noreservation, "noreservation"},
926 {Opt_noload, "noload"},
929 {Opt_commit, "commit=%u"},
930 {Opt_journal_update, "journal=update"},
931 {Opt_journal_inum, "journal=%u"},
932 {Opt_journal_dev, "journal_dev=%u"},
933 {Opt_journal_checksum, "journal_checksum"},
934 {Opt_journal_async_commit, "journal_async_commit"},
935 {Opt_abort, "abort"},
936 {Opt_data_journal, "data=journal"},
937 {Opt_data_ordered, "data=ordered"},
938 {Opt_data_writeback, "data=writeback"},
939 {Opt_offusrjquota, "usrjquota="},
940 {Opt_usrjquota, "usrjquota=%s"},
941 {Opt_offgrpjquota, "grpjquota="},
942 {Opt_grpjquota, "grpjquota=%s"},
943 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
944 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
945 {Opt_grpquota, "grpquota"},
946 {Opt_noquota, "noquota"},
947 {Opt_quota, "quota"},
948 {Opt_usrquota, "usrquota"},
949 {Opt_barrier, "barrier=%u"},
950 {Opt_extents, "extents"},
951 {Opt_noextents, "noextents"},
952 {Opt_i_version, "i_version"},
953 {Opt_mballoc, "mballoc"},
954 {Opt_nomballoc, "nomballoc"},
955 {Opt_stripe, "stripe=%u"},
956 {Opt_resize, "resize"},
960 static ext4_fsblk_t get_sb_block(void **data)
962 ext4_fsblk_t sb_block;
963 char *options = (char *) *data;
965 if (!options || strncmp(options, "sb=", 3) != 0)
966 return 1; /* Default location */
968 /*todo: use simple_strtoll with >32bit ext4 */
969 sb_block = simple_strtoul(options, &options, 0);
970 if (*options && *options != ',') {
971 printk("EXT4-fs: Invalid sb specification: %s\n",
977 *data = (void *) options;
981 static int parse_options (char *options, struct super_block *sb,
982 unsigned int *inum, unsigned long *journal_devnum,
983 ext4_fsblk_t *n_blocks_count, int is_remount)
985 struct ext4_sb_info *sbi = EXT4_SB(sb);
987 substring_t args[MAX_OPT_ARGS];
998 while ((p = strsep (&options, ",")) != NULL) {
1003 token = match_token(p, tokens, args);
1006 clear_opt (sbi->s_mount_opt, MINIX_DF);
1009 set_opt (sbi->s_mount_opt, MINIX_DF);
1012 set_opt (sbi->s_mount_opt, GRPID);
1015 clear_opt (sbi->s_mount_opt, GRPID);
1018 if (match_int(&args[0], &option))
1020 sbi->s_resuid = option;
1023 if (match_int(&args[0], &option))
1025 sbi->s_resgid = option;
1028 /* handled by get_sb_block() instead of here */
1029 /* *sb_block = match_int(&args[0]); */
1032 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1033 clear_opt (sbi->s_mount_opt, ERRORS_RO);
1034 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
1037 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1038 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1039 set_opt (sbi->s_mount_opt, ERRORS_RO);
1042 clear_opt (sbi->s_mount_opt, ERRORS_RO);
1043 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1044 set_opt (sbi->s_mount_opt, ERRORS_CONT);
1047 set_opt (sbi->s_mount_opt, NO_UID32);
1050 clear_opt (sbi->s_mount_opt, CHECK);
1053 set_opt (sbi->s_mount_opt, DEBUG);
1056 set_opt (sbi->s_mount_opt, OLDALLOC);
1059 clear_opt (sbi->s_mount_opt, OLDALLOC);
1061 #ifdef CONFIG_EXT4DEV_FS_XATTR
1062 case Opt_user_xattr:
1063 set_opt (sbi->s_mount_opt, XATTR_USER);
1065 case Opt_nouser_xattr:
1066 clear_opt (sbi->s_mount_opt, XATTR_USER);
1069 case Opt_user_xattr:
1070 case Opt_nouser_xattr:
1071 printk("EXT4 (no)user_xattr options not supported\n");
1074 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
1076 set_opt(sbi->s_mount_opt, POSIX_ACL);
1079 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1084 printk("EXT4 (no)acl options not supported\n");
1087 case Opt_reservation:
1088 set_opt(sbi->s_mount_opt, RESERVATION);
1090 case Opt_noreservation:
1091 clear_opt(sbi->s_mount_opt, RESERVATION);
1093 case Opt_journal_update:
1095 /* Eventually we will want to be able to create
1096 a journal file here. For now, only allow the
1097 user to specify an existing inode to be the
1100 printk(KERN_ERR "EXT4-fs: cannot specify "
1101 "journal on remount\n");
1104 set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
1106 case Opt_journal_inum:
1108 printk(KERN_ERR "EXT4-fs: cannot specify "
1109 "journal on remount\n");
1112 if (match_int(&args[0], &option))
1116 case Opt_journal_dev:
1118 printk(KERN_ERR "EXT4-fs: cannot specify "
1119 "journal on remount\n");
1122 if (match_int(&args[0], &option))
1124 *journal_devnum = option;
1126 case Opt_journal_checksum:
1127 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1129 case Opt_journal_async_commit:
1130 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1131 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1134 set_opt (sbi->s_mount_opt, NOLOAD);
1137 if (match_int(&args[0], &option))
1142 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1143 sbi->s_commit_interval = HZ * option;
1145 case Opt_data_journal:
1146 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1148 case Opt_data_ordered:
1149 data_opt = EXT4_MOUNT_ORDERED_DATA;
1151 case Opt_data_writeback:
1152 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1155 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
1158 "EXT4-fs: cannot change data "
1159 "mode on remount\n");
1163 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
1164 sbi->s_mount_opt |= data_opt;
1174 if ((sb_any_quota_enabled(sb) ||
1175 sb_any_quota_suspended(sb)) &&
1176 !sbi->s_qf_names[qtype]) {
1178 "EXT4-fs: Cannot change journaled "
1179 "quota options when quota turned on.\n");
1182 qname = match_strdup(&args[0]);
1185 "EXT4-fs: not enough memory for "
1186 "storing quotafile name.\n");
1189 if (sbi->s_qf_names[qtype] &&
1190 strcmp(sbi->s_qf_names[qtype], qname)) {
1192 "EXT4-fs: %s quota file already "
1193 "specified.\n", QTYPE2NAME(qtype));
1197 sbi->s_qf_names[qtype] = qname;
1198 if (strchr(sbi->s_qf_names[qtype], '/')) {
1200 "EXT4-fs: quotafile must be on "
1201 "filesystem root.\n");
1202 kfree(sbi->s_qf_names[qtype]);
1203 sbi->s_qf_names[qtype] = NULL;
1206 set_opt(sbi->s_mount_opt, QUOTA);
1208 case Opt_offusrjquota:
1211 case Opt_offgrpjquota:
1214 if ((sb_any_quota_enabled(sb) ||
1215 sb_any_quota_suspended(sb)) &&
1216 sbi->s_qf_names[qtype]) {
1217 printk(KERN_ERR "EXT4-fs: Cannot change "
1218 "journaled quota options when "
1219 "quota turned on.\n");
1223 * The space will be released later when all options
1224 * are confirmed to be correct
1226 sbi->s_qf_names[qtype] = NULL;
1228 case Opt_jqfmt_vfsold:
1229 qfmt = QFMT_VFS_OLD;
1231 case Opt_jqfmt_vfsv0:
1234 if ((sb_any_quota_enabled(sb) ||
1235 sb_any_quota_suspended(sb)) &&
1236 sbi->s_jquota_fmt != qfmt) {
1237 printk(KERN_ERR "EXT4-fs: Cannot change "
1238 "journaled quota options when "
1239 "quota turned on.\n");
1242 sbi->s_jquota_fmt = qfmt;
1246 set_opt(sbi->s_mount_opt, QUOTA);
1247 set_opt(sbi->s_mount_opt, USRQUOTA);
1250 set_opt(sbi->s_mount_opt, QUOTA);
1251 set_opt(sbi->s_mount_opt, GRPQUOTA);
1254 if (sb_any_quota_enabled(sb)) {
1255 printk(KERN_ERR "EXT4-fs: Cannot change quota "
1256 "options when quota turned on.\n");
1259 clear_opt(sbi->s_mount_opt, QUOTA);
1260 clear_opt(sbi->s_mount_opt, USRQUOTA);
1261 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1268 "EXT4-fs: quota options not supported.\n");
1272 case Opt_offusrjquota:
1273 case Opt_offgrpjquota:
1274 case Opt_jqfmt_vfsold:
1275 case Opt_jqfmt_vfsv0:
1277 "EXT4-fs: journaled quota options not "
1284 set_opt(sbi->s_mount_opt, ABORT);
1287 if (match_int(&args[0], &option))
1290 set_opt(sbi->s_mount_opt, BARRIER);
1292 clear_opt(sbi->s_mount_opt, BARRIER);
1298 printk("EXT4-fs: resize option only available "
1302 if (match_int(&args[0], &option) != 0)
1304 *n_blocks_count = option;
1307 set_opt(sbi->s_mount_opt, NOBH);
1310 clear_opt(sbi->s_mount_opt, NOBH);
1313 set_opt (sbi->s_mount_opt, EXTENTS);
1316 clear_opt (sbi->s_mount_opt, EXTENTS);
1319 set_opt(sbi->s_mount_opt, I_VERSION);
1320 sb->s_flags |= MS_I_VERSION;
1323 set_opt(sbi->s_mount_opt, MBALLOC);
1326 clear_opt(sbi->s_mount_opt, MBALLOC);
1329 if (match_int(&args[0], &option))
1333 sbi->s_stripe = option;
1337 "EXT4-fs: Unrecognized mount option \"%s\" "
1338 "or missing value\n", p);
1343 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1344 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
1345 sbi->s_qf_names[USRQUOTA])
1346 clear_opt(sbi->s_mount_opt, USRQUOTA);
1348 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
1349 sbi->s_qf_names[GRPQUOTA])
1350 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1352 if ((sbi->s_qf_names[USRQUOTA] &&
1353 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
1354 (sbi->s_qf_names[GRPQUOTA] &&
1355 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1356 printk(KERN_ERR "EXT4-fs: old and new quota "
1357 "format mixing.\n");
1361 if (!sbi->s_jquota_fmt) {
1362 printk(KERN_ERR "EXT4-fs: journaled quota format "
1363 "not specified.\n");
1367 if (sbi->s_jquota_fmt) {
1368 printk(KERN_ERR "EXT4-fs: journaled quota format "
1369 "specified with no journaling "
1378 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1381 struct ext4_sb_info *sbi = EXT4_SB(sb);
1384 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1385 printk (KERN_ERR "EXT4-fs warning: revision level too high, "
1386 "forcing read-only mode\n");
1391 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1392 printk (KERN_WARNING "EXT4-fs warning: mounting unchecked fs, "
1393 "running e2fsck is recommended\n");
1394 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1395 printk (KERN_WARNING
1396 "EXT4-fs warning: mounting fs with errors, "
1397 "running e2fsck is recommended\n");
1398 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1399 le16_to_cpu(es->s_mnt_count) >=
1400 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1401 printk (KERN_WARNING
1402 "EXT4-fs warning: maximal mount count reached, "
1403 "running e2fsck is recommended\n");
1404 else if (le32_to_cpu(es->s_checkinterval) &&
1405 (le32_to_cpu(es->s_lastcheck) +
1406 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1407 printk (KERN_WARNING
1408 "EXT4-fs warning: checktime reached, "
1409 "running e2fsck is recommended\n");
1411 /* @@@ We _will_ want to clear the valid bit if we find
1412 * inconsistencies, to force a fsck at reboot. But for
1413 * a plain journaled filesystem we can keep it set as
1416 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1418 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1419 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1420 le16_add_cpu(&es->s_mnt_count, 1);
1421 es->s_mtime = cpu_to_le32(get_seconds());
1422 ext4_update_dynamic_rev(sb);
1423 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1425 ext4_commit_super(sb, es, 1);
1426 if (test_opt(sb, DEBUG))
1427 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%lu, "
1428 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1430 sbi->s_groups_count,
1431 EXT4_BLOCKS_PER_GROUP(sb),
1432 EXT4_INODES_PER_GROUP(sb),
1435 printk(KERN_INFO "EXT4 FS on %s, ", sb->s_id);
1436 if (EXT4_SB(sb)->s_journal->j_inode == NULL) {
1437 char b[BDEVNAME_SIZE];
1439 printk("external journal on %s\n",
1440 bdevname(EXT4_SB(sb)->s_journal->j_dev, b));
1442 printk("internal journal\n");
1447 static int ext4_fill_flex_info(struct super_block *sb)
1449 struct ext4_sb_info *sbi = EXT4_SB(sb);
1450 struct ext4_group_desc *gdp = NULL;
1451 struct buffer_head *bh;
1452 ext4_group_t flex_group_count;
1453 ext4_group_t flex_group;
1454 int groups_per_flex = 0;
1455 __u64 block_bitmap = 0;
1458 if (!sbi->s_es->s_log_groups_per_flex) {
1459 sbi->s_log_groups_per_flex = 0;
1463 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1464 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1466 flex_group_count = (sbi->s_groups_count + groups_per_flex - 1) /
1468 sbi->s_flex_groups = kmalloc(flex_group_count *
1469 sizeof(struct flex_groups), GFP_KERNEL);
1470 if (sbi->s_flex_groups == NULL) {
1471 printk(KERN_ERR "EXT4-fs: not enough memory\n");
1474 memset(sbi->s_flex_groups, 0, flex_group_count *
1475 sizeof(struct flex_groups));
1477 gdp = ext4_get_group_desc(sb, 1, &bh);
1478 block_bitmap = ext4_block_bitmap(sb, gdp) - 1;
1480 for (i = 0; i < sbi->s_groups_count; i++) {
1481 gdp = ext4_get_group_desc(sb, i, &bh);
1483 flex_group = ext4_flex_group(sbi, i);
1484 sbi->s_flex_groups[flex_group].free_inodes +=
1485 le16_to_cpu(gdp->bg_free_inodes_count);
1486 sbi->s_flex_groups[flex_group].free_blocks +=
1487 le16_to_cpu(gdp->bg_free_blocks_count);
1495 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1496 struct ext4_group_desc *gdp)
1500 if (sbi->s_es->s_feature_ro_compat &
1501 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1502 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1503 __le32 le_group = cpu_to_le32(block_group);
1505 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1506 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1507 crc = crc16(crc, (__u8 *)gdp, offset);
1508 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1509 /* for checksum of struct ext4_group_desc do the rest...*/
1510 if ((sbi->s_es->s_feature_incompat &
1511 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1512 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1513 crc = crc16(crc, (__u8 *)gdp + offset,
1514 le16_to_cpu(sbi->s_es->s_desc_size) -
1518 return cpu_to_le16(crc);
1521 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1522 struct ext4_group_desc *gdp)
1524 if ((sbi->s_es->s_feature_ro_compat &
1525 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1526 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1532 /* Called at mount-time, super-block is locked */
1533 static int ext4_check_descriptors(struct super_block *sb)
1535 struct ext4_sb_info *sbi = EXT4_SB(sb);
1536 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1537 ext4_fsblk_t last_block;
1538 ext4_fsblk_t block_bitmap;
1539 ext4_fsblk_t inode_bitmap;
1540 ext4_fsblk_t inode_table;
1541 int flexbg_flag = 0;
1544 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1547 ext4_debug ("Checking group descriptors");
1549 for (i = 0; i < sbi->s_groups_count; i++) {
1550 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1552 if (i == sbi->s_groups_count - 1 || flexbg_flag)
1553 last_block = ext4_blocks_count(sbi->s_es) - 1;
1555 last_block = first_block +
1556 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1558 block_bitmap = ext4_block_bitmap(sb, gdp);
1559 if (block_bitmap < first_block || block_bitmap > last_block)
1561 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1562 "Block bitmap for group %lu not in group "
1563 "(block %llu)!", i, block_bitmap);
1566 inode_bitmap = ext4_inode_bitmap(sb, gdp);
1567 if (inode_bitmap < first_block || inode_bitmap > last_block)
1569 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1570 "Inode bitmap for group %lu not in group "
1571 "(block %llu)!", i, inode_bitmap);
1574 inode_table = ext4_inode_table(sb, gdp);
1575 if (inode_table < first_block ||
1576 inode_table + sbi->s_itb_per_group - 1 > last_block)
1578 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1579 "Inode table for group %lu not in group "
1580 "(block %llu)!", i, inode_table);
1583 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1584 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1585 "Checksum for group %lu failed (%u!=%u)\n",
1586 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1587 gdp)), le16_to_cpu(gdp->bg_checksum));
1591 first_block += EXT4_BLOCKS_PER_GROUP(sb);
1594 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1595 sbi->s_es->s_free_inodes_count=cpu_to_le32(ext4_count_free_inodes(sb));
1599 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
1600 * the superblock) which were deleted from all directories, but held open by
1601 * a process at the time of a crash. We walk the list and try to delete these
1602 * inodes at recovery time (only with a read-write filesystem).
1604 * In order to keep the orphan inode chain consistent during traversal (in
1605 * case of crash during recovery), we link each inode into the superblock
1606 * orphan list_head and handle it the same way as an inode deletion during
1607 * normal operation (which journals the operations for us).
1609 * We only do an iget() and an iput() on each inode, which is very safe if we
1610 * accidentally point at an in-use or already deleted inode. The worst that
1611 * can happen in this case is that we get a "bit already cleared" message from
1612 * ext4_free_inode(). The only reason we would point at a wrong inode is if
1613 * e2fsck was run on this filesystem, and it must have already done the orphan
1614 * inode cleanup for us, so we can safely abort without any further action.
1616 static void ext4_orphan_cleanup (struct super_block * sb,
1617 struct ext4_super_block * es)
1619 unsigned int s_flags = sb->s_flags;
1620 int nr_orphans = 0, nr_truncates = 0;
1624 if (!es->s_last_orphan) {
1625 jbd_debug(4, "no orphan inodes to clean up\n");
1629 if (bdev_read_only(sb->s_bdev)) {
1630 printk(KERN_ERR "EXT4-fs: write access "
1631 "unavailable, skipping orphan cleanup.\n");
1635 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
1636 if (es->s_last_orphan)
1637 jbd_debug(1, "Errors on filesystem, "
1638 "clearing orphan list.\n");
1639 es->s_last_orphan = 0;
1640 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1644 if (s_flags & MS_RDONLY) {
1645 printk(KERN_INFO "EXT4-fs: %s: orphan cleanup on readonly fs\n",
1647 sb->s_flags &= ~MS_RDONLY;
1650 /* Needed for iput() to work correctly and not trash data */
1651 sb->s_flags |= MS_ACTIVE;
1652 /* Turn on quotas so that they are updated correctly */
1653 for (i = 0; i < MAXQUOTAS; i++) {
1654 if (EXT4_SB(sb)->s_qf_names[i]) {
1655 int ret = ext4_quota_on_mount(sb, i);
1658 "EXT4-fs: Cannot turn on journaled "
1659 "quota: error %d\n", ret);
1664 while (es->s_last_orphan) {
1665 struct inode *inode;
1667 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1668 if (IS_ERR(inode)) {
1669 es->s_last_orphan = 0;
1673 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
1675 if (inode->i_nlink) {
1677 "%s: truncating inode %lu to %Ld bytes\n",
1678 __func__, inode->i_ino, inode->i_size);
1679 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1680 inode->i_ino, inode->i_size);
1681 ext4_truncate(inode);
1685 "%s: deleting unreferenced inode %lu\n",
1686 __func__, inode->i_ino);
1687 jbd_debug(2, "deleting unreferenced inode %lu\n",
1691 iput(inode); /* The delete magic happens here! */
1694 #define PLURAL(x) (x), ((x)==1) ? "" : "s"
1697 printk(KERN_INFO "EXT4-fs: %s: %d orphan inode%s deleted\n",
1698 sb->s_id, PLURAL(nr_orphans));
1700 printk(KERN_INFO "EXT4-fs: %s: %d truncate%s cleaned up\n",
1701 sb->s_id, PLURAL(nr_truncates));
1703 /* Turn quotas off */
1704 for (i = 0; i < MAXQUOTAS; i++) {
1705 if (sb_dqopt(sb)->files[i])
1706 vfs_quota_off(sb, i, 0);
1709 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1712 * Maximal extent format file size.
1713 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1714 * extent format containers, within a sector_t, and within i_blocks
1715 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1716 * so that won't be a limiting factor.
1718 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1720 static loff_t ext4_max_size(int blkbits)
1723 loff_t upper_limit = MAX_LFS_FILESIZE;
1725 /* small i_blocks in vfs inode? */
1726 if (sizeof(blkcnt_t) < sizeof(u64)) {
1728 * CONFIG_LSF is not enabled implies the inode
1729 * i_block represent total blocks in 512 bytes
1730 * 32 == size of vfs inode i_blocks * 8
1732 upper_limit = (1LL << 32) - 1;
1734 /* total blocks in file system block size */
1735 upper_limit >>= (blkbits - 9);
1736 upper_limit <<= blkbits;
1739 /* 32-bit extent-start container, ee_block */
1744 /* Sanity check against vm- & vfs- imposed limits */
1745 if (res > upper_limit)
1752 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
1753 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1754 * We need to be 1 filesystem block less than the 2^48 sector limit.
1756 static loff_t ext4_max_bitmap_size(int bits)
1758 loff_t res = EXT4_NDIR_BLOCKS;
1761 /* This is calculated to be the largest file size for a
1762 * dense, bitmapped file such that the total number of
1763 * sectors in the file, including data and all indirect blocks,
1764 * does not exceed 2^48 -1
1765 * __u32 i_blocks_lo and _u16 i_blocks_high representing the
1766 * total number of 512 bytes blocks of the file
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 >>= (bits - 9);
1782 * We use 48 bit ext4_inode i_blocks
1783 * With EXT4_HUGE_FILE_FL set the i_blocks
1784 * represent total number of blocks in
1785 * file system block size
1787 upper_limit = (1LL << 48) - 1;
1791 /* indirect blocks */
1793 /* double indirect blocks */
1794 meta_blocks += 1 + (1LL << (bits-2));
1795 /* tripple indirect blocks */
1796 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1798 upper_limit -= meta_blocks;
1799 upper_limit <<= bits;
1801 res += 1LL << (bits-2);
1802 res += 1LL << (2*(bits-2));
1803 res += 1LL << (3*(bits-2));
1805 if (res > upper_limit)
1808 if (res > MAX_LFS_FILESIZE)
1809 res = MAX_LFS_FILESIZE;
1814 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
1815 ext4_fsblk_t logical_sb_block, int nr)
1817 struct ext4_sb_info *sbi = EXT4_SB(sb);
1818 ext4_group_t bg, first_meta_bg;
1821 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1823 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
1825 return logical_sb_block + nr + 1;
1826 bg = sbi->s_desc_per_block * nr;
1827 if (ext4_bg_has_super(sb, bg))
1829 return (has_super + ext4_group_first_block_no(sb, bg));
1833 * ext4_get_stripe_size: Get the stripe size.
1834 * @sbi: In memory super block info
1836 * If we have specified it via mount option, then
1837 * use the mount option value. If the value specified at mount time is
1838 * greater than the blocks per group use the super block value.
1839 * If the super block value is greater than blocks per group return 0.
1840 * Allocator needs it be less than blocks per group.
1843 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
1845 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
1846 unsigned long stripe_width =
1847 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
1849 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
1850 return sbi->s_stripe;
1852 if (stripe_width <= sbi->s_blocks_per_group)
1853 return stripe_width;
1855 if (stride <= sbi->s_blocks_per_group)
1861 static int ext4_fill_super (struct super_block *sb, void *data, int silent)
1862 __releases(kernel_lock)
1863 __acquires(kernel_lock)
1866 struct buffer_head * bh;
1867 struct ext4_super_block *es = NULL;
1868 struct ext4_sb_info *sbi;
1870 ext4_fsblk_t sb_block = get_sb_block(&data);
1871 ext4_fsblk_t logical_sb_block;
1872 unsigned long offset = 0;
1873 unsigned int journal_inum = 0;
1874 unsigned long journal_devnum = 0;
1875 unsigned long def_mount_opts;
1886 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1889 sb->s_fs_info = sbi;
1890 sbi->s_mount_opt = 0;
1891 sbi->s_resuid = EXT4_DEF_RESUID;
1892 sbi->s_resgid = EXT4_DEF_RESGID;
1893 sbi->s_sb_block = sb_block;
1897 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
1899 printk(KERN_ERR "EXT4-fs: unable to set blocksize\n");
1904 * The ext4 superblock will not be buffer aligned for other than 1kB
1905 * block sizes. We need to calculate the offset from buffer start.
1907 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
1908 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
1909 offset = do_div(logical_sb_block, blocksize);
1911 logical_sb_block = sb_block;
1914 if (!(bh = sb_bread(sb, logical_sb_block))) {
1915 printk (KERN_ERR "EXT4-fs: unable to read superblock\n");
1919 * Note: s_es must be initialized as soon as possible because
1920 * some ext4 macro-instructions depend on its value
1922 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
1924 sb->s_magic = le16_to_cpu(es->s_magic);
1925 if (sb->s_magic != EXT4_SUPER_MAGIC)
1928 /* Set defaults before we parse the mount options */
1929 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1930 if (def_mount_opts & EXT4_DEFM_DEBUG)
1931 set_opt(sbi->s_mount_opt, DEBUG);
1932 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
1933 set_opt(sbi->s_mount_opt, GRPID);
1934 if (def_mount_opts & EXT4_DEFM_UID16)
1935 set_opt(sbi->s_mount_opt, NO_UID32);
1936 #ifdef CONFIG_EXT4DEV_FS_XATTR
1937 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
1938 set_opt(sbi->s_mount_opt, XATTR_USER);
1940 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
1941 if (def_mount_opts & EXT4_DEFM_ACL)
1942 set_opt(sbi->s_mount_opt, POSIX_ACL);
1944 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
1945 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
1946 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
1947 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
1948 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
1949 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
1951 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
1952 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1953 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
1954 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1956 set_opt(sbi->s_mount_opt, ERRORS_RO);
1958 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1959 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1961 set_opt(sbi->s_mount_opt, RESERVATION);
1962 set_opt(sbi->s_mount_opt, BARRIER);
1965 * turn on extents feature by default in ext4 filesystem
1966 * User -o noextents to turn it off
1968 set_opt(sbi->s_mount_opt, EXTENTS);
1970 * turn on mballoc feature by default in ext4 filesystem
1971 * User -o nomballoc to turn it off
1973 set_opt(sbi->s_mount_opt, MBALLOC);
1975 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1979 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1980 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1982 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
1983 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
1984 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1985 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
1987 "EXT4-fs warning: feature flags set on rev 0 fs, "
1988 "running e2fsck is recommended\n");
1991 * Since ext4 is still considered development code, we require
1992 * that the TEST_FILESYS flag in s->flags be set.
1994 if (!(le32_to_cpu(es->s_flags) & EXT2_FLAGS_TEST_FILESYS)) {
1995 printk(KERN_WARNING "EXT4-fs: %s: not marked "
1996 "OK to use with test code.\n", sb->s_id);
2001 * Check feature flags regardless of the revision level, since we
2002 * previously didn't change the revision level when setting the flags,
2003 * so there is a chance incompat flags are set on a rev 0 filesystem.
2005 features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
2007 printk(KERN_ERR "EXT4-fs: %s: couldn't mount because of "
2008 "unsupported optional features (%x).\n",
2009 sb->s_id, le32_to_cpu(features));
2012 features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
2013 if (!(sb->s_flags & MS_RDONLY) && features) {
2014 printk(KERN_ERR "EXT4-fs: %s: couldn't mount RDWR because of "
2015 "unsupported optional features (%x).\n",
2016 sb->s_id, le32_to_cpu(features));
2019 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2021 * Large file size enabled file system can only be
2022 * mount if kernel is build with CONFIG_LSF
2024 if (sizeof(root->i_blocks) < sizeof(u64) &&
2025 !(sb->s_flags & MS_RDONLY)) {
2026 printk(KERN_ERR "EXT4-fs: %s: Filesystem with huge "
2027 "files cannot be mounted read-write "
2028 "without CONFIG_LSF.\n", sb->s_id);
2032 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2034 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2035 blocksize > EXT4_MAX_BLOCK_SIZE) {
2037 "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
2038 blocksize, sb->s_id);
2042 if (sb->s_blocksize != blocksize) {
2044 /* Validate the filesystem blocksize */
2045 if (!sb_set_blocksize(sb, blocksize)) {
2046 printk(KERN_ERR "EXT4-fs: bad block size %d.\n",
2052 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2053 offset = do_div(logical_sb_block, blocksize);
2054 bh = sb_bread(sb, logical_sb_block);
2057 "EXT4-fs: Can't read superblock on 2nd try.\n");
2060 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
2062 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2064 "EXT4-fs: Magic mismatch, very weird !\n");
2069 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits);
2070 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits);
2072 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2073 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2074 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
2076 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2077 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
2078 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
2079 (!is_power_of_2(sbi->s_inode_size)) ||
2080 (sbi->s_inode_size > blocksize)) {
2082 "EXT4-fs: unsupported inode size: %d\n",
2086 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2087 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
2089 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2090 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
2091 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
2092 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
2093 !is_power_of_2(sbi->s_desc_size)) {
2095 "EXT4-fs: unsupported descriptor size %lu\n",
2100 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
2101 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
2102 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
2103 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
2105 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
2106 if (sbi->s_inodes_per_block == 0)
2108 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2109 sbi->s_inodes_per_block;
2110 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
2112 sbi->s_mount_state = le16_to_cpu(es->s_state);
2113 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2114 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2115 for (i=0; i < 4; i++)
2116 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2117 sbi->s_def_hash_version = es->s_def_hash_version;
2119 if (sbi->s_blocks_per_group > blocksize * 8) {
2121 "EXT4-fs: #blocks per group too big: %lu\n",
2122 sbi->s_blocks_per_group);
2125 if (sbi->s_inodes_per_group > blocksize * 8) {
2127 "EXT4-fs: #inodes per group too big: %lu\n",
2128 sbi->s_inodes_per_group);
2132 if (ext4_blocks_count(es) >
2133 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
2134 printk(KERN_ERR "EXT4-fs: filesystem on %s:"
2135 " too large to mount safely\n", sb->s_id);
2136 if (sizeof(sector_t) < 8)
2137 printk(KERN_WARNING "EXT4-fs: CONFIG_LBD not "
2142 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2145 /* ensure blocks_count calculation below doesn't sign-extend */
2146 if (ext4_blocks_count(es) + EXT4_BLOCKS_PER_GROUP(sb) <
2147 le32_to_cpu(es->s_first_data_block) + 1) {
2148 printk(KERN_WARNING "EXT4-fs: bad geometry: block count %llu, "
2149 "first data block %u, blocks per group %lu\n",
2150 ext4_blocks_count(es),
2151 le32_to_cpu(es->s_first_data_block),
2152 EXT4_BLOCKS_PER_GROUP(sb));
2155 blocks_count = (ext4_blocks_count(es) -
2156 le32_to_cpu(es->s_first_data_block) +
2157 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2158 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2159 sbi->s_groups_count = blocks_count;
2160 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2161 EXT4_DESC_PER_BLOCK(sb);
2162 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
2164 if (sbi->s_group_desc == NULL) {
2165 printk (KERN_ERR "EXT4-fs: not enough memory\n");
2169 bgl_lock_init(&sbi->s_blockgroup_lock);
2171 for (i = 0; i < db_count; i++) {
2172 block = descriptor_loc(sb, logical_sb_block, i);
2173 sbi->s_group_desc[i] = sb_bread(sb, block);
2174 if (!sbi->s_group_desc[i]) {
2175 printk (KERN_ERR "EXT4-fs: "
2176 "can't read group descriptor %d\n", i);
2181 if (!ext4_check_descriptors (sb)) {
2182 printk(KERN_ERR "EXT4-fs: group descriptors corrupted!\n");
2185 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2186 if (!ext4_fill_flex_info(sb)) {
2188 "EXT4-fs: unable to initialize "
2189 "flex_bg meta info!\n");
2193 sbi->s_gdb_count = db_count;
2194 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2195 spin_lock_init(&sbi->s_next_gen_lock);
2197 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2198 ext4_count_free_blocks(sb));
2200 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2201 ext4_count_free_inodes(sb));
2204 err = percpu_counter_init(&sbi->s_dirs_counter,
2205 ext4_count_dirs(sb));
2208 printk(KERN_ERR "EXT4-fs: insufficient memory\n");
2212 /* per fileystem reservation list head & lock */
2213 spin_lock_init(&sbi->s_rsv_window_lock);
2214 sbi->s_rsv_window_root = RB_ROOT;
2215 /* Add a single, static dummy reservation to the start of the
2216 * reservation window list --- it gives us a placeholder for
2217 * append-at-start-of-list which makes the allocation logic
2218 * _much_ simpler. */
2219 sbi->s_rsv_window_head.rsv_start = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
2220 sbi->s_rsv_window_head.rsv_end = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
2221 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
2222 sbi->s_rsv_window_head.rsv_goal_size = 0;
2223 ext4_rsv_window_add(sb, &sbi->s_rsv_window_head);
2225 sbi->s_stripe = ext4_get_stripe_size(sbi);
2228 * set up enough so that it can read an inode
2230 sb->s_op = &ext4_sops;
2231 sb->s_export_op = &ext4_export_ops;
2232 sb->s_xattr = ext4_xattr_handlers;
2234 sb->s_qcop = &ext4_qctl_operations;
2235 sb->dq_op = &ext4_quota_operations;
2237 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2241 needs_recovery = (es->s_last_orphan != 0 ||
2242 EXT4_HAS_INCOMPAT_FEATURE(sb,
2243 EXT4_FEATURE_INCOMPAT_RECOVER));
2246 * The first inode we look at is the journal inode. Don't try
2247 * root first: it may be modified in the journal!
2249 if (!test_opt(sb, NOLOAD) &&
2250 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2251 if (ext4_load_journal(sb, es, journal_devnum))
2253 if (!(sb->s_flags & MS_RDONLY) &&
2254 EXT4_SB(sb)->s_journal->j_failed_commit) {
2255 printk(KERN_CRIT "EXT4-fs error (device %s): "
2256 "ext4_fill_super: Journal transaction "
2257 "%u is corrupt\n", sb->s_id,
2258 EXT4_SB(sb)->s_journal->j_failed_commit);
2259 if (test_opt (sb, ERRORS_RO)) {
2261 "Mounting filesystem read-only\n");
2262 sb->s_flags |= MS_RDONLY;
2263 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2264 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2266 if (test_opt(sb, ERRORS_PANIC)) {
2267 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2268 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2269 ext4_commit_super(sb, es, 1);
2271 "EXT4-fs (device %s): mount failed\n",
2276 } else if (journal_inum) {
2277 if (ext4_create_journal(sb, es, journal_inum))
2282 "ext4: No journal on filesystem on %s\n",
2287 if (ext4_blocks_count(es) > 0xffffffffULL &&
2288 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2289 JBD2_FEATURE_INCOMPAT_64BIT)) {
2290 printk(KERN_ERR "ext4: Failed to set 64-bit journal feature\n");
2294 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2295 jbd2_journal_set_features(sbi->s_journal,
2296 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2297 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2298 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2299 jbd2_journal_set_features(sbi->s_journal,
2300 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2301 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2302 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2304 jbd2_journal_clear_features(sbi->s_journal,
2305 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2306 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2309 /* We have now updated the journal if required, so we can
2310 * validate the data journaling mode. */
2311 switch (test_opt(sb, DATA_FLAGS)) {
2313 /* No mode set, assume a default based on the journal
2314 * capabilities: ORDERED_DATA if the journal can
2315 * cope, else JOURNAL_DATA
2317 if (jbd2_journal_check_available_features
2318 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
2319 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2321 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2324 case EXT4_MOUNT_ORDERED_DATA:
2325 case EXT4_MOUNT_WRITEBACK_DATA:
2326 if (!jbd2_journal_check_available_features
2327 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
2328 printk(KERN_ERR "EXT4-fs: Journal does not support "
2329 "requested data journaling mode\n");
2336 if (test_opt(sb, NOBH)) {
2337 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2338 printk(KERN_WARNING "EXT4-fs: Ignoring nobh option - "
2339 "its supported only with writeback mode\n");
2340 clear_opt(sbi->s_mount_opt, NOBH);
2344 * The jbd2_journal_load will have done any necessary log recovery,
2345 * so we can safely mount the rest of the filesystem now.
2348 root = ext4_iget(sb, EXT4_ROOT_INO);
2350 printk(KERN_ERR "EXT4-fs: get root inode failed\n");
2351 ret = PTR_ERR(root);
2354 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2356 printk(KERN_ERR "EXT4-fs: corrupt root inode, run e2fsck\n");
2359 sb->s_root = d_alloc_root(root);
2361 printk(KERN_ERR "EXT4-fs: get root dentry failed\n");
2367 ext4_setup_super (sb, es, sb->s_flags & MS_RDONLY);
2369 /* determine the minimum size of new large inodes, if present */
2370 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2371 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2372 EXT4_GOOD_OLD_INODE_SIZE;
2373 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2374 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2375 if (sbi->s_want_extra_isize <
2376 le16_to_cpu(es->s_want_extra_isize))
2377 sbi->s_want_extra_isize =
2378 le16_to_cpu(es->s_want_extra_isize);
2379 if (sbi->s_want_extra_isize <
2380 le16_to_cpu(es->s_min_extra_isize))
2381 sbi->s_want_extra_isize =
2382 le16_to_cpu(es->s_min_extra_isize);
2385 /* Check if enough inode space is available */
2386 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2387 sbi->s_inode_size) {
2388 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2389 EXT4_GOOD_OLD_INODE_SIZE;
2390 printk(KERN_INFO "EXT4-fs: required extra inode space not"
2395 * akpm: core read_super() calls in here with the superblock locked.
2396 * That deadlocks, because orphan cleanup needs to lock the superblock
2397 * in numerous places. Here we just pop the lock - it's relatively
2398 * harmless, because we are now ready to accept write_super() requests,
2399 * and aviro says that's the only reason for hanging onto the
2402 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2403 ext4_orphan_cleanup(sb, es);
2404 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
2406 printk (KERN_INFO "EXT4-fs: recovery complete.\n");
2407 ext4_mark_recovery_complete(sb, es);
2408 printk (KERN_INFO "EXT4-fs: mounted filesystem with %s data mode.\n",
2409 test_opt(sb,DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA ? "journal":
2410 test_opt(sb,DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA ? "ordered":
2414 ext4_mb_init(sb, needs_recovery);
2421 printk(KERN_ERR "VFS: Can't find ext4 filesystem on dev %s.\n",
2426 jbd2_journal_destroy(sbi->s_journal);
2427 sbi->s_journal = NULL;
2429 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2430 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2431 percpu_counter_destroy(&sbi->s_dirs_counter);
2433 for (i = 0; i < db_count; i++)
2434 brelse(sbi->s_group_desc[i]);
2435 kfree(sbi->s_group_desc);
2438 for (i = 0; i < MAXQUOTAS; i++)
2439 kfree(sbi->s_qf_names[i]);
2441 ext4_blkdev_remove(sbi);
2444 sb->s_fs_info = NULL;
2451 * Setup any per-fs journal parameters now. We'll do this both on
2452 * initial mount, once the journal has been initialised but before we've
2453 * done any recovery; and again on any subsequent remount.
2455 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
2457 struct ext4_sb_info *sbi = EXT4_SB(sb);
2459 if (sbi->s_commit_interval)
2460 journal->j_commit_interval = sbi->s_commit_interval;
2461 /* We could also set up an ext4-specific default for the commit
2462 * interval here, but for now we'll just fall back to the jbd
2465 spin_lock(&journal->j_state_lock);
2466 if (test_opt(sb, BARRIER))
2467 journal->j_flags |= JBD2_BARRIER;
2469 journal->j_flags &= ~JBD2_BARRIER;
2470 spin_unlock(&journal->j_state_lock);
2473 static journal_t *ext4_get_journal(struct super_block *sb,
2474 unsigned int journal_inum)
2476 struct inode *journal_inode;
2479 /* First, test for the existence of a valid inode on disk. Bad
2480 * things happen if we iget() an unused inode, as the subsequent
2481 * iput() will try to delete it. */
2483 journal_inode = ext4_iget(sb, journal_inum);
2484 if (IS_ERR(journal_inode)) {
2485 printk(KERN_ERR "EXT4-fs: no journal found.\n");
2488 if (!journal_inode->i_nlink) {
2489 make_bad_inode(journal_inode);
2490 iput(journal_inode);
2491 printk(KERN_ERR "EXT4-fs: journal inode is deleted.\n");
2495 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
2496 journal_inode, journal_inode->i_size);
2497 if (!S_ISREG(journal_inode->i_mode)) {
2498 printk(KERN_ERR "EXT4-fs: invalid journal inode.\n");
2499 iput(journal_inode);
2503 journal = jbd2_journal_init_inode(journal_inode);
2505 printk(KERN_ERR "EXT4-fs: Could not load journal inode\n");
2506 iput(journal_inode);
2509 journal->j_private = sb;
2510 ext4_init_journal_params(sb, journal);
2514 static journal_t *ext4_get_dev_journal(struct super_block *sb,
2517 struct buffer_head * bh;
2521 int hblock, blocksize;
2522 ext4_fsblk_t sb_block;
2523 unsigned long offset;
2524 struct ext4_super_block * es;
2525 struct block_device *bdev;
2527 bdev = ext4_blkdev_get(j_dev);
2531 if (bd_claim(bdev, sb)) {
2533 "EXT4: failed to claim external journal device.\n");
2538 blocksize = sb->s_blocksize;
2539 hblock = bdev_hardsect_size(bdev);
2540 if (blocksize < hblock) {
2542 "EXT4-fs: blocksize too small for journal device.\n");
2546 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
2547 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
2548 set_blocksize(bdev, blocksize);
2549 if (!(bh = __bread(bdev, sb_block, blocksize))) {
2550 printk(KERN_ERR "EXT4-fs: couldn't read superblock of "
2551 "external journal\n");
2555 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2556 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
2557 !(le32_to_cpu(es->s_feature_incompat) &
2558 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2559 printk(KERN_ERR "EXT4-fs: external journal has "
2560 "bad superblock\n");
2565 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2566 printk(KERN_ERR "EXT4-fs: journal UUID does not match\n");
2571 len = ext4_blocks_count(es);
2572 start = sb_block + 1;
2573 brelse(bh); /* we're done with the superblock */
2575 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
2576 start, len, blocksize);
2578 printk(KERN_ERR "EXT4-fs: failed to create device journal\n");
2581 journal->j_private = sb;
2582 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2583 wait_on_buffer(journal->j_sb_buffer);
2584 if (!buffer_uptodate(journal->j_sb_buffer)) {
2585 printk(KERN_ERR "EXT4-fs: I/O error on journal device\n");
2588 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
2589 printk(KERN_ERR "EXT4-fs: External journal has more than one "
2590 "user (unsupported) - %d\n",
2591 be32_to_cpu(journal->j_superblock->s_nr_users));
2594 EXT4_SB(sb)->journal_bdev = bdev;
2595 ext4_init_journal_params(sb, journal);
2598 jbd2_journal_destroy(journal);
2600 ext4_blkdev_put(bdev);
2604 static int ext4_load_journal(struct super_block *sb,
2605 struct ext4_super_block *es,
2606 unsigned long journal_devnum)
2609 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2612 int really_read_only;
2614 if (journal_devnum &&
2615 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2616 printk(KERN_INFO "EXT4-fs: external journal device major/minor "
2617 "numbers have changed\n");
2618 journal_dev = new_decode_dev(journal_devnum);
2620 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2622 really_read_only = bdev_read_only(sb->s_bdev);
2625 * Are we loading a blank journal or performing recovery after a
2626 * crash? For recovery, we need to check in advance whether we
2627 * can get read-write access to the device.
2630 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2631 if (sb->s_flags & MS_RDONLY) {
2632 printk(KERN_INFO "EXT4-fs: INFO: recovery "
2633 "required on readonly filesystem.\n");
2634 if (really_read_only) {
2635 printk(KERN_ERR "EXT4-fs: write access "
2636 "unavailable, cannot proceed.\n");
2639 printk (KERN_INFO "EXT4-fs: write access will "
2640 "be enabled during recovery.\n");
2644 if (journal_inum && journal_dev) {
2645 printk(KERN_ERR "EXT4-fs: filesystem has both journal "
2646 "and inode journals!\n");
2651 if (!(journal = ext4_get_journal(sb, journal_inum)))
2654 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
2658 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
2659 err = jbd2_journal_update_format(journal);
2661 printk(KERN_ERR "EXT4-fs: error updating journal.\n");
2662 jbd2_journal_destroy(journal);
2667 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
2668 err = jbd2_journal_wipe(journal, !really_read_only);
2670 err = jbd2_journal_load(journal);
2673 printk(KERN_ERR "EXT4-fs: error loading journal.\n");
2674 jbd2_journal_destroy(journal);
2678 EXT4_SB(sb)->s_journal = journal;
2679 ext4_clear_journal_err(sb, es);
2681 if (journal_devnum &&
2682 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2683 es->s_journal_dev = cpu_to_le32(journal_devnum);
2686 /* Make sure we flush the recovery flag to disk. */
2687 ext4_commit_super(sb, es, 1);
2693 static int ext4_create_journal(struct super_block * sb,
2694 struct ext4_super_block * es,
2695 unsigned int journal_inum)
2700 if (sb->s_flags & MS_RDONLY) {
2701 printk(KERN_ERR "EXT4-fs: readonly filesystem when trying to "
2702 "create journal.\n");
2706 journal = ext4_get_journal(sb, journal_inum);
2710 printk(KERN_INFO "EXT4-fs: creating new journal on inode %u\n",
2713 err = jbd2_journal_create(journal);
2715 printk(KERN_ERR "EXT4-fs: error creating journal.\n");
2716 jbd2_journal_destroy(journal);
2720 EXT4_SB(sb)->s_journal = journal;
2722 ext4_update_dynamic_rev(sb);
2723 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2724 EXT4_SET_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL);
2726 es->s_journal_inum = cpu_to_le32(journal_inum);
2729 /* Make sure we flush the recovery flag to disk. */
2730 ext4_commit_super(sb, es, 1);
2735 static void ext4_commit_super (struct super_block * sb,
2736 struct ext4_super_block * es,
2739 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
2743 es->s_wtime = cpu_to_le32(get_seconds());
2744 ext4_free_blocks_count_set(es, ext4_count_free_blocks(sb));
2745 es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
2746 BUFFER_TRACE(sbh, "marking dirty");
2747 mark_buffer_dirty(sbh);
2749 sync_dirty_buffer(sbh);
2754 * Have we just finished recovery? If so, and if we are mounting (or
2755 * remounting) the filesystem readonly, then we will end up with a
2756 * consistent fs on disk. Record that fact.
2758 static void ext4_mark_recovery_complete(struct super_block * sb,
2759 struct ext4_super_block * es)
2761 journal_t *journal = EXT4_SB(sb)->s_journal;
2763 jbd2_journal_lock_updates(journal);
2764 jbd2_journal_flush(journal);
2766 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
2767 sb->s_flags & MS_RDONLY) {
2768 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2770 ext4_commit_super(sb, es, 1);
2773 jbd2_journal_unlock_updates(journal);
2777 * If we are mounting (or read-write remounting) a filesystem whose journal
2778 * has recorded an error from a previous lifetime, move that error to the
2779 * main filesystem now.
2781 static void ext4_clear_journal_err(struct super_block * sb,
2782 struct ext4_super_block * es)
2788 journal = EXT4_SB(sb)->s_journal;
2791 * Now check for any error status which may have been recorded in the
2792 * journal by a prior ext4_error() or ext4_abort()
2795 j_errno = jbd2_journal_errno(journal);
2799 errstr = ext4_decode_error(sb, j_errno, nbuf);
2800 ext4_warning(sb, __func__, "Filesystem error recorded "
2801 "from previous mount: %s", errstr);
2802 ext4_warning(sb, __func__, "Marking fs in need of "
2803 "filesystem check.");
2805 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2806 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2807 ext4_commit_super (sb, es, 1);
2809 jbd2_journal_clear_err(journal);
2814 * Force the running and committing transactions to commit,
2815 * and wait on the commit.
2817 int ext4_force_commit(struct super_block *sb)
2822 if (sb->s_flags & MS_RDONLY)
2825 journal = EXT4_SB(sb)->s_journal;
2827 ret = ext4_journal_force_commit(journal);
2832 * Ext4 always journals updates to the superblock itself, so we don't
2833 * have to propagate any other updates to the superblock on disk at this
2834 * point. Just start an async writeback to get the buffers on their way
2837 * This implicitly triggers the writebehind on sync().
2840 static void ext4_write_super (struct super_block * sb)
2842 if (mutex_trylock(&sb->s_lock) != 0)
2847 static int ext4_sync_fs(struct super_block *sb, int wait)
2852 if (jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, &target)) {
2854 jbd2_log_wait_commit(EXT4_SB(sb)->s_journal, target);
2860 * LVM calls this function before a (read-only) snapshot is created. This
2861 * gives us a chance to flush the journal completely and mark the fs clean.
2863 static void ext4_write_super_lockfs(struct super_block *sb)
2867 if (!(sb->s_flags & MS_RDONLY)) {
2868 journal_t *journal = EXT4_SB(sb)->s_journal;
2870 /* Now we set up the journal barrier. */
2871 jbd2_journal_lock_updates(journal);
2872 jbd2_journal_flush(journal);
2874 /* Journal blocked and flushed, clear needs_recovery flag. */
2875 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2876 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
2881 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2882 * flag here, even though the filesystem is not technically dirty yet.
2884 static void ext4_unlockfs(struct super_block *sb)
2886 if (!(sb->s_flags & MS_RDONLY)) {
2888 /* Reser the needs_recovery flag before the fs is unlocked. */
2889 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2890 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
2892 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
2896 static int ext4_remount (struct super_block * sb, int * flags, char * data)
2898 struct ext4_super_block * es;
2899 struct ext4_sb_info *sbi = EXT4_SB(sb);
2900 ext4_fsblk_t n_blocks_count = 0;
2901 unsigned long old_sb_flags;
2902 struct ext4_mount_options old_opts;
2908 /* Store the original options */
2909 old_sb_flags = sb->s_flags;
2910 old_opts.s_mount_opt = sbi->s_mount_opt;
2911 old_opts.s_resuid = sbi->s_resuid;
2912 old_opts.s_resgid = sbi->s_resgid;
2913 old_opts.s_commit_interval = sbi->s_commit_interval;
2915 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2916 for (i = 0; i < MAXQUOTAS; i++)
2917 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2921 * Allow the "check" option to be passed as a remount option.
2923 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
2928 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT)
2929 ext4_abort(sb, __func__, "Abort forced by user");
2931 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2932 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
2936 ext4_init_journal_params(sb, sbi->s_journal);
2938 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
2939 n_blocks_count > ext4_blocks_count(es)) {
2940 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) {
2945 if (*flags & MS_RDONLY) {
2947 * First of all, the unconditional stuff we have to do
2948 * to disable replay of the journal when we next remount
2950 sb->s_flags |= MS_RDONLY;
2953 * OK, test if we are remounting a valid rw partition
2954 * readonly, and if so set the rdonly flag and then
2955 * mark the partition as valid again.
2957 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
2958 (sbi->s_mount_state & EXT4_VALID_FS))
2959 es->s_state = cpu_to_le16(sbi->s_mount_state);
2962 * We have to unlock super so that we can wait for
2966 ext4_mark_recovery_complete(sb, es);
2970 if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2971 ~EXT4_FEATURE_RO_COMPAT_SUPP))) {
2972 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
2973 "remount RDWR because of unsupported "
2974 "optional features (%x).\n",
2975 sb->s_id, le32_to_cpu(ret));
2981 * If we have an unprocessed orphan list hanging
2982 * around from a previously readonly bdev mount,
2983 * require a full umount/remount for now.
2985 if (es->s_last_orphan) {
2986 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
2987 "remount RDWR because of unprocessed "
2988 "orphan inode list. Please "
2989 "umount/remount instead.\n",
2996 * Mounting a RDONLY partition read-write, so reread
2997 * and store the current valid flag. (It may have
2998 * been changed by e2fsck since we originally mounted
3001 ext4_clear_journal_err(sb, es);
3002 sbi->s_mount_state = le16_to_cpu(es->s_state);
3003 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
3005 if (!ext4_setup_super (sb, es, 0))
3006 sb->s_flags &= ~MS_RDONLY;
3010 /* Release old quota file names */
3011 for (i = 0; i < MAXQUOTAS; i++)
3012 if (old_opts.s_qf_names[i] &&
3013 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3014 kfree(old_opts.s_qf_names[i]);
3018 sb->s_flags = old_sb_flags;
3019 sbi->s_mount_opt = old_opts.s_mount_opt;
3020 sbi->s_resuid = old_opts.s_resuid;
3021 sbi->s_resgid = old_opts.s_resgid;
3022 sbi->s_commit_interval = old_opts.s_commit_interval;
3024 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3025 for (i = 0; i < MAXQUOTAS; i++) {
3026 if (sbi->s_qf_names[i] &&
3027 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3028 kfree(sbi->s_qf_names[i]);
3029 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3035 static int ext4_statfs (struct dentry * dentry, struct kstatfs * buf)
3037 struct super_block *sb = dentry->d_sb;
3038 struct ext4_sb_info *sbi = EXT4_SB(sb);
3039 struct ext4_super_block *es = sbi->s_es;
3042 if (test_opt(sb, MINIX_DF)) {
3043 sbi->s_overhead_last = 0;
3044 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
3045 ext4_group_t ngroups = sbi->s_groups_count, i;
3046 ext4_fsblk_t overhead = 0;
3050 * Compute the overhead (FS structures). This is constant
3051 * for a given filesystem unless the number of block groups
3052 * changes so we cache the previous value until it does.
3056 * All of the blocks before first_data_block are
3059 overhead = le32_to_cpu(es->s_first_data_block);
3062 * Add the overhead attributed to the superblock and
3063 * block group descriptors. If the sparse superblocks
3064 * feature is turned on, then not all groups have this.
3066 for (i = 0; i < ngroups; i++) {
3067 overhead += ext4_bg_has_super(sb, i) +
3068 ext4_bg_num_gdb(sb, i);
3073 * Every block group has an inode bitmap, a block
3074 * bitmap, and an inode table.
3076 overhead += ngroups * (2 + sbi->s_itb_per_group);
3077 sbi->s_overhead_last = overhead;
3079 sbi->s_blocks_last = ext4_blocks_count(es);
3082 buf->f_type = EXT4_SUPER_MAGIC;
3083 buf->f_bsize = sb->s_blocksize;
3084 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
3085 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter);
3086 ext4_free_blocks_count_set(es, buf->f_bfree);
3087 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3088 if (buf->f_bfree < ext4_r_blocks_count(es))
3090 buf->f_files = le32_to_cpu(es->s_inodes_count);
3091 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
3092 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
3093 buf->f_namelen = EXT4_NAME_LEN;
3094 fsid = le64_to_cpup((void *)es->s_uuid) ^
3095 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3096 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3097 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
3101 /* Helper function for writing quotas on sync - we need to start transaction before quota file
3102 * is locked for write. Otherwise the are possible deadlocks:
3103 * Process 1 Process 2
3104 * ext4_create() quota_sync()
3105 * jbd2_journal_start() write_dquot()
3106 * DQUOT_INIT() down(dqio_mutex)
3107 * down(dqio_mutex) jbd2_journal_start()
3113 static inline struct inode *dquot_to_inode(struct dquot *dquot)
3115 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3118 static int ext4_dquot_initialize(struct inode *inode, int type)
3123 /* We may create quota structure so we need to reserve enough blocks */
3124 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_INIT_BLOCKS(inode->i_sb));
3126 return PTR_ERR(handle);
3127 ret = dquot_initialize(inode, type);
3128 err = ext4_journal_stop(handle);
3134 static int ext4_dquot_drop(struct inode *inode)
3139 /* We may delete quota structure so we need to reserve enough blocks */
3140 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_DEL_BLOCKS(inode->i_sb));
3141 if (IS_ERR(handle)) {
3143 * We call dquot_drop() anyway to at least release references
3144 * to quota structures so that umount does not hang.
3147 return PTR_ERR(handle);
3149 ret = dquot_drop(inode);
3150 err = ext4_journal_stop(handle);
3156 static int ext4_write_dquot(struct dquot *dquot)
3160 struct inode *inode;
3162 inode = dquot_to_inode(dquot);
3163 handle = ext4_journal_start(inode,
3164 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
3166 return PTR_ERR(handle);
3167 ret = dquot_commit(dquot);
3168 err = ext4_journal_stop(handle);
3174 static int ext4_acquire_dquot(struct dquot *dquot)
3179 handle = ext4_journal_start(dquot_to_inode(dquot),
3180 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
3182 return PTR_ERR(handle);
3183 ret = dquot_acquire(dquot);
3184 err = ext4_journal_stop(handle);
3190 static int ext4_release_dquot(struct dquot *dquot)
3195 handle = ext4_journal_start(dquot_to_inode(dquot),
3196 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
3197 if (IS_ERR(handle)) {
3198 /* Release dquot anyway to avoid endless cycle in dqput() */
3199 dquot_release(dquot);
3200 return PTR_ERR(handle);
3202 ret = dquot_release(dquot);
3203 err = ext4_journal_stop(handle);
3209 static int ext4_mark_dquot_dirty(struct dquot *dquot)
3211 /* Are we journaling quotas? */
3212 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3213 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
3214 dquot_mark_dquot_dirty(dquot);
3215 return ext4_write_dquot(dquot);
3217 return dquot_mark_dquot_dirty(dquot);
3221 static int ext4_write_info(struct super_block *sb, int type)
3226 /* Data block + inode block */
3227 handle = ext4_journal_start(sb->s_root->d_inode, 2);
3229 return PTR_ERR(handle);
3230 ret = dquot_commit_info(sb, type);
3231 err = ext4_journal_stop(handle);
3238 * Turn on quotas during mount time - we need to find
3239 * the quota file and such...
3241 static int ext4_quota_on_mount(struct super_block *sb, int type)
3243 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3244 EXT4_SB(sb)->s_jquota_fmt, type);
3248 * Standard function to be called on quota_on
3250 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
3251 char *path, int remount)
3254 struct nameidata nd;
3256 if (!test_opt(sb, QUOTA))
3258 /* When remounting, no checks are needed and in fact, path is NULL */
3260 return vfs_quota_on(sb, type, format_id, path, remount);
3262 err = path_lookup(path, LOOKUP_FOLLOW, &nd);
3266 /* Quotafile not on the same filesystem? */
3267 if (nd.path.mnt->mnt_sb != sb) {
3271 /* Journaling quota? */
3272 if (EXT4_SB(sb)->s_qf_names[type]) {
3273 /* Quotafile not of fs root? */
3274 if (nd.path.dentry->d_parent->d_inode != sb->s_root->d_inode)
3276 "EXT4-fs: Quota file not on filesystem root. "
3277 "Journaled quota will not work.\n");
3281 * When we journal data on quota file, we have to flush journal to see
3282 * all updates to the file when we bypass pagecache...
3284 if (ext4_should_journal_data(nd.path.dentry->d_inode)) {
3286 * We don't need to lock updates but journal_flush() could
3287 * otherwise be livelocked...
3289 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
3290 jbd2_journal_flush(EXT4_SB(sb)->s_journal);
3291 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3295 return vfs_quota_on(sb, type, format_id, path, remount);
3298 /* Read data from quotafile - avoid pagecache and such because we cannot afford
3299 * acquiring the locks... As quota files are never truncated and quota code
3300 * itself serializes the operations (and noone else should touch the files)
3301 * we don't have to be afraid of races */
3302 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
3303 size_t len, loff_t off)
3305 struct inode *inode = sb_dqopt(sb)->files[type];
3306 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3308 int offset = off & (sb->s_blocksize - 1);
3311 struct buffer_head *bh;
3312 loff_t i_size = i_size_read(inode);
3316 if (off+len > i_size)
3319 while (toread > 0) {
3320 tocopy = sb->s_blocksize - offset < toread ?
3321 sb->s_blocksize - offset : toread;
3322 bh = ext4_bread(NULL, inode, blk, 0, &err);
3325 if (!bh) /* A hole? */
3326 memset(data, 0, tocopy);
3328 memcpy(data, bh->b_data+offset, tocopy);
3338 /* Write to quotafile (we know the transaction is already started and has
3339 * enough credits) */
3340 static ssize_t ext4_quota_write(struct super_block *sb, int type,
3341 const char *data, size_t len, loff_t off)
3343 struct inode *inode = sb_dqopt(sb)->files[type];
3344 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
3346 int offset = off & (sb->s_blocksize - 1);
3348 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
3349 size_t towrite = len;
3350 struct buffer_head *bh;
3351 handle_t *handle = journal_current_handle();
3354 printk(KERN_WARNING "EXT4-fs: Quota write (off=%Lu, len=%Lu)"
3355 " cancelled because transaction is not started.\n",
3356 (unsigned long long)off, (unsigned long long)len);
3359 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3360 while (towrite > 0) {
3361 tocopy = sb->s_blocksize - offset < towrite ?
3362 sb->s_blocksize - offset : towrite;
3363 bh = ext4_bread(handle, inode, blk, 1, &err);
3366 if (journal_quota) {
3367 err = ext4_journal_get_write_access(handle, bh);
3374 memcpy(bh->b_data+offset, data, tocopy);
3375 flush_dcache_page(bh->b_page);
3378 err = ext4_journal_dirty_metadata(handle, bh);
3380 /* Always do at least ordered writes for quotas */
3381 err = ext4_journal_dirty_data(handle, bh);
3382 mark_buffer_dirty(bh);
3393 if (len == towrite) {
3394 mutex_unlock(&inode->i_mutex);
3397 if (inode->i_size < off+len-towrite) {
3398 i_size_write(inode, off+len-towrite);
3399 EXT4_I(inode)->i_disksize = inode->i_size;
3401 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3402 ext4_mark_inode_dirty(handle, inode);
3403 mutex_unlock(&inode->i_mutex);
3404 return len - towrite;
3409 static int ext4_get_sb(struct file_system_type *fs_type,
3410 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3412 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
3415 static struct file_system_type ext4dev_fs_type = {
3416 .owner = THIS_MODULE,
3418 .get_sb = ext4_get_sb,
3419 .kill_sb = kill_block_super,
3420 .fs_flags = FS_REQUIRES_DEV,
3423 static int __init init_ext4_fs(void)
3427 err = init_ext4_mballoc();
3431 err = init_ext4_xattr();
3434 err = init_inodecache();
3437 err = register_filesystem(&ext4dev_fs_type);
3442 destroy_inodecache();
3446 exit_ext4_mballoc();
3450 static void __exit exit_ext4_fs(void)
3452 unregister_filesystem(&ext4dev_fs_type);
3453 destroy_inodecache();
3455 exit_ext4_mballoc();
3458 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
3459 MODULE_DESCRIPTION("Fourth Extended Filesystem with extents");
3460 MODULE_LICENSE("GPL");
3461 module_init(init_ext4_fs)
3462 module_exit(exit_ext4_fs)