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, ...)
602 save_error_info(sb, function, line);
604 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
610 if ((sb->s_flags & MS_RDONLY) == 0) {
611 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
612 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
614 * Make sure updated value of ->s_mount_flags will be visible
615 * before ->s_flags update
618 sb->s_flags |= MS_RDONLY;
619 if (EXT4_SB(sb)->s_journal)
620 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
621 save_error_info(sb, function, line);
623 if (test_opt(sb, ERRORS_PANIC)) {
624 if (EXT4_SB(sb)->s_journal &&
625 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
627 panic("EXT4-fs panic from previous error\n");
631 void __ext4_msg(struct super_block *sb,
632 const char *prefix, const char *fmt, ...)
634 struct va_format vaf;
637 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
643 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
647 #define ext4_warning_ratelimit(sb) \
648 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \
651 void __ext4_warning(struct super_block *sb, const char *function,
652 unsigned int line, const char *fmt, ...)
654 struct va_format vaf;
657 if (!ext4_warning_ratelimit(sb))
663 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
664 sb->s_id, function, line, &vaf);
668 void __ext4_warning_inode(const struct inode *inode, const char *function,
669 unsigned int line, const char *fmt, ...)
671 struct va_format vaf;
674 if (!ext4_warning_ratelimit(inode->i_sb))
680 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
681 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
682 function, line, inode->i_ino, current->comm, &vaf);
686 void __ext4_grp_locked_error(const char *function, unsigned int line,
687 struct super_block *sb, ext4_group_t grp,
688 unsigned long ino, ext4_fsblk_t block,
689 const char *fmt, ...)
693 struct va_format vaf;
695 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
697 es->s_last_error_ino = cpu_to_le32(ino);
698 es->s_last_error_block = cpu_to_le64(block);
699 __save_error_info(sb, function, line);
701 if (ext4_error_ratelimit(sb)) {
705 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
706 sb->s_id, function, line, grp);
708 printk(KERN_CONT "inode %lu: ", ino);
710 printk(KERN_CONT "block %llu:",
711 (unsigned long long) block);
712 printk(KERN_CONT "%pV\n", &vaf);
716 if (test_opt(sb, ERRORS_CONT)) {
717 ext4_commit_super(sb, 0);
721 ext4_unlock_group(sb, grp);
722 ext4_handle_error(sb);
724 * We only get here in the ERRORS_RO case; relocking the group
725 * may be dangerous, but nothing bad will happen since the
726 * filesystem will have already been marked read/only and the
727 * journal has been aborted. We return 1 as a hint to callers
728 * who might what to use the return value from
729 * ext4_grp_locked_error() to distinguish between the
730 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
731 * aggressively from the ext4 function in question, with a
732 * more appropriate error code.
734 ext4_lock_group(sb, grp);
738 void ext4_update_dynamic_rev(struct super_block *sb)
740 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
742 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
746 "updating to rev %d because of new feature flag, "
747 "running e2fsck is recommended",
750 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
751 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
752 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
753 /* leave es->s_feature_*compat flags alone */
754 /* es->s_uuid will be set by e2fsck if empty */
757 * The rest of the superblock fields should be zero, and if not it
758 * means they are likely already in use, so leave them alone. We
759 * can leave it up to e2fsck to clean up any inconsistencies there.
764 * Open the external journal device
766 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
768 struct block_device *bdev;
769 char b[BDEVNAME_SIZE];
771 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
777 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
778 __bdevname(dev, b), PTR_ERR(bdev));
783 * Release the journal device
785 static void ext4_blkdev_put(struct block_device *bdev)
787 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
790 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
792 struct block_device *bdev;
793 bdev = sbi->journal_bdev;
795 ext4_blkdev_put(bdev);
796 sbi->journal_bdev = NULL;
800 static inline struct inode *orphan_list_entry(struct list_head *l)
802 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
805 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
809 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
810 le32_to_cpu(sbi->s_es->s_last_orphan));
812 printk(KERN_ERR "sb_info orphan list:\n");
813 list_for_each(l, &sbi->s_orphan) {
814 struct inode *inode = orphan_list_entry(l);
816 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
817 inode->i_sb->s_id, inode->i_ino, inode,
818 inode->i_mode, inode->i_nlink,
823 static void ext4_put_super(struct super_block *sb)
825 struct ext4_sb_info *sbi = EXT4_SB(sb);
826 struct ext4_super_block *es = sbi->s_es;
829 ext4_unregister_li_request(sb);
830 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
832 flush_workqueue(sbi->rsv_conversion_wq);
833 destroy_workqueue(sbi->rsv_conversion_wq);
835 if (sbi->s_journal) {
836 err = jbd2_journal_destroy(sbi->s_journal);
837 sbi->s_journal = NULL;
839 ext4_abort(sb, "Couldn't clean up the journal");
842 ext4_unregister_sysfs(sb);
843 ext4_es_unregister_shrinker(sbi);
844 del_timer_sync(&sbi->s_err_report);
845 ext4_release_system_zone(sb);
847 ext4_ext_release(sb);
849 if (!(sb->s_flags & MS_RDONLY)) {
850 ext4_clear_feature_journal_needs_recovery(sb);
851 es->s_state = cpu_to_le16(sbi->s_mount_state);
853 if (!(sb->s_flags & MS_RDONLY))
854 ext4_commit_super(sb, 1);
856 for (i = 0; i < sbi->s_gdb_count; i++)
857 brelse(sbi->s_group_desc[i]);
858 kvfree(sbi->s_group_desc);
859 kvfree(sbi->s_flex_groups);
860 percpu_counter_destroy(&sbi->s_freeclusters_counter);
861 percpu_counter_destroy(&sbi->s_freeinodes_counter);
862 percpu_counter_destroy(&sbi->s_dirs_counter);
863 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
864 percpu_free_rwsem(&sbi->s_journal_flag_rwsem);
867 for (i = 0; i < EXT4_MAXQUOTAS; i++)
868 kfree(sbi->s_qf_names[i]);
871 /* Debugging code just in case the in-memory inode orphan list
872 * isn't empty. The on-disk one can be non-empty if we've
873 * detected an error and taken the fs readonly, but the
874 * in-memory list had better be clean by this point. */
875 if (!list_empty(&sbi->s_orphan))
876 dump_orphan_list(sb, sbi);
877 J_ASSERT(list_empty(&sbi->s_orphan));
879 sync_blockdev(sb->s_bdev);
880 invalidate_bdev(sb->s_bdev);
881 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
883 * Invalidate the journal device's buffers. We don't want them
884 * floating about in memory - the physical journal device may
885 * hotswapped, and it breaks the `ro-after' testing code.
887 sync_blockdev(sbi->journal_bdev);
888 invalidate_bdev(sbi->journal_bdev);
889 ext4_blkdev_remove(sbi);
891 if (sbi->s_mb_cache) {
892 ext4_xattr_destroy_cache(sbi->s_mb_cache);
893 sbi->s_mb_cache = NULL;
896 kthread_stop(sbi->s_mmp_tsk);
897 sb->s_fs_info = NULL;
899 * Now that we are completely done shutting down the
900 * superblock, we need to actually destroy the kobject.
902 kobject_put(&sbi->s_kobj);
903 wait_for_completion(&sbi->s_kobj_unregister);
904 if (sbi->s_chksum_driver)
905 crypto_free_shash(sbi->s_chksum_driver);
906 kfree(sbi->s_blockgroup_lock);
910 static struct kmem_cache *ext4_inode_cachep;
913 * Called inside transaction, so use GFP_NOFS
915 static struct inode *ext4_alloc_inode(struct super_block *sb)
917 struct ext4_inode_info *ei;
919 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
923 ei->vfs_inode.i_version = 1;
924 spin_lock_init(&ei->i_raw_lock);
925 INIT_LIST_HEAD(&ei->i_prealloc_list);
926 spin_lock_init(&ei->i_prealloc_lock);
927 ext4_es_init_tree(&ei->i_es_tree);
928 rwlock_init(&ei->i_es_lock);
929 INIT_LIST_HEAD(&ei->i_es_list);
932 ei->i_es_shrink_lblk = 0;
933 ei->i_reserved_data_blocks = 0;
934 ei->i_reserved_meta_blocks = 0;
935 ei->i_allocated_meta_blocks = 0;
936 ei->i_da_metadata_calc_len = 0;
937 ei->i_da_metadata_calc_last_lblock = 0;
938 spin_lock_init(&(ei->i_block_reservation_lock));
940 ei->i_reserved_quota = 0;
941 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
944 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
945 spin_lock_init(&ei->i_completed_io_lock);
947 ei->i_datasync_tid = 0;
948 atomic_set(&ei->i_unwritten, 0);
949 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
950 return &ei->vfs_inode;
953 static int ext4_drop_inode(struct inode *inode)
955 int drop = generic_drop_inode(inode);
957 trace_ext4_drop_inode(inode, drop);
961 static void ext4_i_callback(struct rcu_head *head)
963 struct inode *inode = container_of(head, struct inode, i_rcu);
964 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
967 static void ext4_destroy_inode(struct inode *inode)
969 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
970 ext4_msg(inode->i_sb, KERN_ERR,
971 "Inode %lu (%p): orphan list check failed!",
972 inode->i_ino, EXT4_I(inode));
973 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
974 EXT4_I(inode), sizeof(struct ext4_inode_info),
978 call_rcu(&inode->i_rcu, ext4_i_callback);
981 static void init_once(void *foo)
983 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
985 INIT_LIST_HEAD(&ei->i_orphan);
986 init_rwsem(&ei->xattr_sem);
987 init_rwsem(&ei->i_data_sem);
988 init_rwsem(&ei->i_mmap_sem);
989 inode_init_once(&ei->vfs_inode);
992 static int __init init_inodecache(void)
994 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
995 sizeof(struct ext4_inode_info),
996 0, (SLAB_RECLAIM_ACCOUNT|
997 SLAB_MEM_SPREAD|SLAB_ACCOUNT),
999 if (ext4_inode_cachep == NULL)
1004 static void destroy_inodecache(void)
1007 * Make sure all delayed rcu free inodes are flushed before we
1011 kmem_cache_destroy(ext4_inode_cachep);
1014 void ext4_clear_inode(struct inode *inode)
1016 invalidate_inode_buffers(inode);
1019 ext4_discard_preallocations(inode);
1020 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
1021 if (EXT4_I(inode)->jinode) {
1022 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1023 EXT4_I(inode)->jinode);
1024 jbd2_free_inode(EXT4_I(inode)->jinode);
1025 EXT4_I(inode)->jinode = NULL;
1027 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1028 fscrypt_put_encryption_info(inode, NULL);
1032 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1033 u64 ino, u32 generation)
1035 struct inode *inode;
1037 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1038 return ERR_PTR(-ESTALE);
1039 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1040 return ERR_PTR(-ESTALE);
1042 /* iget isn't really right if the inode is currently unallocated!!
1044 * ext4_read_inode will return a bad_inode if the inode had been
1045 * deleted, so we should be safe.
1047 * Currently we don't know the generation for parent directory, so
1048 * a generation of 0 means "accept any"
1050 inode = ext4_iget_normal(sb, ino);
1052 return ERR_CAST(inode);
1053 if (generation && inode->i_generation != generation) {
1055 return ERR_PTR(-ESTALE);
1061 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1062 int fh_len, int fh_type)
1064 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1065 ext4_nfs_get_inode);
1068 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1069 int fh_len, int fh_type)
1071 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1072 ext4_nfs_get_inode);
1076 * Try to release metadata pages (indirect blocks, directories) which are
1077 * mapped via the block device. Since these pages could have journal heads
1078 * which would prevent try_to_free_buffers() from freeing them, we must use
1079 * jbd2 layer's try_to_free_buffers() function to release them.
1081 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1084 journal_t *journal = EXT4_SB(sb)->s_journal;
1086 WARN_ON(PageChecked(page));
1087 if (!page_has_buffers(page))
1090 return jbd2_journal_try_to_free_buffers(journal, page,
1091 wait & ~__GFP_DIRECT_RECLAIM);
1092 return try_to_free_buffers(page);
1095 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1096 static int ext4_get_context(struct inode *inode, void *ctx, size_t len)
1098 return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1099 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len);
1102 static int ext4_key_prefix(struct inode *inode, u8 **key)
1104 *key = EXT4_SB(inode->i_sb)->key_prefix;
1105 return EXT4_SB(inode->i_sb)->key_prefix_size;
1108 static int ext4_prepare_context(struct inode *inode)
1110 return ext4_convert_inline_data(inode);
1113 static int ext4_set_context(struct inode *inode, const void *ctx, size_t len,
1119 /* fs_data is null when internally used. */
1121 res = ext4_xattr_set(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1122 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx,
1125 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1126 ext4_clear_inode_state(inode,
1127 EXT4_STATE_MAY_INLINE_DATA);
1132 handle = ext4_journal_start(inode, EXT4_HT_MISC,
1133 ext4_jbd2_credits_xattr(inode));
1135 return PTR_ERR(handle);
1137 res = ext4_xattr_set(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1138 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx,
1141 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1142 res = ext4_mark_inode_dirty(handle, inode);
1144 EXT4_ERROR_INODE(inode, "Failed to mark inode dirty");
1146 res2 = ext4_journal_stop(handle);
1152 static int ext4_dummy_context(struct inode *inode)
1154 return DUMMY_ENCRYPTION_ENABLED(EXT4_SB(inode->i_sb));
1157 static unsigned ext4_max_namelen(struct inode *inode)
1159 return S_ISLNK(inode->i_mode) ? inode->i_sb->s_blocksize :
1163 static struct fscrypt_operations ext4_cryptops = {
1164 .get_context = ext4_get_context,
1165 .key_prefix = ext4_key_prefix,
1166 .prepare_context = ext4_prepare_context,
1167 .set_context = ext4_set_context,
1168 .dummy_context = ext4_dummy_context,
1169 .is_encrypted = ext4_encrypted_inode,
1170 .empty_dir = ext4_empty_dir,
1171 .max_namelen = ext4_max_namelen,
1174 static struct fscrypt_operations ext4_cryptops = {
1175 .is_encrypted = ext4_encrypted_inode,
1180 static char *quotatypes[] = INITQFNAMES;
1181 #define QTYPE2NAME(t) (quotatypes[t])
1183 static int ext4_write_dquot(struct dquot *dquot);
1184 static int ext4_acquire_dquot(struct dquot *dquot);
1185 static int ext4_release_dquot(struct dquot *dquot);
1186 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1187 static int ext4_write_info(struct super_block *sb, int type);
1188 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1190 static int ext4_quota_off(struct super_block *sb, int type);
1191 static int ext4_quota_on_mount(struct super_block *sb, int type);
1192 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1193 size_t len, loff_t off);
1194 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1195 const char *data, size_t len, loff_t off);
1196 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1197 unsigned int flags);
1198 static int ext4_enable_quotas(struct super_block *sb);
1199 static int ext4_get_next_id(struct super_block *sb, struct kqid *qid);
1201 static struct dquot **ext4_get_dquots(struct inode *inode)
1203 return EXT4_I(inode)->i_dquot;
1206 static const struct dquot_operations ext4_quota_operations = {
1207 .get_reserved_space = ext4_get_reserved_space,
1208 .write_dquot = ext4_write_dquot,
1209 .acquire_dquot = ext4_acquire_dquot,
1210 .release_dquot = ext4_release_dquot,
1211 .mark_dirty = ext4_mark_dquot_dirty,
1212 .write_info = ext4_write_info,
1213 .alloc_dquot = dquot_alloc,
1214 .destroy_dquot = dquot_destroy,
1215 .get_projid = ext4_get_projid,
1216 .get_next_id = ext4_get_next_id,
1219 static const struct quotactl_ops ext4_qctl_operations = {
1220 .quota_on = ext4_quota_on,
1221 .quota_off = ext4_quota_off,
1222 .quota_sync = dquot_quota_sync,
1223 .get_state = dquot_get_state,
1224 .set_info = dquot_set_dqinfo,
1225 .get_dqblk = dquot_get_dqblk,
1226 .set_dqblk = dquot_set_dqblk,
1227 .get_nextdqblk = dquot_get_next_dqblk,
1231 static const struct super_operations ext4_sops = {
1232 .alloc_inode = ext4_alloc_inode,
1233 .destroy_inode = ext4_destroy_inode,
1234 .write_inode = ext4_write_inode,
1235 .dirty_inode = ext4_dirty_inode,
1236 .drop_inode = ext4_drop_inode,
1237 .evict_inode = ext4_evict_inode,
1238 .put_super = ext4_put_super,
1239 .sync_fs = ext4_sync_fs,
1240 .freeze_fs = ext4_freeze,
1241 .unfreeze_fs = ext4_unfreeze,
1242 .statfs = ext4_statfs,
1243 .remount_fs = ext4_remount,
1244 .show_options = ext4_show_options,
1246 .quota_read = ext4_quota_read,
1247 .quota_write = ext4_quota_write,
1248 .get_dquots = ext4_get_dquots,
1250 .bdev_try_to_free_page = bdev_try_to_free_page,
1253 static const struct export_operations ext4_export_ops = {
1254 .fh_to_dentry = ext4_fh_to_dentry,
1255 .fh_to_parent = ext4_fh_to_parent,
1256 .get_parent = ext4_get_parent,
1260 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1261 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1262 Opt_nouid32, Opt_debug, Opt_removed,
1263 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1264 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1265 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1266 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1267 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1268 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1269 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1270 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1271 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1272 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version, Opt_dax,
1273 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1274 Opt_lazytime, Opt_nolazytime,
1275 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1276 Opt_inode_readahead_blks, Opt_journal_ioprio,
1277 Opt_dioread_nolock, Opt_dioread_lock,
1278 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1279 Opt_max_dir_size_kb, Opt_nojournal_checksum,
1282 static const match_table_t tokens = {
1283 {Opt_bsd_df, "bsddf"},
1284 {Opt_minix_df, "minixdf"},
1285 {Opt_grpid, "grpid"},
1286 {Opt_grpid, "bsdgroups"},
1287 {Opt_nogrpid, "nogrpid"},
1288 {Opt_nogrpid, "sysvgroups"},
1289 {Opt_resgid, "resgid=%u"},
1290 {Opt_resuid, "resuid=%u"},
1292 {Opt_err_cont, "errors=continue"},
1293 {Opt_err_panic, "errors=panic"},
1294 {Opt_err_ro, "errors=remount-ro"},
1295 {Opt_nouid32, "nouid32"},
1296 {Opt_debug, "debug"},
1297 {Opt_removed, "oldalloc"},
1298 {Opt_removed, "orlov"},
1299 {Opt_user_xattr, "user_xattr"},
1300 {Opt_nouser_xattr, "nouser_xattr"},
1302 {Opt_noacl, "noacl"},
1303 {Opt_noload, "norecovery"},
1304 {Opt_noload, "noload"},
1305 {Opt_removed, "nobh"},
1306 {Opt_removed, "bh"},
1307 {Opt_commit, "commit=%u"},
1308 {Opt_min_batch_time, "min_batch_time=%u"},
1309 {Opt_max_batch_time, "max_batch_time=%u"},
1310 {Opt_journal_dev, "journal_dev=%u"},
1311 {Opt_journal_path, "journal_path=%s"},
1312 {Opt_journal_checksum, "journal_checksum"},
1313 {Opt_nojournal_checksum, "nojournal_checksum"},
1314 {Opt_journal_async_commit, "journal_async_commit"},
1315 {Opt_abort, "abort"},
1316 {Opt_data_journal, "data=journal"},
1317 {Opt_data_ordered, "data=ordered"},
1318 {Opt_data_writeback, "data=writeback"},
1319 {Opt_data_err_abort, "data_err=abort"},
1320 {Opt_data_err_ignore, "data_err=ignore"},
1321 {Opt_offusrjquota, "usrjquota="},
1322 {Opt_usrjquota, "usrjquota=%s"},
1323 {Opt_offgrpjquota, "grpjquota="},
1324 {Opt_grpjquota, "grpjquota=%s"},
1325 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1326 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1327 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1328 {Opt_grpquota, "grpquota"},
1329 {Opt_noquota, "noquota"},
1330 {Opt_quota, "quota"},
1331 {Opt_usrquota, "usrquota"},
1332 {Opt_prjquota, "prjquota"},
1333 {Opt_barrier, "barrier=%u"},
1334 {Opt_barrier, "barrier"},
1335 {Opt_nobarrier, "nobarrier"},
1336 {Opt_i_version, "i_version"},
1338 {Opt_stripe, "stripe=%u"},
1339 {Opt_delalloc, "delalloc"},
1340 {Opt_lazytime, "lazytime"},
1341 {Opt_nolazytime, "nolazytime"},
1342 {Opt_nodelalloc, "nodelalloc"},
1343 {Opt_removed, "mblk_io_submit"},
1344 {Opt_removed, "nomblk_io_submit"},
1345 {Opt_block_validity, "block_validity"},
1346 {Opt_noblock_validity, "noblock_validity"},
1347 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1348 {Opt_journal_ioprio, "journal_ioprio=%u"},
1349 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1350 {Opt_auto_da_alloc, "auto_da_alloc"},
1351 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1352 {Opt_dioread_nolock, "dioread_nolock"},
1353 {Opt_dioread_lock, "dioread_lock"},
1354 {Opt_discard, "discard"},
1355 {Opt_nodiscard, "nodiscard"},
1356 {Opt_init_itable, "init_itable=%u"},
1357 {Opt_init_itable, "init_itable"},
1358 {Opt_noinit_itable, "noinit_itable"},
1359 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1360 {Opt_test_dummy_encryption, "test_dummy_encryption"},
1361 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1362 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1363 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1364 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1365 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1369 static ext4_fsblk_t get_sb_block(void **data)
1371 ext4_fsblk_t sb_block;
1372 char *options = (char *) *data;
1374 if (!options || strncmp(options, "sb=", 3) != 0)
1375 return 1; /* Default location */
1378 /* TODO: use simple_strtoll with >32bit ext4 */
1379 sb_block = simple_strtoul(options, &options, 0);
1380 if (*options && *options != ',') {
1381 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1385 if (*options == ',')
1387 *data = (void *) options;
1392 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1393 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1394 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1397 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1399 struct ext4_sb_info *sbi = EXT4_SB(sb);
1403 if (sb_any_quota_loaded(sb) &&
1404 !sbi->s_qf_names[qtype]) {
1405 ext4_msg(sb, KERN_ERR,
1406 "Cannot change journaled "
1407 "quota options when quota turned on");
1410 if (ext4_has_feature_quota(sb)) {
1411 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1412 "ignored when QUOTA feature is enabled");
1415 qname = match_strdup(args);
1417 ext4_msg(sb, KERN_ERR,
1418 "Not enough memory for storing quotafile name");
1421 if (sbi->s_qf_names[qtype]) {
1422 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1425 ext4_msg(sb, KERN_ERR,
1426 "%s quota file already specified",
1430 if (strchr(qname, '/')) {
1431 ext4_msg(sb, KERN_ERR,
1432 "quotafile must be on filesystem root");
1435 sbi->s_qf_names[qtype] = qname;
1443 static int clear_qf_name(struct super_block *sb, int qtype)
1446 struct ext4_sb_info *sbi = EXT4_SB(sb);
1448 if (sb_any_quota_loaded(sb) &&
1449 sbi->s_qf_names[qtype]) {
1450 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1451 " when quota turned on");
1454 kfree(sbi->s_qf_names[qtype]);
1455 sbi->s_qf_names[qtype] = NULL;
1460 #define MOPT_SET 0x0001
1461 #define MOPT_CLEAR 0x0002
1462 #define MOPT_NOSUPPORT 0x0004
1463 #define MOPT_EXPLICIT 0x0008
1464 #define MOPT_CLEAR_ERR 0x0010
1465 #define MOPT_GTE0 0x0020
1468 #define MOPT_QFMT 0x0040
1470 #define MOPT_Q MOPT_NOSUPPORT
1471 #define MOPT_QFMT MOPT_NOSUPPORT
1473 #define MOPT_DATAJ 0x0080
1474 #define MOPT_NO_EXT2 0x0100
1475 #define MOPT_NO_EXT3 0x0200
1476 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1477 #define MOPT_STRING 0x0400
1479 static const struct mount_opts {
1483 } ext4_mount_opts[] = {
1484 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1485 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1486 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1487 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1488 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1489 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1490 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1491 MOPT_EXT4_ONLY | MOPT_SET},
1492 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1493 MOPT_EXT4_ONLY | MOPT_CLEAR},
1494 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1495 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1496 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1497 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1498 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1499 MOPT_EXT4_ONLY | MOPT_CLEAR},
1500 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1501 MOPT_EXT4_ONLY | MOPT_CLEAR},
1502 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1503 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1504 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1505 EXT4_MOUNT_JOURNAL_CHECKSUM),
1506 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1507 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1508 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1509 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1510 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1511 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1513 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1515 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1516 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1517 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1518 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1519 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1520 {Opt_commit, 0, MOPT_GTE0},
1521 {Opt_max_batch_time, 0, MOPT_GTE0},
1522 {Opt_min_batch_time, 0, MOPT_GTE0},
1523 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1524 {Opt_init_itable, 0, MOPT_GTE0},
1525 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
1526 {Opt_stripe, 0, MOPT_GTE0},
1527 {Opt_resuid, 0, MOPT_GTE0},
1528 {Opt_resgid, 0, MOPT_GTE0},
1529 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1530 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1531 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1532 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1533 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1534 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1535 MOPT_NO_EXT2 | MOPT_DATAJ},
1536 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1537 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1538 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1539 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1540 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1542 {Opt_acl, 0, MOPT_NOSUPPORT},
1543 {Opt_noacl, 0, MOPT_NOSUPPORT},
1545 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1546 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1547 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1548 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1550 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1552 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1554 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1555 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1556 MOPT_CLEAR | MOPT_Q},
1557 {Opt_usrjquota, 0, MOPT_Q},
1558 {Opt_grpjquota, 0, MOPT_Q},
1559 {Opt_offusrjquota, 0, MOPT_Q},
1560 {Opt_offgrpjquota, 0, MOPT_Q},
1561 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1562 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1563 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1564 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1565 {Opt_test_dummy_encryption, 0, MOPT_GTE0},
1569 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1570 substring_t *args, unsigned long *journal_devnum,
1571 unsigned int *journal_ioprio, int is_remount)
1573 struct ext4_sb_info *sbi = EXT4_SB(sb);
1574 const struct mount_opts *m;
1580 if (token == Opt_usrjquota)
1581 return set_qf_name(sb, USRQUOTA, &args[0]);
1582 else if (token == Opt_grpjquota)
1583 return set_qf_name(sb, GRPQUOTA, &args[0]);
1584 else if (token == Opt_offusrjquota)
1585 return clear_qf_name(sb, USRQUOTA);
1586 else if (token == Opt_offgrpjquota)
1587 return clear_qf_name(sb, GRPQUOTA);
1591 case Opt_nouser_xattr:
1592 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1595 return 1; /* handled by get_sb_block() */
1597 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1600 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1603 sb->s_flags |= MS_I_VERSION;
1606 sb->s_flags |= MS_LAZYTIME;
1608 case Opt_nolazytime:
1609 sb->s_flags &= ~MS_LAZYTIME;
1613 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1614 if (token == m->token)
1617 if (m->token == Opt_err) {
1618 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1619 "or missing value", opt);
1623 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1624 ext4_msg(sb, KERN_ERR,
1625 "Mount option \"%s\" incompatible with ext2", opt);
1628 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1629 ext4_msg(sb, KERN_ERR,
1630 "Mount option \"%s\" incompatible with ext3", opt);
1634 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
1636 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1638 if (m->flags & MOPT_EXPLICIT) {
1639 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
1640 set_opt2(sb, EXPLICIT_DELALLOC);
1641 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
1642 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
1646 if (m->flags & MOPT_CLEAR_ERR)
1647 clear_opt(sb, ERRORS_MASK);
1648 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1649 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1650 "options when quota turned on");
1654 if (m->flags & MOPT_NOSUPPORT) {
1655 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1656 } else if (token == Opt_commit) {
1658 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1659 sbi->s_commit_interval = HZ * arg;
1660 } else if (token == Opt_max_batch_time) {
1661 sbi->s_max_batch_time = arg;
1662 } else if (token == Opt_min_batch_time) {
1663 sbi->s_min_batch_time = arg;
1664 } else if (token == Opt_inode_readahead_blks) {
1665 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1666 ext4_msg(sb, KERN_ERR,
1667 "EXT4-fs: inode_readahead_blks must be "
1668 "0 or a power of 2 smaller than 2^31");
1671 sbi->s_inode_readahead_blks = arg;
1672 } else if (token == Opt_init_itable) {
1673 set_opt(sb, INIT_INODE_TABLE);
1675 arg = EXT4_DEF_LI_WAIT_MULT;
1676 sbi->s_li_wait_mult = arg;
1677 } else if (token == Opt_max_dir_size_kb) {
1678 sbi->s_max_dir_size_kb = arg;
1679 } else if (token == Opt_stripe) {
1680 sbi->s_stripe = arg;
1681 } else if (token == Opt_resuid) {
1682 uid = make_kuid(current_user_ns(), arg);
1683 if (!uid_valid(uid)) {
1684 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1687 sbi->s_resuid = uid;
1688 } else if (token == Opt_resgid) {
1689 gid = make_kgid(current_user_ns(), arg);
1690 if (!gid_valid(gid)) {
1691 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1694 sbi->s_resgid = gid;
1695 } else if (token == Opt_journal_dev) {
1697 ext4_msg(sb, KERN_ERR,
1698 "Cannot specify journal on remount");
1701 *journal_devnum = arg;
1702 } else if (token == Opt_journal_path) {
1704 struct inode *journal_inode;
1709 ext4_msg(sb, KERN_ERR,
1710 "Cannot specify journal on remount");
1713 journal_path = match_strdup(&args[0]);
1714 if (!journal_path) {
1715 ext4_msg(sb, KERN_ERR, "error: could not dup "
1716 "journal device string");
1720 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1722 ext4_msg(sb, KERN_ERR, "error: could not find "
1723 "journal device path: error %d", error);
1724 kfree(journal_path);
1728 journal_inode = d_inode(path.dentry);
1729 if (!S_ISBLK(journal_inode->i_mode)) {
1730 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1731 "is not a block device", journal_path);
1733 kfree(journal_path);
1737 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1739 kfree(journal_path);
1740 } else if (token == Opt_journal_ioprio) {
1742 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1747 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1748 } else if (token == Opt_test_dummy_encryption) {
1749 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1750 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1751 ext4_msg(sb, KERN_WARNING,
1752 "Test dummy encryption mode enabled");
1754 ext4_msg(sb, KERN_WARNING,
1755 "Test dummy encryption mount option ignored");
1757 } else if (m->flags & MOPT_DATAJ) {
1759 if (!sbi->s_journal)
1760 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1761 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1762 ext4_msg(sb, KERN_ERR,
1763 "Cannot change data mode on remount");
1767 clear_opt(sb, DATA_FLAGS);
1768 sbi->s_mount_opt |= m->mount_opt;
1771 } else if (m->flags & MOPT_QFMT) {
1772 if (sb_any_quota_loaded(sb) &&
1773 sbi->s_jquota_fmt != m->mount_opt) {
1774 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1775 "quota options when quota turned on");
1778 if (ext4_has_feature_quota(sb)) {
1779 ext4_msg(sb, KERN_INFO,
1780 "Quota format mount options ignored "
1781 "when QUOTA feature is enabled");
1784 sbi->s_jquota_fmt = m->mount_opt;
1786 } else if (token == Opt_dax) {
1787 #ifdef CONFIG_FS_DAX
1788 ext4_msg(sb, KERN_WARNING,
1789 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
1790 sbi->s_mount_opt |= m->mount_opt;
1792 ext4_msg(sb, KERN_INFO, "dax option not supported");
1795 } else if (token == Opt_data_err_abort) {
1796 sbi->s_mount_opt |= m->mount_opt;
1797 } else if (token == Opt_data_err_ignore) {
1798 sbi->s_mount_opt &= ~m->mount_opt;
1802 if (m->flags & MOPT_CLEAR)
1804 else if (unlikely(!(m->flags & MOPT_SET))) {
1805 ext4_msg(sb, KERN_WARNING,
1806 "buggy handling of option %s", opt);
1811 sbi->s_mount_opt |= m->mount_opt;
1813 sbi->s_mount_opt &= ~m->mount_opt;
1818 static int parse_options(char *options, struct super_block *sb,
1819 unsigned long *journal_devnum,
1820 unsigned int *journal_ioprio,
1823 struct ext4_sb_info *sbi = EXT4_SB(sb);
1825 substring_t args[MAX_OPT_ARGS];
1831 while ((p = strsep(&options, ",")) != NULL) {
1835 * Initialize args struct so we know whether arg was
1836 * found; some options take optional arguments.
1838 args[0].to = args[0].from = NULL;
1839 token = match_token(p, tokens, args);
1840 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1841 journal_ioprio, is_remount) < 0)
1846 * We do the test below only for project quotas. 'usrquota' and
1847 * 'grpquota' mount options are allowed even without quota feature
1848 * to support legacy quotas in quota files.
1850 if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) {
1851 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. "
1852 "Cannot enable project quota enforcement.");
1855 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1856 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1857 clear_opt(sb, USRQUOTA);
1859 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1860 clear_opt(sb, GRPQUOTA);
1862 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1863 ext4_msg(sb, KERN_ERR, "old and new quota "
1868 if (!sbi->s_jquota_fmt) {
1869 ext4_msg(sb, KERN_ERR, "journaled quota format "
1875 if (test_opt(sb, DIOREAD_NOLOCK)) {
1877 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1879 if (blocksize < PAGE_SIZE) {
1880 ext4_msg(sb, KERN_ERR, "can't mount with "
1881 "dioread_nolock if block size != PAGE_SIZE");
1885 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
1886 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
1887 ext4_msg(sb, KERN_ERR, "can't mount with journal_async_commit "
1888 "in data=ordered mode");
1894 static inline void ext4_show_quota_options(struct seq_file *seq,
1895 struct super_block *sb)
1897 #if defined(CONFIG_QUOTA)
1898 struct ext4_sb_info *sbi = EXT4_SB(sb);
1900 if (sbi->s_jquota_fmt) {
1903 switch (sbi->s_jquota_fmt) {
1914 seq_printf(seq, ",jqfmt=%s", fmtname);
1917 if (sbi->s_qf_names[USRQUOTA])
1918 seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]);
1920 if (sbi->s_qf_names[GRPQUOTA])
1921 seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]);
1925 static const char *token2str(int token)
1927 const struct match_token *t;
1929 for (t = tokens; t->token != Opt_err; t++)
1930 if (t->token == token && !strchr(t->pattern, '='))
1937 * - it's set to a non-default value OR
1938 * - if the per-sb default is different from the global default
1940 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1943 struct ext4_sb_info *sbi = EXT4_SB(sb);
1944 struct ext4_super_block *es = sbi->s_es;
1945 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1946 const struct mount_opts *m;
1947 char sep = nodefs ? '\n' : ',';
1949 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1950 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1952 if (sbi->s_sb_block != 1)
1953 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1955 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1956 int want_set = m->flags & MOPT_SET;
1957 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1958 (m->flags & MOPT_CLEAR_ERR))
1960 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
1961 continue; /* skip if same as the default */
1963 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1964 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1965 continue; /* select Opt_noFoo vs Opt_Foo */
1966 SEQ_OPTS_PRINT("%s", token2str(m->token));
1969 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
1970 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1971 SEQ_OPTS_PRINT("resuid=%u",
1972 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1973 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
1974 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1975 SEQ_OPTS_PRINT("resgid=%u",
1976 from_kgid_munged(&init_user_ns, sbi->s_resgid));
1977 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
1978 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1979 SEQ_OPTS_PUTS("errors=remount-ro");
1980 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1981 SEQ_OPTS_PUTS("errors=continue");
1982 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1983 SEQ_OPTS_PUTS("errors=panic");
1984 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
1985 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
1986 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
1987 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
1988 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
1989 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
1990 if (sb->s_flags & MS_I_VERSION)
1991 SEQ_OPTS_PUTS("i_version");
1992 if (nodefs || sbi->s_stripe)
1993 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
1994 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
1995 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1996 SEQ_OPTS_PUTS("data=journal");
1997 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1998 SEQ_OPTS_PUTS("data=ordered");
1999 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2000 SEQ_OPTS_PUTS("data=writeback");
2003 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
2004 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2005 sbi->s_inode_readahead_blks);
2007 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
2008 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
2009 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
2010 if (nodefs || sbi->s_max_dir_size_kb)
2011 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2012 if (test_opt(sb, DATA_ERR_ABORT))
2013 SEQ_OPTS_PUTS("data_err=abort");
2015 ext4_show_quota_options(seq, sb);
2019 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2021 return _ext4_show_options(seq, root->d_sb, 0);
2024 int ext4_seq_options_show(struct seq_file *seq, void *offset)
2026 struct super_block *sb = seq->private;
2029 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
2030 rc = _ext4_show_options(seq, sb, 1);
2031 seq_puts(seq, "\n");
2035 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
2038 struct ext4_sb_info *sbi = EXT4_SB(sb);
2041 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
2042 ext4_msg(sb, KERN_ERR, "revision level too high, "
2043 "forcing read-only mode");
2048 if (!(sbi->s_mount_state & EXT4_VALID_FS))
2049 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
2050 "running e2fsck is recommended");
2051 else if (sbi->s_mount_state & EXT4_ERROR_FS)
2052 ext4_msg(sb, KERN_WARNING,
2053 "warning: mounting fs with errors, "
2054 "running e2fsck is recommended");
2055 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
2056 le16_to_cpu(es->s_mnt_count) >=
2057 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
2058 ext4_msg(sb, KERN_WARNING,
2059 "warning: maximal mount count reached, "
2060 "running e2fsck is recommended");
2061 else if (le32_to_cpu(es->s_checkinterval) &&
2062 (le32_to_cpu(es->s_lastcheck) +
2063 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
2064 ext4_msg(sb, KERN_WARNING,
2065 "warning: checktime reached, "
2066 "running e2fsck is recommended");
2067 if (!sbi->s_journal)
2068 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
2069 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
2070 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
2071 le16_add_cpu(&es->s_mnt_count, 1);
2072 es->s_mtime = cpu_to_le32(get_seconds());
2073 ext4_update_dynamic_rev(sb);
2075 ext4_set_feature_journal_needs_recovery(sb);
2077 ext4_commit_super(sb, 1);
2079 if (test_opt(sb, DEBUG))
2080 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
2081 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
2083 sbi->s_groups_count,
2084 EXT4_BLOCKS_PER_GROUP(sb),
2085 EXT4_INODES_PER_GROUP(sb),
2086 sbi->s_mount_opt, sbi->s_mount_opt2);
2088 cleancache_init_fs(sb);
2092 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2094 struct ext4_sb_info *sbi = EXT4_SB(sb);
2095 struct flex_groups *new_groups;
2098 if (!sbi->s_log_groups_per_flex)
2101 size = ext4_flex_group(sbi, ngroup - 1) + 1;
2102 if (size <= sbi->s_flex_groups_allocated)
2105 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
2106 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
2108 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
2109 size / (int) sizeof(struct flex_groups));
2113 if (sbi->s_flex_groups) {
2114 memcpy(new_groups, sbi->s_flex_groups,
2115 (sbi->s_flex_groups_allocated *
2116 sizeof(struct flex_groups)));
2117 kvfree(sbi->s_flex_groups);
2119 sbi->s_flex_groups = new_groups;
2120 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
2124 static int ext4_fill_flex_info(struct super_block *sb)
2126 struct ext4_sb_info *sbi = EXT4_SB(sb);
2127 struct ext4_group_desc *gdp = NULL;
2128 ext4_group_t flex_group;
2131 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2132 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
2133 sbi->s_log_groups_per_flex = 0;
2137 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2141 for (i = 0; i < sbi->s_groups_count; i++) {
2142 gdp = ext4_get_group_desc(sb, i, NULL);
2144 flex_group = ext4_flex_group(sbi, i);
2145 atomic_add(ext4_free_inodes_count(sb, gdp),
2146 &sbi->s_flex_groups[flex_group].free_inodes);
2147 atomic64_add(ext4_free_group_clusters(sb, gdp),
2148 &sbi->s_flex_groups[flex_group].free_clusters);
2149 atomic_add(ext4_used_dirs_count(sb, gdp),
2150 &sbi->s_flex_groups[flex_group].used_dirs);
2158 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
2159 struct ext4_group_desc *gdp)
2161 int offset = offsetof(struct ext4_group_desc, bg_checksum);
2163 __le32 le_group = cpu_to_le32(block_group);
2164 struct ext4_sb_info *sbi = EXT4_SB(sb);
2166 if (ext4_has_metadata_csum(sbi->s_sb)) {
2167 /* Use new metadata_csum algorithm */
2169 __u16 dummy_csum = 0;
2171 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2173 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2174 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2175 sizeof(dummy_csum));
2176 offset += sizeof(dummy_csum);
2177 if (offset < sbi->s_desc_size)
2178 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2179 sbi->s_desc_size - offset);
2181 crc = csum32 & 0xFFFF;
2185 /* old crc16 code */
2186 if (!ext4_has_feature_gdt_csum(sb))
2189 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2190 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2191 crc = crc16(crc, (__u8 *)gdp, offset);
2192 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2193 /* for checksum of struct ext4_group_desc do the rest...*/
2194 if (ext4_has_feature_64bit(sb) &&
2195 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2196 crc = crc16(crc, (__u8 *)gdp + offset,
2197 le16_to_cpu(sbi->s_es->s_desc_size) -
2201 return cpu_to_le16(crc);
2204 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2205 struct ext4_group_desc *gdp)
2207 if (ext4_has_group_desc_csum(sb) &&
2208 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
2214 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2215 struct ext4_group_desc *gdp)
2217 if (!ext4_has_group_desc_csum(sb))
2219 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
2222 /* Called at mount-time, super-block is locked */
2223 static int ext4_check_descriptors(struct super_block *sb,
2224 ext4_fsblk_t sb_block,
2225 ext4_group_t *first_not_zeroed)
2227 struct ext4_sb_info *sbi = EXT4_SB(sb);
2228 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2229 ext4_fsblk_t last_block;
2230 ext4_fsblk_t block_bitmap;
2231 ext4_fsblk_t inode_bitmap;
2232 ext4_fsblk_t inode_table;
2233 int flexbg_flag = 0;
2234 ext4_group_t i, grp = sbi->s_groups_count;
2236 if (ext4_has_feature_flex_bg(sb))
2239 ext4_debug("Checking group descriptors");
2241 for (i = 0; i < sbi->s_groups_count; i++) {
2242 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2244 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2245 last_block = ext4_blocks_count(sbi->s_es) - 1;
2247 last_block = first_block +
2248 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2250 if ((grp == sbi->s_groups_count) &&
2251 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2254 block_bitmap = ext4_block_bitmap(sb, gdp);
2255 if (block_bitmap == sb_block) {
2256 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2257 "Block bitmap for group %u overlaps "
2260 if (block_bitmap < first_block || block_bitmap > last_block) {
2261 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2262 "Block bitmap for group %u not in group "
2263 "(block %llu)!", i, block_bitmap);
2266 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2267 if (inode_bitmap == sb_block) {
2268 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2269 "Inode bitmap for group %u overlaps "
2272 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2273 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2274 "Inode bitmap for group %u not in group "
2275 "(block %llu)!", i, inode_bitmap);
2278 inode_table = ext4_inode_table(sb, gdp);
2279 if (inode_table == sb_block) {
2280 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2281 "Inode table for group %u overlaps "
2284 if (inode_table < first_block ||
2285 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2286 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2287 "Inode table for group %u not in group "
2288 "(block %llu)!", i, inode_table);
2291 ext4_lock_group(sb, i);
2292 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2293 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2294 "Checksum for group %u failed (%u!=%u)",
2295 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
2296 gdp)), le16_to_cpu(gdp->bg_checksum));
2297 if (!(sb->s_flags & MS_RDONLY)) {
2298 ext4_unlock_group(sb, i);
2302 ext4_unlock_group(sb, i);
2304 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2306 if (NULL != first_not_zeroed)
2307 *first_not_zeroed = grp;
2311 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2312 * the superblock) which were deleted from all directories, but held open by
2313 * a process at the time of a crash. We walk the list and try to delete these
2314 * inodes at recovery time (only with a read-write filesystem).
2316 * In order to keep the orphan inode chain consistent during traversal (in
2317 * case of crash during recovery), we link each inode into the superblock
2318 * orphan list_head and handle it the same way as an inode deletion during
2319 * normal operation (which journals the operations for us).
2321 * We only do an iget() and an iput() on each inode, which is very safe if we
2322 * accidentally point at an in-use or already deleted inode. The worst that
2323 * can happen in this case is that we get a "bit already cleared" message from
2324 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2325 * e2fsck was run on this filesystem, and it must have already done the orphan
2326 * inode cleanup for us, so we can safely abort without any further action.
2328 static void ext4_orphan_cleanup(struct super_block *sb,
2329 struct ext4_super_block *es)
2331 unsigned int s_flags = sb->s_flags;
2332 int nr_orphans = 0, nr_truncates = 0;
2336 if (!es->s_last_orphan) {
2337 jbd_debug(4, "no orphan inodes to clean up\n");
2341 if (bdev_read_only(sb->s_bdev)) {
2342 ext4_msg(sb, KERN_ERR, "write access "
2343 "unavailable, skipping orphan cleanup");
2347 /* Check if feature set would not allow a r/w mount */
2348 if (!ext4_feature_set_ok(sb, 0)) {
2349 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2350 "unknown ROCOMPAT features");
2354 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2355 /* don't clear list on RO mount w/ errors */
2356 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
2357 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
2358 "clearing orphan list.\n");
2359 es->s_last_orphan = 0;
2361 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2365 if (s_flags & MS_RDONLY) {
2366 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2367 sb->s_flags &= ~MS_RDONLY;
2370 /* Needed for iput() to work correctly and not trash data */
2371 sb->s_flags |= MS_ACTIVE;
2372 /* Turn on quotas so that they are updated correctly */
2373 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2374 if (EXT4_SB(sb)->s_qf_names[i]) {
2375 int ret = ext4_quota_on_mount(sb, i);
2377 ext4_msg(sb, KERN_ERR,
2378 "Cannot turn on journaled "
2379 "quota: error %d", ret);
2384 while (es->s_last_orphan) {
2385 struct inode *inode;
2388 * We may have encountered an error during cleanup; if
2389 * so, skip the rest.
2391 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2392 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2393 es->s_last_orphan = 0;
2397 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2398 if (IS_ERR(inode)) {
2399 es->s_last_orphan = 0;
2403 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2404 dquot_initialize(inode);
2405 if (inode->i_nlink) {
2406 if (test_opt(sb, DEBUG))
2407 ext4_msg(sb, KERN_DEBUG,
2408 "%s: truncating inode %lu to %lld bytes",
2409 __func__, inode->i_ino, inode->i_size);
2410 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2411 inode->i_ino, inode->i_size);
2413 truncate_inode_pages(inode->i_mapping, inode->i_size);
2414 ext4_truncate(inode);
2415 inode_unlock(inode);
2418 if (test_opt(sb, DEBUG))
2419 ext4_msg(sb, KERN_DEBUG,
2420 "%s: deleting unreferenced inode %lu",
2421 __func__, inode->i_ino);
2422 jbd_debug(2, "deleting unreferenced inode %lu\n",
2426 iput(inode); /* The delete magic happens here! */
2429 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2432 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2433 PLURAL(nr_orphans));
2435 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2436 PLURAL(nr_truncates));
2438 /* Turn quotas off */
2439 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2440 if (sb_dqopt(sb)->files[i])
2441 dquot_quota_off(sb, i);
2444 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2448 * Maximal extent format file size.
2449 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2450 * extent format containers, within a sector_t, and within i_blocks
2451 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2452 * so that won't be a limiting factor.
2454 * However there is other limiting factor. We do store extents in the form
2455 * of starting block and length, hence the resulting length of the extent
2456 * covering maximum file size must fit into on-disk format containers as
2457 * well. Given that length is always by 1 unit bigger than max unit (because
2458 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2460 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2462 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2465 loff_t upper_limit = MAX_LFS_FILESIZE;
2467 /* small i_blocks in vfs inode? */
2468 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2470 * CONFIG_LBDAF is not enabled implies the inode
2471 * i_block represent total blocks in 512 bytes
2472 * 32 == size of vfs inode i_blocks * 8
2474 upper_limit = (1LL << 32) - 1;
2476 /* total blocks in file system block size */
2477 upper_limit >>= (blkbits - 9);
2478 upper_limit <<= blkbits;
2482 * 32-bit extent-start container, ee_block. We lower the maxbytes
2483 * by one fs block, so ee_len can cover the extent of maximum file
2486 res = (1LL << 32) - 1;
2489 /* Sanity check against vm- & vfs- imposed limits */
2490 if (res > upper_limit)
2497 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2498 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2499 * We need to be 1 filesystem block less than the 2^48 sector limit.
2501 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2503 loff_t res = EXT4_NDIR_BLOCKS;
2506 /* This is calculated to be the largest file size for a dense, block
2507 * mapped file such that the file's total number of 512-byte sectors,
2508 * including data and all indirect blocks, does not exceed (2^48 - 1).
2510 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2511 * number of 512-byte sectors of the file.
2514 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2516 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2517 * the inode i_block field represents total file blocks in
2518 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2520 upper_limit = (1LL << 32) - 1;
2522 /* total blocks in file system block size */
2523 upper_limit >>= (bits - 9);
2527 * We use 48 bit ext4_inode i_blocks
2528 * With EXT4_HUGE_FILE_FL set the i_blocks
2529 * represent total number of blocks in
2530 * file system block size
2532 upper_limit = (1LL << 48) - 1;
2536 /* indirect blocks */
2538 /* double indirect blocks */
2539 meta_blocks += 1 + (1LL << (bits-2));
2540 /* tripple indirect blocks */
2541 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2543 upper_limit -= meta_blocks;
2544 upper_limit <<= bits;
2546 res += 1LL << (bits-2);
2547 res += 1LL << (2*(bits-2));
2548 res += 1LL << (3*(bits-2));
2550 if (res > upper_limit)
2553 if (res > MAX_LFS_FILESIZE)
2554 res = MAX_LFS_FILESIZE;
2559 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2560 ext4_fsblk_t logical_sb_block, int nr)
2562 struct ext4_sb_info *sbi = EXT4_SB(sb);
2563 ext4_group_t bg, first_meta_bg;
2566 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2568 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
2569 return logical_sb_block + nr + 1;
2570 bg = sbi->s_desc_per_block * nr;
2571 if (ext4_bg_has_super(sb, bg))
2575 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2576 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2577 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2580 if (sb->s_blocksize == 1024 && nr == 0 &&
2581 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2584 return (has_super + ext4_group_first_block_no(sb, bg));
2588 * ext4_get_stripe_size: Get the stripe size.
2589 * @sbi: In memory super block info
2591 * If we have specified it via mount option, then
2592 * use the mount option value. If the value specified at mount time is
2593 * greater than the blocks per group use the super block value.
2594 * If the super block value is greater than blocks per group return 0.
2595 * Allocator needs it be less than blocks per group.
2598 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2600 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2601 unsigned long stripe_width =
2602 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2605 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2606 ret = sbi->s_stripe;
2607 else if (stripe_width <= sbi->s_blocks_per_group)
2609 else if (stride <= sbi->s_blocks_per_group)
2615 * If the stripe width is 1, this makes no sense and
2616 * we set it to 0 to turn off stripe handling code.
2625 * Check whether this filesystem can be mounted based on
2626 * the features present and the RDONLY/RDWR mount requested.
2627 * Returns 1 if this filesystem can be mounted as requested,
2628 * 0 if it cannot be.
2630 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2632 if (ext4_has_unknown_ext4_incompat_features(sb)) {
2633 ext4_msg(sb, KERN_ERR,
2634 "Couldn't mount because of "
2635 "unsupported optional features (%x)",
2636 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2637 ~EXT4_FEATURE_INCOMPAT_SUPP));
2644 if (ext4_has_feature_readonly(sb)) {
2645 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2646 sb->s_flags |= MS_RDONLY;
2650 /* Check that feature set is OK for a read-write mount */
2651 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
2652 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2653 "unsupported optional features (%x)",
2654 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2655 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2659 * Large file size enabled file system can only be mounted
2660 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2662 if (ext4_has_feature_huge_file(sb)) {
2663 if (sizeof(blkcnt_t) < sizeof(u64)) {
2664 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2665 "cannot be mounted RDWR without "
2670 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
2671 ext4_msg(sb, KERN_ERR,
2672 "Can't support bigalloc feature without "
2673 "extents feature\n");
2677 #ifndef CONFIG_QUOTA
2678 if (ext4_has_feature_quota(sb) && !readonly) {
2679 ext4_msg(sb, KERN_ERR,
2680 "Filesystem with quota feature cannot be mounted RDWR "
2681 "without CONFIG_QUOTA");
2684 if (ext4_has_feature_project(sb) && !readonly) {
2685 ext4_msg(sb, KERN_ERR,
2686 "Filesystem with project quota feature cannot be mounted RDWR "
2687 "without CONFIG_QUOTA");
2690 #endif /* CONFIG_QUOTA */
2695 * This function is called once a day if we have errors logged
2696 * on the file system
2698 static void print_daily_error_info(unsigned long arg)
2700 struct super_block *sb = (struct super_block *) arg;
2701 struct ext4_sb_info *sbi;
2702 struct ext4_super_block *es;
2707 if (es->s_error_count)
2708 /* fsck newer than v1.41.13 is needed to clean this condition. */
2709 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
2710 le32_to_cpu(es->s_error_count));
2711 if (es->s_first_error_time) {
2712 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
2713 sb->s_id, le32_to_cpu(es->s_first_error_time),
2714 (int) sizeof(es->s_first_error_func),
2715 es->s_first_error_func,
2716 le32_to_cpu(es->s_first_error_line));
2717 if (es->s_first_error_ino)
2718 printk(": inode %u",
2719 le32_to_cpu(es->s_first_error_ino));
2720 if (es->s_first_error_block)
2721 printk(": block %llu", (unsigned long long)
2722 le64_to_cpu(es->s_first_error_block));
2725 if (es->s_last_error_time) {
2726 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
2727 sb->s_id, le32_to_cpu(es->s_last_error_time),
2728 (int) sizeof(es->s_last_error_func),
2729 es->s_last_error_func,
2730 le32_to_cpu(es->s_last_error_line));
2731 if (es->s_last_error_ino)
2732 printk(": inode %u",
2733 le32_to_cpu(es->s_last_error_ino));
2734 if (es->s_last_error_block)
2735 printk(": block %llu", (unsigned long long)
2736 le64_to_cpu(es->s_last_error_block));
2739 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2742 /* Find next suitable group and run ext4_init_inode_table */
2743 static int ext4_run_li_request(struct ext4_li_request *elr)
2745 struct ext4_group_desc *gdp = NULL;
2746 ext4_group_t group, ngroups;
2747 struct super_block *sb;
2748 unsigned long timeout = 0;
2752 ngroups = EXT4_SB(sb)->s_groups_count;
2754 for (group = elr->lr_next_group; group < ngroups; group++) {
2755 gdp = ext4_get_group_desc(sb, group, NULL);
2761 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2765 if (group >= ngroups)
2770 ret = ext4_init_inode_table(sb, group,
2771 elr->lr_timeout ? 0 : 1);
2772 if (elr->lr_timeout == 0) {
2773 timeout = (jiffies - timeout) *
2774 elr->lr_sbi->s_li_wait_mult;
2775 elr->lr_timeout = timeout;
2777 elr->lr_next_sched = jiffies + elr->lr_timeout;
2778 elr->lr_next_group = group + 1;
2784 * Remove lr_request from the list_request and free the
2785 * request structure. Should be called with li_list_mtx held
2787 static void ext4_remove_li_request(struct ext4_li_request *elr)
2789 struct ext4_sb_info *sbi;
2796 list_del(&elr->lr_request);
2797 sbi->s_li_request = NULL;
2801 static void ext4_unregister_li_request(struct super_block *sb)
2803 mutex_lock(&ext4_li_mtx);
2804 if (!ext4_li_info) {
2805 mutex_unlock(&ext4_li_mtx);
2809 mutex_lock(&ext4_li_info->li_list_mtx);
2810 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2811 mutex_unlock(&ext4_li_info->li_list_mtx);
2812 mutex_unlock(&ext4_li_mtx);
2815 static struct task_struct *ext4_lazyinit_task;
2818 * This is the function where ext4lazyinit thread lives. It walks
2819 * through the request list searching for next scheduled filesystem.
2820 * When such a fs is found, run the lazy initialization request
2821 * (ext4_rn_li_request) and keep track of the time spend in this
2822 * function. Based on that time we compute next schedule time of
2823 * the request. When walking through the list is complete, compute
2824 * next waking time and put itself into sleep.
2826 static int ext4_lazyinit_thread(void *arg)
2828 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2829 struct list_head *pos, *n;
2830 struct ext4_li_request *elr;
2831 unsigned long next_wakeup, cur;
2833 BUG_ON(NULL == eli);
2837 next_wakeup = MAX_JIFFY_OFFSET;
2839 mutex_lock(&eli->li_list_mtx);
2840 if (list_empty(&eli->li_request_list)) {
2841 mutex_unlock(&eli->li_list_mtx);
2844 list_for_each_safe(pos, n, &eli->li_request_list) {
2847 elr = list_entry(pos, struct ext4_li_request,
2850 if (time_before(jiffies, elr->lr_next_sched)) {
2851 if (time_before(elr->lr_next_sched, next_wakeup))
2852 next_wakeup = elr->lr_next_sched;
2855 if (down_read_trylock(&elr->lr_super->s_umount)) {
2856 if (sb_start_write_trylock(elr->lr_super)) {
2859 * We hold sb->s_umount, sb can not
2860 * be removed from the list, it is
2861 * now safe to drop li_list_mtx
2863 mutex_unlock(&eli->li_list_mtx);
2864 err = ext4_run_li_request(elr);
2865 sb_end_write(elr->lr_super);
2866 mutex_lock(&eli->li_list_mtx);
2869 up_read((&elr->lr_super->s_umount));
2871 /* error, remove the lazy_init job */
2873 ext4_remove_li_request(elr);
2877 elr->lr_next_sched = jiffies +
2879 % (EXT4_DEF_LI_MAX_START_DELAY * HZ));
2881 if (time_before(elr->lr_next_sched, next_wakeup))
2882 next_wakeup = elr->lr_next_sched;
2884 mutex_unlock(&eli->li_list_mtx);
2889 if ((time_after_eq(cur, next_wakeup)) ||
2890 (MAX_JIFFY_OFFSET == next_wakeup)) {
2895 schedule_timeout_interruptible(next_wakeup - cur);
2897 if (kthread_should_stop()) {
2898 ext4_clear_request_list();
2905 * It looks like the request list is empty, but we need
2906 * to check it under the li_list_mtx lock, to prevent any
2907 * additions into it, and of course we should lock ext4_li_mtx
2908 * to atomically free the list and ext4_li_info, because at
2909 * this point another ext4 filesystem could be registering
2912 mutex_lock(&ext4_li_mtx);
2913 mutex_lock(&eli->li_list_mtx);
2914 if (!list_empty(&eli->li_request_list)) {
2915 mutex_unlock(&eli->li_list_mtx);
2916 mutex_unlock(&ext4_li_mtx);
2919 mutex_unlock(&eli->li_list_mtx);
2920 kfree(ext4_li_info);
2921 ext4_li_info = NULL;
2922 mutex_unlock(&ext4_li_mtx);
2927 static void ext4_clear_request_list(void)
2929 struct list_head *pos, *n;
2930 struct ext4_li_request *elr;
2932 mutex_lock(&ext4_li_info->li_list_mtx);
2933 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2934 elr = list_entry(pos, struct ext4_li_request,
2936 ext4_remove_li_request(elr);
2938 mutex_unlock(&ext4_li_info->li_list_mtx);
2941 static int ext4_run_lazyinit_thread(void)
2943 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2944 ext4_li_info, "ext4lazyinit");
2945 if (IS_ERR(ext4_lazyinit_task)) {
2946 int err = PTR_ERR(ext4_lazyinit_task);
2947 ext4_clear_request_list();
2948 kfree(ext4_li_info);
2949 ext4_li_info = NULL;
2950 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
2951 "initialization thread\n",
2955 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2960 * Check whether it make sense to run itable init. thread or not.
2961 * If there is at least one uninitialized inode table, return
2962 * corresponding group number, else the loop goes through all
2963 * groups and return total number of groups.
2965 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2967 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2968 struct ext4_group_desc *gdp = NULL;
2970 for (group = 0; group < ngroups; group++) {
2971 gdp = ext4_get_group_desc(sb, group, NULL);
2975 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2982 static int ext4_li_info_new(void)
2984 struct ext4_lazy_init *eli = NULL;
2986 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2990 INIT_LIST_HEAD(&eli->li_request_list);
2991 mutex_init(&eli->li_list_mtx);
2993 eli->li_state |= EXT4_LAZYINIT_QUIT;
3000 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3003 struct ext4_sb_info *sbi = EXT4_SB(sb);
3004 struct ext4_li_request *elr;
3006 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3012 elr->lr_next_group = start;
3015 * Randomize first schedule time of the request to
3016 * spread the inode table initialization requests
3019 elr->lr_next_sched = jiffies + (prandom_u32() %
3020 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3024 int ext4_register_li_request(struct super_block *sb,
3025 ext4_group_t first_not_zeroed)
3027 struct ext4_sb_info *sbi = EXT4_SB(sb);
3028 struct ext4_li_request *elr = NULL;
3029 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3032 mutex_lock(&ext4_li_mtx);
3033 if (sbi->s_li_request != NULL) {
3035 * Reset timeout so it can be computed again, because
3036 * s_li_wait_mult might have changed.
3038 sbi->s_li_request->lr_timeout = 0;
3042 if (first_not_zeroed == ngroups ||
3043 (sb->s_flags & MS_RDONLY) ||
3044 !test_opt(sb, INIT_INODE_TABLE))
3047 elr = ext4_li_request_new(sb, first_not_zeroed);
3053 if (NULL == ext4_li_info) {
3054 ret = ext4_li_info_new();
3059 mutex_lock(&ext4_li_info->li_list_mtx);
3060 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3061 mutex_unlock(&ext4_li_info->li_list_mtx);
3063 sbi->s_li_request = elr;
3065 * set elr to NULL here since it has been inserted to
3066 * the request_list and the removal and free of it is
3067 * handled by ext4_clear_request_list from now on.
3071 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3072 ret = ext4_run_lazyinit_thread();
3077 mutex_unlock(&ext4_li_mtx);
3084 * We do not need to lock anything since this is called on
3087 static void ext4_destroy_lazyinit_thread(void)
3090 * If thread exited earlier
3091 * there's nothing to be done.
3093 if (!ext4_li_info || !ext4_lazyinit_task)
3096 kthread_stop(ext4_lazyinit_task);
3099 static int set_journal_csum_feature_set(struct super_block *sb)
3102 int compat, incompat;
3103 struct ext4_sb_info *sbi = EXT4_SB(sb);
3105 if (ext4_has_metadata_csum(sb)) {
3106 /* journal checksum v3 */
3108 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
3110 /* journal checksum v1 */
3111 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3115 jbd2_journal_clear_features(sbi->s_journal,
3116 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3117 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3118 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3119 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3120 ret = jbd2_journal_set_features(sbi->s_journal,
3122 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3124 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3125 ret = jbd2_journal_set_features(sbi->s_journal,
3128 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3129 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3131 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3132 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3139 * Note: calculating the overhead so we can be compatible with
3140 * historical BSD practice is quite difficult in the face of
3141 * clusters/bigalloc. This is because multiple metadata blocks from
3142 * different block group can end up in the same allocation cluster.
3143 * Calculating the exact overhead in the face of clustered allocation
3144 * requires either O(all block bitmaps) in memory or O(number of block
3145 * groups**2) in time. We will still calculate the superblock for
3146 * older file systems --- and if we come across with a bigalloc file
3147 * system with zero in s_overhead_clusters the estimate will be close to
3148 * correct especially for very large cluster sizes --- but for newer
3149 * file systems, it's better to calculate this figure once at mkfs
3150 * time, and store it in the superblock. If the superblock value is
3151 * present (even for non-bigalloc file systems), we will use it.
3153 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3156 struct ext4_sb_info *sbi = EXT4_SB(sb);
3157 struct ext4_group_desc *gdp;
3158 ext4_fsblk_t first_block, last_block, b;
3159 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3160 int s, j, count = 0;
3162 if (!ext4_has_feature_bigalloc(sb))
3163 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3164 sbi->s_itb_per_group + 2);
3166 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3167 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3168 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3169 for (i = 0; i < ngroups; i++) {
3170 gdp = ext4_get_group_desc(sb, i, NULL);
3171 b = ext4_block_bitmap(sb, gdp);
3172 if (b >= first_block && b <= last_block) {
3173 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3176 b = ext4_inode_bitmap(sb, gdp);
3177 if (b >= first_block && b <= last_block) {
3178 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3181 b = ext4_inode_table(sb, gdp);
3182 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3183 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3184 int c = EXT4_B2C(sbi, b - first_block);
3185 ext4_set_bit(c, buf);
3191 if (ext4_bg_has_super(sb, grp)) {
3192 ext4_set_bit(s++, buf);
3195 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3196 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3202 return EXT4_CLUSTERS_PER_GROUP(sb) -
3203 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3207 * Compute the overhead and stash it in sbi->s_overhead
3209 int ext4_calculate_overhead(struct super_block *sb)
3211 struct ext4_sb_info *sbi = EXT4_SB(sb);
3212 struct ext4_super_block *es = sbi->s_es;
3213 struct inode *j_inode;
3214 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
3215 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3216 ext4_fsblk_t overhead = 0;
3217 char *buf = (char *) get_zeroed_page(GFP_NOFS);
3223 * Compute the overhead (FS structures). This is constant
3224 * for a given filesystem unless the number of block groups
3225 * changes so we cache the previous value until it does.
3229 * All of the blocks before first_data_block are overhead
3231 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3234 * Add the overhead found in each block group
3236 for (i = 0; i < ngroups; i++) {
3239 blks = count_overhead(sb, i, buf);
3242 memset(buf, 0, PAGE_SIZE);
3247 * Add the internal journal blocks whether the journal has been
3250 if (sbi->s_journal && !sbi->journal_bdev)
3251 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3252 else if (ext4_has_feature_journal(sb) && !sbi->s_journal) {
3253 j_inode = ext4_get_journal_inode(sb, j_inum);
3255 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
3256 overhead += EXT4_NUM_B2C(sbi, j_blocks);
3259 ext4_msg(sb, KERN_ERR, "can't get journal size");
3262 sbi->s_overhead = overhead;
3264 free_page((unsigned long) buf);
3268 static void ext4_set_resv_clusters(struct super_block *sb)
3270 ext4_fsblk_t resv_clusters;
3271 struct ext4_sb_info *sbi = EXT4_SB(sb);
3274 * There's no need to reserve anything when we aren't using extents.
3275 * The space estimates are exact, there are no unwritten extents,
3276 * hole punching doesn't need new metadata... This is needed especially
3277 * to keep ext2/3 backward compatibility.
3279 if (!ext4_has_feature_extents(sb))
3282 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3283 * This should cover the situations where we can not afford to run
3284 * out of space like for example punch hole, or converting
3285 * unwritten extents in delalloc path. In most cases such
3286 * allocation would require 1, or 2 blocks, higher numbers are
3289 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3290 sbi->s_cluster_bits);
3292 do_div(resv_clusters, 50);
3293 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3295 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
3298 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3300 char *orig_data = kstrdup(data, GFP_KERNEL);
3301 struct buffer_head *bh;
3302 struct ext4_super_block *es = NULL;
3303 struct ext4_sb_info *sbi;
3305 ext4_fsblk_t sb_block = get_sb_block(&data);
3306 ext4_fsblk_t logical_sb_block;
3307 unsigned long offset = 0;
3308 unsigned long journal_devnum = 0;
3309 unsigned long def_mount_opts;
3313 int blocksize, clustersize;
3314 unsigned int db_count;
3316 int needs_recovery, has_huge_files, has_bigalloc;
3319 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3320 ext4_group_t first_not_zeroed;
3322 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3326 sbi->s_blockgroup_lock =
3327 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3328 if (!sbi->s_blockgroup_lock) {
3332 sb->s_fs_info = sbi;
3334 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3335 sbi->s_sb_block = sb_block;
3336 if (sb->s_bdev->bd_part)
3337 sbi->s_sectors_written_start =
3338 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3340 /* Cleanup superblock name */
3341 strreplace(sb->s_id, '/', '!');
3343 /* -EINVAL is default */
3345 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3347 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3352 * The ext4 superblock will not be buffer aligned for other than 1kB
3353 * block sizes. We need to calculate the offset from buffer start.
3355 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3356 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3357 offset = do_div(logical_sb_block, blocksize);
3359 logical_sb_block = sb_block;
3362 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
3363 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3367 * Note: s_es must be initialized as soon as possible because
3368 * some ext4 macro-instructions depend on its value
3370 es = (struct ext4_super_block *) (bh->b_data + offset);
3372 sb->s_magic = le16_to_cpu(es->s_magic);
3373 if (sb->s_magic != EXT4_SUPER_MAGIC)
3375 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3377 /* Warn if metadata_csum and gdt_csum are both set. */
3378 if (ext4_has_feature_metadata_csum(sb) &&
3379 ext4_has_feature_gdt_csum(sb))
3380 ext4_warning(sb, "metadata_csum and uninit_bg are "
3381 "redundant flags; please run fsck.");
3383 /* Check for a known checksum algorithm */
3384 if (!ext4_verify_csum_type(sb, es)) {
3385 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3386 "unknown checksum algorithm.");
3391 /* Load the checksum driver */
3392 if (ext4_has_feature_metadata_csum(sb)) {
3393 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3394 if (IS_ERR(sbi->s_chksum_driver)) {
3395 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3396 ret = PTR_ERR(sbi->s_chksum_driver);
3397 sbi->s_chksum_driver = NULL;
3402 /* Check superblock checksum */
3403 if (!ext4_superblock_csum_verify(sb, es)) {
3404 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3405 "invalid superblock checksum. Run e2fsck?");
3411 /* Precompute checksum seed for all metadata */
3412 if (ext4_has_feature_csum_seed(sb))
3413 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
3414 else if (ext4_has_metadata_csum(sb))
3415 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3416 sizeof(es->s_uuid));
3418 /* Set defaults before we parse the mount options */
3419 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3420 set_opt(sb, INIT_INODE_TABLE);
3421 if (def_mount_opts & EXT4_DEFM_DEBUG)
3423 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3425 if (def_mount_opts & EXT4_DEFM_UID16)
3426 set_opt(sb, NO_UID32);
3427 /* xattr user namespace & acls are now defaulted on */
3428 set_opt(sb, XATTR_USER);
3429 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3430 set_opt(sb, POSIX_ACL);
3432 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3433 if (ext4_has_metadata_csum(sb))
3434 set_opt(sb, JOURNAL_CHECKSUM);
3436 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3437 set_opt(sb, JOURNAL_DATA);
3438 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3439 set_opt(sb, ORDERED_DATA);
3440 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3441 set_opt(sb, WRITEBACK_DATA);
3443 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3444 set_opt(sb, ERRORS_PANIC);
3445 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3446 set_opt(sb, ERRORS_CONT);
3448 set_opt(sb, ERRORS_RO);
3449 /* block_validity enabled by default; disable with noblock_validity */
3450 set_opt(sb, BLOCK_VALIDITY);
3451 if (def_mount_opts & EXT4_DEFM_DISCARD)
3452 set_opt(sb, DISCARD);
3454 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3455 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3456 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3457 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3458 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3460 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3461 set_opt(sb, BARRIER);
3464 * enable delayed allocation by default
3465 * Use -o nodelalloc to turn it off
3467 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3468 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3469 set_opt(sb, DELALLOC);
3472 * set default s_li_wait_mult for lazyinit, for the case there is
3473 * no mount option specified.
3475 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3477 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3478 &journal_devnum, &journal_ioprio, 0)) {
3479 ext4_msg(sb, KERN_WARNING,
3480 "failed to parse options in superblock: %s",
3481 sbi->s_es->s_mount_opts);
3483 sbi->s_def_mount_opt = sbi->s_mount_opt;
3484 if (!parse_options((char *) data, sb, &journal_devnum,
3485 &journal_ioprio, 0))
3488 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3489 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3490 "with data=journal disables delayed "
3491 "allocation and O_DIRECT support!\n");
3492 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3493 ext4_msg(sb, KERN_ERR, "can't mount with "
3494 "both data=journal and delalloc");
3497 if (test_opt(sb, DIOREAD_NOLOCK)) {
3498 ext4_msg(sb, KERN_ERR, "can't mount with "
3499 "both data=journal and dioread_nolock");
3502 if (test_opt(sb, DAX)) {
3503 ext4_msg(sb, KERN_ERR, "can't mount with "
3504 "both data=journal and dax");
3507 if (test_opt(sb, DELALLOC))
3508 clear_opt(sb, DELALLOC);
3510 sb->s_iflags |= SB_I_CGROUPWB;
3513 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3514 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3516 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3517 (ext4_has_compat_features(sb) ||
3518 ext4_has_ro_compat_features(sb) ||
3519 ext4_has_incompat_features(sb)))
3520 ext4_msg(sb, KERN_WARNING,
3521 "feature flags set on rev 0 fs, "
3522 "running e2fsck is recommended");
3524 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3525 set_opt2(sb, HURD_COMPAT);
3526 if (ext4_has_feature_64bit(sb)) {
3527 ext4_msg(sb, KERN_ERR,
3528 "The Hurd can't support 64-bit file systems");
3533 if (IS_EXT2_SB(sb)) {
3534 if (ext2_feature_set_ok(sb))
3535 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3536 "using the ext4 subsystem");
3538 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3539 "to feature incompatibilities");
3544 if (IS_EXT3_SB(sb)) {
3545 if (ext3_feature_set_ok(sb))
3546 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3547 "using the ext4 subsystem");
3549 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3550 "to feature incompatibilities");
3556 * Check feature flags regardless of the revision level, since we
3557 * previously didn't change the revision level when setting the flags,
3558 * so there is a chance incompat flags are set on a rev 0 filesystem.
3560 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3563 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3564 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3565 blocksize > EXT4_MAX_BLOCK_SIZE) {
3566 ext4_msg(sb, KERN_ERR,
3567 "Unsupported filesystem blocksize %d", blocksize);
3571 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
3572 ext4_msg(sb, KERN_ERR,
3573 "Number of reserved GDT blocks insanely large: %d",
3574 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
3578 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
3579 err = bdev_dax_supported(sb, blocksize);
3584 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
3585 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
3586 es->s_encryption_level);
3590 if (sb->s_blocksize != blocksize) {
3591 /* Validate the filesystem blocksize */
3592 if (!sb_set_blocksize(sb, blocksize)) {
3593 ext4_msg(sb, KERN_ERR, "bad block size %d",
3599 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3600 offset = do_div(logical_sb_block, blocksize);
3601 bh = sb_bread_unmovable(sb, logical_sb_block);
3603 ext4_msg(sb, KERN_ERR,
3604 "Can't read superblock on 2nd try");
3607 es = (struct ext4_super_block *)(bh->b_data + offset);
3609 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3610 ext4_msg(sb, KERN_ERR,
3611 "Magic mismatch, very weird!");
3616 has_huge_files = ext4_has_feature_huge_file(sb);
3617 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3619 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3621 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3622 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3623 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3625 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3626 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3627 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3628 (!is_power_of_2(sbi->s_inode_size)) ||
3629 (sbi->s_inode_size > blocksize)) {
3630 ext4_msg(sb, KERN_ERR,
3631 "unsupported inode size: %d",
3635 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3636 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3639 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3640 if (ext4_has_feature_64bit(sb)) {
3641 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3642 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3643 !is_power_of_2(sbi->s_desc_size)) {
3644 ext4_msg(sb, KERN_ERR,
3645 "unsupported descriptor size %lu",
3650 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3652 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3653 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3654 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3657 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3658 if (sbi->s_inodes_per_block == 0)
3660 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3661 sbi->s_inodes_per_block;
3662 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3664 sbi->s_mount_state = le16_to_cpu(es->s_state);
3665 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3666 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3668 for (i = 0; i < 4; i++)
3669 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3670 sbi->s_def_hash_version = es->s_def_hash_version;
3671 if (ext4_has_feature_dir_index(sb)) {
3672 i = le32_to_cpu(es->s_flags);
3673 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3674 sbi->s_hash_unsigned = 3;
3675 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3676 #ifdef __CHAR_UNSIGNED__
3677 if (!(sb->s_flags & MS_RDONLY))
3679 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3680 sbi->s_hash_unsigned = 3;
3682 if (!(sb->s_flags & MS_RDONLY))
3684 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3689 /* Handle clustersize */
3690 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3691 has_bigalloc = ext4_has_feature_bigalloc(sb);
3693 if (clustersize < blocksize) {
3694 ext4_msg(sb, KERN_ERR,
3695 "cluster size (%d) smaller than "
3696 "block size (%d)", clustersize, blocksize);
3699 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3700 le32_to_cpu(es->s_log_block_size);
3701 sbi->s_clusters_per_group =
3702 le32_to_cpu(es->s_clusters_per_group);
3703 if (sbi->s_clusters_per_group > blocksize * 8) {
3704 ext4_msg(sb, KERN_ERR,
3705 "#clusters per group too big: %lu",
3706 sbi->s_clusters_per_group);
3709 if (sbi->s_blocks_per_group !=
3710 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3711 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3712 "clusters per group (%lu) inconsistent",
3713 sbi->s_blocks_per_group,
3714 sbi->s_clusters_per_group);
3718 if (clustersize != blocksize) {
3719 ext4_warning(sb, "fragment/cluster size (%d) != "
3720 "block size (%d)", clustersize,
3722 clustersize = blocksize;
3724 if (sbi->s_blocks_per_group > blocksize * 8) {
3725 ext4_msg(sb, KERN_ERR,
3726 "#blocks per group too big: %lu",
3727 sbi->s_blocks_per_group);
3730 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3731 sbi->s_cluster_bits = 0;
3733 sbi->s_cluster_ratio = clustersize / blocksize;
3735 if (sbi->s_inodes_per_group > blocksize * 8) {
3736 ext4_msg(sb, KERN_ERR,
3737 "#inodes per group too big: %lu",
3738 sbi->s_inodes_per_group);
3742 /* Do we have standard group size of clustersize * 8 blocks ? */
3743 if (sbi->s_blocks_per_group == clustersize << 3)
3744 set_opt2(sb, STD_GROUP_SIZE);
3747 * Test whether we have more sectors than will fit in sector_t,
3748 * and whether the max offset is addressable by the page cache.
3750 err = generic_check_addressable(sb->s_blocksize_bits,
3751 ext4_blocks_count(es));
3753 ext4_msg(sb, KERN_ERR, "filesystem"
3754 " too large to mount safely on this system");
3755 if (sizeof(sector_t) < 8)
3756 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3760 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3763 /* check blocks count against device size */
3764 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3765 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3766 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3767 "exceeds size of device (%llu blocks)",
3768 ext4_blocks_count(es), blocks_count);
3773 * It makes no sense for the first data block to be beyond the end
3774 * of the filesystem.
3776 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3777 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3778 "block %u is beyond end of filesystem (%llu)",
3779 le32_to_cpu(es->s_first_data_block),
3780 ext4_blocks_count(es));
3783 blocks_count = (ext4_blocks_count(es) -
3784 le32_to_cpu(es->s_first_data_block) +
3785 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3786 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3787 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3788 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3789 "(block count %llu, first data block %u, "
3790 "blocks per group %lu)", sbi->s_groups_count,
3791 ext4_blocks_count(es),
3792 le32_to_cpu(es->s_first_data_block),
3793 EXT4_BLOCKS_PER_GROUP(sb));
3796 sbi->s_groups_count = blocks_count;
3797 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3798 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3799 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3800 EXT4_DESC_PER_BLOCK(sb);
3801 sbi->s_group_desc = ext4_kvmalloc(db_count *
3802 sizeof(struct buffer_head *),
3804 if (sbi->s_group_desc == NULL) {
3805 ext4_msg(sb, KERN_ERR, "not enough memory");
3810 bgl_lock_init(sbi->s_blockgroup_lock);
3812 for (i = 0; i < db_count; i++) {
3813 block = descriptor_loc(sb, logical_sb_block, i);
3814 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
3815 if (!sbi->s_group_desc[i]) {
3816 ext4_msg(sb, KERN_ERR,
3817 "can't read group descriptor %d", i);
3822 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
3823 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3824 ret = -EFSCORRUPTED;
3828 sbi->s_gdb_count = db_count;
3829 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3830 spin_lock_init(&sbi->s_next_gen_lock);
3832 setup_timer(&sbi->s_err_report, print_daily_error_info,
3833 (unsigned long) sb);
3835 /* Register extent status tree shrinker */
3836 if (ext4_es_register_shrinker(sbi))
3839 sbi->s_stripe = ext4_get_stripe_size(sbi);
3840 sbi->s_extent_max_zeroout_kb = 32;
3843 * set up enough so that it can read an inode
3845 sb->s_op = &ext4_sops;
3846 sb->s_export_op = &ext4_export_ops;
3847 sb->s_xattr = ext4_xattr_handlers;
3848 sb->s_cop = &ext4_cryptops;
3850 sb->dq_op = &ext4_quota_operations;
3851 if (ext4_has_feature_quota(sb))
3852 sb->s_qcop = &dquot_quotactl_sysfile_ops;
3854 sb->s_qcop = &ext4_qctl_operations;
3855 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
3857 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3859 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3860 mutex_init(&sbi->s_orphan_lock);
3864 needs_recovery = (es->s_last_orphan != 0 ||
3865 ext4_has_feature_journal_needs_recovery(sb));
3867 if (ext4_has_feature_mmp(sb) && !(sb->s_flags & MS_RDONLY))
3868 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3869 goto failed_mount3a;
3872 * The first inode we look at is the journal inode. Don't try
3873 * root first: it may be modified in the journal!
3875 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
3876 if (ext4_load_journal(sb, es, journal_devnum))
3877 goto failed_mount3a;
3878 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3879 ext4_has_feature_journal_needs_recovery(sb)) {
3880 ext4_msg(sb, KERN_ERR, "required journal recovery "
3881 "suppressed and not mounted read-only");
3882 goto failed_mount_wq;
3884 /* Nojournal mode, all journal mount options are illegal */
3885 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
3886 ext4_msg(sb, KERN_ERR, "can't mount with "
3887 "journal_checksum, fs mounted w/o journal");
3888 goto failed_mount_wq;
3890 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3891 ext4_msg(sb, KERN_ERR, "can't mount with "
3892 "journal_async_commit, fs mounted w/o journal");
3893 goto failed_mount_wq;
3895 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
3896 ext4_msg(sb, KERN_ERR, "can't mount with "
3897 "commit=%lu, fs mounted w/o journal",
3898 sbi->s_commit_interval / HZ);
3899 goto failed_mount_wq;
3901 if (EXT4_MOUNT_DATA_FLAGS &
3902 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
3903 ext4_msg(sb, KERN_ERR, "can't mount with "
3904 "data=, fs mounted w/o journal");
3905 goto failed_mount_wq;
3907 sbi->s_def_mount_opt &= EXT4_MOUNT_JOURNAL_CHECKSUM;
3908 clear_opt(sb, JOURNAL_CHECKSUM);
3909 clear_opt(sb, DATA_FLAGS);
3910 sbi->s_journal = NULL;
3915 if (ext4_has_feature_64bit(sb) &&
3916 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3917 JBD2_FEATURE_INCOMPAT_64BIT)) {
3918 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3919 goto failed_mount_wq;
3922 if (!set_journal_csum_feature_set(sb)) {
3923 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3925 goto failed_mount_wq;
3928 /* We have now updated the journal if required, so we can
3929 * validate the data journaling mode. */
3930 switch (test_opt(sb, DATA_FLAGS)) {
3932 /* No mode set, assume a default based on the journal
3933 * capabilities: ORDERED_DATA if the journal can
3934 * cope, else JOURNAL_DATA
3936 if (jbd2_journal_check_available_features
3937 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3938 set_opt(sb, ORDERED_DATA);
3940 set_opt(sb, JOURNAL_DATA);
3943 case EXT4_MOUNT_ORDERED_DATA:
3944 case EXT4_MOUNT_WRITEBACK_DATA:
3945 if (!jbd2_journal_check_available_features
3946 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3947 ext4_msg(sb, KERN_ERR, "Journal does not support "
3948 "requested data journaling mode");
3949 goto failed_mount_wq;
3954 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3956 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3959 sbi->s_mb_cache = ext4_xattr_create_cache();
3960 if (!sbi->s_mb_cache) {
3961 ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
3962 goto failed_mount_wq;
3965 if ((DUMMY_ENCRYPTION_ENABLED(sbi) || ext4_has_feature_encrypt(sb)) &&
3966 (blocksize != PAGE_SIZE)) {
3967 ext4_msg(sb, KERN_ERR,
3968 "Unsupported blocksize for fs encryption");
3969 goto failed_mount_wq;
3972 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !(sb->s_flags & MS_RDONLY) &&
3973 !ext4_has_feature_encrypt(sb)) {
3974 ext4_set_feature_encrypt(sb);
3975 ext4_commit_super(sb, 1);
3979 * Get the # of file system overhead blocks from the
3980 * superblock if present.
3982 if (es->s_overhead_clusters)
3983 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
3985 err = ext4_calculate_overhead(sb);
3987 goto failed_mount_wq;
3991 * The maximum number of concurrent works can be high and
3992 * concurrency isn't really necessary. Limit it to 1.
3994 EXT4_SB(sb)->rsv_conversion_wq =
3995 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
3996 if (!EXT4_SB(sb)->rsv_conversion_wq) {
3997 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
4003 * The jbd2_journal_load will have done any necessary log recovery,
4004 * so we can safely mount the rest of the filesystem now.
4007 root = ext4_iget(sb, EXT4_ROOT_INO);
4009 ext4_msg(sb, KERN_ERR, "get root inode failed");
4010 ret = PTR_ERR(root);
4014 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4015 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4019 sb->s_root = d_make_root(root);
4021 ext4_msg(sb, KERN_ERR, "get root dentry failed");
4026 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
4027 sb->s_flags |= MS_RDONLY;
4029 /* determine the minimum size of new large inodes, if present */
4030 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4031 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4032 EXT4_GOOD_OLD_INODE_SIZE;
4033 if (ext4_has_feature_extra_isize(sb)) {
4034 if (sbi->s_want_extra_isize <
4035 le16_to_cpu(es->s_want_extra_isize))
4036 sbi->s_want_extra_isize =
4037 le16_to_cpu(es->s_want_extra_isize);
4038 if (sbi->s_want_extra_isize <
4039 le16_to_cpu(es->s_min_extra_isize))
4040 sbi->s_want_extra_isize =
4041 le16_to_cpu(es->s_min_extra_isize);
4044 /* Check if enough inode space is available */
4045 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4046 sbi->s_inode_size) {
4047 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4048 EXT4_GOOD_OLD_INODE_SIZE;
4049 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4053 ext4_set_resv_clusters(sb);
4055 err = ext4_setup_system_zone(sb);
4057 ext4_msg(sb, KERN_ERR, "failed to initialize system "
4059 goto failed_mount4a;
4063 err = ext4_mb_init(sb);
4065 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4070 block = ext4_count_free_clusters(sb);
4071 ext4_free_blocks_count_set(sbi->s_es,
4072 EXT4_C2B(sbi, block));
4073 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4076 unsigned long freei = ext4_count_free_inodes(sb);
4077 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
4078 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4082 err = percpu_counter_init(&sbi->s_dirs_counter,
4083 ext4_count_dirs(sb), GFP_KERNEL);
4085 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4088 err = percpu_init_rwsem(&sbi->s_journal_flag_rwsem);
4091 ext4_msg(sb, KERN_ERR, "insufficient memory");
4095 if (ext4_has_feature_flex_bg(sb))
4096 if (!ext4_fill_flex_info(sb)) {
4097 ext4_msg(sb, KERN_ERR,
4098 "unable to initialize "
4099 "flex_bg meta info!");
4103 err = ext4_register_li_request(sb, first_not_zeroed);
4107 err = ext4_register_sysfs(sb);
4112 /* Enable quota usage during mount. */
4113 if (ext4_has_feature_quota(sb) && !(sb->s_flags & MS_RDONLY)) {
4114 err = ext4_enable_quotas(sb);
4118 #endif /* CONFIG_QUOTA */
4120 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4121 ext4_orphan_cleanup(sb, es);
4122 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4123 if (needs_recovery) {
4124 ext4_msg(sb, KERN_INFO, "recovery complete");
4125 ext4_mark_recovery_complete(sb, es);
4127 if (EXT4_SB(sb)->s_journal) {
4128 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4129 descr = " journalled data mode";
4130 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4131 descr = " ordered data mode";
4133 descr = " writeback data mode";
4135 descr = "out journal";
4137 if (test_opt(sb, DISCARD)) {
4138 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4139 if (!blk_queue_discard(q))
4140 ext4_msg(sb, KERN_WARNING,
4141 "mounting with \"discard\" option, but "
4142 "the device does not support discard");
4145 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
4146 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4147 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
4148 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4150 if (es->s_error_count)
4151 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4153 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4154 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4155 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4156 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4159 #ifdef CONFIG_EXT4_FS_ENCRYPTION
4160 memcpy(sbi->key_prefix, EXT4_KEY_DESC_PREFIX,
4161 EXT4_KEY_DESC_PREFIX_SIZE);
4162 sbi->key_prefix_size = EXT4_KEY_DESC_PREFIX_SIZE;
4168 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4173 ext4_unregister_sysfs(sb);
4176 ext4_unregister_li_request(sb);
4178 ext4_mb_release(sb);
4179 if (sbi->s_flex_groups)
4180 kvfree(sbi->s_flex_groups);
4181 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4182 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4183 percpu_counter_destroy(&sbi->s_dirs_counter);
4184 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4186 ext4_ext_release(sb);
4187 ext4_release_system_zone(sb);
4192 ext4_msg(sb, KERN_ERR, "mount failed");
4193 if (EXT4_SB(sb)->rsv_conversion_wq)
4194 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4196 if (sbi->s_mb_cache) {
4197 ext4_xattr_destroy_cache(sbi->s_mb_cache);
4198 sbi->s_mb_cache = NULL;
4200 if (sbi->s_journal) {
4201 jbd2_journal_destroy(sbi->s_journal);
4202 sbi->s_journal = NULL;
4205 ext4_es_unregister_shrinker(sbi);
4207 del_timer_sync(&sbi->s_err_report);
4209 kthread_stop(sbi->s_mmp_tsk);
4211 for (i = 0; i < db_count; i++)
4212 brelse(sbi->s_group_desc[i]);
4213 kvfree(sbi->s_group_desc);
4215 if (sbi->s_chksum_driver)
4216 crypto_free_shash(sbi->s_chksum_driver);
4218 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4219 kfree(sbi->s_qf_names[i]);
4221 ext4_blkdev_remove(sbi);
4224 sb->s_fs_info = NULL;
4225 kfree(sbi->s_blockgroup_lock);
4229 return err ? err : ret;
4233 * Setup any per-fs journal parameters now. We'll do this both on
4234 * initial mount, once the journal has been initialised but before we've
4235 * done any recovery; and again on any subsequent remount.
4237 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4239 struct ext4_sb_info *sbi = EXT4_SB(sb);
4241 journal->j_commit_interval = sbi->s_commit_interval;
4242 journal->j_min_batch_time = sbi->s_min_batch_time;
4243 journal->j_max_batch_time = sbi->s_max_batch_time;
4245 write_lock(&journal->j_state_lock);
4246 if (test_opt(sb, BARRIER))
4247 journal->j_flags |= JBD2_BARRIER;
4249 journal->j_flags &= ~JBD2_BARRIER;
4250 if (test_opt(sb, DATA_ERR_ABORT))
4251 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4253 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4254 write_unlock(&journal->j_state_lock);
4257 static struct inode *ext4_get_journal_inode(struct super_block *sb,
4258 unsigned int journal_inum)
4260 struct inode *journal_inode;
4263 * Test for the existence of a valid inode on disk. Bad things
4264 * happen if we iget() an unused inode, as the subsequent iput()
4265 * will try to delete it.
4267 journal_inode = ext4_iget(sb, journal_inum);
4268 if (IS_ERR(journal_inode)) {
4269 ext4_msg(sb, KERN_ERR, "no journal found");
4272 if (!journal_inode->i_nlink) {
4273 make_bad_inode(journal_inode);
4274 iput(journal_inode);
4275 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4279 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4280 journal_inode, journal_inode->i_size);
4281 if (!S_ISREG(journal_inode->i_mode)) {
4282 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4283 iput(journal_inode);
4286 return journal_inode;
4289 static journal_t *ext4_get_journal(struct super_block *sb,
4290 unsigned int journal_inum)
4292 struct inode *journal_inode;
4295 BUG_ON(!ext4_has_feature_journal(sb));
4297 journal_inode = ext4_get_journal_inode(sb, journal_inum);
4301 journal = jbd2_journal_init_inode(journal_inode);
4303 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4304 iput(journal_inode);
4307 journal->j_private = sb;
4308 ext4_init_journal_params(sb, journal);
4312 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4315 struct buffer_head *bh;
4319 int hblock, blocksize;
4320 ext4_fsblk_t sb_block;
4321 unsigned long offset;
4322 struct ext4_super_block *es;
4323 struct block_device *bdev;
4325 BUG_ON(!ext4_has_feature_journal(sb));
4327 bdev = ext4_blkdev_get(j_dev, sb);
4331 blocksize = sb->s_blocksize;
4332 hblock = bdev_logical_block_size(bdev);
4333 if (blocksize < hblock) {
4334 ext4_msg(sb, KERN_ERR,
4335 "blocksize too small for journal device");
4339 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4340 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4341 set_blocksize(bdev, blocksize);
4342 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4343 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4344 "external journal");
4348 es = (struct ext4_super_block *) (bh->b_data + offset);
4349 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4350 !(le32_to_cpu(es->s_feature_incompat) &
4351 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4352 ext4_msg(sb, KERN_ERR, "external journal has "
4358 if ((le32_to_cpu(es->s_feature_ro_compat) &
4359 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4360 es->s_checksum != ext4_superblock_csum(sb, es)) {
4361 ext4_msg(sb, KERN_ERR, "external journal has "
4362 "corrupt superblock");
4367 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4368 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4373 len = ext4_blocks_count(es);
4374 start = sb_block + 1;
4375 brelse(bh); /* we're done with the superblock */
4377 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4378 start, len, blocksize);
4380 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4383 journal->j_private = sb;
4384 ll_rw_block(REQ_OP_READ, REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4385 wait_on_buffer(journal->j_sb_buffer);
4386 if (!buffer_uptodate(journal->j_sb_buffer)) {
4387 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4390 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4391 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4392 "user (unsupported) - %d",
4393 be32_to_cpu(journal->j_superblock->s_nr_users));
4396 EXT4_SB(sb)->journal_bdev = bdev;
4397 ext4_init_journal_params(sb, journal);
4401 jbd2_journal_destroy(journal);
4403 ext4_blkdev_put(bdev);
4407 static int ext4_load_journal(struct super_block *sb,
4408 struct ext4_super_block *es,
4409 unsigned long journal_devnum)
4412 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4415 int really_read_only;
4417 BUG_ON(!ext4_has_feature_journal(sb));
4419 if (journal_devnum &&
4420 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4421 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4422 "numbers have changed");
4423 journal_dev = new_decode_dev(journal_devnum);
4425 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4427 really_read_only = bdev_read_only(sb->s_bdev);
4430 * Are we loading a blank journal or performing recovery after a
4431 * crash? For recovery, we need to check in advance whether we
4432 * can get read-write access to the device.
4434 if (ext4_has_feature_journal_needs_recovery(sb)) {
4435 if (sb->s_flags & MS_RDONLY) {
4436 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4437 "required on readonly filesystem");
4438 if (really_read_only) {
4439 ext4_msg(sb, KERN_ERR, "write access "
4440 "unavailable, cannot proceed");
4443 ext4_msg(sb, KERN_INFO, "write access will "
4444 "be enabled during recovery");
4448 if (journal_inum && journal_dev) {
4449 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4450 "and inode journals!");
4455 if (!(journal = ext4_get_journal(sb, journal_inum)))
4458 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4462 if (!(journal->j_flags & JBD2_BARRIER))
4463 ext4_msg(sb, KERN_INFO, "barriers disabled");
4465 if (!ext4_has_feature_journal_needs_recovery(sb))
4466 err = jbd2_journal_wipe(journal, !really_read_only);
4468 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4470 memcpy(save, ((char *) es) +
4471 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4472 err = jbd2_journal_load(journal);
4474 memcpy(((char *) es) + EXT4_S_ERR_START,
4475 save, EXT4_S_ERR_LEN);
4480 ext4_msg(sb, KERN_ERR, "error loading journal");
4481 jbd2_journal_destroy(journal);
4485 EXT4_SB(sb)->s_journal = journal;
4486 ext4_clear_journal_err(sb, es);
4488 if (!really_read_only && journal_devnum &&
4489 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4490 es->s_journal_dev = cpu_to_le32(journal_devnum);
4492 /* Make sure we flush the recovery flag to disk. */
4493 ext4_commit_super(sb, 1);
4499 static int ext4_commit_super(struct super_block *sb, int sync)
4501 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4502 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4505 if (!sbh || block_device_ejected(sb))
4508 * If the file system is mounted read-only, don't update the
4509 * superblock write time. This avoids updating the superblock
4510 * write time when we are mounting the root file system
4511 * read/only but we need to replay the journal; at that point,
4512 * for people who are east of GMT and who make their clock
4513 * tick in localtime for Windows bug-for-bug compatibility,
4514 * the clock is set in the future, and this will cause e2fsck
4515 * to complain and force a full file system check.
4517 if (!(sb->s_flags & MS_RDONLY))
4518 es->s_wtime = cpu_to_le32(get_seconds());
4519 if (sb->s_bdev->bd_part)
4520 es->s_kbytes_written =
4521 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4522 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4523 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4525 es->s_kbytes_written =
4526 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4527 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4528 ext4_free_blocks_count_set(es,
4529 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4530 &EXT4_SB(sb)->s_freeclusters_counter)));
4531 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4532 es->s_free_inodes_count =
4533 cpu_to_le32(percpu_counter_sum_positive(
4534 &EXT4_SB(sb)->s_freeinodes_counter));
4535 BUFFER_TRACE(sbh, "marking dirty");
4536 ext4_superblock_csum_set(sb);
4538 if (buffer_write_io_error(sbh)) {
4540 * Oh, dear. A previous attempt to write the
4541 * superblock failed. This could happen because the
4542 * USB device was yanked out. Or it could happen to
4543 * be a transient write error and maybe the block will
4544 * be remapped. Nothing we can do but to retry the
4545 * write and hope for the best.
4547 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4548 "superblock detected");
4549 clear_buffer_write_io_error(sbh);
4550 set_buffer_uptodate(sbh);
4552 mark_buffer_dirty(sbh);
4555 error = __sync_dirty_buffer(sbh,
4556 test_opt(sb, BARRIER) ? WRITE_FUA : WRITE_SYNC);
4560 error = buffer_write_io_error(sbh);
4562 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4564 clear_buffer_write_io_error(sbh);
4565 set_buffer_uptodate(sbh);
4572 * Have we just finished recovery? If so, and if we are mounting (or
4573 * remounting) the filesystem readonly, then we will end up with a
4574 * consistent fs on disk. Record that fact.
4576 static void ext4_mark_recovery_complete(struct super_block *sb,
4577 struct ext4_super_block *es)
4579 journal_t *journal = EXT4_SB(sb)->s_journal;
4581 if (!ext4_has_feature_journal(sb)) {
4582 BUG_ON(journal != NULL);
4585 jbd2_journal_lock_updates(journal);
4586 if (jbd2_journal_flush(journal) < 0)
4589 if (ext4_has_feature_journal_needs_recovery(sb) &&
4590 sb->s_flags & MS_RDONLY) {
4591 ext4_clear_feature_journal_needs_recovery(sb);
4592 ext4_commit_super(sb, 1);
4596 jbd2_journal_unlock_updates(journal);
4600 * If we are mounting (or read-write remounting) a filesystem whose journal
4601 * has recorded an error from a previous lifetime, move that error to the
4602 * main filesystem now.
4604 static void ext4_clear_journal_err(struct super_block *sb,
4605 struct ext4_super_block *es)
4611 BUG_ON(!ext4_has_feature_journal(sb));
4613 journal = EXT4_SB(sb)->s_journal;
4616 * Now check for any error status which may have been recorded in the
4617 * journal by a prior ext4_error() or ext4_abort()
4620 j_errno = jbd2_journal_errno(journal);
4624 errstr = ext4_decode_error(sb, j_errno, nbuf);
4625 ext4_warning(sb, "Filesystem error recorded "
4626 "from previous mount: %s", errstr);
4627 ext4_warning(sb, "Marking fs in need of filesystem check.");
4629 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4630 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4631 ext4_commit_super(sb, 1);
4633 jbd2_journal_clear_err(journal);
4634 jbd2_journal_update_sb_errno(journal);
4639 * Force the running and committing transactions to commit,
4640 * and wait on the commit.
4642 int ext4_force_commit(struct super_block *sb)
4646 if (sb->s_flags & MS_RDONLY)
4649 journal = EXT4_SB(sb)->s_journal;
4650 return ext4_journal_force_commit(journal);
4653 static int ext4_sync_fs(struct super_block *sb, int wait)
4657 bool needs_barrier = false;
4658 struct ext4_sb_info *sbi = EXT4_SB(sb);
4660 trace_ext4_sync_fs(sb, wait);
4661 flush_workqueue(sbi->rsv_conversion_wq);
4663 * Writeback quota in non-journalled quota case - journalled quota has
4666 dquot_writeback_dquots(sb, -1);
4668 * Data writeback is possible w/o journal transaction, so barrier must
4669 * being sent at the end of the function. But we can skip it if
4670 * transaction_commit will do it for us.
4672 if (sbi->s_journal) {
4673 target = jbd2_get_latest_transaction(sbi->s_journal);
4674 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4675 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4676 needs_barrier = true;
4678 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4680 ret = jbd2_log_wait_commit(sbi->s_journal,
4683 } else if (wait && test_opt(sb, BARRIER))
4684 needs_barrier = true;
4685 if (needs_barrier) {
4687 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4696 * LVM calls this function before a (read-only) snapshot is created. This
4697 * gives us a chance to flush the journal completely and mark the fs clean.
4699 * Note that only this function cannot bring a filesystem to be in a clean
4700 * state independently. It relies on upper layer to stop all data & metadata
4703 static int ext4_freeze(struct super_block *sb)
4708 if (sb->s_flags & MS_RDONLY)
4711 journal = EXT4_SB(sb)->s_journal;
4714 /* Now we set up the journal barrier. */
4715 jbd2_journal_lock_updates(journal);
4718 * Don't clear the needs_recovery flag if we failed to
4719 * flush the journal.
4721 error = jbd2_journal_flush(journal);
4725 /* Journal blocked and flushed, clear needs_recovery flag. */
4726 ext4_clear_feature_journal_needs_recovery(sb);
4729 error = ext4_commit_super(sb, 1);
4732 /* we rely on upper layer to stop further updates */
4733 jbd2_journal_unlock_updates(journal);
4738 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4739 * flag here, even though the filesystem is not technically dirty yet.
4741 static int ext4_unfreeze(struct super_block *sb)
4743 if (sb->s_flags & MS_RDONLY)
4746 if (EXT4_SB(sb)->s_journal) {
4747 /* Reset the needs_recovery flag before the fs is unlocked. */
4748 ext4_set_feature_journal_needs_recovery(sb);
4751 ext4_commit_super(sb, 1);
4756 * Structure to save mount options for ext4_remount's benefit
4758 struct ext4_mount_options {
4759 unsigned long s_mount_opt;
4760 unsigned long s_mount_opt2;
4763 unsigned long s_commit_interval;
4764 u32 s_min_batch_time, s_max_batch_time;
4767 char *s_qf_names[EXT4_MAXQUOTAS];
4771 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4773 struct ext4_super_block *es;
4774 struct ext4_sb_info *sbi = EXT4_SB(sb);
4775 unsigned long old_sb_flags;
4776 struct ext4_mount_options old_opts;
4777 int enable_quota = 0;
4779 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4784 char *orig_data = kstrdup(data, GFP_KERNEL);
4786 /* Store the original options */
4787 old_sb_flags = sb->s_flags;
4788 old_opts.s_mount_opt = sbi->s_mount_opt;
4789 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4790 old_opts.s_resuid = sbi->s_resuid;
4791 old_opts.s_resgid = sbi->s_resgid;
4792 old_opts.s_commit_interval = sbi->s_commit_interval;
4793 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4794 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4796 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4797 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4798 if (sbi->s_qf_names[i]) {
4799 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4801 if (!old_opts.s_qf_names[i]) {
4802 for (j = 0; j < i; j++)
4803 kfree(old_opts.s_qf_names[j]);
4808 old_opts.s_qf_names[i] = NULL;
4810 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4811 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4813 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4818 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
4819 test_opt(sb, JOURNAL_CHECKSUM)) {
4820 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
4821 "during remount not supported; ignoring");
4822 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
4825 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4826 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4827 ext4_msg(sb, KERN_ERR, "can't mount with "
4828 "both data=journal and delalloc");
4832 if (test_opt(sb, DIOREAD_NOLOCK)) {
4833 ext4_msg(sb, KERN_ERR, "can't mount with "
4834 "both data=journal and dioread_nolock");
4838 if (test_opt(sb, DAX)) {
4839 ext4_msg(sb, KERN_ERR, "can't mount with "
4840 "both data=journal and dax");
4846 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
4847 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
4848 "dax flag with busy inodes while remounting");
4849 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
4852 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4853 ext4_abort(sb, "Abort forced by user");
4855 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4856 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4860 if (sbi->s_journal) {
4861 ext4_init_journal_params(sb, sbi->s_journal);
4862 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4865 if (*flags & MS_LAZYTIME)
4866 sb->s_flags |= MS_LAZYTIME;
4868 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4869 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4874 if (*flags & MS_RDONLY) {
4875 err = sync_filesystem(sb);
4878 err = dquot_suspend(sb, -1);
4883 * First of all, the unconditional stuff we have to do
4884 * to disable replay of the journal when we next remount
4886 sb->s_flags |= MS_RDONLY;
4889 * OK, test if we are remounting a valid rw partition
4890 * readonly, and if so set the rdonly flag and then
4891 * mark the partition as valid again.
4893 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4894 (sbi->s_mount_state & EXT4_VALID_FS))
4895 es->s_state = cpu_to_le16(sbi->s_mount_state);
4898 ext4_mark_recovery_complete(sb, es);
4900 /* Make sure we can mount this feature set readwrite */
4901 if (ext4_has_feature_readonly(sb) ||
4902 !ext4_feature_set_ok(sb, 0)) {
4907 * Make sure the group descriptor checksums
4908 * are sane. If they aren't, refuse to remount r/w.
4910 for (g = 0; g < sbi->s_groups_count; g++) {
4911 struct ext4_group_desc *gdp =
4912 ext4_get_group_desc(sb, g, NULL);
4914 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
4915 ext4_msg(sb, KERN_ERR,
4916 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4917 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
4918 le16_to_cpu(gdp->bg_checksum));
4925 * If we have an unprocessed orphan list hanging
4926 * around from a previously readonly bdev mount,
4927 * require a full umount/remount for now.
4929 if (es->s_last_orphan) {
4930 ext4_msg(sb, KERN_WARNING, "Couldn't "
4931 "remount RDWR because of unprocessed "
4932 "orphan inode list. Please "
4933 "umount/remount instead");
4939 * Mounting a RDONLY partition read-write, so reread
4940 * and store the current valid flag. (It may have
4941 * been changed by e2fsck since we originally mounted
4945 ext4_clear_journal_err(sb, es);
4946 sbi->s_mount_state = le16_to_cpu(es->s_state);
4947 if (!ext4_setup_super(sb, es, 0))
4948 sb->s_flags &= ~MS_RDONLY;
4949 if (ext4_has_feature_mmp(sb))
4950 if (ext4_multi_mount_protect(sb,
4951 le64_to_cpu(es->s_mmp_block))) {
4960 * Reinitialize lazy itable initialization thread based on
4963 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4964 ext4_unregister_li_request(sb);
4966 ext4_group_t first_not_zeroed;
4967 first_not_zeroed = ext4_has_uninit_itable(sb);
4968 ext4_register_li_request(sb, first_not_zeroed);
4971 ext4_setup_system_zone(sb);
4972 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
4973 ext4_commit_super(sb, 1);
4976 /* Release old quota file names */
4977 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4978 kfree(old_opts.s_qf_names[i]);
4980 if (sb_any_quota_suspended(sb))
4981 dquot_resume(sb, -1);
4982 else if (ext4_has_feature_quota(sb)) {
4983 err = ext4_enable_quotas(sb);
4990 *flags = (*flags & ~MS_LAZYTIME) | (sb->s_flags & MS_LAZYTIME);
4991 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4996 sb->s_flags = old_sb_flags;
4997 sbi->s_mount_opt = old_opts.s_mount_opt;
4998 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
4999 sbi->s_resuid = old_opts.s_resuid;
5000 sbi->s_resgid = old_opts.s_resgid;
5001 sbi->s_commit_interval = old_opts.s_commit_interval;
5002 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5003 sbi->s_max_batch_time = old_opts.s_max_batch_time;
5005 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
5006 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
5007 kfree(sbi->s_qf_names[i]);
5008 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
5016 static int ext4_statfs_project(struct super_block *sb,
5017 kprojid_t projid, struct kstatfs *buf)
5020 struct dquot *dquot;
5024 qid = make_kqid_projid(projid);
5025 dquot = dqget(sb, qid);
5027 return PTR_ERR(dquot);
5028 spin_lock(&dq_data_lock);
5030 limit = (dquot->dq_dqb.dqb_bsoftlimit ?
5031 dquot->dq_dqb.dqb_bsoftlimit :
5032 dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits;
5033 if (limit && buf->f_blocks > limit) {
5034 curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits;
5035 buf->f_blocks = limit;
5036 buf->f_bfree = buf->f_bavail =
5037 (buf->f_blocks > curblock) ?
5038 (buf->f_blocks - curblock) : 0;
5041 limit = dquot->dq_dqb.dqb_isoftlimit ?
5042 dquot->dq_dqb.dqb_isoftlimit :
5043 dquot->dq_dqb.dqb_ihardlimit;
5044 if (limit && buf->f_files > limit) {
5045 buf->f_files = limit;
5047 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
5048 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
5051 spin_unlock(&dq_data_lock);
5057 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
5059 struct super_block *sb = dentry->d_sb;
5060 struct ext4_sb_info *sbi = EXT4_SB(sb);
5061 struct ext4_super_block *es = sbi->s_es;
5062 ext4_fsblk_t overhead = 0, resv_blocks;
5065 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
5067 if (!test_opt(sb, MINIX_DF))
5068 overhead = sbi->s_overhead;
5070 buf->f_type = EXT4_SUPER_MAGIC;
5071 buf->f_bsize = sb->s_blocksize;
5072 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
5073 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5074 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
5075 /* prevent underflow in case that few free space is available */
5076 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
5077 buf->f_bavail = buf->f_bfree -
5078 (ext4_r_blocks_count(es) + resv_blocks);
5079 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
5081 buf->f_files = le32_to_cpu(es->s_inodes_count);
5082 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5083 buf->f_namelen = EXT4_NAME_LEN;
5084 fsid = le64_to_cpup((void *)es->s_uuid) ^
5085 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5086 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5087 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
5090 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
5091 sb_has_quota_limits_enabled(sb, PRJQUOTA))
5092 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
5097 /* Helper function for writing quotas on sync - we need to start transaction
5098 * before quota file is locked for write. Otherwise the are possible deadlocks:
5099 * Process 1 Process 2
5100 * ext4_create() quota_sync()
5101 * jbd2_journal_start() write_dquot()
5102 * dquot_initialize() down(dqio_mutex)
5103 * down(dqio_mutex) jbd2_journal_start()
5109 static inline struct inode *dquot_to_inode(struct dquot *dquot)
5111 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
5114 static int ext4_write_dquot(struct dquot *dquot)
5118 struct inode *inode;
5120 inode = dquot_to_inode(dquot);
5121 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
5122 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
5124 return PTR_ERR(handle);
5125 ret = dquot_commit(dquot);
5126 err = ext4_journal_stop(handle);
5132 static int ext4_acquire_dquot(struct dquot *dquot)
5137 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5138 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
5140 return PTR_ERR(handle);
5141 ret = dquot_acquire(dquot);
5142 err = ext4_journal_stop(handle);
5148 static int ext4_release_dquot(struct dquot *dquot)
5153 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5154 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
5155 if (IS_ERR(handle)) {
5156 /* Release dquot anyway to avoid endless cycle in dqput() */
5157 dquot_release(dquot);
5158 return PTR_ERR(handle);
5160 ret = dquot_release(dquot);
5161 err = ext4_journal_stop(handle);
5167 static int ext4_mark_dquot_dirty(struct dquot *dquot)
5169 struct super_block *sb = dquot->dq_sb;
5170 struct ext4_sb_info *sbi = EXT4_SB(sb);
5172 /* Are we journaling quotas? */
5173 if (ext4_has_feature_quota(sb) ||
5174 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
5175 dquot_mark_dquot_dirty(dquot);
5176 return ext4_write_dquot(dquot);
5178 return dquot_mark_dquot_dirty(dquot);
5182 static int ext4_write_info(struct super_block *sb, int type)
5187 /* Data block + inode block */
5188 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
5190 return PTR_ERR(handle);
5191 ret = dquot_commit_info(sb, type);
5192 err = ext4_journal_stop(handle);
5199 * Turn on quotas during mount time - we need to find
5200 * the quota file and such...
5202 static int ext4_quota_on_mount(struct super_block *sb, int type)
5204 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5205 EXT4_SB(sb)->s_jquota_fmt, type);
5208 static void lockdep_set_quota_inode(struct inode *inode, int subclass)
5210 struct ext4_inode_info *ei = EXT4_I(inode);
5212 /* The first argument of lockdep_set_subclass has to be
5213 * *exactly* the same as the argument to init_rwsem() --- in
5214 * this case, in init_once() --- or lockdep gets unhappy
5215 * because the name of the lock is set using the
5216 * stringification of the argument to init_rwsem().
5218 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
5219 lockdep_set_subclass(&ei->i_data_sem, subclass);
5223 * Standard function to be called on quota_on
5225 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
5230 if (!test_opt(sb, QUOTA))
5233 /* Quotafile not on the same filesystem? */
5234 if (path->dentry->d_sb != sb)
5236 /* Journaling quota? */
5237 if (EXT4_SB(sb)->s_qf_names[type]) {
5238 /* Quotafile not in fs root? */
5239 if (path->dentry->d_parent != sb->s_root)
5240 ext4_msg(sb, KERN_WARNING,
5241 "Quota file not on filesystem root. "
5242 "Journaled quota will not work");
5246 * When we journal data on quota file, we have to flush journal to see
5247 * all updates to the file when we bypass pagecache...
5249 if (EXT4_SB(sb)->s_journal &&
5250 ext4_should_journal_data(d_inode(path->dentry))) {
5252 * We don't need to lock updates but journal_flush() could
5253 * otherwise be livelocked...
5255 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
5256 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
5257 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
5261 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
5262 err = dquot_quota_on(sb, type, format_id, path);
5264 lockdep_set_quota_inode(path->dentry->d_inode,
5269 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5273 struct inode *qf_inode;
5274 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5275 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5276 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5277 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
5280 BUG_ON(!ext4_has_feature_quota(sb));
5282 if (!qf_inums[type])
5285 qf_inode = ext4_iget(sb, qf_inums[type]);
5286 if (IS_ERR(qf_inode)) {
5287 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5288 return PTR_ERR(qf_inode);
5291 /* Don't account quota for quota files to avoid recursion */
5292 qf_inode->i_flags |= S_NOQUOTA;
5293 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
5294 err = dquot_enable(qf_inode, type, format_id, flags);
5297 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
5302 /* Enable usage tracking for all quota types. */
5303 static int ext4_enable_quotas(struct super_block *sb)
5306 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5307 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5308 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5309 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
5311 bool quota_mopt[EXT4_MAXQUOTAS] = {
5312 test_opt(sb, USRQUOTA),
5313 test_opt(sb, GRPQUOTA),
5314 test_opt(sb, PRJQUOTA),
5317 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5318 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
5319 if (qf_inums[type]) {
5320 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5321 DQUOT_USAGE_ENABLED |
5322 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
5325 "Failed to enable quota tracking "
5326 "(type=%d, err=%d). Please run "
5327 "e2fsck to fix.", type, err);
5335 static int ext4_quota_off(struct super_block *sb, int type)
5337 struct inode *inode = sb_dqopt(sb)->files[type];
5340 /* Force all delayed allocation blocks to be allocated.
5341 * Caller already holds s_umount sem */
5342 if (test_opt(sb, DELALLOC))
5343 sync_filesystem(sb);
5348 /* Update modification times of quota files when userspace can
5349 * start looking at them */
5350 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5353 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5354 ext4_mark_inode_dirty(handle, inode);
5355 ext4_journal_stop(handle);
5358 return dquot_quota_off(sb, type);
5361 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5362 * acquiring the locks... As quota files are never truncated and quota code
5363 * itself serializes the operations (and no one else should touch the files)
5364 * we don't have to be afraid of races */
5365 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5366 size_t len, loff_t off)
5368 struct inode *inode = sb_dqopt(sb)->files[type];
5369 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5370 int offset = off & (sb->s_blocksize - 1);
5373 struct buffer_head *bh;
5374 loff_t i_size = i_size_read(inode);
5378 if (off+len > i_size)
5381 while (toread > 0) {
5382 tocopy = sb->s_blocksize - offset < toread ?
5383 sb->s_blocksize - offset : toread;
5384 bh = ext4_bread(NULL, inode, blk, 0);
5387 if (!bh) /* A hole? */
5388 memset(data, 0, tocopy);
5390 memcpy(data, bh->b_data+offset, tocopy);
5400 /* Write to quotafile (we know the transaction is already started and has
5401 * enough credits) */
5402 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5403 const char *data, size_t len, loff_t off)
5405 struct inode *inode = sb_dqopt(sb)->files[type];
5406 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5407 int err, offset = off & (sb->s_blocksize - 1);
5409 struct buffer_head *bh;
5410 handle_t *handle = journal_current_handle();
5412 if (EXT4_SB(sb)->s_journal && !handle) {
5413 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5414 " cancelled because transaction is not started",
5415 (unsigned long long)off, (unsigned long long)len);
5419 * Since we account only one data block in transaction credits,
5420 * then it is impossible to cross a block boundary.
5422 if (sb->s_blocksize - offset < len) {
5423 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5424 " cancelled because not block aligned",
5425 (unsigned long long)off, (unsigned long long)len);
5430 bh = ext4_bread(handle, inode, blk,
5431 EXT4_GET_BLOCKS_CREATE |
5432 EXT4_GET_BLOCKS_METADATA_NOFAIL);
5433 } while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) &&
5434 ext4_should_retry_alloc(inode->i_sb, &retries));
5439 BUFFER_TRACE(bh, "get write access");
5440 err = ext4_journal_get_write_access(handle, bh);
5446 memcpy(bh->b_data+offset, data, len);
5447 flush_dcache_page(bh->b_page);
5449 err = ext4_handle_dirty_metadata(handle, NULL, bh);
5452 if (inode->i_size < off + len) {
5453 i_size_write(inode, off + len);
5454 EXT4_I(inode)->i_disksize = inode->i_size;
5455 ext4_mark_inode_dirty(handle, inode);
5460 static int ext4_get_next_id(struct super_block *sb, struct kqid *qid)
5462 const struct quota_format_ops *ops;
5464 if (!sb_has_quota_loaded(sb, qid->type))
5466 ops = sb_dqopt(sb)->ops[qid->type];
5467 if (!ops || !ops->get_next_id)
5469 return dquot_get_next_id(sb, qid);
5473 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5474 const char *dev_name, void *data)
5476 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5479 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
5480 static inline void register_as_ext2(void)
5482 int err = register_filesystem(&ext2_fs_type);
5485 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5488 static inline void unregister_as_ext2(void)
5490 unregister_filesystem(&ext2_fs_type);
5493 static inline int ext2_feature_set_ok(struct super_block *sb)
5495 if (ext4_has_unknown_ext2_incompat_features(sb))
5497 if (sb->s_flags & MS_RDONLY)
5499 if (ext4_has_unknown_ext2_ro_compat_features(sb))
5504 static inline void register_as_ext2(void) { }
5505 static inline void unregister_as_ext2(void) { }
5506 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5509 static inline void register_as_ext3(void)
5511 int err = register_filesystem(&ext3_fs_type);
5514 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5517 static inline void unregister_as_ext3(void)
5519 unregister_filesystem(&ext3_fs_type);
5522 static inline int ext3_feature_set_ok(struct super_block *sb)
5524 if (ext4_has_unknown_ext3_incompat_features(sb))
5526 if (!ext4_has_feature_journal(sb))
5528 if (sb->s_flags & MS_RDONLY)
5530 if (ext4_has_unknown_ext3_ro_compat_features(sb))
5535 static struct file_system_type ext4_fs_type = {
5536 .owner = THIS_MODULE,
5538 .mount = ext4_mount,
5539 .kill_sb = kill_block_super,
5540 .fs_flags = FS_REQUIRES_DEV,
5542 MODULE_ALIAS_FS("ext4");
5544 /* Shared across all ext4 file systems */
5545 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5547 static int __init ext4_init_fs(void)
5551 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
5552 ext4_li_info = NULL;
5553 mutex_init(&ext4_li_mtx);
5555 /* Build-time check for flags consistency */
5556 ext4_check_flag_values();
5558 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
5559 init_waitqueue_head(&ext4__ioend_wq[i]);
5561 err = ext4_init_es();
5565 err = ext4_init_pageio();
5569 err = ext4_init_system_zone();
5573 err = ext4_init_sysfs();
5577 err = ext4_init_mballoc();
5580 err = init_inodecache();
5585 err = register_filesystem(&ext4_fs_type);
5591 unregister_as_ext2();
5592 unregister_as_ext3();
5593 destroy_inodecache();
5595 ext4_exit_mballoc();
5599 ext4_exit_system_zone();
5608 static void __exit ext4_exit_fs(void)
5610 ext4_destroy_lazyinit_thread();
5611 unregister_as_ext2();
5612 unregister_as_ext3();
5613 unregister_filesystem(&ext4_fs_type);
5614 destroy_inodecache();
5615 ext4_exit_mballoc();
5617 ext4_exit_system_zone();
5622 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5623 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5624 MODULE_LICENSE("GPL");
5625 module_init(ext4_init_fs)
5626 module_exit(ext4_exit_fs)