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/vmalloc.h>
24 #include <linux/jbd2.h>
25 #include <linux/slab.h>
26 #include <linux/init.h>
27 #include <linux/blkdev.h>
28 #include <linux/parser.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/proc_fs.h>
38 #include <linux/ctype.h>
39 #include <linux/log2.h>
40 #include <linux/crc16.h>
41 #include <linux/cleancache.h>
42 #include <asm/uaccess.h>
44 #include <linux/kthread.h>
45 #include <linux/freezer.h>
48 #include "ext4_extents.h"
49 #include "ext4_jbd2.h"
54 #define CREATE_TRACE_POINTS
55 #include <trace/events/ext4.h>
57 static struct proc_dir_entry *ext4_proc_root;
58 static struct kset *ext4_kset;
59 static struct ext4_lazy_init *ext4_li_info;
60 static struct mutex ext4_li_mtx;
61 static struct ext4_features *ext4_feat;
63 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
64 unsigned long journal_devnum);
65 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
66 static int ext4_commit_super(struct super_block *sb, int sync);
67 static void ext4_mark_recovery_complete(struct super_block *sb,
68 struct ext4_super_block *es);
69 static void ext4_clear_journal_err(struct super_block *sb,
70 struct ext4_super_block *es);
71 static int ext4_sync_fs(struct super_block *sb, int wait);
72 static const char *ext4_decode_error(struct super_block *sb, int errno,
74 static int ext4_remount(struct super_block *sb, int *flags, char *data);
75 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
76 static int ext4_unfreeze(struct super_block *sb);
77 static int ext4_freeze(struct super_block *sb);
78 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
79 const char *dev_name, void *data);
80 static inline int ext2_feature_set_ok(struct super_block *sb);
81 static inline int ext3_feature_set_ok(struct super_block *sb);
82 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
83 static void ext4_destroy_lazyinit_thread(void);
84 static void ext4_unregister_li_request(struct super_block *sb);
85 static void ext4_clear_request_list(void);
87 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
88 static struct file_system_type ext2_fs_type = {
92 .kill_sb = kill_block_super,
93 .fs_flags = FS_REQUIRES_DEV,
95 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
97 #define IS_EXT2_SB(sb) (0)
101 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
102 static struct file_system_type ext3_fs_type = {
103 .owner = THIS_MODULE,
106 .kill_sb = kill_block_super,
107 .fs_flags = FS_REQUIRES_DEV,
109 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
111 #define IS_EXT3_SB(sb) (0)
114 static int ext4_verify_csum_type(struct super_block *sb,
115 struct ext4_super_block *es)
117 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
118 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
121 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
124 static __le32 ext4_superblock_csum(struct super_block *sb,
125 struct ext4_super_block *es)
127 struct ext4_sb_info *sbi = EXT4_SB(sb);
128 int offset = offsetof(struct ext4_super_block, s_checksum);
131 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
133 return cpu_to_le32(csum);
136 int ext4_superblock_csum_verify(struct super_block *sb,
137 struct ext4_super_block *es)
139 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
140 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
143 return es->s_checksum == ext4_superblock_csum(sb, es);
146 void ext4_superblock_csum_set(struct super_block *sb,
147 struct ext4_super_block *es)
149 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
150 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
153 es->s_checksum = ext4_superblock_csum(sb, es);
156 void *ext4_kvmalloc(size_t size, gfp_t flags)
160 ret = kmalloc(size, flags);
162 ret = __vmalloc(size, flags, PAGE_KERNEL);
166 void *ext4_kvzalloc(size_t size, gfp_t flags)
170 ret = kzalloc(size, flags);
172 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
176 void ext4_kvfree(void *ptr)
178 if (is_vmalloc_addr(ptr))
185 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
186 struct ext4_group_desc *bg)
188 return le32_to_cpu(bg->bg_block_bitmap_lo) |
189 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
190 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
193 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
194 struct ext4_group_desc *bg)
196 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
197 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
198 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
201 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
202 struct ext4_group_desc *bg)
204 return le32_to_cpu(bg->bg_inode_table_lo) |
205 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
206 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
209 __u32 ext4_free_group_clusters(struct super_block *sb,
210 struct ext4_group_desc *bg)
212 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
213 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
214 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
217 __u32 ext4_free_inodes_count(struct super_block *sb,
218 struct ext4_group_desc *bg)
220 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
221 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
222 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
225 __u32 ext4_used_dirs_count(struct super_block *sb,
226 struct ext4_group_desc *bg)
228 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
229 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
230 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
233 __u32 ext4_itable_unused_count(struct super_block *sb,
234 struct ext4_group_desc *bg)
236 return le16_to_cpu(bg->bg_itable_unused_lo) |
237 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
238 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
241 void ext4_block_bitmap_set(struct super_block *sb,
242 struct ext4_group_desc *bg, ext4_fsblk_t blk)
244 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
245 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
246 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
249 void ext4_inode_bitmap_set(struct super_block *sb,
250 struct ext4_group_desc *bg, ext4_fsblk_t blk)
252 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
253 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
254 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
257 void ext4_inode_table_set(struct super_block *sb,
258 struct ext4_group_desc *bg, ext4_fsblk_t blk)
260 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
261 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
262 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
265 void ext4_free_group_clusters_set(struct super_block *sb,
266 struct ext4_group_desc *bg, __u32 count)
268 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
269 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
270 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
273 void ext4_free_inodes_set(struct super_block *sb,
274 struct ext4_group_desc *bg, __u32 count)
276 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
277 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
278 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
281 void ext4_used_dirs_set(struct super_block *sb,
282 struct ext4_group_desc *bg, __u32 count)
284 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
285 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
286 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
289 void ext4_itable_unused_set(struct super_block *sb,
290 struct ext4_group_desc *bg, __u32 count)
292 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
293 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
294 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
298 /* Just increment the non-pointer handle value */
299 static handle_t *ext4_get_nojournal(void)
301 handle_t *handle = current->journal_info;
302 unsigned long ref_cnt = (unsigned long)handle;
304 BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT);
307 handle = (handle_t *)ref_cnt;
309 current->journal_info = handle;
314 /* Decrement the non-pointer handle value */
315 static void ext4_put_nojournal(handle_t *handle)
317 unsigned long ref_cnt = (unsigned long)handle;
319 BUG_ON(ref_cnt == 0);
322 handle = (handle_t *)ref_cnt;
324 current->journal_info = handle;
328 * Wrappers for jbd2_journal_start/end.
330 * The only special thing we need to do here is to make sure that all
331 * journal_end calls result in the superblock being marked dirty, so
332 * that sync() will call the filesystem's write_super callback if
335 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
339 trace_ext4_journal_start(sb, nblocks, _RET_IP_);
340 if (sb->s_flags & MS_RDONLY)
341 return ERR_PTR(-EROFS);
343 WARN_ON(sb->s_writers.frozen == SB_FREEZE_COMPLETE);
344 journal = EXT4_SB(sb)->s_journal;
346 return ext4_get_nojournal();
348 * Special case here: if the journal has aborted behind our
349 * backs (eg. EIO in the commit thread), then we still need to
350 * take the FS itself readonly cleanly.
352 if (is_journal_aborted(journal)) {
353 ext4_abort(sb, "Detected aborted journal");
354 return ERR_PTR(-EROFS);
356 return jbd2_journal_start(journal, nblocks);
360 * The only special thing we need to do here is to make sure that all
361 * jbd2_journal_stop calls result in the superblock being marked dirty, so
362 * that sync() will call the filesystem's write_super callback if
365 int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle)
367 struct super_block *sb;
371 if (!ext4_handle_valid(handle)) {
372 ext4_put_nojournal(handle);
375 sb = handle->h_transaction->t_journal->j_private;
377 rc = jbd2_journal_stop(handle);
382 __ext4_std_error(sb, where, line, err);
386 void ext4_journal_abort_handle(const char *caller, unsigned int line,
387 const char *err_fn, struct buffer_head *bh,
388 handle_t *handle, int err)
391 const char *errstr = ext4_decode_error(NULL, err, nbuf);
393 BUG_ON(!ext4_handle_valid(handle));
396 BUFFER_TRACE(bh, "abort");
401 if (is_handle_aborted(handle))
404 printk(KERN_ERR "EXT4-fs: %s:%d: aborting transaction: %s in %s\n",
405 caller, line, errstr, err_fn);
407 jbd2_journal_abort_handle(handle);
410 static void __save_error_info(struct super_block *sb, const char *func,
413 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
415 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
416 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
417 es->s_last_error_time = cpu_to_le32(get_seconds());
418 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
419 es->s_last_error_line = cpu_to_le32(line);
420 if (!es->s_first_error_time) {
421 es->s_first_error_time = es->s_last_error_time;
422 strncpy(es->s_first_error_func, func,
423 sizeof(es->s_first_error_func));
424 es->s_first_error_line = cpu_to_le32(line);
425 es->s_first_error_ino = es->s_last_error_ino;
426 es->s_first_error_block = es->s_last_error_block;
429 * Start the daily error reporting function if it hasn't been
432 if (!es->s_error_count)
433 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
434 le32_add_cpu(&es->s_error_count, 1);
437 static void save_error_info(struct super_block *sb, const char *func,
440 __save_error_info(sb, func, line);
441 ext4_commit_super(sb, 1);
445 * The del_gendisk() function uninitializes the disk-specific data
446 * structures, including the bdi structure, without telling anyone
447 * else. Once this happens, any attempt to call mark_buffer_dirty()
448 * (for example, by ext4_commit_super), will cause a kernel OOPS.
449 * This is a kludge to prevent these oops until we can put in a proper
450 * hook in del_gendisk() to inform the VFS and file system layers.
452 static int block_device_ejected(struct super_block *sb)
454 struct inode *bd_inode = sb->s_bdev->bd_inode;
455 struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info;
457 return bdi->dev == NULL;
460 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
462 struct super_block *sb = journal->j_private;
463 struct ext4_sb_info *sbi = EXT4_SB(sb);
464 int error = is_journal_aborted(journal);
465 struct ext4_journal_cb_entry *jce, *tmp;
467 spin_lock(&sbi->s_md_lock);
468 list_for_each_entry_safe(jce, tmp, &txn->t_private_list, jce_list) {
469 list_del_init(&jce->jce_list);
470 spin_unlock(&sbi->s_md_lock);
471 jce->jce_func(sb, jce, error);
472 spin_lock(&sbi->s_md_lock);
474 spin_unlock(&sbi->s_md_lock);
477 /* Deal with the reporting of failure conditions on a filesystem such as
478 * inconsistencies detected or read IO failures.
480 * On ext2, we can store the error state of the filesystem in the
481 * superblock. That is not possible on ext4, because we may have other
482 * write ordering constraints on the superblock which prevent us from
483 * writing it out straight away; and given that the journal is about to
484 * be aborted, we can't rely on the current, or future, transactions to
485 * write out the superblock safely.
487 * We'll just use the jbd2_journal_abort() error code to record an error in
488 * the journal instead. On recovery, the journal will complain about
489 * that error until we've noted it down and cleared it.
492 static void ext4_handle_error(struct super_block *sb)
494 if (sb->s_flags & MS_RDONLY)
497 if (!test_opt(sb, ERRORS_CONT)) {
498 journal_t *journal = EXT4_SB(sb)->s_journal;
500 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
502 jbd2_journal_abort(journal, -EIO);
504 if (test_opt(sb, ERRORS_RO)) {
505 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
506 sb->s_flags |= MS_RDONLY;
508 if (test_opt(sb, ERRORS_PANIC))
509 panic("EXT4-fs (device %s): panic forced after error\n",
513 void __ext4_error(struct super_block *sb, const char *function,
514 unsigned int line, const char *fmt, ...)
516 struct va_format vaf;
522 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
523 sb->s_id, function, line, current->comm, &vaf);
525 save_error_info(sb, function, line);
527 ext4_handle_error(sb);
530 void ext4_error_inode(struct inode *inode, const char *function,
531 unsigned int line, ext4_fsblk_t block,
532 const char *fmt, ...)
535 struct va_format vaf;
536 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
538 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
539 es->s_last_error_block = cpu_to_le64(block);
540 save_error_info(inode->i_sb, function, line);
545 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
546 "inode #%lu: block %llu: comm %s: %pV\n",
547 inode->i_sb->s_id, function, line, inode->i_ino,
548 block, current->comm, &vaf);
550 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
551 "inode #%lu: comm %s: %pV\n",
552 inode->i_sb->s_id, function, line, inode->i_ino,
553 current->comm, &vaf);
556 ext4_handle_error(inode->i_sb);
559 void ext4_error_file(struct file *file, const char *function,
560 unsigned int line, ext4_fsblk_t block,
561 const char *fmt, ...)
564 struct va_format vaf;
565 struct ext4_super_block *es;
566 struct inode *inode = file->f_dentry->d_inode;
567 char pathname[80], *path;
569 es = EXT4_SB(inode->i_sb)->s_es;
570 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
571 save_error_info(inode->i_sb, function, line);
572 path = d_path(&(file->f_path), pathname, sizeof(pathname));
580 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
581 "block %llu: comm %s: path %s: %pV\n",
582 inode->i_sb->s_id, function, line, inode->i_ino,
583 block, current->comm, path, &vaf);
586 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
587 "comm %s: path %s: %pV\n",
588 inode->i_sb->s_id, function, line, inode->i_ino,
589 current->comm, path, &vaf);
592 ext4_handle_error(inode->i_sb);
595 static const char *ext4_decode_error(struct super_block *sb, int errno,
602 errstr = "IO failure";
605 errstr = "Out of memory";
608 if (!sb || (EXT4_SB(sb)->s_journal &&
609 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
610 errstr = "Journal has aborted";
612 errstr = "Readonly filesystem";
615 /* If the caller passed in an extra buffer for unknown
616 * errors, textualise them now. Else we just return
619 /* Check for truncated error codes... */
620 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
629 /* __ext4_std_error decodes expected errors from journaling functions
630 * automatically and invokes the appropriate error response. */
632 void __ext4_std_error(struct super_block *sb, const char *function,
633 unsigned int line, int errno)
638 /* Special case: if the error is EROFS, and we're not already
639 * inside a transaction, then there's really no point in logging
641 if (errno == -EROFS && journal_current_handle() == NULL &&
642 (sb->s_flags & MS_RDONLY))
645 errstr = ext4_decode_error(sb, errno, nbuf);
646 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
647 sb->s_id, function, line, errstr);
648 save_error_info(sb, function, line);
650 ext4_handle_error(sb);
654 * ext4_abort is a much stronger failure handler than ext4_error. The
655 * abort function may be used to deal with unrecoverable failures such
656 * as journal IO errors or ENOMEM at a critical moment in log management.
658 * We unconditionally force the filesystem into an ABORT|READONLY state,
659 * unless the error response on the fs has been set to panic in which
660 * case we take the easy way out and panic immediately.
663 void __ext4_abort(struct super_block *sb, const char *function,
664 unsigned int line, const char *fmt, ...)
668 save_error_info(sb, function, line);
670 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
676 if ((sb->s_flags & MS_RDONLY) == 0) {
677 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
678 sb->s_flags |= MS_RDONLY;
679 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
680 if (EXT4_SB(sb)->s_journal)
681 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
682 save_error_info(sb, function, line);
684 if (test_opt(sb, ERRORS_PANIC))
685 panic("EXT4-fs panic from previous error\n");
688 void ext4_msg(struct super_block *sb, const char *prefix, const char *fmt, ...)
690 struct va_format vaf;
696 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
700 void __ext4_warning(struct super_block *sb, const char *function,
701 unsigned int line, const char *fmt, ...)
703 struct va_format vaf;
709 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
710 sb->s_id, function, line, &vaf);
714 void __ext4_grp_locked_error(const char *function, unsigned int line,
715 struct super_block *sb, ext4_group_t grp,
716 unsigned long ino, ext4_fsblk_t block,
717 const char *fmt, ...)
721 struct va_format vaf;
723 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
725 es->s_last_error_ino = cpu_to_le32(ino);
726 es->s_last_error_block = cpu_to_le64(block);
727 __save_error_info(sb, function, line);
733 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
734 sb->s_id, function, line, grp);
736 printk(KERN_CONT "inode %lu: ", ino);
738 printk(KERN_CONT "block %llu:", (unsigned long long) block);
739 printk(KERN_CONT "%pV\n", &vaf);
742 if (test_opt(sb, ERRORS_CONT)) {
743 ext4_commit_super(sb, 0);
747 ext4_unlock_group(sb, grp);
748 ext4_handle_error(sb);
750 * We only get here in the ERRORS_RO case; relocking the group
751 * may be dangerous, but nothing bad will happen since the
752 * filesystem will have already been marked read/only and the
753 * journal has been aborted. We return 1 as a hint to callers
754 * who might what to use the return value from
755 * ext4_grp_locked_error() to distinguish between the
756 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
757 * aggressively from the ext4 function in question, with a
758 * more appropriate error code.
760 ext4_lock_group(sb, grp);
764 void ext4_update_dynamic_rev(struct super_block *sb)
766 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
768 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
772 "updating to rev %d because of new feature flag, "
773 "running e2fsck is recommended",
776 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
777 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
778 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
779 /* leave es->s_feature_*compat flags alone */
780 /* es->s_uuid will be set by e2fsck if empty */
783 * The rest of the superblock fields should be zero, and if not it
784 * means they are likely already in use, so leave them alone. We
785 * can leave it up to e2fsck to clean up any inconsistencies there.
790 * Open the external journal device
792 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
794 struct block_device *bdev;
795 char b[BDEVNAME_SIZE];
797 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
803 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
804 __bdevname(dev, b), PTR_ERR(bdev));
809 * Release the journal device
811 static int ext4_blkdev_put(struct block_device *bdev)
813 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
816 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
818 struct block_device *bdev;
821 bdev = sbi->journal_bdev;
823 ret = ext4_blkdev_put(bdev);
824 sbi->journal_bdev = NULL;
829 static inline struct inode *orphan_list_entry(struct list_head *l)
831 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
834 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
838 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
839 le32_to_cpu(sbi->s_es->s_last_orphan));
841 printk(KERN_ERR "sb_info orphan list:\n");
842 list_for_each(l, &sbi->s_orphan) {
843 struct inode *inode = orphan_list_entry(l);
845 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
846 inode->i_sb->s_id, inode->i_ino, inode,
847 inode->i_mode, inode->i_nlink,
852 static void ext4_put_super(struct super_block *sb)
854 struct ext4_sb_info *sbi = EXT4_SB(sb);
855 struct ext4_super_block *es = sbi->s_es;
858 ext4_unregister_li_request(sb);
859 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
861 flush_workqueue(sbi->dio_unwritten_wq);
862 destroy_workqueue(sbi->dio_unwritten_wq);
864 if (sbi->s_journal) {
865 err = jbd2_journal_destroy(sbi->s_journal);
866 sbi->s_journal = NULL;
868 ext4_abort(sb, "Couldn't clean up the journal");
871 del_timer(&sbi->s_err_report);
872 ext4_release_system_zone(sb);
874 ext4_ext_release(sb);
875 ext4_xattr_put_super(sb);
877 if (!(sb->s_flags & MS_RDONLY)) {
878 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
879 es->s_state = cpu_to_le16(sbi->s_mount_state);
881 if (!(sb->s_flags & MS_RDONLY))
882 ext4_commit_super(sb, 1);
885 remove_proc_entry("options", sbi->s_proc);
886 remove_proc_entry(sb->s_id, ext4_proc_root);
888 kobject_del(&sbi->s_kobj);
890 for (i = 0; i < sbi->s_gdb_count; i++)
891 brelse(sbi->s_group_desc[i]);
892 ext4_kvfree(sbi->s_group_desc);
893 ext4_kvfree(sbi->s_flex_groups);
894 percpu_counter_destroy(&sbi->s_freeclusters_counter);
895 percpu_counter_destroy(&sbi->s_freeinodes_counter);
896 percpu_counter_destroy(&sbi->s_dirs_counter);
897 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
900 for (i = 0; i < MAXQUOTAS; i++)
901 kfree(sbi->s_qf_names[i]);
904 /* Debugging code just in case the in-memory inode orphan list
905 * isn't empty. The on-disk one can be non-empty if we've
906 * detected an error and taken the fs readonly, but the
907 * in-memory list had better be clean by this point. */
908 if (!list_empty(&sbi->s_orphan))
909 dump_orphan_list(sb, sbi);
910 J_ASSERT(list_empty(&sbi->s_orphan));
912 invalidate_bdev(sb->s_bdev);
913 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
915 * Invalidate the journal device's buffers. We don't want them
916 * floating about in memory - the physical journal device may
917 * hotswapped, and it breaks the `ro-after' testing code.
919 sync_blockdev(sbi->journal_bdev);
920 invalidate_bdev(sbi->journal_bdev);
921 ext4_blkdev_remove(sbi);
924 kthread_stop(sbi->s_mmp_tsk);
925 sb->s_fs_info = NULL;
927 * Now that we are completely done shutting down the
928 * superblock, we need to actually destroy the kobject.
930 kobject_put(&sbi->s_kobj);
931 wait_for_completion(&sbi->s_kobj_unregister);
932 if (sbi->s_chksum_driver)
933 crypto_free_shash(sbi->s_chksum_driver);
934 kfree(sbi->s_blockgroup_lock);
938 static struct kmem_cache *ext4_inode_cachep;
941 * Called inside transaction, so use GFP_NOFS
943 static struct inode *ext4_alloc_inode(struct super_block *sb)
945 struct ext4_inode_info *ei;
947 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
951 ei->vfs_inode.i_version = 1;
952 ei->vfs_inode.i_data.writeback_index = 0;
953 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
954 INIT_LIST_HEAD(&ei->i_prealloc_list);
955 spin_lock_init(&ei->i_prealloc_lock);
956 ei->i_reserved_data_blocks = 0;
957 ei->i_reserved_meta_blocks = 0;
958 ei->i_allocated_meta_blocks = 0;
959 ei->i_da_metadata_calc_len = 0;
960 ei->i_da_metadata_calc_last_lblock = 0;
961 spin_lock_init(&(ei->i_block_reservation_lock));
963 ei->i_reserved_quota = 0;
966 INIT_LIST_HEAD(&ei->i_completed_io_list);
967 spin_lock_init(&ei->i_completed_io_lock);
969 ei->i_datasync_tid = 0;
970 atomic_set(&ei->i_ioend_count, 0);
971 atomic_set(&ei->i_aiodio_unwritten, 0);
973 return &ei->vfs_inode;
976 static int ext4_drop_inode(struct inode *inode)
978 int drop = generic_drop_inode(inode);
980 trace_ext4_drop_inode(inode, drop);
984 static void ext4_i_callback(struct rcu_head *head)
986 struct inode *inode = container_of(head, struct inode, i_rcu);
987 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
990 static void ext4_destroy_inode(struct inode *inode)
992 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
993 ext4_msg(inode->i_sb, KERN_ERR,
994 "Inode %lu (%p): orphan list check failed!",
995 inode->i_ino, EXT4_I(inode));
996 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
997 EXT4_I(inode), sizeof(struct ext4_inode_info),
1001 call_rcu(&inode->i_rcu, ext4_i_callback);
1004 static void init_once(void *foo)
1006 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
1008 INIT_LIST_HEAD(&ei->i_orphan);
1009 #ifdef CONFIG_EXT4_FS_XATTR
1010 init_rwsem(&ei->xattr_sem);
1012 init_rwsem(&ei->i_data_sem);
1013 inode_init_once(&ei->vfs_inode);
1016 static int init_inodecache(void)
1018 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
1019 sizeof(struct ext4_inode_info),
1020 0, (SLAB_RECLAIM_ACCOUNT|
1023 if (ext4_inode_cachep == NULL)
1028 static void destroy_inodecache(void)
1030 kmem_cache_destroy(ext4_inode_cachep);
1033 void ext4_clear_inode(struct inode *inode)
1035 invalidate_inode_buffers(inode);
1038 ext4_discard_preallocations(inode);
1039 if (EXT4_I(inode)->jinode) {
1040 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1041 EXT4_I(inode)->jinode);
1042 jbd2_free_inode(EXT4_I(inode)->jinode);
1043 EXT4_I(inode)->jinode = NULL;
1047 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1048 u64 ino, u32 generation)
1050 struct inode *inode;
1052 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1053 return ERR_PTR(-ESTALE);
1054 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1055 return ERR_PTR(-ESTALE);
1057 /* iget isn't really right if the inode is currently unallocated!!
1059 * ext4_read_inode will return a bad_inode if the inode had been
1060 * deleted, so we should be safe.
1062 * Currently we don't know the generation for parent directory, so
1063 * a generation of 0 means "accept any"
1065 inode = ext4_iget(sb, ino);
1067 return ERR_CAST(inode);
1068 if (generation && inode->i_generation != generation) {
1070 return ERR_PTR(-ESTALE);
1076 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1077 int fh_len, int fh_type)
1079 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1080 ext4_nfs_get_inode);
1083 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1084 int fh_len, int fh_type)
1086 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1087 ext4_nfs_get_inode);
1091 * Try to release metadata pages (indirect blocks, directories) which are
1092 * mapped via the block device. Since these pages could have journal heads
1093 * which would prevent try_to_free_buffers() from freeing them, we must use
1094 * jbd2 layer's try_to_free_buffers() function to release them.
1096 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1099 journal_t *journal = EXT4_SB(sb)->s_journal;
1101 WARN_ON(PageChecked(page));
1102 if (!page_has_buffers(page))
1105 return jbd2_journal_try_to_free_buffers(journal, page,
1106 wait & ~__GFP_WAIT);
1107 return try_to_free_buffers(page);
1111 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1112 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1114 static int ext4_write_dquot(struct dquot *dquot);
1115 static int ext4_acquire_dquot(struct dquot *dquot);
1116 static int ext4_release_dquot(struct dquot *dquot);
1117 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1118 static int ext4_write_info(struct super_block *sb, int type);
1119 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1121 static int ext4_quota_on_sysfile(struct super_block *sb, int type,
1123 static int ext4_quota_off(struct super_block *sb, int type);
1124 static int ext4_quota_off_sysfile(struct super_block *sb, int type);
1125 static int ext4_quota_on_mount(struct super_block *sb, int type);
1126 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1127 size_t len, loff_t off);
1128 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1129 const char *data, size_t len, loff_t off);
1130 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1131 unsigned int flags);
1132 static int ext4_enable_quotas(struct super_block *sb);
1134 static const struct dquot_operations ext4_quota_operations = {
1135 .get_reserved_space = ext4_get_reserved_space,
1136 .write_dquot = ext4_write_dquot,
1137 .acquire_dquot = ext4_acquire_dquot,
1138 .release_dquot = ext4_release_dquot,
1139 .mark_dirty = ext4_mark_dquot_dirty,
1140 .write_info = ext4_write_info,
1141 .alloc_dquot = dquot_alloc,
1142 .destroy_dquot = dquot_destroy,
1145 static const struct quotactl_ops ext4_qctl_operations = {
1146 .quota_on = ext4_quota_on,
1147 .quota_off = ext4_quota_off,
1148 .quota_sync = dquot_quota_sync,
1149 .get_info = dquot_get_dqinfo,
1150 .set_info = dquot_set_dqinfo,
1151 .get_dqblk = dquot_get_dqblk,
1152 .set_dqblk = dquot_set_dqblk
1155 static const struct quotactl_ops ext4_qctl_sysfile_operations = {
1156 .quota_on_meta = ext4_quota_on_sysfile,
1157 .quota_off = ext4_quota_off_sysfile,
1158 .quota_sync = dquot_quota_sync,
1159 .get_info = dquot_get_dqinfo,
1160 .set_info = dquot_set_dqinfo,
1161 .get_dqblk = dquot_get_dqblk,
1162 .set_dqblk = dquot_set_dqblk
1166 static const struct super_operations ext4_sops = {
1167 .alloc_inode = ext4_alloc_inode,
1168 .destroy_inode = ext4_destroy_inode,
1169 .write_inode = ext4_write_inode,
1170 .dirty_inode = ext4_dirty_inode,
1171 .drop_inode = ext4_drop_inode,
1172 .evict_inode = ext4_evict_inode,
1173 .put_super = ext4_put_super,
1174 .sync_fs = ext4_sync_fs,
1175 .freeze_fs = ext4_freeze,
1176 .unfreeze_fs = ext4_unfreeze,
1177 .statfs = ext4_statfs,
1178 .remount_fs = ext4_remount,
1179 .show_options = ext4_show_options,
1181 .quota_read = ext4_quota_read,
1182 .quota_write = ext4_quota_write,
1184 .bdev_try_to_free_page = bdev_try_to_free_page,
1187 static const struct super_operations ext4_nojournal_sops = {
1188 .alloc_inode = ext4_alloc_inode,
1189 .destroy_inode = ext4_destroy_inode,
1190 .write_inode = ext4_write_inode,
1191 .dirty_inode = ext4_dirty_inode,
1192 .drop_inode = ext4_drop_inode,
1193 .evict_inode = ext4_evict_inode,
1194 .put_super = ext4_put_super,
1195 .statfs = ext4_statfs,
1196 .remount_fs = ext4_remount,
1197 .show_options = ext4_show_options,
1199 .quota_read = ext4_quota_read,
1200 .quota_write = ext4_quota_write,
1202 .bdev_try_to_free_page = bdev_try_to_free_page,
1205 static const struct export_operations ext4_export_ops = {
1206 .fh_to_dentry = ext4_fh_to_dentry,
1207 .fh_to_parent = ext4_fh_to_parent,
1208 .get_parent = ext4_get_parent,
1212 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1213 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1214 Opt_nouid32, Opt_debug, Opt_removed,
1215 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1216 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1217 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1218 Opt_journal_dev, Opt_journal_checksum, Opt_journal_async_commit,
1219 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1220 Opt_data_err_abort, Opt_data_err_ignore,
1221 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1222 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1223 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1224 Opt_usrquota, Opt_grpquota, Opt_i_version,
1225 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1226 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1227 Opt_inode_readahead_blks, Opt_journal_ioprio,
1228 Opt_dioread_nolock, Opt_dioread_lock,
1229 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1230 Opt_max_dir_size_kb,
1233 static const match_table_t tokens = {
1234 {Opt_bsd_df, "bsddf"},
1235 {Opt_minix_df, "minixdf"},
1236 {Opt_grpid, "grpid"},
1237 {Opt_grpid, "bsdgroups"},
1238 {Opt_nogrpid, "nogrpid"},
1239 {Opt_nogrpid, "sysvgroups"},
1240 {Opt_resgid, "resgid=%u"},
1241 {Opt_resuid, "resuid=%u"},
1243 {Opt_err_cont, "errors=continue"},
1244 {Opt_err_panic, "errors=panic"},
1245 {Opt_err_ro, "errors=remount-ro"},
1246 {Opt_nouid32, "nouid32"},
1247 {Opt_debug, "debug"},
1248 {Opt_removed, "oldalloc"},
1249 {Opt_removed, "orlov"},
1250 {Opt_user_xattr, "user_xattr"},
1251 {Opt_nouser_xattr, "nouser_xattr"},
1253 {Opt_noacl, "noacl"},
1254 {Opt_noload, "norecovery"},
1255 {Opt_noload, "noload"},
1256 {Opt_removed, "nobh"},
1257 {Opt_removed, "bh"},
1258 {Opt_commit, "commit=%u"},
1259 {Opt_min_batch_time, "min_batch_time=%u"},
1260 {Opt_max_batch_time, "max_batch_time=%u"},
1261 {Opt_journal_dev, "journal_dev=%u"},
1262 {Opt_journal_checksum, "journal_checksum"},
1263 {Opt_journal_async_commit, "journal_async_commit"},
1264 {Opt_abort, "abort"},
1265 {Opt_data_journal, "data=journal"},
1266 {Opt_data_ordered, "data=ordered"},
1267 {Opt_data_writeback, "data=writeback"},
1268 {Opt_data_err_abort, "data_err=abort"},
1269 {Opt_data_err_ignore, "data_err=ignore"},
1270 {Opt_offusrjquota, "usrjquota="},
1271 {Opt_usrjquota, "usrjquota=%s"},
1272 {Opt_offgrpjquota, "grpjquota="},
1273 {Opt_grpjquota, "grpjquota=%s"},
1274 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1275 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1276 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1277 {Opt_grpquota, "grpquota"},
1278 {Opt_noquota, "noquota"},
1279 {Opt_quota, "quota"},
1280 {Opt_usrquota, "usrquota"},
1281 {Opt_barrier, "barrier=%u"},
1282 {Opt_barrier, "barrier"},
1283 {Opt_nobarrier, "nobarrier"},
1284 {Opt_i_version, "i_version"},
1285 {Opt_stripe, "stripe=%u"},
1286 {Opt_delalloc, "delalloc"},
1287 {Opt_nodelalloc, "nodelalloc"},
1288 {Opt_mblk_io_submit, "mblk_io_submit"},
1289 {Opt_nomblk_io_submit, "nomblk_io_submit"},
1290 {Opt_block_validity, "block_validity"},
1291 {Opt_noblock_validity, "noblock_validity"},
1292 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1293 {Opt_journal_ioprio, "journal_ioprio=%u"},
1294 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1295 {Opt_auto_da_alloc, "auto_da_alloc"},
1296 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1297 {Opt_dioread_nolock, "dioread_nolock"},
1298 {Opt_dioread_lock, "dioread_lock"},
1299 {Opt_discard, "discard"},
1300 {Opt_nodiscard, "nodiscard"},
1301 {Opt_init_itable, "init_itable=%u"},
1302 {Opt_init_itable, "init_itable"},
1303 {Opt_noinit_itable, "noinit_itable"},
1304 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1305 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1306 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1307 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1308 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1309 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1313 static ext4_fsblk_t get_sb_block(void **data)
1315 ext4_fsblk_t sb_block;
1316 char *options = (char *) *data;
1318 if (!options || strncmp(options, "sb=", 3) != 0)
1319 return 1; /* Default location */
1322 /* TODO: use simple_strtoll with >32bit ext4 */
1323 sb_block = simple_strtoul(options, &options, 0);
1324 if (*options && *options != ',') {
1325 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1329 if (*options == ',')
1331 *data = (void *) options;
1336 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1337 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1338 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1341 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1343 struct ext4_sb_info *sbi = EXT4_SB(sb);
1346 if (sb_any_quota_loaded(sb) &&
1347 !sbi->s_qf_names[qtype]) {
1348 ext4_msg(sb, KERN_ERR,
1349 "Cannot change journaled "
1350 "quota options when quota turned on");
1353 qname = match_strdup(args);
1355 ext4_msg(sb, KERN_ERR,
1356 "Not enough memory for storing quotafile name");
1359 if (sbi->s_qf_names[qtype] &&
1360 strcmp(sbi->s_qf_names[qtype], qname)) {
1361 ext4_msg(sb, KERN_ERR,
1362 "%s quota file already specified", QTYPE2NAME(qtype));
1366 sbi->s_qf_names[qtype] = qname;
1367 if (strchr(sbi->s_qf_names[qtype], '/')) {
1368 ext4_msg(sb, KERN_ERR,
1369 "quotafile must be on filesystem root");
1370 kfree(sbi->s_qf_names[qtype]);
1371 sbi->s_qf_names[qtype] = NULL;
1378 static int clear_qf_name(struct super_block *sb, int qtype)
1381 struct ext4_sb_info *sbi = EXT4_SB(sb);
1383 if (sb_any_quota_loaded(sb) &&
1384 sbi->s_qf_names[qtype]) {
1385 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1386 " when quota turned on");
1390 * The space will be released later when all options are confirmed
1393 sbi->s_qf_names[qtype] = NULL;
1398 #define MOPT_SET 0x0001
1399 #define MOPT_CLEAR 0x0002
1400 #define MOPT_NOSUPPORT 0x0004
1401 #define MOPT_EXPLICIT 0x0008
1402 #define MOPT_CLEAR_ERR 0x0010
1403 #define MOPT_GTE0 0x0020
1406 #define MOPT_QFMT 0x0040
1408 #define MOPT_Q MOPT_NOSUPPORT
1409 #define MOPT_QFMT MOPT_NOSUPPORT
1411 #define MOPT_DATAJ 0x0080
1413 static const struct mount_opts {
1417 } ext4_mount_opts[] = {
1418 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1419 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1420 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1421 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1422 {Opt_mblk_io_submit, EXT4_MOUNT_MBLK_IO_SUBMIT, MOPT_SET},
1423 {Opt_nomblk_io_submit, EXT4_MOUNT_MBLK_IO_SUBMIT, MOPT_CLEAR},
1424 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1425 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1426 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK, MOPT_SET},
1427 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK, MOPT_CLEAR},
1428 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1429 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1430 {Opt_delalloc, EXT4_MOUNT_DELALLOC, MOPT_SET | MOPT_EXPLICIT},
1431 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC, MOPT_CLEAR | MOPT_EXPLICIT},
1432 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, MOPT_SET},
1433 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1434 EXT4_MOUNT_JOURNAL_CHECKSUM), MOPT_SET},
1435 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_SET},
1436 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1437 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1438 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1439 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_SET},
1440 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_CLEAR},
1441 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1442 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1443 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1444 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1445 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1446 {Opt_commit, 0, MOPT_GTE0},
1447 {Opt_max_batch_time, 0, MOPT_GTE0},
1448 {Opt_min_batch_time, 0, MOPT_GTE0},
1449 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1450 {Opt_init_itable, 0, MOPT_GTE0},
1451 {Opt_stripe, 0, MOPT_GTE0},
1452 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_DATAJ},
1453 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_DATAJ},
1454 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA, MOPT_DATAJ},
1455 #ifdef CONFIG_EXT4_FS_XATTR
1456 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1457 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1459 {Opt_user_xattr, 0, MOPT_NOSUPPORT},
1460 {Opt_nouser_xattr, 0, MOPT_NOSUPPORT},
1462 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1463 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1464 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1466 {Opt_acl, 0, MOPT_NOSUPPORT},
1467 {Opt_noacl, 0, MOPT_NOSUPPORT},
1469 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1470 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1471 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1472 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1474 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1476 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1477 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1478 {Opt_usrjquota, 0, MOPT_Q},
1479 {Opt_grpjquota, 0, MOPT_Q},
1480 {Opt_offusrjquota, 0, MOPT_Q},
1481 {Opt_offgrpjquota, 0, MOPT_Q},
1482 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1483 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1484 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1485 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1489 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1490 substring_t *args, unsigned long *journal_devnum,
1491 unsigned int *journal_ioprio, int is_remount)
1493 struct ext4_sb_info *sbi = EXT4_SB(sb);
1494 const struct mount_opts *m;
1500 if (token == Opt_usrjquota)
1501 return set_qf_name(sb, USRQUOTA, &args[0]);
1502 else if (token == Opt_grpjquota)
1503 return set_qf_name(sb, GRPQUOTA, &args[0]);
1504 else if (token == Opt_offusrjquota)
1505 return clear_qf_name(sb, USRQUOTA);
1506 else if (token == Opt_offgrpjquota)
1507 return clear_qf_name(sb, GRPQUOTA);
1509 if (args->from && match_int(args, &arg))
1513 case Opt_nouser_xattr:
1514 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1517 return 1; /* handled by get_sb_block() */
1519 ext4_msg(sb, KERN_WARNING,
1520 "Ignoring removed %s option", opt);
1523 uid = make_kuid(current_user_ns(), arg);
1524 if (!uid_valid(uid)) {
1525 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1528 sbi->s_resuid = uid;
1531 gid = make_kgid(current_user_ns(), arg);
1532 if (!gid_valid(gid)) {
1533 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1536 sbi->s_resgid = gid;
1539 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1542 sb->s_flags |= MS_I_VERSION;
1544 case Opt_journal_dev:
1546 ext4_msg(sb, KERN_ERR,
1547 "Cannot specify journal on remount");
1550 *journal_devnum = arg;
1552 case Opt_journal_ioprio:
1553 if (arg < 0 || arg > 7)
1555 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1559 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1560 if (token != m->token)
1562 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1564 if (m->flags & MOPT_EXPLICIT)
1565 set_opt2(sb, EXPLICIT_DELALLOC);
1566 if (m->flags & MOPT_CLEAR_ERR)
1567 clear_opt(sb, ERRORS_MASK);
1568 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1569 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1570 "options when quota turned on");
1574 if (m->flags & MOPT_NOSUPPORT) {
1575 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1576 } else if (token == Opt_commit) {
1578 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1579 sbi->s_commit_interval = HZ * arg;
1580 } else if (token == Opt_max_batch_time) {
1582 arg = EXT4_DEF_MAX_BATCH_TIME;
1583 sbi->s_max_batch_time = arg;
1584 } else if (token == Opt_min_batch_time) {
1585 sbi->s_min_batch_time = arg;
1586 } else if (token == Opt_inode_readahead_blks) {
1587 if (arg > (1 << 30))
1589 if (arg && !is_power_of_2(arg)) {
1590 ext4_msg(sb, KERN_ERR,
1591 "EXT4-fs: inode_readahead_blks"
1592 " must be a power of 2");
1595 sbi->s_inode_readahead_blks = arg;
1596 } else if (token == Opt_init_itable) {
1597 set_opt(sb, INIT_INODE_TABLE);
1599 arg = EXT4_DEF_LI_WAIT_MULT;
1600 sbi->s_li_wait_mult = arg;
1601 } else if (token == Opt_max_dir_size_kb) {
1602 sbi->s_max_dir_size_kb = arg;
1603 } else if (token == Opt_stripe) {
1604 sbi->s_stripe = arg;
1605 } else if (m->flags & MOPT_DATAJ) {
1607 if (!sbi->s_journal)
1608 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1609 else if (test_opt(sb, DATA_FLAGS) !=
1611 ext4_msg(sb, KERN_ERR,
1612 "Cannot change data mode on remount");
1616 clear_opt(sb, DATA_FLAGS);
1617 sbi->s_mount_opt |= m->mount_opt;
1620 } else if (m->flags & MOPT_QFMT) {
1621 if (sb_any_quota_loaded(sb) &&
1622 sbi->s_jquota_fmt != m->mount_opt) {
1623 ext4_msg(sb, KERN_ERR, "Cannot "
1624 "change journaled quota options "
1625 "when quota turned on");
1628 sbi->s_jquota_fmt = m->mount_opt;
1633 if (m->flags & MOPT_CLEAR)
1635 else if (unlikely(!(m->flags & MOPT_SET))) {
1636 ext4_msg(sb, KERN_WARNING,
1637 "buggy handling of option %s", opt);
1642 sbi->s_mount_opt |= m->mount_opt;
1644 sbi->s_mount_opt &= ~m->mount_opt;
1648 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1649 "or missing value", opt);
1653 static int parse_options(char *options, struct super_block *sb,
1654 unsigned long *journal_devnum,
1655 unsigned int *journal_ioprio,
1659 struct ext4_sb_info *sbi = EXT4_SB(sb);
1662 substring_t args[MAX_OPT_ARGS];
1668 while ((p = strsep(&options, ",")) != NULL) {
1672 * Initialize args struct so we know whether arg was
1673 * found; some options take optional arguments.
1675 args[0].to = args[0].from = NULL;
1676 token = match_token(p, tokens, args);
1677 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1678 journal_ioprio, is_remount) < 0)
1682 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1683 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1684 clear_opt(sb, USRQUOTA);
1686 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1687 clear_opt(sb, GRPQUOTA);
1689 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1690 ext4_msg(sb, KERN_ERR, "old and new quota "
1695 if (!sbi->s_jquota_fmt) {
1696 ext4_msg(sb, KERN_ERR, "journaled quota format "
1701 if (sbi->s_jquota_fmt) {
1702 ext4_msg(sb, KERN_ERR, "journaled quota format "
1703 "specified with no journaling "
1712 static inline void ext4_show_quota_options(struct seq_file *seq,
1713 struct super_block *sb)
1715 #if defined(CONFIG_QUOTA)
1716 struct ext4_sb_info *sbi = EXT4_SB(sb);
1718 if (sbi->s_jquota_fmt) {
1721 switch (sbi->s_jquota_fmt) {
1732 seq_printf(seq, ",jqfmt=%s", fmtname);
1735 if (sbi->s_qf_names[USRQUOTA])
1736 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1738 if (sbi->s_qf_names[GRPQUOTA])
1739 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1741 if (test_opt(sb, USRQUOTA))
1742 seq_puts(seq, ",usrquota");
1744 if (test_opt(sb, GRPQUOTA))
1745 seq_puts(seq, ",grpquota");
1749 static const char *token2str(int token)
1751 const struct match_token *t;
1753 for (t = tokens; t->token != Opt_err; t++)
1754 if (t->token == token && !strchr(t->pattern, '='))
1761 * - it's set to a non-default value OR
1762 * - if the per-sb default is different from the global default
1764 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1767 struct ext4_sb_info *sbi = EXT4_SB(sb);
1768 struct ext4_super_block *es = sbi->s_es;
1769 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1770 const struct mount_opts *m;
1771 char sep = nodefs ? '\n' : ',';
1773 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1774 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1776 if (sbi->s_sb_block != 1)
1777 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1779 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1780 int want_set = m->flags & MOPT_SET;
1781 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1782 (m->flags & MOPT_CLEAR_ERR))
1784 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
1785 continue; /* skip if same as the default */
1787 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1788 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1789 continue; /* select Opt_noFoo vs Opt_Foo */
1790 SEQ_OPTS_PRINT("%s", token2str(m->token));
1793 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
1794 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1795 SEQ_OPTS_PRINT("resuid=%u",
1796 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1797 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
1798 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1799 SEQ_OPTS_PRINT("resgid=%u",
1800 from_kgid_munged(&init_user_ns, sbi->s_resgid));
1801 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
1802 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1803 SEQ_OPTS_PUTS("errors=remount-ro");
1804 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1805 SEQ_OPTS_PUTS("errors=continue");
1806 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1807 SEQ_OPTS_PUTS("errors=panic");
1808 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
1809 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
1810 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
1811 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
1812 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
1813 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
1814 if (sb->s_flags & MS_I_VERSION)
1815 SEQ_OPTS_PUTS("i_version");
1816 if (nodefs || sbi->s_stripe)
1817 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
1818 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
1819 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1820 SEQ_OPTS_PUTS("data=journal");
1821 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1822 SEQ_OPTS_PUTS("data=ordered");
1823 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1824 SEQ_OPTS_PUTS("data=writeback");
1827 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1828 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1829 sbi->s_inode_readahead_blks);
1831 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1832 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
1833 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
1834 if (nodefs || sbi->s_max_dir_size_kb)
1835 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
1837 ext4_show_quota_options(seq, sb);
1841 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1843 return _ext4_show_options(seq, root->d_sb, 0);
1846 static int options_seq_show(struct seq_file *seq, void *offset)
1848 struct super_block *sb = seq->private;
1851 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1852 rc = _ext4_show_options(seq, sb, 1);
1853 seq_puts(seq, "\n");
1857 static int options_open_fs(struct inode *inode, struct file *file)
1859 return single_open(file, options_seq_show, PDE(inode)->data);
1862 static const struct file_operations ext4_seq_options_fops = {
1863 .owner = THIS_MODULE,
1864 .open = options_open_fs,
1866 .llseek = seq_lseek,
1867 .release = single_release,
1870 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1873 struct ext4_sb_info *sbi = EXT4_SB(sb);
1876 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1877 ext4_msg(sb, KERN_ERR, "revision level too high, "
1878 "forcing read-only mode");
1883 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1884 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1885 "running e2fsck is recommended");
1886 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1887 ext4_msg(sb, KERN_WARNING,
1888 "warning: mounting fs with errors, "
1889 "running e2fsck is recommended");
1890 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1891 le16_to_cpu(es->s_mnt_count) >=
1892 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1893 ext4_msg(sb, KERN_WARNING,
1894 "warning: maximal mount count reached, "
1895 "running e2fsck is recommended");
1896 else if (le32_to_cpu(es->s_checkinterval) &&
1897 (le32_to_cpu(es->s_lastcheck) +
1898 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1899 ext4_msg(sb, KERN_WARNING,
1900 "warning: checktime reached, "
1901 "running e2fsck is recommended");
1902 if (!sbi->s_journal)
1903 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1904 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1905 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1906 le16_add_cpu(&es->s_mnt_count, 1);
1907 es->s_mtime = cpu_to_le32(get_seconds());
1908 ext4_update_dynamic_rev(sb);
1910 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1912 ext4_commit_super(sb, 1);
1914 if (test_opt(sb, DEBUG))
1915 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1916 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1918 sbi->s_groups_count,
1919 EXT4_BLOCKS_PER_GROUP(sb),
1920 EXT4_INODES_PER_GROUP(sb),
1921 sbi->s_mount_opt, sbi->s_mount_opt2);
1923 cleancache_init_fs(sb);
1927 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1929 struct ext4_sb_info *sbi = EXT4_SB(sb);
1930 struct flex_groups *new_groups;
1933 if (!sbi->s_log_groups_per_flex)
1936 size = ext4_flex_group(sbi, ngroup - 1) + 1;
1937 if (size <= sbi->s_flex_groups_allocated)
1940 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1941 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1943 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1944 size / (int) sizeof(struct flex_groups));
1948 if (sbi->s_flex_groups) {
1949 memcpy(new_groups, sbi->s_flex_groups,
1950 (sbi->s_flex_groups_allocated *
1951 sizeof(struct flex_groups)));
1952 ext4_kvfree(sbi->s_flex_groups);
1954 sbi->s_flex_groups = new_groups;
1955 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1959 static int ext4_fill_flex_info(struct super_block *sb)
1961 struct ext4_sb_info *sbi = EXT4_SB(sb);
1962 struct ext4_group_desc *gdp = NULL;
1963 ext4_group_t flex_group;
1964 unsigned int groups_per_flex = 0;
1967 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1968 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
1969 sbi->s_log_groups_per_flex = 0;
1972 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1974 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
1978 for (i = 0; i < sbi->s_groups_count; i++) {
1979 gdp = ext4_get_group_desc(sb, i, NULL);
1981 flex_group = ext4_flex_group(sbi, i);
1982 atomic_add(ext4_free_inodes_count(sb, gdp),
1983 &sbi->s_flex_groups[flex_group].free_inodes);
1984 atomic_add(ext4_free_group_clusters(sb, gdp),
1985 &sbi->s_flex_groups[flex_group].free_clusters);
1986 atomic_add(ext4_used_dirs_count(sb, gdp),
1987 &sbi->s_flex_groups[flex_group].used_dirs);
1995 static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1996 struct ext4_group_desc *gdp)
2000 __le32 le_group = cpu_to_le32(block_group);
2002 if ((sbi->s_es->s_feature_ro_compat &
2003 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))) {
2004 /* Use new metadata_csum algorithm */
2008 old_csum = gdp->bg_checksum;
2009 gdp->bg_checksum = 0;
2010 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2012 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
2014 gdp->bg_checksum = old_csum;
2016 crc = csum32 & 0xFFFF;
2020 /* old crc16 code */
2021 offset = offsetof(struct ext4_group_desc, bg_checksum);
2023 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2024 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2025 crc = crc16(crc, (__u8 *)gdp, offset);
2026 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2027 /* for checksum of struct ext4_group_desc do the rest...*/
2028 if ((sbi->s_es->s_feature_incompat &
2029 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2030 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2031 crc = crc16(crc, (__u8 *)gdp + offset,
2032 le16_to_cpu(sbi->s_es->s_desc_size) -
2036 return cpu_to_le16(crc);
2039 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2040 struct ext4_group_desc *gdp)
2042 if (ext4_has_group_desc_csum(sb) &&
2043 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
2050 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2051 struct ext4_group_desc *gdp)
2053 if (!ext4_has_group_desc_csum(sb))
2055 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2058 /* Called at mount-time, super-block is locked */
2059 static int ext4_check_descriptors(struct super_block *sb,
2060 ext4_group_t *first_not_zeroed)
2062 struct ext4_sb_info *sbi = EXT4_SB(sb);
2063 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2064 ext4_fsblk_t last_block;
2065 ext4_fsblk_t block_bitmap;
2066 ext4_fsblk_t inode_bitmap;
2067 ext4_fsblk_t inode_table;
2068 int flexbg_flag = 0;
2069 ext4_group_t i, grp = sbi->s_groups_count;
2071 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2074 ext4_debug("Checking group descriptors");
2076 for (i = 0; i < sbi->s_groups_count; i++) {
2077 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2079 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2080 last_block = ext4_blocks_count(sbi->s_es) - 1;
2082 last_block = first_block +
2083 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2085 if ((grp == sbi->s_groups_count) &&
2086 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2089 block_bitmap = ext4_block_bitmap(sb, gdp);
2090 if (block_bitmap < first_block || block_bitmap > last_block) {
2091 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2092 "Block bitmap for group %u not in group "
2093 "(block %llu)!", i, block_bitmap);
2096 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2097 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2098 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2099 "Inode bitmap for group %u not in group "
2100 "(block %llu)!", i, inode_bitmap);
2103 inode_table = ext4_inode_table(sb, gdp);
2104 if (inode_table < first_block ||
2105 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2106 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2107 "Inode table for group %u not in group "
2108 "(block %llu)!", i, inode_table);
2111 ext4_lock_group(sb, i);
2112 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2113 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2114 "Checksum for group %u failed (%u!=%u)",
2115 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2116 gdp)), le16_to_cpu(gdp->bg_checksum));
2117 if (!(sb->s_flags & MS_RDONLY)) {
2118 ext4_unlock_group(sb, i);
2122 ext4_unlock_group(sb, i);
2124 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2126 if (NULL != first_not_zeroed)
2127 *first_not_zeroed = grp;
2129 ext4_free_blocks_count_set(sbi->s_es,
2130 EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
2131 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
2135 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2136 * the superblock) which were deleted from all directories, but held open by
2137 * a process at the time of a crash. We walk the list and try to delete these
2138 * inodes at recovery time (only with a read-write filesystem).
2140 * In order to keep the orphan inode chain consistent during traversal (in
2141 * case of crash during recovery), we link each inode into the superblock
2142 * orphan list_head and handle it the same way as an inode deletion during
2143 * normal operation (which journals the operations for us).
2145 * We only do an iget() and an iput() on each inode, which is very safe if we
2146 * accidentally point at an in-use or already deleted inode. The worst that
2147 * can happen in this case is that we get a "bit already cleared" message from
2148 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2149 * e2fsck was run on this filesystem, and it must have already done the orphan
2150 * inode cleanup for us, so we can safely abort without any further action.
2152 static void ext4_orphan_cleanup(struct super_block *sb,
2153 struct ext4_super_block *es)
2155 unsigned int s_flags = sb->s_flags;
2156 int nr_orphans = 0, nr_truncates = 0;
2160 if (!es->s_last_orphan) {
2161 jbd_debug(4, "no orphan inodes to clean up\n");
2165 if (bdev_read_only(sb->s_bdev)) {
2166 ext4_msg(sb, KERN_ERR, "write access "
2167 "unavailable, skipping orphan cleanup");
2171 /* Check if feature set would not allow a r/w mount */
2172 if (!ext4_feature_set_ok(sb, 0)) {
2173 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2174 "unknown ROCOMPAT features");
2178 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2179 /* don't clear list on RO mount w/ errors */
2180 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
2181 jbd_debug(1, "Errors on filesystem, "
2182 "clearing orphan list.\n");
2183 es->s_last_orphan = 0;
2185 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2189 if (s_flags & MS_RDONLY) {
2190 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2191 sb->s_flags &= ~MS_RDONLY;
2194 /* Needed for iput() to work correctly and not trash data */
2195 sb->s_flags |= MS_ACTIVE;
2196 /* Turn on quotas so that they are updated correctly */
2197 for (i = 0; i < MAXQUOTAS; i++) {
2198 if (EXT4_SB(sb)->s_qf_names[i]) {
2199 int ret = ext4_quota_on_mount(sb, i);
2201 ext4_msg(sb, KERN_ERR,
2202 "Cannot turn on journaled "
2203 "quota: error %d", ret);
2208 while (es->s_last_orphan) {
2209 struct inode *inode;
2211 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2212 if (IS_ERR(inode)) {
2213 es->s_last_orphan = 0;
2217 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2218 dquot_initialize(inode);
2219 if (inode->i_nlink) {
2220 ext4_msg(sb, KERN_DEBUG,
2221 "%s: truncating inode %lu to %lld bytes",
2222 __func__, inode->i_ino, inode->i_size);
2223 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2224 inode->i_ino, inode->i_size);
2225 ext4_truncate(inode);
2228 ext4_msg(sb, KERN_DEBUG,
2229 "%s: deleting unreferenced inode %lu",
2230 __func__, inode->i_ino);
2231 jbd_debug(2, "deleting unreferenced inode %lu\n",
2235 iput(inode); /* The delete magic happens here! */
2238 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2241 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2242 PLURAL(nr_orphans));
2244 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2245 PLURAL(nr_truncates));
2247 /* Turn quotas off */
2248 for (i = 0; i < MAXQUOTAS; i++) {
2249 if (sb_dqopt(sb)->files[i])
2250 dquot_quota_off(sb, i);
2253 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2257 * Maximal extent format file size.
2258 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2259 * extent format containers, within a sector_t, and within i_blocks
2260 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2261 * so that won't be a limiting factor.
2263 * However there is other limiting factor. We do store extents in the form
2264 * of starting block and length, hence the resulting length of the extent
2265 * covering maximum file size must fit into on-disk format containers as
2266 * well. Given that length is always by 1 unit bigger than max unit (because
2267 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2269 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2271 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2274 loff_t upper_limit = MAX_LFS_FILESIZE;
2276 /* small i_blocks in vfs inode? */
2277 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2279 * CONFIG_LBDAF is not enabled implies the inode
2280 * i_block represent total blocks in 512 bytes
2281 * 32 == size of vfs inode i_blocks * 8
2283 upper_limit = (1LL << 32) - 1;
2285 /* total blocks in file system block size */
2286 upper_limit >>= (blkbits - 9);
2287 upper_limit <<= blkbits;
2291 * 32-bit extent-start container, ee_block. We lower the maxbytes
2292 * by one fs block, so ee_len can cover the extent of maximum file
2295 res = (1LL << 32) - 1;
2298 /* Sanity check against vm- & vfs- imposed limits */
2299 if (res > upper_limit)
2306 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2307 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2308 * We need to be 1 filesystem block less than the 2^48 sector limit.
2310 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2312 loff_t res = EXT4_NDIR_BLOCKS;
2315 /* This is calculated to be the largest file size for a dense, block
2316 * mapped file such that the file's total number of 512-byte sectors,
2317 * including data and all indirect blocks, does not exceed (2^48 - 1).
2319 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2320 * number of 512-byte sectors of the file.
2323 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2325 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2326 * the inode i_block field represents total file blocks in
2327 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2329 upper_limit = (1LL << 32) - 1;
2331 /* total blocks in file system block size */
2332 upper_limit >>= (bits - 9);
2336 * We use 48 bit ext4_inode i_blocks
2337 * With EXT4_HUGE_FILE_FL set the i_blocks
2338 * represent total number of blocks in
2339 * file system block size
2341 upper_limit = (1LL << 48) - 1;
2345 /* indirect blocks */
2347 /* double indirect blocks */
2348 meta_blocks += 1 + (1LL << (bits-2));
2349 /* tripple indirect blocks */
2350 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2352 upper_limit -= meta_blocks;
2353 upper_limit <<= bits;
2355 res += 1LL << (bits-2);
2356 res += 1LL << (2*(bits-2));
2357 res += 1LL << (3*(bits-2));
2359 if (res > upper_limit)
2362 if (res > MAX_LFS_FILESIZE)
2363 res = MAX_LFS_FILESIZE;
2368 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2369 ext4_fsblk_t logical_sb_block, int nr)
2371 struct ext4_sb_info *sbi = EXT4_SB(sb);
2372 ext4_group_t bg, first_meta_bg;
2375 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2377 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2379 return logical_sb_block + nr + 1;
2380 bg = sbi->s_desc_per_block * nr;
2381 if (ext4_bg_has_super(sb, bg))
2384 return (has_super + ext4_group_first_block_no(sb, bg));
2388 * ext4_get_stripe_size: Get the stripe size.
2389 * @sbi: In memory super block info
2391 * If we have specified it via mount option, then
2392 * use the mount option value. If the value specified at mount time is
2393 * greater than the blocks per group use the super block value.
2394 * If the super block value is greater than blocks per group return 0.
2395 * Allocator needs it be less than blocks per group.
2398 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2400 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2401 unsigned long stripe_width =
2402 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2405 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2406 ret = sbi->s_stripe;
2407 else if (stripe_width <= sbi->s_blocks_per_group)
2409 else if (stride <= sbi->s_blocks_per_group)
2415 * If the stripe width is 1, this makes no sense and
2416 * we set it to 0 to turn off stripe handling code.
2427 struct attribute attr;
2428 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2429 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2430 const char *, size_t);
2434 static int parse_strtoul(const char *buf,
2435 unsigned long max, unsigned long *value)
2439 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2440 endp = skip_spaces(endp);
2441 if (*endp || *value > max)
2447 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2448 struct ext4_sb_info *sbi,
2451 return snprintf(buf, PAGE_SIZE, "%llu\n",
2453 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
2456 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2457 struct ext4_sb_info *sbi, char *buf)
2459 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2461 if (!sb->s_bdev->bd_part)
2462 return snprintf(buf, PAGE_SIZE, "0\n");
2463 return snprintf(buf, PAGE_SIZE, "%lu\n",
2464 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2465 sbi->s_sectors_written_start) >> 1);
2468 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2469 struct ext4_sb_info *sbi, char *buf)
2471 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2473 if (!sb->s_bdev->bd_part)
2474 return snprintf(buf, PAGE_SIZE, "0\n");
2475 return snprintf(buf, PAGE_SIZE, "%llu\n",
2476 (unsigned long long)(sbi->s_kbytes_written +
2477 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2478 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2481 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2482 struct ext4_sb_info *sbi,
2483 const char *buf, size_t count)
2487 if (parse_strtoul(buf, 0x40000000, &t))
2490 if (t && !is_power_of_2(t))
2493 sbi->s_inode_readahead_blks = t;
2497 static ssize_t sbi_ui_show(struct ext4_attr *a,
2498 struct ext4_sb_info *sbi, char *buf)
2500 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2502 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2505 static ssize_t sbi_ui_store(struct ext4_attr *a,
2506 struct ext4_sb_info *sbi,
2507 const char *buf, size_t count)
2509 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2512 if (parse_strtoul(buf, 0xffffffff, &t))
2518 static ssize_t trigger_test_error(struct ext4_attr *a,
2519 struct ext4_sb_info *sbi,
2520 const char *buf, size_t count)
2524 if (!capable(CAP_SYS_ADMIN))
2527 if (len && buf[len-1] == '\n')
2531 ext4_error(sbi->s_sb, "%.*s", len, buf);
2535 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2536 static struct ext4_attr ext4_attr_##_name = { \
2537 .attr = {.name = __stringify(_name), .mode = _mode }, \
2540 .offset = offsetof(struct ext4_sb_info, _elname), \
2542 #define EXT4_ATTR(name, mode, show, store) \
2543 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2545 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2546 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2547 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2548 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2549 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2550 #define ATTR_LIST(name) &ext4_attr_##name.attr
2552 EXT4_RO_ATTR(delayed_allocation_blocks);
2553 EXT4_RO_ATTR(session_write_kbytes);
2554 EXT4_RO_ATTR(lifetime_write_kbytes);
2555 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2556 inode_readahead_blks_store, s_inode_readahead_blks);
2557 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2558 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2559 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2560 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2561 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2562 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2563 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2564 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2565 EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb);
2566 EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
2568 static struct attribute *ext4_attrs[] = {
2569 ATTR_LIST(delayed_allocation_blocks),
2570 ATTR_LIST(session_write_kbytes),
2571 ATTR_LIST(lifetime_write_kbytes),
2572 ATTR_LIST(inode_readahead_blks),
2573 ATTR_LIST(inode_goal),
2574 ATTR_LIST(mb_stats),
2575 ATTR_LIST(mb_max_to_scan),
2576 ATTR_LIST(mb_min_to_scan),
2577 ATTR_LIST(mb_order2_req),
2578 ATTR_LIST(mb_stream_req),
2579 ATTR_LIST(mb_group_prealloc),
2580 ATTR_LIST(max_writeback_mb_bump),
2581 ATTR_LIST(extent_max_zeroout_kb),
2582 ATTR_LIST(trigger_fs_error),
2586 /* Features this copy of ext4 supports */
2587 EXT4_INFO_ATTR(lazy_itable_init);
2588 EXT4_INFO_ATTR(batched_discard);
2589 EXT4_INFO_ATTR(meta_bg_resize);
2591 static struct attribute *ext4_feat_attrs[] = {
2592 ATTR_LIST(lazy_itable_init),
2593 ATTR_LIST(batched_discard),
2594 ATTR_LIST(meta_bg_resize),
2598 static ssize_t ext4_attr_show(struct kobject *kobj,
2599 struct attribute *attr, char *buf)
2601 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2603 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2605 return a->show ? a->show(a, sbi, buf) : 0;
2608 static ssize_t ext4_attr_store(struct kobject *kobj,
2609 struct attribute *attr,
2610 const char *buf, size_t len)
2612 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2614 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2616 return a->store ? a->store(a, sbi, buf, len) : 0;
2619 static void ext4_sb_release(struct kobject *kobj)
2621 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2623 complete(&sbi->s_kobj_unregister);
2626 static const struct sysfs_ops ext4_attr_ops = {
2627 .show = ext4_attr_show,
2628 .store = ext4_attr_store,
2631 static struct kobj_type ext4_ktype = {
2632 .default_attrs = ext4_attrs,
2633 .sysfs_ops = &ext4_attr_ops,
2634 .release = ext4_sb_release,
2637 static void ext4_feat_release(struct kobject *kobj)
2639 complete(&ext4_feat->f_kobj_unregister);
2642 static struct kobj_type ext4_feat_ktype = {
2643 .default_attrs = ext4_feat_attrs,
2644 .sysfs_ops = &ext4_attr_ops,
2645 .release = ext4_feat_release,
2649 * Check whether this filesystem can be mounted based on
2650 * the features present and the RDONLY/RDWR mount requested.
2651 * Returns 1 if this filesystem can be mounted as requested,
2652 * 0 if it cannot be.
2654 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2656 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2657 ext4_msg(sb, KERN_ERR,
2658 "Couldn't mount because of "
2659 "unsupported optional features (%x)",
2660 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2661 ~EXT4_FEATURE_INCOMPAT_SUPP));
2668 /* Check that feature set is OK for a read-write mount */
2669 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2670 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2671 "unsupported optional features (%x)",
2672 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2673 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2677 * Large file size enabled file system can only be mounted
2678 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2680 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2681 if (sizeof(blkcnt_t) < sizeof(u64)) {
2682 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2683 "cannot be mounted RDWR without "
2688 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2689 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2690 ext4_msg(sb, KERN_ERR,
2691 "Can't support bigalloc feature without "
2692 "extents feature\n");
2696 #ifndef CONFIG_QUOTA
2697 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2699 ext4_msg(sb, KERN_ERR,
2700 "Filesystem with quota feature cannot be mounted RDWR "
2701 "without CONFIG_QUOTA");
2704 #endif /* CONFIG_QUOTA */
2709 * This function is called once a day if we have errors logged
2710 * on the file system
2712 static void print_daily_error_info(unsigned long arg)
2714 struct super_block *sb = (struct super_block *) arg;
2715 struct ext4_sb_info *sbi;
2716 struct ext4_super_block *es;
2721 if (es->s_error_count)
2722 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2723 le32_to_cpu(es->s_error_count));
2724 if (es->s_first_error_time) {
2725 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2726 sb->s_id, le32_to_cpu(es->s_first_error_time),
2727 (int) sizeof(es->s_first_error_func),
2728 es->s_first_error_func,
2729 le32_to_cpu(es->s_first_error_line));
2730 if (es->s_first_error_ino)
2731 printk(": inode %u",
2732 le32_to_cpu(es->s_first_error_ino));
2733 if (es->s_first_error_block)
2734 printk(": block %llu", (unsigned long long)
2735 le64_to_cpu(es->s_first_error_block));
2738 if (es->s_last_error_time) {
2739 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2740 sb->s_id, le32_to_cpu(es->s_last_error_time),
2741 (int) sizeof(es->s_last_error_func),
2742 es->s_last_error_func,
2743 le32_to_cpu(es->s_last_error_line));
2744 if (es->s_last_error_ino)
2745 printk(": inode %u",
2746 le32_to_cpu(es->s_last_error_ino));
2747 if (es->s_last_error_block)
2748 printk(": block %llu", (unsigned long long)
2749 le64_to_cpu(es->s_last_error_block));
2752 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2755 /* Find next suitable group and run ext4_init_inode_table */
2756 static int ext4_run_li_request(struct ext4_li_request *elr)
2758 struct ext4_group_desc *gdp = NULL;
2759 ext4_group_t group, ngroups;
2760 struct super_block *sb;
2761 unsigned long timeout = 0;
2765 ngroups = EXT4_SB(sb)->s_groups_count;
2768 for (group = elr->lr_next_group; group < ngroups; group++) {
2769 gdp = ext4_get_group_desc(sb, group, NULL);
2775 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2779 if (group == ngroups)
2784 ret = ext4_init_inode_table(sb, group,
2785 elr->lr_timeout ? 0 : 1);
2786 if (elr->lr_timeout == 0) {
2787 timeout = (jiffies - timeout) *
2788 elr->lr_sbi->s_li_wait_mult;
2789 elr->lr_timeout = timeout;
2791 elr->lr_next_sched = jiffies + elr->lr_timeout;
2792 elr->lr_next_group = group + 1;
2800 * Remove lr_request from the list_request and free the
2801 * request structure. Should be called with li_list_mtx held
2803 static void ext4_remove_li_request(struct ext4_li_request *elr)
2805 struct ext4_sb_info *sbi;
2812 list_del(&elr->lr_request);
2813 sbi->s_li_request = NULL;
2817 static void ext4_unregister_li_request(struct super_block *sb)
2819 mutex_lock(&ext4_li_mtx);
2820 if (!ext4_li_info) {
2821 mutex_unlock(&ext4_li_mtx);
2825 mutex_lock(&ext4_li_info->li_list_mtx);
2826 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2827 mutex_unlock(&ext4_li_info->li_list_mtx);
2828 mutex_unlock(&ext4_li_mtx);
2831 static struct task_struct *ext4_lazyinit_task;
2834 * This is the function where ext4lazyinit thread lives. It walks
2835 * through the request list searching for next scheduled filesystem.
2836 * When such a fs is found, run the lazy initialization request
2837 * (ext4_rn_li_request) and keep track of the time spend in this
2838 * function. Based on that time we compute next schedule time of
2839 * the request. When walking through the list is complete, compute
2840 * next waking time and put itself into sleep.
2842 static int ext4_lazyinit_thread(void *arg)
2844 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2845 struct list_head *pos, *n;
2846 struct ext4_li_request *elr;
2847 unsigned long next_wakeup, cur;
2849 BUG_ON(NULL == eli);
2853 next_wakeup = MAX_JIFFY_OFFSET;
2855 mutex_lock(&eli->li_list_mtx);
2856 if (list_empty(&eli->li_request_list)) {
2857 mutex_unlock(&eli->li_list_mtx);
2861 list_for_each_safe(pos, n, &eli->li_request_list) {
2862 elr = list_entry(pos, struct ext4_li_request,
2865 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2866 if (ext4_run_li_request(elr) != 0) {
2867 /* error, remove the lazy_init job */
2868 ext4_remove_li_request(elr);
2873 if (time_before(elr->lr_next_sched, next_wakeup))
2874 next_wakeup = elr->lr_next_sched;
2876 mutex_unlock(&eli->li_list_mtx);
2881 if ((time_after_eq(cur, next_wakeup)) ||
2882 (MAX_JIFFY_OFFSET == next_wakeup)) {
2887 schedule_timeout_interruptible(next_wakeup - cur);
2889 if (kthread_should_stop()) {
2890 ext4_clear_request_list();
2897 * It looks like the request list is empty, but we need
2898 * to check it under the li_list_mtx lock, to prevent any
2899 * additions into it, and of course we should lock ext4_li_mtx
2900 * to atomically free the list and ext4_li_info, because at
2901 * this point another ext4 filesystem could be registering
2904 mutex_lock(&ext4_li_mtx);
2905 mutex_lock(&eli->li_list_mtx);
2906 if (!list_empty(&eli->li_request_list)) {
2907 mutex_unlock(&eli->li_list_mtx);
2908 mutex_unlock(&ext4_li_mtx);
2911 mutex_unlock(&eli->li_list_mtx);
2912 kfree(ext4_li_info);
2913 ext4_li_info = NULL;
2914 mutex_unlock(&ext4_li_mtx);
2919 static void ext4_clear_request_list(void)
2921 struct list_head *pos, *n;
2922 struct ext4_li_request *elr;
2924 mutex_lock(&ext4_li_info->li_list_mtx);
2925 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2926 elr = list_entry(pos, struct ext4_li_request,
2928 ext4_remove_li_request(elr);
2930 mutex_unlock(&ext4_li_info->li_list_mtx);
2933 static int ext4_run_lazyinit_thread(void)
2935 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2936 ext4_li_info, "ext4lazyinit");
2937 if (IS_ERR(ext4_lazyinit_task)) {
2938 int err = PTR_ERR(ext4_lazyinit_task);
2939 ext4_clear_request_list();
2940 kfree(ext4_li_info);
2941 ext4_li_info = NULL;
2942 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
2943 "initialization thread\n",
2947 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2952 * Check whether it make sense to run itable init. thread or not.
2953 * If there is at least one uninitialized inode table, return
2954 * corresponding group number, else the loop goes through all
2955 * groups and return total number of groups.
2957 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2959 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2960 struct ext4_group_desc *gdp = NULL;
2962 for (group = 0; group < ngroups; group++) {
2963 gdp = ext4_get_group_desc(sb, group, NULL);
2967 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2974 static int ext4_li_info_new(void)
2976 struct ext4_lazy_init *eli = NULL;
2978 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2982 INIT_LIST_HEAD(&eli->li_request_list);
2983 mutex_init(&eli->li_list_mtx);
2985 eli->li_state |= EXT4_LAZYINIT_QUIT;
2992 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2995 struct ext4_sb_info *sbi = EXT4_SB(sb);
2996 struct ext4_li_request *elr;
2999 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3005 elr->lr_next_group = start;
3008 * Randomize first schedule time of the request to
3009 * spread the inode table initialization requests
3012 get_random_bytes(&rnd, sizeof(rnd));
3013 elr->lr_next_sched = jiffies + (unsigned long)rnd %
3014 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
3019 static int ext4_register_li_request(struct super_block *sb,
3020 ext4_group_t first_not_zeroed)
3022 struct ext4_sb_info *sbi = EXT4_SB(sb);
3023 struct ext4_li_request *elr;
3024 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3027 if (sbi->s_li_request != NULL) {
3029 * Reset timeout so it can be computed again, because
3030 * s_li_wait_mult might have changed.
3032 sbi->s_li_request->lr_timeout = 0;
3036 if (first_not_zeroed == ngroups ||
3037 (sb->s_flags & MS_RDONLY) ||
3038 !test_opt(sb, INIT_INODE_TABLE))
3041 elr = ext4_li_request_new(sb, first_not_zeroed);
3045 mutex_lock(&ext4_li_mtx);
3047 if (NULL == ext4_li_info) {
3048 ret = ext4_li_info_new();
3053 mutex_lock(&ext4_li_info->li_list_mtx);
3054 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3055 mutex_unlock(&ext4_li_info->li_list_mtx);
3057 sbi->s_li_request = elr;
3059 * set elr to NULL here since it has been inserted to
3060 * the request_list and the removal and free of it is
3061 * handled by ext4_clear_request_list from now on.
3065 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3066 ret = ext4_run_lazyinit_thread();
3071 mutex_unlock(&ext4_li_mtx);
3078 * We do not need to lock anything since this is called on
3081 static void ext4_destroy_lazyinit_thread(void)
3084 * If thread exited earlier
3085 * there's nothing to be done.
3087 if (!ext4_li_info || !ext4_lazyinit_task)
3090 kthread_stop(ext4_lazyinit_task);
3093 static int set_journal_csum_feature_set(struct super_block *sb)
3096 int compat, incompat;
3097 struct ext4_sb_info *sbi = EXT4_SB(sb);
3099 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3100 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3101 /* journal checksum v2 */
3103 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V2;
3105 /* journal checksum v1 */
3106 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3110 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3111 ret = jbd2_journal_set_features(sbi->s_journal,
3113 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3115 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3116 ret = jbd2_journal_set_features(sbi->s_journal,
3119 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3120 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3122 jbd2_journal_clear_features(sbi->s_journal,
3123 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3124 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3125 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3132 * Note: calculating the overhead so we can be compatible with
3133 * historical BSD practice is quite difficult in the face of
3134 * clusters/bigalloc. This is because multiple metadata blocks from
3135 * different block group can end up in the same allocation cluster.
3136 * Calculating the exact overhead in the face of clustered allocation
3137 * requires either O(all block bitmaps) in memory or O(number of block
3138 * groups**2) in time. We will still calculate the superblock for
3139 * older file systems --- and if we come across with a bigalloc file
3140 * system with zero in s_overhead_clusters the estimate will be close to
3141 * correct especially for very large cluster sizes --- but for newer
3142 * file systems, it's better to calculate this figure once at mkfs
3143 * time, and store it in the superblock. If the superblock value is
3144 * present (even for non-bigalloc file systems), we will use it.
3146 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3149 struct ext4_sb_info *sbi = EXT4_SB(sb);
3150 struct ext4_group_desc *gdp;
3151 ext4_fsblk_t first_block, last_block, b;
3152 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3153 int s, j, count = 0;
3155 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3156 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3157 sbi->s_itb_per_group + 2);
3159 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3160 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3161 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3162 for (i = 0; i < ngroups; i++) {
3163 gdp = ext4_get_group_desc(sb, i, NULL);
3164 b = ext4_block_bitmap(sb, gdp);
3165 if (b >= first_block && b <= last_block) {
3166 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3169 b = ext4_inode_bitmap(sb, gdp);
3170 if (b >= first_block && b <= last_block) {
3171 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3174 b = ext4_inode_table(sb, gdp);
3175 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3176 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3177 int c = EXT4_B2C(sbi, b - first_block);
3178 ext4_set_bit(c, buf);
3184 if (ext4_bg_has_super(sb, grp)) {
3185 ext4_set_bit(s++, buf);
3188 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3189 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3195 return EXT4_CLUSTERS_PER_GROUP(sb) -
3196 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3200 * Compute the overhead and stash it in sbi->s_overhead
3202 int ext4_calculate_overhead(struct super_block *sb)
3204 struct ext4_sb_info *sbi = EXT4_SB(sb);
3205 struct ext4_super_block *es = sbi->s_es;
3206 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3207 ext4_fsblk_t overhead = 0;
3208 char *buf = (char *) get_zeroed_page(GFP_KERNEL);
3210 memset(buf, 0, PAGE_SIZE);
3215 * Compute the overhead (FS structures). This is constant
3216 * for a given filesystem unless the number of block groups
3217 * changes so we cache the previous value until it does.
3221 * All of the blocks before first_data_block are overhead
3223 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3226 * Add the overhead found in each block group
3228 for (i = 0; i < ngroups; i++) {
3231 blks = count_overhead(sb, i, buf);
3234 memset(buf, 0, PAGE_SIZE);
3237 sbi->s_overhead = overhead;
3239 free_page((unsigned long) buf);
3243 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3245 char *orig_data = kstrdup(data, GFP_KERNEL);
3246 struct buffer_head *bh;
3247 struct ext4_super_block *es = NULL;
3248 struct ext4_sb_info *sbi;
3250 ext4_fsblk_t sb_block = get_sb_block(&data);
3251 ext4_fsblk_t logical_sb_block;
3252 unsigned long offset = 0;
3253 unsigned long journal_devnum = 0;
3254 unsigned long def_mount_opts;
3259 int blocksize, clustersize;
3260 unsigned int db_count;
3262 int needs_recovery, has_huge_files, has_bigalloc;
3265 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3266 ext4_group_t first_not_zeroed;
3268 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3272 sbi->s_blockgroup_lock =
3273 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3274 if (!sbi->s_blockgroup_lock) {
3278 sb->s_fs_info = sbi;
3280 sbi->s_mount_opt = 0;
3281 sbi->s_resuid = make_kuid(&init_user_ns, EXT4_DEF_RESUID);
3282 sbi->s_resgid = make_kgid(&init_user_ns, EXT4_DEF_RESGID);
3283 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3284 sbi->s_sb_block = sb_block;
3285 if (sb->s_bdev->bd_part)
3286 sbi->s_sectors_written_start =
3287 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3289 /* Cleanup superblock name */
3290 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3294 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3296 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3301 * The ext4 superblock will not be buffer aligned for other than 1kB
3302 * block sizes. We need to calculate the offset from buffer start.
3304 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3305 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3306 offset = do_div(logical_sb_block, blocksize);
3308 logical_sb_block = sb_block;
3311 if (!(bh = sb_bread(sb, logical_sb_block))) {
3312 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3316 * Note: s_es must be initialized as soon as possible because
3317 * some ext4 macro-instructions depend on its value
3319 es = (struct ext4_super_block *) (bh->b_data + offset);
3321 sb->s_magic = le16_to_cpu(es->s_magic);
3322 if (sb->s_magic != EXT4_SUPER_MAGIC)
3324 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3326 /* Warn if metadata_csum and gdt_csum are both set. */
3327 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3328 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3329 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
3330 ext4_warning(sb, KERN_INFO "metadata_csum and uninit_bg are "
3331 "redundant flags; please run fsck.");
3333 /* Check for a known checksum algorithm */
3334 if (!ext4_verify_csum_type(sb, es)) {
3335 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3336 "unknown checksum algorithm.");
3341 /* Load the checksum driver */
3342 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3343 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3344 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3345 if (IS_ERR(sbi->s_chksum_driver)) {
3346 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3347 ret = PTR_ERR(sbi->s_chksum_driver);
3348 sbi->s_chksum_driver = NULL;
3353 /* Check superblock checksum */
3354 if (!ext4_superblock_csum_verify(sb, es)) {
3355 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3356 "invalid superblock checksum. Run e2fsck?");
3361 /* Precompute checksum seed for all metadata */
3362 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3363 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
3364 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3365 sizeof(es->s_uuid));
3367 /* Set defaults before we parse the mount options */
3368 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3369 set_opt(sb, INIT_INODE_TABLE);
3370 if (def_mount_opts & EXT4_DEFM_DEBUG)
3372 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3374 if (def_mount_opts & EXT4_DEFM_UID16)
3375 set_opt(sb, NO_UID32);
3376 /* xattr user namespace & acls are now defaulted on */
3377 #ifdef CONFIG_EXT4_FS_XATTR
3378 set_opt(sb, XATTR_USER);
3380 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3381 set_opt(sb, POSIX_ACL);
3383 set_opt(sb, MBLK_IO_SUBMIT);
3384 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3385 set_opt(sb, JOURNAL_DATA);
3386 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3387 set_opt(sb, ORDERED_DATA);
3388 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3389 set_opt(sb, WRITEBACK_DATA);
3391 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3392 set_opt(sb, ERRORS_PANIC);
3393 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3394 set_opt(sb, ERRORS_CONT);
3396 set_opt(sb, ERRORS_RO);
3397 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
3398 set_opt(sb, BLOCK_VALIDITY);
3399 if (def_mount_opts & EXT4_DEFM_DISCARD)
3400 set_opt(sb, DISCARD);
3402 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3403 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3404 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3405 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3406 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3408 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3409 set_opt(sb, BARRIER);
3412 * enable delayed allocation by default
3413 * Use -o nodelalloc to turn it off
3415 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3416 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3417 set_opt(sb, DELALLOC);
3420 * set default s_li_wait_mult for lazyinit, for the case there is
3421 * no mount option specified.
3423 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3425 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3426 &journal_devnum, &journal_ioprio, 0)) {
3427 ext4_msg(sb, KERN_WARNING,
3428 "failed to parse options in superblock: %s",
3429 sbi->s_es->s_mount_opts);
3431 sbi->s_def_mount_opt = sbi->s_mount_opt;
3432 if (!parse_options((char *) data, sb, &journal_devnum,
3433 &journal_ioprio, 0))
3436 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3437 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3438 "with data=journal disables delayed "
3439 "allocation and O_DIRECT support!\n");
3440 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3441 ext4_msg(sb, KERN_ERR, "can't mount with "
3442 "both data=journal and delalloc");
3445 if (test_opt(sb, DIOREAD_NOLOCK)) {
3446 ext4_msg(sb, KERN_ERR, "can't mount with "
3447 "both data=journal and delalloc");
3450 if (test_opt(sb, DELALLOC))
3451 clear_opt(sb, DELALLOC);
3454 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3455 if (test_opt(sb, DIOREAD_NOLOCK)) {
3456 if (blocksize < PAGE_SIZE) {
3457 ext4_msg(sb, KERN_ERR, "can't mount with "
3458 "dioread_nolock if block size != PAGE_SIZE");
3463 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3464 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3466 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3467 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3468 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3469 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3470 ext4_msg(sb, KERN_WARNING,
3471 "feature flags set on rev 0 fs, "
3472 "running e2fsck is recommended");
3474 if (IS_EXT2_SB(sb)) {
3475 if (ext2_feature_set_ok(sb))
3476 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3477 "using the ext4 subsystem");
3479 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3480 "to feature incompatibilities");
3485 if (IS_EXT3_SB(sb)) {
3486 if (ext3_feature_set_ok(sb))
3487 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3488 "using the ext4 subsystem");
3490 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3491 "to feature incompatibilities");
3497 * Check feature flags regardless of the revision level, since we
3498 * previously didn't change the revision level when setting the flags,
3499 * so there is a chance incompat flags are set on a rev 0 filesystem.
3501 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3504 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3505 blocksize > EXT4_MAX_BLOCK_SIZE) {
3506 ext4_msg(sb, KERN_ERR,
3507 "Unsupported filesystem blocksize %d", blocksize);
3511 if (sb->s_blocksize != blocksize) {
3512 /* Validate the filesystem blocksize */
3513 if (!sb_set_blocksize(sb, blocksize)) {
3514 ext4_msg(sb, KERN_ERR, "bad block size %d",
3520 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3521 offset = do_div(logical_sb_block, blocksize);
3522 bh = sb_bread(sb, logical_sb_block);
3524 ext4_msg(sb, KERN_ERR,
3525 "Can't read superblock on 2nd try");
3528 es = (struct ext4_super_block *)(bh->b_data + offset);
3530 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3531 ext4_msg(sb, KERN_ERR,
3532 "Magic mismatch, very weird!");
3537 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3538 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3539 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3541 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3543 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3544 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3545 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3547 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3548 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3549 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3550 (!is_power_of_2(sbi->s_inode_size)) ||
3551 (sbi->s_inode_size > blocksize)) {
3552 ext4_msg(sb, KERN_ERR,
3553 "unsupported inode size: %d",
3557 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3558 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3561 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3562 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3563 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3564 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3565 !is_power_of_2(sbi->s_desc_size)) {
3566 ext4_msg(sb, KERN_ERR,
3567 "unsupported descriptor size %lu",
3572 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3574 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3575 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3576 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3579 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3580 if (sbi->s_inodes_per_block == 0)
3582 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3583 sbi->s_inodes_per_block;
3584 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3586 sbi->s_mount_state = le16_to_cpu(es->s_state);
3587 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3588 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3590 for (i = 0; i < 4; i++)
3591 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3592 sbi->s_def_hash_version = es->s_def_hash_version;
3593 i = le32_to_cpu(es->s_flags);
3594 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3595 sbi->s_hash_unsigned = 3;
3596 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3597 #ifdef __CHAR_UNSIGNED__
3598 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3599 sbi->s_hash_unsigned = 3;
3601 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3605 /* Handle clustersize */
3606 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3607 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3608 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3610 if (clustersize < blocksize) {
3611 ext4_msg(sb, KERN_ERR,
3612 "cluster size (%d) smaller than "
3613 "block size (%d)", clustersize, blocksize);
3616 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3617 le32_to_cpu(es->s_log_block_size);
3618 sbi->s_clusters_per_group =
3619 le32_to_cpu(es->s_clusters_per_group);
3620 if (sbi->s_clusters_per_group > blocksize * 8) {
3621 ext4_msg(sb, KERN_ERR,
3622 "#clusters per group too big: %lu",
3623 sbi->s_clusters_per_group);
3626 if (sbi->s_blocks_per_group !=
3627 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3628 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3629 "clusters per group (%lu) inconsistent",
3630 sbi->s_blocks_per_group,
3631 sbi->s_clusters_per_group);
3635 if (clustersize != blocksize) {
3636 ext4_warning(sb, "fragment/cluster size (%d) != "
3637 "block size (%d)", clustersize,
3639 clustersize = blocksize;
3641 if (sbi->s_blocks_per_group > blocksize * 8) {
3642 ext4_msg(sb, KERN_ERR,
3643 "#blocks per group too big: %lu",
3644 sbi->s_blocks_per_group);
3647 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3648 sbi->s_cluster_bits = 0;
3650 sbi->s_cluster_ratio = clustersize / blocksize;
3652 if (sbi->s_inodes_per_group > blocksize * 8) {
3653 ext4_msg(sb, KERN_ERR,
3654 "#inodes per group too big: %lu",
3655 sbi->s_inodes_per_group);
3660 * Test whether we have more sectors than will fit in sector_t,
3661 * and whether the max offset is addressable by the page cache.
3663 err = generic_check_addressable(sb->s_blocksize_bits,
3664 ext4_blocks_count(es));
3666 ext4_msg(sb, KERN_ERR, "filesystem"
3667 " too large to mount safely on this system");
3668 if (sizeof(sector_t) < 8)
3669 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3674 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3677 /* check blocks count against device size */
3678 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3679 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3680 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3681 "exceeds size of device (%llu blocks)",
3682 ext4_blocks_count(es), blocks_count);
3687 * It makes no sense for the first data block to be beyond the end
3688 * of the filesystem.
3690 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3691 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3692 "block %u is beyond end of filesystem (%llu)",
3693 le32_to_cpu(es->s_first_data_block),
3694 ext4_blocks_count(es));
3697 blocks_count = (ext4_blocks_count(es) -
3698 le32_to_cpu(es->s_first_data_block) +
3699 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3700 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3701 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3702 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3703 "(block count %llu, first data block %u, "
3704 "blocks per group %lu)", sbi->s_groups_count,
3705 ext4_blocks_count(es),
3706 le32_to_cpu(es->s_first_data_block),
3707 EXT4_BLOCKS_PER_GROUP(sb));
3710 sbi->s_groups_count = blocks_count;
3711 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3712 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3713 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3714 EXT4_DESC_PER_BLOCK(sb);
3715 sbi->s_group_desc = ext4_kvmalloc(db_count *
3716 sizeof(struct buffer_head *),
3718 if (sbi->s_group_desc == NULL) {
3719 ext4_msg(sb, KERN_ERR, "not enough memory");
3725 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3728 proc_create_data("options", S_IRUGO, sbi->s_proc,
3729 &ext4_seq_options_fops, sb);
3731 bgl_lock_init(sbi->s_blockgroup_lock);
3733 for (i = 0; i < db_count; i++) {
3734 block = descriptor_loc(sb, logical_sb_block, i);
3735 sbi->s_group_desc[i] = sb_bread(sb, block);
3736 if (!sbi->s_group_desc[i]) {
3737 ext4_msg(sb, KERN_ERR,
3738 "can't read group descriptor %d", i);
3743 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3744 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3747 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3748 if (!ext4_fill_flex_info(sb)) {
3749 ext4_msg(sb, KERN_ERR,
3750 "unable to initialize "
3751 "flex_bg meta info!");
3755 sbi->s_gdb_count = db_count;
3756 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3757 spin_lock_init(&sbi->s_next_gen_lock);
3759 init_timer(&sbi->s_err_report);
3760 sbi->s_err_report.function = print_daily_error_info;
3761 sbi->s_err_report.data = (unsigned long) sb;
3763 err = percpu_counter_init(&sbi->s_freeclusters_counter,
3764 ext4_count_free_clusters(sb));
3766 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3767 ext4_count_free_inodes(sb));
3770 err = percpu_counter_init(&sbi->s_dirs_counter,
3771 ext4_count_dirs(sb));
3774 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
3777 ext4_msg(sb, KERN_ERR, "insufficient memory");
3782 sbi->s_stripe = ext4_get_stripe_size(sbi);
3783 sbi->s_max_writeback_mb_bump = 128;
3784 sbi->s_extent_max_zeroout_kb = 32;
3787 * set up enough so that it can read an inode
3789 if (!test_opt(sb, NOLOAD) &&
3790 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3791 sb->s_op = &ext4_sops;
3793 sb->s_op = &ext4_nojournal_sops;
3794 sb->s_export_op = &ext4_export_ops;
3795 sb->s_xattr = ext4_xattr_handlers;
3797 sb->s_qcop = &ext4_qctl_operations;
3798 sb->dq_op = &ext4_quota_operations;
3800 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
3801 /* Use qctl operations for hidden quota files. */
3802 sb->s_qcop = &ext4_qctl_sysfile_operations;
3805 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3807 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3808 mutex_init(&sbi->s_orphan_lock);
3809 sbi->s_resize_flags = 0;
3813 needs_recovery = (es->s_last_orphan != 0 ||
3814 EXT4_HAS_INCOMPAT_FEATURE(sb,
3815 EXT4_FEATURE_INCOMPAT_RECOVER));
3817 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3818 !(sb->s_flags & MS_RDONLY))
3819 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3823 * The first inode we look at is the journal inode. Don't try
3824 * root first: it may be modified in the journal!
3826 if (!test_opt(sb, NOLOAD) &&
3827 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3828 if (ext4_load_journal(sb, es, journal_devnum))
3830 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3831 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3832 ext4_msg(sb, KERN_ERR, "required journal recovery "
3833 "suppressed and not mounted read-only");
3834 goto failed_mount_wq;
3836 clear_opt(sb, DATA_FLAGS);
3837 sbi->s_journal = NULL;
3842 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
3843 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3844 JBD2_FEATURE_INCOMPAT_64BIT)) {
3845 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3846 goto failed_mount_wq;
3849 if (!set_journal_csum_feature_set(sb)) {
3850 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3852 goto failed_mount_wq;
3855 /* We have now updated the journal if required, so we can
3856 * validate the data journaling mode. */
3857 switch (test_opt(sb, DATA_FLAGS)) {
3859 /* No mode set, assume a default based on the journal
3860 * capabilities: ORDERED_DATA if the journal can
3861 * cope, else JOURNAL_DATA
3863 if (jbd2_journal_check_available_features
3864 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3865 set_opt(sb, ORDERED_DATA);
3867 set_opt(sb, JOURNAL_DATA);
3870 case EXT4_MOUNT_ORDERED_DATA:
3871 case EXT4_MOUNT_WRITEBACK_DATA:
3872 if (!jbd2_journal_check_available_features
3873 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3874 ext4_msg(sb, KERN_ERR, "Journal does not support "
3875 "requested data journaling mode");
3876 goto failed_mount_wq;
3881 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3883 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3886 * The journal may have updated the bg summary counts, so we
3887 * need to update the global counters.
3889 percpu_counter_set(&sbi->s_freeclusters_counter,
3890 ext4_count_free_clusters(sb));
3891 percpu_counter_set(&sbi->s_freeinodes_counter,
3892 ext4_count_free_inodes(sb));
3893 percpu_counter_set(&sbi->s_dirs_counter,
3894 ext4_count_dirs(sb));
3895 percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
3899 * Get the # of file system overhead blocks from the
3900 * superblock if present.
3902 if (es->s_overhead_clusters)
3903 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
3905 ret = ext4_calculate_overhead(sb);
3907 goto failed_mount_wq;
3911 * The maximum number of concurrent works can be high and
3912 * concurrency isn't really necessary. Limit it to 1.
3914 EXT4_SB(sb)->dio_unwritten_wq =
3915 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
3916 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3917 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3918 goto failed_mount_wq;
3922 * The jbd2_journal_load will have done any necessary log recovery,
3923 * so we can safely mount the rest of the filesystem now.
3926 root = ext4_iget(sb, EXT4_ROOT_INO);
3928 ext4_msg(sb, KERN_ERR, "get root inode failed");
3929 ret = PTR_ERR(root);
3933 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
3934 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
3938 sb->s_root = d_make_root(root);
3940 ext4_msg(sb, KERN_ERR, "get root dentry failed");
3945 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
3946 sb->s_flags |= MS_RDONLY;
3948 /* determine the minimum size of new large inodes, if present */
3949 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3950 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3951 EXT4_GOOD_OLD_INODE_SIZE;
3952 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3953 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3954 if (sbi->s_want_extra_isize <
3955 le16_to_cpu(es->s_want_extra_isize))
3956 sbi->s_want_extra_isize =
3957 le16_to_cpu(es->s_want_extra_isize);
3958 if (sbi->s_want_extra_isize <
3959 le16_to_cpu(es->s_min_extra_isize))
3960 sbi->s_want_extra_isize =
3961 le16_to_cpu(es->s_min_extra_isize);
3964 /* Check if enough inode space is available */
3965 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3966 sbi->s_inode_size) {
3967 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3968 EXT4_GOOD_OLD_INODE_SIZE;
3969 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3973 err = ext4_setup_system_zone(sb);
3975 ext4_msg(sb, KERN_ERR, "failed to initialize system "
3977 goto failed_mount4a;
3981 err = ext4_mb_init(sb);
3983 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
3988 err = ext4_register_li_request(sb, first_not_zeroed);
3992 sbi->s_kobj.kset = ext4_kset;
3993 init_completion(&sbi->s_kobj_unregister);
3994 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3999 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4000 ext4_orphan_cleanup(sb, es);
4001 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4002 if (needs_recovery) {
4003 ext4_msg(sb, KERN_INFO, "recovery complete");
4004 ext4_mark_recovery_complete(sb, es);
4006 if (EXT4_SB(sb)->s_journal) {
4007 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4008 descr = " journalled data mode";
4009 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4010 descr = " ordered data mode";
4012 descr = " writeback data mode";
4014 descr = "out journal";
4017 /* Enable quota usage during mount. */
4018 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
4019 !(sb->s_flags & MS_RDONLY)) {
4020 ret = ext4_enable_quotas(sb);
4024 #endif /* CONFIG_QUOTA */
4026 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4027 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
4028 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4030 if (es->s_error_count)
4031 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4038 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4042 ext4_unregister_li_request(sb);
4044 ext4_mb_release(sb);
4046 ext4_ext_release(sb);
4047 ext4_release_system_zone(sb);
4052 ext4_msg(sb, KERN_ERR, "mount failed");
4053 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
4055 if (sbi->s_journal) {
4056 jbd2_journal_destroy(sbi->s_journal);
4057 sbi->s_journal = NULL;
4060 del_timer(&sbi->s_err_report);
4061 if (sbi->s_flex_groups)
4062 ext4_kvfree(sbi->s_flex_groups);
4063 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4064 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4065 percpu_counter_destroy(&sbi->s_dirs_counter);
4066 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4068 kthread_stop(sbi->s_mmp_tsk);
4070 for (i = 0; i < db_count; i++)
4071 brelse(sbi->s_group_desc[i]);
4072 ext4_kvfree(sbi->s_group_desc);
4074 if (sbi->s_chksum_driver)
4075 crypto_free_shash(sbi->s_chksum_driver);
4077 remove_proc_entry("options", sbi->s_proc);
4078 remove_proc_entry(sb->s_id, ext4_proc_root);
4081 for (i = 0; i < MAXQUOTAS; i++)
4082 kfree(sbi->s_qf_names[i]);
4084 ext4_blkdev_remove(sbi);
4087 sb->s_fs_info = NULL;
4088 kfree(sbi->s_blockgroup_lock);
4096 * Setup any per-fs journal parameters now. We'll do this both on
4097 * initial mount, once the journal has been initialised but before we've
4098 * done any recovery; and again on any subsequent remount.
4100 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4102 struct ext4_sb_info *sbi = EXT4_SB(sb);
4104 journal->j_commit_interval = sbi->s_commit_interval;
4105 journal->j_min_batch_time = sbi->s_min_batch_time;
4106 journal->j_max_batch_time = sbi->s_max_batch_time;
4108 write_lock(&journal->j_state_lock);
4109 if (test_opt(sb, BARRIER))
4110 journal->j_flags |= JBD2_BARRIER;
4112 journal->j_flags &= ~JBD2_BARRIER;
4113 if (test_opt(sb, DATA_ERR_ABORT))
4114 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4116 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4117 write_unlock(&journal->j_state_lock);
4120 static journal_t *ext4_get_journal(struct super_block *sb,
4121 unsigned int journal_inum)
4123 struct inode *journal_inode;
4126 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4128 /* First, test for the existence of a valid inode on disk. Bad
4129 * things happen if we iget() an unused inode, as the subsequent
4130 * iput() will try to delete it. */
4132 journal_inode = ext4_iget(sb, journal_inum);
4133 if (IS_ERR(journal_inode)) {
4134 ext4_msg(sb, KERN_ERR, "no journal found");
4137 if (!journal_inode->i_nlink) {
4138 make_bad_inode(journal_inode);
4139 iput(journal_inode);
4140 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4144 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4145 journal_inode, journal_inode->i_size);
4146 if (!S_ISREG(journal_inode->i_mode)) {
4147 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4148 iput(journal_inode);
4152 journal = jbd2_journal_init_inode(journal_inode);
4154 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4155 iput(journal_inode);
4158 journal->j_private = sb;
4159 ext4_init_journal_params(sb, journal);
4163 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4166 struct buffer_head *bh;
4170 int hblock, blocksize;
4171 ext4_fsblk_t sb_block;
4172 unsigned long offset;
4173 struct ext4_super_block *es;
4174 struct block_device *bdev;
4176 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4178 bdev = ext4_blkdev_get(j_dev, sb);
4182 blocksize = sb->s_blocksize;
4183 hblock = bdev_logical_block_size(bdev);
4184 if (blocksize < hblock) {
4185 ext4_msg(sb, KERN_ERR,
4186 "blocksize too small for journal device");
4190 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4191 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4192 set_blocksize(bdev, blocksize);
4193 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4194 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4195 "external journal");
4199 es = (struct ext4_super_block *) (bh->b_data + offset);
4200 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4201 !(le32_to_cpu(es->s_feature_incompat) &
4202 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4203 ext4_msg(sb, KERN_ERR, "external journal has "
4209 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4210 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4215 len = ext4_blocks_count(es);
4216 start = sb_block + 1;
4217 brelse(bh); /* we're done with the superblock */
4219 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4220 start, len, blocksize);
4222 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4225 journal->j_private = sb;
4226 ll_rw_block(READ, 1, &journal->j_sb_buffer);
4227 wait_on_buffer(journal->j_sb_buffer);
4228 if (!buffer_uptodate(journal->j_sb_buffer)) {
4229 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4232 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4233 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4234 "user (unsupported) - %d",
4235 be32_to_cpu(journal->j_superblock->s_nr_users));
4238 EXT4_SB(sb)->journal_bdev = bdev;
4239 ext4_init_journal_params(sb, journal);
4243 jbd2_journal_destroy(journal);
4245 ext4_blkdev_put(bdev);
4249 static int ext4_load_journal(struct super_block *sb,
4250 struct ext4_super_block *es,
4251 unsigned long journal_devnum)
4254 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4257 int really_read_only;
4259 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4261 if (journal_devnum &&
4262 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4263 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4264 "numbers have changed");
4265 journal_dev = new_decode_dev(journal_devnum);
4267 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4269 really_read_only = bdev_read_only(sb->s_bdev);
4272 * Are we loading a blank journal or performing recovery after a
4273 * crash? For recovery, we need to check in advance whether we
4274 * can get read-write access to the device.
4276 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4277 if (sb->s_flags & MS_RDONLY) {
4278 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4279 "required on readonly filesystem");
4280 if (really_read_only) {
4281 ext4_msg(sb, KERN_ERR, "write access "
4282 "unavailable, cannot proceed");
4285 ext4_msg(sb, KERN_INFO, "write access will "
4286 "be enabled during recovery");
4290 if (journal_inum && journal_dev) {
4291 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4292 "and inode journals!");
4297 if (!(journal = ext4_get_journal(sb, journal_inum)))
4300 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4304 if (!(journal->j_flags & JBD2_BARRIER))
4305 ext4_msg(sb, KERN_INFO, "barriers disabled");
4307 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
4308 err = jbd2_journal_wipe(journal, !really_read_only);
4310 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4312 memcpy(save, ((char *) es) +
4313 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4314 err = jbd2_journal_load(journal);
4316 memcpy(((char *) es) + EXT4_S_ERR_START,
4317 save, EXT4_S_ERR_LEN);
4322 ext4_msg(sb, KERN_ERR, "error loading journal");
4323 jbd2_journal_destroy(journal);
4327 EXT4_SB(sb)->s_journal = journal;
4328 ext4_clear_journal_err(sb, es);
4330 if (!really_read_only && journal_devnum &&
4331 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4332 es->s_journal_dev = cpu_to_le32(journal_devnum);
4334 /* Make sure we flush the recovery flag to disk. */
4335 ext4_commit_super(sb, 1);
4341 static int ext4_commit_super(struct super_block *sb, int sync)
4343 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4344 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4347 if (!sbh || block_device_ejected(sb))
4349 if (buffer_write_io_error(sbh)) {
4351 * Oh, dear. A previous attempt to write the
4352 * superblock failed. This could happen because the
4353 * USB device was yanked out. Or it could happen to
4354 * be a transient write error and maybe the block will
4355 * be remapped. Nothing we can do but to retry the
4356 * write and hope for the best.
4358 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4359 "superblock detected");
4360 clear_buffer_write_io_error(sbh);
4361 set_buffer_uptodate(sbh);
4364 * If the file system is mounted read-only, don't update the
4365 * superblock write time. This avoids updating the superblock
4366 * write time when we are mounting the root file system
4367 * read/only but we need to replay the journal; at that point,
4368 * for people who are east of GMT and who make their clock
4369 * tick in localtime for Windows bug-for-bug compatibility,
4370 * the clock is set in the future, and this will cause e2fsck
4371 * to complain and force a full file system check.
4373 if (!(sb->s_flags & MS_RDONLY))
4374 es->s_wtime = cpu_to_le32(get_seconds());
4375 if (sb->s_bdev->bd_part)
4376 es->s_kbytes_written =
4377 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4378 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4379 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4381 es->s_kbytes_written =
4382 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4383 ext4_free_blocks_count_set(es,
4384 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4385 &EXT4_SB(sb)->s_freeclusters_counter)));
4386 es->s_free_inodes_count =
4387 cpu_to_le32(percpu_counter_sum_positive(
4388 &EXT4_SB(sb)->s_freeinodes_counter));
4389 BUFFER_TRACE(sbh, "marking dirty");
4390 ext4_superblock_csum_set(sb, es);
4391 mark_buffer_dirty(sbh);
4393 error = sync_dirty_buffer(sbh);
4397 error = buffer_write_io_error(sbh);
4399 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4401 clear_buffer_write_io_error(sbh);
4402 set_buffer_uptodate(sbh);
4409 * Have we just finished recovery? If so, and if we are mounting (or
4410 * remounting) the filesystem readonly, then we will end up with a
4411 * consistent fs on disk. Record that fact.
4413 static void ext4_mark_recovery_complete(struct super_block *sb,
4414 struct ext4_super_block *es)
4416 journal_t *journal = EXT4_SB(sb)->s_journal;
4418 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4419 BUG_ON(journal != NULL);
4422 jbd2_journal_lock_updates(journal);
4423 if (jbd2_journal_flush(journal) < 0)
4426 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4427 sb->s_flags & MS_RDONLY) {
4428 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4429 ext4_commit_super(sb, 1);
4433 jbd2_journal_unlock_updates(journal);
4437 * If we are mounting (or read-write remounting) a filesystem whose journal
4438 * has recorded an error from a previous lifetime, move that error to the
4439 * main filesystem now.
4441 static void ext4_clear_journal_err(struct super_block *sb,
4442 struct ext4_super_block *es)
4448 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4450 journal = EXT4_SB(sb)->s_journal;
4453 * Now check for any error status which may have been recorded in the
4454 * journal by a prior ext4_error() or ext4_abort()
4457 j_errno = jbd2_journal_errno(journal);
4461 errstr = ext4_decode_error(sb, j_errno, nbuf);
4462 ext4_warning(sb, "Filesystem error recorded "
4463 "from previous mount: %s", errstr);
4464 ext4_warning(sb, "Marking fs in need of filesystem check.");
4466 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4467 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4468 ext4_commit_super(sb, 1);
4470 jbd2_journal_clear_err(journal);
4471 jbd2_journal_update_sb_errno(journal);
4476 * Force the running and committing transactions to commit,
4477 * and wait on the commit.
4479 int ext4_force_commit(struct super_block *sb)
4484 if (sb->s_flags & MS_RDONLY)
4487 journal = EXT4_SB(sb)->s_journal;
4489 ret = ext4_journal_force_commit(journal);
4494 static int ext4_sync_fs(struct super_block *sb, int wait)
4498 struct ext4_sb_info *sbi = EXT4_SB(sb);
4500 trace_ext4_sync_fs(sb, wait);
4501 flush_workqueue(sbi->dio_unwritten_wq);
4503 * Writeback quota in non-journalled quota case - journalled quota has
4506 dquot_writeback_dquots(sb, -1);
4507 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4509 jbd2_log_wait_commit(sbi->s_journal, target);
4515 * LVM calls this function before a (read-only) snapshot is created. This
4516 * gives us a chance to flush the journal completely and mark the fs clean.
4518 * Note that only this function cannot bring a filesystem to be in a clean
4519 * state independently. It relies on upper layer to stop all data & metadata
4522 static int ext4_freeze(struct super_block *sb)
4527 if (sb->s_flags & MS_RDONLY)
4530 journal = EXT4_SB(sb)->s_journal;
4532 /* Now we set up the journal barrier. */
4533 jbd2_journal_lock_updates(journal);
4536 * Don't clear the needs_recovery flag if we failed to flush
4539 error = jbd2_journal_flush(journal);
4543 /* Journal blocked and flushed, clear needs_recovery flag. */
4544 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4545 error = ext4_commit_super(sb, 1);
4547 /* we rely on upper layer to stop further updates */
4548 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4553 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4554 * flag here, even though the filesystem is not technically dirty yet.
4556 static int ext4_unfreeze(struct super_block *sb)
4558 if (sb->s_flags & MS_RDONLY)
4561 /* Reset the needs_recovery flag before the fs is unlocked. */
4562 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4563 ext4_commit_super(sb, 1);
4568 * Structure to save mount options for ext4_remount's benefit
4570 struct ext4_mount_options {
4571 unsigned long s_mount_opt;
4572 unsigned long s_mount_opt2;
4575 unsigned long s_commit_interval;
4576 u32 s_min_batch_time, s_max_batch_time;
4579 char *s_qf_names[MAXQUOTAS];
4583 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4585 struct ext4_super_block *es;
4586 struct ext4_sb_info *sbi = EXT4_SB(sb);
4587 unsigned long old_sb_flags;
4588 struct ext4_mount_options old_opts;
4589 int enable_quota = 0;
4591 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4596 char *orig_data = kstrdup(data, GFP_KERNEL);
4598 /* Store the original options */
4599 old_sb_flags = sb->s_flags;
4600 old_opts.s_mount_opt = sbi->s_mount_opt;
4601 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4602 old_opts.s_resuid = sbi->s_resuid;
4603 old_opts.s_resgid = sbi->s_resgid;
4604 old_opts.s_commit_interval = sbi->s_commit_interval;
4605 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4606 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4608 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4609 for (i = 0; i < MAXQUOTAS; i++)
4610 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
4612 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4613 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4616 * Allow the "check" option to be passed as a remount option.
4618 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4623 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4624 ext4_abort(sb, "Abort forced by user");
4626 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4627 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4631 if (sbi->s_journal) {
4632 ext4_init_journal_params(sb, sbi->s_journal);
4633 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4636 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4637 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4642 if (*flags & MS_RDONLY) {
4643 err = dquot_suspend(sb, -1);
4648 * First of all, the unconditional stuff we have to do
4649 * to disable replay of the journal when we next remount
4651 sb->s_flags |= MS_RDONLY;
4654 * OK, test if we are remounting a valid rw partition
4655 * readonly, and if so set the rdonly flag and then
4656 * mark the partition as valid again.
4658 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4659 (sbi->s_mount_state & EXT4_VALID_FS))
4660 es->s_state = cpu_to_le16(sbi->s_mount_state);
4663 ext4_mark_recovery_complete(sb, es);
4665 /* Make sure we can mount this feature set readwrite */
4666 if (!ext4_feature_set_ok(sb, 0)) {
4671 * Make sure the group descriptor checksums
4672 * are sane. If they aren't, refuse to remount r/w.
4674 for (g = 0; g < sbi->s_groups_count; g++) {
4675 struct ext4_group_desc *gdp =
4676 ext4_get_group_desc(sb, g, NULL);
4678 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
4679 ext4_msg(sb, KERN_ERR,
4680 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4681 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4682 le16_to_cpu(gdp->bg_checksum));
4689 * If we have an unprocessed orphan list hanging
4690 * around from a previously readonly bdev mount,
4691 * require a full umount/remount for now.
4693 if (es->s_last_orphan) {
4694 ext4_msg(sb, KERN_WARNING, "Couldn't "
4695 "remount RDWR because of unprocessed "
4696 "orphan inode list. Please "
4697 "umount/remount instead");
4703 * Mounting a RDONLY partition read-write, so reread
4704 * and store the current valid flag. (It may have
4705 * been changed by e2fsck since we originally mounted
4709 ext4_clear_journal_err(sb, es);
4710 sbi->s_mount_state = le16_to_cpu(es->s_state);
4711 if (!ext4_setup_super(sb, es, 0))
4712 sb->s_flags &= ~MS_RDONLY;
4713 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4714 EXT4_FEATURE_INCOMPAT_MMP))
4715 if (ext4_multi_mount_protect(sb,
4716 le64_to_cpu(es->s_mmp_block))) {
4725 * Reinitialize lazy itable initialization thread based on
4728 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4729 ext4_unregister_li_request(sb);
4731 ext4_group_t first_not_zeroed;
4732 first_not_zeroed = ext4_has_uninit_itable(sb);
4733 ext4_register_li_request(sb, first_not_zeroed);
4736 ext4_setup_system_zone(sb);
4737 if (sbi->s_journal == NULL)
4738 ext4_commit_super(sb, 1);
4741 /* Release old quota file names */
4742 for (i = 0; i < MAXQUOTAS; i++)
4743 if (old_opts.s_qf_names[i] &&
4744 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4745 kfree(old_opts.s_qf_names[i]);
4747 if (sb_any_quota_suspended(sb))
4748 dquot_resume(sb, -1);
4749 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4750 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
4751 err = ext4_enable_quotas(sb);
4758 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4763 sb->s_flags = old_sb_flags;
4764 sbi->s_mount_opt = old_opts.s_mount_opt;
4765 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
4766 sbi->s_resuid = old_opts.s_resuid;
4767 sbi->s_resgid = old_opts.s_resgid;
4768 sbi->s_commit_interval = old_opts.s_commit_interval;
4769 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4770 sbi->s_max_batch_time = old_opts.s_max_batch_time;
4772 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4773 for (i = 0; i < MAXQUOTAS; i++) {
4774 if (sbi->s_qf_names[i] &&
4775 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4776 kfree(sbi->s_qf_names[i]);
4777 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4784 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
4786 struct super_block *sb = dentry->d_sb;
4787 struct ext4_sb_info *sbi = EXT4_SB(sb);
4788 struct ext4_super_block *es = sbi->s_es;
4789 ext4_fsblk_t overhead = 0;
4793 if (!test_opt(sb, MINIX_DF))
4794 overhead = sbi->s_overhead;
4796 buf->f_type = EXT4_SUPER_MAGIC;
4797 buf->f_bsize = sb->s_blocksize;
4798 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, sbi->s_overhead);
4799 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4800 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
4801 /* prevent underflow in case that few free space is available */
4802 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
4803 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4804 if (buf->f_bfree < ext4_r_blocks_count(es))
4806 buf->f_files = le32_to_cpu(es->s_inodes_count);
4807 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
4808 buf->f_namelen = EXT4_NAME_LEN;
4809 fsid = le64_to_cpup((void *)es->s_uuid) ^
4810 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4811 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4812 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
4817 /* Helper function for writing quotas on sync - we need to start transaction
4818 * before quota file is locked for write. Otherwise the are possible deadlocks:
4819 * Process 1 Process 2
4820 * ext4_create() quota_sync()
4821 * jbd2_journal_start() write_dquot()
4822 * dquot_initialize() down(dqio_mutex)
4823 * down(dqio_mutex) jbd2_journal_start()
4829 static inline struct inode *dquot_to_inode(struct dquot *dquot)
4831 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
4834 static int ext4_write_dquot(struct dquot *dquot)
4838 struct inode *inode;
4840 inode = dquot_to_inode(dquot);
4841 handle = ext4_journal_start(inode,
4842 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
4844 return PTR_ERR(handle);
4845 ret = dquot_commit(dquot);
4846 err = ext4_journal_stop(handle);
4852 static int ext4_acquire_dquot(struct dquot *dquot)
4857 handle = ext4_journal_start(dquot_to_inode(dquot),
4858 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
4860 return PTR_ERR(handle);
4861 ret = dquot_acquire(dquot);
4862 err = ext4_journal_stop(handle);
4868 static int ext4_release_dquot(struct dquot *dquot)
4873 handle = ext4_journal_start(dquot_to_inode(dquot),
4874 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
4875 if (IS_ERR(handle)) {
4876 /* Release dquot anyway to avoid endless cycle in dqput() */
4877 dquot_release(dquot);
4878 return PTR_ERR(handle);
4880 ret = dquot_release(dquot);
4881 err = ext4_journal_stop(handle);
4887 static int ext4_mark_dquot_dirty(struct dquot *dquot)
4889 /* Are we journaling quotas? */
4890 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4891 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
4892 dquot_mark_dquot_dirty(dquot);
4893 return ext4_write_dquot(dquot);
4895 return dquot_mark_dquot_dirty(dquot);
4899 static int ext4_write_info(struct super_block *sb, int type)
4904 /* Data block + inode block */
4905 handle = ext4_journal_start(sb->s_root->d_inode, 2);
4907 return PTR_ERR(handle);
4908 ret = dquot_commit_info(sb, type);
4909 err = ext4_journal_stop(handle);
4916 * Turn on quotas during mount time - we need to find
4917 * the quota file and such...
4919 static int ext4_quota_on_mount(struct super_block *sb, int type)
4921 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4922 EXT4_SB(sb)->s_jquota_fmt, type);
4926 * Standard function to be called on quota_on
4928 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
4933 if (!test_opt(sb, QUOTA))
4936 /* Quotafile not on the same filesystem? */
4937 if (path->dentry->d_sb != sb)
4939 /* Journaling quota? */
4940 if (EXT4_SB(sb)->s_qf_names[type]) {
4941 /* Quotafile not in fs root? */
4942 if (path->dentry->d_parent != sb->s_root)
4943 ext4_msg(sb, KERN_WARNING,
4944 "Quota file not on filesystem root. "
4945 "Journaled quota will not work");
4949 * When we journal data on quota file, we have to flush journal to see
4950 * all updates to the file when we bypass pagecache...
4952 if (EXT4_SB(sb)->s_journal &&
4953 ext4_should_journal_data(path->dentry->d_inode)) {
4955 * We don't need to lock updates but journal_flush() could
4956 * otherwise be livelocked...
4958 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
4959 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
4960 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4965 return dquot_quota_on(sb, type, format_id, path);
4968 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
4972 struct inode *qf_inode;
4973 unsigned long qf_inums[MAXQUOTAS] = {
4974 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
4975 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
4978 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
4980 if (!qf_inums[type])
4983 qf_inode = ext4_iget(sb, qf_inums[type]);
4984 if (IS_ERR(qf_inode)) {
4985 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
4986 return PTR_ERR(qf_inode);
4989 err = dquot_enable(qf_inode, type, format_id, flags);
4995 /* Enable usage tracking for all quota types. */
4996 static int ext4_enable_quotas(struct super_block *sb)
4999 unsigned long qf_inums[MAXQUOTAS] = {
5000 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5001 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5004 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5005 for (type = 0; type < MAXQUOTAS; type++) {
5006 if (qf_inums[type]) {
5007 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5008 DQUOT_USAGE_ENABLED);
5011 "Failed to enable quota (type=%d) "
5012 "tracking. Please run e2fsck to fix.",
5022 * quota_on function that is used when QUOTA feature is set.
5024 static int ext4_quota_on_sysfile(struct super_block *sb, int type,
5027 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5031 * USAGE was enabled at mount time. Only need to enable LIMITS now.
5033 return ext4_quota_enable(sb, type, format_id, DQUOT_LIMITS_ENABLED);
5036 static int ext4_quota_off(struct super_block *sb, int type)
5038 struct inode *inode = sb_dqopt(sb)->files[type];
5041 /* Force all delayed allocation blocks to be allocated.
5042 * Caller already holds s_umount sem */
5043 if (test_opt(sb, DELALLOC))
5044 sync_filesystem(sb);
5049 /* Update modification times of quota files when userspace can
5050 * start looking at them */
5051 handle = ext4_journal_start(inode, 1);
5054 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5055 ext4_mark_inode_dirty(handle, inode);
5056 ext4_journal_stop(handle);
5059 return dquot_quota_off(sb, type);
5063 * quota_off function that is used when QUOTA feature is set.
5065 static int ext4_quota_off_sysfile(struct super_block *sb, int type)
5067 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5070 /* Disable only the limits. */
5071 return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
5074 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5075 * acquiring the locks... As quota files are never truncated and quota code
5076 * itself serializes the operations (and no one else should touch the files)
5077 * we don't have to be afraid of races */
5078 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5079 size_t len, loff_t off)
5081 struct inode *inode = sb_dqopt(sb)->files[type];
5082 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5084 int offset = off & (sb->s_blocksize - 1);
5087 struct buffer_head *bh;
5088 loff_t i_size = i_size_read(inode);
5092 if (off+len > i_size)
5095 while (toread > 0) {
5096 tocopy = sb->s_blocksize - offset < toread ?
5097 sb->s_blocksize - offset : toread;
5098 bh = ext4_bread(NULL, inode, blk, 0, &err);
5101 if (!bh) /* A hole? */
5102 memset(data, 0, tocopy);
5104 memcpy(data, bh->b_data+offset, tocopy);
5114 /* Write to quotafile (we know the transaction is already started and has
5115 * enough credits) */
5116 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5117 const char *data, size_t len, loff_t off)
5119 struct inode *inode = sb_dqopt(sb)->files[type];
5120 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5122 int offset = off & (sb->s_blocksize - 1);
5123 struct buffer_head *bh;
5124 handle_t *handle = journal_current_handle();
5126 if (EXT4_SB(sb)->s_journal && !handle) {
5127 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5128 " cancelled because transaction is not started",
5129 (unsigned long long)off, (unsigned long long)len);
5133 * Since we account only one data block in transaction credits,
5134 * then it is impossible to cross a block boundary.
5136 if (sb->s_blocksize - offset < len) {
5137 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5138 " cancelled because not block aligned",
5139 (unsigned long long)off, (unsigned long long)len);
5143 bh = ext4_bread(handle, inode, blk, 1, &err);
5146 err = ext4_journal_get_write_access(handle, bh);
5152 memcpy(bh->b_data+offset, data, len);
5153 flush_dcache_page(bh->b_page);
5155 err = ext4_handle_dirty_metadata(handle, NULL, bh);
5160 if (inode->i_size < off + len) {
5161 i_size_write(inode, off + len);
5162 EXT4_I(inode)->i_disksize = inode->i_size;
5163 ext4_mark_inode_dirty(handle, inode);
5170 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5171 const char *dev_name, void *data)
5173 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5176 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5177 static inline void register_as_ext2(void)
5179 int err = register_filesystem(&ext2_fs_type);
5182 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5185 static inline void unregister_as_ext2(void)
5187 unregister_filesystem(&ext2_fs_type);
5190 static inline int ext2_feature_set_ok(struct super_block *sb)
5192 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5194 if (sb->s_flags & MS_RDONLY)
5196 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5200 MODULE_ALIAS("ext2");
5202 static inline void register_as_ext2(void) { }
5203 static inline void unregister_as_ext2(void) { }
5204 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5207 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5208 static inline void register_as_ext3(void)
5210 int err = register_filesystem(&ext3_fs_type);
5213 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5216 static inline void unregister_as_ext3(void)
5218 unregister_filesystem(&ext3_fs_type);
5221 static inline int ext3_feature_set_ok(struct super_block *sb)
5223 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5225 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5227 if (sb->s_flags & MS_RDONLY)
5229 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5233 MODULE_ALIAS("ext3");
5235 static inline void register_as_ext3(void) { }
5236 static inline void unregister_as_ext3(void) { }
5237 static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
5240 static struct file_system_type ext4_fs_type = {
5241 .owner = THIS_MODULE,
5243 .mount = ext4_mount,
5244 .kill_sb = kill_block_super,
5245 .fs_flags = FS_REQUIRES_DEV,
5248 static int __init ext4_init_feat_adverts(void)
5250 struct ext4_features *ef;
5253 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5257 ef->f_kobj.kset = ext4_kset;
5258 init_completion(&ef->f_kobj_unregister);
5259 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5272 static void ext4_exit_feat_adverts(void)
5274 kobject_put(&ext4_feat->f_kobj);
5275 wait_for_completion(&ext4_feat->f_kobj_unregister);
5279 /* Shared across all ext4 file systems */
5280 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5281 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5283 static int __init ext4_init_fs(void)
5287 ext4_li_info = NULL;
5288 mutex_init(&ext4_li_mtx);
5290 ext4_check_flag_values();
5292 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5293 mutex_init(&ext4__aio_mutex[i]);
5294 init_waitqueue_head(&ext4__ioend_wq[i]);
5297 err = ext4_init_pageio();
5300 err = ext4_init_system_zone();
5303 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
5308 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
5310 err = ext4_init_feat_adverts();
5314 err = ext4_init_mballoc();
5318 err = ext4_init_xattr();
5321 err = init_inodecache();
5326 err = register_filesystem(&ext4_fs_type);
5332 unregister_as_ext2();
5333 unregister_as_ext3();
5334 destroy_inodecache();
5338 ext4_exit_mballoc();
5340 ext4_exit_feat_adverts();
5343 remove_proc_entry("fs/ext4", NULL);
5344 kset_unregister(ext4_kset);
5346 ext4_exit_system_zone();
5352 static void __exit ext4_exit_fs(void)
5354 ext4_destroy_lazyinit_thread();
5355 unregister_as_ext2();
5356 unregister_as_ext3();
5357 unregister_filesystem(&ext4_fs_type);
5358 destroy_inodecache();
5360 ext4_exit_mballoc();
5361 ext4_exit_feat_adverts();
5362 remove_proc_entry("fs/ext4", NULL);
5363 kset_unregister(ext4_kset);
5364 ext4_exit_system_zone();
5368 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5369 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5370 MODULE_LICENSE("GPL");
5371 module_init(ext4_init_fs)
5372 module_exit(ext4_exit_fs)