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/slab.h>
25 #include <linux/init.h>
26 #include <linux/blkdev.h>
27 #include <linux/backing-dev.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/ctype.h>
38 #include <linux/log2.h>
39 #include <linux/crc16.h>
40 #include <linux/cleancache.h>
41 #include <asm/uaccess.h>
43 #include <linux/kthread.h>
44 #include <linux/freezer.h>
47 #include "ext4_extents.h" /* Needed for trace points definition */
48 #include "ext4_jbd2.h"
53 #define CREATE_TRACE_POINTS
54 #include <trace/events/ext4.h>
56 static struct ext4_lazy_init *ext4_li_info;
57 static struct mutex ext4_li_mtx;
58 static int ext4_mballoc_ready;
59 static struct ratelimit_state ext4_mount_msg_ratelimit;
61 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
62 unsigned long journal_devnum);
63 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
64 static int ext4_commit_super(struct super_block *sb, int sync);
65 static void ext4_mark_recovery_complete(struct super_block *sb,
66 struct ext4_super_block *es);
67 static void ext4_clear_journal_err(struct super_block *sb,
68 struct ext4_super_block *es);
69 static int ext4_sync_fs(struct super_block *sb, int wait);
70 static int ext4_remount(struct super_block *sb, int *flags, char *data);
71 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
72 static int ext4_unfreeze(struct super_block *sb);
73 static int ext4_freeze(struct super_block *sb);
74 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
75 const char *dev_name, void *data);
76 static inline int ext2_feature_set_ok(struct super_block *sb);
77 static inline int ext3_feature_set_ok(struct super_block *sb);
78 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
79 static void ext4_destroy_lazyinit_thread(void);
80 static void ext4_unregister_li_request(struct super_block *sb);
81 static void ext4_clear_request_list(void);
86 * Note the difference between i_mmap_sem (EXT4_I(inode)->i_mmap_sem) and
87 * i_mmap_rwsem (inode->i_mmap_rwsem)!
90 * mmap_sem -> sb_start_pagefault -> i_mmap_sem (r) -> transaction start ->
91 * page lock -> i_data_sem (rw)
93 * buffered write path:
94 * sb_start_write -> i_mutex -> mmap_sem
95 * sb_start_write -> i_mutex -> transaction start -> page lock ->
99 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) ->
100 * i_mmap_rwsem (w) -> page lock
101 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) ->
102 * transaction start -> i_data_sem (rw)
105 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) -> mmap_sem
106 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) ->
107 * transaction start -> i_data_sem (rw)
110 * transaction start -> page lock(s) -> i_data_sem (rw)
113 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
114 static struct file_system_type ext2_fs_type = {
115 .owner = THIS_MODULE,
118 .kill_sb = kill_block_super,
119 .fs_flags = FS_REQUIRES_DEV,
121 MODULE_ALIAS_FS("ext2");
122 MODULE_ALIAS("ext2");
123 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
125 #define IS_EXT2_SB(sb) (0)
129 static struct file_system_type ext3_fs_type = {
130 .owner = THIS_MODULE,
133 .kill_sb = kill_block_super,
134 .fs_flags = FS_REQUIRES_DEV,
136 MODULE_ALIAS_FS("ext3");
137 MODULE_ALIAS("ext3");
138 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
140 static int ext4_verify_csum_type(struct super_block *sb,
141 struct ext4_super_block *es)
143 if (!ext4_has_feature_metadata_csum(sb))
146 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
149 static __le32 ext4_superblock_csum(struct super_block *sb,
150 struct ext4_super_block *es)
152 struct ext4_sb_info *sbi = EXT4_SB(sb);
153 int offset = offsetof(struct ext4_super_block, s_checksum);
156 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
158 return cpu_to_le32(csum);
161 static int ext4_superblock_csum_verify(struct super_block *sb,
162 struct ext4_super_block *es)
164 if (!ext4_has_metadata_csum(sb))
167 return es->s_checksum == ext4_superblock_csum(sb, es);
170 void ext4_superblock_csum_set(struct super_block *sb)
172 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
174 if (!ext4_has_metadata_csum(sb))
177 es->s_checksum = ext4_superblock_csum(sb, es);
180 void *ext4_kvmalloc(size_t size, gfp_t flags)
184 ret = kmalloc(size, flags | __GFP_NOWARN);
186 ret = __vmalloc(size, flags, PAGE_KERNEL);
190 void *ext4_kvzalloc(size_t size, gfp_t flags)
194 ret = kzalloc(size, flags | __GFP_NOWARN);
196 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
200 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
201 struct ext4_group_desc *bg)
203 return le32_to_cpu(bg->bg_block_bitmap_lo) |
204 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
205 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
208 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
209 struct ext4_group_desc *bg)
211 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
212 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
213 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
216 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
217 struct ext4_group_desc *bg)
219 return le32_to_cpu(bg->bg_inode_table_lo) |
220 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
221 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
224 __u32 ext4_free_group_clusters(struct super_block *sb,
225 struct ext4_group_desc *bg)
227 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
228 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
229 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
232 __u32 ext4_free_inodes_count(struct super_block *sb,
233 struct ext4_group_desc *bg)
235 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
236 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
237 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
240 __u32 ext4_used_dirs_count(struct super_block *sb,
241 struct ext4_group_desc *bg)
243 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
244 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
245 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
248 __u32 ext4_itable_unused_count(struct super_block *sb,
249 struct ext4_group_desc *bg)
251 return le16_to_cpu(bg->bg_itable_unused_lo) |
252 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
253 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
256 void ext4_block_bitmap_set(struct super_block *sb,
257 struct ext4_group_desc *bg, ext4_fsblk_t blk)
259 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
260 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
261 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
264 void ext4_inode_bitmap_set(struct super_block *sb,
265 struct ext4_group_desc *bg, ext4_fsblk_t blk)
267 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
268 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
269 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
272 void ext4_inode_table_set(struct super_block *sb,
273 struct ext4_group_desc *bg, ext4_fsblk_t blk)
275 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
276 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
277 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
280 void ext4_free_group_clusters_set(struct super_block *sb,
281 struct ext4_group_desc *bg, __u32 count)
283 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
284 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
285 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
288 void ext4_free_inodes_set(struct super_block *sb,
289 struct ext4_group_desc *bg, __u32 count)
291 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
292 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
293 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
296 void ext4_used_dirs_set(struct super_block *sb,
297 struct ext4_group_desc *bg, __u32 count)
299 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
300 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
301 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
304 void ext4_itable_unused_set(struct super_block *sb,
305 struct ext4_group_desc *bg, __u32 count)
307 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
308 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
309 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
313 static void __save_error_info(struct super_block *sb, const char *func,
316 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
318 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
319 if (bdev_read_only(sb->s_bdev))
321 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
322 es->s_last_error_time = cpu_to_le32(get_seconds());
323 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
324 es->s_last_error_line = cpu_to_le32(line);
325 if (!es->s_first_error_time) {
326 es->s_first_error_time = es->s_last_error_time;
327 strncpy(es->s_first_error_func, func,
328 sizeof(es->s_first_error_func));
329 es->s_first_error_line = cpu_to_le32(line);
330 es->s_first_error_ino = es->s_last_error_ino;
331 es->s_first_error_block = es->s_last_error_block;
334 * Start the daily error reporting function if it hasn't been
337 if (!es->s_error_count)
338 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
339 le32_add_cpu(&es->s_error_count, 1);
342 static void save_error_info(struct super_block *sb, const char *func,
345 __save_error_info(sb, func, line);
346 ext4_commit_super(sb, 1);
350 * The del_gendisk() function uninitializes the disk-specific data
351 * structures, including the bdi structure, without telling anyone
352 * else. Once this happens, any attempt to call mark_buffer_dirty()
353 * (for example, by ext4_commit_super), will cause a kernel OOPS.
354 * This is a kludge to prevent these oops until we can put in a proper
355 * hook in del_gendisk() to inform the VFS and file system layers.
357 static int block_device_ejected(struct super_block *sb)
359 struct inode *bd_inode = sb->s_bdev->bd_inode;
360 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
362 return bdi->dev == NULL;
365 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
367 struct super_block *sb = journal->j_private;
368 struct ext4_sb_info *sbi = EXT4_SB(sb);
369 int error = is_journal_aborted(journal);
370 struct ext4_journal_cb_entry *jce;
372 BUG_ON(txn->t_state == T_FINISHED);
373 spin_lock(&sbi->s_md_lock);
374 while (!list_empty(&txn->t_private_list)) {
375 jce = list_entry(txn->t_private_list.next,
376 struct ext4_journal_cb_entry, jce_list);
377 list_del_init(&jce->jce_list);
378 spin_unlock(&sbi->s_md_lock);
379 jce->jce_func(sb, jce, error);
380 spin_lock(&sbi->s_md_lock);
382 spin_unlock(&sbi->s_md_lock);
385 /* Deal with the reporting of failure conditions on a filesystem such as
386 * inconsistencies detected or read IO failures.
388 * On ext2, we can store the error state of the filesystem in the
389 * superblock. That is not possible on ext4, because we may have other
390 * write ordering constraints on the superblock which prevent us from
391 * writing it out straight away; and given that the journal is about to
392 * be aborted, we can't rely on the current, or future, transactions to
393 * write out the superblock safely.
395 * We'll just use the jbd2_journal_abort() error code to record an error in
396 * the journal instead. On recovery, the journal will complain about
397 * that error until we've noted it down and cleared it.
400 static void ext4_handle_error(struct super_block *sb)
402 if (sb->s_flags & MS_RDONLY)
405 if (!test_opt(sb, ERRORS_CONT)) {
406 journal_t *journal = EXT4_SB(sb)->s_journal;
408 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
410 jbd2_journal_abort(journal, -EIO);
412 if (test_opt(sb, ERRORS_RO)) {
413 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
415 * Make sure updated value of ->s_mount_flags will be visible
416 * before ->s_flags update
419 sb->s_flags |= MS_RDONLY;
421 if (test_opt(sb, ERRORS_PANIC)) {
422 if (EXT4_SB(sb)->s_journal &&
423 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
425 panic("EXT4-fs (device %s): panic forced after error\n",
430 #define ext4_error_ratelimit(sb) \
431 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
434 void __ext4_error(struct super_block *sb, const char *function,
435 unsigned int line, const char *fmt, ...)
437 struct va_format vaf;
440 if (ext4_error_ratelimit(sb)) {
445 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
446 sb->s_id, function, line, current->comm, &vaf);
449 save_error_info(sb, function, line);
450 ext4_handle_error(sb);
453 void __ext4_error_inode(struct inode *inode, const char *function,
454 unsigned int line, ext4_fsblk_t block,
455 const char *fmt, ...)
458 struct va_format vaf;
459 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
461 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
462 es->s_last_error_block = cpu_to_le64(block);
463 if (ext4_error_ratelimit(inode->i_sb)) {
468 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
469 "inode #%lu: block %llu: comm %s: %pV\n",
470 inode->i_sb->s_id, function, line, inode->i_ino,
471 block, current->comm, &vaf);
473 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
474 "inode #%lu: comm %s: %pV\n",
475 inode->i_sb->s_id, function, line, inode->i_ino,
476 current->comm, &vaf);
479 save_error_info(inode->i_sb, function, line);
480 ext4_handle_error(inode->i_sb);
483 void __ext4_error_file(struct file *file, const char *function,
484 unsigned int line, ext4_fsblk_t block,
485 const char *fmt, ...)
488 struct va_format vaf;
489 struct ext4_super_block *es;
490 struct inode *inode = file_inode(file);
491 char pathname[80], *path;
493 es = EXT4_SB(inode->i_sb)->s_es;
494 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
495 if (ext4_error_ratelimit(inode->i_sb)) {
496 path = file_path(file, pathname, sizeof(pathname));
504 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
505 "block %llu: comm %s: path %s: %pV\n",
506 inode->i_sb->s_id, function, line, inode->i_ino,
507 block, current->comm, path, &vaf);
510 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
511 "comm %s: path %s: %pV\n",
512 inode->i_sb->s_id, function, line, inode->i_ino,
513 current->comm, path, &vaf);
516 save_error_info(inode->i_sb, function, line);
517 ext4_handle_error(inode->i_sb);
520 const char *ext4_decode_error(struct super_block *sb, int errno,
527 errstr = "Corrupt filesystem";
530 errstr = "Filesystem failed CRC";
533 errstr = "IO failure";
536 errstr = "Out of memory";
539 if (!sb || (EXT4_SB(sb)->s_journal &&
540 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
541 errstr = "Journal has aborted";
543 errstr = "Readonly filesystem";
546 /* If the caller passed in an extra buffer for unknown
547 * errors, textualise them now. Else we just return
550 /* Check for truncated error codes... */
551 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
560 /* __ext4_std_error decodes expected errors from journaling functions
561 * automatically and invokes the appropriate error response. */
563 void __ext4_std_error(struct super_block *sb, const char *function,
564 unsigned int line, int errno)
569 /* Special case: if the error is EROFS, and we're not already
570 * inside a transaction, then there's really no point in logging
572 if (errno == -EROFS && journal_current_handle() == NULL &&
573 (sb->s_flags & MS_RDONLY))
576 if (ext4_error_ratelimit(sb)) {
577 errstr = ext4_decode_error(sb, errno, nbuf);
578 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
579 sb->s_id, function, line, errstr);
582 save_error_info(sb, function, line);
583 ext4_handle_error(sb);
587 * ext4_abort is a much stronger failure handler than ext4_error. The
588 * abort function may be used to deal with unrecoverable failures such
589 * as journal IO errors or ENOMEM at a critical moment in log management.
591 * We unconditionally force the filesystem into an ABORT|READONLY state,
592 * unless the error response on the fs has been set to panic in which
593 * case we take the easy way out and panic immediately.
596 void __ext4_abort(struct super_block *sb, const char *function,
597 unsigned int line, const char *fmt, ...)
601 save_error_info(sb, function, line);
603 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
609 if ((sb->s_flags & MS_RDONLY) == 0) {
610 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
611 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
613 * Make sure updated value of ->s_mount_flags will be visible
614 * before ->s_flags update
617 sb->s_flags |= MS_RDONLY;
618 if (EXT4_SB(sb)->s_journal)
619 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
620 save_error_info(sb, function, line);
622 if (test_opt(sb, ERRORS_PANIC)) {
623 if (EXT4_SB(sb)->s_journal &&
624 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
626 panic("EXT4-fs panic from previous error\n");
630 void __ext4_msg(struct super_block *sb,
631 const char *prefix, const char *fmt, ...)
633 struct va_format vaf;
636 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
642 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
646 #define ext4_warning_ratelimit(sb) \
647 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \
650 void __ext4_warning(struct super_block *sb, const char *function,
651 unsigned int line, const char *fmt, ...)
653 struct va_format vaf;
656 if (!ext4_warning_ratelimit(sb))
662 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
663 sb->s_id, function, line, &vaf);
667 void __ext4_warning_inode(const struct inode *inode, const char *function,
668 unsigned int line, const char *fmt, ...)
670 struct va_format vaf;
673 if (!ext4_warning_ratelimit(inode->i_sb))
679 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
680 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
681 function, line, inode->i_ino, current->comm, &vaf);
685 void __ext4_grp_locked_error(const char *function, unsigned int line,
686 struct super_block *sb, ext4_group_t grp,
687 unsigned long ino, ext4_fsblk_t block,
688 const char *fmt, ...)
692 struct va_format vaf;
694 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
696 es->s_last_error_ino = cpu_to_le32(ino);
697 es->s_last_error_block = cpu_to_le64(block);
698 __save_error_info(sb, function, line);
700 if (ext4_error_ratelimit(sb)) {
704 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
705 sb->s_id, function, line, grp);
707 printk(KERN_CONT "inode %lu: ", ino);
709 printk(KERN_CONT "block %llu:",
710 (unsigned long long) block);
711 printk(KERN_CONT "%pV\n", &vaf);
715 if (test_opt(sb, ERRORS_CONT)) {
716 ext4_commit_super(sb, 0);
720 ext4_unlock_group(sb, grp);
721 ext4_handle_error(sb);
723 * We only get here in the ERRORS_RO case; relocking the group
724 * may be dangerous, but nothing bad will happen since the
725 * filesystem will have already been marked read/only and the
726 * journal has been aborted. We return 1 as a hint to callers
727 * who might what to use the return value from
728 * ext4_grp_locked_error() to distinguish between the
729 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
730 * aggressively from the ext4 function in question, with a
731 * more appropriate error code.
733 ext4_lock_group(sb, grp);
737 void ext4_update_dynamic_rev(struct super_block *sb)
739 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
741 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
745 "updating to rev %d because of new feature flag, "
746 "running e2fsck is recommended",
749 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
750 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
751 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
752 /* leave es->s_feature_*compat flags alone */
753 /* es->s_uuid will be set by e2fsck if empty */
756 * The rest of the superblock fields should be zero, and if not it
757 * means they are likely already in use, so leave them alone. We
758 * can leave it up to e2fsck to clean up any inconsistencies there.
763 * Open the external journal device
765 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
767 struct block_device *bdev;
768 char b[BDEVNAME_SIZE];
770 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
776 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
777 __bdevname(dev, b), PTR_ERR(bdev));
782 * Release the journal device
784 static void ext4_blkdev_put(struct block_device *bdev)
786 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
789 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
791 struct block_device *bdev;
792 bdev = sbi->journal_bdev;
794 ext4_blkdev_put(bdev);
795 sbi->journal_bdev = NULL;
799 static inline struct inode *orphan_list_entry(struct list_head *l)
801 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
804 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
808 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
809 le32_to_cpu(sbi->s_es->s_last_orphan));
811 printk(KERN_ERR "sb_info orphan list:\n");
812 list_for_each(l, &sbi->s_orphan) {
813 struct inode *inode = orphan_list_entry(l);
815 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
816 inode->i_sb->s_id, inode->i_ino, inode,
817 inode->i_mode, inode->i_nlink,
822 static void ext4_put_super(struct super_block *sb)
824 struct ext4_sb_info *sbi = EXT4_SB(sb);
825 struct ext4_super_block *es = sbi->s_es;
828 ext4_unregister_li_request(sb);
829 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
831 flush_workqueue(sbi->rsv_conversion_wq);
832 destroy_workqueue(sbi->rsv_conversion_wq);
834 if (sbi->s_journal) {
835 err = jbd2_journal_destroy(sbi->s_journal);
836 sbi->s_journal = NULL;
838 ext4_abort(sb, "Couldn't clean up the journal");
841 ext4_unregister_sysfs(sb);
842 ext4_es_unregister_shrinker(sbi);
843 del_timer_sync(&sbi->s_err_report);
844 ext4_release_system_zone(sb);
846 ext4_ext_release(sb);
847 ext4_xattr_put_super(sb);
849 if (!(sb->s_flags & MS_RDONLY)) {
850 ext4_clear_feature_journal_needs_recovery(sb);
851 es->s_state = cpu_to_le16(sbi->s_mount_state);
853 if (!(sb->s_flags & MS_RDONLY))
854 ext4_commit_super(sb, 1);
856 for (i = 0; i < sbi->s_gdb_count; i++)
857 brelse(sbi->s_group_desc[i]);
858 kvfree(sbi->s_group_desc);
859 kvfree(sbi->s_flex_groups);
860 percpu_counter_destroy(&sbi->s_freeclusters_counter);
861 percpu_counter_destroy(&sbi->s_freeinodes_counter);
862 percpu_counter_destroy(&sbi->s_dirs_counter);
863 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
866 for (i = 0; i < EXT4_MAXQUOTAS; i++)
867 kfree(sbi->s_qf_names[i]);
870 /* Debugging code just in case the in-memory inode orphan list
871 * isn't empty. The on-disk one can be non-empty if we've
872 * detected an error and taken the fs readonly, but the
873 * in-memory list had better be clean by this point. */
874 if (!list_empty(&sbi->s_orphan))
875 dump_orphan_list(sb, sbi);
876 J_ASSERT(list_empty(&sbi->s_orphan));
878 sync_blockdev(sb->s_bdev);
879 invalidate_bdev(sb->s_bdev);
880 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
882 * Invalidate the journal device's buffers. We don't want them
883 * floating about in memory - the physical journal device may
884 * hotswapped, and it breaks the `ro-after' testing code.
886 sync_blockdev(sbi->journal_bdev);
887 invalidate_bdev(sbi->journal_bdev);
888 ext4_blkdev_remove(sbi);
890 if (sbi->s_mb_cache) {
891 ext4_xattr_destroy_cache(sbi->s_mb_cache);
892 sbi->s_mb_cache = NULL;
895 kthread_stop(sbi->s_mmp_tsk);
896 sb->s_fs_info = NULL;
898 * Now that we are completely done shutting down the
899 * superblock, we need to actually destroy the kobject.
901 kobject_put(&sbi->s_kobj);
902 wait_for_completion(&sbi->s_kobj_unregister);
903 if (sbi->s_chksum_driver)
904 crypto_free_shash(sbi->s_chksum_driver);
905 kfree(sbi->s_blockgroup_lock);
909 static struct kmem_cache *ext4_inode_cachep;
912 * Called inside transaction, so use GFP_NOFS
914 static struct inode *ext4_alloc_inode(struct super_block *sb)
916 struct ext4_inode_info *ei;
918 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
922 ei->vfs_inode.i_version = 1;
923 spin_lock_init(&ei->i_raw_lock);
924 INIT_LIST_HEAD(&ei->i_prealloc_list);
925 spin_lock_init(&ei->i_prealloc_lock);
926 ext4_es_init_tree(&ei->i_es_tree);
927 rwlock_init(&ei->i_es_lock);
928 INIT_LIST_HEAD(&ei->i_es_list);
931 ei->i_es_shrink_lblk = 0;
932 ei->i_reserved_data_blocks = 0;
933 ei->i_reserved_meta_blocks = 0;
934 ei->i_allocated_meta_blocks = 0;
935 ei->i_da_metadata_calc_len = 0;
936 ei->i_da_metadata_calc_last_lblock = 0;
937 spin_lock_init(&(ei->i_block_reservation_lock));
939 ei->i_reserved_quota = 0;
940 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
943 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
944 spin_lock_init(&ei->i_completed_io_lock);
946 ei->i_datasync_tid = 0;
947 atomic_set(&ei->i_ioend_count, 0);
948 atomic_set(&ei->i_unwritten, 0);
949 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
950 #ifdef CONFIG_EXT4_FS_ENCRYPTION
951 ei->i_crypt_info = NULL;
953 return &ei->vfs_inode;
956 static int ext4_drop_inode(struct inode *inode)
958 int drop = generic_drop_inode(inode);
960 trace_ext4_drop_inode(inode, drop);
964 static void ext4_i_callback(struct rcu_head *head)
966 struct inode *inode = container_of(head, struct inode, i_rcu);
967 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
970 static void ext4_destroy_inode(struct inode *inode)
972 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
973 ext4_msg(inode->i_sb, KERN_ERR,
974 "Inode %lu (%p): orphan list check failed!",
975 inode->i_ino, EXT4_I(inode));
976 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
977 EXT4_I(inode), sizeof(struct ext4_inode_info),
981 call_rcu(&inode->i_rcu, ext4_i_callback);
984 static void init_once(void *foo)
986 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
988 INIT_LIST_HEAD(&ei->i_orphan);
989 init_rwsem(&ei->xattr_sem);
990 init_rwsem(&ei->i_data_sem);
991 init_rwsem(&ei->i_mmap_sem);
992 inode_init_once(&ei->vfs_inode);
995 static int __init init_inodecache(void)
997 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
998 sizeof(struct ext4_inode_info),
999 0, (SLAB_RECLAIM_ACCOUNT|
1000 SLAB_MEM_SPREAD|SLAB_ACCOUNT),
1002 if (ext4_inode_cachep == NULL)
1007 static void destroy_inodecache(void)
1010 * Make sure all delayed rcu free inodes are flushed before we
1014 kmem_cache_destroy(ext4_inode_cachep);
1017 void ext4_clear_inode(struct inode *inode)
1019 invalidate_inode_buffers(inode);
1022 ext4_discard_preallocations(inode);
1023 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
1024 if (EXT4_I(inode)->jinode) {
1025 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1026 EXT4_I(inode)->jinode);
1027 jbd2_free_inode(EXT4_I(inode)->jinode);
1028 EXT4_I(inode)->jinode = NULL;
1030 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1031 if (EXT4_I(inode)->i_crypt_info)
1032 ext4_free_encryption_info(inode, EXT4_I(inode)->i_crypt_info);
1036 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1037 u64 ino, u32 generation)
1039 struct inode *inode;
1041 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1042 return ERR_PTR(-ESTALE);
1043 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1044 return ERR_PTR(-ESTALE);
1046 /* iget isn't really right if the inode is currently unallocated!!
1048 * ext4_read_inode will return a bad_inode if the inode had been
1049 * deleted, so we should be safe.
1051 * Currently we don't know the generation for parent directory, so
1052 * a generation of 0 means "accept any"
1054 inode = ext4_iget_normal(sb, ino);
1056 return ERR_CAST(inode);
1057 if (generation && inode->i_generation != generation) {
1059 return ERR_PTR(-ESTALE);
1065 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1066 int fh_len, int fh_type)
1068 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1069 ext4_nfs_get_inode);
1072 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1073 int fh_len, int fh_type)
1075 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1076 ext4_nfs_get_inode);
1080 * Try to release metadata pages (indirect blocks, directories) which are
1081 * mapped via the block device. Since these pages could have journal heads
1082 * which would prevent try_to_free_buffers() from freeing them, we must use
1083 * jbd2 layer's try_to_free_buffers() function to release them.
1085 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1088 journal_t *journal = EXT4_SB(sb)->s_journal;
1090 WARN_ON(PageChecked(page));
1091 if (!page_has_buffers(page))
1094 return jbd2_journal_try_to_free_buffers(journal, page,
1095 wait & ~__GFP_DIRECT_RECLAIM);
1096 return try_to_free_buffers(page);
1100 static char *quotatypes[] = INITQFNAMES;
1101 #define QTYPE2NAME(t) (quotatypes[t])
1103 static int ext4_write_dquot(struct dquot *dquot);
1104 static int ext4_acquire_dquot(struct dquot *dquot);
1105 static int ext4_release_dquot(struct dquot *dquot);
1106 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1107 static int ext4_write_info(struct super_block *sb, int type);
1108 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1110 static int ext4_quota_off(struct super_block *sb, int type);
1111 static int ext4_quota_on_mount(struct super_block *sb, int type);
1112 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1113 size_t len, loff_t off);
1114 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1115 const char *data, size_t len, loff_t off);
1116 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1117 unsigned int flags);
1118 static int ext4_enable_quotas(struct super_block *sb);
1120 static struct dquot **ext4_get_dquots(struct inode *inode)
1122 return EXT4_I(inode)->i_dquot;
1125 static const struct dquot_operations ext4_quota_operations = {
1126 .get_reserved_space = ext4_get_reserved_space,
1127 .write_dquot = ext4_write_dquot,
1128 .acquire_dquot = ext4_acquire_dquot,
1129 .release_dquot = ext4_release_dquot,
1130 .mark_dirty = ext4_mark_dquot_dirty,
1131 .write_info = ext4_write_info,
1132 .alloc_dquot = dquot_alloc,
1133 .destroy_dquot = dquot_destroy,
1134 .get_projid = ext4_get_projid,
1135 .get_next_id = dquot_get_next_id,
1138 static const struct quotactl_ops ext4_qctl_operations = {
1139 .quota_on = ext4_quota_on,
1140 .quota_off = ext4_quota_off,
1141 .quota_sync = dquot_quota_sync,
1142 .get_state = dquot_get_state,
1143 .set_info = dquot_set_dqinfo,
1144 .get_dqblk = dquot_get_dqblk,
1145 .set_dqblk = dquot_set_dqblk
1149 static const struct super_operations ext4_sops = {
1150 .alloc_inode = ext4_alloc_inode,
1151 .destroy_inode = ext4_destroy_inode,
1152 .write_inode = ext4_write_inode,
1153 .dirty_inode = ext4_dirty_inode,
1154 .drop_inode = ext4_drop_inode,
1155 .evict_inode = ext4_evict_inode,
1156 .put_super = ext4_put_super,
1157 .sync_fs = ext4_sync_fs,
1158 .freeze_fs = ext4_freeze,
1159 .unfreeze_fs = ext4_unfreeze,
1160 .statfs = ext4_statfs,
1161 .remount_fs = ext4_remount,
1162 .show_options = ext4_show_options,
1164 .quota_read = ext4_quota_read,
1165 .quota_write = ext4_quota_write,
1166 .get_dquots = ext4_get_dquots,
1168 .bdev_try_to_free_page = bdev_try_to_free_page,
1171 static const struct export_operations ext4_export_ops = {
1172 .fh_to_dentry = ext4_fh_to_dentry,
1173 .fh_to_parent = ext4_fh_to_parent,
1174 .get_parent = ext4_get_parent,
1178 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1179 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1180 Opt_nouid32, Opt_debug, Opt_removed,
1181 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1182 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1183 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1184 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1185 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1186 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1187 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1188 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1189 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1190 Opt_usrquota, Opt_grpquota, Opt_i_version, Opt_dax,
1191 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1192 Opt_lazytime, Opt_nolazytime,
1193 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1194 Opt_inode_readahead_blks, Opt_journal_ioprio,
1195 Opt_dioread_nolock, Opt_dioread_lock,
1196 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1197 Opt_max_dir_size_kb, Opt_nojournal_checksum,
1200 static const match_table_t tokens = {
1201 {Opt_bsd_df, "bsddf"},
1202 {Opt_minix_df, "minixdf"},
1203 {Opt_grpid, "grpid"},
1204 {Opt_grpid, "bsdgroups"},
1205 {Opt_nogrpid, "nogrpid"},
1206 {Opt_nogrpid, "sysvgroups"},
1207 {Opt_resgid, "resgid=%u"},
1208 {Opt_resuid, "resuid=%u"},
1210 {Opt_err_cont, "errors=continue"},
1211 {Opt_err_panic, "errors=panic"},
1212 {Opt_err_ro, "errors=remount-ro"},
1213 {Opt_nouid32, "nouid32"},
1214 {Opt_debug, "debug"},
1215 {Opt_removed, "oldalloc"},
1216 {Opt_removed, "orlov"},
1217 {Opt_user_xattr, "user_xattr"},
1218 {Opt_nouser_xattr, "nouser_xattr"},
1220 {Opt_noacl, "noacl"},
1221 {Opt_noload, "norecovery"},
1222 {Opt_noload, "noload"},
1223 {Opt_removed, "nobh"},
1224 {Opt_removed, "bh"},
1225 {Opt_commit, "commit=%u"},
1226 {Opt_min_batch_time, "min_batch_time=%u"},
1227 {Opt_max_batch_time, "max_batch_time=%u"},
1228 {Opt_journal_dev, "journal_dev=%u"},
1229 {Opt_journal_path, "journal_path=%s"},
1230 {Opt_journal_checksum, "journal_checksum"},
1231 {Opt_nojournal_checksum, "nojournal_checksum"},
1232 {Opt_journal_async_commit, "journal_async_commit"},
1233 {Opt_abort, "abort"},
1234 {Opt_data_journal, "data=journal"},
1235 {Opt_data_ordered, "data=ordered"},
1236 {Opt_data_writeback, "data=writeback"},
1237 {Opt_data_err_abort, "data_err=abort"},
1238 {Opt_data_err_ignore, "data_err=ignore"},
1239 {Opt_offusrjquota, "usrjquota="},
1240 {Opt_usrjquota, "usrjquota=%s"},
1241 {Opt_offgrpjquota, "grpjquota="},
1242 {Opt_grpjquota, "grpjquota=%s"},
1243 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1244 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1245 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1246 {Opt_grpquota, "grpquota"},
1247 {Opt_noquota, "noquota"},
1248 {Opt_quota, "quota"},
1249 {Opt_usrquota, "usrquota"},
1250 {Opt_barrier, "barrier=%u"},
1251 {Opt_barrier, "barrier"},
1252 {Opt_nobarrier, "nobarrier"},
1253 {Opt_i_version, "i_version"},
1255 {Opt_stripe, "stripe=%u"},
1256 {Opt_delalloc, "delalloc"},
1257 {Opt_lazytime, "lazytime"},
1258 {Opt_nolazytime, "nolazytime"},
1259 {Opt_nodelalloc, "nodelalloc"},
1260 {Opt_removed, "mblk_io_submit"},
1261 {Opt_removed, "nomblk_io_submit"},
1262 {Opt_block_validity, "block_validity"},
1263 {Opt_noblock_validity, "noblock_validity"},
1264 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1265 {Opt_journal_ioprio, "journal_ioprio=%u"},
1266 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1267 {Opt_auto_da_alloc, "auto_da_alloc"},
1268 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1269 {Opt_dioread_nolock, "dioread_nolock"},
1270 {Opt_dioread_lock, "dioread_lock"},
1271 {Opt_discard, "discard"},
1272 {Opt_nodiscard, "nodiscard"},
1273 {Opt_init_itable, "init_itable=%u"},
1274 {Opt_init_itable, "init_itable"},
1275 {Opt_noinit_itable, "noinit_itable"},
1276 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1277 {Opt_test_dummy_encryption, "test_dummy_encryption"},
1278 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1279 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1280 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1281 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1282 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1286 static ext4_fsblk_t get_sb_block(void **data)
1288 ext4_fsblk_t sb_block;
1289 char *options = (char *) *data;
1291 if (!options || strncmp(options, "sb=", 3) != 0)
1292 return 1; /* Default location */
1295 /* TODO: use simple_strtoll with >32bit ext4 */
1296 sb_block = simple_strtoul(options, &options, 0);
1297 if (*options && *options != ',') {
1298 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1302 if (*options == ',')
1304 *data = (void *) options;
1309 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1310 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1311 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1314 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1316 struct ext4_sb_info *sbi = EXT4_SB(sb);
1320 if (sb_any_quota_loaded(sb) &&
1321 !sbi->s_qf_names[qtype]) {
1322 ext4_msg(sb, KERN_ERR,
1323 "Cannot change journaled "
1324 "quota options when quota turned on");
1327 if (ext4_has_feature_quota(sb)) {
1328 ext4_msg(sb, KERN_ERR, "Cannot set journaled quota options "
1329 "when QUOTA feature is enabled");
1332 qname = match_strdup(args);
1334 ext4_msg(sb, KERN_ERR,
1335 "Not enough memory for storing quotafile name");
1338 if (sbi->s_qf_names[qtype]) {
1339 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1342 ext4_msg(sb, KERN_ERR,
1343 "%s quota file already specified",
1347 if (strchr(qname, '/')) {
1348 ext4_msg(sb, KERN_ERR,
1349 "quotafile must be on filesystem root");
1352 sbi->s_qf_names[qtype] = qname;
1360 static int clear_qf_name(struct super_block *sb, int qtype)
1363 struct ext4_sb_info *sbi = EXT4_SB(sb);
1365 if (sb_any_quota_loaded(sb) &&
1366 sbi->s_qf_names[qtype]) {
1367 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1368 " when quota turned on");
1371 kfree(sbi->s_qf_names[qtype]);
1372 sbi->s_qf_names[qtype] = NULL;
1377 #define MOPT_SET 0x0001
1378 #define MOPT_CLEAR 0x0002
1379 #define MOPT_NOSUPPORT 0x0004
1380 #define MOPT_EXPLICIT 0x0008
1381 #define MOPT_CLEAR_ERR 0x0010
1382 #define MOPT_GTE0 0x0020
1385 #define MOPT_QFMT 0x0040
1387 #define MOPT_Q MOPT_NOSUPPORT
1388 #define MOPT_QFMT MOPT_NOSUPPORT
1390 #define MOPT_DATAJ 0x0080
1391 #define MOPT_NO_EXT2 0x0100
1392 #define MOPT_NO_EXT3 0x0200
1393 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1394 #define MOPT_STRING 0x0400
1396 static const struct mount_opts {
1400 } ext4_mount_opts[] = {
1401 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1402 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1403 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1404 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1405 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1406 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1407 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1408 MOPT_EXT4_ONLY | MOPT_SET},
1409 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1410 MOPT_EXT4_ONLY | MOPT_CLEAR},
1411 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1412 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1413 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1414 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1415 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1416 MOPT_EXT4_ONLY | MOPT_CLEAR},
1417 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1418 MOPT_EXT4_ONLY | MOPT_CLEAR},
1419 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1420 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1421 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1422 EXT4_MOUNT_JOURNAL_CHECKSUM),
1423 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1424 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1425 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1426 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1427 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1428 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1429 MOPT_NO_EXT2 | MOPT_SET},
1430 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1431 MOPT_NO_EXT2 | MOPT_CLEAR},
1432 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1433 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1434 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1435 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1436 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1437 {Opt_commit, 0, MOPT_GTE0},
1438 {Opt_max_batch_time, 0, MOPT_GTE0},
1439 {Opt_min_batch_time, 0, MOPT_GTE0},
1440 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1441 {Opt_init_itable, 0, MOPT_GTE0},
1442 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
1443 {Opt_stripe, 0, MOPT_GTE0},
1444 {Opt_resuid, 0, MOPT_GTE0},
1445 {Opt_resgid, 0, MOPT_GTE0},
1446 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1447 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1448 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1449 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1450 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1451 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1452 MOPT_NO_EXT2 | MOPT_DATAJ},
1453 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1454 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1455 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1456 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1457 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1459 {Opt_acl, 0, MOPT_NOSUPPORT},
1460 {Opt_noacl, 0, MOPT_NOSUPPORT},
1462 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1463 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1464 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1465 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1467 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1469 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1470 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1471 {Opt_usrjquota, 0, MOPT_Q},
1472 {Opt_grpjquota, 0, MOPT_Q},
1473 {Opt_offusrjquota, 0, MOPT_Q},
1474 {Opt_offgrpjquota, 0, MOPT_Q},
1475 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1476 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1477 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1478 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1479 {Opt_test_dummy_encryption, 0, MOPT_GTE0},
1483 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1484 substring_t *args, unsigned long *journal_devnum,
1485 unsigned int *journal_ioprio, int is_remount)
1487 struct ext4_sb_info *sbi = EXT4_SB(sb);
1488 const struct mount_opts *m;
1494 if (token == Opt_usrjquota)
1495 return set_qf_name(sb, USRQUOTA, &args[0]);
1496 else if (token == Opt_grpjquota)
1497 return set_qf_name(sb, GRPQUOTA, &args[0]);
1498 else if (token == Opt_offusrjquota)
1499 return clear_qf_name(sb, USRQUOTA);
1500 else if (token == Opt_offgrpjquota)
1501 return clear_qf_name(sb, GRPQUOTA);
1505 case Opt_nouser_xattr:
1506 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1509 return 1; /* handled by get_sb_block() */
1511 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1514 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1517 sb->s_flags |= MS_I_VERSION;
1520 sb->s_flags |= MS_LAZYTIME;
1522 case Opt_nolazytime:
1523 sb->s_flags &= ~MS_LAZYTIME;
1527 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1528 if (token == m->token)
1531 if (m->token == Opt_err) {
1532 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1533 "or missing value", opt);
1537 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1538 ext4_msg(sb, KERN_ERR,
1539 "Mount option \"%s\" incompatible with ext2", opt);
1542 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1543 ext4_msg(sb, KERN_ERR,
1544 "Mount option \"%s\" incompatible with ext3", opt);
1548 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
1550 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1552 if (m->flags & MOPT_EXPLICIT) {
1553 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
1554 set_opt2(sb, EXPLICIT_DELALLOC);
1555 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
1556 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
1560 if (m->flags & MOPT_CLEAR_ERR)
1561 clear_opt(sb, ERRORS_MASK);
1562 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1563 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1564 "options when quota turned on");
1568 if (m->flags & MOPT_NOSUPPORT) {
1569 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1570 } else if (token == Opt_commit) {
1572 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1573 sbi->s_commit_interval = HZ * arg;
1574 } else if (token == Opt_max_batch_time) {
1575 sbi->s_max_batch_time = arg;
1576 } else if (token == Opt_min_batch_time) {
1577 sbi->s_min_batch_time = arg;
1578 } else if (token == Opt_inode_readahead_blks) {
1579 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1580 ext4_msg(sb, KERN_ERR,
1581 "EXT4-fs: inode_readahead_blks must be "
1582 "0 or a power of 2 smaller than 2^31");
1585 sbi->s_inode_readahead_blks = arg;
1586 } else if (token == Opt_init_itable) {
1587 set_opt(sb, INIT_INODE_TABLE);
1589 arg = EXT4_DEF_LI_WAIT_MULT;
1590 sbi->s_li_wait_mult = arg;
1591 } else if (token == Opt_max_dir_size_kb) {
1592 sbi->s_max_dir_size_kb = arg;
1593 } else if (token == Opt_stripe) {
1594 sbi->s_stripe = arg;
1595 } else if (token == Opt_resuid) {
1596 uid = make_kuid(current_user_ns(), arg);
1597 if (!uid_valid(uid)) {
1598 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1601 sbi->s_resuid = uid;
1602 } else if (token == Opt_resgid) {
1603 gid = make_kgid(current_user_ns(), arg);
1604 if (!gid_valid(gid)) {
1605 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1608 sbi->s_resgid = gid;
1609 } else if (token == Opt_journal_dev) {
1611 ext4_msg(sb, KERN_ERR,
1612 "Cannot specify journal on remount");
1615 *journal_devnum = arg;
1616 } else if (token == Opt_journal_path) {
1618 struct inode *journal_inode;
1623 ext4_msg(sb, KERN_ERR,
1624 "Cannot specify journal on remount");
1627 journal_path = match_strdup(&args[0]);
1628 if (!journal_path) {
1629 ext4_msg(sb, KERN_ERR, "error: could not dup "
1630 "journal device string");
1634 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1636 ext4_msg(sb, KERN_ERR, "error: could not find "
1637 "journal device path: error %d", error);
1638 kfree(journal_path);
1642 journal_inode = d_inode(path.dentry);
1643 if (!S_ISBLK(journal_inode->i_mode)) {
1644 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1645 "is not a block device", journal_path);
1647 kfree(journal_path);
1651 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1653 kfree(journal_path);
1654 } else if (token == Opt_journal_ioprio) {
1656 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1661 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1662 } else if (token == Opt_test_dummy_encryption) {
1663 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1664 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1665 ext4_msg(sb, KERN_WARNING,
1666 "Test dummy encryption mode enabled");
1668 ext4_msg(sb, KERN_WARNING,
1669 "Test dummy encryption mount option ignored");
1671 } else if (m->flags & MOPT_DATAJ) {
1673 if (!sbi->s_journal)
1674 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1675 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1676 ext4_msg(sb, KERN_ERR,
1677 "Cannot change data mode on remount");
1681 clear_opt(sb, DATA_FLAGS);
1682 sbi->s_mount_opt |= m->mount_opt;
1685 } else if (m->flags & MOPT_QFMT) {
1686 if (sb_any_quota_loaded(sb) &&
1687 sbi->s_jquota_fmt != m->mount_opt) {
1688 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1689 "quota options when quota turned on");
1692 if (ext4_has_feature_quota(sb)) {
1693 ext4_msg(sb, KERN_ERR,
1694 "Cannot set journaled quota options "
1695 "when QUOTA feature is enabled");
1698 sbi->s_jquota_fmt = m->mount_opt;
1700 } else if (token == Opt_dax) {
1701 #ifdef CONFIG_FS_DAX
1702 ext4_msg(sb, KERN_WARNING,
1703 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
1704 sbi->s_mount_opt |= m->mount_opt;
1706 ext4_msg(sb, KERN_INFO, "dax option not supported");
1712 if (m->flags & MOPT_CLEAR)
1714 else if (unlikely(!(m->flags & MOPT_SET))) {
1715 ext4_msg(sb, KERN_WARNING,
1716 "buggy handling of option %s", opt);
1721 sbi->s_mount_opt |= m->mount_opt;
1723 sbi->s_mount_opt &= ~m->mount_opt;
1728 static int parse_options(char *options, struct super_block *sb,
1729 unsigned long *journal_devnum,
1730 unsigned int *journal_ioprio,
1733 struct ext4_sb_info *sbi = EXT4_SB(sb);
1735 substring_t args[MAX_OPT_ARGS];
1741 while ((p = strsep(&options, ",")) != NULL) {
1745 * Initialize args struct so we know whether arg was
1746 * found; some options take optional arguments.
1748 args[0].to = args[0].from = NULL;
1749 token = match_token(p, tokens, args);
1750 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1751 journal_ioprio, is_remount) < 0)
1755 if (ext4_has_feature_quota(sb) &&
1756 (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) {
1757 ext4_msg(sb, KERN_ERR, "Cannot set quota options when QUOTA "
1758 "feature is enabled");
1761 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1762 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1763 clear_opt(sb, USRQUOTA);
1765 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1766 clear_opt(sb, GRPQUOTA);
1768 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1769 ext4_msg(sb, KERN_ERR, "old and new quota "
1774 if (!sbi->s_jquota_fmt) {
1775 ext4_msg(sb, KERN_ERR, "journaled quota format "
1781 if (test_opt(sb, DIOREAD_NOLOCK)) {
1783 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1785 if (blocksize < PAGE_CACHE_SIZE) {
1786 ext4_msg(sb, KERN_ERR, "can't mount with "
1787 "dioread_nolock if block size != PAGE_SIZE");
1791 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
1792 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
1793 ext4_msg(sb, KERN_ERR, "can't mount with journal_async_commit "
1794 "in data=ordered mode");
1800 static inline void ext4_show_quota_options(struct seq_file *seq,
1801 struct super_block *sb)
1803 #if defined(CONFIG_QUOTA)
1804 struct ext4_sb_info *sbi = EXT4_SB(sb);
1806 if (sbi->s_jquota_fmt) {
1809 switch (sbi->s_jquota_fmt) {
1820 seq_printf(seq, ",jqfmt=%s", fmtname);
1823 if (sbi->s_qf_names[USRQUOTA])
1824 seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]);
1826 if (sbi->s_qf_names[GRPQUOTA])
1827 seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]);
1831 static const char *token2str(int token)
1833 const struct match_token *t;
1835 for (t = tokens; t->token != Opt_err; t++)
1836 if (t->token == token && !strchr(t->pattern, '='))
1843 * - it's set to a non-default value OR
1844 * - if the per-sb default is different from the global default
1846 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1849 struct ext4_sb_info *sbi = EXT4_SB(sb);
1850 struct ext4_super_block *es = sbi->s_es;
1851 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1852 const struct mount_opts *m;
1853 char sep = nodefs ? '\n' : ',';
1855 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1856 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1858 if (sbi->s_sb_block != 1)
1859 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1861 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1862 int want_set = m->flags & MOPT_SET;
1863 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1864 (m->flags & MOPT_CLEAR_ERR))
1866 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
1867 continue; /* skip if same as the default */
1869 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1870 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1871 continue; /* select Opt_noFoo vs Opt_Foo */
1872 SEQ_OPTS_PRINT("%s", token2str(m->token));
1875 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
1876 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1877 SEQ_OPTS_PRINT("resuid=%u",
1878 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1879 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
1880 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1881 SEQ_OPTS_PRINT("resgid=%u",
1882 from_kgid_munged(&init_user_ns, sbi->s_resgid));
1883 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
1884 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1885 SEQ_OPTS_PUTS("errors=remount-ro");
1886 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1887 SEQ_OPTS_PUTS("errors=continue");
1888 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1889 SEQ_OPTS_PUTS("errors=panic");
1890 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
1891 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
1892 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
1893 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
1894 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
1895 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
1896 if (sb->s_flags & MS_I_VERSION)
1897 SEQ_OPTS_PUTS("i_version");
1898 if (nodefs || sbi->s_stripe)
1899 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
1900 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
1901 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1902 SEQ_OPTS_PUTS("data=journal");
1903 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1904 SEQ_OPTS_PUTS("data=ordered");
1905 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1906 SEQ_OPTS_PUTS("data=writeback");
1909 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1910 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1911 sbi->s_inode_readahead_blks);
1913 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1914 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
1915 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
1916 if (nodefs || sbi->s_max_dir_size_kb)
1917 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
1919 ext4_show_quota_options(seq, sb);
1923 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1925 return _ext4_show_options(seq, root->d_sb, 0);
1928 int ext4_seq_options_show(struct seq_file *seq, void *offset)
1930 struct super_block *sb = seq->private;
1933 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1934 rc = _ext4_show_options(seq, sb, 1);
1935 seq_puts(seq, "\n");
1939 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1942 struct ext4_sb_info *sbi = EXT4_SB(sb);
1945 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1946 ext4_msg(sb, KERN_ERR, "revision level too high, "
1947 "forcing read-only mode");
1952 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1953 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1954 "running e2fsck is recommended");
1955 else if (sbi->s_mount_state & EXT4_ERROR_FS)
1956 ext4_msg(sb, KERN_WARNING,
1957 "warning: mounting fs with errors, "
1958 "running e2fsck is recommended");
1959 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1960 le16_to_cpu(es->s_mnt_count) >=
1961 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1962 ext4_msg(sb, KERN_WARNING,
1963 "warning: maximal mount count reached, "
1964 "running e2fsck is recommended");
1965 else if (le32_to_cpu(es->s_checkinterval) &&
1966 (le32_to_cpu(es->s_lastcheck) +
1967 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1968 ext4_msg(sb, KERN_WARNING,
1969 "warning: checktime reached, "
1970 "running e2fsck is recommended");
1971 if (!sbi->s_journal)
1972 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1973 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1974 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1975 le16_add_cpu(&es->s_mnt_count, 1);
1976 es->s_mtime = cpu_to_le32(get_seconds());
1977 ext4_update_dynamic_rev(sb);
1979 ext4_set_feature_journal_needs_recovery(sb);
1981 ext4_commit_super(sb, 1);
1983 if (test_opt(sb, DEBUG))
1984 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1985 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1987 sbi->s_groups_count,
1988 EXT4_BLOCKS_PER_GROUP(sb),
1989 EXT4_INODES_PER_GROUP(sb),
1990 sbi->s_mount_opt, sbi->s_mount_opt2);
1992 cleancache_init_fs(sb);
1996 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1998 struct ext4_sb_info *sbi = EXT4_SB(sb);
1999 struct flex_groups *new_groups;
2002 if (!sbi->s_log_groups_per_flex)
2005 size = ext4_flex_group(sbi, ngroup - 1) + 1;
2006 if (size <= sbi->s_flex_groups_allocated)
2009 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
2010 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
2012 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
2013 size / (int) sizeof(struct flex_groups));
2017 if (sbi->s_flex_groups) {
2018 memcpy(new_groups, sbi->s_flex_groups,
2019 (sbi->s_flex_groups_allocated *
2020 sizeof(struct flex_groups)));
2021 kvfree(sbi->s_flex_groups);
2023 sbi->s_flex_groups = new_groups;
2024 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
2028 static int ext4_fill_flex_info(struct super_block *sb)
2030 struct ext4_sb_info *sbi = EXT4_SB(sb);
2031 struct ext4_group_desc *gdp = NULL;
2032 ext4_group_t flex_group;
2035 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2036 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
2037 sbi->s_log_groups_per_flex = 0;
2041 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2045 for (i = 0; i < sbi->s_groups_count; i++) {
2046 gdp = ext4_get_group_desc(sb, i, NULL);
2048 flex_group = ext4_flex_group(sbi, i);
2049 atomic_add(ext4_free_inodes_count(sb, gdp),
2050 &sbi->s_flex_groups[flex_group].free_inodes);
2051 atomic64_add(ext4_free_group_clusters(sb, gdp),
2052 &sbi->s_flex_groups[flex_group].free_clusters);
2053 atomic_add(ext4_used_dirs_count(sb, gdp),
2054 &sbi->s_flex_groups[flex_group].used_dirs);
2062 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
2063 struct ext4_group_desc *gdp)
2067 __le32 le_group = cpu_to_le32(block_group);
2068 struct ext4_sb_info *sbi = EXT4_SB(sb);
2070 if (ext4_has_metadata_csum(sbi->s_sb)) {
2071 /* Use new metadata_csum algorithm */
2075 save_csum = gdp->bg_checksum;
2076 gdp->bg_checksum = 0;
2077 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2079 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
2081 gdp->bg_checksum = save_csum;
2083 crc = csum32 & 0xFFFF;
2087 /* old crc16 code */
2088 if (!ext4_has_feature_gdt_csum(sb))
2091 offset = offsetof(struct ext4_group_desc, bg_checksum);
2093 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2094 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2095 crc = crc16(crc, (__u8 *)gdp, offset);
2096 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2097 /* for checksum of struct ext4_group_desc do the rest...*/
2098 if (ext4_has_feature_64bit(sb) &&
2099 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2100 crc = crc16(crc, (__u8 *)gdp + offset,
2101 le16_to_cpu(sbi->s_es->s_desc_size) -
2105 return cpu_to_le16(crc);
2108 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2109 struct ext4_group_desc *gdp)
2111 if (ext4_has_group_desc_csum(sb) &&
2112 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
2118 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2119 struct ext4_group_desc *gdp)
2121 if (!ext4_has_group_desc_csum(sb))
2123 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
2126 /* Called at mount-time, super-block is locked */
2127 static int ext4_check_descriptors(struct super_block *sb,
2128 ext4_group_t *first_not_zeroed)
2130 struct ext4_sb_info *sbi = EXT4_SB(sb);
2131 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2132 ext4_fsblk_t last_block;
2133 ext4_fsblk_t block_bitmap;
2134 ext4_fsblk_t inode_bitmap;
2135 ext4_fsblk_t inode_table;
2136 int flexbg_flag = 0;
2137 ext4_group_t i, grp = sbi->s_groups_count;
2139 if (ext4_has_feature_flex_bg(sb))
2142 ext4_debug("Checking group descriptors");
2144 for (i = 0; i < sbi->s_groups_count; i++) {
2145 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2147 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2148 last_block = ext4_blocks_count(sbi->s_es) - 1;
2150 last_block = first_block +
2151 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2153 if ((grp == sbi->s_groups_count) &&
2154 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2157 block_bitmap = ext4_block_bitmap(sb, gdp);
2158 if (block_bitmap < first_block || block_bitmap > last_block) {
2159 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2160 "Block bitmap for group %u not in group "
2161 "(block %llu)!", i, block_bitmap);
2164 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2165 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2166 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2167 "Inode bitmap for group %u not in group "
2168 "(block %llu)!", i, inode_bitmap);
2171 inode_table = ext4_inode_table(sb, gdp);
2172 if (inode_table < first_block ||
2173 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2174 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2175 "Inode table for group %u not in group "
2176 "(block %llu)!", i, inode_table);
2179 ext4_lock_group(sb, i);
2180 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2181 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2182 "Checksum for group %u failed (%u!=%u)",
2183 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
2184 gdp)), le16_to_cpu(gdp->bg_checksum));
2185 if (!(sb->s_flags & MS_RDONLY)) {
2186 ext4_unlock_group(sb, i);
2190 ext4_unlock_group(sb, i);
2192 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2194 if (NULL != first_not_zeroed)
2195 *first_not_zeroed = grp;
2199 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2200 * the superblock) which were deleted from all directories, but held open by
2201 * a process at the time of a crash. We walk the list and try to delete these
2202 * inodes at recovery time (only with a read-write filesystem).
2204 * In order to keep the orphan inode chain consistent during traversal (in
2205 * case of crash during recovery), we link each inode into the superblock
2206 * orphan list_head and handle it the same way as an inode deletion during
2207 * normal operation (which journals the operations for us).
2209 * We only do an iget() and an iput() on each inode, which is very safe if we
2210 * accidentally point at an in-use or already deleted inode. The worst that
2211 * can happen in this case is that we get a "bit already cleared" message from
2212 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2213 * e2fsck was run on this filesystem, and it must have already done the orphan
2214 * inode cleanup for us, so we can safely abort without any further action.
2216 static void ext4_orphan_cleanup(struct super_block *sb,
2217 struct ext4_super_block *es)
2219 unsigned int s_flags = sb->s_flags;
2220 int nr_orphans = 0, nr_truncates = 0;
2224 if (!es->s_last_orphan) {
2225 jbd_debug(4, "no orphan inodes to clean up\n");
2229 if (bdev_read_only(sb->s_bdev)) {
2230 ext4_msg(sb, KERN_ERR, "write access "
2231 "unavailable, skipping orphan cleanup");
2235 /* Check if feature set would not allow a r/w mount */
2236 if (!ext4_feature_set_ok(sb, 0)) {
2237 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2238 "unknown ROCOMPAT features");
2242 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2243 /* don't clear list on RO mount w/ errors */
2244 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
2245 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
2246 "clearing orphan list.\n");
2247 es->s_last_orphan = 0;
2249 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2253 if (s_flags & MS_RDONLY) {
2254 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2255 sb->s_flags &= ~MS_RDONLY;
2258 /* Needed for iput() to work correctly and not trash data */
2259 sb->s_flags |= MS_ACTIVE;
2260 /* Turn on quotas so that they are updated correctly */
2261 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2262 if (EXT4_SB(sb)->s_qf_names[i]) {
2263 int ret = ext4_quota_on_mount(sb, i);
2265 ext4_msg(sb, KERN_ERR,
2266 "Cannot turn on journaled "
2267 "quota: error %d", ret);
2272 while (es->s_last_orphan) {
2273 struct inode *inode;
2275 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2276 if (IS_ERR(inode)) {
2277 es->s_last_orphan = 0;
2281 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2282 dquot_initialize(inode);
2283 if (inode->i_nlink) {
2284 if (test_opt(sb, DEBUG))
2285 ext4_msg(sb, KERN_DEBUG,
2286 "%s: truncating inode %lu to %lld bytes",
2287 __func__, inode->i_ino, inode->i_size);
2288 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2289 inode->i_ino, inode->i_size);
2291 truncate_inode_pages(inode->i_mapping, inode->i_size);
2292 ext4_truncate(inode);
2293 inode_unlock(inode);
2296 if (test_opt(sb, DEBUG))
2297 ext4_msg(sb, KERN_DEBUG,
2298 "%s: deleting unreferenced inode %lu",
2299 __func__, inode->i_ino);
2300 jbd_debug(2, "deleting unreferenced inode %lu\n",
2304 iput(inode); /* The delete magic happens here! */
2307 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2310 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2311 PLURAL(nr_orphans));
2313 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2314 PLURAL(nr_truncates));
2316 /* Turn quotas off */
2317 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2318 if (sb_dqopt(sb)->files[i])
2319 dquot_quota_off(sb, i);
2322 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2326 * Maximal extent format file size.
2327 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2328 * extent format containers, within a sector_t, and within i_blocks
2329 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2330 * so that won't be a limiting factor.
2332 * However there is other limiting factor. We do store extents in the form
2333 * of starting block and length, hence the resulting length of the extent
2334 * covering maximum file size must fit into on-disk format containers as
2335 * well. Given that length is always by 1 unit bigger than max unit (because
2336 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2338 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2340 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2343 loff_t upper_limit = MAX_LFS_FILESIZE;
2345 /* small i_blocks in vfs inode? */
2346 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2348 * CONFIG_LBDAF is not enabled implies the inode
2349 * i_block represent total blocks in 512 bytes
2350 * 32 == size of vfs inode i_blocks * 8
2352 upper_limit = (1LL << 32) - 1;
2354 /* total blocks in file system block size */
2355 upper_limit >>= (blkbits - 9);
2356 upper_limit <<= blkbits;
2360 * 32-bit extent-start container, ee_block. We lower the maxbytes
2361 * by one fs block, so ee_len can cover the extent of maximum file
2364 res = (1LL << 32) - 1;
2367 /* Sanity check against vm- & vfs- imposed limits */
2368 if (res > upper_limit)
2375 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2376 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2377 * We need to be 1 filesystem block less than the 2^48 sector limit.
2379 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2381 loff_t res = EXT4_NDIR_BLOCKS;
2384 /* This is calculated to be the largest file size for a dense, block
2385 * mapped file such that the file's total number of 512-byte sectors,
2386 * including data and all indirect blocks, does not exceed (2^48 - 1).
2388 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2389 * number of 512-byte sectors of the file.
2392 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2394 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2395 * the inode i_block field represents total file blocks in
2396 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2398 upper_limit = (1LL << 32) - 1;
2400 /* total blocks in file system block size */
2401 upper_limit >>= (bits - 9);
2405 * We use 48 bit ext4_inode i_blocks
2406 * With EXT4_HUGE_FILE_FL set the i_blocks
2407 * represent total number of blocks in
2408 * file system block size
2410 upper_limit = (1LL << 48) - 1;
2414 /* indirect blocks */
2416 /* double indirect blocks */
2417 meta_blocks += 1 + (1LL << (bits-2));
2418 /* tripple indirect blocks */
2419 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2421 upper_limit -= meta_blocks;
2422 upper_limit <<= bits;
2424 res += 1LL << (bits-2);
2425 res += 1LL << (2*(bits-2));
2426 res += 1LL << (3*(bits-2));
2428 if (res > upper_limit)
2431 if (res > MAX_LFS_FILESIZE)
2432 res = MAX_LFS_FILESIZE;
2437 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2438 ext4_fsblk_t logical_sb_block, int nr)
2440 struct ext4_sb_info *sbi = EXT4_SB(sb);
2441 ext4_group_t bg, first_meta_bg;
2444 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2446 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
2447 return logical_sb_block + nr + 1;
2448 bg = sbi->s_desc_per_block * nr;
2449 if (ext4_bg_has_super(sb, bg))
2453 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2454 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2455 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2458 if (sb->s_blocksize == 1024 && nr == 0 &&
2459 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2462 return (has_super + ext4_group_first_block_no(sb, bg));
2466 * ext4_get_stripe_size: Get the stripe size.
2467 * @sbi: In memory super block info
2469 * If we have specified it via mount option, then
2470 * use the mount option value. If the value specified at mount time is
2471 * greater than the blocks per group use the super block value.
2472 * If the super block value is greater than blocks per group return 0.
2473 * Allocator needs it be less than blocks per group.
2476 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2478 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2479 unsigned long stripe_width =
2480 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2483 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2484 ret = sbi->s_stripe;
2485 else if (stripe_width <= sbi->s_blocks_per_group)
2487 else if (stride <= sbi->s_blocks_per_group)
2493 * If the stripe width is 1, this makes no sense and
2494 * we set it to 0 to turn off stripe handling code.
2503 * Check whether this filesystem can be mounted based on
2504 * the features present and the RDONLY/RDWR mount requested.
2505 * Returns 1 if this filesystem can be mounted as requested,
2506 * 0 if it cannot be.
2508 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2510 if (ext4_has_unknown_ext4_incompat_features(sb)) {
2511 ext4_msg(sb, KERN_ERR,
2512 "Couldn't mount because of "
2513 "unsupported optional features (%x)",
2514 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2515 ~EXT4_FEATURE_INCOMPAT_SUPP));
2522 if (ext4_has_feature_readonly(sb)) {
2523 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2524 sb->s_flags |= MS_RDONLY;
2528 /* Check that feature set is OK for a read-write mount */
2529 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
2530 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2531 "unsupported optional features (%x)",
2532 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2533 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2537 * Large file size enabled file system can only be mounted
2538 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2540 if (ext4_has_feature_huge_file(sb)) {
2541 if (sizeof(blkcnt_t) < sizeof(u64)) {
2542 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2543 "cannot be mounted RDWR without "
2548 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
2549 ext4_msg(sb, KERN_ERR,
2550 "Can't support bigalloc feature without "
2551 "extents feature\n");
2555 #ifndef CONFIG_QUOTA
2556 if (ext4_has_feature_quota(sb) && !readonly) {
2557 ext4_msg(sb, KERN_ERR,
2558 "Filesystem with quota feature cannot be mounted RDWR "
2559 "without CONFIG_QUOTA");
2562 if (ext4_has_feature_project(sb) && !readonly) {
2563 ext4_msg(sb, KERN_ERR,
2564 "Filesystem with project quota feature cannot be mounted RDWR "
2565 "without CONFIG_QUOTA");
2568 #endif /* CONFIG_QUOTA */
2573 * This function is called once a day if we have errors logged
2574 * on the file system
2576 static void print_daily_error_info(unsigned long arg)
2578 struct super_block *sb = (struct super_block *) arg;
2579 struct ext4_sb_info *sbi;
2580 struct ext4_super_block *es;
2585 if (es->s_error_count)
2586 /* fsck newer than v1.41.13 is needed to clean this condition. */
2587 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
2588 le32_to_cpu(es->s_error_count));
2589 if (es->s_first_error_time) {
2590 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
2591 sb->s_id, le32_to_cpu(es->s_first_error_time),
2592 (int) sizeof(es->s_first_error_func),
2593 es->s_first_error_func,
2594 le32_to_cpu(es->s_first_error_line));
2595 if (es->s_first_error_ino)
2596 printk(": inode %u",
2597 le32_to_cpu(es->s_first_error_ino));
2598 if (es->s_first_error_block)
2599 printk(": block %llu", (unsigned long long)
2600 le64_to_cpu(es->s_first_error_block));
2603 if (es->s_last_error_time) {
2604 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
2605 sb->s_id, le32_to_cpu(es->s_last_error_time),
2606 (int) sizeof(es->s_last_error_func),
2607 es->s_last_error_func,
2608 le32_to_cpu(es->s_last_error_line));
2609 if (es->s_last_error_ino)
2610 printk(": inode %u",
2611 le32_to_cpu(es->s_last_error_ino));
2612 if (es->s_last_error_block)
2613 printk(": block %llu", (unsigned long long)
2614 le64_to_cpu(es->s_last_error_block));
2617 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2620 /* Find next suitable group and run ext4_init_inode_table */
2621 static int ext4_run_li_request(struct ext4_li_request *elr)
2623 struct ext4_group_desc *gdp = NULL;
2624 ext4_group_t group, ngroups;
2625 struct super_block *sb;
2626 unsigned long timeout = 0;
2630 ngroups = EXT4_SB(sb)->s_groups_count;
2633 for (group = elr->lr_next_group; group < ngroups; group++) {
2634 gdp = ext4_get_group_desc(sb, group, NULL);
2640 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2644 if (group >= ngroups)
2649 ret = ext4_init_inode_table(sb, group,
2650 elr->lr_timeout ? 0 : 1);
2651 if (elr->lr_timeout == 0) {
2652 timeout = (jiffies - timeout) *
2653 elr->lr_sbi->s_li_wait_mult;
2654 elr->lr_timeout = timeout;
2656 elr->lr_next_sched = jiffies + elr->lr_timeout;
2657 elr->lr_next_group = group + 1;
2665 * Remove lr_request from the list_request and free the
2666 * request structure. Should be called with li_list_mtx held
2668 static void ext4_remove_li_request(struct ext4_li_request *elr)
2670 struct ext4_sb_info *sbi;
2677 list_del(&elr->lr_request);
2678 sbi->s_li_request = NULL;
2682 static void ext4_unregister_li_request(struct super_block *sb)
2684 mutex_lock(&ext4_li_mtx);
2685 if (!ext4_li_info) {
2686 mutex_unlock(&ext4_li_mtx);
2690 mutex_lock(&ext4_li_info->li_list_mtx);
2691 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2692 mutex_unlock(&ext4_li_info->li_list_mtx);
2693 mutex_unlock(&ext4_li_mtx);
2696 static struct task_struct *ext4_lazyinit_task;
2699 * This is the function where ext4lazyinit thread lives. It walks
2700 * through the request list searching for next scheduled filesystem.
2701 * When such a fs is found, run the lazy initialization request
2702 * (ext4_rn_li_request) and keep track of the time spend in this
2703 * function. Based on that time we compute next schedule time of
2704 * the request. When walking through the list is complete, compute
2705 * next waking time and put itself into sleep.
2707 static int ext4_lazyinit_thread(void *arg)
2709 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2710 struct list_head *pos, *n;
2711 struct ext4_li_request *elr;
2712 unsigned long next_wakeup, cur;
2714 BUG_ON(NULL == eli);
2718 next_wakeup = MAX_JIFFY_OFFSET;
2720 mutex_lock(&eli->li_list_mtx);
2721 if (list_empty(&eli->li_request_list)) {
2722 mutex_unlock(&eli->li_list_mtx);
2726 list_for_each_safe(pos, n, &eli->li_request_list) {
2727 elr = list_entry(pos, struct ext4_li_request,
2730 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2731 if (ext4_run_li_request(elr) != 0) {
2732 /* error, remove the lazy_init job */
2733 ext4_remove_li_request(elr);
2738 if (time_before(elr->lr_next_sched, next_wakeup))
2739 next_wakeup = elr->lr_next_sched;
2741 mutex_unlock(&eli->li_list_mtx);
2746 if ((time_after_eq(cur, next_wakeup)) ||
2747 (MAX_JIFFY_OFFSET == next_wakeup)) {
2752 schedule_timeout_interruptible(next_wakeup - cur);
2754 if (kthread_should_stop()) {
2755 ext4_clear_request_list();
2762 * It looks like the request list is empty, but we need
2763 * to check it under the li_list_mtx lock, to prevent any
2764 * additions into it, and of course we should lock ext4_li_mtx
2765 * to atomically free the list and ext4_li_info, because at
2766 * this point another ext4 filesystem could be registering
2769 mutex_lock(&ext4_li_mtx);
2770 mutex_lock(&eli->li_list_mtx);
2771 if (!list_empty(&eli->li_request_list)) {
2772 mutex_unlock(&eli->li_list_mtx);
2773 mutex_unlock(&ext4_li_mtx);
2776 mutex_unlock(&eli->li_list_mtx);
2777 kfree(ext4_li_info);
2778 ext4_li_info = NULL;
2779 mutex_unlock(&ext4_li_mtx);
2784 static void ext4_clear_request_list(void)
2786 struct list_head *pos, *n;
2787 struct ext4_li_request *elr;
2789 mutex_lock(&ext4_li_info->li_list_mtx);
2790 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2791 elr = list_entry(pos, struct ext4_li_request,
2793 ext4_remove_li_request(elr);
2795 mutex_unlock(&ext4_li_info->li_list_mtx);
2798 static int ext4_run_lazyinit_thread(void)
2800 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2801 ext4_li_info, "ext4lazyinit");
2802 if (IS_ERR(ext4_lazyinit_task)) {
2803 int err = PTR_ERR(ext4_lazyinit_task);
2804 ext4_clear_request_list();
2805 kfree(ext4_li_info);
2806 ext4_li_info = NULL;
2807 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
2808 "initialization thread\n",
2812 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2817 * Check whether it make sense to run itable init. thread or not.
2818 * If there is at least one uninitialized inode table, return
2819 * corresponding group number, else the loop goes through all
2820 * groups and return total number of groups.
2822 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2824 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2825 struct ext4_group_desc *gdp = NULL;
2827 for (group = 0; group < ngroups; group++) {
2828 gdp = ext4_get_group_desc(sb, group, NULL);
2832 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2839 static int ext4_li_info_new(void)
2841 struct ext4_lazy_init *eli = NULL;
2843 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2847 INIT_LIST_HEAD(&eli->li_request_list);
2848 mutex_init(&eli->li_list_mtx);
2850 eli->li_state |= EXT4_LAZYINIT_QUIT;
2857 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2860 struct ext4_sb_info *sbi = EXT4_SB(sb);
2861 struct ext4_li_request *elr;
2863 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2869 elr->lr_next_group = start;
2872 * Randomize first schedule time of the request to
2873 * spread the inode table initialization requests
2876 elr->lr_next_sched = jiffies + (prandom_u32() %
2877 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
2881 int ext4_register_li_request(struct super_block *sb,
2882 ext4_group_t first_not_zeroed)
2884 struct ext4_sb_info *sbi = EXT4_SB(sb);
2885 struct ext4_li_request *elr = NULL;
2886 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
2889 mutex_lock(&ext4_li_mtx);
2890 if (sbi->s_li_request != NULL) {
2892 * Reset timeout so it can be computed again, because
2893 * s_li_wait_mult might have changed.
2895 sbi->s_li_request->lr_timeout = 0;
2899 if (first_not_zeroed == ngroups ||
2900 (sb->s_flags & MS_RDONLY) ||
2901 !test_opt(sb, INIT_INODE_TABLE))
2904 elr = ext4_li_request_new(sb, first_not_zeroed);
2910 if (NULL == ext4_li_info) {
2911 ret = ext4_li_info_new();
2916 mutex_lock(&ext4_li_info->li_list_mtx);
2917 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
2918 mutex_unlock(&ext4_li_info->li_list_mtx);
2920 sbi->s_li_request = elr;
2922 * set elr to NULL here since it has been inserted to
2923 * the request_list and the removal and free of it is
2924 * handled by ext4_clear_request_list from now on.
2928 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
2929 ret = ext4_run_lazyinit_thread();
2934 mutex_unlock(&ext4_li_mtx);
2941 * We do not need to lock anything since this is called on
2944 static void ext4_destroy_lazyinit_thread(void)
2947 * If thread exited earlier
2948 * there's nothing to be done.
2950 if (!ext4_li_info || !ext4_lazyinit_task)
2953 kthread_stop(ext4_lazyinit_task);
2956 static int set_journal_csum_feature_set(struct super_block *sb)
2959 int compat, incompat;
2960 struct ext4_sb_info *sbi = EXT4_SB(sb);
2962 if (ext4_has_metadata_csum(sb)) {
2963 /* journal checksum v3 */
2965 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
2967 /* journal checksum v1 */
2968 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
2972 jbd2_journal_clear_features(sbi->s_journal,
2973 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2974 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
2975 JBD2_FEATURE_INCOMPAT_CSUM_V2);
2976 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2977 ret = jbd2_journal_set_features(sbi->s_journal,
2979 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
2981 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2982 ret = jbd2_journal_set_features(sbi->s_journal,
2985 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2986 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2988 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2989 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2996 * Note: calculating the overhead so we can be compatible with
2997 * historical BSD practice is quite difficult in the face of
2998 * clusters/bigalloc. This is because multiple metadata blocks from
2999 * different block group can end up in the same allocation cluster.
3000 * Calculating the exact overhead in the face of clustered allocation
3001 * requires either O(all block bitmaps) in memory or O(number of block
3002 * groups**2) in time. We will still calculate the superblock for
3003 * older file systems --- and if we come across with a bigalloc file
3004 * system with zero in s_overhead_clusters the estimate will be close to
3005 * correct especially for very large cluster sizes --- but for newer
3006 * file systems, it's better to calculate this figure once at mkfs
3007 * time, and store it in the superblock. If the superblock value is
3008 * present (even for non-bigalloc file systems), we will use it.
3010 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3013 struct ext4_sb_info *sbi = EXT4_SB(sb);
3014 struct ext4_group_desc *gdp;
3015 ext4_fsblk_t first_block, last_block, b;
3016 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3017 int s, j, count = 0;
3019 if (!ext4_has_feature_bigalloc(sb))
3020 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3021 sbi->s_itb_per_group + 2);
3023 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3024 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3025 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3026 for (i = 0; i < ngroups; i++) {
3027 gdp = ext4_get_group_desc(sb, i, NULL);
3028 b = ext4_block_bitmap(sb, gdp);
3029 if (b >= first_block && b <= last_block) {
3030 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3033 b = ext4_inode_bitmap(sb, gdp);
3034 if (b >= first_block && b <= last_block) {
3035 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3038 b = ext4_inode_table(sb, gdp);
3039 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3040 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3041 int c = EXT4_B2C(sbi, b - first_block);
3042 ext4_set_bit(c, buf);
3048 if (ext4_bg_has_super(sb, grp)) {
3049 ext4_set_bit(s++, buf);
3052 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3053 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3059 return EXT4_CLUSTERS_PER_GROUP(sb) -
3060 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3064 * Compute the overhead and stash it in sbi->s_overhead
3066 int ext4_calculate_overhead(struct super_block *sb)
3068 struct ext4_sb_info *sbi = EXT4_SB(sb);
3069 struct ext4_super_block *es = sbi->s_es;
3070 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3071 ext4_fsblk_t overhead = 0;
3072 char *buf = (char *) get_zeroed_page(GFP_NOFS);
3078 * Compute the overhead (FS structures). This is constant
3079 * for a given filesystem unless the number of block groups
3080 * changes so we cache the previous value until it does.
3084 * All of the blocks before first_data_block are overhead
3086 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3089 * Add the overhead found in each block group
3091 for (i = 0; i < ngroups; i++) {
3094 blks = count_overhead(sb, i, buf);
3097 memset(buf, 0, PAGE_SIZE);
3100 /* Add the internal journal blocks as well */
3101 if (sbi->s_journal && !sbi->journal_bdev)
3102 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3104 sbi->s_overhead = overhead;
3106 free_page((unsigned long) buf);
3110 static void ext4_set_resv_clusters(struct super_block *sb)
3112 ext4_fsblk_t resv_clusters;
3113 struct ext4_sb_info *sbi = EXT4_SB(sb);
3116 * There's no need to reserve anything when we aren't using extents.
3117 * The space estimates are exact, there are no unwritten extents,
3118 * hole punching doesn't need new metadata... This is needed especially
3119 * to keep ext2/3 backward compatibility.
3121 if (!ext4_has_feature_extents(sb))
3124 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3125 * This should cover the situations where we can not afford to run
3126 * out of space like for example punch hole, or converting
3127 * unwritten extents in delalloc path. In most cases such
3128 * allocation would require 1, or 2 blocks, higher numbers are
3131 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3132 sbi->s_cluster_bits);
3134 do_div(resv_clusters, 50);
3135 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3137 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
3140 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3142 char *orig_data = kstrdup(data, GFP_KERNEL);
3143 struct buffer_head *bh;
3144 struct ext4_super_block *es = NULL;
3145 struct ext4_sb_info *sbi;
3147 ext4_fsblk_t sb_block = get_sb_block(&data);
3148 ext4_fsblk_t logical_sb_block;
3149 unsigned long offset = 0;
3150 unsigned long journal_devnum = 0;
3151 unsigned long def_mount_opts;
3155 int blocksize, clustersize;
3156 unsigned int db_count;
3158 int needs_recovery, has_huge_files, has_bigalloc;
3161 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3162 ext4_group_t first_not_zeroed;
3164 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3168 sbi->s_blockgroup_lock =
3169 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3170 if (!sbi->s_blockgroup_lock) {
3174 sb->s_fs_info = sbi;
3176 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3177 sbi->s_sb_block = sb_block;
3178 if (sb->s_bdev->bd_part)
3179 sbi->s_sectors_written_start =
3180 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3182 /* Cleanup superblock name */
3183 strreplace(sb->s_id, '/', '!');
3185 /* -EINVAL is default */
3187 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3189 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3194 * The ext4 superblock will not be buffer aligned for other than 1kB
3195 * block sizes. We need to calculate the offset from buffer start.
3197 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3198 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3199 offset = do_div(logical_sb_block, blocksize);
3201 logical_sb_block = sb_block;
3204 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
3205 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3209 * Note: s_es must be initialized as soon as possible because
3210 * some ext4 macro-instructions depend on its value
3212 es = (struct ext4_super_block *) (bh->b_data + offset);
3214 sb->s_magic = le16_to_cpu(es->s_magic);
3215 if (sb->s_magic != EXT4_SUPER_MAGIC)
3217 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3219 /* Warn if metadata_csum and gdt_csum are both set. */
3220 if (ext4_has_feature_metadata_csum(sb) &&
3221 ext4_has_feature_gdt_csum(sb))
3222 ext4_warning(sb, "metadata_csum and uninit_bg are "
3223 "redundant flags; please run fsck.");
3225 /* Check for a known checksum algorithm */
3226 if (!ext4_verify_csum_type(sb, es)) {
3227 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3228 "unknown checksum algorithm.");
3233 /* Load the checksum driver */
3234 if (ext4_has_feature_metadata_csum(sb)) {
3235 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3236 if (IS_ERR(sbi->s_chksum_driver)) {
3237 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3238 ret = PTR_ERR(sbi->s_chksum_driver);
3239 sbi->s_chksum_driver = NULL;
3244 /* Check superblock checksum */
3245 if (!ext4_superblock_csum_verify(sb, es)) {
3246 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3247 "invalid superblock checksum. Run e2fsck?");
3253 /* Precompute checksum seed for all metadata */
3254 if (ext4_has_feature_csum_seed(sb))
3255 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
3256 else if (ext4_has_metadata_csum(sb))
3257 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3258 sizeof(es->s_uuid));
3260 /* Set defaults before we parse the mount options */
3261 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3262 set_opt(sb, INIT_INODE_TABLE);
3263 if (def_mount_opts & EXT4_DEFM_DEBUG)
3265 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3267 if (def_mount_opts & EXT4_DEFM_UID16)
3268 set_opt(sb, NO_UID32);
3269 /* xattr user namespace & acls are now defaulted on */
3270 set_opt(sb, XATTR_USER);
3271 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3272 set_opt(sb, POSIX_ACL);
3274 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3275 if (ext4_has_metadata_csum(sb))
3276 set_opt(sb, JOURNAL_CHECKSUM);
3278 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3279 set_opt(sb, JOURNAL_DATA);
3280 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3281 set_opt(sb, ORDERED_DATA);
3282 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3283 set_opt(sb, WRITEBACK_DATA);
3285 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3286 set_opt(sb, ERRORS_PANIC);
3287 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3288 set_opt(sb, ERRORS_CONT);
3290 set_opt(sb, ERRORS_RO);
3291 /* block_validity enabled by default; disable with noblock_validity */
3292 set_opt(sb, BLOCK_VALIDITY);
3293 if (def_mount_opts & EXT4_DEFM_DISCARD)
3294 set_opt(sb, DISCARD);
3296 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3297 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3298 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3299 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3300 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3302 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3303 set_opt(sb, BARRIER);
3306 * enable delayed allocation by default
3307 * Use -o nodelalloc to turn it off
3309 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3310 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3311 set_opt(sb, DELALLOC);
3314 * set default s_li_wait_mult for lazyinit, for the case there is
3315 * no mount option specified.
3317 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3319 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3320 &journal_devnum, &journal_ioprio, 0)) {
3321 ext4_msg(sb, KERN_WARNING,
3322 "failed to parse options in superblock: %s",
3323 sbi->s_es->s_mount_opts);
3325 sbi->s_def_mount_opt = sbi->s_mount_opt;
3326 if (!parse_options((char *) data, sb, &journal_devnum,
3327 &journal_ioprio, 0))
3330 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3331 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3332 "with data=journal disables delayed "
3333 "allocation and O_DIRECT support!\n");
3334 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3335 ext4_msg(sb, KERN_ERR, "can't mount with "
3336 "both data=journal and delalloc");
3339 if (test_opt(sb, DIOREAD_NOLOCK)) {
3340 ext4_msg(sb, KERN_ERR, "can't mount with "
3341 "both data=journal and dioread_nolock");
3344 if (test_opt(sb, DAX)) {
3345 ext4_msg(sb, KERN_ERR, "can't mount with "
3346 "both data=journal and dax");
3349 if (test_opt(sb, DELALLOC))
3350 clear_opt(sb, DELALLOC);
3352 sb->s_iflags |= SB_I_CGROUPWB;
3355 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3356 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3358 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3359 (ext4_has_compat_features(sb) ||
3360 ext4_has_ro_compat_features(sb) ||
3361 ext4_has_incompat_features(sb)))
3362 ext4_msg(sb, KERN_WARNING,
3363 "feature flags set on rev 0 fs, "
3364 "running e2fsck is recommended");
3366 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3367 set_opt2(sb, HURD_COMPAT);
3368 if (ext4_has_feature_64bit(sb)) {
3369 ext4_msg(sb, KERN_ERR,
3370 "The Hurd can't support 64-bit file systems");
3375 if (IS_EXT2_SB(sb)) {
3376 if (ext2_feature_set_ok(sb))
3377 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3378 "using the ext4 subsystem");
3380 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3381 "to feature incompatibilities");
3386 if (IS_EXT3_SB(sb)) {
3387 if (ext3_feature_set_ok(sb))
3388 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3389 "using the ext4 subsystem");
3391 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3392 "to feature incompatibilities");
3398 * Check feature flags regardless of the revision level, since we
3399 * previously didn't change the revision level when setting the flags,
3400 * so there is a chance incompat flags are set on a rev 0 filesystem.
3402 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3405 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3406 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3407 blocksize > EXT4_MAX_BLOCK_SIZE) {
3408 ext4_msg(sb, KERN_ERR,
3409 "Unsupported filesystem blocksize %d", blocksize);
3413 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
3414 if (blocksize != PAGE_SIZE) {
3415 ext4_msg(sb, KERN_ERR,
3416 "error: unsupported blocksize for dax");
3419 if (!sb->s_bdev->bd_disk->fops->direct_access) {
3420 ext4_msg(sb, KERN_ERR,
3421 "error: device does not support dax");
3426 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
3427 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
3428 es->s_encryption_level);
3432 if (sb->s_blocksize != blocksize) {
3433 /* Validate the filesystem blocksize */
3434 if (!sb_set_blocksize(sb, blocksize)) {
3435 ext4_msg(sb, KERN_ERR, "bad block size %d",
3441 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3442 offset = do_div(logical_sb_block, blocksize);
3443 bh = sb_bread_unmovable(sb, logical_sb_block);
3445 ext4_msg(sb, KERN_ERR,
3446 "Can't read superblock on 2nd try");
3449 es = (struct ext4_super_block *)(bh->b_data + offset);
3451 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3452 ext4_msg(sb, KERN_ERR,
3453 "Magic mismatch, very weird!");
3458 has_huge_files = ext4_has_feature_huge_file(sb);
3459 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3461 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3463 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3464 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3465 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3467 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3468 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3469 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3470 (!is_power_of_2(sbi->s_inode_size)) ||
3471 (sbi->s_inode_size > blocksize)) {
3472 ext4_msg(sb, KERN_ERR,
3473 "unsupported inode size: %d",
3477 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3478 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3481 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3482 if (ext4_has_feature_64bit(sb)) {
3483 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3484 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3485 !is_power_of_2(sbi->s_desc_size)) {
3486 ext4_msg(sb, KERN_ERR,
3487 "unsupported descriptor size %lu",
3492 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3494 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3495 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3496 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3499 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3500 if (sbi->s_inodes_per_block == 0)
3502 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3503 sbi->s_inodes_per_block;
3504 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3506 sbi->s_mount_state = le16_to_cpu(es->s_state);
3507 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3508 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3510 for (i = 0; i < 4; i++)
3511 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3512 sbi->s_def_hash_version = es->s_def_hash_version;
3513 if (ext4_has_feature_dir_index(sb)) {
3514 i = le32_to_cpu(es->s_flags);
3515 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3516 sbi->s_hash_unsigned = 3;
3517 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3518 #ifdef __CHAR_UNSIGNED__
3519 if (!(sb->s_flags & MS_RDONLY))
3521 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3522 sbi->s_hash_unsigned = 3;
3524 if (!(sb->s_flags & MS_RDONLY))
3526 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3531 /* Handle clustersize */
3532 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3533 has_bigalloc = ext4_has_feature_bigalloc(sb);
3535 if (clustersize < blocksize) {
3536 ext4_msg(sb, KERN_ERR,
3537 "cluster size (%d) smaller than "
3538 "block size (%d)", clustersize, blocksize);
3541 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3542 le32_to_cpu(es->s_log_block_size);
3543 sbi->s_clusters_per_group =
3544 le32_to_cpu(es->s_clusters_per_group);
3545 if (sbi->s_clusters_per_group > blocksize * 8) {
3546 ext4_msg(sb, KERN_ERR,
3547 "#clusters per group too big: %lu",
3548 sbi->s_clusters_per_group);
3551 if (sbi->s_blocks_per_group !=
3552 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3553 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3554 "clusters per group (%lu) inconsistent",
3555 sbi->s_blocks_per_group,
3556 sbi->s_clusters_per_group);
3560 if (clustersize != blocksize) {
3561 ext4_warning(sb, "fragment/cluster size (%d) != "
3562 "block size (%d)", clustersize,
3564 clustersize = blocksize;
3566 if (sbi->s_blocks_per_group > blocksize * 8) {
3567 ext4_msg(sb, KERN_ERR,
3568 "#blocks per group too big: %lu",
3569 sbi->s_blocks_per_group);
3572 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3573 sbi->s_cluster_bits = 0;
3575 sbi->s_cluster_ratio = clustersize / blocksize;
3577 if (sbi->s_inodes_per_group > blocksize * 8) {
3578 ext4_msg(sb, KERN_ERR,
3579 "#inodes per group too big: %lu",
3580 sbi->s_inodes_per_group);
3584 /* Do we have standard group size of clustersize * 8 blocks ? */
3585 if (sbi->s_blocks_per_group == clustersize << 3)
3586 set_opt2(sb, STD_GROUP_SIZE);
3589 * Test whether we have more sectors than will fit in sector_t,
3590 * and whether the max offset is addressable by the page cache.
3592 err = generic_check_addressable(sb->s_blocksize_bits,
3593 ext4_blocks_count(es));
3595 ext4_msg(sb, KERN_ERR, "filesystem"
3596 " too large to mount safely on this system");
3597 if (sizeof(sector_t) < 8)
3598 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3602 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3605 /* check blocks count against device size */
3606 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3607 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3608 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3609 "exceeds size of device (%llu blocks)",
3610 ext4_blocks_count(es), blocks_count);
3615 * It makes no sense for the first data block to be beyond the end
3616 * of the filesystem.
3618 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3619 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3620 "block %u is beyond end of filesystem (%llu)",
3621 le32_to_cpu(es->s_first_data_block),
3622 ext4_blocks_count(es));
3625 blocks_count = (ext4_blocks_count(es) -
3626 le32_to_cpu(es->s_first_data_block) +
3627 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3628 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3629 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3630 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3631 "(block count %llu, first data block %u, "
3632 "blocks per group %lu)", sbi->s_groups_count,
3633 ext4_blocks_count(es),
3634 le32_to_cpu(es->s_first_data_block),
3635 EXT4_BLOCKS_PER_GROUP(sb));
3638 sbi->s_groups_count = blocks_count;
3639 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3640 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3641 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3642 EXT4_DESC_PER_BLOCK(sb);
3643 sbi->s_group_desc = ext4_kvmalloc(db_count *
3644 sizeof(struct buffer_head *),
3646 if (sbi->s_group_desc == NULL) {
3647 ext4_msg(sb, KERN_ERR, "not enough memory");
3652 bgl_lock_init(sbi->s_blockgroup_lock);
3654 for (i = 0; i < db_count; i++) {
3655 block = descriptor_loc(sb, logical_sb_block, i);
3656 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
3657 if (!sbi->s_group_desc[i]) {
3658 ext4_msg(sb, KERN_ERR,
3659 "can't read group descriptor %d", i);
3664 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3665 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3666 ret = -EFSCORRUPTED;
3670 sbi->s_gdb_count = db_count;
3671 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3672 spin_lock_init(&sbi->s_next_gen_lock);
3674 setup_timer(&sbi->s_err_report, print_daily_error_info,
3675 (unsigned long) sb);
3677 /* Register extent status tree shrinker */
3678 if (ext4_es_register_shrinker(sbi))
3681 sbi->s_stripe = ext4_get_stripe_size(sbi);
3682 sbi->s_extent_max_zeroout_kb = 32;
3685 * set up enough so that it can read an inode
3687 sb->s_op = &ext4_sops;
3688 sb->s_export_op = &ext4_export_ops;
3689 sb->s_xattr = ext4_xattr_handlers;
3691 sb->dq_op = &ext4_quota_operations;
3692 if (ext4_has_feature_quota(sb))
3693 sb->s_qcop = &dquot_quotactl_sysfile_ops;
3695 sb->s_qcop = &ext4_qctl_operations;
3696 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
3698 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3700 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3701 mutex_init(&sbi->s_orphan_lock);
3705 needs_recovery = (es->s_last_orphan != 0 ||
3706 ext4_has_feature_journal_needs_recovery(sb));
3708 if (ext4_has_feature_mmp(sb) && !(sb->s_flags & MS_RDONLY))
3709 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3710 goto failed_mount3a;
3713 * The first inode we look at is the journal inode. Don't try
3714 * root first: it may be modified in the journal!
3716 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
3717 if (ext4_load_journal(sb, es, journal_devnum))
3718 goto failed_mount3a;
3719 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3720 ext4_has_feature_journal_needs_recovery(sb)) {
3721 ext4_msg(sb, KERN_ERR, "required journal recovery "
3722 "suppressed and not mounted read-only");
3723 goto failed_mount_wq;
3725 /* Nojournal mode, all journal mount options are illegal */
3726 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
3727 ext4_msg(sb, KERN_ERR, "can't mount with "
3728 "journal_checksum, fs mounted w/o journal");
3729 goto failed_mount_wq;
3731 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3732 ext4_msg(sb, KERN_ERR, "can't mount with "
3733 "journal_async_commit, fs mounted w/o journal");
3734 goto failed_mount_wq;
3736 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
3737 ext4_msg(sb, KERN_ERR, "can't mount with "
3738 "commit=%lu, fs mounted w/o journal",
3739 sbi->s_commit_interval / HZ);
3740 goto failed_mount_wq;
3742 if (EXT4_MOUNT_DATA_FLAGS &
3743 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
3744 ext4_msg(sb, KERN_ERR, "can't mount with "
3745 "data=, fs mounted w/o journal");
3746 goto failed_mount_wq;
3748 sbi->s_def_mount_opt &= EXT4_MOUNT_JOURNAL_CHECKSUM;
3749 clear_opt(sb, JOURNAL_CHECKSUM);
3750 clear_opt(sb, DATA_FLAGS);
3751 sbi->s_journal = NULL;
3756 if (ext4_has_feature_64bit(sb) &&
3757 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3758 JBD2_FEATURE_INCOMPAT_64BIT)) {
3759 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3760 goto failed_mount_wq;
3763 if (!set_journal_csum_feature_set(sb)) {
3764 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3766 goto failed_mount_wq;
3769 /* We have now updated the journal if required, so we can
3770 * validate the data journaling mode. */
3771 switch (test_opt(sb, DATA_FLAGS)) {
3773 /* No mode set, assume a default based on the journal
3774 * capabilities: ORDERED_DATA if the journal can
3775 * cope, else JOURNAL_DATA
3777 if (jbd2_journal_check_available_features
3778 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3779 set_opt(sb, ORDERED_DATA);
3781 set_opt(sb, JOURNAL_DATA);
3784 case EXT4_MOUNT_ORDERED_DATA:
3785 case EXT4_MOUNT_WRITEBACK_DATA:
3786 if (!jbd2_journal_check_available_features
3787 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3788 ext4_msg(sb, KERN_ERR, "Journal does not support "
3789 "requested data journaling mode");
3790 goto failed_mount_wq;
3795 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3797 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3800 if (ext4_mballoc_ready) {
3801 sbi->s_mb_cache = ext4_xattr_create_cache(sb->s_id);
3802 if (!sbi->s_mb_cache) {
3803 ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
3804 goto failed_mount_wq;
3808 if ((DUMMY_ENCRYPTION_ENABLED(sbi) || ext4_has_feature_encrypt(sb)) &&
3809 (blocksize != PAGE_CACHE_SIZE)) {
3810 ext4_msg(sb, KERN_ERR,
3811 "Unsupported blocksize for fs encryption");
3812 goto failed_mount_wq;
3815 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !(sb->s_flags & MS_RDONLY) &&
3816 !ext4_has_feature_encrypt(sb)) {
3817 ext4_set_feature_encrypt(sb);
3818 ext4_commit_super(sb, 1);
3822 * Get the # of file system overhead blocks from the
3823 * superblock if present.
3825 if (es->s_overhead_clusters)
3826 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
3828 err = ext4_calculate_overhead(sb);
3830 goto failed_mount_wq;
3834 * The maximum number of concurrent works can be high and
3835 * concurrency isn't really necessary. Limit it to 1.
3837 EXT4_SB(sb)->rsv_conversion_wq =
3838 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
3839 if (!EXT4_SB(sb)->rsv_conversion_wq) {
3840 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
3846 * The jbd2_journal_load will have done any necessary log recovery,
3847 * so we can safely mount the rest of the filesystem now.
3850 root = ext4_iget(sb, EXT4_ROOT_INO);
3852 ext4_msg(sb, KERN_ERR, "get root inode failed");
3853 ret = PTR_ERR(root);
3857 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
3858 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
3862 sb->s_root = d_make_root(root);
3864 ext4_msg(sb, KERN_ERR, "get root dentry failed");
3869 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
3870 sb->s_flags |= MS_RDONLY;
3872 /* determine the minimum size of new large inodes, if present */
3873 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3874 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3875 EXT4_GOOD_OLD_INODE_SIZE;
3876 if (ext4_has_feature_extra_isize(sb)) {
3877 if (sbi->s_want_extra_isize <
3878 le16_to_cpu(es->s_want_extra_isize))
3879 sbi->s_want_extra_isize =
3880 le16_to_cpu(es->s_want_extra_isize);
3881 if (sbi->s_want_extra_isize <
3882 le16_to_cpu(es->s_min_extra_isize))
3883 sbi->s_want_extra_isize =
3884 le16_to_cpu(es->s_min_extra_isize);
3887 /* Check if enough inode space is available */
3888 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3889 sbi->s_inode_size) {
3890 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3891 EXT4_GOOD_OLD_INODE_SIZE;
3892 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3896 ext4_set_resv_clusters(sb);
3898 err = ext4_setup_system_zone(sb);
3900 ext4_msg(sb, KERN_ERR, "failed to initialize system "
3902 goto failed_mount4a;
3906 err = ext4_mb_init(sb);
3908 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
3913 block = ext4_count_free_clusters(sb);
3914 ext4_free_blocks_count_set(sbi->s_es,
3915 EXT4_C2B(sbi, block));
3916 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
3919 unsigned long freei = ext4_count_free_inodes(sb);
3920 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
3921 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
3925 err = percpu_counter_init(&sbi->s_dirs_counter,
3926 ext4_count_dirs(sb), GFP_KERNEL);
3928 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
3931 ext4_msg(sb, KERN_ERR, "insufficient memory");
3935 if (ext4_has_feature_flex_bg(sb))
3936 if (!ext4_fill_flex_info(sb)) {
3937 ext4_msg(sb, KERN_ERR,
3938 "unable to initialize "
3939 "flex_bg meta info!");
3943 err = ext4_register_li_request(sb, first_not_zeroed);
3947 err = ext4_register_sysfs(sb);
3952 /* Enable quota usage during mount. */
3953 if (ext4_has_feature_quota(sb) && !(sb->s_flags & MS_RDONLY)) {
3954 err = ext4_enable_quotas(sb);
3958 #endif /* CONFIG_QUOTA */
3960 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3961 ext4_orphan_cleanup(sb, es);
3962 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
3963 if (needs_recovery) {
3964 ext4_msg(sb, KERN_INFO, "recovery complete");
3965 ext4_mark_recovery_complete(sb, es);
3967 if (EXT4_SB(sb)->s_journal) {
3968 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3969 descr = " journalled data mode";
3970 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3971 descr = " ordered data mode";
3973 descr = " writeback data mode";
3975 descr = "out journal";
3977 if (test_opt(sb, DISCARD)) {
3978 struct request_queue *q = bdev_get_queue(sb->s_bdev);
3979 if (!blk_queue_discard(q))
3980 ext4_msg(sb, KERN_WARNING,
3981 "mounting with \"discard\" option, but "
3982 "the device does not support discard");
3985 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
3986 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
3987 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3988 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
3990 if (es->s_error_count)
3991 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
3993 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
3994 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
3995 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
3996 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4003 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4008 ext4_unregister_sysfs(sb);
4011 ext4_unregister_li_request(sb);
4013 ext4_mb_release(sb);
4014 if (sbi->s_flex_groups)
4015 kvfree(sbi->s_flex_groups);
4016 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4017 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4018 percpu_counter_destroy(&sbi->s_dirs_counter);
4019 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4021 ext4_ext_release(sb);
4022 ext4_release_system_zone(sb);
4027 ext4_msg(sb, KERN_ERR, "mount failed");
4028 if (EXT4_SB(sb)->rsv_conversion_wq)
4029 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4031 if (sbi->s_journal) {
4032 jbd2_journal_destroy(sbi->s_journal);
4033 sbi->s_journal = NULL;
4036 ext4_es_unregister_shrinker(sbi);
4038 del_timer_sync(&sbi->s_err_report);
4040 kthread_stop(sbi->s_mmp_tsk);
4042 for (i = 0; i < db_count; i++)
4043 brelse(sbi->s_group_desc[i]);
4044 kvfree(sbi->s_group_desc);
4046 if (sbi->s_chksum_driver)
4047 crypto_free_shash(sbi->s_chksum_driver);
4049 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4050 kfree(sbi->s_qf_names[i]);
4052 ext4_blkdev_remove(sbi);
4055 sb->s_fs_info = NULL;
4056 kfree(sbi->s_blockgroup_lock);
4060 return err ? err : ret;
4064 * Setup any per-fs journal parameters now. We'll do this both on
4065 * initial mount, once the journal has been initialised but before we've
4066 * done any recovery; and again on any subsequent remount.
4068 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4070 struct ext4_sb_info *sbi = EXT4_SB(sb);
4072 journal->j_commit_interval = sbi->s_commit_interval;
4073 journal->j_min_batch_time = sbi->s_min_batch_time;
4074 journal->j_max_batch_time = sbi->s_max_batch_time;
4076 write_lock(&journal->j_state_lock);
4077 if (test_opt(sb, BARRIER))
4078 journal->j_flags |= JBD2_BARRIER;
4080 journal->j_flags &= ~JBD2_BARRIER;
4081 if (test_opt(sb, DATA_ERR_ABORT))
4082 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4084 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4085 write_unlock(&journal->j_state_lock);
4088 static journal_t *ext4_get_journal(struct super_block *sb,
4089 unsigned int journal_inum)
4091 struct inode *journal_inode;
4094 BUG_ON(!ext4_has_feature_journal(sb));
4096 /* First, test for the existence of a valid inode on disk. Bad
4097 * things happen if we iget() an unused inode, as the subsequent
4098 * iput() will try to delete it. */
4100 journal_inode = ext4_iget(sb, journal_inum);
4101 if (IS_ERR(journal_inode)) {
4102 ext4_msg(sb, KERN_ERR, "no journal found");
4105 if (!journal_inode->i_nlink) {
4106 make_bad_inode(journal_inode);
4107 iput(journal_inode);
4108 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4112 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4113 journal_inode, journal_inode->i_size);
4114 if (!S_ISREG(journal_inode->i_mode)) {
4115 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4116 iput(journal_inode);
4120 journal = jbd2_journal_init_inode(journal_inode);
4122 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4123 iput(journal_inode);
4126 journal->j_private = sb;
4127 ext4_init_journal_params(sb, journal);
4131 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4134 struct buffer_head *bh;
4138 int hblock, blocksize;
4139 ext4_fsblk_t sb_block;
4140 unsigned long offset;
4141 struct ext4_super_block *es;
4142 struct block_device *bdev;
4144 BUG_ON(!ext4_has_feature_journal(sb));
4146 bdev = ext4_blkdev_get(j_dev, sb);
4150 blocksize = sb->s_blocksize;
4151 hblock = bdev_logical_block_size(bdev);
4152 if (blocksize < hblock) {
4153 ext4_msg(sb, KERN_ERR,
4154 "blocksize too small for journal device");
4158 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4159 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4160 set_blocksize(bdev, blocksize);
4161 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4162 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4163 "external journal");
4167 es = (struct ext4_super_block *) (bh->b_data + offset);
4168 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4169 !(le32_to_cpu(es->s_feature_incompat) &
4170 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4171 ext4_msg(sb, KERN_ERR, "external journal has "
4177 if ((le32_to_cpu(es->s_feature_ro_compat) &
4178 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4179 es->s_checksum != ext4_superblock_csum(sb, es)) {
4180 ext4_msg(sb, KERN_ERR, "external journal has "
4181 "corrupt superblock");
4186 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4187 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4192 len = ext4_blocks_count(es);
4193 start = sb_block + 1;
4194 brelse(bh); /* we're done with the superblock */
4196 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4197 start, len, blocksize);
4199 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4202 journal->j_private = sb;
4203 ll_rw_block(READ | REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4204 wait_on_buffer(journal->j_sb_buffer);
4205 if (!buffer_uptodate(journal->j_sb_buffer)) {
4206 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4209 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4210 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4211 "user (unsupported) - %d",
4212 be32_to_cpu(journal->j_superblock->s_nr_users));
4215 EXT4_SB(sb)->journal_bdev = bdev;
4216 ext4_init_journal_params(sb, journal);
4220 jbd2_journal_destroy(journal);
4222 ext4_blkdev_put(bdev);
4226 static int ext4_load_journal(struct super_block *sb,
4227 struct ext4_super_block *es,
4228 unsigned long journal_devnum)
4231 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4234 int really_read_only;
4236 BUG_ON(!ext4_has_feature_journal(sb));
4238 if (journal_devnum &&
4239 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4240 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4241 "numbers have changed");
4242 journal_dev = new_decode_dev(journal_devnum);
4244 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4246 really_read_only = bdev_read_only(sb->s_bdev);
4249 * Are we loading a blank journal or performing recovery after a
4250 * crash? For recovery, we need to check in advance whether we
4251 * can get read-write access to the device.
4253 if (ext4_has_feature_journal_needs_recovery(sb)) {
4254 if (sb->s_flags & MS_RDONLY) {
4255 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4256 "required on readonly filesystem");
4257 if (really_read_only) {
4258 ext4_msg(sb, KERN_ERR, "write access "
4259 "unavailable, cannot proceed");
4262 ext4_msg(sb, KERN_INFO, "write access will "
4263 "be enabled during recovery");
4267 if (journal_inum && journal_dev) {
4268 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4269 "and inode journals!");
4274 if (!(journal = ext4_get_journal(sb, journal_inum)))
4277 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4281 if (!(journal->j_flags & JBD2_BARRIER))
4282 ext4_msg(sb, KERN_INFO, "barriers disabled");
4284 if (!ext4_has_feature_journal_needs_recovery(sb))
4285 err = jbd2_journal_wipe(journal, !really_read_only);
4287 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4289 memcpy(save, ((char *) es) +
4290 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4291 err = jbd2_journal_load(journal);
4293 memcpy(((char *) es) + EXT4_S_ERR_START,
4294 save, EXT4_S_ERR_LEN);
4299 ext4_msg(sb, KERN_ERR, "error loading journal");
4300 jbd2_journal_destroy(journal);
4304 EXT4_SB(sb)->s_journal = journal;
4305 ext4_clear_journal_err(sb, es);
4307 if (!really_read_only && journal_devnum &&
4308 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4309 es->s_journal_dev = cpu_to_le32(journal_devnum);
4311 /* Make sure we flush the recovery flag to disk. */
4312 ext4_commit_super(sb, 1);
4318 static int ext4_commit_super(struct super_block *sb, int sync)
4320 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4321 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4324 if (!sbh || block_device_ejected(sb))
4326 if (buffer_write_io_error(sbh)) {
4328 * Oh, dear. A previous attempt to write the
4329 * superblock failed. This could happen because the
4330 * USB device was yanked out. Or it could happen to
4331 * be a transient write error and maybe the block will
4332 * be remapped. Nothing we can do but to retry the
4333 * write and hope for the best.
4335 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4336 "superblock detected");
4337 clear_buffer_write_io_error(sbh);
4338 set_buffer_uptodate(sbh);
4341 * If the file system is mounted read-only, don't update the
4342 * superblock write time. This avoids updating the superblock
4343 * write time when we are mounting the root file system
4344 * read/only but we need to replay the journal; at that point,
4345 * for people who are east of GMT and who make their clock
4346 * tick in localtime for Windows bug-for-bug compatibility,
4347 * the clock is set in the future, and this will cause e2fsck
4348 * to complain and force a full file system check.
4350 if (!(sb->s_flags & MS_RDONLY))
4351 es->s_wtime = cpu_to_le32(get_seconds());
4352 if (sb->s_bdev->bd_part)
4353 es->s_kbytes_written =
4354 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4355 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4356 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4358 es->s_kbytes_written =
4359 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4360 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4361 ext4_free_blocks_count_set(es,
4362 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4363 &EXT4_SB(sb)->s_freeclusters_counter)));
4364 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4365 es->s_free_inodes_count =
4366 cpu_to_le32(percpu_counter_sum_positive(
4367 &EXT4_SB(sb)->s_freeinodes_counter));
4368 BUFFER_TRACE(sbh, "marking dirty");
4369 ext4_superblock_csum_set(sb);
4370 mark_buffer_dirty(sbh);
4372 error = __sync_dirty_buffer(sbh,
4373 test_opt(sb, BARRIER) ? WRITE_FUA : WRITE_SYNC);
4377 error = buffer_write_io_error(sbh);
4379 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4381 clear_buffer_write_io_error(sbh);
4382 set_buffer_uptodate(sbh);
4389 * Have we just finished recovery? If so, and if we are mounting (or
4390 * remounting) the filesystem readonly, then we will end up with a
4391 * consistent fs on disk. Record that fact.
4393 static void ext4_mark_recovery_complete(struct super_block *sb,
4394 struct ext4_super_block *es)
4396 journal_t *journal = EXT4_SB(sb)->s_journal;
4398 if (!ext4_has_feature_journal(sb)) {
4399 BUG_ON(journal != NULL);
4402 jbd2_journal_lock_updates(journal);
4403 if (jbd2_journal_flush(journal) < 0)
4406 if (ext4_has_feature_journal_needs_recovery(sb) &&
4407 sb->s_flags & MS_RDONLY) {
4408 ext4_clear_feature_journal_needs_recovery(sb);
4409 ext4_commit_super(sb, 1);
4413 jbd2_journal_unlock_updates(journal);
4417 * If we are mounting (or read-write remounting) a filesystem whose journal
4418 * has recorded an error from a previous lifetime, move that error to the
4419 * main filesystem now.
4421 static void ext4_clear_journal_err(struct super_block *sb,
4422 struct ext4_super_block *es)
4428 BUG_ON(!ext4_has_feature_journal(sb));
4430 journal = EXT4_SB(sb)->s_journal;
4433 * Now check for any error status which may have been recorded in the
4434 * journal by a prior ext4_error() or ext4_abort()
4437 j_errno = jbd2_journal_errno(journal);
4441 errstr = ext4_decode_error(sb, j_errno, nbuf);
4442 ext4_warning(sb, "Filesystem error recorded "
4443 "from previous mount: %s", errstr);
4444 ext4_warning(sb, "Marking fs in need of filesystem check.");
4446 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4447 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4448 ext4_commit_super(sb, 1);
4450 jbd2_journal_clear_err(journal);
4451 jbd2_journal_update_sb_errno(journal);
4456 * Force the running and committing transactions to commit,
4457 * and wait on the commit.
4459 int ext4_force_commit(struct super_block *sb)
4463 if (sb->s_flags & MS_RDONLY)
4466 journal = EXT4_SB(sb)->s_journal;
4467 return ext4_journal_force_commit(journal);
4470 static int ext4_sync_fs(struct super_block *sb, int wait)
4474 bool needs_barrier = false;
4475 struct ext4_sb_info *sbi = EXT4_SB(sb);
4477 trace_ext4_sync_fs(sb, wait);
4478 flush_workqueue(sbi->rsv_conversion_wq);
4480 * Writeback quota in non-journalled quota case - journalled quota has
4483 dquot_writeback_dquots(sb, -1);
4485 * Data writeback is possible w/o journal transaction, so barrier must
4486 * being sent at the end of the function. But we can skip it if
4487 * transaction_commit will do it for us.
4489 if (sbi->s_journal) {
4490 target = jbd2_get_latest_transaction(sbi->s_journal);
4491 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4492 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4493 needs_barrier = true;
4495 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4497 ret = jbd2_log_wait_commit(sbi->s_journal,
4500 } else if (wait && test_opt(sb, BARRIER))
4501 needs_barrier = true;
4502 if (needs_barrier) {
4504 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4513 * LVM calls this function before a (read-only) snapshot is created. This
4514 * gives us a chance to flush the journal completely and mark the fs clean.
4516 * Note that only this function cannot bring a filesystem to be in a clean
4517 * state independently. It relies on upper layer to stop all data & metadata
4520 static int ext4_freeze(struct super_block *sb)
4525 if (sb->s_flags & MS_RDONLY)
4528 journal = EXT4_SB(sb)->s_journal;
4531 /* Now we set up the journal barrier. */
4532 jbd2_journal_lock_updates(journal);
4535 * Don't clear the needs_recovery flag if we failed to
4536 * flush the journal.
4538 error = jbd2_journal_flush(journal);
4542 /* Journal blocked and flushed, clear needs_recovery flag. */
4543 ext4_clear_feature_journal_needs_recovery(sb);
4546 error = ext4_commit_super(sb, 1);
4549 /* we rely on upper layer to stop further updates */
4550 jbd2_journal_unlock_updates(journal);
4555 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4556 * flag here, even though the filesystem is not technically dirty yet.
4558 static int ext4_unfreeze(struct super_block *sb)
4560 if (sb->s_flags & MS_RDONLY)
4563 if (EXT4_SB(sb)->s_journal) {
4564 /* Reset the needs_recovery flag before the fs is unlocked. */
4565 ext4_set_feature_journal_needs_recovery(sb);
4568 ext4_commit_super(sb, 1);
4573 * Structure to save mount options for ext4_remount's benefit
4575 struct ext4_mount_options {
4576 unsigned long s_mount_opt;
4577 unsigned long s_mount_opt2;
4580 unsigned long s_commit_interval;
4581 u32 s_min_batch_time, s_max_batch_time;
4584 char *s_qf_names[EXT4_MAXQUOTAS];
4588 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4590 struct ext4_super_block *es;
4591 struct ext4_sb_info *sbi = EXT4_SB(sb);
4592 unsigned long old_sb_flags;
4593 struct ext4_mount_options old_opts;
4594 int enable_quota = 0;
4596 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4601 char *orig_data = kstrdup(data, GFP_KERNEL);
4603 /* Store the original options */
4604 old_sb_flags = sb->s_flags;
4605 old_opts.s_mount_opt = sbi->s_mount_opt;
4606 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4607 old_opts.s_resuid = sbi->s_resuid;
4608 old_opts.s_resgid = sbi->s_resgid;
4609 old_opts.s_commit_interval = sbi->s_commit_interval;
4610 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4611 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4613 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4614 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4615 if (sbi->s_qf_names[i]) {
4616 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4618 if (!old_opts.s_qf_names[i]) {
4619 for (j = 0; j < i; j++)
4620 kfree(old_opts.s_qf_names[j]);
4625 old_opts.s_qf_names[i] = NULL;
4627 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4628 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4630 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4635 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
4636 test_opt(sb, JOURNAL_CHECKSUM)) {
4637 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
4638 "during remount not supported; ignoring");
4639 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
4642 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4643 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4644 ext4_msg(sb, KERN_ERR, "can't mount with "
4645 "both data=journal and delalloc");
4649 if (test_opt(sb, DIOREAD_NOLOCK)) {
4650 ext4_msg(sb, KERN_ERR, "can't mount with "
4651 "both data=journal and dioread_nolock");
4655 if (test_opt(sb, DAX)) {
4656 ext4_msg(sb, KERN_ERR, "can't mount with "
4657 "both data=journal and dax");
4663 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
4664 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
4665 "dax flag with busy inodes while remounting");
4666 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
4669 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4670 ext4_abort(sb, "Abort forced by user");
4672 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4673 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4677 if (sbi->s_journal) {
4678 ext4_init_journal_params(sb, sbi->s_journal);
4679 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4682 if (*flags & MS_LAZYTIME)
4683 sb->s_flags |= MS_LAZYTIME;
4685 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4686 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4691 if (*flags & MS_RDONLY) {
4692 err = sync_filesystem(sb);
4695 err = dquot_suspend(sb, -1);
4700 * First of all, the unconditional stuff we have to do
4701 * to disable replay of the journal when we next remount
4703 sb->s_flags |= MS_RDONLY;
4706 * OK, test if we are remounting a valid rw partition
4707 * readonly, and if so set the rdonly flag and then
4708 * mark the partition as valid again.
4710 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4711 (sbi->s_mount_state & EXT4_VALID_FS))
4712 es->s_state = cpu_to_le16(sbi->s_mount_state);
4715 ext4_mark_recovery_complete(sb, es);
4717 /* Make sure we can mount this feature set readwrite */
4718 if (ext4_has_feature_readonly(sb) ||
4719 !ext4_feature_set_ok(sb, 0)) {
4724 * Make sure the group descriptor checksums
4725 * are sane. If they aren't, refuse to remount r/w.
4727 for (g = 0; g < sbi->s_groups_count; g++) {
4728 struct ext4_group_desc *gdp =
4729 ext4_get_group_desc(sb, g, NULL);
4731 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
4732 ext4_msg(sb, KERN_ERR,
4733 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4734 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
4735 le16_to_cpu(gdp->bg_checksum));
4742 * If we have an unprocessed orphan list hanging
4743 * around from a previously readonly bdev mount,
4744 * require a full umount/remount for now.
4746 if (es->s_last_orphan) {
4747 ext4_msg(sb, KERN_WARNING, "Couldn't "
4748 "remount RDWR because of unprocessed "
4749 "orphan inode list. Please "
4750 "umount/remount instead");
4756 * Mounting a RDONLY partition read-write, so reread
4757 * and store the current valid flag. (It may have
4758 * been changed by e2fsck since we originally mounted
4762 ext4_clear_journal_err(sb, es);
4763 sbi->s_mount_state = le16_to_cpu(es->s_state);
4764 if (!ext4_setup_super(sb, es, 0))
4765 sb->s_flags &= ~MS_RDONLY;
4766 if (ext4_has_feature_mmp(sb))
4767 if (ext4_multi_mount_protect(sb,
4768 le64_to_cpu(es->s_mmp_block))) {
4777 * Reinitialize lazy itable initialization thread based on
4780 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4781 ext4_unregister_li_request(sb);
4783 ext4_group_t first_not_zeroed;
4784 first_not_zeroed = ext4_has_uninit_itable(sb);
4785 ext4_register_li_request(sb, first_not_zeroed);
4788 ext4_setup_system_zone(sb);
4789 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
4790 ext4_commit_super(sb, 1);
4793 /* Release old quota file names */
4794 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4795 kfree(old_opts.s_qf_names[i]);
4797 if (sb_any_quota_suspended(sb))
4798 dquot_resume(sb, -1);
4799 else if (ext4_has_feature_quota(sb)) {
4800 err = ext4_enable_quotas(sb);
4807 *flags = (*flags & ~MS_LAZYTIME) | (sb->s_flags & MS_LAZYTIME);
4808 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4813 sb->s_flags = old_sb_flags;
4814 sbi->s_mount_opt = old_opts.s_mount_opt;
4815 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
4816 sbi->s_resuid = old_opts.s_resuid;
4817 sbi->s_resgid = old_opts.s_resgid;
4818 sbi->s_commit_interval = old_opts.s_commit_interval;
4819 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4820 sbi->s_max_batch_time = old_opts.s_max_batch_time;
4822 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4823 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
4824 kfree(sbi->s_qf_names[i]);
4825 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4833 static int ext4_statfs_project(struct super_block *sb,
4834 kprojid_t projid, struct kstatfs *buf)
4837 struct dquot *dquot;
4841 qid = make_kqid_projid(projid);
4842 dquot = dqget(sb, qid);
4844 return PTR_ERR(dquot);
4845 spin_lock(&dq_data_lock);
4847 limit = (dquot->dq_dqb.dqb_bsoftlimit ?
4848 dquot->dq_dqb.dqb_bsoftlimit :
4849 dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits;
4850 if (limit && buf->f_blocks > limit) {
4851 curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits;
4852 buf->f_blocks = limit;
4853 buf->f_bfree = buf->f_bavail =
4854 (buf->f_blocks > curblock) ?
4855 (buf->f_blocks - curblock) : 0;
4858 limit = dquot->dq_dqb.dqb_isoftlimit ?
4859 dquot->dq_dqb.dqb_isoftlimit :
4860 dquot->dq_dqb.dqb_ihardlimit;
4861 if (limit && buf->f_files > limit) {
4862 buf->f_files = limit;
4864 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
4865 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
4868 spin_unlock(&dq_data_lock);
4874 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
4876 struct super_block *sb = dentry->d_sb;
4877 struct ext4_sb_info *sbi = EXT4_SB(sb);
4878 struct ext4_super_block *es = sbi->s_es;
4879 ext4_fsblk_t overhead = 0, resv_blocks;
4882 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
4884 if (!test_opt(sb, MINIX_DF))
4885 overhead = sbi->s_overhead;
4887 buf->f_type = EXT4_SUPER_MAGIC;
4888 buf->f_bsize = sb->s_blocksize;
4889 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
4890 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4891 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
4892 /* prevent underflow in case that few free space is available */
4893 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
4894 buf->f_bavail = buf->f_bfree -
4895 (ext4_r_blocks_count(es) + resv_blocks);
4896 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
4898 buf->f_files = le32_to_cpu(es->s_inodes_count);
4899 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
4900 buf->f_namelen = EXT4_NAME_LEN;
4901 fsid = le64_to_cpup((void *)es->s_uuid) ^
4902 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4903 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4904 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
4907 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
4908 sb_has_quota_limits_enabled(sb, PRJQUOTA))
4909 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
4914 /* Helper function for writing quotas on sync - we need to start transaction
4915 * before quota file is locked for write. Otherwise the are possible deadlocks:
4916 * Process 1 Process 2
4917 * ext4_create() quota_sync()
4918 * jbd2_journal_start() write_dquot()
4919 * dquot_initialize() down(dqio_mutex)
4920 * down(dqio_mutex) jbd2_journal_start()
4926 static inline struct inode *dquot_to_inode(struct dquot *dquot)
4928 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
4931 static int ext4_write_dquot(struct dquot *dquot)
4935 struct inode *inode;
4937 inode = dquot_to_inode(dquot);
4938 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
4939 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
4941 return PTR_ERR(handle);
4942 ret = dquot_commit(dquot);
4943 err = ext4_journal_stop(handle);
4949 static int ext4_acquire_dquot(struct dquot *dquot)
4954 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
4955 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
4957 return PTR_ERR(handle);
4958 ret = dquot_acquire(dquot);
4959 err = ext4_journal_stop(handle);
4965 static int ext4_release_dquot(struct dquot *dquot)
4970 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
4971 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
4972 if (IS_ERR(handle)) {
4973 /* Release dquot anyway to avoid endless cycle in dqput() */
4974 dquot_release(dquot);
4975 return PTR_ERR(handle);
4977 ret = dquot_release(dquot);
4978 err = ext4_journal_stop(handle);
4984 static int ext4_mark_dquot_dirty(struct dquot *dquot)
4986 struct super_block *sb = dquot->dq_sb;
4987 struct ext4_sb_info *sbi = EXT4_SB(sb);
4989 /* Are we journaling quotas? */
4990 if (ext4_has_feature_quota(sb) ||
4991 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
4992 dquot_mark_dquot_dirty(dquot);
4993 return ext4_write_dquot(dquot);
4995 return dquot_mark_dquot_dirty(dquot);
4999 static int ext4_write_info(struct super_block *sb, int type)
5004 /* Data block + inode block */
5005 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
5007 return PTR_ERR(handle);
5008 ret = dquot_commit_info(sb, type);
5009 err = ext4_journal_stop(handle);
5016 * Turn on quotas during mount time - we need to find
5017 * the quota file and such...
5019 static int ext4_quota_on_mount(struct super_block *sb, int type)
5021 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5022 EXT4_SB(sb)->s_jquota_fmt, type);
5026 * Standard function to be called on quota_on
5028 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
5033 if (!test_opt(sb, QUOTA))
5036 /* Quotafile not on the same filesystem? */
5037 if (path->dentry->d_sb != sb)
5039 /* Journaling quota? */
5040 if (EXT4_SB(sb)->s_qf_names[type]) {
5041 /* Quotafile not in fs root? */
5042 if (path->dentry->d_parent != sb->s_root)
5043 ext4_msg(sb, KERN_WARNING,
5044 "Quota file not on filesystem root. "
5045 "Journaled quota will not work");
5049 * When we journal data on quota file, we have to flush journal to see
5050 * all updates to the file when we bypass pagecache...
5052 if (EXT4_SB(sb)->s_journal &&
5053 ext4_should_journal_data(d_inode(path->dentry))) {
5055 * We don't need to lock updates but journal_flush() could
5056 * otherwise be livelocked...
5058 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
5059 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
5060 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
5065 return dquot_quota_on(sb, type, format_id, path);
5068 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5072 struct inode *qf_inode;
5073 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5074 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5075 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5076 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
5079 BUG_ON(!ext4_has_feature_quota(sb));
5081 if (!qf_inums[type])
5084 qf_inode = ext4_iget(sb, qf_inums[type]);
5085 if (IS_ERR(qf_inode)) {
5086 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5087 return PTR_ERR(qf_inode);
5090 /* Don't account quota for quota files to avoid recursion */
5091 qf_inode->i_flags |= S_NOQUOTA;
5092 err = dquot_enable(qf_inode, type, format_id, flags);
5098 /* Enable usage tracking for all quota types. */
5099 static int ext4_enable_quotas(struct super_block *sb)
5102 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5103 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5104 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5105 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
5108 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5109 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
5110 if (qf_inums[type]) {
5111 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5112 DQUOT_USAGE_ENABLED);
5115 "Failed to enable quota tracking "
5116 "(type=%d, err=%d). Please run "
5117 "e2fsck to fix.", type, err);
5125 static int ext4_quota_off(struct super_block *sb, int type)
5127 struct inode *inode = sb_dqopt(sb)->files[type];
5130 /* Force all delayed allocation blocks to be allocated.
5131 * Caller already holds s_umount sem */
5132 if (test_opt(sb, DELALLOC))
5133 sync_filesystem(sb);
5138 /* Update modification times of quota files when userspace can
5139 * start looking at them */
5140 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5143 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5144 ext4_mark_inode_dirty(handle, inode);
5145 ext4_journal_stop(handle);
5148 return dquot_quota_off(sb, type);
5151 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5152 * acquiring the locks... As quota files are never truncated and quota code
5153 * itself serializes the operations (and no one else should touch the files)
5154 * we don't have to be afraid of races */
5155 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5156 size_t len, loff_t off)
5158 struct inode *inode = sb_dqopt(sb)->files[type];
5159 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5160 int offset = off & (sb->s_blocksize - 1);
5163 struct buffer_head *bh;
5164 loff_t i_size = i_size_read(inode);
5168 if (off+len > i_size)
5171 while (toread > 0) {
5172 tocopy = sb->s_blocksize - offset < toread ?
5173 sb->s_blocksize - offset : toread;
5174 bh = ext4_bread(NULL, inode, blk, 0);
5177 if (!bh) /* A hole? */
5178 memset(data, 0, tocopy);
5180 memcpy(data, bh->b_data+offset, tocopy);
5190 /* Write to quotafile (we know the transaction is already started and has
5191 * enough credits) */
5192 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5193 const char *data, size_t len, loff_t off)
5195 struct inode *inode = sb_dqopt(sb)->files[type];
5196 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5197 int err, offset = off & (sb->s_blocksize - 1);
5199 struct buffer_head *bh;
5200 handle_t *handle = journal_current_handle();
5202 if (EXT4_SB(sb)->s_journal && !handle) {
5203 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5204 " cancelled because transaction is not started",
5205 (unsigned long long)off, (unsigned long long)len);
5209 * Since we account only one data block in transaction credits,
5210 * then it is impossible to cross a block boundary.
5212 if (sb->s_blocksize - offset < len) {
5213 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5214 " cancelled because not block aligned",
5215 (unsigned long long)off, (unsigned long long)len);
5220 bh = ext4_bread(handle, inode, blk,
5221 EXT4_GET_BLOCKS_CREATE |
5222 EXT4_GET_BLOCKS_METADATA_NOFAIL);
5223 } while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) &&
5224 ext4_should_retry_alloc(inode->i_sb, &retries));
5229 BUFFER_TRACE(bh, "get write access");
5230 err = ext4_journal_get_write_access(handle, bh);
5236 memcpy(bh->b_data+offset, data, len);
5237 flush_dcache_page(bh->b_page);
5239 err = ext4_handle_dirty_metadata(handle, NULL, bh);
5242 if (inode->i_size < off + len) {
5243 i_size_write(inode, off + len);
5244 EXT4_I(inode)->i_disksize = inode->i_size;
5245 ext4_mark_inode_dirty(handle, inode);
5252 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5253 const char *dev_name, void *data)
5255 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5258 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
5259 static inline void register_as_ext2(void)
5261 int err = register_filesystem(&ext2_fs_type);
5264 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5267 static inline void unregister_as_ext2(void)
5269 unregister_filesystem(&ext2_fs_type);
5272 static inline int ext2_feature_set_ok(struct super_block *sb)
5274 if (ext4_has_unknown_ext2_incompat_features(sb))
5276 if (sb->s_flags & MS_RDONLY)
5278 if (ext4_has_unknown_ext2_ro_compat_features(sb))
5283 static inline void register_as_ext2(void) { }
5284 static inline void unregister_as_ext2(void) { }
5285 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5288 static inline void register_as_ext3(void)
5290 int err = register_filesystem(&ext3_fs_type);
5293 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5296 static inline void unregister_as_ext3(void)
5298 unregister_filesystem(&ext3_fs_type);
5301 static inline int ext3_feature_set_ok(struct super_block *sb)
5303 if (ext4_has_unknown_ext3_incompat_features(sb))
5305 if (!ext4_has_feature_journal(sb))
5307 if (sb->s_flags & MS_RDONLY)
5309 if (ext4_has_unknown_ext3_ro_compat_features(sb))
5314 static struct file_system_type ext4_fs_type = {
5315 .owner = THIS_MODULE,
5317 .mount = ext4_mount,
5318 .kill_sb = kill_block_super,
5319 .fs_flags = FS_REQUIRES_DEV,
5321 MODULE_ALIAS_FS("ext4");
5323 /* Shared across all ext4 file systems */
5324 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5325 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5327 static int __init ext4_init_fs(void)
5331 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
5332 ext4_li_info = NULL;
5333 mutex_init(&ext4_li_mtx);
5335 /* Build-time check for flags consistency */
5336 ext4_check_flag_values();
5338 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5339 mutex_init(&ext4__aio_mutex[i]);
5340 init_waitqueue_head(&ext4__ioend_wq[i]);
5343 err = ext4_init_es();
5347 err = ext4_init_pageio();
5351 err = ext4_init_system_zone();
5355 err = ext4_init_sysfs();
5359 err = ext4_init_mballoc();
5363 ext4_mballoc_ready = 1;
5364 err = init_inodecache();
5369 err = register_filesystem(&ext4_fs_type);
5375 unregister_as_ext2();
5376 unregister_as_ext3();
5377 destroy_inodecache();
5379 ext4_mballoc_ready = 0;
5380 ext4_exit_mballoc();
5384 ext4_exit_system_zone();
5393 static void __exit ext4_exit_fs(void)
5396 ext4_destroy_lazyinit_thread();
5397 unregister_as_ext2();
5398 unregister_as_ext3();
5399 unregister_filesystem(&ext4_fs_type);
5400 destroy_inodecache();
5401 ext4_exit_mballoc();
5403 ext4_exit_system_zone();
5408 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5409 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5410 MODULE_LICENSE("GPL");
5411 module_init(ext4_init_fs)
5412 module_exit(ext4_exit_fs)