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 <linux/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 (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
444 if (ext4_error_ratelimit(sb)) {
449 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
450 sb->s_id, function, line, current->comm, &vaf);
453 save_error_info(sb, function, line);
454 ext4_handle_error(sb);
457 void __ext4_error_inode(struct inode *inode, const char *function,
458 unsigned int line, ext4_fsblk_t block,
459 const char *fmt, ...)
462 struct va_format vaf;
463 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
465 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
468 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
469 es->s_last_error_block = cpu_to_le64(block);
470 if (ext4_error_ratelimit(inode->i_sb)) {
475 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
476 "inode #%lu: block %llu: comm %s: %pV\n",
477 inode->i_sb->s_id, function, line, inode->i_ino,
478 block, current->comm, &vaf);
480 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
481 "inode #%lu: comm %s: %pV\n",
482 inode->i_sb->s_id, function, line, inode->i_ino,
483 current->comm, &vaf);
486 save_error_info(inode->i_sb, function, line);
487 ext4_handle_error(inode->i_sb);
490 void __ext4_error_file(struct file *file, const char *function,
491 unsigned int line, ext4_fsblk_t block,
492 const char *fmt, ...)
495 struct va_format vaf;
496 struct ext4_super_block *es;
497 struct inode *inode = file_inode(file);
498 char pathname[80], *path;
500 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
503 es = EXT4_SB(inode->i_sb)->s_es;
504 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
505 if (ext4_error_ratelimit(inode->i_sb)) {
506 path = file_path(file, pathname, sizeof(pathname));
514 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
515 "block %llu: comm %s: path %s: %pV\n",
516 inode->i_sb->s_id, function, line, inode->i_ino,
517 block, current->comm, path, &vaf);
520 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
521 "comm %s: path %s: %pV\n",
522 inode->i_sb->s_id, function, line, inode->i_ino,
523 current->comm, path, &vaf);
526 save_error_info(inode->i_sb, function, line);
527 ext4_handle_error(inode->i_sb);
530 const char *ext4_decode_error(struct super_block *sb, int errno,
537 errstr = "Corrupt filesystem";
540 errstr = "Filesystem failed CRC";
543 errstr = "IO failure";
546 errstr = "Out of memory";
549 if (!sb || (EXT4_SB(sb)->s_journal &&
550 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
551 errstr = "Journal has aborted";
553 errstr = "Readonly filesystem";
556 /* If the caller passed in an extra buffer for unknown
557 * errors, textualise them now. Else we just return
560 /* Check for truncated error codes... */
561 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
570 /* __ext4_std_error decodes expected errors from journaling functions
571 * automatically and invokes the appropriate error response. */
573 void __ext4_std_error(struct super_block *sb, const char *function,
574 unsigned int line, int errno)
579 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
582 /* Special case: if the error is EROFS, and we're not already
583 * inside a transaction, then there's really no point in logging
585 if (errno == -EROFS && journal_current_handle() == NULL &&
586 (sb->s_flags & MS_RDONLY))
589 if (ext4_error_ratelimit(sb)) {
590 errstr = ext4_decode_error(sb, errno, nbuf);
591 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
592 sb->s_id, function, line, errstr);
595 save_error_info(sb, function, line);
596 ext4_handle_error(sb);
600 * ext4_abort is a much stronger failure handler than ext4_error. The
601 * abort function may be used to deal with unrecoverable failures such
602 * as journal IO errors or ENOMEM at a critical moment in log management.
604 * We unconditionally force the filesystem into an ABORT|READONLY state,
605 * unless the error response on the fs has been set to panic in which
606 * case we take the easy way out and panic immediately.
609 void __ext4_abort(struct super_block *sb, const char *function,
610 unsigned int line, const char *fmt, ...)
612 struct va_format vaf;
615 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
618 save_error_info(sb, function, line);
622 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: %pV\n",
623 sb->s_id, function, line, &vaf);
626 if ((sb->s_flags & MS_RDONLY) == 0) {
627 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
628 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
630 * Make sure updated value of ->s_mount_flags will be visible
631 * before ->s_flags update
634 sb->s_flags |= MS_RDONLY;
635 if (EXT4_SB(sb)->s_journal)
636 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
637 save_error_info(sb, function, line);
639 if (test_opt(sb, ERRORS_PANIC)) {
640 if (EXT4_SB(sb)->s_journal &&
641 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
643 panic("EXT4-fs panic from previous error\n");
647 void __ext4_msg(struct super_block *sb,
648 const char *prefix, const char *fmt, ...)
650 struct va_format vaf;
653 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
659 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
663 #define ext4_warning_ratelimit(sb) \
664 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \
667 void __ext4_warning(struct super_block *sb, const char *function,
668 unsigned int line, const char *fmt, ...)
670 struct va_format vaf;
673 if (!ext4_warning_ratelimit(sb))
679 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
680 sb->s_id, function, line, &vaf);
684 void __ext4_warning_inode(const struct inode *inode, const char *function,
685 unsigned int line, const char *fmt, ...)
687 struct va_format vaf;
690 if (!ext4_warning_ratelimit(inode->i_sb))
696 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
697 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
698 function, line, inode->i_ino, current->comm, &vaf);
702 void __ext4_grp_locked_error(const char *function, unsigned int line,
703 struct super_block *sb, ext4_group_t grp,
704 unsigned long ino, ext4_fsblk_t block,
705 const char *fmt, ...)
709 struct va_format vaf;
711 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
713 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
716 es->s_last_error_ino = cpu_to_le32(ino);
717 es->s_last_error_block = cpu_to_le64(block);
718 __save_error_info(sb, function, line);
720 if (ext4_error_ratelimit(sb)) {
724 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
725 sb->s_id, function, line, grp);
727 printk(KERN_CONT "inode %lu: ", ino);
729 printk(KERN_CONT "block %llu:",
730 (unsigned long long) block);
731 printk(KERN_CONT "%pV\n", &vaf);
735 if (test_opt(sb, ERRORS_CONT)) {
736 ext4_commit_super(sb, 0);
740 ext4_unlock_group(sb, grp);
741 ext4_handle_error(sb);
743 * We only get here in the ERRORS_RO case; relocking the group
744 * may be dangerous, but nothing bad will happen since the
745 * filesystem will have already been marked read/only and the
746 * journal has been aborted. We return 1 as a hint to callers
747 * who might what to use the return value from
748 * ext4_grp_locked_error() to distinguish between the
749 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
750 * aggressively from the ext4 function in question, with a
751 * more appropriate error code.
753 ext4_lock_group(sb, grp);
757 void ext4_update_dynamic_rev(struct super_block *sb)
759 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
761 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
765 "updating to rev %d because of new feature flag, "
766 "running e2fsck is recommended",
769 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
770 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
771 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
772 /* leave es->s_feature_*compat flags alone */
773 /* es->s_uuid will be set by e2fsck if empty */
776 * The rest of the superblock fields should be zero, and if not it
777 * means they are likely already in use, so leave them alone. We
778 * can leave it up to e2fsck to clean up any inconsistencies there.
783 * Open the external journal device
785 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
787 struct block_device *bdev;
788 char b[BDEVNAME_SIZE];
790 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
796 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
797 __bdevname(dev, b), PTR_ERR(bdev));
802 * Release the journal device
804 static void ext4_blkdev_put(struct block_device *bdev)
806 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
809 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
811 struct block_device *bdev;
812 bdev = sbi->journal_bdev;
814 ext4_blkdev_put(bdev);
815 sbi->journal_bdev = NULL;
819 static inline struct inode *orphan_list_entry(struct list_head *l)
821 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
824 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
828 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
829 le32_to_cpu(sbi->s_es->s_last_orphan));
831 printk(KERN_ERR "sb_info orphan list:\n");
832 list_for_each(l, &sbi->s_orphan) {
833 struct inode *inode = orphan_list_entry(l);
835 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
836 inode->i_sb->s_id, inode->i_ino, inode,
837 inode->i_mode, inode->i_nlink,
843 static int ext4_quota_off(struct super_block *sb, int type);
845 static inline void ext4_quota_off_umount(struct super_block *sb)
849 if (ext4_has_feature_quota(sb)) {
850 dquot_disable(sb, -1,
851 DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
853 /* Use our quota_off function to clear inode flags etc. */
854 for (type = 0; type < EXT4_MAXQUOTAS; type++)
855 ext4_quota_off(sb, type);
859 static inline void ext4_quota_off_umount(struct super_block *sb)
864 static void ext4_put_super(struct super_block *sb)
866 struct ext4_sb_info *sbi = EXT4_SB(sb);
867 struct ext4_super_block *es = sbi->s_es;
871 ext4_unregister_li_request(sb);
872 ext4_quota_off_umount(sb);
874 flush_workqueue(sbi->rsv_conversion_wq);
875 destroy_workqueue(sbi->rsv_conversion_wq);
877 if (sbi->s_journal) {
878 aborted = is_journal_aborted(sbi->s_journal);
879 err = jbd2_journal_destroy(sbi->s_journal);
880 sbi->s_journal = NULL;
881 if ((err < 0) && !aborted)
882 ext4_abort(sb, "Couldn't clean up the journal");
885 ext4_unregister_sysfs(sb);
886 ext4_es_unregister_shrinker(sbi);
887 del_timer_sync(&sbi->s_err_report);
888 ext4_release_system_zone(sb);
890 ext4_ext_release(sb);
892 if (!(sb->s_flags & MS_RDONLY) && !aborted) {
893 ext4_clear_feature_journal_needs_recovery(sb);
894 es->s_state = cpu_to_le16(sbi->s_mount_state);
896 if (!(sb->s_flags & MS_RDONLY))
897 ext4_commit_super(sb, 1);
899 for (i = 0; i < sbi->s_gdb_count; i++)
900 brelse(sbi->s_group_desc[i]);
901 kvfree(sbi->s_group_desc);
902 kvfree(sbi->s_flex_groups);
903 percpu_counter_destroy(&sbi->s_freeclusters_counter);
904 percpu_counter_destroy(&sbi->s_freeinodes_counter);
905 percpu_counter_destroy(&sbi->s_dirs_counter);
906 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
907 percpu_free_rwsem(&sbi->s_journal_flag_rwsem);
909 for (i = 0; i < EXT4_MAXQUOTAS; i++)
910 kfree(sbi->s_qf_names[i]);
913 /* Debugging code just in case the in-memory inode orphan list
914 * isn't empty. The on-disk one can be non-empty if we've
915 * detected an error and taken the fs readonly, but the
916 * in-memory list had better be clean by this point. */
917 if (!list_empty(&sbi->s_orphan))
918 dump_orphan_list(sb, sbi);
919 J_ASSERT(list_empty(&sbi->s_orphan));
921 sync_blockdev(sb->s_bdev);
922 invalidate_bdev(sb->s_bdev);
923 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
925 * Invalidate the journal device's buffers. We don't want them
926 * floating about in memory - the physical journal device may
927 * hotswapped, and it breaks the `ro-after' testing code.
929 sync_blockdev(sbi->journal_bdev);
930 invalidate_bdev(sbi->journal_bdev);
931 ext4_blkdev_remove(sbi);
933 if (sbi->s_mb_cache) {
934 ext4_xattr_destroy_cache(sbi->s_mb_cache);
935 sbi->s_mb_cache = NULL;
938 kthread_stop(sbi->s_mmp_tsk);
940 sb->s_fs_info = NULL;
942 * Now that we are completely done shutting down the
943 * superblock, we need to actually destroy the kobject.
945 kobject_put(&sbi->s_kobj);
946 wait_for_completion(&sbi->s_kobj_unregister);
947 if (sbi->s_chksum_driver)
948 crypto_free_shash(sbi->s_chksum_driver);
949 kfree(sbi->s_blockgroup_lock);
953 static struct kmem_cache *ext4_inode_cachep;
956 * Called inside transaction, so use GFP_NOFS
958 static struct inode *ext4_alloc_inode(struct super_block *sb)
960 struct ext4_inode_info *ei;
962 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
966 ei->vfs_inode.i_version = 1;
967 spin_lock_init(&ei->i_raw_lock);
968 INIT_LIST_HEAD(&ei->i_prealloc_list);
969 spin_lock_init(&ei->i_prealloc_lock);
970 ext4_es_init_tree(&ei->i_es_tree);
971 rwlock_init(&ei->i_es_lock);
972 INIT_LIST_HEAD(&ei->i_es_list);
975 ei->i_es_shrink_lblk = 0;
976 ei->i_reserved_data_blocks = 0;
977 ei->i_reserved_meta_blocks = 0;
978 ei->i_allocated_meta_blocks = 0;
979 ei->i_da_metadata_calc_len = 0;
980 ei->i_da_metadata_calc_last_lblock = 0;
981 spin_lock_init(&(ei->i_block_reservation_lock));
983 ei->i_reserved_quota = 0;
984 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
987 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
988 spin_lock_init(&ei->i_completed_io_lock);
990 ei->i_datasync_tid = 0;
991 atomic_set(&ei->i_unwritten, 0);
992 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
993 return &ei->vfs_inode;
996 static int ext4_drop_inode(struct inode *inode)
998 int drop = generic_drop_inode(inode);
1000 trace_ext4_drop_inode(inode, drop);
1004 static void ext4_i_callback(struct rcu_head *head)
1006 struct inode *inode = container_of(head, struct inode, i_rcu);
1007 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
1010 static void ext4_destroy_inode(struct inode *inode)
1012 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
1013 ext4_msg(inode->i_sb, KERN_ERR,
1014 "Inode %lu (%p): orphan list check failed!",
1015 inode->i_ino, EXT4_I(inode));
1016 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1017 EXT4_I(inode), sizeof(struct ext4_inode_info),
1021 call_rcu(&inode->i_rcu, ext4_i_callback);
1024 static void init_once(void *foo)
1026 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
1028 INIT_LIST_HEAD(&ei->i_orphan);
1029 init_rwsem(&ei->xattr_sem);
1030 init_rwsem(&ei->i_data_sem);
1031 init_rwsem(&ei->i_mmap_sem);
1032 inode_init_once(&ei->vfs_inode);
1035 static int __init init_inodecache(void)
1037 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
1038 sizeof(struct ext4_inode_info),
1039 0, (SLAB_RECLAIM_ACCOUNT|
1040 SLAB_MEM_SPREAD|SLAB_ACCOUNT),
1042 if (ext4_inode_cachep == NULL)
1047 static void destroy_inodecache(void)
1050 * Make sure all delayed rcu free inodes are flushed before we
1054 kmem_cache_destroy(ext4_inode_cachep);
1057 void ext4_clear_inode(struct inode *inode)
1059 invalidate_inode_buffers(inode);
1062 ext4_discard_preallocations(inode);
1063 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
1064 if (EXT4_I(inode)->jinode) {
1065 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1066 EXT4_I(inode)->jinode);
1067 jbd2_free_inode(EXT4_I(inode)->jinode);
1068 EXT4_I(inode)->jinode = NULL;
1070 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1071 fscrypt_put_encryption_info(inode, NULL);
1075 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1076 u64 ino, u32 generation)
1078 struct inode *inode;
1080 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1081 return ERR_PTR(-ESTALE);
1082 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1083 return ERR_PTR(-ESTALE);
1085 /* iget isn't really right if the inode is currently unallocated!!
1087 * ext4_read_inode will return a bad_inode if the inode had been
1088 * deleted, so we should be safe.
1090 * Currently we don't know the generation for parent directory, so
1091 * a generation of 0 means "accept any"
1093 inode = ext4_iget_normal(sb, ino);
1095 return ERR_CAST(inode);
1096 if (generation && inode->i_generation != generation) {
1098 return ERR_PTR(-ESTALE);
1104 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1105 int fh_len, int fh_type)
1107 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1108 ext4_nfs_get_inode);
1111 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1112 int fh_len, int fh_type)
1114 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1115 ext4_nfs_get_inode);
1119 * Try to release metadata pages (indirect blocks, directories) which are
1120 * mapped via the block device. Since these pages could have journal heads
1121 * which would prevent try_to_free_buffers() from freeing them, we must use
1122 * jbd2 layer's try_to_free_buffers() function to release them.
1124 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1127 journal_t *journal = EXT4_SB(sb)->s_journal;
1129 WARN_ON(PageChecked(page));
1130 if (!page_has_buffers(page))
1133 return jbd2_journal_try_to_free_buffers(journal, page,
1134 wait & ~__GFP_DIRECT_RECLAIM);
1135 return try_to_free_buffers(page);
1138 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1139 static int ext4_get_context(struct inode *inode, void *ctx, size_t len)
1141 return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1142 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len);
1145 static int ext4_set_context(struct inode *inode, const void *ctx, size_t len,
1148 handle_t *handle = fs_data;
1149 int res, res2, retries = 0;
1151 res = ext4_convert_inline_data(inode);
1156 * If a journal handle was specified, then the encryption context is
1157 * being set on a new inode via inheritance and is part of a larger
1158 * transaction to create the inode. Otherwise the encryption context is
1159 * being set on an existing inode in its own transaction. Only in the
1160 * latter case should the "retry on ENOSPC" logic be used.
1164 res = ext4_xattr_set_handle(handle, inode,
1165 EXT4_XATTR_INDEX_ENCRYPTION,
1166 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1169 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1170 ext4_clear_inode_state(inode,
1171 EXT4_STATE_MAY_INLINE_DATA);
1173 * Update inode->i_flags - e.g. S_DAX may get disabled
1175 ext4_set_inode_flags(inode);
1181 handle = ext4_journal_start(inode, EXT4_HT_MISC,
1182 ext4_jbd2_credits_xattr(inode));
1184 return PTR_ERR(handle);
1186 res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION,
1187 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1190 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1191 /* Update inode->i_flags - e.g. S_DAX may get disabled */
1192 ext4_set_inode_flags(inode);
1193 res = ext4_mark_inode_dirty(handle, inode);
1195 EXT4_ERROR_INODE(inode, "Failed to mark inode dirty");
1197 res2 = ext4_journal_stop(handle);
1199 if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
1206 static int ext4_dummy_context(struct inode *inode)
1208 return DUMMY_ENCRYPTION_ENABLED(EXT4_SB(inode->i_sb));
1211 static unsigned ext4_max_namelen(struct inode *inode)
1213 return S_ISLNK(inode->i_mode) ? inode->i_sb->s_blocksize :
1217 static const struct fscrypt_operations ext4_cryptops = {
1218 .key_prefix = "ext4:",
1219 .get_context = ext4_get_context,
1220 .set_context = ext4_set_context,
1221 .dummy_context = ext4_dummy_context,
1222 .is_encrypted = ext4_encrypted_inode,
1223 .empty_dir = ext4_empty_dir,
1224 .max_namelen = ext4_max_namelen,
1227 static const struct fscrypt_operations ext4_cryptops = {
1228 .is_encrypted = ext4_encrypted_inode,
1233 static char *quotatypes[] = INITQFNAMES;
1234 #define QTYPE2NAME(t) (quotatypes[t])
1236 static int ext4_write_dquot(struct dquot *dquot);
1237 static int ext4_acquire_dquot(struct dquot *dquot);
1238 static int ext4_release_dquot(struct dquot *dquot);
1239 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1240 static int ext4_write_info(struct super_block *sb, int type);
1241 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1242 const struct path *path);
1243 static int ext4_quota_on_mount(struct super_block *sb, int type);
1244 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1245 size_t len, loff_t off);
1246 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1247 const char *data, size_t len, loff_t off);
1248 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1249 unsigned int flags);
1250 static int ext4_enable_quotas(struct super_block *sb);
1251 static int ext4_get_next_id(struct super_block *sb, struct kqid *qid);
1253 static struct dquot **ext4_get_dquots(struct inode *inode)
1255 return EXT4_I(inode)->i_dquot;
1258 static const struct dquot_operations ext4_quota_operations = {
1259 .get_reserved_space = ext4_get_reserved_space,
1260 .write_dquot = ext4_write_dquot,
1261 .acquire_dquot = ext4_acquire_dquot,
1262 .release_dquot = ext4_release_dquot,
1263 .mark_dirty = ext4_mark_dquot_dirty,
1264 .write_info = ext4_write_info,
1265 .alloc_dquot = dquot_alloc,
1266 .destroy_dquot = dquot_destroy,
1267 .get_projid = ext4_get_projid,
1268 .get_next_id = ext4_get_next_id,
1271 static const struct quotactl_ops ext4_qctl_operations = {
1272 .quota_on = ext4_quota_on,
1273 .quota_off = ext4_quota_off,
1274 .quota_sync = dquot_quota_sync,
1275 .get_state = dquot_get_state,
1276 .set_info = dquot_set_dqinfo,
1277 .get_dqblk = dquot_get_dqblk,
1278 .set_dqblk = dquot_set_dqblk,
1279 .get_nextdqblk = dquot_get_next_dqblk,
1283 static const struct super_operations ext4_sops = {
1284 .alloc_inode = ext4_alloc_inode,
1285 .destroy_inode = ext4_destroy_inode,
1286 .write_inode = ext4_write_inode,
1287 .dirty_inode = ext4_dirty_inode,
1288 .drop_inode = ext4_drop_inode,
1289 .evict_inode = ext4_evict_inode,
1290 .put_super = ext4_put_super,
1291 .sync_fs = ext4_sync_fs,
1292 .freeze_fs = ext4_freeze,
1293 .unfreeze_fs = ext4_unfreeze,
1294 .statfs = ext4_statfs,
1295 .remount_fs = ext4_remount,
1296 .show_options = ext4_show_options,
1298 .quota_read = ext4_quota_read,
1299 .quota_write = ext4_quota_write,
1300 .get_dquots = ext4_get_dquots,
1302 .bdev_try_to_free_page = bdev_try_to_free_page,
1305 static const struct export_operations ext4_export_ops = {
1306 .fh_to_dentry = ext4_fh_to_dentry,
1307 .fh_to_parent = ext4_fh_to_parent,
1308 .get_parent = ext4_get_parent,
1312 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1313 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1314 Opt_nouid32, Opt_debug, Opt_removed,
1315 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1316 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1317 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1318 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1319 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1320 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1321 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1322 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1323 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1324 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version, Opt_dax,
1325 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1326 Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize,
1327 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1328 Opt_inode_readahead_blks, Opt_journal_ioprio,
1329 Opt_dioread_nolock, Opt_dioread_lock,
1330 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1331 Opt_max_dir_size_kb, Opt_nojournal_checksum,
1334 static const match_table_t tokens = {
1335 {Opt_bsd_df, "bsddf"},
1336 {Opt_minix_df, "minixdf"},
1337 {Opt_grpid, "grpid"},
1338 {Opt_grpid, "bsdgroups"},
1339 {Opt_nogrpid, "nogrpid"},
1340 {Opt_nogrpid, "sysvgroups"},
1341 {Opt_resgid, "resgid=%u"},
1342 {Opt_resuid, "resuid=%u"},
1344 {Opt_err_cont, "errors=continue"},
1345 {Opt_err_panic, "errors=panic"},
1346 {Opt_err_ro, "errors=remount-ro"},
1347 {Opt_nouid32, "nouid32"},
1348 {Opt_debug, "debug"},
1349 {Opt_removed, "oldalloc"},
1350 {Opt_removed, "orlov"},
1351 {Opt_user_xattr, "user_xattr"},
1352 {Opt_nouser_xattr, "nouser_xattr"},
1354 {Opt_noacl, "noacl"},
1355 {Opt_noload, "norecovery"},
1356 {Opt_noload, "noload"},
1357 {Opt_removed, "nobh"},
1358 {Opt_removed, "bh"},
1359 {Opt_commit, "commit=%u"},
1360 {Opt_min_batch_time, "min_batch_time=%u"},
1361 {Opt_max_batch_time, "max_batch_time=%u"},
1362 {Opt_journal_dev, "journal_dev=%u"},
1363 {Opt_journal_path, "journal_path=%s"},
1364 {Opt_journal_checksum, "journal_checksum"},
1365 {Opt_nojournal_checksum, "nojournal_checksum"},
1366 {Opt_journal_async_commit, "journal_async_commit"},
1367 {Opt_abort, "abort"},
1368 {Opt_data_journal, "data=journal"},
1369 {Opt_data_ordered, "data=ordered"},
1370 {Opt_data_writeback, "data=writeback"},
1371 {Opt_data_err_abort, "data_err=abort"},
1372 {Opt_data_err_ignore, "data_err=ignore"},
1373 {Opt_offusrjquota, "usrjquota="},
1374 {Opt_usrjquota, "usrjquota=%s"},
1375 {Opt_offgrpjquota, "grpjquota="},
1376 {Opt_grpjquota, "grpjquota=%s"},
1377 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1378 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1379 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1380 {Opt_grpquota, "grpquota"},
1381 {Opt_noquota, "noquota"},
1382 {Opt_quota, "quota"},
1383 {Opt_usrquota, "usrquota"},
1384 {Opt_prjquota, "prjquota"},
1385 {Opt_barrier, "barrier=%u"},
1386 {Opt_barrier, "barrier"},
1387 {Opt_nobarrier, "nobarrier"},
1388 {Opt_i_version, "i_version"},
1390 {Opt_stripe, "stripe=%u"},
1391 {Opt_delalloc, "delalloc"},
1392 {Opt_lazytime, "lazytime"},
1393 {Opt_nolazytime, "nolazytime"},
1394 {Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"},
1395 {Opt_nodelalloc, "nodelalloc"},
1396 {Opt_removed, "mblk_io_submit"},
1397 {Opt_removed, "nomblk_io_submit"},
1398 {Opt_block_validity, "block_validity"},
1399 {Opt_noblock_validity, "noblock_validity"},
1400 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1401 {Opt_journal_ioprio, "journal_ioprio=%u"},
1402 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1403 {Opt_auto_da_alloc, "auto_da_alloc"},
1404 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1405 {Opt_dioread_nolock, "dioread_nolock"},
1406 {Opt_dioread_lock, "dioread_lock"},
1407 {Opt_discard, "discard"},
1408 {Opt_nodiscard, "nodiscard"},
1409 {Opt_init_itable, "init_itable=%u"},
1410 {Opt_init_itable, "init_itable"},
1411 {Opt_noinit_itable, "noinit_itable"},
1412 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1413 {Opt_test_dummy_encryption, "test_dummy_encryption"},
1414 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1415 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1416 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1417 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1418 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1422 static ext4_fsblk_t get_sb_block(void **data)
1424 ext4_fsblk_t sb_block;
1425 char *options = (char *) *data;
1427 if (!options || strncmp(options, "sb=", 3) != 0)
1428 return 1; /* Default location */
1431 /* TODO: use simple_strtoll with >32bit ext4 */
1432 sb_block = simple_strtoul(options, &options, 0);
1433 if (*options && *options != ',') {
1434 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1438 if (*options == ',')
1440 *data = (void *) options;
1445 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1446 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1447 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1450 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1452 struct ext4_sb_info *sbi = EXT4_SB(sb);
1456 if (sb_any_quota_loaded(sb) &&
1457 !sbi->s_qf_names[qtype]) {
1458 ext4_msg(sb, KERN_ERR,
1459 "Cannot change journaled "
1460 "quota options when quota turned on");
1463 if (ext4_has_feature_quota(sb)) {
1464 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1465 "ignored when QUOTA feature is enabled");
1468 qname = match_strdup(args);
1470 ext4_msg(sb, KERN_ERR,
1471 "Not enough memory for storing quotafile name");
1474 if (sbi->s_qf_names[qtype]) {
1475 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1478 ext4_msg(sb, KERN_ERR,
1479 "%s quota file already specified",
1483 if (strchr(qname, '/')) {
1484 ext4_msg(sb, KERN_ERR,
1485 "quotafile must be on filesystem root");
1488 sbi->s_qf_names[qtype] = qname;
1496 static int clear_qf_name(struct super_block *sb, int qtype)
1499 struct ext4_sb_info *sbi = EXT4_SB(sb);
1501 if (sb_any_quota_loaded(sb) &&
1502 sbi->s_qf_names[qtype]) {
1503 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1504 " when quota turned on");
1507 kfree(sbi->s_qf_names[qtype]);
1508 sbi->s_qf_names[qtype] = NULL;
1513 #define MOPT_SET 0x0001
1514 #define MOPT_CLEAR 0x0002
1515 #define MOPT_NOSUPPORT 0x0004
1516 #define MOPT_EXPLICIT 0x0008
1517 #define MOPT_CLEAR_ERR 0x0010
1518 #define MOPT_GTE0 0x0020
1521 #define MOPT_QFMT 0x0040
1523 #define MOPT_Q MOPT_NOSUPPORT
1524 #define MOPT_QFMT MOPT_NOSUPPORT
1526 #define MOPT_DATAJ 0x0080
1527 #define MOPT_NO_EXT2 0x0100
1528 #define MOPT_NO_EXT3 0x0200
1529 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1530 #define MOPT_STRING 0x0400
1532 static const struct mount_opts {
1536 } ext4_mount_opts[] = {
1537 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1538 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1539 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1540 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1541 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1542 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1543 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1544 MOPT_EXT4_ONLY | MOPT_SET},
1545 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1546 MOPT_EXT4_ONLY | MOPT_CLEAR},
1547 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1548 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1549 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1550 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1551 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1552 MOPT_EXT4_ONLY | MOPT_CLEAR},
1553 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1554 MOPT_EXT4_ONLY | MOPT_CLEAR},
1555 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1556 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1557 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1558 EXT4_MOUNT_JOURNAL_CHECKSUM),
1559 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1560 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1561 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1562 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1563 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1564 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1566 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1568 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1569 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1570 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1571 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1572 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1573 {Opt_commit, 0, MOPT_GTE0},
1574 {Opt_max_batch_time, 0, MOPT_GTE0},
1575 {Opt_min_batch_time, 0, MOPT_GTE0},
1576 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1577 {Opt_init_itable, 0, MOPT_GTE0},
1578 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
1579 {Opt_stripe, 0, MOPT_GTE0},
1580 {Opt_resuid, 0, MOPT_GTE0},
1581 {Opt_resgid, 0, MOPT_GTE0},
1582 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1583 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1584 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1585 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1586 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1587 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1588 MOPT_NO_EXT2 | MOPT_DATAJ},
1589 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1590 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1591 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1592 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1593 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1595 {Opt_acl, 0, MOPT_NOSUPPORT},
1596 {Opt_noacl, 0, MOPT_NOSUPPORT},
1598 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1599 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1600 {Opt_debug_want_extra_isize, 0, MOPT_GTE0},
1601 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1602 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1604 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1606 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1608 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1609 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1610 MOPT_CLEAR | MOPT_Q},
1611 {Opt_usrjquota, 0, MOPT_Q},
1612 {Opt_grpjquota, 0, MOPT_Q},
1613 {Opt_offusrjquota, 0, MOPT_Q},
1614 {Opt_offgrpjquota, 0, MOPT_Q},
1615 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1616 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1617 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1618 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1619 {Opt_test_dummy_encryption, 0, MOPT_GTE0},
1623 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1624 substring_t *args, unsigned long *journal_devnum,
1625 unsigned int *journal_ioprio, int is_remount)
1627 struct ext4_sb_info *sbi = EXT4_SB(sb);
1628 const struct mount_opts *m;
1634 if (token == Opt_usrjquota)
1635 return set_qf_name(sb, USRQUOTA, &args[0]);
1636 else if (token == Opt_grpjquota)
1637 return set_qf_name(sb, GRPQUOTA, &args[0]);
1638 else if (token == Opt_offusrjquota)
1639 return clear_qf_name(sb, USRQUOTA);
1640 else if (token == Opt_offgrpjquota)
1641 return clear_qf_name(sb, GRPQUOTA);
1645 case Opt_nouser_xattr:
1646 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1649 return 1; /* handled by get_sb_block() */
1651 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1654 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1657 sb->s_flags |= MS_I_VERSION;
1660 sb->s_flags |= MS_LAZYTIME;
1662 case Opt_nolazytime:
1663 sb->s_flags &= ~MS_LAZYTIME;
1667 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1668 if (token == m->token)
1671 if (m->token == Opt_err) {
1672 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1673 "or missing value", opt);
1677 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1678 ext4_msg(sb, KERN_ERR,
1679 "Mount option \"%s\" incompatible with ext2", opt);
1682 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1683 ext4_msg(sb, KERN_ERR,
1684 "Mount option \"%s\" incompatible with ext3", opt);
1688 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
1690 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1692 if (m->flags & MOPT_EXPLICIT) {
1693 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
1694 set_opt2(sb, EXPLICIT_DELALLOC);
1695 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
1696 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
1700 if (m->flags & MOPT_CLEAR_ERR)
1701 clear_opt(sb, ERRORS_MASK);
1702 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1703 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1704 "options when quota turned on");
1708 if (m->flags & MOPT_NOSUPPORT) {
1709 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1710 } else if (token == Opt_commit) {
1712 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1713 sbi->s_commit_interval = HZ * arg;
1714 } else if (token == Opt_debug_want_extra_isize) {
1715 sbi->s_want_extra_isize = arg;
1716 } else if (token == Opt_max_batch_time) {
1717 sbi->s_max_batch_time = arg;
1718 } else if (token == Opt_min_batch_time) {
1719 sbi->s_min_batch_time = arg;
1720 } else if (token == Opt_inode_readahead_blks) {
1721 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1722 ext4_msg(sb, KERN_ERR,
1723 "EXT4-fs: inode_readahead_blks must be "
1724 "0 or a power of 2 smaller than 2^31");
1727 sbi->s_inode_readahead_blks = arg;
1728 } else if (token == Opt_init_itable) {
1729 set_opt(sb, INIT_INODE_TABLE);
1731 arg = EXT4_DEF_LI_WAIT_MULT;
1732 sbi->s_li_wait_mult = arg;
1733 } else if (token == Opt_max_dir_size_kb) {
1734 sbi->s_max_dir_size_kb = arg;
1735 } else if (token == Opt_stripe) {
1736 sbi->s_stripe = arg;
1737 } else if (token == Opt_resuid) {
1738 uid = make_kuid(current_user_ns(), arg);
1739 if (!uid_valid(uid)) {
1740 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1743 sbi->s_resuid = uid;
1744 } else if (token == Opt_resgid) {
1745 gid = make_kgid(current_user_ns(), arg);
1746 if (!gid_valid(gid)) {
1747 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1750 sbi->s_resgid = gid;
1751 } else if (token == Opt_journal_dev) {
1753 ext4_msg(sb, KERN_ERR,
1754 "Cannot specify journal on remount");
1757 *journal_devnum = arg;
1758 } else if (token == Opt_journal_path) {
1760 struct inode *journal_inode;
1765 ext4_msg(sb, KERN_ERR,
1766 "Cannot specify journal on remount");
1769 journal_path = match_strdup(&args[0]);
1770 if (!journal_path) {
1771 ext4_msg(sb, KERN_ERR, "error: could not dup "
1772 "journal device string");
1776 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1778 ext4_msg(sb, KERN_ERR, "error: could not find "
1779 "journal device path: error %d", error);
1780 kfree(journal_path);
1784 journal_inode = d_inode(path.dentry);
1785 if (!S_ISBLK(journal_inode->i_mode)) {
1786 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1787 "is not a block device", journal_path);
1789 kfree(journal_path);
1793 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1795 kfree(journal_path);
1796 } else if (token == Opt_journal_ioprio) {
1798 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1803 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1804 } else if (token == Opt_test_dummy_encryption) {
1805 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1806 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1807 ext4_msg(sb, KERN_WARNING,
1808 "Test dummy encryption mode enabled");
1810 ext4_msg(sb, KERN_WARNING,
1811 "Test dummy encryption mount option ignored");
1813 } else if (m->flags & MOPT_DATAJ) {
1815 if (!sbi->s_journal)
1816 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1817 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1818 ext4_msg(sb, KERN_ERR,
1819 "Cannot change data mode on remount");
1823 clear_opt(sb, DATA_FLAGS);
1824 sbi->s_mount_opt |= m->mount_opt;
1827 } else if (m->flags & MOPT_QFMT) {
1828 if (sb_any_quota_loaded(sb) &&
1829 sbi->s_jquota_fmt != m->mount_opt) {
1830 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1831 "quota options when quota turned on");
1834 if (ext4_has_feature_quota(sb)) {
1835 ext4_msg(sb, KERN_INFO,
1836 "Quota format mount options ignored "
1837 "when QUOTA feature is enabled");
1840 sbi->s_jquota_fmt = m->mount_opt;
1842 } else if (token == Opt_dax) {
1843 #ifdef CONFIG_FS_DAX
1844 ext4_msg(sb, KERN_WARNING,
1845 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
1846 sbi->s_mount_opt |= m->mount_opt;
1848 ext4_msg(sb, KERN_INFO, "dax option not supported");
1851 } else if (token == Opt_data_err_abort) {
1852 sbi->s_mount_opt |= m->mount_opt;
1853 } else if (token == Opt_data_err_ignore) {
1854 sbi->s_mount_opt &= ~m->mount_opt;
1858 if (m->flags & MOPT_CLEAR)
1860 else if (unlikely(!(m->flags & MOPT_SET))) {
1861 ext4_msg(sb, KERN_WARNING,
1862 "buggy handling of option %s", opt);
1867 sbi->s_mount_opt |= m->mount_opt;
1869 sbi->s_mount_opt &= ~m->mount_opt;
1874 static int parse_options(char *options, struct super_block *sb,
1875 unsigned long *journal_devnum,
1876 unsigned int *journal_ioprio,
1879 struct ext4_sb_info *sbi = EXT4_SB(sb);
1881 substring_t args[MAX_OPT_ARGS];
1887 while ((p = strsep(&options, ",")) != NULL) {
1891 * Initialize args struct so we know whether arg was
1892 * found; some options take optional arguments.
1894 args[0].to = args[0].from = NULL;
1895 token = match_token(p, tokens, args);
1896 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1897 journal_ioprio, is_remount) < 0)
1902 * We do the test below only for project quotas. 'usrquota' and
1903 * 'grpquota' mount options are allowed even without quota feature
1904 * to support legacy quotas in quota files.
1906 if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) {
1907 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. "
1908 "Cannot enable project quota enforcement.");
1911 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1912 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1913 clear_opt(sb, USRQUOTA);
1915 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1916 clear_opt(sb, GRPQUOTA);
1918 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1919 ext4_msg(sb, KERN_ERR, "old and new quota "
1924 if (!sbi->s_jquota_fmt) {
1925 ext4_msg(sb, KERN_ERR, "journaled quota format "
1931 if (test_opt(sb, DIOREAD_NOLOCK)) {
1933 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1935 if (blocksize < PAGE_SIZE) {
1936 ext4_msg(sb, KERN_ERR, "can't mount with "
1937 "dioread_nolock if block size != PAGE_SIZE");
1944 static inline void ext4_show_quota_options(struct seq_file *seq,
1945 struct super_block *sb)
1947 #if defined(CONFIG_QUOTA)
1948 struct ext4_sb_info *sbi = EXT4_SB(sb);
1950 if (sbi->s_jquota_fmt) {
1953 switch (sbi->s_jquota_fmt) {
1964 seq_printf(seq, ",jqfmt=%s", fmtname);
1967 if (sbi->s_qf_names[USRQUOTA])
1968 seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]);
1970 if (sbi->s_qf_names[GRPQUOTA])
1971 seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]);
1975 static const char *token2str(int token)
1977 const struct match_token *t;
1979 for (t = tokens; t->token != Opt_err; t++)
1980 if (t->token == token && !strchr(t->pattern, '='))
1987 * - it's set to a non-default value OR
1988 * - if the per-sb default is different from the global default
1990 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1993 struct ext4_sb_info *sbi = EXT4_SB(sb);
1994 struct ext4_super_block *es = sbi->s_es;
1995 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1996 const struct mount_opts *m;
1997 char sep = nodefs ? '\n' : ',';
1999 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2000 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2002 if (sbi->s_sb_block != 1)
2003 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
2005 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
2006 int want_set = m->flags & MOPT_SET;
2007 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
2008 (m->flags & MOPT_CLEAR_ERR))
2010 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
2011 continue; /* skip if same as the default */
2013 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
2014 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
2015 continue; /* select Opt_noFoo vs Opt_Foo */
2016 SEQ_OPTS_PRINT("%s", token2str(m->token));
2019 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
2020 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
2021 SEQ_OPTS_PRINT("resuid=%u",
2022 from_kuid_munged(&init_user_ns, sbi->s_resuid));
2023 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
2024 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
2025 SEQ_OPTS_PRINT("resgid=%u",
2026 from_kgid_munged(&init_user_ns, sbi->s_resgid));
2027 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
2028 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
2029 SEQ_OPTS_PUTS("errors=remount-ro");
2030 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
2031 SEQ_OPTS_PUTS("errors=continue");
2032 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
2033 SEQ_OPTS_PUTS("errors=panic");
2034 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
2035 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
2036 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
2037 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
2038 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
2039 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2040 if (sb->s_flags & MS_I_VERSION)
2041 SEQ_OPTS_PUTS("i_version");
2042 if (nodefs || sbi->s_stripe)
2043 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
2044 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
2045 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2046 SEQ_OPTS_PUTS("data=journal");
2047 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2048 SEQ_OPTS_PUTS("data=ordered");
2049 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2050 SEQ_OPTS_PUTS("data=writeback");
2053 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
2054 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2055 sbi->s_inode_readahead_blks);
2057 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
2058 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
2059 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
2060 if (nodefs || sbi->s_max_dir_size_kb)
2061 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2062 if (test_opt(sb, DATA_ERR_ABORT))
2063 SEQ_OPTS_PUTS("data_err=abort");
2065 ext4_show_quota_options(seq, sb);
2069 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2071 return _ext4_show_options(seq, root->d_sb, 0);
2074 int ext4_seq_options_show(struct seq_file *seq, void *offset)
2076 struct super_block *sb = seq->private;
2079 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
2080 rc = _ext4_show_options(seq, sb, 1);
2081 seq_puts(seq, "\n");
2085 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
2088 struct ext4_sb_info *sbi = EXT4_SB(sb);
2091 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
2092 ext4_msg(sb, KERN_ERR, "revision level too high, "
2093 "forcing read-only mode");
2098 if (!(sbi->s_mount_state & EXT4_VALID_FS))
2099 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
2100 "running e2fsck is recommended");
2101 else if (sbi->s_mount_state & EXT4_ERROR_FS)
2102 ext4_msg(sb, KERN_WARNING,
2103 "warning: mounting fs with errors, "
2104 "running e2fsck is recommended");
2105 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
2106 le16_to_cpu(es->s_mnt_count) >=
2107 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
2108 ext4_msg(sb, KERN_WARNING,
2109 "warning: maximal mount count reached, "
2110 "running e2fsck is recommended");
2111 else if (le32_to_cpu(es->s_checkinterval) &&
2112 (le32_to_cpu(es->s_lastcheck) +
2113 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
2114 ext4_msg(sb, KERN_WARNING,
2115 "warning: checktime reached, "
2116 "running e2fsck is recommended");
2117 if (!sbi->s_journal)
2118 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
2119 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
2120 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
2121 le16_add_cpu(&es->s_mnt_count, 1);
2122 es->s_mtime = cpu_to_le32(get_seconds());
2123 ext4_update_dynamic_rev(sb);
2125 ext4_set_feature_journal_needs_recovery(sb);
2127 ext4_commit_super(sb, 1);
2129 if (test_opt(sb, DEBUG))
2130 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
2131 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
2133 sbi->s_groups_count,
2134 EXT4_BLOCKS_PER_GROUP(sb),
2135 EXT4_INODES_PER_GROUP(sb),
2136 sbi->s_mount_opt, sbi->s_mount_opt2);
2138 cleancache_init_fs(sb);
2142 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2144 struct ext4_sb_info *sbi = EXT4_SB(sb);
2145 struct flex_groups *new_groups;
2148 if (!sbi->s_log_groups_per_flex)
2151 size = ext4_flex_group(sbi, ngroup - 1) + 1;
2152 if (size <= sbi->s_flex_groups_allocated)
2155 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
2156 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
2158 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
2159 size / (int) sizeof(struct flex_groups));
2163 if (sbi->s_flex_groups) {
2164 memcpy(new_groups, sbi->s_flex_groups,
2165 (sbi->s_flex_groups_allocated *
2166 sizeof(struct flex_groups)));
2167 kvfree(sbi->s_flex_groups);
2169 sbi->s_flex_groups = new_groups;
2170 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
2174 static int ext4_fill_flex_info(struct super_block *sb)
2176 struct ext4_sb_info *sbi = EXT4_SB(sb);
2177 struct ext4_group_desc *gdp = NULL;
2178 ext4_group_t flex_group;
2181 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2182 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
2183 sbi->s_log_groups_per_flex = 0;
2187 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2191 for (i = 0; i < sbi->s_groups_count; i++) {
2192 gdp = ext4_get_group_desc(sb, i, NULL);
2194 flex_group = ext4_flex_group(sbi, i);
2195 atomic_add(ext4_free_inodes_count(sb, gdp),
2196 &sbi->s_flex_groups[flex_group].free_inodes);
2197 atomic64_add(ext4_free_group_clusters(sb, gdp),
2198 &sbi->s_flex_groups[flex_group].free_clusters);
2199 atomic_add(ext4_used_dirs_count(sb, gdp),
2200 &sbi->s_flex_groups[flex_group].used_dirs);
2208 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
2209 struct ext4_group_desc *gdp)
2211 int offset = offsetof(struct ext4_group_desc, bg_checksum);
2213 __le32 le_group = cpu_to_le32(block_group);
2214 struct ext4_sb_info *sbi = EXT4_SB(sb);
2216 if (ext4_has_metadata_csum(sbi->s_sb)) {
2217 /* Use new metadata_csum algorithm */
2219 __u16 dummy_csum = 0;
2221 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2223 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2224 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2225 sizeof(dummy_csum));
2226 offset += sizeof(dummy_csum);
2227 if (offset < sbi->s_desc_size)
2228 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2229 sbi->s_desc_size - offset);
2231 crc = csum32 & 0xFFFF;
2235 /* old crc16 code */
2236 if (!ext4_has_feature_gdt_csum(sb))
2239 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2240 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2241 crc = crc16(crc, (__u8 *)gdp, offset);
2242 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2243 /* for checksum of struct ext4_group_desc do the rest...*/
2244 if (ext4_has_feature_64bit(sb) &&
2245 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2246 crc = crc16(crc, (__u8 *)gdp + offset,
2247 le16_to_cpu(sbi->s_es->s_desc_size) -
2251 return cpu_to_le16(crc);
2254 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2255 struct ext4_group_desc *gdp)
2257 if (ext4_has_group_desc_csum(sb) &&
2258 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
2264 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2265 struct ext4_group_desc *gdp)
2267 if (!ext4_has_group_desc_csum(sb))
2269 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
2272 /* Called at mount-time, super-block is locked */
2273 static int ext4_check_descriptors(struct super_block *sb,
2274 ext4_fsblk_t sb_block,
2275 ext4_group_t *first_not_zeroed)
2277 struct ext4_sb_info *sbi = EXT4_SB(sb);
2278 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2279 ext4_fsblk_t last_block;
2280 ext4_fsblk_t block_bitmap;
2281 ext4_fsblk_t inode_bitmap;
2282 ext4_fsblk_t inode_table;
2283 int flexbg_flag = 0;
2284 ext4_group_t i, grp = sbi->s_groups_count;
2286 if (ext4_has_feature_flex_bg(sb))
2289 ext4_debug("Checking group descriptors");
2291 for (i = 0; i < sbi->s_groups_count; i++) {
2292 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2294 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2295 last_block = ext4_blocks_count(sbi->s_es) - 1;
2297 last_block = first_block +
2298 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2300 if ((grp == sbi->s_groups_count) &&
2301 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2304 block_bitmap = ext4_block_bitmap(sb, gdp);
2305 if (block_bitmap == sb_block) {
2306 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2307 "Block bitmap for group %u overlaps "
2310 if (block_bitmap < first_block || block_bitmap > last_block) {
2311 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2312 "Block bitmap for group %u not in group "
2313 "(block %llu)!", i, block_bitmap);
2316 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2317 if (inode_bitmap == sb_block) {
2318 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2319 "Inode bitmap for group %u overlaps "
2322 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2323 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2324 "Inode bitmap for group %u not in group "
2325 "(block %llu)!", i, inode_bitmap);
2328 inode_table = ext4_inode_table(sb, gdp);
2329 if (inode_table == sb_block) {
2330 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2331 "Inode table for group %u overlaps "
2334 if (inode_table < first_block ||
2335 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2336 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2337 "Inode table for group %u not in group "
2338 "(block %llu)!", i, inode_table);
2341 ext4_lock_group(sb, i);
2342 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2343 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2344 "Checksum for group %u failed (%u!=%u)",
2345 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
2346 gdp)), le16_to_cpu(gdp->bg_checksum));
2347 if (!(sb->s_flags & MS_RDONLY)) {
2348 ext4_unlock_group(sb, i);
2352 ext4_unlock_group(sb, i);
2354 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2356 if (NULL != first_not_zeroed)
2357 *first_not_zeroed = grp;
2361 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2362 * the superblock) which were deleted from all directories, but held open by
2363 * a process at the time of a crash. We walk the list and try to delete these
2364 * inodes at recovery time (only with a read-write filesystem).
2366 * In order to keep the orphan inode chain consistent during traversal (in
2367 * case of crash during recovery), we link each inode into the superblock
2368 * orphan list_head and handle it the same way as an inode deletion during
2369 * normal operation (which journals the operations for us).
2371 * We only do an iget() and an iput() on each inode, which is very safe if we
2372 * accidentally point at an in-use or already deleted inode. The worst that
2373 * can happen in this case is that we get a "bit already cleared" message from
2374 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2375 * e2fsck was run on this filesystem, and it must have already done the orphan
2376 * inode cleanup for us, so we can safely abort without any further action.
2378 static void ext4_orphan_cleanup(struct super_block *sb,
2379 struct ext4_super_block *es)
2381 unsigned int s_flags = sb->s_flags;
2382 int ret, nr_orphans = 0, nr_truncates = 0;
2386 if (!es->s_last_orphan) {
2387 jbd_debug(4, "no orphan inodes to clean up\n");
2391 if (bdev_read_only(sb->s_bdev)) {
2392 ext4_msg(sb, KERN_ERR, "write access "
2393 "unavailable, skipping orphan cleanup");
2397 /* Check if feature set would not allow a r/w mount */
2398 if (!ext4_feature_set_ok(sb, 0)) {
2399 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2400 "unknown ROCOMPAT features");
2404 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2405 /* don't clear list on RO mount w/ errors */
2406 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
2407 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
2408 "clearing orphan list.\n");
2409 es->s_last_orphan = 0;
2411 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2415 if (s_flags & MS_RDONLY) {
2416 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2417 sb->s_flags &= ~MS_RDONLY;
2420 /* Needed for iput() to work correctly and not trash data */
2421 sb->s_flags |= MS_ACTIVE;
2422 /* Turn on quotas so that they are updated correctly */
2423 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2424 if (EXT4_SB(sb)->s_qf_names[i]) {
2425 int ret = ext4_quota_on_mount(sb, i);
2427 ext4_msg(sb, KERN_ERR,
2428 "Cannot turn on journaled "
2429 "quota: error %d", ret);
2434 while (es->s_last_orphan) {
2435 struct inode *inode;
2438 * We may have encountered an error during cleanup; if
2439 * so, skip the rest.
2441 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2442 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2443 es->s_last_orphan = 0;
2447 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2448 if (IS_ERR(inode)) {
2449 es->s_last_orphan = 0;
2453 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2454 dquot_initialize(inode);
2455 if (inode->i_nlink) {
2456 if (test_opt(sb, DEBUG))
2457 ext4_msg(sb, KERN_DEBUG,
2458 "%s: truncating inode %lu to %lld bytes",
2459 __func__, inode->i_ino, inode->i_size);
2460 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2461 inode->i_ino, inode->i_size);
2463 truncate_inode_pages(inode->i_mapping, inode->i_size);
2464 ret = ext4_truncate(inode);
2466 ext4_std_error(inode->i_sb, ret);
2467 inode_unlock(inode);
2470 if (test_opt(sb, DEBUG))
2471 ext4_msg(sb, KERN_DEBUG,
2472 "%s: deleting unreferenced inode %lu",
2473 __func__, inode->i_ino);
2474 jbd_debug(2, "deleting unreferenced inode %lu\n",
2478 iput(inode); /* The delete magic happens here! */
2481 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2484 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2485 PLURAL(nr_orphans));
2487 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2488 PLURAL(nr_truncates));
2490 /* Turn quotas off */
2491 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2492 if (sb_dqopt(sb)->files[i])
2493 dquot_quota_off(sb, i);
2496 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2500 * Maximal extent format file size.
2501 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2502 * extent format containers, within a sector_t, and within i_blocks
2503 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2504 * so that won't be a limiting factor.
2506 * However there is other limiting factor. We do store extents in the form
2507 * of starting block and length, hence the resulting length of the extent
2508 * covering maximum file size must fit into on-disk format containers as
2509 * well. Given that length is always by 1 unit bigger than max unit (because
2510 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2512 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2514 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2517 loff_t upper_limit = MAX_LFS_FILESIZE;
2519 /* small i_blocks in vfs inode? */
2520 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2522 * CONFIG_LBDAF is not enabled implies the inode
2523 * i_block represent total blocks in 512 bytes
2524 * 32 == size of vfs inode i_blocks * 8
2526 upper_limit = (1LL << 32) - 1;
2528 /* total blocks in file system block size */
2529 upper_limit >>= (blkbits - 9);
2530 upper_limit <<= blkbits;
2534 * 32-bit extent-start container, ee_block. We lower the maxbytes
2535 * by one fs block, so ee_len can cover the extent of maximum file
2538 res = (1LL << 32) - 1;
2541 /* Sanity check against vm- & vfs- imposed limits */
2542 if (res > upper_limit)
2549 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2550 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2551 * We need to be 1 filesystem block less than the 2^48 sector limit.
2553 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2555 loff_t res = EXT4_NDIR_BLOCKS;
2558 /* This is calculated to be the largest file size for a dense, block
2559 * mapped file such that the file's total number of 512-byte sectors,
2560 * including data and all indirect blocks, does not exceed (2^48 - 1).
2562 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2563 * number of 512-byte sectors of the file.
2566 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2568 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2569 * the inode i_block field represents total file blocks in
2570 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2572 upper_limit = (1LL << 32) - 1;
2574 /* total blocks in file system block size */
2575 upper_limit >>= (bits - 9);
2579 * We use 48 bit ext4_inode i_blocks
2580 * With EXT4_HUGE_FILE_FL set the i_blocks
2581 * represent total number of blocks in
2582 * file system block size
2584 upper_limit = (1LL << 48) - 1;
2588 /* indirect blocks */
2590 /* double indirect blocks */
2591 meta_blocks += 1 + (1LL << (bits-2));
2592 /* tripple indirect blocks */
2593 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2595 upper_limit -= meta_blocks;
2596 upper_limit <<= bits;
2598 res += 1LL << (bits-2);
2599 res += 1LL << (2*(bits-2));
2600 res += 1LL << (3*(bits-2));
2602 if (res > upper_limit)
2605 if (res > MAX_LFS_FILESIZE)
2606 res = MAX_LFS_FILESIZE;
2611 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2612 ext4_fsblk_t logical_sb_block, int nr)
2614 struct ext4_sb_info *sbi = EXT4_SB(sb);
2615 ext4_group_t bg, first_meta_bg;
2618 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2620 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
2621 return logical_sb_block + nr + 1;
2622 bg = sbi->s_desc_per_block * nr;
2623 if (ext4_bg_has_super(sb, bg))
2627 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2628 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2629 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2632 if (sb->s_blocksize == 1024 && nr == 0 &&
2633 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2636 return (has_super + ext4_group_first_block_no(sb, bg));
2640 * ext4_get_stripe_size: Get the stripe size.
2641 * @sbi: In memory super block info
2643 * If we have specified it via mount option, then
2644 * use the mount option value. If the value specified at mount time is
2645 * greater than the blocks per group use the super block value.
2646 * If the super block value is greater than blocks per group return 0.
2647 * Allocator needs it be less than blocks per group.
2650 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2652 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2653 unsigned long stripe_width =
2654 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2657 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2658 ret = sbi->s_stripe;
2659 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
2661 else if (stride && stride <= sbi->s_blocks_per_group)
2667 * If the stripe width is 1, this makes no sense and
2668 * we set it to 0 to turn off stripe handling code.
2677 * Check whether this filesystem can be mounted based on
2678 * the features present and the RDONLY/RDWR mount requested.
2679 * Returns 1 if this filesystem can be mounted as requested,
2680 * 0 if it cannot be.
2682 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2684 if (ext4_has_unknown_ext4_incompat_features(sb)) {
2685 ext4_msg(sb, KERN_ERR,
2686 "Couldn't mount because of "
2687 "unsupported optional features (%x)",
2688 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2689 ~EXT4_FEATURE_INCOMPAT_SUPP));
2696 if (ext4_has_feature_readonly(sb)) {
2697 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2698 sb->s_flags |= MS_RDONLY;
2702 /* Check that feature set is OK for a read-write mount */
2703 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
2704 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2705 "unsupported optional features (%x)",
2706 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2707 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2711 * Large file size enabled file system can only be mounted
2712 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2714 if (ext4_has_feature_huge_file(sb)) {
2715 if (sizeof(blkcnt_t) < sizeof(u64)) {
2716 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2717 "cannot be mounted RDWR without "
2722 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
2723 ext4_msg(sb, KERN_ERR,
2724 "Can't support bigalloc feature without "
2725 "extents feature\n");
2729 #ifndef CONFIG_QUOTA
2730 if (ext4_has_feature_quota(sb) && !readonly) {
2731 ext4_msg(sb, KERN_ERR,
2732 "Filesystem with quota feature cannot be mounted RDWR "
2733 "without CONFIG_QUOTA");
2736 if (ext4_has_feature_project(sb) && !readonly) {
2737 ext4_msg(sb, KERN_ERR,
2738 "Filesystem with project quota feature cannot be mounted RDWR "
2739 "without CONFIG_QUOTA");
2742 #endif /* CONFIG_QUOTA */
2747 * This function is called once a day if we have errors logged
2748 * on the file system
2750 static void print_daily_error_info(unsigned long arg)
2752 struct super_block *sb = (struct super_block *) arg;
2753 struct ext4_sb_info *sbi;
2754 struct ext4_super_block *es;
2759 if (es->s_error_count)
2760 /* fsck newer than v1.41.13 is needed to clean this condition. */
2761 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
2762 le32_to_cpu(es->s_error_count));
2763 if (es->s_first_error_time) {
2764 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
2765 sb->s_id, le32_to_cpu(es->s_first_error_time),
2766 (int) sizeof(es->s_first_error_func),
2767 es->s_first_error_func,
2768 le32_to_cpu(es->s_first_error_line));
2769 if (es->s_first_error_ino)
2770 printk(KERN_CONT ": inode %u",
2771 le32_to_cpu(es->s_first_error_ino));
2772 if (es->s_first_error_block)
2773 printk(KERN_CONT ": block %llu", (unsigned long long)
2774 le64_to_cpu(es->s_first_error_block));
2775 printk(KERN_CONT "\n");
2777 if (es->s_last_error_time) {
2778 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
2779 sb->s_id, le32_to_cpu(es->s_last_error_time),
2780 (int) sizeof(es->s_last_error_func),
2781 es->s_last_error_func,
2782 le32_to_cpu(es->s_last_error_line));
2783 if (es->s_last_error_ino)
2784 printk(KERN_CONT ": inode %u",
2785 le32_to_cpu(es->s_last_error_ino));
2786 if (es->s_last_error_block)
2787 printk(KERN_CONT ": block %llu", (unsigned long long)
2788 le64_to_cpu(es->s_last_error_block));
2789 printk(KERN_CONT "\n");
2791 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2794 /* Find next suitable group and run ext4_init_inode_table */
2795 static int ext4_run_li_request(struct ext4_li_request *elr)
2797 struct ext4_group_desc *gdp = NULL;
2798 ext4_group_t group, ngroups;
2799 struct super_block *sb;
2800 unsigned long timeout = 0;
2804 ngroups = EXT4_SB(sb)->s_groups_count;
2806 for (group = elr->lr_next_group; group < ngroups; group++) {
2807 gdp = ext4_get_group_desc(sb, group, NULL);
2813 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2817 if (group >= ngroups)
2822 ret = ext4_init_inode_table(sb, group,
2823 elr->lr_timeout ? 0 : 1);
2824 if (elr->lr_timeout == 0) {
2825 timeout = (jiffies - timeout) *
2826 elr->lr_sbi->s_li_wait_mult;
2827 elr->lr_timeout = timeout;
2829 elr->lr_next_sched = jiffies + elr->lr_timeout;
2830 elr->lr_next_group = group + 1;
2836 * Remove lr_request from the list_request and free the
2837 * request structure. Should be called with li_list_mtx held
2839 static void ext4_remove_li_request(struct ext4_li_request *elr)
2841 struct ext4_sb_info *sbi;
2848 list_del(&elr->lr_request);
2849 sbi->s_li_request = NULL;
2853 static void ext4_unregister_li_request(struct super_block *sb)
2855 mutex_lock(&ext4_li_mtx);
2856 if (!ext4_li_info) {
2857 mutex_unlock(&ext4_li_mtx);
2861 mutex_lock(&ext4_li_info->li_list_mtx);
2862 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2863 mutex_unlock(&ext4_li_info->li_list_mtx);
2864 mutex_unlock(&ext4_li_mtx);
2867 static struct task_struct *ext4_lazyinit_task;
2870 * This is the function where ext4lazyinit thread lives. It walks
2871 * through the request list searching for next scheduled filesystem.
2872 * When such a fs is found, run the lazy initialization request
2873 * (ext4_rn_li_request) and keep track of the time spend in this
2874 * function. Based on that time we compute next schedule time of
2875 * the request. When walking through the list is complete, compute
2876 * next waking time and put itself into sleep.
2878 static int ext4_lazyinit_thread(void *arg)
2880 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2881 struct list_head *pos, *n;
2882 struct ext4_li_request *elr;
2883 unsigned long next_wakeup, cur;
2885 BUG_ON(NULL == eli);
2889 next_wakeup = MAX_JIFFY_OFFSET;
2891 mutex_lock(&eli->li_list_mtx);
2892 if (list_empty(&eli->li_request_list)) {
2893 mutex_unlock(&eli->li_list_mtx);
2896 list_for_each_safe(pos, n, &eli->li_request_list) {
2899 elr = list_entry(pos, struct ext4_li_request,
2902 if (time_before(jiffies, elr->lr_next_sched)) {
2903 if (time_before(elr->lr_next_sched, next_wakeup))
2904 next_wakeup = elr->lr_next_sched;
2907 if (down_read_trylock(&elr->lr_super->s_umount)) {
2908 if (sb_start_write_trylock(elr->lr_super)) {
2911 * We hold sb->s_umount, sb can not
2912 * be removed from the list, it is
2913 * now safe to drop li_list_mtx
2915 mutex_unlock(&eli->li_list_mtx);
2916 err = ext4_run_li_request(elr);
2917 sb_end_write(elr->lr_super);
2918 mutex_lock(&eli->li_list_mtx);
2921 up_read((&elr->lr_super->s_umount));
2923 /* error, remove the lazy_init job */
2925 ext4_remove_li_request(elr);
2929 elr->lr_next_sched = jiffies +
2931 % (EXT4_DEF_LI_MAX_START_DELAY * HZ));
2933 if (time_before(elr->lr_next_sched, next_wakeup))
2934 next_wakeup = elr->lr_next_sched;
2936 mutex_unlock(&eli->li_list_mtx);
2941 if ((time_after_eq(cur, next_wakeup)) ||
2942 (MAX_JIFFY_OFFSET == next_wakeup)) {
2947 schedule_timeout_interruptible(next_wakeup - cur);
2949 if (kthread_should_stop()) {
2950 ext4_clear_request_list();
2957 * It looks like the request list is empty, but we need
2958 * to check it under the li_list_mtx lock, to prevent any
2959 * additions into it, and of course we should lock ext4_li_mtx
2960 * to atomically free the list and ext4_li_info, because at
2961 * this point another ext4 filesystem could be registering
2964 mutex_lock(&ext4_li_mtx);
2965 mutex_lock(&eli->li_list_mtx);
2966 if (!list_empty(&eli->li_request_list)) {
2967 mutex_unlock(&eli->li_list_mtx);
2968 mutex_unlock(&ext4_li_mtx);
2971 mutex_unlock(&eli->li_list_mtx);
2972 kfree(ext4_li_info);
2973 ext4_li_info = NULL;
2974 mutex_unlock(&ext4_li_mtx);
2979 static void ext4_clear_request_list(void)
2981 struct list_head *pos, *n;
2982 struct ext4_li_request *elr;
2984 mutex_lock(&ext4_li_info->li_list_mtx);
2985 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2986 elr = list_entry(pos, struct ext4_li_request,
2988 ext4_remove_li_request(elr);
2990 mutex_unlock(&ext4_li_info->li_list_mtx);
2993 static int ext4_run_lazyinit_thread(void)
2995 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2996 ext4_li_info, "ext4lazyinit");
2997 if (IS_ERR(ext4_lazyinit_task)) {
2998 int err = PTR_ERR(ext4_lazyinit_task);
2999 ext4_clear_request_list();
3000 kfree(ext4_li_info);
3001 ext4_li_info = NULL;
3002 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3003 "initialization thread\n",
3007 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3012 * Check whether it make sense to run itable init. thread or not.
3013 * If there is at least one uninitialized inode table, return
3014 * corresponding group number, else the loop goes through all
3015 * groups and return total number of groups.
3017 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3019 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3020 struct ext4_group_desc *gdp = NULL;
3022 for (group = 0; group < ngroups; group++) {
3023 gdp = ext4_get_group_desc(sb, group, NULL);
3027 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3034 static int ext4_li_info_new(void)
3036 struct ext4_lazy_init *eli = NULL;
3038 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3042 INIT_LIST_HEAD(&eli->li_request_list);
3043 mutex_init(&eli->li_list_mtx);
3045 eli->li_state |= EXT4_LAZYINIT_QUIT;
3052 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3055 struct ext4_sb_info *sbi = EXT4_SB(sb);
3056 struct ext4_li_request *elr;
3058 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3064 elr->lr_next_group = start;
3067 * Randomize first schedule time of the request to
3068 * spread the inode table initialization requests
3071 elr->lr_next_sched = jiffies + (prandom_u32() %
3072 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3076 int ext4_register_li_request(struct super_block *sb,
3077 ext4_group_t first_not_zeroed)
3079 struct ext4_sb_info *sbi = EXT4_SB(sb);
3080 struct ext4_li_request *elr = NULL;
3081 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3084 mutex_lock(&ext4_li_mtx);
3085 if (sbi->s_li_request != NULL) {
3087 * Reset timeout so it can be computed again, because
3088 * s_li_wait_mult might have changed.
3090 sbi->s_li_request->lr_timeout = 0;
3094 if (first_not_zeroed == ngroups ||
3095 (sb->s_flags & MS_RDONLY) ||
3096 !test_opt(sb, INIT_INODE_TABLE))
3099 elr = ext4_li_request_new(sb, first_not_zeroed);
3105 if (NULL == ext4_li_info) {
3106 ret = ext4_li_info_new();
3111 mutex_lock(&ext4_li_info->li_list_mtx);
3112 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3113 mutex_unlock(&ext4_li_info->li_list_mtx);
3115 sbi->s_li_request = elr;
3117 * set elr to NULL here since it has been inserted to
3118 * the request_list and the removal and free of it is
3119 * handled by ext4_clear_request_list from now on.
3123 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3124 ret = ext4_run_lazyinit_thread();
3129 mutex_unlock(&ext4_li_mtx);
3136 * We do not need to lock anything since this is called on
3139 static void ext4_destroy_lazyinit_thread(void)
3142 * If thread exited earlier
3143 * there's nothing to be done.
3145 if (!ext4_li_info || !ext4_lazyinit_task)
3148 kthread_stop(ext4_lazyinit_task);
3151 static int set_journal_csum_feature_set(struct super_block *sb)
3154 int compat, incompat;
3155 struct ext4_sb_info *sbi = EXT4_SB(sb);
3157 if (ext4_has_metadata_csum(sb)) {
3158 /* journal checksum v3 */
3160 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
3162 /* journal checksum v1 */
3163 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3167 jbd2_journal_clear_features(sbi->s_journal,
3168 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3169 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3170 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3171 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3172 ret = jbd2_journal_set_features(sbi->s_journal,
3174 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3176 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3177 ret = jbd2_journal_set_features(sbi->s_journal,
3180 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3181 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3183 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3184 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3191 * Note: calculating the overhead so we can be compatible with
3192 * historical BSD practice is quite difficult in the face of
3193 * clusters/bigalloc. This is because multiple metadata blocks from
3194 * different block group can end up in the same allocation cluster.
3195 * Calculating the exact overhead in the face of clustered allocation
3196 * requires either O(all block bitmaps) in memory or O(number of block
3197 * groups**2) in time. We will still calculate the superblock for
3198 * older file systems --- and if we come across with a bigalloc file
3199 * system with zero in s_overhead_clusters the estimate will be close to
3200 * correct especially for very large cluster sizes --- but for newer
3201 * file systems, it's better to calculate this figure once at mkfs
3202 * time, and store it in the superblock. If the superblock value is
3203 * present (even for non-bigalloc file systems), we will use it.
3205 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3208 struct ext4_sb_info *sbi = EXT4_SB(sb);
3209 struct ext4_group_desc *gdp;
3210 ext4_fsblk_t first_block, last_block, b;
3211 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3212 int s, j, count = 0;
3214 if (!ext4_has_feature_bigalloc(sb))
3215 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3216 sbi->s_itb_per_group + 2);
3218 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3219 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3220 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3221 for (i = 0; i < ngroups; i++) {
3222 gdp = ext4_get_group_desc(sb, i, NULL);
3223 b = ext4_block_bitmap(sb, gdp);
3224 if (b >= first_block && b <= last_block) {
3225 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3228 b = ext4_inode_bitmap(sb, gdp);
3229 if (b >= first_block && b <= last_block) {
3230 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3233 b = ext4_inode_table(sb, gdp);
3234 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3235 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3236 int c = EXT4_B2C(sbi, b - first_block);
3237 ext4_set_bit(c, buf);
3243 if (ext4_bg_has_super(sb, grp)) {
3244 ext4_set_bit(s++, buf);
3247 j = ext4_bg_num_gdb(sb, grp);
3248 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
3249 ext4_error(sb, "Invalid number of block group "
3250 "descriptor blocks: %d", j);
3251 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
3255 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3259 return EXT4_CLUSTERS_PER_GROUP(sb) -
3260 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3264 * Compute the overhead and stash it in sbi->s_overhead
3266 int ext4_calculate_overhead(struct super_block *sb)
3268 struct ext4_sb_info *sbi = EXT4_SB(sb);
3269 struct ext4_super_block *es = sbi->s_es;
3270 struct inode *j_inode;
3271 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
3272 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3273 ext4_fsblk_t overhead = 0;
3274 char *buf = (char *) get_zeroed_page(GFP_NOFS);
3280 * Compute the overhead (FS structures). This is constant
3281 * for a given filesystem unless the number of block groups
3282 * changes so we cache the previous value until it does.
3286 * All of the blocks before first_data_block are overhead
3288 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3291 * Add the overhead found in each block group
3293 for (i = 0; i < ngroups; i++) {
3296 blks = count_overhead(sb, i, buf);
3299 memset(buf, 0, PAGE_SIZE);
3304 * Add the internal journal blocks whether the journal has been
3307 if (sbi->s_journal && !sbi->journal_bdev)
3308 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3309 else if (ext4_has_feature_journal(sb) && !sbi->s_journal) {
3310 j_inode = ext4_get_journal_inode(sb, j_inum);
3312 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
3313 overhead += EXT4_NUM_B2C(sbi, j_blocks);
3316 ext4_msg(sb, KERN_ERR, "can't get journal size");
3319 sbi->s_overhead = overhead;
3321 free_page((unsigned long) buf);
3325 static void ext4_set_resv_clusters(struct super_block *sb)
3327 ext4_fsblk_t resv_clusters;
3328 struct ext4_sb_info *sbi = EXT4_SB(sb);
3331 * There's no need to reserve anything when we aren't using extents.
3332 * The space estimates are exact, there are no unwritten extents,
3333 * hole punching doesn't need new metadata... This is needed especially
3334 * to keep ext2/3 backward compatibility.
3336 if (!ext4_has_feature_extents(sb))
3339 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3340 * This should cover the situations where we can not afford to run
3341 * out of space like for example punch hole, or converting
3342 * unwritten extents in delalloc path. In most cases such
3343 * allocation would require 1, or 2 blocks, higher numbers are
3346 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3347 sbi->s_cluster_bits);
3349 do_div(resv_clusters, 50);
3350 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3352 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
3355 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3357 char *orig_data = kstrdup(data, GFP_KERNEL);
3358 struct buffer_head *bh;
3359 struct ext4_super_block *es = NULL;
3360 struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3362 ext4_fsblk_t sb_block = get_sb_block(&data);
3363 ext4_fsblk_t logical_sb_block;
3364 unsigned long offset = 0;
3365 unsigned long journal_devnum = 0;
3366 unsigned long def_mount_opts;
3370 int blocksize, clustersize;
3371 unsigned int db_count;
3373 int needs_recovery, has_huge_files, has_bigalloc;
3376 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3377 ext4_group_t first_not_zeroed;
3379 if ((data && !orig_data) || !sbi)
3382 sbi->s_blockgroup_lock =
3383 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3384 if (!sbi->s_blockgroup_lock)
3387 sb->s_fs_info = sbi;
3389 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3390 sbi->s_sb_block = sb_block;
3391 if (sb->s_bdev->bd_part)
3392 sbi->s_sectors_written_start =
3393 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3395 /* Cleanup superblock name */
3396 strreplace(sb->s_id, '/', '!');
3398 /* -EINVAL is default */
3400 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3402 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3407 * The ext4 superblock will not be buffer aligned for other than 1kB
3408 * block sizes. We need to calculate the offset from buffer start.
3410 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3411 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3412 offset = do_div(logical_sb_block, blocksize);
3414 logical_sb_block = sb_block;
3417 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
3418 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3422 * Note: s_es must be initialized as soon as possible because
3423 * some ext4 macro-instructions depend on its value
3425 es = (struct ext4_super_block *) (bh->b_data + offset);
3427 sb->s_magic = le16_to_cpu(es->s_magic);
3428 if (sb->s_magic != EXT4_SUPER_MAGIC)
3430 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3432 /* Warn if metadata_csum and gdt_csum are both set. */
3433 if (ext4_has_feature_metadata_csum(sb) &&
3434 ext4_has_feature_gdt_csum(sb))
3435 ext4_warning(sb, "metadata_csum and uninit_bg are "
3436 "redundant flags; please run fsck.");
3438 /* Check for a known checksum algorithm */
3439 if (!ext4_verify_csum_type(sb, es)) {
3440 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3441 "unknown checksum algorithm.");
3446 /* Load the checksum driver */
3447 if (ext4_has_feature_metadata_csum(sb)) {
3448 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3449 if (IS_ERR(sbi->s_chksum_driver)) {
3450 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3451 ret = PTR_ERR(sbi->s_chksum_driver);
3452 sbi->s_chksum_driver = NULL;
3457 /* Check superblock checksum */
3458 if (!ext4_superblock_csum_verify(sb, es)) {
3459 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3460 "invalid superblock checksum. Run e2fsck?");
3466 /* Precompute checksum seed for all metadata */
3467 if (ext4_has_feature_csum_seed(sb))
3468 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
3469 else if (ext4_has_metadata_csum(sb))
3470 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3471 sizeof(es->s_uuid));
3473 /* Set defaults before we parse the mount options */
3474 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3475 set_opt(sb, INIT_INODE_TABLE);
3476 if (def_mount_opts & EXT4_DEFM_DEBUG)
3478 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3480 if (def_mount_opts & EXT4_DEFM_UID16)
3481 set_opt(sb, NO_UID32);
3482 /* xattr user namespace & acls are now defaulted on */
3483 set_opt(sb, XATTR_USER);
3484 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3485 set_opt(sb, POSIX_ACL);
3487 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3488 if (ext4_has_metadata_csum(sb))
3489 set_opt(sb, JOURNAL_CHECKSUM);
3491 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3492 set_opt(sb, JOURNAL_DATA);
3493 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3494 set_opt(sb, ORDERED_DATA);
3495 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3496 set_opt(sb, WRITEBACK_DATA);
3498 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3499 set_opt(sb, ERRORS_PANIC);
3500 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3501 set_opt(sb, ERRORS_CONT);
3503 set_opt(sb, ERRORS_RO);
3504 /* block_validity enabled by default; disable with noblock_validity */
3505 set_opt(sb, BLOCK_VALIDITY);
3506 if (def_mount_opts & EXT4_DEFM_DISCARD)
3507 set_opt(sb, DISCARD);
3509 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3510 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3511 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3512 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3513 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3515 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3516 set_opt(sb, BARRIER);
3519 * enable delayed allocation by default
3520 * Use -o nodelalloc to turn it off
3522 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3523 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3524 set_opt(sb, DELALLOC);
3527 * set default s_li_wait_mult for lazyinit, for the case there is
3528 * no mount option specified.
3530 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3532 if (sbi->s_es->s_mount_opts[0]) {
3533 char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
3534 sizeof(sbi->s_es->s_mount_opts),
3538 if (!parse_options(s_mount_opts, sb, &journal_devnum,
3539 &journal_ioprio, 0)) {
3540 ext4_msg(sb, KERN_WARNING,
3541 "failed to parse options in superblock: %s",
3544 kfree(s_mount_opts);
3546 sbi->s_def_mount_opt = sbi->s_mount_opt;
3547 if (!parse_options((char *) data, sb, &journal_devnum,
3548 &journal_ioprio, 0))
3551 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3552 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3553 "with data=journal disables delayed "
3554 "allocation and O_DIRECT support!\n");
3555 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3556 ext4_msg(sb, KERN_ERR, "can't mount with "
3557 "both data=journal and delalloc");
3560 if (test_opt(sb, DIOREAD_NOLOCK)) {
3561 ext4_msg(sb, KERN_ERR, "can't mount with "
3562 "both data=journal and dioread_nolock");
3565 if (test_opt(sb, DAX)) {
3566 ext4_msg(sb, KERN_ERR, "can't mount with "
3567 "both data=journal and dax");
3570 if (ext4_has_feature_encrypt(sb)) {
3571 ext4_msg(sb, KERN_WARNING,
3572 "encrypted files will use data=ordered "
3573 "instead of data journaling mode");
3575 if (test_opt(sb, DELALLOC))
3576 clear_opt(sb, DELALLOC);
3578 sb->s_iflags |= SB_I_CGROUPWB;
3581 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3582 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3584 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3585 (ext4_has_compat_features(sb) ||
3586 ext4_has_ro_compat_features(sb) ||
3587 ext4_has_incompat_features(sb)))
3588 ext4_msg(sb, KERN_WARNING,
3589 "feature flags set on rev 0 fs, "
3590 "running e2fsck is recommended");
3592 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3593 set_opt2(sb, HURD_COMPAT);
3594 if (ext4_has_feature_64bit(sb)) {
3595 ext4_msg(sb, KERN_ERR,
3596 "The Hurd can't support 64-bit file systems");
3601 if (IS_EXT2_SB(sb)) {
3602 if (ext2_feature_set_ok(sb))
3603 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3604 "using the ext4 subsystem");
3606 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3607 "to feature incompatibilities");
3612 if (IS_EXT3_SB(sb)) {
3613 if (ext3_feature_set_ok(sb))
3614 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3615 "using the ext4 subsystem");
3617 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3618 "to feature incompatibilities");
3624 * Check feature flags regardless of the revision level, since we
3625 * previously didn't change the revision level when setting the flags,
3626 * so there is a chance incompat flags are set on a rev 0 filesystem.
3628 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3631 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3632 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3633 blocksize > EXT4_MAX_BLOCK_SIZE) {
3634 ext4_msg(sb, KERN_ERR,
3635 "Unsupported filesystem blocksize %d (%d log_block_size)",
3636 blocksize, le32_to_cpu(es->s_log_block_size));
3639 if (le32_to_cpu(es->s_log_block_size) >
3640 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3641 ext4_msg(sb, KERN_ERR,
3642 "Invalid log block size: %u",
3643 le32_to_cpu(es->s_log_block_size));
3647 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
3648 ext4_msg(sb, KERN_ERR,
3649 "Number of reserved GDT blocks insanely large: %d",
3650 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
3654 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
3655 err = bdev_dax_supported(sb, blocksize);
3660 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
3661 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
3662 es->s_encryption_level);
3666 if (sb->s_blocksize != blocksize) {
3667 /* Validate the filesystem blocksize */
3668 if (!sb_set_blocksize(sb, blocksize)) {
3669 ext4_msg(sb, KERN_ERR, "bad block size %d",
3675 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3676 offset = do_div(logical_sb_block, blocksize);
3677 bh = sb_bread_unmovable(sb, logical_sb_block);
3679 ext4_msg(sb, KERN_ERR,
3680 "Can't read superblock on 2nd try");
3683 es = (struct ext4_super_block *)(bh->b_data + offset);
3685 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3686 ext4_msg(sb, KERN_ERR,
3687 "Magic mismatch, very weird!");
3692 has_huge_files = ext4_has_feature_huge_file(sb);
3693 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3695 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3697 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3698 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3699 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3701 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3702 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3703 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3704 (!is_power_of_2(sbi->s_inode_size)) ||
3705 (sbi->s_inode_size > blocksize)) {
3706 ext4_msg(sb, KERN_ERR,
3707 "unsupported inode size: %d",
3711 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3712 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3715 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3716 if (ext4_has_feature_64bit(sb)) {
3717 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3718 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3719 !is_power_of_2(sbi->s_desc_size)) {
3720 ext4_msg(sb, KERN_ERR,
3721 "unsupported descriptor size %lu",
3726 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3728 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3729 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3731 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3732 if (sbi->s_inodes_per_block == 0)
3734 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
3735 sbi->s_inodes_per_group > blocksize * 8) {
3736 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
3737 sbi->s_blocks_per_group);
3740 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3741 sbi->s_inodes_per_block;
3742 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3744 sbi->s_mount_state = le16_to_cpu(es->s_state);
3745 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3746 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3748 for (i = 0; i < 4; i++)
3749 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3750 sbi->s_def_hash_version = es->s_def_hash_version;
3751 if (ext4_has_feature_dir_index(sb)) {
3752 i = le32_to_cpu(es->s_flags);
3753 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3754 sbi->s_hash_unsigned = 3;
3755 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3756 #ifdef __CHAR_UNSIGNED__
3757 if (!(sb->s_flags & MS_RDONLY))
3759 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3760 sbi->s_hash_unsigned = 3;
3762 if (!(sb->s_flags & MS_RDONLY))
3764 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3769 /* Handle clustersize */
3770 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3771 has_bigalloc = ext4_has_feature_bigalloc(sb);
3773 if (clustersize < blocksize) {
3774 ext4_msg(sb, KERN_ERR,
3775 "cluster size (%d) smaller than "
3776 "block size (%d)", clustersize, blocksize);
3779 if (le32_to_cpu(es->s_log_cluster_size) >
3780 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3781 ext4_msg(sb, KERN_ERR,
3782 "Invalid log cluster size: %u",
3783 le32_to_cpu(es->s_log_cluster_size));
3786 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3787 le32_to_cpu(es->s_log_block_size);
3788 sbi->s_clusters_per_group =
3789 le32_to_cpu(es->s_clusters_per_group);
3790 if (sbi->s_clusters_per_group > blocksize * 8) {
3791 ext4_msg(sb, KERN_ERR,
3792 "#clusters per group too big: %lu",
3793 sbi->s_clusters_per_group);
3796 if (sbi->s_blocks_per_group !=
3797 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3798 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3799 "clusters per group (%lu) inconsistent",
3800 sbi->s_blocks_per_group,
3801 sbi->s_clusters_per_group);
3805 if (clustersize != blocksize) {
3806 ext4_warning(sb, "fragment/cluster size (%d) != "
3807 "block size (%d)", clustersize,
3809 clustersize = blocksize;
3811 if (sbi->s_blocks_per_group > blocksize * 8) {
3812 ext4_msg(sb, KERN_ERR,
3813 "#blocks per group too big: %lu",
3814 sbi->s_blocks_per_group);
3817 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3818 sbi->s_cluster_bits = 0;
3820 sbi->s_cluster_ratio = clustersize / blocksize;
3822 /* Do we have standard group size of clustersize * 8 blocks ? */
3823 if (sbi->s_blocks_per_group == clustersize << 3)
3824 set_opt2(sb, STD_GROUP_SIZE);
3827 * Test whether we have more sectors than will fit in sector_t,
3828 * and whether the max offset is addressable by the page cache.
3830 err = generic_check_addressable(sb->s_blocksize_bits,
3831 ext4_blocks_count(es));
3833 ext4_msg(sb, KERN_ERR, "filesystem"
3834 " too large to mount safely on this system");
3835 if (sizeof(sector_t) < 8)
3836 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3840 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3843 /* check blocks count against device size */
3844 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3845 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3846 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3847 "exceeds size of device (%llu blocks)",
3848 ext4_blocks_count(es), blocks_count);
3853 * It makes no sense for the first data block to be beyond the end
3854 * of the filesystem.
3856 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3857 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3858 "block %u is beyond end of filesystem (%llu)",
3859 le32_to_cpu(es->s_first_data_block),
3860 ext4_blocks_count(es));
3863 blocks_count = (ext4_blocks_count(es) -
3864 le32_to_cpu(es->s_first_data_block) +
3865 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3866 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3867 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3868 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3869 "(block count %llu, first data block %u, "
3870 "blocks per group %lu)", sbi->s_groups_count,
3871 ext4_blocks_count(es),
3872 le32_to_cpu(es->s_first_data_block),
3873 EXT4_BLOCKS_PER_GROUP(sb));
3876 sbi->s_groups_count = blocks_count;
3877 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3878 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3879 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3880 EXT4_DESC_PER_BLOCK(sb);
3881 if (ext4_has_feature_meta_bg(sb)) {
3882 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
3883 ext4_msg(sb, KERN_WARNING,
3884 "first meta block group too large: %u "
3885 "(group descriptor block count %u)",
3886 le32_to_cpu(es->s_first_meta_bg), db_count);
3890 sbi->s_group_desc = ext4_kvmalloc(db_count *
3891 sizeof(struct buffer_head *),
3893 if (sbi->s_group_desc == NULL) {
3894 ext4_msg(sb, KERN_ERR, "not enough memory");
3899 bgl_lock_init(sbi->s_blockgroup_lock);
3901 for (i = 0; i < db_count; i++) {
3902 block = descriptor_loc(sb, logical_sb_block, i);
3903 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
3904 if (!sbi->s_group_desc[i]) {
3905 ext4_msg(sb, KERN_ERR,
3906 "can't read group descriptor %d", i);
3911 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
3912 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3913 ret = -EFSCORRUPTED;
3917 sbi->s_gdb_count = db_count;
3918 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3919 spin_lock_init(&sbi->s_next_gen_lock);
3921 setup_timer(&sbi->s_err_report, print_daily_error_info,
3922 (unsigned long) sb);
3924 /* Register extent status tree shrinker */
3925 if (ext4_es_register_shrinker(sbi))
3928 sbi->s_stripe = ext4_get_stripe_size(sbi);
3929 sbi->s_extent_max_zeroout_kb = 32;
3932 * set up enough so that it can read an inode
3934 sb->s_op = &ext4_sops;
3935 sb->s_export_op = &ext4_export_ops;
3936 sb->s_xattr = ext4_xattr_handlers;
3937 sb->s_cop = &ext4_cryptops;
3939 sb->dq_op = &ext4_quota_operations;
3940 if (ext4_has_feature_quota(sb))
3941 sb->s_qcop = &dquot_quotactl_sysfile_ops;
3943 sb->s_qcop = &ext4_qctl_operations;
3944 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
3946 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3948 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3949 mutex_init(&sbi->s_orphan_lock);
3953 needs_recovery = (es->s_last_orphan != 0 ||
3954 ext4_has_feature_journal_needs_recovery(sb));
3956 if (ext4_has_feature_mmp(sb) && !(sb->s_flags & MS_RDONLY))
3957 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3958 goto failed_mount3a;
3961 * The first inode we look at is the journal inode. Don't try
3962 * root first: it may be modified in the journal!
3964 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
3965 err = ext4_load_journal(sb, es, journal_devnum);
3967 goto failed_mount3a;
3968 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3969 ext4_has_feature_journal_needs_recovery(sb)) {
3970 ext4_msg(sb, KERN_ERR, "required journal recovery "
3971 "suppressed and not mounted read-only");
3972 goto failed_mount_wq;
3974 /* Nojournal mode, all journal mount options are illegal */
3975 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
3976 ext4_msg(sb, KERN_ERR, "can't mount with "
3977 "journal_checksum, fs mounted w/o journal");
3978 goto failed_mount_wq;
3980 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3981 ext4_msg(sb, KERN_ERR, "can't mount with "
3982 "journal_async_commit, fs mounted w/o journal");
3983 goto failed_mount_wq;
3985 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
3986 ext4_msg(sb, KERN_ERR, "can't mount with "
3987 "commit=%lu, fs mounted w/o journal",
3988 sbi->s_commit_interval / HZ);
3989 goto failed_mount_wq;
3991 if (EXT4_MOUNT_DATA_FLAGS &
3992 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
3993 ext4_msg(sb, KERN_ERR, "can't mount with "
3994 "data=, fs mounted w/o journal");
3995 goto failed_mount_wq;
3997 sbi->s_def_mount_opt &= EXT4_MOUNT_JOURNAL_CHECKSUM;
3998 clear_opt(sb, JOURNAL_CHECKSUM);
3999 clear_opt(sb, DATA_FLAGS);
4000 sbi->s_journal = NULL;
4005 if (ext4_has_feature_64bit(sb) &&
4006 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4007 JBD2_FEATURE_INCOMPAT_64BIT)) {
4008 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
4009 goto failed_mount_wq;
4012 if (!set_journal_csum_feature_set(sb)) {
4013 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4015 goto failed_mount_wq;
4018 /* We have now updated the journal if required, so we can
4019 * validate the data journaling mode. */
4020 switch (test_opt(sb, DATA_FLAGS)) {
4022 /* No mode set, assume a default based on the journal
4023 * capabilities: ORDERED_DATA if the journal can
4024 * cope, else JOURNAL_DATA
4026 if (jbd2_journal_check_available_features
4027 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
4028 set_opt(sb, ORDERED_DATA);
4030 set_opt(sb, JOURNAL_DATA);
4033 case EXT4_MOUNT_ORDERED_DATA:
4034 case EXT4_MOUNT_WRITEBACK_DATA:
4035 if (!jbd2_journal_check_available_features
4036 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4037 ext4_msg(sb, KERN_ERR, "Journal does not support "
4038 "requested data journaling mode");
4039 goto failed_mount_wq;
4045 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
4046 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4047 ext4_msg(sb, KERN_ERR, "can't mount with "
4048 "journal_async_commit in data=ordered mode");
4049 goto failed_mount_wq;
4052 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4054 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4057 sbi->s_mb_cache = ext4_xattr_create_cache();
4058 if (!sbi->s_mb_cache) {
4059 ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
4060 goto failed_mount_wq;
4063 if ((DUMMY_ENCRYPTION_ENABLED(sbi) || ext4_has_feature_encrypt(sb)) &&
4064 (blocksize != PAGE_SIZE)) {
4065 ext4_msg(sb, KERN_ERR,
4066 "Unsupported blocksize for fs encryption");
4067 goto failed_mount_wq;
4070 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !(sb->s_flags & MS_RDONLY) &&
4071 !ext4_has_feature_encrypt(sb)) {
4072 ext4_set_feature_encrypt(sb);
4073 ext4_commit_super(sb, 1);
4077 * Get the # of file system overhead blocks from the
4078 * superblock if present.
4080 if (es->s_overhead_clusters)
4081 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4083 err = ext4_calculate_overhead(sb);
4085 goto failed_mount_wq;
4089 * The maximum number of concurrent works can be high and
4090 * concurrency isn't really necessary. Limit it to 1.
4092 EXT4_SB(sb)->rsv_conversion_wq =
4093 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4094 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4095 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
4101 * The jbd2_journal_load will have done any necessary log recovery,
4102 * so we can safely mount the rest of the filesystem now.
4105 root = ext4_iget(sb, EXT4_ROOT_INO);
4107 ext4_msg(sb, KERN_ERR, "get root inode failed");
4108 ret = PTR_ERR(root);
4112 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4113 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4117 sb->s_root = d_make_root(root);
4119 ext4_msg(sb, KERN_ERR, "get root dentry failed");
4124 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
4125 sb->s_flags |= MS_RDONLY;
4127 /* determine the minimum size of new large inodes, if present */
4128 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE &&
4129 sbi->s_want_extra_isize == 0) {
4130 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4131 EXT4_GOOD_OLD_INODE_SIZE;
4132 if (ext4_has_feature_extra_isize(sb)) {
4133 if (sbi->s_want_extra_isize <
4134 le16_to_cpu(es->s_want_extra_isize))
4135 sbi->s_want_extra_isize =
4136 le16_to_cpu(es->s_want_extra_isize);
4137 if (sbi->s_want_extra_isize <
4138 le16_to_cpu(es->s_min_extra_isize))
4139 sbi->s_want_extra_isize =
4140 le16_to_cpu(es->s_min_extra_isize);
4143 /* Check if enough inode space is available */
4144 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4145 sbi->s_inode_size) {
4146 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4147 EXT4_GOOD_OLD_INODE_SIZE;
4148 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4152 ext4_set_resv_clusters(sb);
4154 err = ext4_setup_system_zone(sb);
4156 ext4_msg(sb, KERN_ERR, "failed to initialize system "
4158 goto failed_mount4a;
4162 err = ext4_mb_init(sb);
4164 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4169 block = ext4_count_free_clusters(sb);
4170 ext4_free_blocks_count_set(sbi->s_es,
4171 EXT4_C2B(sbi, block));
4172 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4175 unsigned long freei = ext4_count_free_inodes(sb);
4176 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
4177 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4181 err = percpu_counter_init(&sbi->s_dirs_counter,
4182 ext4_count_dirs(sb), GFP_KERNEL);
4184 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4187 err = percpu_init_rwsem(&sbi->s_journal_flag_rwsem);
4190 ext4_msg(sb, KERN_ERR, "insufficient memory");
4194 if (ext4_has_feature_flex_bg(sb))
4195 if (!ext4_fill_flex_info(sb)) {
4196 ext4_msg(sb, KERN_ERR,
4197 "unable to initialize "
4198 "flex_bg meta info!");
4202 err = ext4_register_li_request(sb, first_not_zeroed);
4206 err = ext4_register_sysfs(sb);
4211 /* Enable quota usage during mount. */
4212 if (ext4_has_feature_quota(sb) && !(sb->s_flags & MS_RDONLY)) {
4213 err = ext4_enable_quotas(sb);
4217 #endif /* CONFIG_QUOTA */
4219 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4220 ext4_orphan_cleanup(sb, es);
4221 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4222 if (needs_recovery) {
4223 ext4_msg(sb, KERN_INFO, "recovery complete");
4224 ext4_mark_recovery_complete(sb, es);
4226 if (EXT4_SB(sb)->s_journal) {
4227 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4228 descr = " journalled data mode";
4229 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4230 descr = " ordered data mode";
4232 descr = " writeback data mode";
4234 descr = "out journal";
4236 if (test_opt(sb, DISCARD)) {
4237 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4238 if (!blk_queue_discard(q))
4239 ext4_msg(sb, KERN_WARNING,
4240 "mounting with \"discard\" option, but "
4241 "the device does not support discard");
4244 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
4245 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4246 "Opts: %.*s%s%s", descr,
4247 (int) sizeof(sbi->s_es->s_mount_opts),
4248 sbi->s_es->s_mount_opts,
4249 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4251 if (es->s_error_count)
4252 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4254 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4255 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4256 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4257 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4264 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4269 ext4_unregister_sysfs(sb);
4272 ext4_unregister_li_request(sb);
4274 ext4_mb_release(sb);
4275 if (sbi->s_flex_groups)
4276 kvfree(sbi->s_flex_groups);
4277 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4278 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4279 percpu_counter_destroy(&sbi->s_dirs_counter);
4280 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4282 ext4_ext_release(sb);
4283 ext4_release_system_zone(sb);
4288 ext4_msg(sb, KERN_ERR, "mount failed");
4289 if (EXT4_SB(sb)->rsv_conversion_wq)
4290 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4292 if (sbi->s_mb_cache) {
4293 ext4_xattr_destroy_cache(sbi->s_mb_cache);
4294 sbi->s_mb_cache = NULL;
4296 if (sbi->s_journal) {
4297 jbd2_journal_destroy(sbi->s_journal);
4298 sbi->s_journal = NULL;
4301 ext4_es_unregister_shrinker(sbi);
4303 del_timer_sync(&sbi->s_err_report);
4305 kthread_stop(sbi->s_mmp_tsk);
4307 for (i = 0; i < db_count; i++)
4308 brelse(sbi->s_group_desc[i]);
4309 kvfree(sbi->s_group_desc);
4311 if (sbi->s_chksum_driver)
4312 crypto_free_shash(sbi->s_chksum_driver);
4314 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4315 kfree(sbi->s_qf_names[i]);
4317 ext4_blkdev_remove(sbi);
4320 sb->s_fs_info = NULL;
4321 kfree(sbi->s_blockgroup_lock);
4325 return err ? err : ret;
4329 * Setup any per-fs journal parameters now. We'll do this both on
4330 * initial mount, once the journal has been initialised but before we've
4331 * done any recovery; and again on any subsequent remount.
4333 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4335 struct ext4_sb_info *sbi = EXT4_SB(sb);
4337 journal->j_commit_interval = sbi->s_commit_interval;
4338 journal->j_min_batch_time = sbi->s_min_batch_time;
4339 journal->j_max_batch_time = sbi->s_max_batch_time;
4341 write_lock(&journal->j_state_lock);
4342 if (test_opt(sb, BARRIER))
4343 journal->j_flags |= JBD2_BARRIER;
4345 journal->j_flags &= ~JBD2_BARRIER;
4346 if (test_opt(sb, DATA_ERR_ABORT))
4347 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4349 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4350 write_unlock(&journal->j_state_lock);
4353 static struct inode *ext4_get_journal_inode(struct super_block *sb,
4354 unsigned int journal_inum)
4356 struct inode *journal_inode;
4359 * Test for the existence of a valid inode on disk. Bad things
4360 * happen if we iget() an unused inode, as the subsequent iput()
4361 * will try to delete it.
4363 journal_inode = ext4_iget(sb, journal_inum);
4364 if (IS_ERR(journal_inode)) {
4365 ext4_msg(sb, KERN_ERR, "no journal found");
4368 if (!journal_inode->i_nlink) {
4369 make_bad_inode(journal_inode);
4370 iput(journal_inode);
4371 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4375 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4376 journal_inode, journal_inode->i_size);
4377 if (!S_ISREG(journal_inode->i_mode)) {
4378 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4379 iput(journal_inode);
4382 return journal_inode;
4385 static journal_t *ext4_get_journal(struct super_block *sb,
4386 unsigned int journal_inum)
4388 struct inode *journal_inode;
4391 BUG_ON(!ext4_has_feature_journal(sb));
4393 journal_inode = ext4_get_journal_inode(sb, journal_inum);
4397 journal = jbd2_journal_init_inode(journal_inode);
4399 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4400 iput(journal_inode);
4403 journal->j_private = sb;
4404 ext4_init_journal_params(sb, journal);
4408 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4411 struct buffer_head *bh;
4415 int hblock, blocksize;
4416 ext4_fsblk_t sb_block;
4417 unsigned long offset;
4418 struct ext4_super_block *es;
4419 struct block_device *bdev;
4421 BUG_ON(!ext4_has_feature_journal(sb));
4423 bdev = ext4_blkdev_get(j_dev, sb);
4427 blocksize = sb->s_blocksize;
4428 hblock = bdev_logical_block_size(bdev);
4429 if (blocksize < hblock) {
4430 ext4_msg(sb, KERN_ERR,
4431 "blocksize too small for journal device");
4435 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4436 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4437 set_blocksize(bdev, blocksize);
4438 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4439 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4440 "external journal");
4444 es = (struct ext4_super_block *) (bh->b_data + offset);
4445 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4446 !(le32_to_cpu(es->s_feature_incompat) &
4447 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4448 ext4_msg(sb, KERN_ERR, "external journal has "
4454 if ((le32_to_cpu(es->s_feature_ro_compat) &
4455 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4456 es->s_checksum != ext4_superblock_csum(sb, es)) {
4457 ext4_msg(sb, KERN_ERR, "external journal has "
4458 "corrupt superblock");
4463 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4464 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4469 len = ext4_blocks_count(es);
4470 start = sb_block + 1;
4471 brelse(bh); /* we're done with the superblock */
4473 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4474 start, len, blocksize);
4476 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4479 journal->j_private = sb;
4480 ll_rw_block(REQ_OP_READ, REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4481 wait_on_buffer(journal->j_sb_buffer);
4482 if (!buffer_uptodate(journal->j_sb_buffer)) {
4483 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4486 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4487 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4488 "user (unsupported) - %d",
4489 be32_to_cpu(journal->j_superblock->s_nr_users));
4492 EXT4_SB(sb)->journal_bdev = bdev;
4493 ext4_init_journal_params(sb, journal);
4497 jbd2_journal_destroy(journal);
4499 ext4_blkdev_put(bdev);
4503 static int ext4_load_journal(struct super_block *sb,
4504 struct ext4_super_block *es,
4505 unsigned long journal_devnum)
4508 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4511 int really_read_only;
4513 BUG_ON(!ext4_has_feature_journal(sb));
4515 if (journal_devnum &&
4516 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4517 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4518 "numbers have changed");
4519 journal_dev = new_decode_dev(journal_devnum);
4521 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4523 really_read_only = bdev_read_only(sb->s_bdev);
4526 * Are we loading a blank journal or performing recovery after a
4527 * crash? For recovery, we need to check in advance whether we
4528 * can get read-write access to the device.
4530 if (ext4_has_feature_journal_needs_recovery(sb)) {
4531 if (sb->s_flags & MS_RDONLY) {
4532 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4533 "required on readonly filesystem");
4534 if (really_read_only) {
4535 ext4_msg(sb, KERN_ERR, "write access "
4536 "unavailable, cannot proceed");
4539 ext4_msg(sb, KERN_INFO, "write access will "
4540 "be enabled during recovery");
4544 if (journal_inum && journal_dev) {
4545 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4546 "and inode journals!");
4551 if (!(journal = ext4_get_journal(sb, journal_inum)))
4554 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4558 if (!(journal->j_flags & JBD2_BARRIER))
4559 ext4_msg(sb, KERN_INFO, "barriers disabled");
4561 if (!ext4_has_feature_journal_needs_recovery(sb))
4562 err = jbd2_journal_wipe(journal, !really_read_only);
4564 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4566 memcpy(save, ((char *) es) +
4567 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4568 err = jbd2_journal_load(journal);
4570 memcpy(((char *) es) + EXT4_S_ERR_START,
4571 save, EXT4_S_ERR_LEN);
4576 ext4_msg(sb, KERN_ERR, "error loading journal");
4577 jbd2_journal_destroy(journal);
4581 EXT4_SB(sb)->s_journal = journal;
4582 ext4_clear_journal_err(sb, es);
4584 if (!really_read_only && journal_devnum &&
4585 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4586 es->s_journal_dev = cpu_to_le32(journal_devnum);
4588 /* Make sure we flush the recovery flag to disk. */
4589 ext4_commit_super(sb, 1);
4595 static int ext4_commit_super(struct super_block *sb, int sync)
4597 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4598 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4601 if (!sbh || block_device_ejected(sb))
4604 * If the file system is mounted read-only, don't update the
4605 * superblock write time. This avoids updating the superblock
4606 * write time when we are mounting the root file system
4607 * read/only but we need to replay the journal; at that point,
4608 * for people who are east of GMT and who make their clock
4609 * tick in localtime for Windows bug-for-bug compatibility,
4610 * the clock is set in the future, and this will cause e2fsck
4611 * to complain and force a full file system check.
4613 if (!(sb->s_flags & MS_RDONLY))
4614 es->s_wtime = cpu_to_le32(get_seconds());
4615 if (sb->s_bdev->bd_part)
4616 es->s_kbytes_written =
4617 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4618 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4619 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4621 es->s_kbytes_written =
4622 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4623 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4624 ext4_free_blocks_count_set(es,
4625 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4626 &EXT4_SB(sb)->s_freeclusters_counter)));
4627 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4628 es->s_free_inodes_count =
4629 cpu_to_le32(percpu_counter_sum_positive(
4630 &EXT4_SB(sb)->s_freeinodes_counter));
4631 BUFFER_TRACE(sbh, "marking dirty");
4632 ext4_superblock_csum_set(sb);
4635 if (buffer_write_io_error(sbh)) {
4637 * Oh, dear. A previous attempt to write the
4638 * superblock failed. This could happen because the
4639 * USB device was yanked out. Or it could happen to
4640 * be a transient write error and maybe the block will
4641 * be remapped. Nothing we can do but to retry the
4642 * write and hope for the best.
4644 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4645 "superblock detected");
4646 clear_buffer_write_io_error(sbh);
4647 set_buffer_uptodate(sbh);
4649 mark_buffer_dirty(sbh);
4652 error = __sync_dirty_buffer(sbh,
4653 test_opt(sb, BARRIER) ? REQ_FUA : REQ_SYNC);
4657 error = buffer_write_io_error(sbh);
4659 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4661 clear_buffer_write_io_error(sbh);
4662 set_buffer_uptodate(sbh);
4669 * Have we just finished recovery? If so, and if we are mounting (or
4670 * remounting) the filesystem readonly, then we will end up with a
4671 * consistent fs on disk. Record that fact.
4673 static void ext4_mark_recovery_complete(struct super_block *sb,
4674 struct ext4_super_block *es)
4676 journal_t *journal = EXT4_SB(sb)->s_journal;
4678 if (!ext4_has_feature_journal(sb)) {
4679 BUG_ON(journal != NULL);
4682 jbd2_journal_lock_updates(journal);
4683 if (jbd2_journal_flush(journal) < 0)
4686 if (ext4_has_feature_journal_needs_recovery(sb) &&
4687 sb->s_flags & MS_RDONLY) {
4688 ext4_clear_feature_journal_needs_recovery(sb);
4689 ext4_commit_super(sb, 1);
4693 jbd2_journal_unlock_updates(journal);
4697 * If we are mounting (or read-write remounting) a filesystem whose journal
4698 * has recorded an error from a previous lifetime, move that error to the
4699 * main filesystem now.
4701 static void ext4_clear_journal_err(struct super_block *sb,
4702 struct ext4_super_block *es)
4708 BUG_ON(!ext4_has_feature_journal(sb));
4710 journal = EXT4_SB(sb)->s_journal;
4713 * Now check for any error status which may have been recorded in the
4714 * journal by a prior ext4_error() or ext4_abort()
4717 j_errno = jbd2_journal_errno(journal);
4721 errstr = ext4_decode_error(sb, j_errno, nbuf);
4722 ext4_warning(sb, "Filesystem error recorded "
4723 "from previous mount: %s", errstr);
4724 ext4_warning(sb, "Marking fs in need of filesystem check.");
4726 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4727 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4728 ext4_commit_super(sb, 1);
4730 jbd2_journal_clear_err(journal);
4731 jbd2_journal_update_sb_errno(journal);
4736 * Force the running and committing transactions to commit,
4737 * and wait on the commit.
4739 int ext4_force_commit(struct super_block *sb)
4743 if (sb->s_flags & MS_RDONLY)
4746 journal = EXT4_SB(sb)->s_journal;
4747 return ext4_journal_force_commit(journal);
4750 static int ext4_sync_fs(struct super_block *sb, int wait)
4754 bool needs_barrier = false;
4755 struct ext4_sb_info *sbi = EXT4_SB(sb);
4757 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
4760 trace_ext4_sync_fs(sb, wait);
4761 flush_workqueue(sbi->rsv_conversion_wq);
4763 * Writeback quota in non-journalled quota case - journalled quota has
4766 dquot_writeback_dquots(sb, -1);
4768 * Data writeback is possible w/o journal transaction, so barrier must
4769 * being sent at the end of the function. But we can skip it if
4770 * transaction_commit will do it for us.
4772 if (sbi->s_journal) {
4773 target = jbd2_get_latest_transaction(sbi->s_journal);
4774 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4775 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4776 needs_barrier = true;
4778 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4780 ret = jbd2_log_wait_commit(sbi->s_journal,
4783 } else if (wait && test_opt(sb, BARRIER))
4784 needs_barrier = true;
4785 if (needs_barrier) {
4787 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4796 * LVM calls this function before a (read-only) snapshot is created. This
4797 * gives us a chance to flush the journal completely and mark the fs clean.
4799 * Note that only this function cannot bring a filesystem to be in a clean
4800 * state independently. It relies on upper layer to stop all data & metadata
4803 static int ext4_freeze(struct super_block *sb)
4808 if (sb->s_flags & MS_RDONLY)
4811 journal = EXT4_SB(sb)->s_journal;
4814 /* Now we set up the journal barrier. */
4815 jbd2_journal_lock_updates(journal);
4818 * Don't clear the needs_recovery flag if we failed to
4819 * flush the journal.
4821 error = jbd2_journal_flush(journal);
4825 /* Journal blocked and flushed, clear needs_recovery flag. */
4826 ext4_clear_feature_journal_needs_recovery(sb);
4829 error = ext4_commit_super(sb, 1);
4832 /* we rely on upper layer to stop further updates */
4833 jbd2_journal_unlock_updates(journal);
4838 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4839 * flag here, even though the filesystem is not technically dirty yet.
4841 static int ext4_unfreeze(struct super_block *sb)
4843 if ((sb->s_flags & MS_RDONLY) || ext4_forced_shutdown(EXT4_SB(sb)))
4846 if (EXT4_SB(sb)->s_journal) {
4847 /* Reset the needs_recovery flag before the fs is unlocked. */
4848 ext4_set_feature_journal_needs_recovery(sb);
4851 ext4_commit_super(sb, 1);
4856 * Structure to save mount options for ext4_remount's benefit
4858 struct ext4_mount_options {
4859 unsigned long s_mount_opt;
4860 unsigned long s_mount_opt2;
4863 unsigned long s_commit_interval;
4864 u32 s_min_batch_time, s_max_batch_time;
4867 char *s_qf_names[EXT4_MAXQUOTAS];
4871 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4873 struct ext4_super_block *es;
4874 struct ext4_sb_info *sbi = EXT4_SB(sb);
4875 unsigned long old_sb_flags;
4876 struct ext4_mount_options old_opts;
4877 int enable_quota = 0;
4879 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4884 char *orig_data = kstrdup(data, GFP_KERNEL);
4886 /* Store the original options */
4887 old_sb_flags = sb->s_flags;
4888 old_opts.s_mount_opt = sbi->s_mount_opt;
4889 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4890 old_opts.s_resuid = sbi->s_resuid;
4891 old_opts.s_resgid = sbi->s_resgid;
4892 old_opts.s_commit_interval = sbi->s_commit_interval;
4893 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4894 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4896 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4897 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4898 if (sbi->s_qf_names[i]) {
4899 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4901 if (!old_opts.s_qf_names[i]) {
4902 for (j = 0; j < i; j++)
4903 kfree(old_opts.s_qf_names[j]);
4908 old_opts.s_qf_names[i] = NULL;
4910 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4911 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4913 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4918 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
4919 test_opt(sb, JOURNAL_CHECKSUM)) {
4920 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
4921 "during remount not supported; ignoring");
4922 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
4925 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4926 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4927 ext4_msg(sb, KERN_ERR, "can't mount with "
4928 "both data=journal and delalloc");
4932 if (test_opt(sb, DIOREAD_NOLOCK)) {
4933 ext4_msg(sb, KERN_ERR, "can't mount with "
4934 "both data=journal and dioread_nolock");
4938 if (test_opt(sb, DAX)) {
4939 ext4_msg(sb, KERN_ERR, "can't mount with "
4940 "both data=journal and dax");
4944 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
4945 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4946 ext4_msg(sb, KERN_ERR, "can't mount with "
4947 "journal_async_commit in data=ordered mode");
4953 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
4954 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
4955 "dax flag with busy inodes while remounting");
4956 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
4959 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4960 ext4_abort(sb, "Abort forced by user");
4962 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4963 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4967 if (sbi->s_journal) {
4968 ext4_init_journal_params(sb, sbi->s_journal);
4969 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4972 if (*flags & MS_LAZYTIME)
4973 sb->s_flags |= MS_LAZYTIME;
4975 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4976 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4981 if (*flags & MS_RDONLY) {
4982 err = sync_filesystem(sb);
4985 err = dquot_suspend(sb, -1);
4990 * First of all, the unconditional stuff we have to do
4991 * to disable replay of the journal when we next remount
4993 sb->s_flags |= MS_RDONLY;
4996 * OK, test if we are remounting a valid rw partition
4997 * readonly, and if so set the rdonly flag and then
4998 * mark the partition as valid again.
5000 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
5001 (sbi->s_mount_state & EXT4_VALID_FS))
5002 es->s_state = cpu_to_le16(sbi->s_mount_state);
5005 ext4_mark_recovery_complete(sb, es);
5007 /* Make sure we can mount this feature set readwrite */
5008 if (ext4_has_feature_readonly(sb) ||
5009 !ext4_feature_set_ok(sb, 0)) {
5014 * Make sure the group descriptor checksums
5015 * are sane. If they aren't, refuse to remount r/w.
5017 for (g = 0; g < sbi->s_groups_count; g++) {
5018 struct ext4_group_desc *gdp =
5019 ext4_get_group_desc(sb, g, NULL);
5021 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
5022 ext4_msg(sb, KERN_ERR,
5023 "ext4_remount: Checksum for group %u failed (%u!=%u)",
5024 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
5025 le16_to_cpu(gdp->bg_checksum));
5032 * If we have an unprocessed orphan list hanging
5033 * around from a previously readonly bdev mount,
5034 * require a full umount/remount for now.
5036 if (es->s_last_orphan) {
5037 ext4_msg(sb, KERN_WARNING, "Couldn't "
5038 "remount RDWR because of unprocessed "
5039 "orphan inode list. Please "
5040 "umount/remount instead");
5046 * Mounting a RDONLY partition read-write, so reread
5047 * and store the current valid flag. (It may have
5048 * been changed by e2fsck since we originally mounted
5052 ext4_clear_journal_err(sb, es);
5053 sbi->s_mount_state = le16_to_cpu(es->s_state);
5054 if (!ext4_setup_super(sb, es, 0))
5055 sb->s_flags &= ~MS_RDONLY;
5056 if (ext4_has_feature_mmp(sb))
5057 if (ext4_multi_mount_protect(sb,
5058 le64_to_cpu(es->s_mmp_block))) {
5067 * Reinitialize lazy itable initialization thread based on
5070 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
5071 ext4_unregister_li_request(sb);
5073 ext4_group_t first_not_zeroed;
5074 first_not_zeroed = ext4_has_uninit_itable(sb);
5075 ext4_register_li_request(sb, first_not_zeroed);
5078 ext4_setup_system_zone(sb);
5079 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
5080 ext4_commit_super(sb, 1);
5083 /* Release old quota file names */
5084 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5085 kfree(old_opts.s_qf_names[i]);
5087 if (sb_any_quota_suspended(sb))
5088 dquot_resume(sb, -1);
5089 else if (ext4_has_feature_quota(sb)) {
5090 err = ext4_enable_quotas(sb);
5097 *flags = (*flags & ~MS_LAZYTIME) | (sb->s_flags & MS_LAZYTIME);
5098 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5103 sb->s_flags = old_sb_flags;
5104 sbi->s_mount_opt = old_opts.s_mount_opt;
5105 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
5106 sbi->s_resuid = old_opts.s_resuid;
5107 sbi->s_resgid = old_opts.s_resgid;
5108 sbi->s_commit_interval = old_opts.s_commit_interval;
5109 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5110 sbi->s_max_batch_time = old_opts.s_max_batch_time;
5112 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
5113 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
5114 kfree(sbi->s_qf_names[i]);
5115 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
5123 static int ext4_statfs_project(struct super_block *sb,
5124 kprojid_t projid, struct kstatfs *buf)
5127 struct dquot *dquot;
5131 qid = make_kqid_projid(projid);
5132 dquot = dqget(sb, qid);
5134 return PTR_ERR(dquot);
5135 spin_lock(&dq_data_lock);
5137 limit = (dquot->dq_dqb.dqb_bsoftlimit ?
5138 dquot->dq_dqb.dqb_bsoftlimit :
5139 dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits;
5140 if (limit && buf->f_blocks > limit) {
5141 curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits;
5142 buf->f_blocks = limit;
5143 buf->f_bfree = buf->f_bavail =
5144 (buf->f_blocks > curblock) ?
5145 (buf->f_blocks - curblock) : 0;
5148 limit = dquot->dq_dqb.dqb_isoftlimit ?
5149 dquot->dq_dqb.dqb_isoftlimit :
5150 dquot->dq_dqb.dqb_ihardlimit;
5151 if (limit && buf->f_files > limit) {
5152 buf->f_files = limit;
5154 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
5155 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
5158 spin_unlock(&dq_data_lock);
5164 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
5166 struct super_block *sb = dentry->d_sb;
5167 struct ext4_sb_info *sbi = EXT4_SB(sb);
5168 struct ext4_super_block *es = sbi->s_es;
5169 ext4_fsblk_t overhead = 0, resv_blocks;
5172 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
5174 if (!test_opt(sb, MINIX_DF))
5175 overhead = sbi->s_overhead;
5177 buf->f_type = EXT4_SUPER_MAGIC;
5178 buf->f_bsize = sb->s_blocksize;
5179 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
5180 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5181 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
5182 /* prevent underflow in case that few free space is available */
5183 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
5184 buf->f_bavail = buf->f_bfree -
5185 (ext4_r_blocks_count(es) + resv_blocks);
5186 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
5188 buf->f_files = le32_to_cpu(es->s_inodes_count);
5189 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5190 buf->f_namelen = EXT4_NAME_LEN;
5191 fsid = le64_to_cpup((void *)es->s_uuid) ^
5192 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5193 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5194 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
5197 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
5198 sb_has_quota_limits_enabled(sb, PRJQUOTA))
5199 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
5204 /* Helper function for writing quotas on sync - we need to start transaction
5205 * before quota file is locked for write. Otherwise the are possible deadlocks:
5206 * Process 1 Process 2
5207 * ext4_create() quota_sync()
5208 * jbd2_journal_start() write_dquot()
5209 * dquot_initialize() down(dqio_mutex)
5210 * down(dqio_mutex) jbd2_journal_start()
5216 static inline struct inode *dquot_to_inode(struct dquot *dquot)
5218 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
5221 static int ext4_write_dquot(struct dquot *dquot)
5225 struct inode *inode;
5227 inode = dquot_to_inode(dquot);
5228 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
5229 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
5231 return PTR_ERR(handle);
5232 ret = dquot_commit(dquot);
5233 err = ext4_journal_stop(handle);
5239 static int ext4_acquire_dquot(struct dquot *dquot)
5244 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5245 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
5247 return PTR_ERR(handle);
5248 ret = dquot_acquire(dquot);
5249 err = ext4_journal_stop(handle);
5255 static int ext4_release_dquot(struct dquot *dquot)
5260 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5261 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
5262 if (IS_ERR(handle)) {
5263 /* Release dquot anyway to avoid endless cycle in dqput() */
5264 dquot_release(dquot);
5265 return PTR_ERR(handle);
5267 ret = dquot_release(dquot);
5268 err = ext4_journal_stop(handle);
5274 static int ext4_mark_dquot_dirty(struct dquot *dquot)
5276 struct super_block *sb = dquot->dq_sb;
5277 struct ext4_sb_info *sbi = EXT4_SB(sb);
5279 /* Are we journaling quotas? */
5280 if (ext4_has_feature_quota(sb) ||
5281 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
5282 dquot_mark_dquot_dirty(dquot);
5283 return ext4_write_dquot(dquot);
5285 return dquot_mark_dquot_dirty(dquot);
5289 static int ext4_write_info(struct super_block *sb, int type)
5294 /* Data block + inode block */
5295 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
5297 return PTR_ERR(handle);
5298 ret = dquot_commit_info(sb, type);
5299 err = ext4_journal_stop(handle);
5306 * Turn on quotas during mount time - we need to find
5307 * the quota file and such...
5309 static int ext4_quota_on_mount(struct super_block *sb, int type)
5311 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5312 EXT4_SB(sb)->s_jquota_fmt, type);
5315 static void lockdep_set_quota_inode(struct inode *inode, int subclass)
5317 struct ext4_inode_info *ei = EXT4_I(inode);
5319 /* The first argument of lockdep_set_subclass has to be
5320 * *exactly* the same as the argument to init_rwsem() --- in
5321 * this case, in init_once() --- or lockdep gets unhappy
5322 * because the name of the lock is set using the
5323 * stringification of the argument to init_rwsem().
5325 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
5326 lockdep_set_subclass(&ei->i_data_sem, subclass);
5330 * Standard function to be called on quota_on
5332 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
5333 const struct path *path)
5337 if (!test_opt(sb, QUOTA))
5340 /* Quotafile not on the same filesystem? */
5341 if (path->dentry->d_sb != sb)
5343 /* Journaling quota? */
5344 if (EXT4_SB(sb)->s_qf_names[type]) {
5345 /* Quotafile not in fs root? */
5346 if (path->dentry->d_parent != sb->s_root)
5347 ext4_msg(sb, KERN_WARNING,
5348 "Quota file not on filesystem root. "
5349 "Journaled quota will not work");
5353 * When we journal data on quota file, we have to flush journal to see
5354 * all updates to the file when we bypass pagecache...
5356 if (EXT4_SB(sb)->s_journal &&
5357 ext4_should_journal_data(d_inode(path->dentry))) {
5359 * We don't need to lock updates but journal_flush() could
5360 * otherwise be livelocked...
5362 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
5363 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
5364 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
5369 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
5370 err = dquot_quota_on(sb, type, format_id, path);
5372 lockdep_set_quota_inode(path->dentry->d_inode,
5375 struct inode *inode = d_inode(path->dentry);
5379 * Set inode flags to prevent userspace from messing with quota
5380 * files. If this fails, we return success anyway since quotas
5381 * are already enabled and this is not a hard failure.
5384 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5387 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL;
5388 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
5389 S_NOATIME | S_IMMUTABLE);
5390 ext4_mark_inode_dirty(handle, inode);
5391 ext4_journal_stop(handle);
5393 inode_unlock(inode);
5398 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5402 struct inode *qf_inode;
5403 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5404 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5405 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5406 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
5409 BUG_ON(!ext4_has_feature_quota(sb));
5411 if (!qf_inums[type])
5414 qf_inode = ext4_iget(sb, qf_inums[type]);
5415 if (IS_ERR(qf_inode)) {
5416 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5417 return PTR_ERR(qf_inode);
5420 /* Don't account quota for quota files to avoid recursion */
5421 qf_inode->i_flags |= S_NOQUOTA;
5422 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
5423 err = dquot_enable(qf_inode, type, format_id, flags);
5426 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
5431 /* Enable usage tracking for all quota types. */
5432 static int ext4_enable_quotas(struct super_block *sb)
5435 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5436 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5437 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5438 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
5440 bool quota_mopt[EXT4_MAXQUOTAS] = {
5441 test_opt(sb, USRQUOTA),
5442 test_opt(sb, GRPQUOTA),
5443 test_opt(sb, PRJQUOTA),
5446 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5447 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
5448 if (qf_inums[type]) {
5449 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5450 DQUOT_USAGE_ENABLED |
5451 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
5454 "Failed to enable quota tracking "
5455 "(type=%d, err=%d). Please run "
5456 "e2fsck to fix.", type, err);
5464 static int ext4_quota_off(struct super_block *sb, int type)
5466 struct inode *inode = sb_dqopt(sb)->files[type];
5470 /* Force all delayed allocation blocks to be allocated.
5471 * Caller already holds s_umount sem */
5472 if (test_opt(sb, DELALLOC))
5473 sync_filesystem(sb);
5475 if (!inode || !igrab(inode))
5478 err = dquot_quota_off(sb, type);
5484 * Update modification times of quota files when userspace can
5485 * start looking at them. If we fail, we return success anyway since
5486 * this is not a hard failure and quotas are already disabled.
5488 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5491 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL);
5492 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
5493 inode->i_mtime = inode->i_ctime = current_time(inode);
5494 ext4_mark_inode_dirty(handle, inode);
5495 ext4_journal_stop(handle);
5497 inode_unlock(inode);
5502 return dquot_quota_off(sb, type);
5505 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5506 * acquiring the locks... As quota files are never truncated and quota code
5507 * itself serializes the operations (and no one else should touch the files)
5508 * we don't have to be afraid of races */
5509 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5510 size_t len, loff_t off)
5512 struct inode *inode = sb_dqopt(sb)->files[type];
5513 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5514 int offset = off & (sb->s_blocksize - 1);
5517 struct buffer_head *bh;
5518 loff_t i_size = i_size_read(inode);
5522 if (off+len > i_size)
5525 while (toread > 0) {
5526 tocopy = sb->s_blocksize - offset < toread ?
5527 sb->s_blocksize - offset : toread;
5528 bh = ext4_bread(NULL, inode, blk, 0);
5531 if (!bh) /* A hole? */
5532 memset(data, 0, tocopy);
5534 memcpy(data, bh->b_data+offset, tocopy);
5544 /* Write to quotafile (we know the transaction is already started and has
5545 * enough credits) */
5546 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5547 const char *data, size_t len, loff_t off)
5549 struct inode *inode = sb_dqopt(sb)->files[type];
5550 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5551 int err, offset = off & (sb->s_blocksize - 1);
5553 struct buffer_head *bh;
5554 handle_t *handle = journal_current_handle();
5556 if (EXT4_SB(sb)->s_journal && !handle) {
5557 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5558 " cancelled because transaction is not started",
5559 (unsigned long long)off, (unsigned long long)len);
5563 * Since we account only one data block in transaction credits,
5564 * then it is impossible to cross a block boundary.
5566 if (sb->s_blocksize - offset < len) {
5567 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5568 " cancelled because not block aligned",
5569 (unsigned long long)off, (unsigned long long)len);
5574 bh = ext4_bread(handle, inode, blk,
5575 EXT4_GET_BLOCKS_CREATE |
5576 EXT4_GET_BLOCKS_METADATA_NOFAIL);
5577 } while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) &&
5578 ext4_should_retry_alloc(inode->i_sb, &retries));
5583 BUFFER_TRACE(bh, "get write access");
5584 err = ext4_journal_get_write_access(handle, bh);
5590 memcpy(bh->b_data+offset, data, len);
5591 flush_dcache_page(bh->b_page);
5593 err = ext4_handle_dirty_metadata(handle, NULL, bh);
5596 if (inode->i_size < off + len) {
5597 i_size_write(inode, off + len);
5598 EXT4_I(inode)->i_disksize = inode->i_size;
5599 ext4_mark_inode_dirty(handle, inode);
5604 static int ext4_get_next_id(struct super_block *sb, struct kqid *qid)
5606 const struct quota_format_ops *ops;
5608 if (!sb_has_quota_loaded(sb, qid->type))
5610 ops = sb_dqopt(sb)->ops[qid->type];
5611 if (!ops || !ops->get_next_id)
5613 return dquot_get_next_id(sb, qid);
5617 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5618 const char *dev_name, void *data)
5620 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5623 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
5624 static inline void register_as_ext2(void)
5626 int err = register_filesystem(&ext2_fs_type);
5629 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5632 static inline void unregister_as_ext2(void)
5634 unregister_filesystem(&ext2_fs_type);
5637 static inline int ext2_feature_set_ok(struct super_block *sb)
5639 if (ext4_has_unknown_ext2_incompat_features(sb))
5641 if (sb->s_flags & MS_RDONLY)
5643 if (ext4_has_unknown_ext2_ro_compat_features(sb))
5648 static inline void register_as_ext2(void) { }
5649 static inline void unregister_as_ext2(void) { }
5650 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5653 static inline void register_as_ext3(void)
5655 int err = register_filesystem(&ext3_fs_type);
5658 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5661 static inline void unregister_as_ext3(void)
5663 unregister_filesystem(&ext3_fs_type);
5666 static inline int ext3_feature_set_ok(struct super_block *sb)
5668 if (ext4_has_unknown_ext3_incompat_features(sb))
5670 if (!ext4_has_feature_journal(sb))
5672 if (sb->s_flags & MS_RDONLY)
5674 if (ext4_has_unknown_ext3_ro_compat_features(sb))
5679 static struct file_system_type ext4_fs_type = {
5680 .owner = THIS_MODULE,
5682 .mount = ext4_mount,
5683 .kill_sb = kill_block_super,
5684 .fs_flags = FS_REQUIRES_DEV,
5686 MODULE_ALIAS_FS("ext4");
5688 /* Shared across all ext4 file systems */
5689 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5691 static int __init ext4_init_fs(void)
5695 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
5696 ext4_li_info = NULL;
5697 mutex_init(&ext4_li_mtx);
5699 /* Build-time check for flags consistency */
5700 ext4_check_flag_values();
5702 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
5703 init_waitqueue_head(&ext4__ioend_wq[i]);
5705 err = ext4_init_es();
5709 err = ext4_init_pageio();
5713 err = ext4_init_system_zone();
5717 err = ext4_init_sysfs();
5721 err = ext4_init_mballoc();
5724 err = init_inodecache();
5729 err = register_filesystem(&ext4_fs_type);
5735 unregister_as_ext2();
5736 unregister_as_ext3();
5737 destroy_inodecache();
5739 ext4_exit_mballoc();
5743 ext4_exit_system_zone();
5752 static void __exit ext4_exit_fs(void)
5754 ext4_destroy_lazyinit_thread();
5755 unregister_as_ext2();
5756 unregister_as_ext3();
5757 unregister_filesystem(&ext4_fs_type);
5758 destroy_inodecache();
5759 ext4_exit_mballoc();
5761 ext4_exit_system_zone();
5766 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5767 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5768 MODULE_LICENSE("GPL");
5769 module_init(ext4_init_fs)
5770 module_exit(ext4_exit_fs)