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 struct ratelimit_state ext4_mount_msg_ratelimit;
60 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
61 unsigned long journal_devnum);
62 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
63 static int ext4_commit_super(struct super_block *sb, int sync);
64 static void ext4_mark_recovery_complete(struct super_block *sb,
65 struct ext4_super_block *es);
66 static void ext4_clear_journal_err(struct super_block *sb,
67 struct ext4_super_block *es);
68 static int ext4_sync_fs(struct super_block *sb, int wait);
69 static int ext4_remount(struct super_block *sb, int *flags, char *data);
70 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
71 static int ext4_unfreeze(struct super_block *sb);
72 static int ext4_freeze(struct super_block *sb);
73 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
74 const char *dev_name, void *data);
75 static inline int ext2_feature_set_ok(struct super_block *sb);
76 static inline int ext3_feature_set_ok(struct super_block *sb);
77 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
78 static void ext4_destroy_lazyinit_thread(void);
79 static void ext4_unregister_li_request(struct super_block *sb);
80 static void ext4_clear_request_list(void);
81 static struct inode *ext4_get_journal_inode(struct super_block *sb,
82 unsigned int journal_inum);
87 * Note the difference between i_mmap_sem (EXT4_I(inode)->i_mmap_sem) and
88 * i_mmap_rwsem (inode->i_mmap_rwsem)!
91 * mmap_sem -> sb_start_pagefault -> i_mmap_sem (r) -> transaction start ->
92 * page lock -> i_data_sem (rw)
94 * buffered write path:
95 * sb_start_write -> i_mutex -> mmap_sem
96 * sb_start_write -> i_mutex -> transaction start -> page lock ->
100 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) ->
101 * i_mmap_rwsem (w) -> page lock
102 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) ->
103 * transaction start -> i_data_sem (rw)
106 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) -> mmap_sem
107 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) ->
108 * transaction start -> i_data_sem (rw)
111 * transaction start -> page lock(s) -> i_data_sem (rw)
114 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
115 static struct file_system_type ext2_fs_type = {
116 .owner = THIS_MODULE,
119 .kill_sb = kill_block_super,
120 .fs_flags = FS_REQUIRES_DEV,
122 MODULE_ALIAS_FS("ext2");
123 MODULE_ALIAS("ext2");
124 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
126 #define IS_EXT2_SB(sb) (0)
130 static struct file_system_type ext3_fs_type = {
131 .owner = THIS_MODULE,
134 .kill_sb = kill_block_super,
135 .fs_flags = FS_REQUIRES_DEV,
137 MODULE_ALIAS_FS("ext3");
138 MODULE_ALIAS("ext3");
139 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
141 static int ext4_verify_csum_type(struct super_block *sb,
142 struct ext4_super_block *es)
144 if (!ext4_has_feature_metadata_csum(sb))
147 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
150 static __le32 ext4_superblock_csum(struct super_block *sb,
151 struct ext4_super_block *es)
153 struct ext4_sb_info *sbi = EXT4_SB(sb);
154 int offset = offsetof(struct ext4_super_block, s_checksum);
157 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
159 return cpu_to_le32(csum);
162 static int ext4_superblock_csum_verify(struct super_block *sb,
163 struct ext4_super_block *es)
165 if (!ext4_has_metadata_csum(sb))
168 return es->s_checksum == ext4_superblock_csum(sb, es);
171 void ext4_superblock_csum_set(struct super_block *sb)
173 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
175 if (!ext4_has_metadata_csum(sb))
178 es->s_checksum = ext4_superblock_csum(sb, es);
181 void *ext4_kvmalloc(size_t size, gfp_t flags)
185 ret = kmalloc(size, flags | __GFP_NOWARN);
187 ret = __vmalloc(size, flags, PAGE_KERNEL);
191 void *ext4_kvzalloc(size_t size, gfp_t flags)
195 ret = kzalloc(size, flags | __GFP_NOWARN);
197 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
201 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
202 struct ext4_group_desc *bg)
204 return le32_to_cpu(bg->bg_block_bitmap_lo) |
205 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
206 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
209 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
210 struct ext4_group_desc *bg)
212 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
213 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
214 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
217 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
218 struct ext4_group_desc *bg)
220 return le32_to_cpu(bg->bg_inode_table_lo) |
221 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
222 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
225 __u32 ext4_free_group_clusters(struct super_block *sb,
226 struct ext4_group_desc *bg)
228 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
229 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
230 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
233 __u32 ext4_free_inodes_count(struct super_block *sb,
234 struct ext4_group_desc *bg)
236 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
237 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
238 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
241 __u32 ext4_used_dirs_count(struct super_block *sb,
242 struct ext4_group_desc *bg)
244 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
245 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
246 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
249 __u32 ext4_itable_unused_count(struct super_block *sb,
250 struct ext4_group_desc *bg)
252 return le16_to_cpu(bg->bg_itable_unused_lo) |
253 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
254 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
257 void ext4_block_bitmap_set(struct super_block *sb,
258 struct ext4_group_desc *bg, ext4_fsblk_t blk)
260 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
261 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
262 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
265 void ext4_inode_bitmap_set(struct super_block *sb,
266 struct ext4_group_desc *bg, ext4_fsblk_t blk)
268 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
269 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
270 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
273 void ext4_inode_table_set(struct super_block *sb,
274 struct ext4_group_desc *bg, ext4_fsblk_t blk)
276 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
277 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
278 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
281 void ext4_free_group_clusters_set(struct super_block *sb,
282 struct ext4_group_desc *bg, __u32 count)
284 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
285 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
286 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
289 void ext4_free_inodes_set(struct super_block *sb,
290 struct ext4_group_desc *bg, __u32 count)
292 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
293 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
294 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
297 void ext4_used_dirs_set(struct super_block *sb,
298 struct ext4_group_desc *bg, __u32 count)
300 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
301 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
302 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
305 void ext4_itable_unused_set(struct super_block *sb,
306 struct ext4_group_desc *bg, __u32 count)
308 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
309 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
310 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
314 static void __save_error_info(struct super_block *sb, const char *func,
317 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
319 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
320 if (bdev_read_only(sb->s_bdev))
322 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
323 es->s_last_error_time = cpu_to_le32(get_seconds());
324 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
325 es->s_last_error_line = cpu_to_le32(line);
326 if (!es->s_first_error_time) {
327 es->s_first_error_time = es->s_last_error_time;
328 strncpy(es->s_first_error_func, func,
329 sizeof(es->s_first_error_func));
330 es->s_first_error_line = cpu_to_le32(line);
331 es->s_first_error_ino = es->s_last_error_ino;
332 es->s_first_error_block = es->s_last_error_block;
335 * Start the daily error reporting function if it hasn't been
338 if (!es->s_error_count)
339 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
340 le32_add_cpu(&es->s_error_count, 1);
343 static void save_error_info(struct super_block *sb, const char *func,
346 __save_error_info(sb, func, line);
347 ext4_commit_super(sb, 1);
351 * The del_gendisk() function uninitializes the disk-specific data
352 * structures, including the bdi structure, without telling anyone
353 * else. Once this happens, any attempt to call mark_buffer_dirty()
354 * (for example, by ext4_commit_super), will cause a kernel OOPS.
355 * This is a kludge to prevent these oops until we can put in a proper
356 * hook in del_gendisk() to inform the VFS and file system layers.
358 static int block_device_ejected(struct super_block *sb)
360 struct inode *bd_inode = sb->s_bdev->bd_inode;
361 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
363 return bdi->dev == NULL;
366 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
368 struct super_block *sb = journal->j_private;
369 struct ext4_sb_info *sbi = EXT4_SB(sb);
370 int error = is_journal_aborted(journal);
371 struct ext4_journal_cb_entry *jce;
373 BUG_ON(txn->t_state == T_FINISHED);
374 spin_lock(&sbi->s_md_lock);
375 while (!list_empty(&txn->t_private_list)) {
376 jce = list_entry(txn->t_private_list.next,
377 struct ext4_journal_cb_entry, jce_list);
378 list_del_init(&jce->jce_list);
379 spin_unlock(&sbi->s_md_lock);
380 jce->jce_func(sb, jce, error);
381 spin_lock(&sbi->s_md_lock);
383 spin_unlock(&sbi->s_md_lock);
386 /* Deal with the reporting of failure conditions on a filesystem such as
387 * inconsistencies detected or read IO failures.
389 * On ext2, we can store the error state of the filesystem in the
390 * superblock. That is not possible on ext4, because we may have other
391 * write ordering constraints on the superblock which prevent us from
392 * writing it out straight away; and given that the journal is about to
393 * be aborted, we can't rely on the current, or future, transactions to
394 * write out the superblock safely.
396 * We'll just use the jbd2_journal_abort() error code to record an error in
397 * the journal instead. On recovery, the journal will complain about
398 * that error until we've noted it down and cleared it.
401 static void ext4_handle_error(struct super_block *sb)
403 if (sb->s_flags & MS_RDONLY)
406 if (!test_opt(sb, ERRORS_CONT)) {
407 journal_t *journal = EXT4_SB(sb)->s_journal;
409 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
411 jbd2_journal_abort(journal, -EIO);
413 if (test_opt(sb, ERRORS_RO)) {
414 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
416 * Make sure updated value of ->s_mount_flags will be visible
417 * before ->s_flags update
420 sb->s_flags |= MS_RDONLY;
422 if (test_opt(sb, ERRORS_PANIC)) {
423 if (EXT4_SB(sb)->s_journal &&
424 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
426 panic("EXT4-fs (device %s): panic forced after error\n",
431 #define ext4_error_ratelimit(sb) \
432 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
435 void __ext4_error(struct super_block *sb, const char *function,
436 unsigned int line, const char *fmt, ...)
438 struct va_format vaf;
441 if (ext4_error_ratelimit(sb)) {
446 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
447 sb->s_id, function, line, current->comm, &vaf);
450 save_error_info(sb, function, line);
451 ext4_handle_error(sb);
454 void __ext4_error_inode(struct inode *inode, const char *function,
455 unsigned int line, ext4_fsblk_t block,
456 const char *fmt, ...)
459 struct va_format vaf;
460 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
462 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
463 es->s_last_error_block = cpu_to_le64(block);
464 if (ext4_error_ratelimit(inode->i_sb)) {
469 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
470 "inode #%lu: block %llu: comm %s: %pV\n",
471 inode->i_sb->s_id, function, line, inode->i_ino,
472 block, current->comm, &vaf);
474 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
475 "inode #%lu: comm %s: %pV\n",
476 inode->i_sb->s_id, function, line, inode->i_ino,
477 current->comm, &vaf);
480 save_error_info(inode->i_sb, function, line);
481 ext4_handle_error(inode->i_sb);
484 void __ext4_error_file(struct file *file, const char *function,
485 unsigned int line, ext4_fsblk_t block,
486 const char *fmt, ...)
489 struct va_format vaf;
490 struct ext4_super_block *es;
491 struct inode *inode = file_inode(file);
492 char pathname[80], *path;
494 es = EXT4_SB(inode->i_sb)->s_es;
495 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
496 if (ext4_error_ratelimit(inode->i_sb)) {
497 path = file_path(file, pathname, sizeof(pathname));
505 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
506 "block %llu: comm %s: path %s: %pV\n",
507 inode->i_sb->s_id, function, line, inode->i_ino,
508 block, current->comm, path, &vaf);
511 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
512 "comm %s: path %s: %pV\n",
513 inode->i_sb->s_id, function, line, inode->i_ino,
514 current->comm, path, &vaf);
517 save_error_info(inode->i_sb, function, line);
518 ext4_handle_error(inode->i_sb);
521 const char *ext4_decode_error(struct super_block *sb, int errno,
528 errstr = "Corrupt filesystem";
531 errstr = "Filesystem failed CRC";
534 errstr = "IO failure";
537 errstr = "Out of memory";
540 if (!sb || (EXT4_SB(sb)->s_journal &&
541 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
542 errstr = "Journal has aborted";
544 errstr = "Readonly filesystem";
547 /* If the caller passed in an extra buffer for unknown
548 * errors, textualise them now. Else we just return
551 /* Check for truncated error codes... */
552 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
561 /* __ext4_std_error decodes expected errors from journaling functions
562 * automatically and invokes the appropriate error response. */
564 void __ext4_std_error(struct super_block *sb, const char *function,
565 unsigned int line, int errno)
570 /* Special case: if the error is EROFS, and we're not already
571 * inside a transaction, then there's really no point in logging
573 if (errno == -EROFS && journal_current_handle() == NULL &&
574 (sb->s_flags & MS_RDONLY))
577 if (ext4_error_ratelimit(sb)) {
578 errstr = ext4_decode_error(sb, errno, nbuf);
579 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
580 sb->s_id, function, line, errstr);
583 save_error_info(sb, function, line);
584 ext4_handle_error(sb);
588 * ext4_abort is a much stronger failure handler than ext4_error. The
589 * abort function may be used to deal with unrecoverable failures such
590 * as journal IO errors or ENOMEM at a critical moment in log management.
592 * We unconditionally force the filesystem into an ABORT|READONLY state,
593 * unless the error response on the fs has been set to panic in which
594 * case we take the easy way out and panic immediately.
597 void __ext4_abort(struct super_block *sb, const char *function,
598 unsigned int line, const char *fmt, ...)
600 struct va_format vaf;
603 save_error_info(sb, function, line);
607 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: %pV\n",
608 sb->s_id, function, line, &vaf);
611 if ((sb->s_flags & MS_RDONLY) == 0) {
612 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
613 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
615 * Make sure updated value of ->s_mount_flags will be visible
616 * before ->s_flags update
619 sb->s_flags |= MS_RDONLY;
620 if (EXT4_SB(sb)->s_journal)
621 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
622 save_error_info(sb, function, line);
624 if (test_opt(sb, ERRORS_PANIC)) {
625 if (EXT4_SB(sb)->s_journal &&
626 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
628 panic("EXT4-fs panic from previous error\n");
632 void __ext4_msg(struct super_block *sb,
633 const char *prefix, const char *fmt, ...)
635 struct va_format vaf;
638 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
644 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
648 #define ext4_warning_ratelimit(sb) \
649 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \
652 void __ext4_warning(struct super_block *sb, const char *function,
653 unsigned int line, const char *fmt, ...)
655 struct va_format vaf;
658 if (!ext4_warning_ratelimit(sb))
664 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
665 sb->s_id, function, line, &vaf);
669 void __ext4_warning_inode(const struct inode *inode, const char *function,
670 unsigned int line, const char *fmt, ...)
672 struct va_format vaf;
675 if (!ext4_warning_ratelimit(inode->i_sb))
681 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
682 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
683 function, line, inode->i_ino, current->comm, &vaf);
687 void __ext4_grp_locked_error(const char *function, unsigned int line,
688 struct super_block *sb, ext4_group_t grp,
689 unsigned long ino, ext4_fsblk_t block,
690 const char *fmt, ...)
694 struct va_format vaf;
696 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
698 es->s_last_error_ino = cpu_to_le32(ino);
699 es->s_last_error_block = cpu_to_le64(block);
700 __save_error_info(sb, function, line);
702 if (ext4_error_ratelimit(sb)) {
706 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
707 sb->s_id, function, line, grp);
709 printk(KERN_CONT "inode %lu: ", ino);
711 printk(KERN_CONT "block %llu:",
712 (unsigned long long) block);
713 printk(KERN_CONT "%pV\n", &vaf);
717 if (test_opt(sb, ERRORS_CONT)) {
718 ext4_commit_super(sb, 0);
722 ext4_unlock_group(sb, grp);
723 ext4_handle_error(sb);
725 * We only get here in the ERRORS_RO case; relocking the group
726 * may be dangerous, but nothing bad will happen since the
727 * filesystem will have already been marked read/only and the
728 * journal has been aborted. We return 1 as a hint to callers
729 * who might what to use the return value from
730 * ext4_grp_locked_error() to distinguish between the
731 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
732 * aggressively from the ext4 function in question, with a
733 * more appropriate error code.
735 ext4_lock_group(sb, grp);
739 void ext4_update_dynamic_rev(struct super_block *sb)
741 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
743 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
747 "updating to rev %d because of new feature flag, "
748 "running e2fsck is recommended",
751 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
752 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
753 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
754 /* leave es->s_feature_*compat flags alone */
755 /* es->s_uuid will be set by e2fsck if empty */
758 * The rest of the superblock fields should be zero, and if not it
759 * means they are likely already in use, so leave them alone. We
760 * can leave it up to e2fsck to clean up any inconsistencies there.
765 * Open the external journal device
767 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
769 struct block_device *bdev;
770 char b[BDEVNAME_SIZE];
772 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
778 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
779 __bdevname(dev, b), PTR_ERR(bdev));
784 * Release the journal device
786 static void ext4_blkdev_put(struct block_device *bdev)
788 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
791 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
793 struct block_device *bdev;
794 bdev = sbi->journal_bdev;
796 ext4_blkdev_put(bdev);
797 sbi->journal_bdev = NULL;
801 static inline struct inode *orphan_list_entry(struct list_head *l)
803 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
806 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
810 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
811 le32_to_cpu(sbi->s_es->s_last_orphan));
813 printk(KERN_ERR "sb_info orphan list:\n");
814 list_for_each(l, &sbi->s_orphan) {
815 struct inode *inode = orphan_list_entry(l);
817 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
818 inode->i_sb->s_id, inode->i_ino, inode,
819 inode->i_mode, inode->i_nlink,
824 static void ext4_put_super(struct super_block *sb)
826 struct ext4_sb_info *sbi = EXT4_SB(sb);
827 struct ext4_super_block *es = sbi->s_es;
830 ext4_unregister_li_request(sb);
831 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
833 flush_workqueue(sbi->rsv_conversion_wq);
834 destroy_workqueue(sbi->rsv_conversion_wq);
836 if (sbi->s_journal) {
837 err = jbd2_journal_destroy(sbi->s_journal);
838 sbi->s_journal = NULL;
840 ext4_abort(sb, "Couldn't clean up the journal");
843 ext4_unregister_sysfs(sb);
844 ext4_es_unregister_shrinker(sbi);
845 del_timer_sync(&sbi->s_err_report);
846 ext4_release_system_zone(sb);
848 ext4_ext_release(sb);
850 if (!(sb->s_flags & MS_RDONLY)) {
851 ext4_clear_feature_journal_needs_recovery(sb);
852 es->s_state = cpu_to_le16(sbi->s_mount_state);
854 if (!(sb->s_flags & MS_RDONLY))
855 ext4_commit_super(sb, 1);
857 for (i = 0; i < sbi->s_gdb_count; i++)
858 brelse(sbi->s_group_desc[i]);
859 kvfree(sbi->s_group_desc);
860 kvfree(sbi->s_flex_groups);
861 percpu_counter_destroy(&sbi->s_freeclusters_counter);
862 percpu_counter_destroy(&sbi->s_freeinodes_counter);
863 percpu_counter_destroy(&sbi->s_dirs_counter);
864 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
865 percpu_free_rwsem(&sbi->s_journal_flag_rwsem);
868 for (i = 0; i < EXT4_MAXQUOTAS; i++)
869 kfree(sbi->s_qf_names[i]);
872 /* Debugging code just in case the in-memory inode orphan list
873 * isn't empty. The on-disk one can be non-empty if we've
874 * detected an error and taken the fs readonly, but the
875 * in-memory list had better be clean by this point. */
876 if (!list_empty(&sbi->s_orphan))
877 dump_orphan_list(sb, sbi);
878 J_ASSERT(list_empty(&sbi->s_orphan));
880 sync_blockdev(sb->s_bdev);
881 invalidate_bdev(sb->s_bdev);
882 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
884 * Invalidate the journal device's buffers. We don't want them
885 * floating about in memory - the physical journal device may
886 * hotswapped, and it breaks the `ro-after' testing code.
888 sync_blockdev(sbi->journal_bdev);
889 invalidate_bdev(sbi->journal_bdev);
890 ext4_blkdev_remove(sbi);
892 if (sbi->s_mb_cache) {
893 ext4_xattr_destroy_cache(sbi->s_mb_cache);
894 sbi->s_mb_cache = NULL;
897 kthread_stop(sbi->s_mmp_tsk);
898 sb->s_fs_info = NULL;
900 * Now that we are completely done shutting down the
901 * superblock, we need to actually destroy the kobject.
903 kobject_put(&sbi->s_kobj);
904 wait_for_completion(&sbi->s_kobj_unregister);
905 if (sbi->s_chksum_driver)
906 crypto_free_shash(sbi->s_chksum_driver);
907 kfree(sbi->s_blockgroup_lock);
911 static struct kmem_cache *ext4_inode_cachep;
914 * Called inside transaction, so use GFP_NOFS
916 static struct inode *ext4_alloc_inode(struct super_block *sb)
918 struct ext4_inode_info *ei;
920 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
924 ei->vfs_inode.i_version = 1;
925 spin_lock_init(&ei->i_raw_lock);
926 INIT_LIST_HEAD(&ei->i_prealloc_list);
927 spin_lock_init(&ei->i_prealloc_lock);
928 ext4_es_init_tree(&ei->i_es_tree);
929 rwlock_init(&ei->i_es_lock);
930 INIT_LIST_HEAD(&ei->i_es_list);
933 ei->i_es_shrink_lblk = 0;
934 ei->i_reserved_data_blocks = 0;
935 ei->i_reserved_meta_blocks = 0;
936 ei->i_allocated_meta_blocks = 0;
937 ei->i_da_metadata_calc_len = 0;
938 ei->i_da_metadata_calc_last_lblock = 0;
939 spin_lock_init(&(ei->i_block_reservation_lock));
941 ei->i_reserved_quota = 0;
942 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
945 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
946 spin_lock_init(&ei->i_completed_io_lock);
948 ei->i_datasync_tid = 0;
949 atomic_set(&ei->i_unwritten, 0);
950 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
951 return &ei->vfs_inode;
954 static int ext4_drop_inode(struct inode *inode)
956 int drop = generic_drop_inode(inode);
958 trace_ext4_drop_inode(inode, drop);
962 static void ext4_i_callback(struct rcu_head *head)
964 struct inode *inode = container_of(head, struct inode, i_rcu);
965 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
968 static void ext4_destroy_inode(struct inode *inode)
970 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
971 ext4_msg(inode->i_sb, KERN_ERR,
972 "Inode %lu (%p): orphan list check failed!",
973 inode->i_ino, EXT4_I(inode));
974 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
975 EXT4_I(inode), sizeof(struct ext4_inode_info),
979 call_rcu(&inode->i_rcu, ext4_i_callback);
982 static void init_once(void *foo)
984 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
986 INIT_LIST_HEAD(&ei->i_orphan);
987 init_rwsem(&ei->xattr_sem);
988 init_rwsem(&ei->i_data_sem);
989 init_rwsem(&ei->i_mmap_sem);
990 inode_init_once(&ei->vfs_inode);
993 static int __init init_inodecache(void)
995 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
996 sizeof(struct ext4_inode_info),
997 0, (SLAB_RECLAIM_ACCOUNT|
998 SLAB_MEM_SPREAD|SLAB_ACCOUNT),
1000 if (ext4_inode_cachep == NULL)
1005 static void destroy_inodecache(void)
1008 * Make sure all delayed rcu free inodes are flushed before we
1012 kmem_cache_destroy(ext4_inode_cachep);
1015 void ext4_clear_inode(struct inode *inode)
1017 invalidate_inode_buffers(inode);
1020 ext4_discard_preallocations(inode);
1021 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
1022 if (EXT4_I(inode)->jinode) {
1023 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1024 EXT4_I(inode)->jinode);
1025 jbd2_free_inode(EXT4_I(inode)->jinode);
1026 EXT4_I(inode)->jinode = NULL;
1028 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1029 fscrypt_put_encryption_info(inode, NULL);
1033 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1034 u64 ino, u32 generation)
1036 struct inode *inode;
1038 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1039 return ERR_PTR(-ESTALE);
1040 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1041 return ERR_PTR(-ESTALE);
1043 /* iget isn't really right if the inode is currently unallocated!!
1045 * ext4_read_inode will return a bad_inode if the inode had been
1046 * deleted, so we should be safe.
1048 * Currently we don't know the generation for parent directory, so
1049 * a generation of 0 means "accept any"
1051 inode = ext4_iget_normal(sb, ino);
1053 return ERR_CAST(inode);
1054 if (generation && inode->i_generation != generation) {
1056 return ERR_PTR(-ESTALE);
1062 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1063 int fh_len, int fh_type)
1065 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1066 ext4_nfs_get_inode);
1069 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1070 int fh_len, int fh_type)
1072 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1073 ext4_nfs_get_inode);
1077 * Try to release metadata pages (indirect blocks, directories) which are
1078 * mapped via the block device. Since these pages could have journal heads
1079 * which would prevent try_to_free_buffers() from freeing them, we must use
1080 * jbd2 layer's try_to_free_buffers() function to release them.
1082 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1085 journal_t *journal = EXT4_SB(sb)->s_journal;
1087 WARN_ON(PageChecked(page));
1088 if (!page_has_buffers(page))
1091 return jbd2_journal_try_to_free_buffers(journal, page,
1092 wait & ~__GFP_DIRECT_RECLAIM);
1093 return try_to_free_buffers(page);
1096 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1097 static int ext4_get_context(struct inode *inode, void *ctx, size_t len)
1099 return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1100 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len);
1103 static int ext4_key_prefix(struct inode *inode, u8 **key)
1105 *key = EXT4_SB(inode->i_sb)->key_prefix;
1106 return EXT4_SB(inode->i_sb)->key_prefix_size;
1109 static int ext4_prepare_context(struct inode *inode)
1111 return ext4_convert_inline_data(inode);
1114 static int ext4_set_context(struct inode *inode, const void *ctx, size_t len,
1120 /* fs_data is null when internally used. */
1122 res = ext4_xattr_set(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1123 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx,
1126 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1127 ext4_clear_inode_state(inode,
1128 EXT4_STATE_MAY_INLINE_DATA);
1133 handle = ext4_journal_start(inode, EXT4_HT_MISC,
1134 ext4_jbd2_credits_xattr(inode));
1136 return PTR_ERR(handle);
1138 res = ext4_xattr_set(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1139 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx,
1142 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1143 res = ext4_mark_inode_dirty(handle, inode);
1145 EXT4_ERROR_INODE(inode, "Failed to mark inode dirty");
1147 res2 = ext4_journal_stop(handle);
1153 static int ext4_dummy_context(struct inode *inode)
1155 return DUMMY_ENCRYPTION_ENABLED(EXT4_SB(inode->i_sb));
1158 static unsigned ext4_max_namelen(struct inode *inode)
1160 return S_ISLNK(inode->i_mode) ? inode->i_sb->s_blocksize :
1164 static struct fscrypt_operations ext4_cryptops = {
1165 .get_context = ext4_get_context,
1166 .key_prefix = ext4_key_prefix,
1167 .prepare_context = ext4_prepare_context,
1168 .set_context = ext4_set_context,
1169 .dummy_context = ext4_dummy_context,
1170 .is_encrypted = ext4_encrypted_inode,
1171 .empty_dir = ext4_empty_dir,
1172 .max_namelen = ext4_max_namelen,
1175 static struct fscrypt_operations ext4_cryptops = {
1176 .is_encrypted = ext4_encrypted_inode,
1181 static char *quotatypes[] = INITQFNAMES;
1182 #define QTYPE2NAME(t) (quotatypes[t])
1184 static int ext4_write_dquot(struct dquot *dquot);
1185 static int ext4_acquire_dquot(struct dquot *dquot);
1186 static int ext4_release_dquot(struct dquot *dquot);
1187 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1188 static int ext4_write_info(struct super_block *sb, int type);
1189 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1191 static int ext4_quota_off(struct super_block *sb, int type);
1192 static int ext4_quota_on_mount(struct super_block *sb, int type);
1193 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1194 size_t len, loff_t off);
1195 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1196 const char *data, size_t len, loff_t off);
1197 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1198 unsigned int flags);
1199 static int ext4_enable_quotas(struct super_block *sb);
1200 static int ext4_get_next_id(struct super_block *sb, struct kqid *qid);
1202 static struct dquot **ext4_get_dquots(struct inode *inode)
1204 return EXT4_I(inode)->i_dquot;
1207 static const struct dquot_operations ext4_quota_operations = {
1208 .get_reserved_space = ext4_get_reserved_space,
1209 .write_dquot = ext4_write_dquot,
1210 .acquire_dquot = ext4_acquire_dquot,
1211 .release_dquot = ext4_release_dquot,
1212 .mark_dirty = ext4_mark_dquot_dirty,
1213 .write_info = ext4_write_info,
1214 .alloc_dquot = dquot_alloc,
1215 .destroy_dquot = dquot_destroy,
1216 .get_projid = ext4_get_projid,
1217 .get_next_id = ext4_get_next_id,
1220 static const struct quotactl_ops ext4_qctl_operations = {
1221 .quota_on = ext4_quota_on,
1222 .quota_off = ext4_quota_off,
1223 .quota_sync = dquot_quota_sync,
1224 .get_state = dquot_get_state,
1225 .set_info = dquot_set_dqinfo,
1226 .get_dqblk = dquot_get_dqblk,
1227 .set_dqblk = dquot_set_dqblk,
1228 .get_nextdqblk = dquot_get_next_dqblk,
1232 static const struct super_operations ext4_sops = {
1233 .alloc_inode = ext4_alloc_inode,
1234 .destroy_inode = ext4_destroy_inode,
1235 .write_inode = ext4_write_inode,
1236 .dirty_inode = ext4_dirty_inode,
1237 .drop_inode = ext4_drop_inode,
1238 .evict_inode = ext4_evict_inode,
1239 .put_super = ext4_put_super,
1240 .sync_fs = ext4_sync_fs,
1241 .freeze_fs = ext4_freeze,
1242 .unfreeze_fs = ext4_unfreeze,
1243 .statfs = ext4_statfs,
1244 .remount_fs = ext4_remount,
1245 .show_options = ext4_show_options,
1247 .quota_read = ext4_quota_read,
1248 .quota_write = ext4_quota_write,
1249 .get_dquots = ext4_get_dquots,
1251 .bdev_try_to_free_page = bdev_try_to_free_page,
1254 static const struct export_operations ext4_export_ops = {
1255 .fh_to_dentry = ext4_fh_to_dentry,
1256 .fh_to_parent = ext4_fh_to_parent,
1257 .get_parent = ext4_get_parent,
1261 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1262 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1263 Opt_nouid32, Opt_debug, Opt_removed,
1264 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1265 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1266 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1267 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1268 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1269 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1270 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1271 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1272 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1273 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version, Opt_dax,
1274 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1275 Opt_lazytime, Opt_nolazytime,
1276 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1277 Opt_inode_readahead_blks, Opt_journal_ioprio,
1278 Opt_dioread_nolock, Opt_dioread_lock,
1279 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1280 Opt_max_dir_size_kb, Opt_nojournal_checksum,
1283 static const match_table_t tokens = {
1284 {Opt_bsd_df, "bsddf"},
1285 {Opt_minix_df, "minixdf"},
1286 {Opt_grpid, "grpid"},
1287 {Opt_grpid, "bsdgroups"},
1288 {Opt_nogrpid, "nogrpid"},
1289 {Opt_nogrpid, "sysvgroups"},
1290 {Opt_resgid, "resgid=%u"},
1291 {Opt_resuid, "resuid=%u"},
1293 {Opt_err_cont, "errors=continue"},
1294 {Opt_err_panic, "errors=panic"},
1295 {Opt_err_ro, "errors=remount-ro"},
1296 {Opt_nouid32, "nouid32"},
1297 {Opt_debug, "debug"},
1298 {Opt_removed, "oldalloc"},
1299 {Opt_removed, "orlov"},
1300 {Opt_user_xattr, "user_xattr"},
1301 {Opt_nouser_xattr, "nouser_xattr"},
1303 {Opt_noacl, "noacl"},
1304 {Opt_noload, "norecovery"},
1305 {Opt_noload, "noload"},
1306 {Opt_removed, "nobh"},
1307 {Opt_removed, "bh"},
1308 {Opt_commit, "commit=%u"},
1309 {Opt_min_batch_time, "min_batch_time=%u"},
1310 {Opt_max_batch_time, "max_batch_time=%u"},
1311 {Opt_journal_dev, "journal_dev=%u"},
1312 {Opt_journal_path, "journal_path=%s"},
1313 {Opt_journal_checksum, "journal_checksum"},
1314 {Opt_nojournal_checksum, "nojournal_checksum"},
1315 {Opt_journal_async_commit, "journal_async_commit"},
1316 {Opt_abort, "abort"},
1317 {Opt_data_journal, "data=journal"},
1318 {Opt_data_ordered, "data=ordered"},
1319 {Opt_data_writeback, "data=writeback"},
1320 {Opt_data_err_abort, "data_err=abort"},
1321 {Opt_data_err_ignore, "data_err=ignore"},
1322 {Opt_offusrjquota, "usrjquota="},
1323 {Opt_usrjquota, "usrjquota=%s"},
1324 {Opt_offgrpjquota, "grpjquota="},
1325 {Opt_grpjquota, "grpjquota=%s"},
1326 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1327 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1328 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1329 {Opt_grpquota, "grpquota"},
1330 {Opt_noquota, "noquota"},
1331 {Opt_quota, "quota"},
1332 {Opt_usrquota, "usrquota"},
1333 {Opt_prjquota, "prjquota"},
1334 {Opt_barrier, "barrier=%u"},
1335 {Opt_barrier, "barrier"},
1336 {Opt_nobarrier, "nobarrier"},
1337 {Opt_i_version, "i_version"},
1339 {Opt_stripe, "stripe=%u"},
1340 {Opt_delalloc, "delalloc"},
1341 {Opt_lazytime, "lazytime"},
1342 {Opt_nolazytime, "nolazytime"},
1343 {Opt_nodelalloc, "nodelalloc"},
1344 {Opt_removed, "mblk_io_submit"},
1345 {Opt_removed, "nomblk_io_submit"},
1346 {Opt_block_validity, "block_validity"},
1347 {Opt_noblock_validity, "noblock_validity"},
1348 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1349 {Opt_journal_ioprio, "journal_ioprio=%u"},
1350 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1351 {Opt_auto_da_alloc, "auto_da_alloc"},
1352 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1353 {Opt_dioread_nolock, "dioread_nolock"},
1354 {Opt_dioread_lock, "dioread_lock"},
1355 {Opt_discard, "discard"},
1356 {Opt_nodiscard, "nodiscard"},
1357 {Opt_init_itable, "init_itable=%u"},
1358 {Opt_init_itable, "init_itable"},
1359 {Opt_noinit_itable, "noinit_itable"},
1360 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1361 {Opt_test_dummy_encryption, "test_dummy_encryption"},
1362 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1363 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1364 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1365 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1366 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1370 static ext4_fsblk_t get_sb_block(void **data)
1372 ext4_fsblk_t sb_block;
1373 char *options = (char *) *data;
1375 if (!options || strncmp(options, "sb=", 3) != 0)
1376 return 1; /* Default location */
1379 /* TODO: use simple_strtoll with >32bit ext4 */
1380 sb_block = simple_strtoul(options, &options, 0);
1381 if (*options && *options != ',') {
1382 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1386 if (*options == ',')
1388 *data = (void *) options;
1393 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1394 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1395 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1398 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1400 struct ext4_sb_info *sbi = EXT4_SB(sb);
1404 if (sb_any_quota_loaded(sb) &&
1405 !sbi->s_qf_names[qtype]) {
1406 ext4_msg(sb, KERN_ERR,
1407 "Cannot change journaled "
1408 "quota options when quota turned on");
1411 if (ext4_has_feature_quota(sb)) {
1412 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1413 "ignored when QUOTA feature is enabled");
1416 qname = match_strdup(args);
1418 ext4_msg(sb, KERN_ERR,
1419 "Not enough memory for storing quotafile name");
1422 if (sbi->s_qf_names[qtype]) {
1423 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1426 ext4_msg(sb, KERN_ERR,
1427 "%s quota file already specified",
1431 if (strchr(qname, '/')) {
1432 ext4_msg(sb, KERN_ERR,
1433 "quotafile must be on filesystem root");
1436 sbi->s_qf_names[qtype] = qname;
1444 static int clear_qf_name(struct super_block *sb, int qtype)
1447 struct ext4_sb_info *sbi = EXT4_SB(sb);
1449 if (sb_any_quota_loaded(sb) &&
1450 sbi->s_qf_names[qtype]) {
1451 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1452 " when quota turned on");
1455 kfree(sbi->s_qf_names[qtype]);
1456 sbi->s_qf_names[qtype] = NULL;
1461 #define MOPT_SET 0x0001
1462 #define MOPT_CLEAR 0x0002
1463 #define MOPT_NOSUPPORT 0x0004
1464 #define MOPT_EXPLICIT 0x0008
1465 #define MOPT_CLEAR_ERR 0x0010
1466 #define MOPT_GTE0 0x0020
1469 #define MOPT_QFMT 0x0040
1471 #define MOPT_Q MOPT_NOSUPPORT
1472 #define MOPT_QFMT MOPT_NOSUPPORT
1474 #define MOPT_DATAJ 0x0080
1475 #define MOPT_NO_EXT2 0x0100
1476 #define MOPT_NO_EXT3 0x0200
1477 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1478 #define MOPT_STRING 0x0400
1480 static const struct mount_opts {
1484 } ext4_mount_opts[] = {
1485 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1486 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1487 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1488 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1489 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1490 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1491 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1492 MOPT_EXT4_ONLY | MOPT_SET},
1493 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1494 MOPT_EXT4_ONLY | MOPT_CLEAR},
1495 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1496 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1497 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1498 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1499 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1500 MOPT_EXT4_ONLY | MOPT_CLEAR},
1501 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1502 MOPT_EXT4_ONLY | MOPT_CLEAR},
1503 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1504 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1505 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1506 EXT4_MOUNT_JOURNAL_CHECKSUM),
1507 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1508 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1509 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1510 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1511 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1512 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1514 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1516 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1517 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1518 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1519 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1520 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1521 {Opt_commit, 0, MOPT_GTE0},
1522 {Opt_max_batch_time, 0, MOPT_GTE0},
1523 {Opt_min_batch_time, 0, MOPT_GTE0},
1524 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1525 {Opt_init_itable, 0, MOPT_GTE0},
1526 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
1527 {Opt_stripe, 0, MOPT_GTE0},
1528 {Opt_resuid, 0, MOPT_GTE0},
1529 {Opt_resgid, 0, MOPT_GTE0},
1530 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1531 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1532 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1533 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1534 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1535 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1536 MOPT_NO_EXT2 | MOPT_DATAJ},
1537 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1538 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1539 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1540 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1541 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1543 {Opt_acl, 0, MOPT_NOSUPPORT},
1544 {Opt_noacl, 0, MOPT_NOSUPPORT},
1546 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1547 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1548 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1549 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1551 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1553 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1555 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1556 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1557 MOPT_CLEAR | MOPT_Q},
1558 {Opt_usrjquota, 0, MOPT_Q},
1559 {Opt_grpjquota, 0, MOPT_Q},
1560 {Opt_offusrjquota, 0, MOPT_Q},
1561 {Opt_offgrpjquota, 0, MOPT_Q},
1562 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1563 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1564 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1565 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1566 {Opt_test_dummy_encryption, 0, MOPT_GTE0},
1570 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1571 substring_t *args, unsigned long *journal_devnum,
1572 unsigned int *journal_ioprio, int is_remount)
1574 struct ext4_sb_info *sbi = EXT4_SB(sb);
1575 const struct mount_opts *m;
1581 if (token == Opt_usrjquota)
1582 return set_qf_name(sb, USRQUOTA, &args[0]);
1583 else if (token == Opt_grpjquota)
1584 return set_qf_name(sb, GRPQUOTA, &args[0]);
1585 else if (token == Opt_offusrjquota)
1586 return clear_qf_name(sb, USRQUOTA);
1587 else if (token == Opt_offgrpjquota)
1588 return clear_qf_name(sb, GRPQUOTA);
1592 case Opt_nouser_xattr:
1593 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1596 return 1; /* handled by get_sb_block() */
1598 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1601 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1604 sb->s_flags |= MS_I_VERSION;
1607 sb->s_flags |= MS_LAZYTIME;
1609 case Opt_nolazytime:
1610 sb->s_flags &= ~MS_LAZYTIME;
1614 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1615 if (token == m->token)
1618 if (m->token == Opt_err) {
1619 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1620 "or missing value", opt);
1624 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1625 ext4_msg(sb, KERN_ERR,
1626 "Mount option \"%s\" incompatible with ext2", opt);
1629 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1630 ext4_msg(sb, KERN_ERR,
1631 "Mount option \"%s\" incompatible with ext3", opt);
1635 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
1637 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1639 if (m->flags & MOPT_EXPLICIT) {
1640 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
1641 set_opt2(sb, EXPLICIT_DELALLOC);
1642 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
1643 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
1647 if (m->flags & MOPT_CLEAR_ERR)
1648 clear_opt(sb, ERRORS_MASK);
1649 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1650 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1651 "options when quota turned on");
1655 if (m->flags & MOPT_NOSUPPORT) {
1656 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1657 } else if (token == Opt_commit) {
1659 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1660 sbi->s_commit_interval = HZ * arg;
1661 } else if (token == Opt_max_batch_time) {
1662 sbi->s_max_batch_time = arg;
1663 } else if (token == Opt_min_batch_time) {
1664 sbi->s_min_batch_time = arg;
1665 } else if (token == Opt_inode_readahead_blks) {
1666 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1667 ext4_msg(sb, KERN_ERR,
1668 "EXT4-fs: inode_readahead_blks must be "
1669 "0 or a power of 2 smaller than 2^31");
1672 sbi->s_inode_readahead_blks = arg;
1673 } else if (token == Opt_init_itable) {
1674 set_opt(sb, INIT_INODE_TABLE);
1676 arg = EXT4_DEF_LI_WAIT_MULT;
1677 sbi->s_li_wait_mult = arg;
1678 } else if (token == Opt_max_dir_size_kb) {
1679 sbi->s_max_dir_size_kb = arg;
1680 } else if (token == Opt_stripe) {
1681 sbi->s_stripe = arg;
1682 } else if (token == Opt_resuid) {
1683 uid = make_kuid(current_user_ns(), arg);
1684 if (!uid_valid(uid)) {
1685 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1688 sbi->s_resuid = uid;
1689 } else if (token == Opt_resgid) {
1690 gid = make_kgid(current_user_ns(), arg);
1691 if (!gid_valid(gid)) {
1692 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1695 sbi->s_resgid = gid;
1696 } else if (token == Opt_journal_dev) {
1698 ext4_msg(sb, KERN_ERR,
1699 "Cannot specify journal on remount");
1702 *journal_devnum = arg;
1703 } else if (token == Opt_journal_path) {
1705 struct inode *journal_inode;
1710 ext4_msg(sb, KERN_ERR,
1711 "Cannot specify journal on remount");
1714 journal_path = match_strdup(&args[0]);
1715 if (!journal_path) {
1716 ext4_msg(sb, KERN_ERR, "error: could not dup "
1717 "journal device string");
1721 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1723 ext4_msg(sb, KERN_ERR, "error: could not find "
1724 "journal device path: error %d", error);
1725 kfree(journal_path);
1729 journal_inode = d_inode(path.dentry);
1730 if (!S_ISBLK(journal_inode->i_mode)) {
1731 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1732 "is not a block device", journal_path);
1734 kfree(journal_path);
1738 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1740 kfree(journal_path);
1741 } else if (token == Opt_journal_ioprio) {
1743 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1748 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1749 } else if (token == Opt_test_dummy_encryption) {
1750 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1751 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1752 ext4_msg(sb, KERN_WARNING,
1753 "Test dummy encryption mode enabled");
1755 ext4_msg(sb, KERN_WARNING,
1756 "Test dummy encryption mount option ignored");
1758 } else if (m->flags & MOPT_DATAJ) {
1760 if (!sbi->s_journal)
1761 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1762 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1763 ext4_msg(sb, KERN_ERR,
1764 "Cannot change data mode on remount");
1768 clear_opt(sb, DATA_FLAGS);
1769 sbi->s_mount_opt |= m->mount_opt;
1772 } else if (m->flags & MOPT_QFMT) {
1773 if (sb_any_quota_loaded(sb) &&
1774 sbi->s_jquota_fmt != m->mount_opt) {
1775 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1776 "quota options when quota turned on");
1779 if (ext4_has_feature_quota(sb)) {
1780 ext4_msg(sb, KERN_INFO,
1781 "Quota format mount options ignored "
1782 "when QUOTA feature is enabled");
1785 sbi->s_jquota_fmt = m->mount_opt;
1787 } else if (token == Opt_dax) {
1788 #ifdef CONFIG_FS_DAX
1789 ext4_msg(sb, KERN_WARNING,
1790 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
1791 sbi->s_mount_opt |= m->mount_opt;
1793 ext4_msg(sb, KERN_INFO, "dax option not supported");
1796 } else if (token == Opt_data_err_abort) {
1797 sbi->s_mount_opt |= m->mount_opt;
1798 } else if (token == Opt_data_err_ignore) {
1799 sbi->s_mount_opt &= ~m->mount_opt;
1803 if (m->flags & MOPT_CLEAR)
1805 else if (unlikely(!(m->flags & MOPT_SET))) {
1806 ext4_msg(sb, KERN_WARNING,
1807 "buggy handling of option %s", opt);
1812 sbi->s_mount_opt |= m->mount_opt;
1814 sbi->s_mount_opt &= ~m->mount_opt;
1819 static int parse_options(char *options, struct super_block *sb,
1820 unsigned long *journal_devnum,
1821 unsigned int *journal_ioprio,
1824 struct ext4_sb_info *sbi = EXT4_SB(sb);
1826 substring_t args[MAX_OPT_ARGS];
1832 while ((p = strsep(&options, ",")) != NULL) {
1836 * Initialize args struct so we know whether arg was
1837 * found; some options take optional arguments.
1839 args[0].to = args[0].from = NULL;
1840 token = match_token(p, tokens, args);
1841 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1842 journal_ioprio, is_remount) < 0)
1847 * We do the test below only for project quotas. 'usrquota' and
1848 * 'grpquota' mount options are allowed even without quota feature
1849 * to support legacy quotas in quota files.
1851 if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) {
1852 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. "
1853 "Cannot enable project quota enforcement.");
1856 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1857 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1858 clear_opt(sb, USRQUOTA);
1860 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1861 clear_opt(sb, GRPQUOTA);
1863 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1864 ext4_msg(sb, KERN_ERR, "old and new quota "
1869 if (!sbi->s_jquota_fmt) {
1870 ext4_msg(sb, KERN_ERR, "journaled quota format "
1876 if (test_opt(sb, DIOREAD_NOLOCK)) {
1878 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1880 if (blocksize < PAGE_SIZE) {
1881 ext4_msg(sb, KERN_ERR, "can't mount with "
1882 "dioread_nolock if block size != PAGE_SIZE");
1886 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
1887 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
1888 ext4_msg(sb, KERN_ERR, "can't mount with journal_async_commit "
1889 "in data=ordered mode");
1895 static inline void ext4_show_quota_options(struct seq_file *seq,
1896 struct super_block *sb)
1898 #if defined(CONFIG_QUOTA)
1899 struct ext4_sb_info *sbi = EXT4_SB(sb);
1901 if (sbi->s_jquota_fmt) {
1904 switch (sbi->s_jquota_fmt) {
1915 seq_printf(seq, ",jqfmt=%s", fmtname);
1918 if (sbi->s_qf_names[USRQUOTA])
1919 seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]);
1921 if (sbi->s_qf_names[GRPQUOTA])
1922 seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]);
1926 static const char *token2str(int token)
1928 const struct match_token *t;
1930 for (t = tokens; t->token != Opt_err; t++)
1931 if (t->token == token && !strchr(t->pattern, '='))
1938 * - it's set to a non-default value OR
1939 * - if the per-sb default is different from the global default
1941 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1944 struct ext4_sb_info *sbi = EXT4_SB(sb);
1945 struct ext4_super_block *es = sbi->s_es;
1946 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1947 const struct mount_opts *m;
1948 char sep = nodefs ? '\n' : ',';
1950 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1951 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1953 if (sbi->s_sb_block != 1)
1954 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1956 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1957 int want_set = m->flags & MOPT_SET;
1958 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1959 (m->flags & MOPT_CLEAR_ERR))
1961 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
1962 continue; /* skip if same as the default */
1964 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1965 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1966 continue; /* select Opt_noFoo vs Opt_Foo */
1967 SEQ_OPTS_PRINT("%s", token2str(m->token));
1970 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
1971 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1972 SEQ_OPTS_PRINT("resuid=%u",
1973 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1974 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
1975 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1976 SEQ_OPTS_PRINT("resgid=%u",
1977 from_kgid_munged(&init_user_ns, sbi->s_resgid));
1978 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
1979 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1980 SEQ_OPTS_PUTS("errors=remount-ro");
1981 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1982 SEQ_OPTS_PUTS("errors=continue");
1983 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1984 SEQ_OPTS_PUTS("errors=panic");
1985 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
1986 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
1987 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
1988 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
1989 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
1990 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
1991 if (sb->s_flags & MS_I_VERSION)
1992 SEQ_OPTS_PUTS("i_version");
1993 if (nodefs || sbi->s_stripe)
1994 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
1995 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
1996 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1997 SEQ_OPTS_PUTS("data=journal");
1998 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1999 SEQ_OPTS_PUTS("data=ordered");
2000 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2001 SEQ_OPTS_PUTS("data=writeback");
2004 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
2005 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2006 sbi->s_inode_readahead_blks);
2008 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
2009 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
2010 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
2011 if (nodefs || sbi->s_max_dir_size_kb)
2012 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2013 if (test_opt(sb, DATA_ERR_ABORT))
2014 SEQ_OPTS_PUTS("data_err=abort");
2016 ext4_show_quota_options(seq, sb);
2020 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2022 return _ext4_show_options(seq, root->d_sb, 0);
2025 int ext4_seq_options_show(struct seq_file *seq, void *offset)
2027 struct super_block *sb = seq->private;
2030 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
2031 rc = _ext4_show_options(seq, sb, 1);
2032 seq_puts(seq, "\n");
2036 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
2039 struct ext4_sb_info *sbi = EXT4_SB(sb);
2042 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
2043 ext4_msg(sb, KERN_ERR, "revision level too high, "
2044 "forcing read-only mode");
2049 if (!(sbi->s_mount_state & EXT4_VALID_FS))
2050 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
2051 "running e2fsck is recommended");
2052 else if (sbi->s_mount_state & EXT4_ERROR_FS)
2053 ext4_msg(sb, KERN_WARNING,
2054 "warning: mounting fs with errors, "
2055 "running e2fsck is recommended");
2056 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
2057 le16_to_cpu(es->s_mnt_count) >=
2058 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
2059 ext4_msg(sb, KERN_WARNING,
2060 "warning: maximal mount count reached, "
2061 "running e2fsck is recommended");
2062 else if (le32_to_cpu(es->s_checkinterval) &&
2063 (le32_to_cpu(es->s_lastcheck) +
2064 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
2065 ext4_msg(sb, KERN_WARNING,
2066 "warning: checktime reached, "
2067 "running e2fsck is recommended");
2068 if (!sbi->s_journal)
2069 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
2070 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
2071 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
2072 le16_add_cpu(&es->s_mnt_count, 1);
2073 es->s_mtime = cpu_to_le32(get_seconds());
2074 ext4_update_dynamic_rev(sb);
2076 ext4_set_feature_journal_needs_recovery(sb);
2078 ext4_commit_super(sb, 1);
2080 if (test_opt(sb, DEBUG))
2081 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
2082 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
2084 sbi->s_groups_count,
2085 EXT4_BLOCKS_PER_GROUP(sb),
2086 EXT4_INODES_PER_GROUP(sb),
2087 sbi->s_mount_opt, sbi->s_mount_opt2);
2089 cleancache_init_fs(sb);
2093 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2095 struct ext4_sb_info *sbi = EXT4_SB(sb);
2096 struct flex_groups *new_groups;
2099 if (!sbi->s_log_groups_per_flex)
2102 size = ext4_flex_group(sbi, ngroup - 1) + 1;
2103 if (size <= sbi->s_flex_groups_allocated)
2106 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
2107 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
2109 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
2110 size / (int) sizeof(struct flex_groups));
2114 if (sbi->s_flex_groups) {
2115 memcpy(new_groups, sbi->s_flex_groups,
2116 (sbi->s_flex_groups_allocated *
2117 sizeof(struct flex_groups)));
2118 kvfree(sbi->s_flex_groups);
2120 sbi->s_flex_groups = new_groups;
2121 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
2125 static int ext4_fill_flex_info(struct super_block *sb)
2127 struct ext4_sb_info *sbi = EXT4_SB(sb);
2128 struct ext4_group_desc *gdp = NULL;
2129 ext4_group_t flex_group;
2132 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2133 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
2134 sbi->s_log_groups_per_flex = 0;
2138 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2142 for (i = 0; i < sbi->s_groups_count; i++) {
2143 gdp = ext4_get_group_desc(sb, i, NULL);
2145 flex_group = ext4_flex_group(sbi, i);
2146 atomic_add(ext4_free_inodes_count(sb, gdp),
2147 &sbi->s_flex_groups[flex_group].free_inodes);
2148 atomic64_add(ext4_free_group_clusters(sb, gdp),
2149 &sbi->s_flex_groups[flex_group].free_clusters);
2150 atomic_add(ext4_used_dirs_count(sb, gdp),
2151 &sbi->s_flex_groups[flex_group].used_dirs);
2159 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
2160 struct ext4_group_desc *gdp)
2162 int offset = offsetof(struct ext4_group_desc, bg_checksum);
2164 __le32 le_group = cpu_to_le32(block_group);
2165 struct ext4_sb_info *sbi = EXT4_SB(sb);
2167 if (ext4_has_metadata_csum(sbi->s_sb)) {
2168 /* Use new metadata_csum algorithm */
2170 __u16 dummy_csum = 0;
2172 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2174 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2175 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2176 sizeof(dummy_csum));
2177 offset += sizeof(dummy_csum);
2178 if (offset < sbi->s_desc_size)
2179 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2180 sbi->s_desc_size - offset);
2182 crc = csum32 & 0xFFFF;
2186 /* old crc16 code */
2187 if (!ext4_has_feature_gdt_csum(sb))
2190 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2191 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2192 crc = crc16(crc, (__u8 *)gdp, offset);
2193 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2194 /* for checksum of struct ext4_group_desc do the rest...*/
2195 if (ext4_has_feature_64bit(sb) &&
2196 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2197 crc = crc16(crc, (__u8 *)gdp + offset,
2198 le16_to_cpu(sbi->s_es->s_desc_size) -
2202 return cpu_to_le16(crc);
2205 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2206 struct ext4_group_desc *gdp)
2208 if (ext4_has_group_desc_csum(sb) &&
2209 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
2215 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2216 struct ext4_group_desc *gdp)
2218 if (!ext4_has_group_desc_csum(sb))
2220 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
2223 /* Called at mount-time, super-block is locked */
2224 static int ext4_check_descriptors(struct super_block *sb,
2225 ext4_fsblk_t sb_block,
2226 ext4_group_t *first_not_zeroed)
2228 struct ext4_sb_info *sbi = EXT4_SB(sb);
2229 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2230 ext4_fsblk_t last_block;
2231 ext4_fsblk_t block_bitmap;
2232 ext4_fsblk_t inode_bitmap;
2233 ext4_fsblk_t inode_table;
2234 int flexbg_flag = 0;
2235 ext4_group_t i, grp = sbi->s_groups_count;
2237 if (ext4_has_feature_flex_bg(sb))
2240 ext4_debug("Checking group descriptors");
2242 for (i = 0; i < sbi->s_groups_count; i++) {
2243 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2245 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2246 last_block = ext4_blocks_count(sbi->s_es) - 1;
2248 last_block = first_block +
2249 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2251 if ((grp == sbi->s_groups_count) &&
2252 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2255 block_bitmap = ext4_block_bitmap(sb, gdp);
2256 if (block_bitmap == sb_block) {
2257 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2258 "Block bitmap for group %u overlaps "
2261 if (block_bitmap < first_block || block_bitmap > last_block) {
2262 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2263 "Block bitmap for group %u not in group "
2264 "(block %llu)!", i, block_bitmap);
2267 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2268 if (inode_bitmap == sb_block) {
2269 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2270 "Inode bitmap for group %u overlaps "
2273 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2274 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2275 "Inode bitmap for group %u not in group "
2276 "(block %llu)!", i, inode_bitmap);
2279 inode_table = ext4_inode_table(sb, gdp);
2280 if (inode_table == sb_block) {
2281 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2282 "Inode table for group %u overlaps "
2285 if (inode_table < first_block ||
2286 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2287 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2288 "Inode table for group %u not in group "
2289 "(block %llu)!", i, inode_table);
2292 ext4_lock_group(sb, i);
2293 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2294 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2295 "Checksum for group %u failed (%u!=%u)",
2296 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
2297 gdp)), le16_to_cpu(gdp->bg_checksum));
2298 if (!(sb->s_flags & MS_RDONLY)) {
2299 ext4_unlock_group(sb, i);
2303 ext4_unlock_group(sb, i);
2305 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2307 if (NULL != first_not_zeroed)
2308 *first_not_zeroed = grp;
2312 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2313 * the superblock) which were deleted from all directories, but held open by
2314 * a process at the time of a crash. We walk the list and try to delete these
2315 * inodes at recovery time (only with a read-write filesystem).
2317 * In order to keep the orphan inode chain consistent during traversal (in
2318 * case of crash during recovery), we link each inode into the superblock
2319 * orphan list_head and handle it the same way as an inode deletion during
2320 * normal operation (which journals the operations for us).
2322 * We only do an iget() and an iput() on each inode, which is very safe if we
2323 * accidentally point at an in-use or already deleted inode. The worst that
2324 * can happen in this case is that we get a "bit already cleared" message from
2325 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2326 * e2fsck was run on this filesystem, and it must have already done the orphan
2327 * inode cleanup for us, so we can safely abort without any further action.
2329 static void ext4_orphan_cleanup(struct super_block *sb,
2330 struct ext4_super_block *es)
2332 unsigned int s_flags = sb->s_flags;
2333 int ret, nr_orphans = 0, nr_truncates = 0;
2337 if (!es->s_last_orphan) {
2338 jbd_debug(4, "no orphan inodes to clean up\n");
2342 if (bdev_read_only(sb->s_bdev)) {
2343 ext4_msg(sb, KERN_ERR, "write access "
2344 "unavailable, skipping orphan cleanup");
2348 /* Check if feature set would not allow a r/w mount */
2349 if (!ext4_feature_set_ok(sb, 0)) {
2350 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2351 "unknown ROCOMPAT features");
2355 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2356 /* don't clear list on RO mount w/ errors */
2357 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
2358 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
2359 "clearing orphan list.\n");
2360 es->s_last_orphan = 0;
2362 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2366 if (s_flags & MS_RDONLY) {
2367 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2368 sb->s_flags &= ~MS_RDONLY;
2371 /* Needed for iput() to work correctly and not trash data */
2372 sb->s_flags |= MS_ACTIVE;
2373 /* Turn on quotas so that they are updated correctly */
2374 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2375 if (EXT4_SB(sb)->s_qf_names[i]) {
2376 int ret = ext4_quota_on_mount(sb, i);
2378 ext4_msg(sb, KERN_ERR,
2379 "Cannot turn on journaled "
2380 "quota: error %d", ret);
2385 while (es->s_last_orphan) {
2386 struct inode *inode;
2389 * We may have encountered an error during cleanup; if
2390 * so, skip the rest.
2392 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2393 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2394 es->s_last_orphan = 0;
2398 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2399 if (IS_ERR(inode)) {
2400 es->s_last_orphan = 0;
2404 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2405 dquot_initialize(inode);
2406 if (inode->i_nlink) {
2407 if (test_opt(sb, DEBUG))
2408 ext4_msg(sb, KERN_DEBUG,
2409 "%s: truncating inode %lu to %lld bytes",
2410 __func__, inode->i_ino, inode->i_size);
2411 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2412 inode->i_ino, inode->i_size);
2414 truncate_inode_pages(inode->i_mapping, inode->i_size);
2415 ret = ext4_truncate(inode);
2417 ext4_std_error(inode->i_sb, ret);
2418 inode_unlock(inode);
2421 if (test_opt(sb, DEBUG))
2422 ext4_msg(sb, KERN_DEBUG,
2423 "%s: deleting unreferenced inode %lu",
2424 __func__, inode->i_ino);
2425 jbd_debug(2, "deleting unreferenced inode %lu\n",
2429 iput(inode); /* The delete magic happens here! */
2432 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2435 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2436 PLURAL(nr_orphans));
2438 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2439 PLURAL(nr_truncates));
2441 /* Turn quotas off */
2442 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2443 if (sb_dqopt(sb)->files[i])
2444 dquot_quota_off(sb, i);
2447 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2451 * Maximal extent format file size.
2452 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2453 * extent format containers, within a sector_t, and within i_blocks
2454 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2455 * so that won't be a limiting factor.
2457 * However there is other limiting factor. We do store extents in the form
2458 * of starting block and length, hence the resulting length of the extent
2459 * covering maximum file size must fit into on-disk format containers as
2460 * well. Given that length is always by 1 unit bigger than max unit (because
2461 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2463 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2465 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2468 loff_t upper_limit = MAX_LFS_FILESIZE;
2470 /* small i_blocks in vfs inode? */
2471 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2473 * CONFIG_LBDAF is not enabled implies the inode
2474 * i_block represent total blocks in 512 bytes
2475 * 32 == size of vfs inode i_blocks * 8
2477 upper_limit = (1LL << 32) - 1;
2479 /* total blocks in file system block size */
2480 upper_limit >>= (blkbits - 9);
2481 upper_limit <<= blkbits;
2485 * 32-bit extent-start container, ee_block. We lower the maxbytes
2486 * by one fs block, so ee_len can cover the extent of maximum file
2489 res = (1LL << 32) - 1;
2492 /* Sanity check against vm- & vfs- imposed limits */
2493 if (res > upper_limit)
2500 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2501 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2502 * We need to be 1 filesystem block less than the 2^48 sector limit.
2504 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2506 loff_t res = EXT4_NDIR_BLOCKS;
2509 /* This is calculated to be the largest file size for a dense, block
2510 * mapped file such that the file's total number of 512-byte sectors,
2511 * including data and all indirect blocks, does not exceed (2^48 - 1).
2513 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2514 * number of 512-byte sectors of the file.
2517 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2519 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2520 * the inode i_block field represents total file blocks in
2521 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2523 upper_limit = (1LL << 32) - 1;
2525 /* total blocks in file system block size */
2526 upper_limit >>= (bits - 9);
2530 * We use 48 bit ext4_inode i_blocks
2531 * With EXT4_HUGE_FILE_FL set the i_blocks
2532 * represent total number of blocks in
2533 * file system block size
2535 upper_limit = (1LL << 48) - 1;
2539 /* indirect blocks */
2541 /* double indirect blocks */
2542 meta_blocks += 1 + (1LL << (bits-2));
2543 /* tripple indirect blocks */
2544 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2546 upper_limit -= meta_blocks;
2547 upper_limit <<= bits;
2549 res += 1LL << (bits-2);
2550 res += 1LL << (2*(bits-2));
2551 res += 1LL << (3*(bits-2));
2553 if (res > upper_limit)
2556 if (res > MAX_LFS_FILESIZE)
2557 res = MAX_LFS_FILESIZE;
2562 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2563 ext4_fsblk_t logical_sb_block, int nr)
2565 struct ext4_sb_info *sbi = EXT4_SB(sb);
2566 ext4_group_t bg, first_meta_bg;
2569 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2571 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
2572 return logical_sb_block + nr + 1;
2573 bg = sbi->s_desc_per_block * nr;
2574 if (ext4_bg_has_super(sb, bg))
2578 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2579 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2580 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2583 if (sb->s_blocksize == 1024 && nr == 0 &&
2584 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2587 return (has_super + ext4_group_first_block_no(sb, bg));
2591 * ext4_get_stripe_size: Get the stripe size.
2592 * @sbi: In memory super block info
2594 * If we have specified it via mount option, then
2595 * use the mount option value. If the value specified at mount time is
2596 * greater than the blocks per group use the super block value.
2597 * If the super block value is greater than blocks per group return 0.
2598 * Allocator needs it be less than blocks per group.
2601 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2603 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2604 unsigned long stripe_width =
2605 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2608 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2609 ret = sbi->s_stripe;
2610 else if (stripe_width <= sbi->s_blocks_per_group)
2612 else if (stride <= sbi->s_blocks_per_group)
2618 * If the stripe width is 1, this makes no sense and
2619 * we set it to 0 to turn off stripe handling code.
2628 * Check whether this filesystem can be mounted based on
2629 * the features present and the RDONLY/RDWR mount requested.
2630 * Returns 1 if this filesystem can be mounted as requested,
2631 * 0 if it cannot be.
2633 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2635 if (ext4_has_unknown_ext4_incompat_features(sb)) {
2636 ext4_msg(sb, KERN_ERR,
2637 "Couldn't mount because of "
2638 "unsupported optional features (%x)",
2639 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2640 ~EXT4_FEATURE_INCOMPAT_SUPP));
2647 if (ext4_has_feature_readonly(sb)) {
2648 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2649 sb->s_flags |= MS_RDONLY;
2653 /* Check that feature set is OK for a read-write mount */
2654 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
2655 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2656 "unsupported optional features (%x)",
2657 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2658 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2662 * Large file size enabled file system can only be mounted
2663 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2665 if (ext4_has_feature_huge_file(sb)) {
2666 if (sizeof(blkcnt_t) < sizeof(u64)) {
2667 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2668 "cannot be mounted RDWR without "
2673 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
2674 ext4_msg(sb, KERN_ERR,
2675 "Can't support bigalloc feature without "
2676 "extents feature\n");
2680 #ifndef CONFIG_QUOTA
2681 if (ext4_has_feature_quota(sb) && !readonly) {
2682 ext4_msg(sb, KERN_ERR,
2683 "Filesystem with quota feature cannot be mounted RDWR "
2684 "without CONFIG_QUOTA");
2687 if (ext4_has_feature_project(sb) && !readonly) {
2688 ext4_msg(sb, KERN_ERR,
2689 "Filesystem with project quota feature cannot be mounted RDWR "
2690 "without CONFIG_QUOTA");
2693 #endif /* CONFIG_QUOTA */
2698 * This function is called once a day if we have errors logged
2699 * on the file system
2701 static void print_daily_error_info(unsigned long arg)
2703 struct super_block *sb = (struct super_block *) arg;
2704 struct ext4_sb_info *sbi;
2705 struct ext4_super_block *es;
2710 if (es->s_error_count)
2711 /* fsck newer than v1.41.13 is needed to clean this condition. */
2712 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
2713 le32_to_cpu(es->s_error_count));
2714 if (es->s_first_error_time) {
2715 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
2716 sb->s_id, le32_to_cpu(es->s_first_error_time),
2717 (int) sizeof(es->s_first_error_func),
2718 es->s_first_error_func,
2719 le32_to_cpu(es->s_first_error_line));
2720 if (es->s_first_error_ino)
2721 printk(KERN_CONT ": inode %u",
2722 le32_to_cpu(es->s_first_error_ino));
2723 if (es->s_first_error_block)
2724 printk(KERN_CONT ": block %llu", (unsigned long long)
2725 le64_to_cpu(es->s_first_error_block));
2726 printk(KERN_CONT "\n");
2728 if (es->s_last_error_time) {
2729 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
2730 sb->s_id, le32_to_cpu(es->s_last_error_time),
2731 (int) sizeof(es->s_last_error_func),
2732 es->s_last_error_func,
2733 le32_to_cpu(es->s_last_error_line));
2734 if (es->s_last_error_ino)
2735 printk(KERN_CONT ": inode %u",
2736 le32_to_cpu(es->s_last_error_ino));
2737 if (es->s_last_error_block)
2738 printk(KERN_CONT ": block %llu", (unsigned long long)
2739 le64_to_cpu(es->s_last_error_block));
2740 printk(KERN_CONT "\n");
2742 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2745 /* Find next suitable group and run ext4_init_inode_table */
2746 static int ext4_run_li_request(struct ext4_li_request *elr)
2748 struct ext4_group_desc *gdp = NULL;
2749 ext4_group_t group, ngroups;
2750 struct super_block *sb;
2751 unsigned long timeout = 0;
2755 ngroups = EXT4_SB(sb)->s_groups_count;
2757 for (group = elr->lr_next_group; group < ngroups; group++) {
2758 gdp = ext4_get_group_desc(sb, group, NULL);
2764 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2768 if (group >= ngroups)
2773 ret = ext4_init_inode_table(sb, group,
2774 elr->lr_timeout ? 0 : 1);
2775 if (elr->lr_timeout == 0) {
2776 timeout = (jiffies - timeout) *
2777 elr->lr_sbi->s_li_wait_mult;
2778 elr->lr_timeout = timeout;
2780 elr->lr_next_sched = jiffies + elr->lr_timeout;
2781 elr->lr_next_group = group + 1;
2787 * Remove lr_request from the list_request and free the
2788 * request structure. Should be called with li_list_mtx held
2790 static void ext4_remove_li_request(struct ext4_li_request *elr)
2792 struct ext4_sb_info *sbi;
2799 list_del(&elr->lr_request);
2800 sbi->s_li_request = NULL;
2804 static void ext4_unregister_li_request(struct super_block *sb)
2806 mutex_lock(&ext4_li_mtx);
2807 if (!ext4_li_info) {
2808 mutex_unlock(&ext4_li_mtx);
2812 mutex_lock(&ext4_li_info->li_list_mtx);
2813 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2814 mutex_unlock(&ext4_li_info->li_list_mtx);
2815 mutex_unlock(&ext4_li_mtx);
2818 static struct task_struct *ext4_lazyinit_task;
2821 * This is the function where ext4lazyinit thread lives. It walks
2822 * through the request list searching for next scheduled filesystem.
2823 * When such a fs is found, run the lazy initialization request
2824 * (ext4_rn_li_request) and keep track of the time spend in this
2825 * function. Based on that time we compute next schedule time of
2826 * the request. When walking through the list is complete, compute
2827 * next waking time and put itself into sleep.
2829 static int ext4_lazyinit_thread(void *arg)
2831 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2832 struct list_head *pos, *n;
2833 struct ext4_li_request *elr;
2834 unsigned long next_wakeup, cur;
2836 BUG_ON(NULL == eli);
2840 next_wakeup = MAX_JIFFY_OFFSET;
2842 mutex_lock(&eli->li_list_mtx);
2843 if (list_empty(&eli->li_request_list)) {
2844 mutex_unlock(&eli->li_list_mtx);
2847 list_for_each_safe(pos, n, &eli->li_request_list) {
2850 elr = list_entry(pos, struct ext4_li_request,
2853 if (time_before(jiffies, elr->lr_next_sched)) {
2854 if (time_before(elr->lr_next_sched, next_wakeup))
2855 next_wakeup = elr->lr_next_sched;
2858 if (down_read_trylock(&elr->lr_super->s_umount)) {
2859 if (sb_start_write_trylock(elr->lr_super)) {
2862 * We hold sb->s_umount, sb can not
2863 * be removed from the list, it is
2864 * now safe to drop li_list_mtx
2866 mutex_unlock(&eli->li_list_mtx);
2867 err = ext4_run_li_request(elr);
2868 sb_end_write(elr->lr_super);
2869 mutex_lock(&eli->li_list_mtx);
2872 up_read((&elr->lr_super->s_umount));
2874 /* error, remove the lazy_init job */
2876 ext4_remove_li_request(elr);
2880 elr->lr_next_sched = jiffies +
2882 % (EXT4_DEF_LI_MAX_START_DELAY * HZ));
2884 if (time_before(elr->lr_next_sched, next_wakeup))
2885 next_wakeup = elr->lr_next_sched;
2887 mutex_unlock(&eli->li_list_mtx);
2892 if ((time_after_eq(cur, next_wakeup)) ||
2893 (MAX_JIFFY_OFFSET == next_wakeup)) {
2898 schedule_timeout_interruptible(next_wakeup - cur);
2900 if (kthread_should_stop()) {
2901 ext4_clear_request_list();
2908 * It looks like the request list is empty, but we need
2909 * to check it under the li_list_mtx lock, to prevent any
2910 * additions into it, and of course we should lock ext4_li_mtx
2911 * to atomically free the list and ext4_li_info, because at
2912 * this point another ext4 filesystem could be registering
2915 mutex_lock(&ext4_li_mtx);
2916 mutex_lock(&eli->li_list_mtx);
2917 if (!list_empty(&eli->li_request_list)) {
2918 mutex_unlock(&eli->li_list_mtx);
2919 mutex_unlock(&ext4_li_mtx);
2922 mutex_unlock(&eli->li_list_mtx);
2923 kfree(ext4_li_info);
2924 ext4_li_info = NULL;
2925 mutex_unlock(&ext4_li_mtx);
2930 static void ext4_clear_request_list(void)
2932 struct list_head *pos, *n;
2933 struct ext4_li_request *elr;
2935 mutex_lock(&ext4_li_info->li_list_mtx);
2936 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2937 elr = list_entry(pos, struct ext4_li_request,
2939 ext4_remove_li_request(elr);
2941 mutex_unlock(&ext4_li_info->li_list_mtx);
2944 static int ext4_run_lazyinit_thread(void)
2946 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2947 ext4_li_info, "ext4lazyinit");
2948 if (IS_ERR(ext4_lazyinit_task)) {
2949 int err = PTR_ERR(ext4_lazyinit_task);
2950 ext4_clear_request_list();
2951 kfree(ext4_li_info);
2952 ext4_li_info = NULL;
2953 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
2954 "initialization thread\n",
2958 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2963 * Check whether it make sense to run itable init. thread or not.
2964 * If there is at least one uninitialized inode table, return
2965 * corresponding group number, else the loop goes through all
2966 * groups and return total number of groups.
2968 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2970 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2971 struct ext4_group_desc *gdp = NULL;
2973 for (group = 0; group < ngroups; group++) {
2974 gdp = ext4_get_group_desc(sb, group, NULL);
2978 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2985 static int ext4_li_info_new(void)
2987 struct ext4_lazy_init *eli = NULL;
2989 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2993 INIT_LIST_HEAD(&eli->li_request_list);
2994 mutex_init(&eli->li_list_mtx);
2996 eli->li_state |= EXT4_LAZYINIT_QUIT;
3003 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3006 struct ext4_sb_info *sbi = EXT4_SB(sb);
3007 struct ext4_li_request *elr;
3009 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3015 elr->lr_next_group = start;
3018 * Randomize first schedule time of the request to
3019 * spread the inode table initialization requests
3022 elr->lr_next_sched = jiffies + (prandom_u32() %
3023 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3027 int ext4_register_li_request(struct super_block *sb,
3028 ext4_group_t first_not_zeroed)
3030 struct ext4_sb_info *sbi = EXT4_SB(sb);
3031 struct ext4_li_request *elr = NULL;
3032 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3035 mutex_lock(&ext4_li_mtx);
3036 if (sbi->s_li_request != NULL) {
3038 * Reset timeout so it can be computed again, because
3039 * s_li_wait_mult might have changed.
3041 sbi->s_li_request->lr_timeout = 0;
3045 if (first_not_zeroed == ngroups ||
3046 (sb->s_flags & MS_RDONLY) ||
3047 !test_opt(sb, INIT_INODE_TABLE))
3050 elr = ext4_li_request_new(sb, first_not_zeroed);
3056 if (NULL == ext4_li_info) {
3057 ret = ext4_li_info_new();
3062 mutex_lock(&ext4_li_info->li_list_mtx);
3063 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3064 mutex_unlock(&ext4_li_info->li_list_mtx);
3066 sbi->s_li_request = elr;
3068 * set elr to NULL here since it has been inserted to
3069 * the request_list and the removal and free of it is
3070 * handled by ext4_clear_request_list from now on.
3074 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3075 ret = ext4_run_lazyinit_thread();
3080 mutex_unlock(&ext4_li_mtx);
3087 * We do not need to lock anything since this is called on
3090 static void ext4_destroy_lazyinit_thread(void)
3093 * If thread exited earlier
3094 * there's nothing to be done.
3096 if (!ext4_li_info || !ext4_lazyinit_task)
3099 kthread_stop(ext4_lazyinit_task);
3102 static int set_journal_csum_feature_set(struct super_block *sb)
3105 int compat, incompat;
3106 struct ext4_sb_info *sbi = EXT4_SB(sb);
3108 if (ext4_has_metadata_csum(sb)) {
3109 /* journal checksum v3 */
3111 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
3113 /* journal checksum v1 */
3114 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3118 jbd2_journal_clear_features(sbi->s_journal,
3119 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3120 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3121 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3122 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3123 ret = jbd2_journal_set_features(sbi->s_journal,
3125 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3127 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3128 ret = jbd2_journal_set_features(sbi->s_journal,
3131 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3132 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3134 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3135 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3142 * Note: calculating the overhead so we can be compatible with
3143 * historical BSD practice is quite difficult in the face of
3144 * clusters/bigalloc. This is because multiple metadata blocks from
3145 * different block group can end up in the same allocation cluster.
3146 * Calculating the exact overhead in the face of clustered allocation
3147 * requires either O(all block bitmaps) in memory or O(number of block
3148 * groups**2) in time. We will still calculate the superblock for
3149 * older file systems --- and if we come across with a bigalloc file
3150 * system with zero in s_overhead_clusters the estimate will be close to
3151 * correct especially for very large cluster sizes --- but for newer
3152 * file systems, it's better to calculate this figure once at mkfs
3153 * time, and store it in the superblock. If the superblock value is
3154 * present (even for non-bigalloc file systems), we will use it.
3156 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3159 struct ext4_sb_info *sbi = EXT4_SB(sb);
3160 struct ext4_group_desc *gdp;
3161 ext4_fsblk_t first_block, last_block, b;
3162 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3163 int s, j, count = 0;
3165 if (!ext4_has_feature_bigalloc(sb))
3166 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3167 sbi->s_itb_per_group + 2);
3169 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3170 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3171 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3172 for (i = 0; i < ngroups; i++) {
3173 gdp = ext4_get_group_desc(sb, i, NULL);
3174 b = ext4_block_bitmap(sb, gdp);
3175 if (b >= first_block && b <= last_block) {
3176 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3179 b = ext4_inode_bitmap(sb, gdp);
3180 if (b >= first_block && b <= last_block) {
3181 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3184 b = ext4_inode_table(sb, gdp);
3185 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3186 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3187 int c = EXT4_B2C(sbi, b - first_block);
3188 ext4_set_bit(c, buf);
3194 if (ext4_bg_has_super(sb, grp)) {
3195 ext4_set_bit(s++, buf);
3198 j = ext4_bg_num_gdb(sb, grp);
3199 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
3200 ext4_error(sb, "Invalid number of block group "
3201 "descriptor blocks: %d", j);
3202 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
3206 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3210 return EXT4_CLUSTERS_PER_GROUP(sb) -
3211 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3215 * Compute the overhead and stash it in sbi->s_overhead
3217 int ext4_calculate_overhead(struct super_block *sb)
3219 struct ext4_sb_info *sbi = EXT4_SB(sb);
3220 struct ext4_super_block *es = sbi->s_es;
3221 struct inode *j_inode;
3222 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
3223 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3224 ext4_fsblk_t overhead = 0;
3225 char *buf = (char *) get_zeroed_page(GFP_NOFS);
3231 * Compute the overhead (FS structures). This is constant
3232 * for a given filesystem unless the number of block groups
3233 * changes so we cache the previous value until it does.
3237 * All of the blocks before first_data_block are overhead
3239 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3242 * Add the overhead found in each block group
3244 for (i = 0; i < ngroups; i++) {
3247 blks = count_overhead(sb, i, buf);
3250 memset(buf, 0, PAGE_SIZE);
3255 * Add the internal journal blocks whether the journal has been
3258 if (sbi->s_journal && !sbi->journal_bdev)
3259 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3260 else if (ext4_has_feature_journal(sb) && !sbi->s_journal) {
3261 j_inode = ext4_get_journal_inode(sb, j_inum);
3263 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
3264 overhead += EXT4_NUM_B2C(sbi, j_blocks);
3267 ext4_msg(sb, KERN_ERR, "can't get journal size");
3270 sbi->s_overhead = overhead;
3272 free_page((unsigned long) buf);
3276 static void ext4_set_resv_clusters(struct super_block *sb)
3278 ext4_fsblk_t resv_clusters;
3279 struct ext4_sb_info *sbi = EXT4_SB(sb);
3282 * There's no need to reserve anything when we aren't using extents.
3283 * The space estimates are exact, there are no unwritten extents,
3284 * hole punching doesn't need new metadata... This is needed especially
3285 * to keep ext2/3 backward compatibility.
3287 if (!ext4_has_feature_extents(sb))
3290 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3291 * This should cover the situations where we can not afford to run
3292 * out of space like for example punch hole, or converting
3293 * unwritten extents in delalloc path. In most cases such
3294 * allocation would require 1, or 2 blocks, higher numbers are
3297 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3298 sbi->s_cluster_bits);
3300 do_div(resv_clusters, 50);
3301 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3303 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
3306 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3308 char *orig_data = kstrdup(data, GFP_KERNEL);
3309 struct buffer_head *bh;
3310 struct ext4_super_block *es = NULL;
3311 struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3313 ext4_fsblk_t sb_block = get_sb_block(&data);
3314 ext4_fsblk_t logical_sb_block;
3315 unsigned long offset = 0;
3316 unsigned long journal_devnum = 0;
3317 unsigned long def_mount_opts;
3321 int blocksize, clustersize;
3322 unsigned int db_count;
3324 int needs_recovery, has_huge_files, has_bigalloc;
3327 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3328 ext4_group_t first_not_zeroed;
3330 if ((data && !orig_data) || !sbi)
3333 sbi->s_blockgroup_lock =
3334 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3335 if (!sbi->s_blockgroup_lock)
3338 sb->s_fs_info = sbi;
3340 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3341 sbi->s_sb_block = sb_block;
3342 if (sb->s_bdev->bd_part)
3343 sbi->s_sectors_written_start =
3344 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3346 /* Cleanup superblock name */
3347 strreplace(sb->s_id, '/', '!');
3349 /* -EINVAL is default */
3351 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3353 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3358 * The ext4 superblock will not be buffer aligned for other than 1kB
3359 * block sizes. We need to calculate the offset from buffer start.
3361 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3362 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3363 offset = do_div(logical_sb_block, blocksize);
3365 logical_sb_block = sb_block;
3368 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
3369 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3373 * Note: s_es must be initialized as soon as possible because
3374 * some ext4 macro-instructions depend on its value
3376 es = (struct ext4_super_block *) (bh->b_data + offset);
3378 sb->s_magic = le16_to_cpu(es->s_magic);
3379 if (sb->s_magic != EXT4_SUPER_MAGIC)
3381 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3383 /* Warn if metadata_csum and gdt_csum are both set. */
3384 if (ext4_has_feature_metadata_csum(sb) &&
3385 ext4_has_feature_gdt_csum(sb))
3386 ext4_warning(sb, "metadata_csum and uninit_bg are "
3387 "redundant flags; please run fsck.");
3389 /* Check for a known checksum algorithm */
3390 if (!ext4_verify_csum_type(sb, es)) {
3391 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3392 "unknown checksum algorithm.");
3397 /* Load the checksum driver */
3398 if (ext4_has_feature_metadata_csum(sb)) {
3399 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3400 if (IS_ERR(sbi->s_chksum_driver)) {
3401 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3402 ret = PTR_ERR(sbi->s_chksum_driver);
3403 sbi->s_chksum_driver = NULL;
3408 /* Check superblock checksum */
3409 if (!ext4_superblock_csum_verify(sb, es)) {
3410 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3411 "invalid superblock checksum. Run e2fsck?");
3417 /* Precompute checksum seed for all metadata */
3418 if (ext4_has_feature_csum_seed(sb))
3419 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
3420 else if (ext4_has_metadata_csum(sb))
3421 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3422 sizeof(es->s_uuid));
3424 /* Set defaults before we parse the mount options */
3425 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3426 set_opt(sb, INIT_INODE_TABLE);
3427 if (def_mount_opts & EXT4_DEFM_DEBUG)
3429 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3431 if (def_mount_opts & EXT4_DEFM_UID16)
3432 set_opt(sb, NO_UID32);
3433 /* xattr user namespace & acls are now defaulted on */
3434 set_opt(sb, XATTR_USER);
3435 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3436 set_opt(sb, POSIX_ACL);
3438 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3439 if (ext4_has_metadata_csum(sb))
3440 set_opt(sb, JOURNAL_CHECKSUM);
3442 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3443 set_opt(sb, JOURNAL_DATA);
3444 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3445 set_opt(sb, ORDERED_DATA);
3446 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3447 set_opt(sb, WRITEBACK_DATA);
3449 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3450 set_opt(sb, ERRORS_PANIC);
3451 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3452 set_opt(sb, ERRORS_CONT);
3454 set_opt(sb, ERRORS_RO);
3455 /* block_validity enabled by default; disable with noblock_validity */
3456 set_opt(sb, BLOCK_VALIDITY);
3457 if (def_mount_opts & EXT4_DEFM_DISCARD)
3458 set_opt(sb, DISCARD);
3460 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3461 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3462 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3463 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3464 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3466 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3467 set_opt(sb, BARRIER);
3470 * enable delayed allocation by default
3471 * Use -o nodelalloc to turn it off
3473 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3474 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3475 set_opt(sb, DELALLOC);
3478 * set default s_li_wait_mult for lazyinit, for the case there is
3479 * no mount option specified.
3481 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3483 if (sbi->s_es->s_mount_opts[0]) {
3484 char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
3485 sizeof(sbi->s_es->s_mount_opts),
3489 if (!parse_options(s_mount_opts, sb, &journal_devnum,
3490 &journal_ioprio, 0)) {
3491 ext4_msg(sb, KERN_WARNING,
3492 "failed to parse options in superblock: %s",
3495 kfree(s_mount_opts);
3497 sbi->s_def_mount_opt = sbi->s_mount_opt;
3498 if (!parse_options((char *) data, sb, &journal_devnum,
3499 &journal_ioprio, 0))
3502 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3503 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3504 "with data=journal disables delayed "
3505 "allocation and O_DIRECT support!\n");
3506 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3507 ext4_msg(sb, KERN_ERR, "can't mount with "
3508 "both data=journal and delalloc");
3511 if (test_opt(sb, DIOREAD_NOLOCK)) {
3512 ext4_msg(sb, KERN_ERR, "can't mount with "
3513 "both data=journal and dioread_nolock");
3516 if (test_opt(sb, DAX)) {
3517 ext4_msg(sb, KERN_ERR, "can't mount with "
3518 "both data=journal and dax");
3521 if (test_opt(sb, DELALLOC))
3522 clear_opt(sb, DELALLOC);
3524 sb->s_iflags |= SB_I_CGROUPWB;
3527 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3528 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3530 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3531 (ext4_has_compat_features(sb) ||
3532 ext4_has_ro_compat_features(sb) ||
3533 ext4_has_incompat_features(sb)))
3534 ext4_msg(sb, KERN_WARNING,
3535 "feature flags set on rev 0 fs, "
3536 "running e2fsck is recommended");
3538 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3539 set_opt2(sb, HURD_COMPAT);
3540 if (ext4_has_feature_64bit(sb)) {
3541 ext4_msg(sb, KERN_ERR,
3542 "The Hurd can't support 64-bit file systems");
3547 if (IS_EXT2_SB(sb)) {
3548 if (ext2_feature_set_ok(sb))
3549 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3550 "using the ext4 subsystem");
3552 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3553 "to feature incompatibilities");
3558 if (IS_EXT3_SB(sb)) {
3559 if (ext3_feature_set_ok(sb))
3560 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3561 "using the ext4 subsystem");
3563 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3564 "to feature incompatibilities");
3570 * Check feature flags regardless of the revision level, since we
3571 * previously didn't change the revision level when setting the flags,
3572 * so there is a chance incompat flags are set on a rev 0 filesystem.
3574 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3577 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3578 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3579 blocksize > EXT4_MAX_BLOCK_SIZE) {
3580 ext4_msg(sb, KERN_ERR,
3581 "Unsupported filesystem blocksize %d (%d log_block_size)",
3582 blocksize, le32_to_cpu(es->s_log_block_size));
3585 if (le32_to_cpu(es->s_log_block_size) >
3586 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3587 ext4_msg(sb, KERN_ERR,
3588 "Invalid log block size: %u",
3589 le32_to_cpu(es->s_log_block_size));
3593 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
3594 ext4_msg(sb, KERN_ERR,
3595 "Number of reserved GDT blocks insanely large: %d",
3596 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
3600 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
3601 err = bdev_dax_supported(sb, blocksize);
3606 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
3607 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
3608 es->s_encryption_level);
3612 if (sb->s_blocksize != blocksize) {
3613 /* Validate the filesystem blocksize */
3614 if (!sb_set_blocksize(sb, blocksize)) {
3615 ext4_msg(sb, KERN_ERR, "bad block size %d",
3621 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3622 offset = do_div(logical_sb_block, blocksize);
3623 bh = sb_bread_unmovable(sb, logical_sb_block);
3625 ext4_msg(sb, KERN_ERR,
3626 "Can't read superblock on 2nd try");
3629 es = (struct ext4_super_block *)(bh->b_data + offset);
3631 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3632 ext4_msg(sb, KERN_ERR,
3633 "Magic mismatch, very weird!");
3638 has_huge_files = ext4_has_feature_huge_file(sb);
3639 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3641 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3643 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3644 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3645 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3647 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3648 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3649 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3650 (!is_power_of_2(sbi->s_inode_size)) ||
3651 (sbi->s_inode_size > blocksize)) {
3652 ext4_msg(sb, KERN_ERR,
3653 "unsupported inode size: %d",
3657 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3658 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3661 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3662 if (ext4_has_feature_64bit(sb)) {
3663 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3664 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3665 !is_power_of_2(sbi->s_desc_size)) {
3666 ext4_msg(sb, KERN_ERR,
3667 "unsupported descriptor size %lu",
3672 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3674 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3675 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3677 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3678 if (sbi->s_inodes_per_block == 0)
3680 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
3681 sbi->s_inodes_per_group > blocksize * 8) {
3682 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
3683 sbi->s_blocks_per_group);
3686 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3687 sbi->s_inodes_per_block;
3688 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3690 sbi->s_mount_state = le16_to_cpu(es->s_state);
3691 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3692 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3694 for (i = 0; i < 4; i++)
3695 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3696 sbi->s_def_hash_version = es->s_def_hash_version;
3697 if (ext4_has_feature_dir_index(sb)) {
3698 i = le32_to_cpu(es->s_flags);
3699 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3700 sbi->s_hash_unsigned = 3;
3701 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3702 #ifdef __CHAR_UNSIGNED__
3703 if (!(sb->s_flags & MS_RDONLY))
3705 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3706 sbi->s_hash_unsigned = 3;
3708 if (!(sb->s_flags & MS_RDONLY))
3710 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3715 /* Handle clustersize */
3716 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3717 has_bigalloc = ext4_has_feature_bigalloc(sb);
3719 if (clustersize < blocksize) {
3720 ext4_msg(sb, KERN_ERR,
3721 "cluster size (%d) smaller than "
3722 "block size (%d)", clustersize, blocksize);
3725 if (le32_to_cpu(es->s_log_cluster_size) >
3726 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3727 ext4_msg(sb, KERN_ERR,
3728 "Invalid log cluster size: %u",
3729 le32_to_cpu(es->s_log_cluster_size));
3732 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3733 le32_to_cpu(es->s_log_block_size);
3734 sbi->s_clusters_per_group =
3735 le32_to_cpu(es->s_clusters_per_group);
3736 if (sbi->s_clusters_per_group > blocksize * 8) {
3737 ext4_msg(sb, KERN_ERR,
3738 "#clusters per group too big: %lu",
3739 sbi->s_clusters_per_group);
3742 if (sbi->s_blocks_per_group !=
3743 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3744 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3745 "clusters per group (%lu) inconsistent",
3746 sbi->s_blocks_per_group,
3747 sbi->s_clusters_per_group);
3751 if (clustersize != blocksize) {
3752 ext4_warning(sb, "fragment/cluster size (%d) != "
3753 "block size (%d)", clustersize,
3755 clustersize = blocksize;
3757 if (sbi->s_blocks_per_group > blocksize * 8) {
3758 ext4_msg(sb, KERN_ERR,
3759 "#blocks per group too big: %lu",
3760 sbi->s_blocks_per_group);
3763 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3764 sbi->s_cluster_bits = 0;
3766 sbi->s_cluster_ratio = clustersize / blocksize;
3768 /* Do we have standard group size of clustersize * 8 blocks ? */
3769 if (sbi->s_blocks_per_group == clustersize << 3)
3770 set_opt2(sb, STD_GROUP_SIZE);
3773 * Test whether we have more sectors than will fit in sector_t,
3774 * and whether the max offset is addressable by the page cache.
3776 err = generic_check_addressable(sb->s_blocksize_bits,
3777 ext4_blocks_count(es));
3779 ext4_msg(sb, KERN_ERR, "filesystem"
3780 " too large to mount safely on this system");
3781 if (sizeof(sector_t) < 8)
3782 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3786 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3789 /* check blocks count against device size */
3790 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3791 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3792 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3793 "exceeds size of device (%llu blocks)",
3794 ext4_blocks_count(es), blocks_count);
3799 * It makes no sense for the first data block to be beyond the end
3800 * of the filesystem.
3802 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3803 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3804 "block %u is beyond end of filesystem (%llu)",
3805 le32_to_cpu(es->s_first_data_block),
3806 ext4_blocks_count(es));
3809 blocks_count = (ext4_blocks_count(es) -
3810 le32_to_cpu(es->s_first_data_block) +
3811 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3812 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3813 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3814 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3815 "(block count %llu, first data block %u, "
3816 "blocks per group %lu)", sbi->s_groups_count,
3817 ext4_blocks_count(es),
3818 le32_to_cpu(es->s_first_data_block),
3819 EXT4_BLOCKS_PER_GROUP(sb));
3822 sbi->s_groups_count = blocks_count;
3823 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3824 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3825 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3826 EXT4_DESC_PER_BLOCK(sb);
3827 sbi->s_group_desc = ext4_kvmalloc(db_count *
3828 sizeof(struct buffer_head *),
3830 if (sbi->s_group_desc == NULL) {
3831 ext4_msg(sb, KERN_ERR, "not enough memory");
3836 bgl_lock_init(sbi->s_blockgroup_lock);
3838 for (i = 0; i < db_count; i++) {
3839 block = descriptor_loc(sb, logical_sb_block, i);
3840 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
3841 if (!sbi->s_group_desc[i]) {
3842 ext4_msg(sb, KERN_ERR,
3843 "can't read group descriptor %d", i);
3848 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
3849 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3850 ret = -EFSCORRUPTED;
3854 sbi->s_gdb_count = db_count;
3855 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3856 spin_lock_init(&sbi->s_next_gen_lock);
3858 setup_timer(&sbi->s_err_report, print_daily_error_info,
3859 (unsigned long) sb);
3861 /* Register extent status tree shrinker */
3862 if (ext4_es_register_shrinker(sbi))
3865 sbi->s_stripe = ext4_get_stripe_size(sbi);
3866 sbi->s_extent_max_zeroout_kb = 32;
3869 * set up enough so that it can read an inode
3871 sb->s_op = &ext4_sops;
3872 sb->s_export_op = &ext4_export_ops;
3873 sb->s_xattr = ext4_xattr_handlers;
3874 sb->s_cop = &ext4_cryptops;
3876 sb->dq_op = &ext4_quota_operations;
3877 if (ext4_has_feature_quota(sb))
3878 sb->s_qcop = &dquot_quotactl_sysfile_ops;
3880 sb->s_qcop = &ext4_qctl_operations;
3881 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
3883 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3885 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3886 mutex_init(&sbi->s_orphan_lock);
3890 needs_recovery = (es->s_last_orphan != 0 ||
3891 ext4_has_feature_journal_needs_recovery(sb));
3893 if (ext4_has_feature_mmp(sb) && !(sb->s_flags & MS_RDONLY))
3894 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3895 goto failed_mount3a;
3898 * The first inode we look at is the journal inode. Don't try
3899 * root first: it may be modified in the journal!
3901 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
3902 if (ext4_load_journal(sb, es, journal_devnum))
3903 goto failed_mount3a;
3904 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3905 ext4_has_feature_journal_needs_recovery(sb)) {
3906 ext4_msg(sb, KERN_ERR, "required journal recovery "
3907 "suppressed and not mounted read-only");
3908 goto failed_mount_wq;
3910 /* Nojournal mode, all journal mount options are illegal */
3911 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
3912 ext4_msg(sb, KERN_ERR, "can't mount with "
3913 "journal_checksum, fs mounted w/o journal");
3914 goto failed_mount_wq;
3916 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3917 ext4_msg(sb, KERN_ERR, "can't mount with "
3918 "journal_async_commit, fs mounted w/o journal");
3919 goto failed_mount_wq;
3921 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
3922 ext4_msg(sb, KERN_ERR, "can't mount with "
3923 "commit=%lu, fs mounted w/o journal",
3924 sbi->s_commit_interval / HZ);
3925 goto failed_mount_wq;
3927 if (EXT4_MOUNT_DATA_FLAGS &
3928 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
3929 ext4_msg(sb, KERN_ERR, "can't mount with "
3930 "data=, fs mounted w/o journal");
3931 goto failed_mount_wq;
3933 sbi->s_def_mount_opt &= EXT4_MOUNT_JOURNAL_CHECKSUM;
3934 clear_opt(sb, JOURNAL_CHECKSUM);
3935 clear_opt(sb, DATA_FLAGS);
3936 sbi->s_journal = NULL;
3941 if (ext4_has_feature_64bit(sb) &&
3942 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3943 JBD2_FEATURE_INCOMPAT_64BIT)) {
3944 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3945 goto failed_mount_wq;
3948 if (!set_journal_csum_feature_set(sb)) {
3949 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3951 goto failed_mount_wq;
3954 /* We have now updated the journal if required, so we can
3955 * validate the data journaling mode. */
3956 switch (test_opt(sb, DATA_FLAGS)) {
3958 /* No mode set, assume a default based on the journal
3959 * capabilities: ORDERED_DATA if the journal can
3960 * cope, else JOURNAL_DATA
3962 if (jbd2_journal_check_available_features
3963 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3964 set_opt(sb, ORDERED_DATA);
3966 set_opt(sb, JOURNAL_DATA);
3969 case EXT4_MOUNT_ORDERED_DATA:
3970 case EXT4_MOUNT_WRITEBACK_DATA:
3971 if (!jbd2_journal_check_available_features
3972 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3973 ext4_msg(sb, KERN_ERR, "Journal does not support "
3974 "requested data journaling mode");
3975 goto failed_mount_wq;
3980 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3982 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3985 sbi->s_mb_cache = ext4_xattr_create_cache();
3986 if (!sbi->s_mb_cache) {
3987 ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
3988 goto failed_mount_wq;
3991 if ((DUMMY_ENCRYPTION_ENABLED(sbi) || ext4_has_feature_encrypt(sb)) &&
3992 (blocksize != PAGE_SIZE)) {
3993 ext4_msg(sb, KERN_ERR,
3994 "Unsupported blocksize for fs encryption");
3995 goto failed_mount_wq;
3998 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !(sb->s_flags & MS_RDONLY) &&
3999 !ext4_has_feature_encrypt(sb)) {
4000 ext4_set_feature_encrypt(sb);
4001 ext4_commit_super(sb, 1);
4005 * Get the # of file system overhead blocks from the
4006 * superblock if present.
4008 if (es->s_overhead_clusters)
4009 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4011 err = ext4_calculate_overhead(sb);
4013 goto failed_mount_wq;
4017 * The maximum number of concurrent works can be high and
4018 * concurrency isn't really necessary. Limit it to 1.
4020 EXT4_SB(sb)->rsv_conversion_wq =
4021 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4022 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4023 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
4029 * The jbd2_journal_load will have done any necessary log recovery,
4030 * so we can safely mount the rest of the filesystem now.
4033 root = ext4_iget(sb, EXT4_ROOT_INO);
4035 ext4_msg(sb, KERN_ERR, "get root inode failed");
4036 ret = PTR_ERR(root);
4040 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4041 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4045 sb->s_root = d_make_root(root);
4047 ext4_msg(sb, KERN_ERR, "get root dentry failed");
4052 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
4053 sb->s_flags |= MS_RDONLY;
4055 /* determine the minimum size of new large inodes, if present */
4056 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4057 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4058 EXT4_GOOD_OLD_INODE_SIZE;
4059 if (ext4_has_feature_extra_isize(sb)) {
4060 if (sbi->s_want_extra_isize <
4061 le16_to_cpu(es->s_want_extra_isize))
4062 sbi->s_want_extra_isize =
4063 le16_to_cpu(es->s_want_extra_isize);
4064 if (sbi->s_want_extra_isize <
4065 le16_to_cpu(es->s_min_extra_isize))
4066 sbi->s_want_extra_isize =
4067 le16_to_cpu(es->s_min_extra_isize);
4070 /* Check if enough inode space is available */
4071 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4072 sbi->s_inode_size) {
4073 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4074 EXT4_GOOD_OLD_INODE_SIZE;
4075 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4079 ext4_set_resv_clusters(sb);
4081 err = ext4_setup_system_zone(sb);
4083 ext4_msg(sb, KERN_ERR, "failed to initialize system "
4085 goto failed_mount4a;
4089 err = ext4_mb_init(sb);
4091 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4096 block = ext4_count_free_clusters(sb);
4097 ext4_free_blocks_count_set(sbi->s_es,
4098 EXT4_C2B(sbi, block));
4099 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4102 unsigned long freei = ext4_count_free_inodes(sb);
4103 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
4104 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4108 err = percpu_counter_init(&sbi->s_dirs_counter,
4109 ext4_count_dirs(sb), GFP_KERNEL);
4111 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4114 err = percpu_init_rwsem(&sbi->s_journal_flag_rwsem);
4117 ext4_msg(sb, KERN_ERR, "insufficient memory");
4121 if (ext4_has_feature_flex_bg(sb))
4122 if (!ext4_fill_flex_info(sb)) {
4123 ext4_msg(sb, KERN_ERR,
4124 "unable to initialize "
4125 "flex_bg meta info!");
4129 err = ext4_register_li_request(sb, first_not_zeroed);
4133 err = ext4_register_sysfs(sb);
4138 /* Enable quota usage during mount. */
4139 if (ext4_has_feature_quota(sb) && !(sb->s_flags & MS_RDONLY)) {
4140 err = ext4_enable_quotas(sb);
4144 #endif /* CONFIG_QUOTA */
4146 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4147 ext4_orphan_cleanup(sb, es);
4148 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4149 if (needs_recovery) {
4150 ext4_msg(sb, KERN_INFO, "recovery complete");
4151 ext4_mark_recovery_complete(sb, es);
4153 if (EXT4_SB(sb)->s_journal) {
4154 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4155 descr = " journalled data mode";
4156 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4157 descr = " ordered data mode";
4159 descr = " writeback data mode";
4161 descr = "out journal";
4163 if (test_opt(sb, DISCARD)) {
4164 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4165 if (!blk_queue_discard(q))
4166 ext4_msg(sb, KERN_WARNING,
4167 "mounting with \"discard\" option, but "
4168 "the device does not support discard");
4171 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
4172 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4173 "Opts: %.*s%s%s", descr,
4174 (int) sizeof(sbi->s_es->s_mount_opts),
4175 sbi->s_es->s_mount_opts,
4176 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4178 if (es->s_error_count)
4179 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4181 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4182 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4183 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4184 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4187 #ifdef CONFIG_EXT4_FS_ENCRYPTION
4188 memcpy(sbi->key_prefix, EXT4_KEY_DESC_PREFIX,
4189 EXT4_KEY_DESC_PREFIX_SIZE);
4190 sbi->key_prefix_size = EXT4_KEY_DESC_PREFIX_SIZE;
4196 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4201 ext4_unregister_sysfs(sb);
4204 ext4_unregister_li_request(sb);
4206 ext4_mb_release(sb);
4207 if (sbi->s_flex_groups)
4208 kvfree(sbi->s_flex_groups);
4209 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4210 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4211 percpu_counter_destroy(&sbi->s_dirs_counter);
4212 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4214 ext4_ext_release(sb);
4215 ext4_release_system_zone(sb);
4220 ext4_msg(sb, KERN_ERR, "mount failed");
4221 if (EXT4_SB(sb)->rsv_conversion_wq)
4222 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4224 if (sbi->s_mb_cache) {
4225 ext4_xattr_destroy_cache(sbi->s_mb_cache);
4226 sbi->s_mb_cache = NULL;
4228 if (sbi->s_journal) {
4229 jbd2_journal_destroy(sbi->s_journal);
4230 sbi->s_journal = NULL;
4233 ext4_es_unregister_shrinker(sbi);
4235 del_timer_sync(&sbi->s_err_report);
4237 kthread_stop(sbi->s_mmp_tsk);
4239 for (i = 0; i < db_count; i++)
4240 brelse(sbi->s_group_desc[i]);
4241 kvfree(sbi->s_group_desc);
4243 if (sbi->s_chksum_driver)
4244 crypto_free_shash(sbi->s_chksum_driver);
4246 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4247 kfree(sbi->s_qf_names[i]);
4249 ext4_blkdev_remove(sbi);
4252 sb->s_fs_info = NULL;
4253 kfree(sbi->s_blockgroup_lock);
4257 return err ? err : ret;
4261 * Setup any per-fs journal parameters now. We'll do this both on
4262 * initial mount, once the journal has been initialised but before we've
4263 * done any recovery; and again on any subsequent remount.
4265 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4267 struct ext4_sb_info *sbi = EXT4_SB(sb);
4269 journal->j_commit_interval = sbi->s_commit_interval;
4270 journal->j_min_batch_time = sbi->s_min_batch_time;
4271 journal->j_max_batch_time = sbi->s_max_batch_time;
4273 write_lock(&journal->j_state_lock);
4274 if (test_opt(sb, BARRIER))
4275 journal->j_flags |= JBD2_BARRIER;
4277 journal->j_flags &= ~JBD2_BARRIER;
4278 if (test_opt(sb, DATA_ERR_ABORT))
4279 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4281 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4282 write_unlock(&journal->j_state_lock);
4285 static struct inode *ext4_get_journal_inode(struct super_block *sb,
4286 unsigned int journal_inum)
4288 struct inode *journal_inode;
4291 * Test for the existence of a valid inode on disk. Bad things
4292 * happen if we iget() an unused inode, as the subsequent iput()
4293 * will try to delete it.
4295 journal_inode = ext4_iget(sb, journal_inum);
4296 if (IS_ERR(journal_inode)) {
4297 ext4_msg(sb, KERN_ERR, "no journal found");
4300 if (!journal_inode->i_nlink) {
4301 make_bad_inode(journal_inode);
4302 iput(journal_inode);
4303 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4307 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4308 journal_inode, journal_inode->i_size);
4309 if (!S_ISREG(journal_inode->i_mode)) {
4310 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4311 iput(journal_inode);
4314 return journal_inode;
4317 static journal_t *ext4_get_journal(struct super_block *sb,
4318 unsigned int journal_inum)
4320 struct inode *journal_inode;
4323 BUG_ON(!ext4_has_feature_journal(sb));
4325 journal_inode = ext4_get_journal_inode(sb, journal_inum);
4329 journal = jbd2_journal_init_inode(journal_inode);
4331 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4332 iput(journal_inode);
4335 journal->j_private = sb;
4336 ext4_init_journal_params(sb, journal);
4340 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4343 struct buffer_head *bh;
4347 int hblock, blocksize;
4348 ext4_fsblk_t sb_block;
4349 unsigned long offset;
4350 struct ext4_super_block *es;
4351 struct block_device *bdev;
4353 BUG_ON(!ext4_has_feature_journal(sb));
4355 bdev = ext4_blkdev_get(j_dev, sb);
4359 blocksize = sb->s_blocksize;
4360 hblock = bdev_logical_block_size(bdev);
4361 if (blocksize < hblock) {
4362 ext4_msg(sb, KERN_ERR,
4363 "blocksize too small for journal device");
4367 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4368 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4369 set_blocksize(bdev, blocksize);
4370 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4371 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4372 "external journal");
4376 es = (struct ext4_super_block *) (bh->b_data + offset);
4377 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4378 !(le32_to_cpu(es->s_feature_incompat) &
4379 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4380 ext4_msg(sb, KERN_ERR, "external journal has "
4386 if ((le32_to_cpu(es->s_feature_ro_compat) &
4387 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4388 es->s_checksum != ext4_superblock_csum(sb, es)) {
4389 ext4_msg(sb, KERN_ERR, "external journal has "
4390 "corrupt superblock");
4395 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4396 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4401 len = ext4_blocks_count(es);
4402 start = sb_block + 1;
4403 brelse(bh); /* we're done with the superblock */
4405 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4406 start, len, blocksize);
4408 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4411 journal->j_private = sb;
4412 ll_rw_block(REQ_OP_READ, REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4413 wait_on_buffer(journal->j_sb_buffer);
4414 if (!buffer_uptodate(journal->j_sb_buffer)) {
4415 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4418 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4419 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4420 "user (unsupported) - %d",
4421 be32_to_cpu(journal->j_superblock->s_nr_users));
4424 EXT4_SB(sb)->journal_bdev = bdev;
4425 ext4_init_journal_params(sb, journal);
4429 jbd2_journal_destroy(journal);
4431 ext4_blkdev_put(bdev);
4435 static int ext4_load_journal(struct super_block *sb,
4436 struct ext4_super_block *es,
4437 unsigned long journal_devnum)
4440 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4443 int really_read_only;
4445 BUG_ON(!ext4_has_feature_journal(sb));
4447 if (journal_devnum &&
4448 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4449 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4450 "numbers have changed");
4451 journal_dev = new_decode_dev(journal_devnum);
4453 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4455 really_read_only = bdev_read_only(sb->s_bdev);
4458 * Are we loading a blank journal or performing recovery after a
4459 * crash? For recovery, we need to check in advance whether we
4460 * can get read-write access to the device.
4462 if (ext4_has_feature_journal_needs_recovery(sb)) {
4463 if (sb->s_flags & MS_RDONLY) {
4464 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4465 "required on readonly filesystem");
4466 if (really_read_only) {
4467 ext4_msg(sb, KERN_ERR, "write access "
4468 "unavailable, cannot proceed");
4471 ext4_msg(sb, KERN_INFO, "write access will "
4472 "be enabled during recovery");
4476 if (journal_inum && journal_dev) {
4477 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4478 "and inode journals!");
4483 if (!(journal = ext4_get_journal(sb, journal_inum)))
4486 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4490 if (!(journal->j_flags & JBD2_BARRIER))
4491 ext4_msg(sb, KERN_INFO, "barriers disabled");
4493 if (!ext4_has_feature_journal_needs_recovery(sb))
4494 err = jbd2_journal_wipe(journal, !really_read_only);
4496 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4498 memcpy(save, ((char *) es) +
4499 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4500 err = jbd2_journal_load(journal);
4502 memcpy(((char *) es) + EXT4_S_ERR_START,
4503 save, EXT4_S_ERR_LEN);
4508 ext4_msg(sb, KERN_ERR, "error loading journal");
4509 jbd2_journal_destroy(journal);
4513 EXT4_SB(sb)->s_journal = journal;
4514 ext4_clear_journal_err(sb, es);
4516 if (!really_read_only && journal_devnum &&
4517 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4518 es->s_journal_dev = cpu_to_le32(journal_devnum);
4520 /* Make sure we flush the recovery flag to disk. */
4521 ext4_commit_super(sb, 1);
4527 static int ext4_commit_super(struct super_block *sb, int sync)
4529 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4530 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4533 if (!sbh || block_device_ejected(sb))
4536 * If the file system is mounted read-only, don't update the
4537 * superblock write time. This avoids updating the superblock
4538 * write time when we are mounting the root file system
4539 * read/only but we need to replay the journal; at that point,
4540 * for people who are east of GMT and who make their clock
4541 * tick in localtime for Windows bug-for-bug compatibility,
4542 * the clock is set in the future, and this will cause e2fsck
4543 * to complain and force a full file system check.
4545 if (!(sb->s_flags & MS_RDONLY))
4546 es->s_wtime = cpu_to_le32(get_seconds());
4547 if (sb->s_bdev->bd_part)
4548 es->s_kbytes_written =
4549 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4550 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4551 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4553 es->s_kbytes_written =
4554 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4555 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4556 ext4_free_blocks_count_set(es,
4557 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4558 &EXT4_SB(sb)->s_freeclusters_counter)));
4559 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4560 es->s_free_inodes_count =
4561 cpu_to_le32(percpu_counter_sum_positive(
4562 &EXT4_SB(sb)->s_freeinodes_counter));
4563 BUFFER_TRACE(sbh, "marking dirty");
4564 ext4_superblock_csum_set(sb);
4567 if (buffer_write_io_error(sbh)) {
4569 * Oh, dear. A previous attempt to write the
4570 * superblock failed. This could happen because the
4571 * USB device was yanked out. Or it could happen to
4572 * be a transient write error and maybe the block will
4573 * be remapped. Nothing we can do but to retry the
4574 * write and hope for the best.
4576 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4577 "superblock detected");
4578 clear_buffer_write_io_error(sbh);
4579 set_buffer_uptodate(sbh);
4581 mark_buffer_dirty(sbh);
4584 error = __sync_dirty_buffer(sbh,
4585 test_opt(sb, BARRIER) ? WRITE_FUA : WRITE_SYNC);
4589 error = buffer_write_io_error(sbh);
4591 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4593 clear_buffer_write_io_error(sbh);
4594 set_buffer_uptodate(sbh);
4601 * Have we just finished recovery? If so, and if we are mounting (or
4602 * remounting) the filesystem readonly, then we will end up with a
4603 * consistent fs on disk. Record that fact.
4605 static void ext4_mark_recovery_complete(struct super_block *sb,
4606 struct ext4_super_block *es)
4608 journal_t *journal = EXT4_SB(sb)->s_journal;
4610 if (!ext4_has_feature_journal(sb)) {
4611 BUG_ON(journal != NULL);
4614 jbd2_journal_lock_updates(journal);
4615 if (jbd2_journal_flush(journal) < 0)
4618 if (ext4_has_feature_journal_needs_recovery(sb) &&
4619 sb->s_flags & MS_RDONLY) {
4620 ext4_clear_feature_journal_needs_recovery(sb);
4621 ext4_commit_super(sb, 1);
4625 jbd2_journal_unlock_updates(journal);
4629 * If we are mounting (or read-write remounting) a filesystem whose journal
4630 * has recorded an error from a previous lifetime, move that error to the
4631 * main filesystem now.
4633 static void ext4_clear_journal_err(struct super_block *sb,
4634 struct ext4_super_block *es)
4640 BUG_ON(!ext4_has_feature_journal(sb));
4642 journal = EXT4_SB(sb)->s_journal;
4645 * Now check for any error status which may have been recorded in the
4646 * journal by a prior ext4_error() or ext4_abort()
4649 j_errno = jbd2_journal_errno(journal);
4653 errstr = ext4_decode_error(sb, j_errno, nbuf);
4654 ext4_warning(sb, "Filesystem error recorded "
4655 "from previous mount: %s", errstr);
4656 ext4_warning(sb, "Marking fs in need of filesystem check.");
4658 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4659 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4660 ext4_commit_super(sb, 1);
4662 jbd2_journal_clear_err(journal);
4663 jbd2_journal_update_sb_errno(journal);
4668 * Force the running and committing transactions to commit,
4669 * and wait on the commit.
4671 int ext4_force_commit(struct super_block *sb)
4675 if (sb->s_flags & MS_RDONLY)
4678 journal = EXT4_SB(sb)->s_journal;
4679 return ext4_journal_force_commit(journal);
4682 static int ext4_sync_fs(struct super_block *sb, int wait)
4686 bool needs_barrier = false;
4687 struct ext4_sb_info *sbi = EXT4_SB(sb);
4689 trace_ext4_sync_fs(sb, wait);
4690 flush_workqueue(sbi->rsv_conversion_wq);
4692 * Writeback quota in non-journalled quota case - journalled quota has
4695 dquot_writeback_dquots(sb, -1);
4697 * Data writeback is possible w/o journal transaction, so barrier must
4698 * being sent at the end of the function. But we can skip it if
4699 * transaction_commit will do it for us.
4701 if (sbi->s_journal) {
4702 target = jbd2_get_latest_transaction(sbi->s_journal);
4703 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4704 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4705 needs_barrier = true;
4707 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4709 ret = jbd2_log_wait_commit(sbi->s_journal,
4712 } else if (wait && test_opt(sb, BARRIER))
4713 needs_barrier = true;
4714 if (needs_barrier) {
4716 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4725 * LVM calls this function before a (read-only) snapshot is created. This
4726 * gives us a chance to flush the journal completely and mark the fs clean.
4728 * Note that only this function cannot bring a filesystem to be in a clean
4729 * state independently. It relies on upper layer to stop all data & metadata
4732 static int ext4_freeze(struct super_block *sb)
4737 if (sb->s_flags & MS_RDONLY)
4740 journal = EXT4_SB(sb)->s_journal;
4743 /* Now we set up the journal barrier. */
4744 jbd2_journal_lock_updates(journal);
4747 * Don't clear the needs_recovery flag if we failed to
4748 * flush the journal.
4750 error = jbd2_journal_flush(journal);
4754 /* Journal blocked and flushed, clear needs_recovery flag. */
4755 ext4_clear_feature_journal_needs_recovery(sb);
4758 error = ext4_commit_super(sb, 1);
4761 /* we rely on upper layer to stop further updates */
4762 jbd2_journal_unlock_updates(journal);
4767 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4768 * flag here, even though the filesystem is not technically dirty yet.
4770 static int ext4_unfreeze(struct super_block *sb)
4772 if (sb->s_flags & MS_RDONLY)
4775 if (EXT4_SB(sb)->s_journal) {
4776 /* Reset the needs_recovery flag before the fs is unlocked. */
4777 ext4_set_feature_journal_needs_recovery(sb);
4780 ext4_commit_super(sb, 1);
4785 * Structure to save mount options for ext4_remount's benefit
4787 struct ext4_mount_options {
4788 unsigned long s_mount_opt;
4789 unsigned long s_mount_opt2;
4792 unsigned long s_commit_interval;
4793 u32 s_min_batch_time, s_max_batch_time;
4796 char *s_qf_names[EXT4_MAXQUOTAS];
4800 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4802 struct ext4_super_block *es;
4803 struct ext4_sb_info *sbi = EXT4_SB(sb);
4804 unsigned long old_sb_flags;
4805 struct ext4_mount_options old_opts;
4806 int enable_quota = 0;
4808 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4813 char *orig_data = kstrdup(data, GFP_KERNEL);
4815 /* Store the original options */
4816 old_sb_flags = sb->s_flags;
4817 old_opts.s_mount_opt = sbi->s_mount_opt;
4818 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4819 old_opts.s_resuid = sbi->s_resuid;
4820 old_opts.s_resgid = sbi->s_resgid;
4821 old_opts.s_commit_interval = sbi->s_commit_interval;
4822 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4823 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4825 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4826 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4827 if (sbi->s_qf_names[i]) {
4828 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4830 if (!old_opts.s_qf_names[i]) {
4831 for (j = 0; j < i; j++)
4832 kfree(old_opts.s_qf_names[j]);
4837 old_opts.s_qf_names[i] = NULL;
4839 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4840 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4842 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4847 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
4848 test_opt(sb, JOURNAL_CHECKSUM)) {
4849 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
4850 "during remount not supported; ignoring");
4851 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
4854 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4855 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4856 ext4_msg(sb, KERN_ERR, "can't mount with "
4857 "both data=journal and delalloc");
4861 if (test_opt(sb, DIOREAD_NOLOCK)) {
4862 ext4_msg(sb, KERN_ERR, "can't mount with "
4863 "both data=journal and dioread_nolock");
4867 if (test_opt(sb, DAX)) {
4868 ext4_msg(sb, KERN_ERR, "can't mount with "
4869 "both data=journal and dax");
4875 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
4876 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
4877 "dax flag with busy inodes while remounting");
4878 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
4881 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4882 ext4_abort(sb, "Abort forced by user");
4884 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4885 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4889 if (sbi->s_journal) {
4890 ext4_init_journal_params(sb, sbi->s_journal);
4891 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4894 if (*flags & MS_LAZYTIME)
4895 sb->s_flags |= MS_LAZYTIME;
4897 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4898 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4903 if (*flags & MS_RDONLY) {
4904 err = sync_filesystem(sb);
4907 err = dquot_suspend(sb, -1);
4912 * First of all, the unconditional stuff we have to do
4913 * to disable replay of the journal when we next remount
4915 sb->s_flags |= MS_RDONLY;
4918 * OK, test if we are remounting a valid rw partition
4919 * readonly, and if so set the rdonly flag and then
4920 * mark the partition as valid again.
4922 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4923 (sbi->s_mount_state & EXT4_VALID_FS))
4924 es->s_state = cpu_to_le16(sbi->s_mount_state);
4927 ext4_mark_recovery_complete(sb, es);
4929 /* Make sure we can mount this feature set readwrite */
4930 if (ext4_has_feature_readonly(sb) ||
4931 !ext4_feature_set_ok(sb, 0)) {
4936 * Make sure the group descriptor checksums
4937 * are sane. If they aren't, refuse to remount r/w.
4939 for (g = 0; g < sbi->s_groups_count; g++) {
4940 struct ext4_group_desc *gdp =
4941 ext4_get_group_desc(sb, g, NULL);
4943 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
4944 ext4_msg(sb, KERN_ERR,
4945 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4946 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
4947 le16_to_cpu(gdp->bg_checksum));
4954 * If we have an unprocessed orphan list hanging
4955 * around from a previously readonly bdev mount,
4956 * require a full umount/remount for now.
4958 if (es->s_last_orphan) {
4959 ext4_msg(sb, KERN_WARNING, "Couldn't "
4960 "remount RDWR because of unprocessed "
4961 "orphan inode list. Please "
4962 "umount/remount instead");
4968 * Mounting a RDONLY partition read-write, so reread
4969 * and store the current valid flag. (It may have
4970 * been changed by e2fsck since we originally mounted
4974 ext4_clear_journal_err(sb, es);
4975 sbi->s_mount_state = le16_to_cpu(es->s_state);
4976 if (!ext4_setup_super(sb, es, 0))
4977 sb->s_flags &= ~MS_RDONLY;
4978 if (ext4_has_feature_mmp(sb))
4979 if (ext4_multi_mount_protect(sb,
4980 le64_to_cpu(es->s_mmp_block))) {
4989 * Reinitialize lazy itable initialization thread based on
4992 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4993 ext4_unregister_li_request(sb);
4995 ext4_group_t first_not_zeroed;
4996 first_not_zeroed = ext4_has_uninit_itable(sb);
4997 ext4_register_li_request(sb, first_not_zeroed);
5000 ext4_setup_system_zone(sb);
5001 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
5002 ext4_commit_super(sb, 1);
5005 /* Release old quota file names */
5006 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5007 kfree(old_opts.s_qf_names[i]);
5009 if (sb_any_quota_suspended(sb))
5010 dquot_resume(sb, -1);
5011 else if (ext4_has_feature_quota(sb)) {
5012 err = ext4_enable_quotas(sb);
5019 *flags = (*flags & ~MS_LAZYTIME) | (sb->s_flags & MS_LAZYTIME);
5020 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5025 sb->s_flags = old_sb_flags;
5026 sbi->s_mount_opt = old_opts.s_mount_opt;
5027 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
5028 sbi->s_resuid = old_opts.s_resuid;
5029 sbi->s_resgid = old_opts.s_resgid;
5030 sbi->s_commit_interval = old_opts.s_commit_interval;
5031 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5032 sbi->s_max_batch_time = old_opts.s_max_batch_time;
5034 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
5035 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
5036 kfree(sbi->s_qf_names[i]);
5037 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
5045 static int ext4_statfs_project(struct super_block *sb,
5046 kprojid_t projid, struct kstatfs *buf)
5049 struct dquot *dquot;
5053 qid = make_kqid_projid(projid);
5054 dquot = dqget(sb, qid);
5056 return PTR_ERR(dquot);
5057 spin_lock(&dq_data_lock);
5059 limit = (dquot->dq_dqb.dqb_bsoftlimit ?
5060 dquot->dq_dqb.dqb_bsoftlimit :
5061 dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits;
5062 if (limit && buf->f_blocks > limit) {
5063 curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits;
5064 buf->f_blocks = limit;
5065 buf->f_bfree = buf->f_bavail =
5066 (buf->f_blocks > curblock) ?
5067 (buf->f_blocks - curblock) : 0;
5070 limit = dquot->dq_dqb.dqb_isoftlimit ?
5071 dquot->dq_dqb.dqb_isoftlimit :
5072 dquot->dq_dqb.dqb_ihardlimit;
5073 if (limit && buf->f_files > limit) {
5074 buf->f_files = limit;
5076 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
5077 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
5080 spin_unlock(&dq_data_lock);
5086 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
5088 struct super_block *sb = dentry->d_sb;
5089 struct ext4_sb_info *sbi = EXT4_SB(sb);
5090 struct ext4_super_block *es = sbi->s_es;
5091 ext4_fsblk_t overhead = 0, resv_blocks;
5094 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
5096 if (!test_opt(sb, MINIX_DF))
5097 overhead = sbi->s_overhead;
5099 buf->f_type = EXT4_SUPER_MAGIC;
5100 buf->f_bsize = sb->s_blocksize;
5101 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
5102 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5103 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
5104 /* prevent underflow in case that few free space is available */
5105 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
5106 buf->f_bavail = buf->f_bfree -
5107 (ext4_r_blocks_count(es) + resv_blocks);
5108 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
5110 buf->f_files = le32_to_cpu(es->s_inodes_count);
5111 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5112 buf->f_namelen = EXT4_NAME_LEN;
5113 fsid = le64_to_cpup((void *)es->s_uuid) ^
5114 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5115 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5116 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
5119 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
5120 sb_has_quota_limits_enabled(sb, PRJQUOTA))
5121 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
5126 /* Helper function for writing quotas on sync - we need to start transaction
5127 * before quota file is locked for write. Otherwise the are possible deadlocks:
5128 * Process 1 Process 2
5129 * ext4_create() quota_sync()
5130 * jbd2_journal_start() write_dquot()
5131 * dquot_initialize() down(dqio_mutex)
5132 * down(dqio_mutex) jbd2_journal_start()
5138 static inline struct inode *dquot_to_inode(struct dquot *dquot)
5140 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
5143 static int ext4_write_dquot(struct dquot *dquot)
5147 struct inode *inode;
5149 inode = dquot_to_inode(dquot);
5150 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
5151 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
5153 return PTR_ERR(handle);
5154 ret = dquot_commit(dquot);
5155 err = ext4_journal_stop(handle);
5161 static int ext4_acquire_dquot(struct dquot *dquot)
5166 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5167 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
5169 return PTR_ERR(handle);
5170 ret = dquot_acquire(dquot);
5171 err = ext4_journal_stop(handle);
5177 static int ext4_release_dquot(struct dquot *dquot)
5182 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5183 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
5184 if (IS_ERR(handle)) {
5185 /* Release dquot anyway to avoid endless cycle in dqput() */
5186 dquot_release(dquot);
5187 return PTR_ERR(handle);
5189 ret = dquot_release(dquot);
5190 err = ext4_journal_stop(handle);
5196 static int ext4_mark_dquot_dirty(struct dquot *dquot)
5198 struct super_block *sb = dquot->dq_sb;
5199 struct ext4_sb_info *sbi = EXT4_SB(sb);
5201 /* Are we journaling quotas? */
5202 if (ext4_has_feature_quota(sb) ||
5203 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
5204 dquot_mark_dquot_dirty(dquot);
5205 return ext4_write_dquot(dquot);
5207 return dquot_mark_dquot_dirty(dquot);
5211 static int ext4_write_info(struct super_block *sb, int type)
5216 /* Data block + inode block */
5217 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
5219 return PTR_ERR(handle);
5220 ret = dquot_commit_info(sb, type);
5221 err = ext4_journal_stop(handle);
5228 * Turn on quotas during mount time - we need to find
5229 * the quota file and such...
5231 static int ext4_quota_on_mount(struct super_block *sb, int type)
5233 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5234 EXT4_SB(sb)->s_jquota_fmt, type);
5237 static void lockdep_set_quota_inode(struct inode *inode, int subclass)
5239 struct ext4_inode_info *ei = EXT4_I(inode);
5241 /* The first argument of lockdep_set_subclass has to be
5242 * *exactly* the same as the argument to init_rwsem() --- in
5243 * this case, in init_once() --- or lockdep gets unhappy
5244 * because the name of the lock is set using the
5245 * stringification of the argument to init_rwsem().
5247 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
5248 lockdep_set_subclass(&ei->i_data_sem, subclass);
5252 * Standard function to be called on quota_on
5254 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
5259 if (!test_opt(sb, QUOTA))
5262 /* Quotafile not on the same filesystem? */
5263 if (path->dentry->d_sb != sb)
5265 /* Journaling quota? */
5266 if (EXT4_SB(sb)->s_qf_names[type]) {
5267 /* Quotafile not in fs root? */
5268 if (path->dentry->d_parent != sb->s_root)
5269 ext4_msg(sb, KERN_WARNING,
5270 "Quota file not on filesystem root. "
5271 "Journaled quota will not work");
5275 * When we journal data on quota file, we have to flush journal to see
5276 * all updates to the file when we bypass pagecache...
5278 if (EXT4_SB(sb)->s_journal &&
5279 ext4_should_journal_data(d_inode(path->dentry))) {
5281 * We don't need to lock updates but journal_flush() could
5282 * otherwise be livelocked...
5284 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
5285 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
5286 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
5290 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
5291 err = dquot_quota_on(sb, type, format_id, path);
5293 lockdep_set_quota_inode(path->dentry->d_inode,
5298 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5302 struct inode *qf_inode;
5303 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5304 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5305 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5306 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
5309 BUG_ON(!ext4_has_feature_quota(sb));
5311 if (!qf_inums[type])
5314 qf_inode = ext4_iget(sb, qf_inums[type]);
5315 if (IS_ERR(qf_inode)) {
5316 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5317 return PTR_ERR(qf_inode);
5320 /* Don't account quota for quota files to avoid recursion */
5321 qf_inode->i_flags |= S_NOQUOTA;
5322 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
5323 err = dquot_enable(qf_inode, type, format_id, flags);
5326 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
5331 /* Enable usage tracking for all quota types. */
5332 static int ext4_enable_quotas(struct super_block *sb)
5335 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5336 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5337 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5338 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
5340 bool quota_mopt[EXT4_MAXQUOTAS] = {
5341 test_opt(sb, USRQUOTA),
5342 test_opt(sb, GRPQUOTA),
5343 test_opt(sb, PRJQUOTA),
5346 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5347 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
5348 if (qf_inums[type]) {
5349 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5350 DQUOT_USAGE_ENABLED |
5351 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
5354 "Failed to enable quota tracking "
5355 "(type=%d, err=%d). Please run "
5356 "e2fsck to fix.", type, err);
5364 static int ext4_quota_off(struct super_block *sb, int type)
5366 struct inode *inode = sb_dqopt(sb)->files[type];
5369 /* Force all delayed allocation blocks to be allocated.
5370 * Caller already holds s_umount sem */
5371 if (test_opt(sb, DELALLOC))
5372 sync_filesystem(sb);
5377 /* Update modification times of quota files when userspace can
5378 * start looking at them */
5379 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5382 inode->i_mtime = inode->i_ctime = current_time(inode);
5383 ext4_mark_inode_dirty(handle, inode);
5384 ext4_journal_stop(handle);
5387 return dquot_quota_off(sb, type);
5390 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5391 * acquiring the locks... As quota files are never truncated and quota code
5392 * itself serializes the operations (and no one else should touch the files)
5393 * we don't have to be afraid of races */
5394 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5395 size_t len, loff_t off)
5397 struct inode *inode = sb_dqopt(sb)->files[type];
5398 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5399 int offset = off & (sb->s_blocksize - 1);
5402 struct buffer_head *bh;
5403 loff_t i_size = i_size_read(inode);
5407 if (off+len > i_size)
5410 while (toread > 0) {
5411 tocopy = sb->s_blocksize - offset < toread ?
5412 sb->s_blocksize - offset : toread;
5413 bh = ext4_bread(NULL, inode, blk, 0);
5416 if (!bh) /* A hole? */
5417 memset(data, 0, tocopy);
5419 memcpy(data, bh->b_data+offset, tocopy);
5429 /* Write to quotafile (we know the transaction is already started and has
5430 * enough credits) */
5431 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5432 const char *data, size_t len, loff_t off)
5434 struct inode *inode = sb_dqopt(sb)->files[type];
5435 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5436 int err, offset = off & (sb->s_blocksize - 1);
5438 struct buffer_head *bh;
5439 handle_t *handle = journal_current_handle();
5441 if (EXT4_SB(sb)->s_journal && !handle) {
5442 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5443 " cancelled because transaction is not started",
5444 (unsigned long long)off, (unsigned long long)len);
5448 * Since we account only one data block in transaction credits,
5449 * then it is impossible to cross a block boundary.
5451 if (sb->s_blocksize - offset < len) {
5452 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5453 " cancelled because not block aligned",
5454 (unsigned long long)off, (unsigned long long)len);
5459 bh = ext4_bread(handle, inode, blk,
5460 EXT4_GET_BLOCKS_CREATE |
5461 EXT4_GET_BLOCKS_METADATA_NOFAIL);
5462 } while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) &&
5463 ext4_should_retry_alloc(inode->i_sb, &retries));
5468 BUFFER_TRACE(bh, "get write access");
5469 err = ext4_journal_get_write_access(handle, bh);
5475 memcpy(bh->b_data+offset, data, len);
5476 flush_dcache_page(bh->b_page);
5478 err = ext4_handle_dirty_metadata(handle, NULL, bh);
5481 if (inode->i_size < off + len) {
5482 i_size_write(inode, off + len);
5483 EXT4_I(inode)->i_disksize = inode->i_size;
5484 ext4_mark_inode_dirty(handle, inode);
5489 static int ext4_get_next_id(struct super_block *sb, struct kqid *qid)
5491 const struct quota_format_ops *ops;
5493 if (!sb_has_quota_loaded(sb, qid->type))
5495 ops = sb_dqopt(sb)->ops[qid->type];
5496 if (!ops || !ops->get_next_id)
5498 return dquot_get_next_id(sb, qid);
5502 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5503 const char *dev_name, void *data)
5505 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5508 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
5509 static inline void register_as_ext2(void)
5511 int err = register_filesystem(&ext2_fs_type);
5514 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5517 static inline void unregister_as_ext2(void)
5519 unregister_filesystem(&ext2_fs_type);
5522 static inline int ext2_feature_set_ok(struct super_block *sb)
5524 if (ext4_has_unknown_ext2_incompat_features(sb))
5526 if (sb->s_flags & MS_RDONLY)
5528 if (ext4_has_unknown_ext2_ro_compat_features(sb))
5533 static inline void register_as_ext2(void) { }
5534 static inline void unregister_as_ext2(void) { }
5535 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5538 static inline void register_as_ext3(void)
5540 int err = register_filesystem(&ext3_fs_type);
5543 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5546 static inline void unregister_as_ext3(void)
5548 unregister_filesystem(&ext3_fs_type);
5551 static inline int ext3_feature_set_ok(struct super_block *sb)
5553 if (ext4_has_unknown_ext3_incompat_features(sb))
5555 if (!ext4_has_feature_journal(sb))
5557 if (sb->s_flags & MS_RDONLY)
5559 if (ext4_has_unknown_ext3_ro_compat_features(sb))
5564 static struct file_system_type ext4_fs_type = {
5565 .owner = THIS_MODULE,
5567 .mount = ext4_mount,
5568 .kill_sb = kill_block_super,
5569 .fs_flags = FS_REQUIRES_DEV,
5571 MODULE_ALIAS_FS("ext4");
5573 /* Shared across all ext4 file systems */
5574 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5576 static int __init ext4_init_fs(void)
5580 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
5581 ext4_li_info = NULL;
5582 mutex_init(&ext4_li_mtx);
5584 /* Build-time check for flags consistency */
5585 ext4_check_flag_values();
5587 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
5588 init_waitqueue_head(&ext4__ioend_wq[i]);
5590 err = ext4_init_es();
5594 err = ext4_init_pageio();
5598 err = ext4_init_system_zone();
5602 err = ext4_init_sysfs();
5606 err = ext4_init_mballoc();
5609 err = init_inodecache();
5614 err = register_filesystem(&ext4_fs_type);
5620 unregister_as_ext2();
5621 unregister_as_ext3();
5622 destroy_inodecache();
5624 ext4_exit_mballoc();
5628 ext4_exit_system_zone();
5637 static void __exit ext4_exit_fs(void)
5639 ext4_destroy_lazyinit_thread();
5640 unregister_as_ext2();
5641 unregister_as_ext3();
5642 unregister_filesystem(&ext4_fs_type);
5643 destroy_inodecache();
5644 ext4_exit_mballoc();
5646 ext4_exit_system_zone();
5651 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5652 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5653 MODULE_LICENSE("GPL");
5654 module_init(ext4_init_fs)
5655 module_exit(ext4_exit_fs)