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_prepare_context(struct inode *inode)
1147 return ext4_convert_inline_data(inode);
1150 static int ext4_set_context(struct inode *inode, const void *ctx, size_t len,
1153 handle_t *handle = fs_data;
1154 int res, res2, retries = 0;
1157 * If a journal handle was specified, then the encryption context is
1158 * being set on a new inode via inheritance and is part of a larger
1159 * transaction to create the inode. Otherwise the encryption context is
1160 * being set on an existing inode in its own transaction. Only in the
1161 * latter case should the "retry on ENOSPC" logic be used.
1165 res = ext4_xattr_set_handle(handle, inode,
1166 EXT4_XATTR_INDEX_ENCRYPTION,
1167 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1170 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1171 ext4_clear_inode_state(inode,
1172 EXT4_STATE_MAY_INLINE_DATA);
1174 * Update inode->i_flags - e.g. S_DAX may get disabled
1176 ext4_set_inode_flags(inode);
1182 handle = ext4_journal_start(inode, EXT4_HT_MISC,
1183 ext4_jbd2_credits_xattr(inode));
1185 return PTR_ERR(handle);
1187 res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION,
1188 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1191 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1192 /* Update inode->i_flags - e.g. S_DAX may get disabled */
1193 ext4_set_inode_flags(inode);
1194 res = ext4_mark_inode_dirty(handle, inode);
1196 EXT4_ERROR_INODE(inode, "Failed to mark inode dirty");
1198 res2 = ext4_journal_stop(handle);
1200 if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
1207 static int ext4_dummy_context(struct inode *inode)
1209 return DUMMY_ENCRYPTION_ENABLED(EXT4_SB(inode->i_sb));
1212 static unsigned ext4_max_namelen(struct inode *inode)
1214 return S_ISLNK(inode->i_mode) ? inode->i_sb->s_blocksize :
1218 static const struct fscrypt_operations ext4_cryptops = {
1219 .key_prefix = "ext4:",
1220 .get_context = ext4_get_context,
1221 .prepare_context = ext4_prepare_context,
1222 .set_context = ext4_set_context,
1223 .dummy_context = ext4_dummy_context,
1224 .is_encrypted = ext4_encrypted_inode,
1225 .empty_dir = ext4_empty_dir,
1226 .max_namelen = ext4_max_namelen,
1229 static const struct fscrypt_operations ext4_cryptops = {
1230 .is_encrypted = ext4_encrypted_inode,
1235 static char *quotatypes[] = INITQFNAMES;
1236 #define QTYPE2NAME(t) (quotatypes[t])
1238 static int ext4_write_dquot(struct dquot *dquot);
1239 static int ext4_acquire_dquot(struct dquot *dquot);
1240 static int ext4_release_dquot(struct dquot *dquot);
1241 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1242 static int ext4_write_info(struct super_block *sb, int type);
1243 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1244 const struct path *path);
1245 static int ext4_quota_on_mount(struct super_block *sb, int type);
1246 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1247 size_t len, loff_t off);
1248 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1249 const char *data, size_t len, loff_t off);
1250 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1251 unsigned int flags);
1252 static int ext4_enable_quotas(struct super_block *sb);
1253 static int ext4_get_next_id(struct super_block *sb, struct kqid *qid);
1255 static struct dquot **ext4_get_dquots(struct inode *inode)
1257 return EXT4_I(inode)->i_dquot;
1260 static const struct dquot_operations ext4_quota_operations = {
1261 .get_reserved_space = ext4_get_reserved_space,
1262 .write_dquot = ext4_write_dquot,
1263 .acquire_dquot = ext4_acquire_dquot,
1264 .release_dquot = ext4_release_dquot,
1265 .mark_dirty = ext4_mark_dquot_dirty,
1266 .write_info = ext4_write_info,
1267 .alloc_dquot = dquot_alloc,
1268 .destroy_dquot = dquot_destroy,
1269 .get_projid = ext4_get_projid,
1270 .get_next_id = ext4_get_next_id,
1273 static const struct quotactl_ops ext4_qctl_operations = {
1274 .quota_on = ext4_quota_on,
1275 .quota_off = ext4_quota_off,
1276 .quota_sync = dquot_quota_sync,
1277 .get_state = dquot_get_state,
1278 .set_info = dquot_set_dqinfo,
1279 .get_dqblk = dquot_get_dqblk,
1280 .set_dqblk = dquot_set_dqblk,
1281 .get_nextdqblk = dquot_get_next_dqblk,
1285 static const struct super_operations ext4_sops = {
1286 .alloc_inode = ext4_alloc_inode,
1287 .destroy_inode = ext4_destroy_inode,
1288 .write_inode = ext4_write_inode,
1289 .dirty_inode = ext4_dirty_inode,
1290 .drop_inode = ext4_drop_inode,
1291 .evict_inode = ext4_evict_inode,
1292 .put_super = ext4_put_super,
1293 .sync_fs = ext4_sync_fs,
1294 .freeze_fs = ext4_freeze,
1295 .unfreeze_fs = ext4_unfreeze,
1296 .statfs = ext4_statfs,
1297 .remount_fs = ext4_remount,
1298 .show_options = ext4_show_options,
1300 .quota_read = ext4_quota_read,
1301 .quota_write = ext4_quota_write,
1302 .get_dquots = ext4_get_dquots,
1304 .bdev_try_to_free_page = bdev_try_to_free_page,
1307 static const struct export_operations ext4_export_ops = {
1308 .fh_to_dentry = ext4_fh_to_dentry,
1309 .fh_to_parent = ext4_fh_to_parent,
1310 .get_parent = ext4_get_parent,
1314 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1315 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1316 Opt_nouid32, Opt_debug, Opt_removed,
1317 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1318 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1319 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1320 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1321 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1322 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1323 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1324 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1325 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1326 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version, Opt_dax,
1327 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1328 Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize,
1329 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1330 Opt_inode_readahead_blks, Opt_journal_ioprio,
1331 Opt_dioread_nolock, Opt_dioread_lock,
1332 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1333 Opt_max_dir_size_kb, Opt_nojournal_checksum,
1336 static const match_table_t tokens = {
1337 {Opt_bsd_df, "bsddf"},
1338 {Opt_minix_df, "minixdf"},
1339 {Opt_grpid, "grpid"},
1340 {Opt_grpid, "bsdgroups"},
1341 {Opt_nogrpid, "nogrpid"},
1342 {Opt_nogrpid, "sysvgroups"},
1343 {Opt_resgid, "resgid=%u"},
1344 {Opt_resuid, "resuid=%u"},
1346 {Opt_err_cont, "errors=continue"},
1347 {Opt_err_panic, "errors=panic"},
1348 {Opt_err_ro, "errors=remount-ro"},
1349 {Opt_nouid32, "nouid32"},
1350 {Opt_debug, "debug"},
1351 {Opt_removed, "oldalloc"},
1352 {Opt_removed, "orlov"},
1353 {Opt_user_xattr, "user_xattr"},
1354 {Opt_nouser_xattr, "nouser_xattr"},
1356 {Opt_noacl, "noacl"},
1357 {Opt_noload, "norecovery"},
1358 {Opt_noload, "noload"},
1359 {Opt_removed, "nobh"},
1360 {Opt_removed, "bh"},
1361 {Opt_commit, "commit=%u"},
1362 {Opt_min_batch_time, "min_batch_time=%u"},
1363 {Opt_max_batch_time, "max_batch_time=%u"},
1364 {Opt_journal_dev, "journal_dev=%u"},
1365 {Opt_journal_path, "journal_path=%s"},
1366 {Opt_journal_checksum, "journal_checksum"},
1367 {Opt_nojournal_checksum, "nojournal_checksum"},
1368 {Opt_journal_async_commit, "journal_async_commit"},
1369 {Opt_abort, "abort"},
1370 {Opt_data_journal, "data=journal"},
1371 {Opt_data_ordered, "data=ordered"},
1372 {Opt_data_writeback, "data=writeback"},
1373 {Opt_data_err_abort, "data_err=abort"},
1374 {Opt_data_err_ignore, "data_err=ignore"},
1375 {Opt_offusrjquota, "usrjquota="},
1376 {Opt_usrjquota, "usrjquota=%s"},
1377 {Opt_offgrpjquota, "grpjquota="},
1378 {Opt_grpjquota, "grpjquota=%s"},
1379 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1380 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1381 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1382 {Opt_grpquota, "grpquota"},
1383 {Opt_noquota, "noquota"},
1384 {Opt_quota, "quota"},
1385 {Opt_usrquota, "usrquota"},
1386 {Opt_prjquota, "prjquota"},
1387 {Opt_barrier, "barrier=%u"},
1388 {Opt_barrier, "barrier"},
1389 {Opt_nobarrier, "nobarrier"},
1390 {Opt_i_version, "i_version"},
1392 {Opt_stripe, "stripe=%u"},
1393 {Opt_delalloc, "delalloc"},
1394 {Opt_lazytime, "lazytime"},
1395 {Opt_nolazytime, "nolazytime"},
1396 {Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"},
1397 {Opt_nodelalloc, "nodelalloc"},
1398 {Opt_removed, "mblk_io_submit"},
1399 {Opt_removed, "nomblk_io_submit"},
1400 {Opt_block_validity, "block_validity"},
1401 {Opt_noblock_validity, "noblock_validity"},
1402 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1403 {Opt_journal_ioprio, "journal_ioprio=%u"},
1404 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1405 {Opt_auto_da_alloc, "auto_da_alloc"},
1406 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1407 {Opt_dioread_nolock, "dioread_nolock"},
1408 {Opt_dioread_lock, "dioread_lock"},
1409 {Opt_discard, "discard"},
1410 {Opt_nodiscard, "nodiscard"},
1411 {Opt_init_itable, "init_itable=%u"},
1412 {Opt_init_itable, "init_itable"},
1413 {Opt_noinit_itable, "noinit_itable"},
1414 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1415 {Opt_test_dummy_encryption, "test_dummy_encryption"},
1416 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1417 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1418 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1419 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1420 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1424 static ext4_fsblk_t get_sb_block(void **data)
1426 ext4_fsblk_t sb_block;
1427 char *options = (char *) *data;
1429 if (!options || strncmp(options, "sb=", 3) != 0)
1430 return 1; /* Default location */
1433 /* TODO: use simple_strtoll with >32bit ext4 */
1434 sb_block = simple_strtoul(options, &options, 0);
1435 if (*options && *options != ',') {
1436 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1440 if (*options == ',')
1442 *data = (void *) options;
1447 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1448 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1449 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1452 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1454 struct ext4_sb_info *sbi = EXT4_SB(sb);
1458 if (sb_any_quota_loaded(sb) &&
1459 !sbi->s_qf_names[qtype]) {
1460 ext4_msg(sb, KERN_ERR,
1461 "Cannot change journaled "
1462 "quota options when quota turned on");
1465 if (ext4_has_feature_quota(sb)) {
1466 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1467 "ignored when QUOTA feature is enabled");
1470 qname = match_strdup(args);
1472 ext4_msg(sb, KERN_ERR,
1473 "Not enough memory for storing quotafile name");
1476 if (sbi->s_qf_names[qtype]) {
1477 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1480 ext4_msg(sb, KERN_ERR,
1481 "%s quota file already specified",
1485 if (strchr(qname, '/')) {
1486 ext4_msg(sb, KERN_ERR,
1487 "quotafile must be on filesystem root");
1490 sbi->s_qf_names[qtype] = qname;
1498 static int clear_qf_name(struct super_block *sb, int qtype)
1501 struct ext4_sb_info *sbi = EXT4_SB(sb);
1503 if (sb_any_quota_loaded(sb) &&
1504 sbi->s_qf_names[qtype]) {
1505 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1506 " when quota turned on");
1509 kfree(sbi->s_qf_names[qtype]);
1510 sbi->s_qf_names[qtype] = NULL;
1515 #define MOPT_SET 0x0001
1516 #define MOPT_CLEAR 0x0002
1517 #define MOPT_NOSUPPORT 0x0004
1518 #define MOPT_EXPLICIT 0x0008
1519 #define MOPT_CLEAR_ERR 0x0010
1520 #define MOPT_GTE0 0x0020
1523 #define MOPT_QFMT 0x0040
1525 #define MOPT_Q MOPT_NOSUPPORT
1526 #define MOPT_QFMT MOPT_NOSUPPORT
1528 #define MOPT_DATAJ 0x0080
1529 #define MOPT_NO_EXT2 0x0100
1530 #define MOPT_NO_EXT3 0x0200
1531 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1532 #define MOPT_STRING 0x0400
1534 static const struct mount_opts {
1538 } ext4_mount_opts[] = {
1539 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1540 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1541 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1542 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1543 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1544 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1545 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1546 MOPT_EXT4_ONLY | MOPT_SET},
1547 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1548 MOPT_EXT4_ONLY | MOPT_CLEAR},
1549 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1550 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1551 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1552 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1553 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1554 MOPT_EXT4_ONLY | MOPT_CLEAR},
1555 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1556 MOPT_EXT4_ONLY | MOPT_CLEAR},
1557 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1558 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1559 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1560 EXT4_MOUNT_JOURNAL_CHECKSUM),
1561 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1562 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1563 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1564 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1565 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1566 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1568 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1570 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1571 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1572 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1573 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1574 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1575 {Opt_commit, 0, MOPT_GTE0},
1576 {Opt_max_batch_time, 0, MOPT_GTE0},
1577 {Opt_min_batch_time, 0, MOPT_GTE0},
1578 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1579 {Opt_init_itable, 0, MOPT_GTE0},
1580 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
1581 {Opt_stripe, 0, MOPT_GTE0},
1582 {Opt_resuid, 0, MOPT_GTE0},
1583 {Opt_resgid, 0, MOPT_GTE0},
1584 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1585 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1586 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1587 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1588 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1589 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1590 MOPT_NO_EXT2 | MOPT_DATAJ},
1591 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1592 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1593 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1594 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1595 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1597 {Opt_acl, 0, MOPT_NOSUPPORT},
1598 {Opt_noacl, 0, MOPT_NOSUPPORT},
1600 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1601 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1602 {Opt_debug_want_extra_isize, 0, MOPT_GTE0},
1603 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1604 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1606 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1608 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1610 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1611 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1612 MOPT_CLEAR | MOPT_Q},
1613 {Opt_usrjquota, 0, MOPT_Q},
1614 {Opt_grpjquota, 0, MOPT_Q},
1615 {Opt_offusrjquota, 0, MOPT_Q},
1616 {Opt_offgrpjquota, 0, MOPT_Q},
1617 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1618 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1619 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1620 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1621 {Opt_test_dummy_encryption, 0, MOPT_GTE0},
1625 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1626 substring_t *args, unsigned long *journal_devnum,
1627 unsigned int *journal_ioprio, int is_remount)
1629 struct ext4_sb_info *sbi = EXT4_SB(sb);
1630 const struct mount_opts *m;
1636 if (token == Opt_usrjquota)
1637 return set_qf_name(sb, USRQUOTA, &args[0]);
1638 else if (token == Opt_grpjquota)
1639 return set_qf_name(sb, GRPQUOTA, &args[0]);
1640 else if (token == Opt_offusrjquota)
1641 return clear_qf_name(sb, USRQUOTA);
1642 else if (token == Opt_offgrpjquota)
1643 return clear_qf_name(sb, GRPQUOTA);
1647 case Opt_nouser_xattr:
1648 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1651 return 1; /* handled by get_sb_block() */
1653 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1656 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1659 sb->s_flags |= MS_I_VERSION;
1662 sb->s_flags |= MS_LAZYTIME;
1664 case Opt_nolazytime:
1665 sb->s_flags &= ~MS_LAZYTIME;
1669 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1670 if (token == m->token)
1673 if (m->token == Opt_err) {
1674 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1675 "or missing value", opt);
1679 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1680 ext4_msg(sb, KERN_ERR,
1681 "Mount option \"%s\" incompatible with ext2", opt);
1684 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1685 ext4_msg(sb, KERN_ERR,
1686 "Mount option \"%s\" incompatible with ext3", opt);
1690 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
1692 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1694 if (m->flags & MOPT_EXPLICIT) {
1695 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
1696 set_opt2(sb, EXPLICIT_DELALLOC);
1697 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
1698 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
1702 if (m->flags & MOPT_CLEAR_ERR)
1703 clear_opt(sb, ERRORS_MASK);
1704 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1705 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1706 "options when quota turned on");
1710 if (m->flags & MOPT_NOSUPPORT) {
1711 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1712 } else if (token == Opt_commit) {
1714 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1715 sbi->s_commit_interval = HZ * arg;
1716 } else if (token == Opt_debug_want_extra_isize) {
1717 sbi->s_want_extra_isize = arg;
1718 } else if (token == Opt_max_batch_time) {
1719 sbi->s_max_batch_time = arg;
1720 } else if (token == Opt_min_batch_time) {
1721 sbi->s_min_batch_time = arg;
1722 } else if (token == Opt_inode_readahead_blks) {
1723 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1724 ext4_msg(sb, KERN_ERR,
1725 "EXT4-fs: inode_readahead_blks must be "
1726 "0 or a power of 2 smaller than 2^31");
1729 sbi->s_inode_readahead_blks = arg;
1730 } else if (token == Opt_init_itable) {
1731 set_opt(sb, INIT_INODE_TABLE);
1733 arg = EXT4_DEF_LI_WAIT_MULT;
1734 sbi->s_li_wait_mult = arg;
1735 } else if (token == Opt_max_dir_size_kb) {
1736 sbi->s_max_dir_size_kb = arg;
1737 } else if (token == Opt_stripe) {
1738 sbi->s_stripe = arg;
1739 } else if (token == Opt_resuid) {
1740 uid = make_kuid(current_user_ns(), arg);
1741 if (!uid_valid(uid)) {
1742 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1745 sbi->s_resuid = uid;
1746 } else if (token == Opt_resgid) {
1747 gid = make_kgid(current_user_ns(), arg);
1748 if (!gid_valid(gid)) {
1749 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1752 sbi->s_resgid = gid;
1753 } else if (token == Opt_journal_dev) {
1755 ext4_msg(sb, KERN_ERR,
1756 "Cannot specify journal on remount");
1759 *journal_devnum = arg;
1760 } else if (token == Opt_journal_path) {
1762 struct inode *journal_inode;
1767 ext4_msg(sb, KERN_ERR,
1768 "Cannot specify journal on remount");
1771 journal_path = match_strdup(&args[0]);
1772 if (!journal_path) {
1773 ext4_msg(sb, KERN_ERR, "error: could not dup "
1774 "journal device string");
1778 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1780 ext4_msg(sb, KERN_ERR, "error: could not find "
1781 "journal device path: error %d", error);
1782 kfree(journal_path);
1786 journal_inode = d_inode(path.dentry);
1787 if (!S_ISBLK(journal_inode->i_mode)) {
1788 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1789 "is not a block device", journal_path);
1791 kfree(journal_path);
1795 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1797 kfree(journal_path);
1798 } else if (token == Opt_journal_ioprio) {
1800 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1805 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1806 } else if (token == Opt_test_dummy_encryption) {
1807 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1808 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1809 ext4_msg(sb, KERN_WARNING,
1810 "Test dummy encryption mode enabled");
1812 ext4_msg(sb, KERN_WARNING,
1813 "Test dummy encryption mount option ignored");
1815 } else if (m->flags & MOPT_DATAJ) {
1817 if (!sbi->s_journal)
1818 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1819 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1820 ext4_msg(sb, KERN_ERR,
1821 "Cannot change data mode on remount");
1825 clear_opt(sb, DATA_FLAGS);
1826 sbi->s_mount_opt |= m->mount_opt;
1829 } else if (m->flags & MOPT_QFMT) {
1830 if (sb_any_quota_loaded(sb) &&
1831 sbi->s_jquota_fmt != m->mount_opt) {
1832 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1833 "quota options when quota turned on");
1836 if (ext4_has_feature_quota(sb)) {
1837 ext4_msg(sb, KERN_INFO,
1838 "Quota format mount options ignored "
1839 "when QUOTA feature is enabled");
1842 sbi->s_jquota_fmt = m->mount_opt;
1844 } else if (token == Opt_dax) {
1845 #ifdef CONFIG_FS_DAX
1846 ext4_msg(sb, KERN_WARNING,
1847 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
1848 sbi->s_mount_opt |= m->mount_opt;
1850 ext4_msg(sb, KERN_INFO, "dax option not supported");
1853 } else if (token == Opt_data_err_abort) {
1854 sbi->s_mount_opt |= m->mount_opt;
1855 } else if (token == Opt_data_err_ignore) {
1856 sbi->s_mount_opt &= ~m->mount_opt;
1860 if (m->flags & MOPT_CLEAR)
1862 else if (unlikely(!(m->flags & MOPT_SET))) {
1863 ext4_msg(sb, KERN_WARNING,
1864 "buggy handling of option %s", opt);
1869 sbi->s_mount_opt |= m->mount_opt;
1871 sbi->s_mount_opt &= ~m->mount_opt;
1876 static int parse_options(char *options, struct super_block *sb,
1877 unsigned long *journal_devnum,
1878 unsigned int *journal_ioprio,
1881 struct ext4_sb_info *sbi = EXT4_SB(sb);
1883 substring_t args[MAX_OPT_ARGS];
1889 while ((p = strsep(&options, ",")) != NULL) {
1893 * Initialize args struct so we know whether arg was
1894 * found; some options take optional arguments.
1896 args[0].to = args[0].from = NULL;
1897 token = match_token(p, tokens, args);
1898 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1899 journal_ioprio, is_remount) < 0)
1904 * We do the test below only for project quotas. 'usrquota' and
1905 * 'grpquota' mount options are allowed even without quota feature
1906 * to support legacy quotas in quota files.
1908 if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) {
1909 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. "
1910 "Cannot enable project quota enforcement.");
1913 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1914 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1915 clear_opt(sb, USRQUOTA);
1917 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1918 clear_opt(sb, GRPQUOTA);
1920 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1921 ext4_msg(sb, KERN_ERR, "old and new quota "
1926 if (!sbi->s_jquota_fmt) {
1927 ext4_msg(sb, KERN_ERR, "journaled quota format "
1933 if (test_opt(sb, DIOREAD_NOLOCK)) {
1935 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1937 if (blocksize < PAGE_SIZE) {
1938 ext4_msg(sb, KERN_ERR, "can't mount with "
1939 "dioread_nolock if block size != PAGE_SIZE");
1946 static inline void ext4_show_quota_options(struct seq_file *seq,
1947 struct super_block *sb)
1949 #if defined(CONFIG_QUOTA)
1950 struct ext4_sb_info *sbi = EXT4_SB(sb);
1952 if (sbi->s_jquota_fmt) {
1955 switch (sbi->s_jquota_fmt) {
1966 seq_printf(seq, ",jqfmt=%s", fmtname);
1969 if (sbi->s_qf_names[USRQUOTA])
1970 seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]);
1972 if (sbi->s_qf_names[GRPQUOTA])
1973 seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]);
1977 static const char *token2str(int token)
1979 const struct match_token *t;
1981 for (t = tokens; t->token != Opt_err; t++)
1982 if (t->token == token && !strchr(t->pattern, '='))
1989 * - it's set to a non-default value OR
1990 * - if the per-sb default is different from the global default
1992 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1995 struct ext4_sb_info *sbi = EXT4_SB(sb);
1996 struct ext4_super_block *es = sbi->s_es;
1997 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1998 const struct mount_opts *m;
1999 char sep = nodefs ? '\n' : ',';
2001 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2002 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2004 if (sbi->s_sb_block != 1)
2005 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
2007 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
2008 int want_set = m->flags & MOPT_SET;
2009 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
2010 (m->flags & MOPT_CLEAR_ERR))
2012 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
2013 continue; /* skip if same as the default */
2015 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
2016 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
2017 continue; /* select Opt_noFoo vs Opt_Foo */
2018 SEQ_OPTS_PRINT("%s", token2str(m->token));
2021 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
2022 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
2023 SEQ_OPTS_PRINT("resuid=%u",
2024 from_kuid_munged(&init_user_ns, sbi->s_resuid));
2025 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
2026 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
2027 SEQ_OPTS_PRINT("resgid=%u",
2028 from_kgid_munged(&init_user_ns, sbi->s_resgid));
2029 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
2030 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
2031 SEQ_OPTS_PUTS("errors=remount-ro");
2032 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
2033 SEQ_OPTS_PUTS("errors=continue");
2034 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
2035 SEQ_OPTS_PUTS("errors=panic");
2036 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
2037 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
2038 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
2039 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
2040 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
2041 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2042 if (sb->s_flags & MS_I_VERSION)
2043 SEQ_OPTS_PUTS("i_version");
2044 if (nodefs || sbi->s_stripe)
2045 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
2046 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
2047 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2048 SEQ_OPTS_PUTS("data=journal");
2049 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2050 SEQ_OPTS_PUTS("data=ordered");
2051 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2052 SEQ_OPTS_PUTS("data=writeback");
2055 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
2056 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2057 sbi->s_inode_readahead_blks);
2059 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
2060 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
2061 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
2062 if (nodefs || sbi->s_max_dir_size_kb)
2063 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2064 if (test_opt(sb, DATA_ERR_ABORT))
2065 SEQ_OPTS_PUTS("data_err=abort");
2067 ext4_show_quota_options(seq, sb);
2071 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2073 return _ext4_show_options(seq, root->d_sb, 0);
2076 int ext4_seq_options_show(struct seq_file *seq, void *offset)
2078 struct super_block *sb = seq->private;
2081 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
2082 rc = _ext4_show_options(seq, sb, 1);
2083 seq_puts(seq, "\n");
2087 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
2090 struct ext4_sb_info *sbi = EXT4_SB(sb);
2093 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
2094 ext4_msg(sb, KERN_ERR, "revision level too high, "
2095 "forcing read-only mode");
2100 if (!(sbi->s_mount_state & EXT4_VALID_FS))
2101 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
2102 "running e2fsck is recommended");
2103 else if (sbi->s_mount_state & EXT4_ERROR_FS)
2104 ext4_msg(sb, KERN_WARNING,
2105 "warning: mounting fs with errors, "
2106 "running e2fsck is recommended");
2107 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
2108 le16_to_cpu(es->s_mnt_count) >=
2109 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
2110 ext4_msg(sb, KERN_WARNING,
2111 "warning: maximal mount count reached, "
2112 "running e2fsck is recommended");
2113 else if (le32_to_cpu(es->s_checkinterval) &&
2114 (le32_to_cpu(es->s_lastcheck) +
2115 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
2116 ext4_msg(sb, KERN_WARNING,
2117 "warning: checktime reached, "
2118 "running e2fsck is recommended");
2119 if (!sbi->s_journal)
2120 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
2121 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
2122 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
2123 le16_add_cpu(&es->s_mnt_count, 1);
2124 es->s_mtime = cpu_to_le32(get_seconds());
2125 ext4_update_dynamic_rev(sb);
2127 ext4_set_feature_journal_needs_recovery(sb);
2129 ext4_commit_super(sb, 1);
2131 if (test_opt(sb, DEBUG))
2132 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
2133 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
2135 sbi->s_groups_count,
2136 EXT4_BLOCKS_PER_GROUP(sb),
2137 EXT4_INODES_PER_GROUP(sb),
2138 sbi->s_mount_opt, sbi->s_mount_opt2);
2140 cleancache_init_fs(sb);
2144 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2146 struct ext4_sb_info *sbi = EXT4_SB(sb);
2147 struct flex_groups *new_groups;
2150 if (!sbi->s_log_groups_per_flex)
2153 size = ext4_flex_group(sbi, ngroup - 1) + 1;
2154 if (size <= sbi->s_flex_groups_allocated)
2157 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
2158 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
2160 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
2161 size / (int) sizeof(struct flex_groups));
2165 if (sbi->s_flex_groups) {
2166 memcpy(new_groups, sbi->s_flex_groups,
2167 (sbi->s_flex_groups_allocated *
2168 sizeof(struct flex_groups)));
2169 kvfree(sbi->s_flex_groups);
2171 sbi->s_flex_groups = new_groups;
2172 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
2176 static int ext4_fill_flex_info(struct super_block *sb)
2178 struct ext4_sb_info *sbi = EXT4_SB(sb);
2179 struct ext4_group_desc *gdp = NULL;
2180 ext4_group_t flex_group;
2183 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2184 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
2185 sbi->s_log_groups_per_flex = 0;
2189 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2193 for (i = 0; i < sbi->s_groups_count; i++) {
2194 gdp = ext4_get_group_desc(sb, i, NULL);
2196 flex_group = ext4_flex_group(sbi, i);
2197 atomic_add(ext4_free_inodes_count(sb, gdp),
2198 &sbi->s_flex_groups[flex_group].free_inodes);
2199 atomic64_add(ext4_free_group_clusters(sb, gdp),
2200 &sbi->s_flex_groups[flex_group].free_clusters);
2201 atomic_add(ext4_used_dirs_count(sb, gdp),
2202 &sbi->s_flex_groups[flex_group].used_dirs);
2210 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
2211 struct ext4_group_desc *gdp)
2213 int offset = offsetof(struct ext4_group_desc, bg_checksum);
2215 __le32 le_group = cpu_to_le32(block_group);
2216 struct ext4_sb_info *sbi = EXT4_SB(sb);
2218 if (ext4_has_metadata_csum(sbi->s_sb)) {
2219 /* Use new metadata_csum algorithm */
2221 __u16 dummy_csum = 0;
2223 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2225 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2226 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2227 sizeof(dummy_csum));
2228 offset += sizeof(dummy_csum);
2229 if (offset < sbi->s_desc_size)
2230 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2231 sbi->s_desc_size - offset);
2233 crc = csum32 & 0xFFFF;
2237 /* old crc16 code */
2238 if (!ext4_has_feature_gdt_csum(sb))
2241 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2242 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2243 crc = crc16(crc, (__u8 *)gdp, offset);
2244 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2245 /* for checksum of struct ext4_group_desc do the rest...*/
2246 if (ext4_has_feature_64bit(sb) &&
2247 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2248 crc = crc16(crc, (__u8 *)gdp + offset,
2249 le16_to_cpu(sbi->s_es->s_desc_size) -
2253 return cpu_to_le16(crc);
2256 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2257 struct ext4_group_desc *gdp)
2259 if (ext4_has_group_desc_csum(sb) &&
2260 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
2266 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2267 struct ext4_group_desc *gdp)
2269 if (!ext4_has_group_desc_csum(sb))
2271 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
2274 /* Called at mount-time, super-block is locked */
2275 static int ext4_check_descriptors(struct super_block *sb,
2276 ext4_fsblk_t sb_block,
2277 ext4_group_t *first_not_zeroed)
2279 struct ext4_sb_info *sbi = EXT4_SB(sb);
2280 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2281 ext4_fsblk_t last_block;
2282 ext4_fsblk_t block_bitmap;
2283 ext4_fsblk_t inode_bitmap;
2284 ext4_fsblk_t inode_table;
2285 int flexbg_flag = 0;
2286 ext4_group_t i, grp = sbi->s_groups_count;
2288 if (ext4_has_feature_flex_bg(sb))
2291 ext4_debug("Checking group descriptors");
2293 for (i = 0; i < sbi->s_groups_count; i++) {
2294 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2296 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2297 last_block = ext4_blocks_count(sbi->s_es) - 1;
2299 last_block = first_block +
2300 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2302 if ((grp == sbi->s_groups_count) &&
2303 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2306 block_bitmap = ext4_block_bitmap(sb, gdp);
2307 if (block_bitmap == sb_block) {
2308 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2309 "Block bitmap for group %u overlaps "
2312 if (block_bitmap < first_block || block_bitmap > last_block) {
2313 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2314 "Block bitmap for group %u not in group "
2315 "(block %llu)!", i, block_bitmap);
2318 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2319 if (inode_bitmap == sb_block) {
2320 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2321 "Inode bitmap for group %u overlaps "
2324 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2325 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2326 "Inode bitmap for group %u not in group "
2327 "(block %llu)!", i, inode_bitmap);
2330 inode_table = ext4_inode_table(sb, gdp);
2331 if (inode_table == sb_block) {
2332 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2333 "Inode table for group %u overlaps "
2336 if (inode_table < first_block ||
2337 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2338 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2339 "Inode table for group %u not in group "
2340 "(block %llu)!", i, inode_table);
2343 ext4_lock_group(sb, i);
2344 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2345 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2346 "Checksum for group %u failed (%u!=%u)",
2347 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
2348 gdp)), le16_to_cpu(gdp->bg_checksum));
2349 if (!(sb->s_flags & MS_RDONLY)) {
2350 ext4_unlock_group(sb, i);
2354 ext4_unlock_group(sb, i);
2356 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2358 if (NULL != first_not_zeroed)
2359 *first_not_zeroed = grp;
2363 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2364 * the superblock) which were deleted from all directories, but held open by
2365 * a process at the time of a crash. We walk the list and try to delete these
2366 * inodes at recovery time (only with a read-write filesystem).
2368 * In order to keep the orphan inode chain consistent during traversal (in
2369 * case of crash during recovery), we link each inode into the superblock
2370 * orphan list_head and handle it the same way as an inode deletion during
2371 * normal operation (which journals the operations for us).
2373 * We only do an iget() and an iput() on each inode, which is very safe if we
2374 * accidentally point at an in-use or already deleted inode. The worst that
2375 * can happen in this case is that we get a "bit already cleared" message from
2376 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2377 * e2fsck was run on this filesystem, and it must have already done the orphan
2378 * inode cleanup for us, so we can safely abort without any further action.
2380 static void ext4_orphan_cleanup(struct super_block *sb,
2381 struct ext4_super_block *es)
2383 unsigned int s_flags = sb->s_flags;
2384 int ret, nr_orphans = 0, nr_truncates = 0;
2388 if (!es->s_last_orphan) {
2389 jbd_debug(4, "no orphan inodes to clean up\n");
2393 if (bdev_read_only(sb->s_bdev)) {
2394 ext4_msg(sb, KERN_ERR, "write access "
2395 "unavailable, skipping orphan cleanup");
2399 /* Check if feature set would not allow a r/w mount */
2400 if (!ext4_feature_set_ok(sb, 0)) {
2401 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2402 "unknown ROCOMPAT features");
2406 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2407 /* don't clear list on RO mount w/ errors */
2408 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
2409 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
2410 "clearing orphan list.\n");
2411 es->s_last_orphan = 0;
2413 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2417 if (s_flags & MS_RDONLY) {
2418 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2419 sb->s_flags &= ~MS_RDONLY;
2422 /* Needed for iput() to work correctly and not trash data */
2423 sb->s_flags |= MS_ACTIVE;
2424 /* Turn on quotas so that they are updated correctly */
2425 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2426 if (EXT4_SB(sb)->s_qf_names[i]) {
2427 int ret = ext4_quota_on_mount(sb, i);
2429 ext4_msg(sb, KERN_ERR,
2430 "Cannot turn on journaled "
2431 "quota: error %d", ret);
2436 while (es->s_last_orphan) {
2437 struct inode *inode;
2440 * We may have encountered an error during cleanup; if
2441 * so, skip the rest.
2443 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2444 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2445 es->s_last_orphan = 0;
2449 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2450 if (IS_ERR(inode)) {
2451 es->s_last_orphan = 0;
2455 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2456 dquot_initialize(inode);
2457 if (inode->i_nlink) {
2458 if (test_opt(sb, DEBUG))
2459 ext4_msg(sb, KERN_DEBUG,
2460 "%s: truncating inode %lu to %lld bytes",
2461 __func__, inode->i_ino, inode->i_size);
2462 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2463 inode->i_ino, inode->i_size);
2465 truncate_inode_pages(inode->i_mapping, inode->i_size);
2466 ret = ext4_truncate(inode);
2468 ext4_std_error(inode->i_sb, ret);
2469 inode_unlock(inode);
2472 if (test_opt(sb, DEBUG))
2473 ext4_msg(sb, KERN_DEBUG,
2474 "%s: deleting unreferenced inode %lu",
2475 __func__, inode->i_ino);
2476 jbd_debug(2, "deleting unreferenced inode %lu\n",
2480 iput(inode); /* The delete magic happens here! */
2483 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2486 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2487 PLURAL(nr_orphans));
2489 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2490 PLURAL(nr_truncates));
2492 /* Turn quotas off */
2493 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2494 if (sb_dqopt(sb)->files[i])
2495 dquot_quota_off(sb, i);
2498 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2502 * Maximal extent format file size.
2503 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2504 * extent format containers, within a sector_t, and within i_blocks
2505 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2506 * so that won't be a limiting factor.
2508 * However there is other limiting factor. We do store extents in the form
2509 * of starting block and length, hence the resulting length of the extent
2510 * covering maximum file size must fit into on-disk format containers as
2511 * well. Given that length is always by 1 unit bigger than max unit (because
2512 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2514 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2516 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2519 loff_t upper_limit = MAX_LFS_FILESIZE;
2521 /* small i_blocks in vfs inode? */
2522 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2524 * CONFIG_LBDAF is not enabled implies the inode
2525 * i_block represent total blocks in 512 bytes
2526 * 32 == size of vfs inode i_blocks * 8
2528 upper_limit = (1LL << 32) - 1;
2530 /* total blocks in file system block size */
2531 upper_limit >>= (blkbits - 9);
2532 upper_limit <<= blkbits;
2536 * 32-bit extent-start container, ee_block. We lower the maxbytes
2537 * by one fs block, so ee_len can cover the extent of maximum file
2540 res = (1LL << 32) - 1;
2543 /* Sanity check against vm- & vfs- imposed limits */
2544 if (res > upper_limit)
2551 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2552 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2553 * We need to be 1 filesystem block less than the 2^48 sector limit.
2555 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2557 loff_t res = EXT4_NDIR_BLOCKS;
2560 /* This is calculated to be the largest file size for a dense, block
2561 * mapped file such that the file's total number of 512-byte sectors,
2562 * including data and all indirect blocks, does not exceed (2^48 - 1).
2564 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2565 * number of 512-byte sectors of the file.
2568 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2570 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2571 * the inode i_block field represents total file blocks in
2572 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2574 upper_limit = (1LL << 32) - 1;
2576 /* total blocks in file system block size */
2577 upper_limit >>= (bits - 9);
2581 * We use 48 bit ext4_inode i_blocks
2582 * With EXT4_HUGE_FILE_FL set the i_blocks
2583 * represent total number of blocks in
2584 * file system block size
2586 upper_limit = (1LL << 48) - 1;
2590 /* indirect blocks */
2592 /* double indirect blocks */
2593 meta_blocks += 1 + (1LL << (bits-2));
2594 /* tripple indirect blocks */
2595 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2597 upper_limit -= meta_blocks;
2598 upper_limit <<= bits;
2600 res += 1LL << (bits-2);
2601 res += 1LL << (2*(bits-2));
2602 res += 1LL << (3*(bits-2));
2604 if (res > upper_limit)
2607 if (res > MAX_LFS_FILESIZE)
2608 res = MAX_LFS_FILESIZE;
2613 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2614 ext4_fsblk_t logical_sb_block, int nr)
2616 struct ext4_sb_info *sbi = EXT4_SB(sb);
2617 ext4_group_t bg, first_meta_bg;
2620 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2622 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
2623 return logical_sb_block + nr + 1;
2624 bg = sbi->s_desc_per_block * nr;
2625 if (ext4_bg_has_super(sb, bg))
2629 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2630 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2631 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2634 if (sb->s_blocksize == 1024 && nr == 0 &&
2635 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2638 return (has_super + ext4_group_first_block_no(sb, bg));
2642 * ext4_get_stripe_size: Get the stripe size.
2643 * @sbi: In memory super block info
2645 * If we have specified it via mount option, then
2646 * use the mount option value. If the value specified at mount time is
2647 * greater than the blocks per group use the super block value.
2648 * If the super block value is greater than blocks per group return 0.
2649 * Allocator needs it be less than blocks per group.
2652 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2654 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2655 unsigned long stripe_width =
2656 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2659 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2660 ret = sbi->s_stripe;
2661 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
2663 else if (stride && stride <= sbi->s_blocks_per_group)
2669 * If the stripe width is 1, this makes no sense and
2670 * we set it to 0 to turn off stripe handling code.
2679 * Check whether this filesystem can be mounted based on
2680 * the features present and the RDONLY/RDWR mount requested.
2681 * Returns 1 if this filesystem can be mounted as requested,
2682 * 0 if it cannot be.
2684 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2686 if (ext4_has_unknown_ext4_incompat_features(sb)) {
2687 ext4_msg(sb, KERN_ERR,
2688 "Couldn't mount because of "
2689 "unsupported optional features (%x)",
2690 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2691 ~EXT4_FEATURE_INCOMPAT_SUPP));
2698 if (ext4_has_feature_readonly(sb)) {
2699 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2700 sb->s_flags |= MS_RDONLY;
2704 /* Check that feature set is OK for a read-write mount */
2705 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
2706 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2707 "unsupported optional features (%x)",
2708 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2709 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2713 * Large file size enabled file system can only be mounted
2714 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2716 if (ext4_has_feature_huge_file(sb)) {
2717 if (sizeof(blkcnt_t) < sizeof(u64)) {
2718 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2719 "cannot be mounted RDWR without "
2724 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
2725 ext4_msg(sb, KERN_ERR,
2726 "Can't support bigalloc feature without "
2727 "extents feature\n");
2731 #ifndef CONFIG_QUOTA
2732 if (ext4_has_feature_quota(sb) && !readonly) {
2733 ext4_msg(sb, KERN_ERR,
2734 "Filesystem with quota feature cannot be mounted RDWR "
2735 "without CONFIG_QUOTA");
2738 if (ext4_has_feature_project(sb) && !readonly) {
2739 ext4_msg(sb, KERN_ERR,
2740 "Filesystem with project quota feature cannot be mounted RDWR "
2741 "without CONFIG_QUOTA");
2744 #endif /* CONFIG_QUOTA */
2749 * This function is called once a day if we have errors logged
2750 * on the file system
2752 static void print_daily_error_info(unsigned long arg)
2754 struct super_block *sb = (struct super_block *) arg;
2755 struct ext4_sb_info *sbi;
2756 struct ext4_super_block *es;
2761 if (es->s_error_count)
2762 /* fsck newer than v1.41.13 is needed to clean this condition. */
2763 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
2764 le32_to_cpu(es->s_error_count));
2765 if (es->s_first_error_time) {
2766 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
2767 sb->s_id, le32_to_cpu(es->s_first_error_time),
2768 (int) sizeof(es->s_first_error_func),
2769 es->s_first_error_func,
2770 le32_to_cpu(es->s_first_error_line));
2771 if (es->s_first_error_ino)
2772 printk(KERN_CONT ": inode %u",
2773 le32_to_cpu(es->s_first_error_ino));
2774 if (es->s_first_error_block)
2775 printk(KERN_CONT ": block %llu", (unsigned long long)
2776 le64_to_cpu(es->s_first_error_block));
2777 printk(KERN_CONT "\n");
2779 if (es->s_last_error_time) {
2780 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
2781 sb->s_id, le32_to_cpu(es->s_last_error_time),
2782 (int) sizeof(es->s_last_error_func),
2783 es->s_last_error_func,
2784 le32_to_cpu(es->s_last_error_line));
2785 if (es->s_last_error_ino)
2786 printk(KERN_CONT ": inode %u",
2787 le32_to_cpu(es->s_last_error_ino));
2788 if (es->s_last_error_block)
2789 printk(KERN_CONT ": block %llu", (unsigned long long)
2790 le64_to_cpu(es->s_last_error_block));
2791 printk(KERN_CONT "\n");
2793 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2796 /* Find next suitable group and run ext4_init_inode_table */
2797 static int ext4_run_li_request(struct ext4_li_request *elr)
2799 struct ext4_group_desc *gdp = NULL;
2800 ext4_group_t group, ngroups;
2801 struct super_block *sb;
2802 unsigned long timeout = 0;
2806 ngroups = EXT4_SB(sb)->s_groups_count;
2808 for (group = elr->lr_next_group; group < ngroups; group++) {
2809 gdp = ext4_get_group_desc(sb, group, NULL);
2815 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2819 if (group >= ngroups)
2824 ret = ext4_init_inode_table(sb, group,
2825 elr->lr_timeout ? 0 : 1);
2826 if (elr->lr_timeout == 0) {
2827 timeout = (jiffies - timeout) *
2828 elr->lr_sbi->s_li_wait_mult;
2829 elr->lr_timeout = timeout;
2831 elr->lr_next_sched = jiffies + elr->lr_timeout;
2832 elr->lr_next_group = group + 1;
2838 * Remove lr_request from the list_request and free the
2839 * request structure. Should be called with li_list_mtx held
2841 static void ext4_remove_li_request(struct ext4_li_request *elr)
2843 struct ext4_sb_info *sbi;
2850 list_del(&elr->lr_request);
2851 sbi->s_li_request = NULL;
2855 static void ext4_unregister_li_request(struct super_block *sb)
2857 mutex_lock(&ext4_li_mtx);
2858 if (!ext4_li_info) {
2859 mutex_unlock(&ext4_li_mtx);
2863 mutex_lock(&ext4_li_info->li_list_mtx);
2864 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2865 mutex_unlock(&ext4_li_info->li_list_mtx);
2866 mutex_unlock(&ext4_li_mtx);
2869 static struct task_struct *ext4_lazyinit_task;
2872 * This is the function where ext4lazyinit thread lives. It walks
2873 * through the request list searching for next scheduled filesystem.
2874 * When such a fs is found, run the lazy initialization request
2875 * (ext4_rn_li_request) and keep track of the time spend in this
2876 * function. Based on that time we compute next schedule time of
2877 * the request. When walking through the list is complete, compute
2878 * next waking time and put itself into sleep.
2880 static int ext4_lazyinit_thread(void *arg)
2882 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2883 struct list_head *pos, *n;
2884 struct ext4_li_request *elr;
2885 unsigned long next_wakeup, cur;
2887 BUG_ON(NULL == eli);
2891 next_wakeup = MAX_JIFFY_OFFSET;
2893 mutex_lock(&eli->li_list_mtx);
2894 if (list_empty(&eli->li_request_list)) {
2895 mutex_unlock(&eli->li_list_mtx);
2898 list_for_each_safe(pos, n, &eli->li_request_list) {
2901 elr = list_entry(pos, struct ext4_li_request,
2904 if (time_before(jiffies, elr->lr_next_sched)) {
2905 if (time_before(elr->lr_next_sched, next_wakeup))
2906 next_wakeup = elr->lr_next_sched;
2909 if (down_read_trylock(&elr->lr_super->s_umount)) {
2910 if (sb_start_write_trylock(elr->lr_super)) {
2913 * We hold sb->s_umount, sb can not
2914 * be removed from the list, it is
2915 * now safe to drop li_list_mtx
2917 mutex_unlock(&eli->li_list_mtx);
2918 err = ext4_run_li_request(elr);
2919 sb_end_write(elr->lr_super);
2920 mutex_lock(&eli->li_list_mtx);
2923 up_read((&elr->lr_super->s_umount));
2925 /* error, remove the lazy_init job */
2927 ext4_remove_li_request(elr);
2931 elr->lr_next_sched = jiffies +
2933 % (EXT4_DEF_LI_MAX_START_DELAY * HZ));
2935 if (time_before(elr->lr_next_sched, next_wakeup))
2936 next_wakeup = elr->lr_next_sched;
2938 mutex_unlock(&eli->li_list_mtx);
2943 if ((time_after_eq(cur, next_wakeup)) ||
2944 (MAX_JIFFY_OFFSET == next_wakeup)) {
2949 schedule_timeout_interruptible(next_wakeup - cur);
2951 if (kthread_should_stop()) {
2952 ext4_clear_request_list();
2959 * It looks like the request list is empty, but we need
2960 * to check it under the li_list_mtx lock, to prevent any
2961 * additions into it, and of course we should lock ext4_li_mtx
2962 * to atomically free the list and ext4_li_info, because at
2963 * this point another ext4 filesystem could be registering
2966 mutex_lock(&ext4_li_mtx);
2967 mutex_lock(&eli->li_list_mtx);
2968 if (!list_empty(&eli->li_request_list)) {
2969 mutex_unlock(&eli->li_list_mtx);
2970 mutex_unlock(&ext4_li_mtx);
2973 mutex_unlock(&eli->li_list_mtx);
2974 kfree(ext4_li_info);
2975 ext4_li_info = NULL;
2976 mutex_unlock(&ext4_li_mtx);
2981 static void ext4_clear_request_list(void)
2983 struct list_head *pos, *n;
2984 struct ext4_li_request *elr;
2986 mutex_lock(&ext4_li_info->li_list_mtx);
2987 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2988 elr = list_entry(pos, struct ext4_li_request,
2990 ext4_remove_li_request(elr);
2992 mutex_unlock(&ext4_li_info->li_list_mtx);
2995 static int ext4_run_lazyinit_thread(void)
2997 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2998 ext4_li_info, "ext4lazyinit");
2999 if (IS_ERR(ext4_lazyinit_task)) {
3000 int err = PTR_ERR(ext4_lazyinit_task);
3001 ext4_clear_request_list();
3002 kfree(ext4_li_info);
3003 ext4_li_info = NULL;
3004 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3005 "initialization thread\n",
3009 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3014 * Check whether it make sense to run itable init. thread or not.
3015 * If there is at least one uninitialized inode table, return
3016 * corresponding group number, else the loop goes through all
3017 * groups and return total number of groups.
3019 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3021 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3022 struct ext4_group_desc *gdp = NULL;
3024 for (group = 0; group < ngroups; group++) {
3025 gdp = ext4_get_group_desc(sb, group, NULL);
3029 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3036 static int ext4_li_info_new(void)
3038 struct ext4_lazy_init *eli = NULL;
3040 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3044 INIT_LIST_HEAD(&eli->li_request_list);
3045 mutex_init(&eli->li_list_mtx);
3047 eli->li_state |= EXT4_LAZYINIT_QUIT;
3054 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3057 struct ext4_sb_info *sbi = EXT4_SB(sb);
3058 struct ext4_li_request *elr;
3060 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3066 elr->lr_next_group = start;
3069 * Randomize first schedule time of the request to
3070 * spread the inode table initialization requests
3073 elr->lr_next_sched = jiffies + (prandom_u32() %
3074 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3078 int ext4_register_li_request(struct super_block *sb,
3079 ext4_group_t first_not_zeroed)
3081 struct ext4_sb_info *sbi = EXT4_SB(sb);
3082 struct ext4_li_request *elr = NULL;
3083 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3086 mutex_lock(&ext4_li_mtx);
3087 if (sbi->s_li_request != NULL) {
3089 * Reset timeout so it can be computed again, because
3090 * s_li_wait_mult might have changed.
3092 sbi->s_li_request->lr_timeout = 0;
3096 if (first_not_zeroed == ngroups ||
3097 (sb->s_flags & MS_RDONLY) ||
3098 !test_opt(sb, INIT_INODE_TABLE))
3101 elr = ext4_li_request_new(sb, first_not_zeroed);
3107 if (NULL == ext4_li_info) {
3108 ret = ext4_li_info_new();
3113 mutex_lock(&ext4_li_info->li_list_mtx);
3114 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3115 mutex_unlock(&ext4_li_info->li_list_mtx);
3117 sbi->s_li_request = elr;
3119 * set elr to NULL here since it has been inserted to
3120 * the request_list and the removal and free of it is
3121 * handled by ext4_clear_request_list from now on.
3125 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3126 ret = ext4_run_lazyinit_thread();
3131 mutex_unlock(&ext4_li_mtx);
3138 * We do not need to lock anything since this is called on
3141 static void ext4_destroy_lazyinit_thread(void)
3144 * If thread exited earlier
3145 * there's nothing to be done.
3147 if (!ext4_li_info || !ext4_lazyinit_task)
3150 kthread_stop(ext4_lazyinit_task);
3153 static int set_journal_csum_feature_set(struct super_block *sb)
3156 int compat, incompat;
3157 struct ext4_sb_info *sbi = EXT4_SB(sb);
3159 if (ext4_has_metadata_csum(sb)) {
3160 /* journal checksum v3 */
3162 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
3164 /* journal checksum v1 */
3165 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3169 jbd2_journal_clear_features(sbi->s_journal,
3170 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3171 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3172 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3173 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3174 ret = jbd2_journal_set_features(sbi->s_journal,
3176 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3178 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3179 ret = jbd2_journal_set_features(sbi->s_journal,
3182 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3183 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3185 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3186 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3193 * Note: calculating the overhead so we can be compatible with
3194 * historical BSD practice is quite difficult in the face of
3195 * clusters/bigalloc. This is because multiple metadata blocks from
3196 * different block group can end up in the same allocation cluster.
3197 * Calculating the exact overhead in the face of clustered allocation
3198 * requires either O(all block bitmaps) in memory or O(number of block
3199 * groups**2) in time. We will still calculate the superblock for
3200 * older file systems --- and if we come across with a bigalloc file
3201 * system with zero in s_overhead_clusters the estimate will be close to
3202 * correct especially for very large cluster sizes --- but for newer
3203 * file systems, it's better to calculate this figure once at mkfs
3204 * time, and store it in the superblock. If the superblock value is
3205 * present (even for non-bigalloc file systems), we will use it.
3207 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3210 struct ext4_sb_info *sbi = EXT4_SB(sb);
3211 struct ext4_group_desc *gdp;
3212 ext4_fsblk_t first_block, last_block, b;
3213 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3214 int s, j, count = 0;
3216 if (!ext4_has_feature_bigalloc(sb))
3217 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3218 sbi->s_itb_per_group + 2);
3220 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3221 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3222 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3223 for (i = 0; i < ngroups; i++) {
3224 gdp = ext4_get_group_desc(sb, i, NULL);
3225 b = ext4_block_bitmap(sb, gdp);
3226 if (b >= first_block && b <= last_block) {
3227 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3230 b = ext4_inode_bitmap(sb, gdp);
3231 if (b >= first_block && b <= last_block) {
3232 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3235 b = ext4_inode_table(sb, gdp);
3236 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3237 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3238 int c = EXT4_B2C(sbi, b - first_block);
3239 ext4_set_bit(c, buf);
3245 if (ext4_bg_has_super(sb, grp)) {
3246 ext4_set_bit(s++, buf);
3249 j = ext4_bg_num_gdb(sb, grp);
3250 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
3251 ext4_error(sb, "Invalid number of block group "
3252 "descriptor blocks: %d", j);
3253 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
3257 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3261 return EXT4_CLUSTERS_PER_GROUP(sb) -
3262 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3266 * Compute the overhead and stash it in sbi->s_overhead
3268 int ext4_calculate_overhead(struct super_block *sb)
3270 struct ext4_sb_info *sbi = EXT4_SB(sb);
3271 struct ext4_super_block *es = sbi->s_es;
3272 struct inode *j_inode;
3273 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
3274 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3275 ext4_fsblk_t overhead = 0;
3276 char *buf = (char *) get_zeroed_page(GFP_NOFS);
3282 * Compute the overhead (FS structures). This is constant
3283 * for a given filesystem unless the number of block groups
3284 * changes so we cache the previous value until it does.
3288 * All of the blocks before first_data_block are overhead
3290 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3293 * Add the overhead found in each block group
3295 for (i = 0; i < ngroups; i++) {
3298 blks = count_overhead(sb, i, buf);
3301 memset(buf, 0, PAGE_SIZE);
3306 * Add the internal journal blocks whether the journal has been
3309 if (sbi->s_journal && !sbi->journal_bdev)
3310 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3311 else if (ext4_has_feature_journal(sb) && !sbi->s_journal) {
3312 j_inode = ext4_get_journal_inode(sb, j_inum);
3314 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
3315 overhead += EXT4_NUM_B2C(sbi, j_blocks);
3318 ext4_msg(sb, KERN_ERR, "can't get journal size");
3321 sbi->s_overhead = overhead;
3323 free_page((unsigned long) buf);
3327 static void ext4_set_resv_clusters(struct super_block *sb)
3329 ext4_fsblk_t resv_clusters;
3330 struct ext4_sb_info *sbi = EXT4_SB(sb);
3333 * There's no need to reserve anything when we aren't using extents.
3334 * The space estimates are exact, there are no unwritten extents,
3335 * hole punching doesn't need new metadata... This is needed especially
3336 * to keep ext2/3 backward compatibility.
3338 if (!ext4_has_feature_extents(sb))
3341 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3342 * This should cover the situations where we can not afford to run
3343 * out of space like for example punch hole, or converting
3344 * unwritten extents in delalloc path. In most cases such
3345 * allocation would require 1, or 2 blocks, higher numbers are
3348 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3349 sbi->s_cluster_bits);
3351 do_div(resv_clusters, 50);
3352 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3354 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
3357 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3359 char *orig_data = kstrdup(data, GFP_KERNEL);
3360 struct buffer_head *bh;
3361 struct ext4_super_block *es = NULL;
3362 struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3364 ext4_fsblk_t sb_block = get_sb_block(&data);
3365 ext4_fsblk_t logical_sb_block;
3366 unsigned long offset = 0;
3367 unsigned long journal_devnum = 0;
3368 unsigned long def_mount_opts;
3372 int blocksize, clustersize;
3373 unsigned int db_count;
3375 int needs_recovery, has_huge_files, has_bigalloc;
3378 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3379 ext4_group_t first_not_zeroed;
3381 if ((data && !orig_data) || !sbi)
3384 sbi->s_blockgroup_lock =
3385 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3386 if (!sbi->s_blockgroup_lock)
3389 sb->s_fs_info = sbi;
3391 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3392 sbi->s_sb_block = sb_block;
3393 if (sb->s_bdev->bd_part)
3394 sbi->s_sectors_written_start =
3395 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3397 /* Cleanup superblock name */
3398 strreplace(sb->s_id, '/', '!');
3400 /* -EINVAL is default */
3402 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3404 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3409 * The ext4 superblock will not be buffer aligned for other than 1kB
3410 * block sizes. We need to calculate the offset from buffer start.
3412 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3413 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3414 offset = do_div(logical_sb_block, blocksize);
3416 logical_sb_block = sb_block;
3419 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
3420 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3424 * Note: s_es must be initialized as soon as possible because
3425 * some ext4 macro-instructions depend on its value
3427 es = (struct ext4_super_block *) (bh->b_data + offset);
3429 sb->s_magic = le16_to_cpu(es->s_magic);
3430 if (sb->s_magic != EXT4_SUPER_MAGIC)
3432 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3434 /* Warn if metadata_csum and gdt_csum are both set. */
3435 if (ext4_has_feature_metadata_csum(sb) &&
3436 ext4_has_feature_gdt_csum(sb))
3437 ext4_warning(sb, "metadata_csum and uninit_bg are "
3438 "redundant flags; please run fsck.");
3440 /* Check for a known checksum algorithm */
3441 if (!ext4_verify_csum_type(sb, es)) {
3442 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3443 "unknown checksum algorithm.");
3448 /* Load the checksum driver */
3449 if (ext4_has_feature_metadata_csum(sb)) {
3450 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3451 if (IS_ERR(sbi->s_chksum_driver)) {
3452 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3453 ret = PTR_ERR(sbi->s_chksum_driver);
3454 sbi->s_chksum_driver = NULL;
3459 /* Check superblock checksum */
3460 if (!ext4_superblock_csum_verify(sb, es)) {
3461 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3462 "invalid superblock checksum. Run e2fsck?");
3468 /* Precompute checksum seed for all metadata */
3469 if (ext4_has_feature_csum_seed(sb))
3470 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
3471 else if (ext4_has_metadata_csum(sb))
3472 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3473 sizeof(es->s_uuid));
3475 /* Set defaults before we parse the mount options */
3476 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3477 set_opt(sb, INIT_INODE_TABLE);
3478 if (def_mount_opts & EXT4_DEFM_DEBUG)
3480 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3482 if (def_mount_opts & EXT4_DEFM_UID16)
3483 set_opt(sb, NO_UID32);
3484 /* xattr user namespace & acls are now defaulted on */
3485 set_opt(sb, XATTR_USER);
3486 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3487 set_opt(sb, POSIX_ACL);
3489 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3490 if (ext4_has_metadata_csum(sb))
3491 set_opt(sb, JOURNAL_CHECKSUM);
3493 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3494 set_opt(sb, JOURNAL_DATA);
3495 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3496 set_opt(sb, ORDERED_DATA);
3497 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3498 set_opt(sb, WRITEBACK_DATA);
3500 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3501 set_opt(sb, ERRORS_PANIC);
3502 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3503 set_opt(sb, ERRORS_CONT);
3505 set_opt(sb, ERRORS_RO);
3506 /* block_validity enabled by default; disable with noblock_validity */
3507 set_opt(sb, BLOCK_VALIDITY);
3508 if (def_mount_opts & EXT4_DEFM_DISCARD)
3509 set_opt(sb, DISCARD);
3511 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3512 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3513 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3514 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3515 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3517 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3518 set_opt(sb, BARRIER);
3521 * enable delayed allocation by default
3522 * Use -o nodelalloc to turn it off
3524 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3525 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3526 set_opt(sb, DELALLOC);
3529 * set default s_li_wait_mult for lazyinit, for the case there is
3530 * no mount option specified.
3532 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3534 if (sbi->s_es->s_mount_opts[0]) {
3535 char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
3536 sizeof(sbi->s_es->s_mount_opts),
3540 if (!parse_options(s_mount_opts, sb, &journal_devnum,
3541 &journal_ioprio, 0)) {
3542 ext4_msg(sb, KERN_WARNING,
3543 "failed to parse options in superblock: %s",
3546 kfree(s_mount_opts);
3548 sbi->s_def_mount_opt = sbi->s_mount_opt;
3549 if (!parse_options((char *) data, sb, &journal_devnum,
3550 &journal_ioprio, 0))
3553 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3554 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3555 "with data=journal disables delayed "
3556 "allocation and O_DIRECT support!\n");
3557 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3558 ext4_msg(sb, KERN_ERR, "can't mount with "
3559 "both data=journal and delalloc");
3562 if (test_opt(sb, DIOREAD_NOLOCK)) {
3563 ext4_msg(sb, KERN_ERR, "can't mount with "
3564 "both data=journal and dioread_nolock");
3567 if (test_opt(sb, DAX)) {
3568 ext4_msg(sb, KERN_ERR, "can't mount with "
3569 "both data=journal and dax");
3572 if (ext4_has_feature_encrypt(sb)) {
3573 ext4_msg(sb, KERN_WARNING,
3574 "encrypted files will use data=ordered "
3575 "instead of data journaling mode");
3577 if (test_opt(sb, DELALLOC))
3578 clear_opt(sb, DELALLOC);
3580 sb->s_iflags |= SB_I_CGROUPWB;
3583 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3584 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3586 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3587 (ext4_has_compat_features(sb) ||
3588 ext4_has_ro_compat_features(sb) ||
3589 ext4_has_incompat_features(sb)))
3590 ext4_msg(sb, KERN_WARNING,
3591 "feature flags set on rev 0 fs, "
3592 "running e2fsck is recommended");
3594 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3595 set_opt2(sb, HURD_COMPAT);
3596 if (ext4_has_feature_64bit(sb)) {
3597 ext4_msg(sb, KERN_ERR,
3598 "The Hurd can't support 64-bit file systems");
3603 if (IS_EXT2_SB(sb)) {
3604 if (ext2_feature_set_ok(sb))
3605 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3606 "using the ext4 subsystem");
3608 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3609 "to feature incompatibilities");
3614 if (IS_EXT3_SB(sb)) {
3615 if (ext3_feature_set_ok(sb))
3616 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3617 "using the ext4 subsystem");
3619 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3620 "to feature incompatibilities");
3626 * Check feature flags regardless of the revision level, since we
3627 * previously didn't change the revision level when setting the flags,
3628 * so there is a chance incompat flags are set on a rev 0 filesystem.
3630 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3633 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3634 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3635 blocksize > EXT4_MAX_BLOCK_SIZE) {
3636 ext4_msg(sb, KERN_ERR,
3637 "Unsupported filesystem blocksize %d (%d log_block_size)",
3638 blocksize, le32_to_cpu(es->s_log_block_size));
3641 if (le32_to_cpu(es->s_log_block_size) >
3642 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3643 ext4_msg(sb, KERN_ERR,
3644 "Invalid log block size: %u",
3645 le32_to_cpu(es->s_log_block_size));
3649 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
3650 ext4_msg(sb, KERN_ERR,
3651 "Number of reserved GDT blocks insanely large: %d",
3652 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
3656 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
3657 err = bdev_dax_supported(sb, blocksize);
3662 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
3663 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
3664 es->s_encryption_level);
3668 if (sb->s_blocksize != blocksize) {
3669 /* Validate the filesystem blocksize */
3670 if (!sb_set_blocksize(sb, blocksize)) {
3671 ext4_msg(sb, KERN_ERR, "bad block size %d",
3677 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3678 offset = do_div(logical_sb_block, blocksize);
3679 bh = sb_bread_unmovable(sb, logical_sb_block);
3681 ext4_msg(sb, KERN_ERR,
3682 "Can't read superblock on 2nd try");
3685 es = (struct ext4_super_block *)(bh->b_data + offset);
3687 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3688 ext4_msg(sb, KERN_ERR,
3689 "Magic mismatch, very weird!");
3694 has_huge_files = ext4_has_feature_huge_file(sb);
3695 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3697 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3699 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3700 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3701 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3703 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3704 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3705 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3706 (!is_power_of_2(sbi->s_inode_size)) ||
3707 (sbi->s_inode_size > blocksize)) {
3708 ext4_msg(sb, KERN_ERR,
3709 "unsupported inode size: %d",
3713 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3714 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3717 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3718 if (ext4_has_feature_64bit(sb)) {
3719 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3720 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3721 !is_power_of_2(sbi->s_desc_size)) {
3722 ext4_msg(sb, KERN_ERR,
3723 "unsupported descriptor size %lu",
3728 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3730 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3731 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3733 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3734 if (sbi->s_inodes_per_block == 0)
3736 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
3737 sbi->s_inodes_per_group > blocksize * 8) {
3738 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
3739 sbi->s_blocks_per_group);
3742 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3743 sbi->s_inodes_per_block;
3744 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3746 sbi->s_mount_state = le16_to_cpu(es->s_state);
3747 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3748 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3750 for (i = 0; i < 4; i++)
3751 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3752 sbi->s_def_hash_version = es->s_def_hash_version;
3753 if (ext4_has_feature_dir_index(sb)) {
3754 i = le32_to_cpu(es->s_flags);
3755 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3756 sbi->s_hash_unsigned = 3;
3757 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3758 #ifdef __CHAR_UNSIGNED__
3759 if (!(sb->s_flags & MS_RDONLY))
3761 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3762 sbi->s_hash_unsigned = 3;
3764 if (!(sb->s_flags & MS_RDONLY))
3766 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3771 /* Handle clustersize */
3772 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3773 has_bigalloc = ext4_has_feature_bigalloc(sb);
3775 if (clustersize < blocksize) {
3776 ext4_msg(sb, KERN_ERR,
3777 "cluster size (%d) smaller than "
3778 "block size (%d)", clustersize, blocksize);
3781 if (le32_to_cpu(es->s_log_cluster_size) >
3782 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3783 ext4_msg(sb, KERN_ERR,
3784 "Invalid log cluster size: %u",
3785 le32_to_cpu(es->s_log_cluster_size));
3788 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3789 le32_to_cpu(es->s_log_block_size);
3790 sbi->s_clusters_per_group =
3791 le32_to_cpu(es->s_clusters_per_group);
3792 if (sbi->s_clusters_per_group > blocksize * 8) {
3793 ext4_msg(sb, KERN_ERR,
3794 "#clusters per group too big: %lu",
3795 sbi->s_clusters_per_group);
3798 if (sbi->s_blocks_per_group !=
3799 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3800 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3801 "clusters per group (%lu) inconsistent",
3802 sbi->s_blocks_per_group,
3803 sbi->s_clusters_per_group);
3807 if (clustersize != blocksize) {
3808 ext4_warning(sb, "fragment/cluster size (%d) != "
3809 "block size (%d)", clustersize,
3811 clustersize = blocksize;
3813 if (sbi->s_blocks_per_group > blocksize * 8) {
3814 ext4_msg(sb, KERN_ERR,
3815 "#blocks per group too big: %lu",
3816 sbi->s_blocks_per_group);
3819 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3820 sbi->s_cluster_bits = 0;
3822 sbi->s_cluster_ratio = clustersize / blocksize;
3824 /* Do we have standard group size of clustersize * 8 blocks ? */
3825 if (sbi->s_blocks_per_group == clustersize << 3)
3826 set_opt2(sb, STD_GROUP_SIZE);
3829 * Test whether we have more sectors than will fit in sector_t,
3830 * and whether the max offset is addressable by the page cache.
3832 err = generic_check_addressable(sb->s_blocksize_bits,
3833 ext4_blocks_count(es));
3835 ext4_msg(sb, KERN_ERR, "filesystem"
3836 " too large to mount safely on this system");
3837 if (sizeof(sector_t) < 8)
3838 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3842 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3845 /* check blocks count against device size */
3846 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3847 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3848 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3849 "exceeds size of device (%llu blocks)",
3850 ext4_blocks_count(es), blocks_count);
3855 * It makes no sense for the first data block to be beyond the end
3856 * of the filesystem.
3858 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3859 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3860 "block %u is beyond end of filesystem (%llu)",
3861 le32_to_cpu(es->s_first_data_block),
3862 ext4_blocks_count(es));
3865 blocks_count = (ext4_blocks_count(es) -
3866 le32_to_cpu(es->s_first_data_block) +
3867 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3868 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3869 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3870 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3871 "(block count %llu, first data block %u, "
3872 "blocks per group %lu)", sbi->s_groups_count,
3873 ext4_blocks_count(es),
3874 le32_to_cpu(es->s_first_data_block),
3875 EXT4_BLOCKS_PER_GROUP(sb));
3878 sbi->s_groups_count = blocks_count;
3879 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3880 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3881 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3882 EXT4_DESC_PER_BLOCK(sb);
3883 if (ext4_has_feature_meta_bg(sb)) {
3884 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
3885 ext4_msg(sb, KERN_WARNING,
3886 "first meta block group too large: %u "
3887 "(group descriptor block count %u)",
3888 le32_to_cpu(es->s_first_meta_bg), db_count);
3892 sbi->s_group_desc = ext4_kvmalloc(db_count *
3893 sizeof(struct buffer_head *),
3895 if (sbi->s_group_desc == NULL) {
3896 ext4_msg(sb, KERN_ERR, "not enough memory");
3901 bgl_lock_init(sbi->s_blockgroup_lock);
3903 for (i = 0; i < db_count; i++) {
3904 block = descriptor_loc(sb, logical_sb_block, i);
3905 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
3906 if (!sbi->s_group_desc[i]) {
3907 ext4_msg(sb, KERN_ERR,
3908 "can't read group descriptor %d", i);
3913 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
3914 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3915 ret = -EFSCORRUPTED;
3919 sbi->s_gdb_count = db_count;
3920 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3921 spin_lock_init(&sbi->s_next_gen_lock);
3923 setup_timer(&sbi->s_err_report, print_daily_error_info,
3924 (unsigned long) sb);
3926 /* Register extent status tree shrinker */
3927 if (ext4_es_register_shrinker(sbi))
3930 sbi->s_stripe = ext4_get_stripe_size(sbi);
3931 sbi->s_extent_max_zeroout_kb = 32;
3934 * set up enough so that it can read an inode
3936 sb->s_op = &ext4_sops;
3937 sb->s_export_op = &ext4_export_ops;
3938 sb->s_xattr = ext4_xattr_handlers;
3939 sb->s_cop = &ext4_cryptops;
3941 sb->dq_op = &ext4_quota_operations;
3942 if (ext4_has_feature_quota(sb))
3943 sb->s_qcop = &dquot_quotactl_sysfile_ops;
3945 sb->s_qcop = &ext4_qctl_operations;
3946 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
3948 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3950 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3951 mutex_init(&sbi->s_orphan_lock);
3955 needs_recovery = (es->s_last_orphan != 0 ||
3956 ext4_has_feature_journal_needs_recovery(sb));
3958 if (ext4_has_feature_mmp(sb) && !(sb->s_flags & MS_RDONLY))
3959 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3960 goto failed_mount3a;
3963 * The first inode we look at is the journal inode. Don't try
3964 * root first: it may be modified in the journal!
3966 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
3967 err = ext4_load_journal(sb, es, journal_devnum);
3969 goto failed_mount3a;
3970 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3971 ext4_has_feature_journal_needs_recovery(sb)) {
3972 ext4_msg(sb, KERN_ERR, "required journal recovery "
3973 "suppressed and not mounted read-only");
3974 goto failed_mount_wq;
3976 /* Nojournal mode, all journal mount options are illegal */
3977 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
3978 ext4_msg(sb, KERN_ERR, "can't mount with "
3979 "journal_checksum, fs mounted w/o journal");
3980 goto failed_mount_wq;
3982 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3983 ext4_msg(sb, KERN_ERR, "can't mount with "
3984 "journal_async_commit, fs mounted w/o journal");
3985 goto failed_mount_wq;
3987 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
3988 ext4_msg(sb, KERN_ERR, "can't mount with "
3989 "commit=%lu, fs mounted w/o journal",
3990 sbi->s_commit_interval / HZ);
3991 goto failed_mount_wq;
3993 if (EXT4_MOUNT_DATA_FLAGS &
3994 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
3995 ext4_msg(sb, KERN_ERR, "can't mount with "
3996 "data=, fs mounted w/o journal");
3997 goto failed_mount_wq;
3999 sbi->s_def_mount_opt &= EXT4_MOUNT_JOURNAL_CHECKSUM;
4000 clear_opt(sb, JOURNAL_CHECKSUM);
4001 clear_opt(sb, DATA_FLAGS);
4002 sbi->s_journal = NULL;
4007 if (ext4_has_feature_64bit(sb) &&
4008 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4009 JBD2_FEATURE_INCOMPAT_64BIT)) {
4010 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
4011 goto failed_mount_wq;
4014 if (!set_journal_csum_feature_set(sb)) {
4015 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4017 goto failed_mount_wq;
4020 /* We have now updated the journal if required, so we can
4021 * validate the data journaling mode. */
4022 switch (test_opt(sb, DATA_FLAGS)) {
4024 /* No mode set, assume a default based on the journal
4025 * capabilities: ORDERED_DATA if the journal can
4026 * cope, else JOURNAL_DATA
4028 if (jbd2_journal_check_available_features
4029 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
4030 set_opt(sb, ORDERED_DATA);
4032 set_opt(sb, JOURNAL_DATA);
4035 case EXT4_MOUNT_ORDERED_DATA:
4036 case EXT4_MOUNT_WRITEBACK_DATA:
4037 if (!jbd2_journal_check_available_features
4038 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4039 ext4_msg(sb, KERN_ERR, "Journal does not support "
4040 "requested data journaling mode");
4041 goto failed_mount_wq;
4047 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
4048 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4049 ext4_msg(sb, KERN_ERR, "can't mount with "
4050 "journal_async_commit in data=ordered mode");
4051 goto failed_mount_wq;
4054 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4056 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4059 sbi->s_mb_cache = ext4_xattr_create_cache();
4060 if (!sbi->s_mb_cache) {
4061 ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
4062 goto failed_mount_wq;
4065 if ((DUMMY_ENCRYPTION_ENABLED(sbi) || ext4_has_feature_encrypt(sb)) &&
4066 (blocksize != PAGE_SIZE)) {
4067 ext4_msg(sb, KERN_ERR,
4068 "Unsupported blocksize for fs encryption");
4069 goto failed_mount_wq;
4072 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !(sb->s_flags & MS_RDONLY) &&
4073 !ext4_has_feature_encrypt(sb)) {
4074 ext4_set_feature_encrypt(sb);
4075 ext4_commit_super(sb, 1);
4079 * Get the # of file system overhead blocks from the
4080 * superblock if present.
4082 if (es->s_overhead_clusters)
4083 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4085 err = ext4_calculate_overhead(sb);
4087 goto failed_mount_wq;
4091 * The maximum number of concurrent works can be high and
4092 * concurrency isn't really necessary. Limit it to 1.
4094 EXT4_SB(sb)->rsv_conversion_wq =
4095 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4096 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4097 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
4103 * The jbd2_journal_load will have done any necessary log recovery,
4104 * so we can safely mount the rest of the filesystem now.
4107 root = ext4_iget(sb, EXT4_ROOT_INO);
4109 ext4_msg(sb, KERN_ERR, "get root inode failed");
4110 ret = PTR_ERR(root);
4114 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4115 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4119 sb->s_root = d_make_root(root);
4121 ext4_msg(sb, KERN_ERR, "get root dentry failed");
4126 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
4127 sb->s_flags |= MS_RDONLY;
4129 /* determine the minimum size of new large inodes, if present */
4130 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE &&
4131 sbi->s_want_extra_isize == 0) {
4132 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4133 EXT4_GOOD_OLD_INODE_SIZE;
4134 if (ext4_has_feature_extra_isize(sb)) {
4135 if (sbi->s_want_extra_isize <
4136 le16_to_cpu(es->s_want_extra_isize))
4137 sbi->s_want_extra_isize =
4138 le16_to_cpu(es->s_want_extra_isize);
4139 if (sbi->s_want_extra_isize <
4140 le16_to_cpu(es->s_min_extra_isize))
4141 sbi->s_want_extra_isize =
4142 le16_to_cpu(es->s_min_extra_isize);
4145 /* Check if enough inode space is available */
4146 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4147 sbi->s_inode_size) {
4148 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4149 EXT4_GOOD_OLD_INODE_SIZE;
4150 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4154 ext4_set_resv_clusters(sb);
4156 err = ext4_setup_system_zone(sb);
4158 ext4_msg(sb, KERN_ERR, "failed to initialize system "
4160 goto failed_mount4a;
4164 err = ext4_mb_init(sb);
4166 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4171 block = ext4_count_free_clusters(sb);
4172 ext4_free_blocks_count_set(sbi->s_es,
4173 EXT4_C2B(sbi, block));
4174 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4177 unsigned long freei = ext4_count_free_inodes(sb);
4178 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
4179 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4183 err = percpu_counter_init(&sbi->s_dirs_counter,
4184 ext4_count_dirs(sb), GFP_KERNEL);
4186 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4189 err = percpu_init_rwsem(&sbi->s_journal_flag_rwsem);
4192 ext4_msg(sb, KERN_ERR, "insufficient memory");
4196 if (ext4_has_feature_flex_bg(sb))
4197 if (!ext4_fill_flex_info(sb)) {
4198 ext4_msg(sb, KERN_ERR,
4199 "unable to initialize "
4200 "flex_bg meta info!");
4204 err = ext4_register_li_request(sb, first_not_zeroed);
4208 err = ext4_register_sysfs(sb);
4213 /* Enable quota usage during mount. */
4214 if (ext4_has_feature_quota(sb) && !(sb->s_flags & MS_RDONLY)) {
4215 err = ext4_enable_quotas(sb);
4219 #endif /* CONFIG_QUOTA */
4221 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4222 ext4_orphan_cleanup(sb, es);
4223 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4224 if (needs_recovery) {
4225 ext4_msg(sb, KERN_INFO, "recovery complete");
4226 ext4_mark_recovery_complete(sb, es);
4228 if (EXT4_SB(sb)->s_journal) {
4229 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4230 descr = " journalled data mode";
4231 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4232 descr = " ordered data mode";
4234 descr = " writeback data mode";
4236 descr = "out journal";
4238 if (test_opt(sb, DISCARD)) {
4239 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4240 if (!blk_queue_discard(q))
4241 ext4_msg(sb, KERN_WARNING,
4242 "mounting with \"discard\" option, but "
4243 "the device does not support discard");
4246 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
4247 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4248 "Opts: %.*s%s%s", descr,
4249 (int) sizeof(sbi->s_es->s_mount_opts),
4250 sbi->s_es->s_mount_opts,
4251 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4253 if (es->s_error_count)
4254 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4256 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4257 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4258 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4259 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4266 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4271 ext4_unregister_sysfs(sb);
4274 ext4_unregister_li_request(sb);
4276 ext4_mb_release(sb);
4277 if (sbi->s_flex_groups)
4278 kvfree(sbi->s_flex_groups);
4279 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4280 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4281 percpu_counter_destroy(&sbi->s_dirs_counter);
4282 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4284 ext4_ext_release(sb);
4285 ext4_release_system_zone(sb);
4290 ext4_msg(sb, KERN_ERR, "mount failed");
4291 if (EXT4_SB(sb)->rsv_conversion_wq)
4292 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4294 if (sbi->s_mb_cache) {
4295 ext4_xattr_destroy_cache(sbi->s_mb_cache);
4296 sbi->s_mb_cache = NULL;
4298 if (sbi->s_journal) {
4299 jbd2_journal_destroy(sbi->s_journal);
4300 sbi->s_journal = NULL;
4303 ext4_es_unregister_shrinker(sbi);
4305 del_timer_sync(&sbi->s_err_report);
4307 kthread_stop(sbi->s_mmp_tsk);
4309 for (i = 0; i < db_count; i++)
4310 brelse(sbi->s_group_desc[i]);
4311 kvfree(sbi->s_group_desc);
4313 if (sbi->s_chksum_driver)
4314 crypto_free_shash(sbi->s_chksum_driver);
4316 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4317 kfree(sbi->s_qf_names[i]);
4319 ext4_blkdev_remove(sbi);
4322 sb->s_fs_info = NULL;
4323 kfree(sbi->s_blockgroup_lock);
4327 return err ? err : ret;
4331 * Setup any per-fs journal parameters now. We'll do this both on
4332 * initial mount, once the journal has been initialised but before we've
4333 * done any recovery; and again on any subsequent remount.
4335 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4337 struct ext4_sb_info *sbi = EXT4_SB(sb);
4339 journal->j_commit_interval = sbi->s_commit_interval;
4340 journal->j_min_batch_time = sbi->s_min_batch_time;
4341 journal->j_max_batch_time = sbi->s_max_batch_time;
4343 write_lock(&journal->j_state_lock);
4344 if (test_opt(sb, BARRIER))
4345 journal->j_flags |= JBD2_BARRIER;
4347 journal->j_flags &= ~JBD2_BARRIER;
4348 if (test_opt(sb, DATA_ERR_ABORT))
4349 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4351 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4352 write_unlock(&journal->j_state_lock);
4355 static struct inode *ext4_get_journal_inode(struct super_block *sb,
4356 unsigned int journal_inum)
4358 struct inode *journal_inode;
4361 * Test for the existence of a valid inode on disk. Bad things
4362 * happen if we iget() an unused inode, as the subsequent iput()
4363 * will try to delete it.
4365 journal_inode = ext4_iget(sb, journal_inum);
4366 if (IS_ERR(journal_inode)) {
4367 ext4_msg(sb, KERN_ERR, "no journal found");
4370 if (!journal_inode->i_nlink) {
4371 make_bad_inode(journal_inode);
4372 iput(journal_inode);
4373 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4377 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4378 journal_inode, journal_inode->i_size);
4379 if (!S_ISREG(journal_inode->i_mode)) {
4380 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4381 iput(journal_inode);
4384 return journal_inode;
4387 static journal_t *ext4_get_journal(struct super_block *sb,
4388 unsigned int journal_inum)
4390 struct inode *journal_inode;
4393 BUG_ON(!ext4_has_feature_journal(sb));
4395 journal_inode = ext4_get_journal_inode(sb, journal_inum);
4399 journal = jbd2_journal_init_inode(journal_inode);
4401 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4402 iput(journal_inode);
4405 journal->j_private = sb;
4406 ext4_init_journal_params(sb, journal);
4410 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4413 struct buffer_head *bh;
4417 int hblock, blocksize;
4418 ext4_fsblk_t sb_block;
4419 unsigned long offset;
4420 struct ext4_super_block *es;
4421 struct block_device *bdev;
4423 BUG_ON(!ext4_has_feature_journal(sb));
4425 bdev = ext4_blkdev_get(j_dev, sb);
4429 blocksize = sb->s_blocksize;
4430 hblock = bdev_logical_block_size(bdev);
4431 if (blocksize < hblock) {
4432 ext4_msg(sb, KERN_ERR,
4433 "blocksize too small for journal device");
4437 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4438 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4439 set_blocksize(bdev, blocksize);
4440 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4441 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4442 "external journal");
4446 es = (struct ext4_super_block *) (bh->b_data + offset);
4447 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4448 !(le32_to_cpu(es->s_feature_incompat) &
4449 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4450 ext4_msg(sb, KERN_ERR, "external journal has "
4456 if ((le32_to_cpu(es->s_feature_ro_compat) &
4457 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4458 es->s_checksum != ext4_superblock_csum(sb, es)) {
4459 ext4_msg(sb, KERN_ERR, "external journal has "
4460 "corrupt superblock");
4465 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4466 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4471 len = ext4_blocks_count(es);
4472 start = sb_block + 1;
4473 brelse(bh); /* we're done with the superblock */
4475 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4476 start, len, blocksize);
4478 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4481 journal->j_private = sb;
4482 ll_rw_block(REQ_OP_READ, REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4483 wait_on_buffer(journal->j_sb_buffer);
4484 if (!buffer_uptodate(journal->j_sb_buffer)) {
4485 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4488 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4489 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4490 "user (unsupported) - %d",
4491 be32_to_cpu(journal->j_superblock->s_nr_users));
4494 EXT4_SB(sb)->journal_bdev = bdev;
4495 ext4_init_journal_params(sb, journal);
4499 jbd2_journal_destroy(journal);
4501 ext4_blkdev_put(bdev);
4505 static int ext4_load_journal(struct super_block *sb,
4506 struct ext4_super_block *es,
4507 unsigned long journal_devnum)
4510 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4513 int really_read_only;
4515 BUG_ON(!ext4_has_feature_journal(sb));
4517 if (journal_devnum &&
4518 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4519 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4520 "numbers have changed");
4521 journal_dev = new_decode_dev(journal_devnum);
4523 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4525 really_read_only = bdev_read_only(sb->s_bdev);
4528 * Are we loading a blank journal or performing recovery after a
4529 * crash? For recovery, we need to check in advance whether we
4530 * can get read-write access to the device.
4532 if (ext4_has_feature_journal_needs_recovery(sb)) {
4533 if (sb->s_flags & MS_RDONLY) {
4534 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4535 "required on readonly filesystem");
4536 if (really_read_only) {
4537 ext4_msg(sb, KERN_ERR, "write access "
4538 "unavailable, cannot proceed");
4541 ext4_msg(sb, KERN_INFO, "write access will "
4542 "be enabled during recovery");
4546 if (journal_inum && journal_dev) {
4547 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4548 "and inode journals!");
4553 if (!(journal = ext4_get_journal(sb, journal_inum)))
4556 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4560 if (!(journal->j_flags & JBD2_BARRIER))
4561 ext4_msg(sb, KERN_INFO, "barriers disabled");
4563 if (!ext4_has_feature_journal_needs_recovery(sb))
4564 err = jbd2_journal_wipe(journal, !really_read_only);
4566 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4568 memcpy(save, ((char *) es) +
4569 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4570 err = jbd2_journal_load(journal);
4572 memcpy(((char *) es) + EXT4_S_ERR_START,
4573 save, EXT4_S_ERR_LEN);
4578 ext4_msg(sb, KERN_ERR, "error loading journal");
4579 jbd2_journal_destroy(journal);
4583 EXT4_SB(sb)->s_journal = journal;
4584 ext4_clear_journal_err(sb, es);
4586 if (!really_read_only && journal_devnum &&
4587 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4588 es->s_journal_dev = cpu_to_le32(journal_devnum);
4590 /* Make sure we flush the recovery flag to disk. */
4591 ext4_commit_super(sb, 1);
4597 static int ext4_commit_super(struct super_block *sb, int sync)
4599 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4600 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4603 if (!sbh || block_device_ejected(sb))
4606 * If the file system is mounted read-only, don't update the
4607 * superblock write time. This avoids updating the superblock
4608 * write time when we are mounting the root file system
4609 * read/only but we need to replay the journal; at that point,
4610 * for people who are east of GMT and who make their clock
4611 * tick in localtime for Windows bug-for-bug compatibility,
4612 * the clock is set in the future, and this will cause e2fsck
4613 * to complain and force a full file system check.
4615 if (!(sb->s_flags & MS_RDONLY))
4616 es->s_wtime = cpu_to_le32(get_seconds());
4617 if (sb->s_bdev->bd_part)
4618 es->s_kbytes_written =
4619 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4620 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4621 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4623 es->s_kbytes_written =
4624 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4625 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4626 ext4_free_blocks_count_set(es,
4627 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4628 &EXT4_SB(sb)->s_freeclusters_counter)));
4629 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4630 es->s_free_inodes_count =
4631 cpu_to_le32(percpu_counter_sum_positive(
4632 &EXT4_SB(sb)->s_freeinodes_counter));
4633 BUFFER_TRACE(sbh, "marking dirty");
4634 ext4_superblock_csum_set(sb);
4637 if (buffer_write_io_error(sbh)) {
4639 * Oh, dear. A previous attempt to write the
4640 * superblock failed. This could happen because the
4641 * USB device was yanked out. Or it could happen to
4642 * be a transient write error and maybe the block will
4643 * be remapped. Nothing we can do but to retry the
4644 * write and hope for the best.
4646 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4647 "superblock detected");
4648 clear_buffer_write_io_error(sbh);
4649 set_buffer_uptodate(sbh);
4651 mark_buffer_dirty(sbh);
4654 error = __sync_dirty_buffer(sbh,
4655 test_opt(sb, BARRIER) ? REQ_FUA : REQ_SYNC);
4659 error = buffer_write_io_error(sbh);
4661 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4663 clear_buffer_write_io_error(sbh);
4664 set_buffer_uptodate(sbh);
4671 * Have we just finished recovery? If so, and if we are mounting (or
4672 * remounting) the filesystem readonly, then we will end up with a
4673 * consistent fs on disk. Record that fact.
4675 static void ext4_mark_recovery_complete(struct super_block *sb,
4676 struct ext4_super_block *es)
4678 journal_t *journal = EXT4_SB(sb)->s_journal;
4680 if (!ext4_has_feature_journal(sb)) {
4681 BUG_ON(journal != NULL);
4684 jbd2_journal_lock_updates(journal);
4685 if (jbd2_journal_flush(journal) < 0)
4688 if (ext4_has_feature_journal_needs_recovery(sb) &&
4689 sb->s_flags & MS_RDONLY) {
4690 ext4_clear_feature_journal_needs_recovery(sb);
4691 ext4_commit_super(sb, 1);
4695 jbd2_journal_unlock_updates(journal);
4699 * If we are mounting (or read-write remounting) a filesystem whose journal
4700 * has recorded an error from a previous lifetime, move that error to the
4701 * main filesystem now.
4703 static void ext4_clear_journal_err(struct super_block *sb,
4704 struct ext4_super_block *es)
4710 BUG_ON(!ext4_has_feature_journal(sb));
4712 journal = EXT4_SB(sb)->s_journal;
4715 * Now check for any error status which may have been recorded in the
4716 * journal by a prior ext4_error() or ext4_abort()
4719 j_errno = jbd2_journal_errno(journal);
4723 errstr = ext4_decode_error(sb, j_errno, nbuf);
4724 ext4_warning(sb, "Filesystem error recorded "
4725 "from previous mount: %s", errstr);
4726 ext4_warning(sb, "Marking fs in need of filesystem check.");
4728 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4729 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4730 ext4_commit_super(sb, 1);
4732 jbd2_journal_clear_err(journal);
4733 jbd2_journal_update_sb_errno(journal);
4738 * Force the running and committing transactions to commit,
4739 * and wait on the commit.
4741 int ext4_force_commit(struct super_block *sb)
4745 if (sb->s_flags & MS_RDONLY)
4748 journal = EXT4_SB(sb)->s_journal;
4749 return ext4_journal_force_commit(journal);
4752 static int ext4_sync_fs(struct super_block *sb, int wait)
4756 bool needs_barrier = false;
4757 struct ext4_sb_info *sbi = EXT4_SB(sb);
4759 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
4762 trace_ext4_sync_fs(sb, wait);
4763 flush_workqueue(sbi->rsv_conversion_wq);
4765 * Writeback quota in non-journalled quota case - journalled quota has
4768 dquot_writeback_dquots(sb, -1);
4770 * Data writeback is possible w/o journal transaction, so barrier must
4771 * being sent at the end of the function. But we can skip it if
4772 * transaction_commit will do it for us.
4774 if (sbi->s_journal) {
4775 target = jbd2_get_latest_transaction(sbi->s_journal);
4776 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4777 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4778 needs_barrier = true;
4780 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4782 ret = jbd2_log_wait_commit(sbi->s_journal,
4785 } else if (wait && test_opt(sb, BARRIER))
4786 needs_barrier = true;
4787 if (needs_barrier) {
4789 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4798 * LVM calls this function before a (read-only) snapshot is created. This
4799 * gives us a chance to flush the journal completely and mark the fs clean.
4801 * Note that only this function cannot bring a filesystem to be in a clean
4802 * state independently. It relies on upper layer to stop all data & metadata
4805 static int ext4_freeze(struct super_block *sb)
4810 if (sb->s_flags & MS_RDONLY)
4813 journal = EXT4_SB(sb)->s_journal;
4816 /* Now we set up the journal barrier. */
4817 jbd2_journal_lock_updates(journal);
4820 * Don't clear the needs_recovery flag if we failed to
4821 * flush the journal.
4823 error = jbd2_journal_flush(journal);
4827 /* Journal blocked and flushed, clear needs_recovery flag. */
4828 ext4_clear_feature_journal_needs_recovery(sb);
4831 error = ext4_commit_super(sb, 1);
4834 /* we rely on upper layer to stop further updates */
4835 jbd2_journal_unlock_updates(journal);
4840 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4841 * flag here, even though the filesystem is not technically dirty yet.
4843 static int ext4_unfreeze(struct super_block *sb)
4845 if ((sb->s_flags & MS_RDONLY) || ext4_forced_shutdown(EXT4_SB(sb)))
4848 if (EXT4_SB(sb)->s_journal) {
4849 /* Reset the needs_recovery flag before the fs is unlocked. */
4850 ext4_set_feature_journal_needs_recovery(sb);
4853 ext4_commit_super(sb, 1);
4858 * Structure to save mount options for ext4_remount's benefit
4860 struct ext4_mount_options {
4861 unsigned long s_mount_opt;
4862 unsigned long s_mount_opt2;
4865 unsigned long s_commit_interval;
4866 u32 s_min_batch_time, s_max_batch_time;
4869 char *s_qf_names[EXT4_MAXQUOTAS];
4873 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4875 struct ext4_super_block *es;
4876 struct ext4_sb_info *sbi = EXT4_SB(sb);
4877 unsigned long old_sb_flags;
4878 struct ext4_mount_options old_opts;
4879 int enable_quota = 0;
4881 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4886 char *orig_data = kstrdup(data, GFP_KERNEL);
4888 /* Store the original options */
4889 old_sb_flags = sb->s_flags;
4890 old_opts.s_mount_opt = sbi->s_mount_opt;
4891 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4892 old_opts.s_resuid = sbi->s_resuid;
4893 old_opts.s_resgid = sbi->s_resgid;
4894 old_opts.s_commit_interval = sbi->s_commit_interval;
4895 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4896 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4898 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4899 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4900 if (sbi->s_qf_names[i]) {
4901 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4903 if (!old_opts.s_qf_names[i]) {
4904 for (j = 0; j < i; j++)
4905 kfree(old_opts.s_qf_names[j]);
4910 old_opts.s_qf_names[i] = NULL;
4912 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4913 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4915 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4920 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
4921 test_opt(sb, JOURNAL_CHECKSUM)) {
4922 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
4923 "during remount not supported; ignoring");
4924 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
4927 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4928 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4929 ext4_msg(sb, KERN_ERR, "can't mount with "
4930 "both data=journal and delalloc");
4934 if (test_opt(sb, DIOREAD_NOLOCK)) {
4935 ext4_msg(sb, KERN_ERR, "can't mount with "
4936 "both data=journal and dioread_nolock");
4940 if (test_opt(sb, DAX)) {
4941 ext4_msg(sb, KERN_ERR, "can't mount with "
4942 "both data=journal and dax");
4946 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
4947 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4948 ext4_msg(sb, KERN_ERR, "can't mount with "
4949 "journal_async_commit in data=ordered mode");
4955 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
4956 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
4957 "dax flag with busy inodes while remounting");
4958 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
4961 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4962 ext4_abort(sb, "Abort forced by user");
4964 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4965 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4969 if (sbi->s_journal) {
4970 ext4_init_journal_params(sb, sbi->s_journal);
4971 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4974 if (*flags & MS_LAZYTIME)
4975 sb->s_flags |= MS_LAZYTIME;
4977 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4978 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4983 if (*flags & MS_RDONLY) {
4984 err = sync_filesystem(sb);
4987 err = dquot_suspend(sb, -1);
4992 * First of all, the unconditional stuff we have to do
4993 * to disable replay of the journal when we next remount
4995 sb->s_flags |= MS_RDONLY;
4998 * OK, test if we are remounting a valid rw partition
4999 * readonly, and if so set the rdonly flag and then
5000 * mark the partition as valid again.
5002 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
5003 (sbi->s_mount_state & EXT4_VALID_FS))
5004 es->s_state = cpu_to_le16(sbi->s_mount_state);
5007 ext4_mark_recovery_complete(sb, es);
5009 /* Make sure we can mount this feature set readwrite */
5010 if (ext4_has_feature_readonly(sb) ||
5011 !ext4_feature_set_ok(sb, 0)) {
5016 * Make sure the group descriptor checksums
5017 * are sane. If they aren't, refuse to remount r/w.
5019 for (g = 0; g < sbi->s_groups_count; g++) {
5020 struct ext4_group_desc *gdp =
5021 ext4_get_group_desc(sb, g, NULL);
5023 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
5024 ext4_msg(sb, KERN_ERR,
5025 "ext4_remount: Checksum for group %u failed (%u!=%u)",
5026 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
5027 le16_to_cpu(gdp->bg_checksum));
5034 * If we have an unprocessed orphan list hanging
5035 * around from a previously readonly bdev mount,
5036 * require a full umount/remount for now.
5038 if (es->s_last_orphan) {
5039 ext4_msg(sb, KERN_WARNING, "Couldn't "
5040 "remount RDWR because of unprocessed "
5041 "orphan inode list. Please "
5042 "umount/remount instead");
5048 * Mounting a RDONLY partition read-write, so reread
5049 * and store the current valid flag. (It may have
5050 * been changed by e2fsck since we originally mounted
5054 ext4_clear_journal_err(sb, es);
5055 sbi->s_mount_state = le16_to_cpu(es->s_state);
5056 if (!ext4_setup_super(sb, es, 0))
5057 sb->s_flags &= ~MS_RDONLY;
5058 if (ext4_has_feature_mmp(sb))
5059 if (ext4_multi_mount_protect(sb,
5060 le64_to_cpu(es->s_mmp_block))) {
5069 * Reinitialize lazy itable initialization thread based on
5072 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
5073 ext4_unregister_li_request(sb);
5075 ext4_group_t first_not_zeroed;
5076 first_not_zeroed = ext4_has_uninit_itable(sb);
5077 ext4_register_li_request(sb, first_not_zeroed);
5080 ext4_setup_system_zone(sb);
5081 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
5082 ext4_commit_super(sb, 1);
5085 /* Release old quota file names */
5086 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5087 kfree(old_opts.s_qf_names[i]);
5089 if (sb_any_quota_suspended(sb))
5090 dquot_resume(sb, -1);
5091 else if (ext4_has_feature_quota(sb)) {
5092 err = ext4_enable_quotas(sb);
5099 *flags = (*flags & ~MS_LAZYTIME) | (sb->s_flags & MS_LAZYTIME);
5100 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5105 sb->s_flags = old_sb_flags;
5106 sbi->s_mount_opt = old_opts.s_mount_opt;
5107 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
5108 sbi->s_resuid = old_opts.s_resuid;
5109 sbi->s_resgid = old_opts.s_resgid;
5110 sbi->s_commit_interval = old_opts.s_commit_interval;
5111 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5112 sbi->s_max_batch_time = old_opts.s_max_batch_time;
5114 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
5115 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
5116 kfree(sbi->s_qf_names[i]);
5117 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
5125 static int ext4_statfs_project(struct super_block *sb,
5126 kprojid_t projid, struct kstatfs *buf)
5129 struct dquot *dquot;
5133 qid = make_kqid_projid(projid);
5134 dquot = dqget(sb, qid);
5136 return PTR_ERR(dquot);
5137 spin_lock(&dq_data_lock);
5139 limit = (dquot->dq_dqb.dqb_bsoftlimit ?
5140 dquot->dq_dqb.dqb_bsoftlimit :
5141 dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits;
5142 if (limit && buf->f_blocks > limit) {
5143 curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits;
5144 buf->f_blocks = limit;
5145 buf->f_bfree = buf->f_bavail =
5146 (buf->f_blocks > curblock) ?
5147 (buf->f_blocks - curblock) : 0;
5150 limit = dquot->dq_dqb.dqb_isoftlimit ?
5151 dquot->dq_dqb.dqb_isoftlimit :
5152 dquot->dq_dqb.dqb_ihardlimit;
5153 if (limit && buf->f_files > limit) {
5154 buf->f_files = limit;
5156 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
5157 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
5160 spin_unlock(&dq_data_lock);
5166 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
5168 struct super_block *sb = dentry->d_sb;
5169 struct ext4_sb_info *sbi = EXT4_SB(sb);
5170 struct ext4_super_block *es = sbi->s_es;
5171 ext4_fsblk_t overhead = 0, resv_blocks;
5174 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
5176 if (!test_opt(sb, MINIX_DF))
5177 overhead = sbi->s_overhead;
5179 buf->f_type = EXT4_SUPER_MAGIC;
5180 buf->f_bsize = sb->s_blocksize;
5181 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
5182 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5183 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
5184 /* prevent underflow in case that few free space is available */
5185 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
5186 buf->f_bavail = buf->f_bfree -
5187 (ext4_r_blocks_count(es) + resv_blocks);
5188 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
5190 buf->f_files = le32_to_cpu(es->s_inodes_count);
5191 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5192 buf->f_namelen = EXT4_NAME_LEN;
5193 fsid = le64_to_cpup((void *)es->s_uuid) ^
5194 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5195 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5196 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
5199 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
5200 sb_has_quota_limits_enabled(sb, PRJQUOTA))
5201 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
5206 /* Helper function for writing quotas on sync - we need to start transaction
5207 * before quota file is locked for write. Otherwise the are possible deadlocks:
5208 * Process 1 Process 2
5209 * ext4_create() quota_sync()
5210 * jbd2_journal_start() write_dquot()
5211 * dquot_initialize() down(dqio_mutex)
5212 * down(dqio_mutex) jbd2_journal_start()
5218 static inline struct inode *dquot_to_inode(struct dquot *dquot)
5220 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
5223 static int ext4_write_dquot(struct dquot *dquot)
5227 struct inode *inode;
5229 inode = dquot_to_inode(dquot);
5230 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
5231 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
5233 return PTR_ERR(handle);
5234 ret = dquot_commit(dquot);
5235 err = ext4_journal_stop(handle);
5241 static int ext4_acquire_dquot(struct dquot *dquot)
5246 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5247 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
5249 return PTR_ERR(handle);
5250 ret = dquot_acquire(dquot);
5251 err = ext4_journal_stop(handle);
5257 static int ext4_release_dquot(struct dquot *dquot)
5262 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5263 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
5264 if (IS_ERR(handle)) {
5265 /* Release dquot anyway to avoid endless cycle in dqput() */
5266 dquot_release(dquot);
5267 return PTR_ERR(handle);
5269 ret = dquot_release(dquot);
5270 err = ext4_journal_stop(handle);
5276 static int ext4_mark_dquot_dirty(struct dquot *dquot)
5278 struct super_block *sb = dquot->dq_sb;
5279 struct ext4_sb_info *sbi = EXT4_SB(sb);
5281 /* Are we journaling quotas? */
5282 if (ext4_has_feature_quota(sb) ||
5283 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
5284 dquot_mark_dquot_dirty(dquot);
5285 return ext4_write_dquot(dquot);
5287 return dquot_mark_dquot_dirty(dquot);
5291 static int ext4_write_info(struct super_block *sb, int type)
5296 /* Data block + inode block */
5297 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
5299 return PTR_ERR(handle);
5300 ret = dquot_commit_info(sb, type);
5301 err = ext4_journal_stop(handle);
5308 * Turn on quotas during mount time - we need to find
5309 * the quota file and such...
5311 static int ext4_quota_on_mount(struct super_block *sb, int type)
5313 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5314 EXT4_SB(sb)->s_jquota_fmt, type);
5317 static void lockdep_set_quota_inode(struct inode *inode, int subclass)
5319 struct ext4_inode_info *ei = EXT4_I(inode);
5321 /* The first argument of lockdep_set_subclass has to be
5322 * *exactly* the same as the argument to init_rwsem() --- in
5323 * this case, in init_once() --- or lockdep gets unhappy
5324 * because the name of the lock is set using the
5325 * stringification of the argument to init_rwsem().
5327 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
5328 lockdep_set_subclass(&ei->i_data_sem, subclass);
5332 * Standard function to be called on quota_on
5334 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
5335 const struct path *path)
5339 if (!test_opt(sb, QUOTA))
5342 /* Quotafile not on the same filesystem? */
5343 if (path->dentry->d_sb != sb)
5345 /* Journaling quota? */
5346 if (EXT4_SB(sb)->s_qf_names[type]) {
5347 /* Quotafile not in fs root? */
5348 if (path->dentry->d_parent != sb->s_root)
5349 ext4_msg(sb, KERN_WARNING,
5350 "Quota file not on filesystem root. "
5351 "Journaled quota will not work");
5355 * When we journal data on quota file, we have to flush journal to see
5356 * all updates to the file when we bypass pagecache...
5358 if (EXT4_SB(sb)->s_journal &&
5359 ext4_should_journal_data(d_inode(path->dentry))) {
5361 * We don't need to lock updates but journal_flush() could
5362 * otherwise be livelocked...
5364 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
5365 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
5366 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
5371 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
5372 err = dquot_quota_on(sb, type, format_id, path);
5374 lockdep_set_quota_inode(path->dentry->d_inode,
5377 struct inode *inode = d_inode(path->dentry);
5381 * Set inode flags to prevent userspace from messing with quota
5382 * files. If this fails, we return success anyway since quotas
5383 * are already enabled and this is not a hard failure.
5386 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5389 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL;
5390 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
5391 S_NOATIME | S_IMMUTABLE);
5392 ext4_mark_inode_dirty(handle, inode);
5393 ext4_journal_stop(handle);
5395 inode_unlock(inode);
5400 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5404 struct inode *qf_inode;
5405 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5406 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5407 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5408 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
5411 BUG_ON(!ext4_has_feature_quota(sb));
5413 if (!qf_inums[type])
5416 qf_inode = ext4_iget(sb, qf_inums[type]);
5417 if (IS_ERR(qf_inode)) {
5418 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5419 return PTR_ERR(qf_inode);
5422 /* Don't account quota for quota files to avoid recursion */
5423 qf_inode->i_flags |= S_NOQUOTA;
5424 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
5425 err = dquot_enable(qf_inode, type, format_id, flags);
5428 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
5433 /* Enable usage tracking for all quota types. */
5434 static int ext4_enable_quotas(struct super_block *sb)
5437 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5438 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5439 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5440 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
5442 bool quota_mopt[EXT4_MAXQUOTAS] = {
5443 test_opt(sb, USRQUOTA),
5444 test_opt(sb, GRPQUOTA),
5445 test_opt(sb, PRJQUOTA),
5448 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5449 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
5450 if (qf_inums[type]) {
5451 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5452 DQUOT_USAGE_ENABLED |
5453 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
5456 "Failed to enable quota tracking "
5457 "(type=%d, err=%d). Please run "
5458 "e2fsck to fix.", type, err);
5466 static int ext4_quota_off(struct super_block *sb, int type)
5468 struct inode *inode = sb_dqopt(sb)->files[type];
5472 /* Force all delayed allocation blocks to be allocated.
5473 * Caller already holds s_umount sem */
5474 if (test_opt(sb, DELALLOC))
5475 sync_filesystem(sb);
5477 if (!inode || !igrab(inode))
5480 err = dquot_quota_off(sb, type);
5486 * Update modification times of quota files when userspace can
5487 * start looking at them. If we fail, we return success anyway since
5488 * this is not a hard failure and quotas are already disabled.
5490 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5493 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL);
5494 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
5495 inode->i_mtime = inode->i_ctime = current_time(inode);
5496 ext4_mark_inode_dirty(handle, inode);
5497 ext4_journal_stop(handle);
5499 inode_unlock(inode);
5504 return dquot_quota_off(sb, type);
5507 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5508 * acquiring the locks... As quota files are never truncated and quota code
5509 * itself serializes the operations (and no one else should touch the files)
5510 * we don't have to be afraid of races */
5511 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5512 size_t len, loff_t off)
5514 struct inode *inode = sb_dqopt(sb)->files[type];
5515 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5516 int offset = off & (sb->s_blocksize - 1);
5519 struct buffer_head *bh;
5520 loff_t i_size = i_size_read(inode);
5524 if (off+len > i_size)
5527 while (toread > 0) {
5528 tocopy = sb->s_blocksize - offset < toread ?
5529 sb->s_blocksize - offset : toread;
5530 bh = ext4_bread(NULL, inode, blk, 0);
5533 if (!bh) /* A hole? */
5534 memset(data, 0, tocopy);
5536 memcpy(data, bh->b_data+offset, tocopy);
5546 /* Write to quotafile (we know the transaction is already started and has
5547 * enough credits) */
5548 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5549 const char *data, size_t len, loff_t off)
5551 struct inode *inode = sb_dqopt(sb)->files[type];
5552 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5553 int err, offset = off & (sb->s_blocksize - 1);
5555 struct buffer_head *bh;
5556 handle_t *handle = journal_current_handle();
5558 if (EXT4_SB(sb)->s_journal && !handle) {
5559 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5560 " cancelled because transaction is not started",
5561 (unsigned long long)off, (unsigned long long)len);
5565 * Since we account only one data block in transaction credits,
5566 * then it is impossible to cross a block boundary.
5568 if (sb->s_blocksize - offset < len) {
5569 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5570 " cancelled because not block aligned",
5571 (unsigned long long)off, (unsigned long long)len);
5576 bh = ext4_bread(handle, inode, blk,
5577 EXT4_GET_BLOCKS_CREATE |
5578 EXT4_GET_BLOCKS_METADATA_NOFAIL);
5579 } while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) &&
5580 ext4_should_retry_alloc(inode->i_sb, &retries));
5585 BUFFER_TRACE(bh, "get write access");
5586 err = ext4_journal_get_write_access(handle, bh);
5592 memcpy(bh->b_data+offset, data, len);
5593 flush_dcache_page(bh->b_page);
5595 err = ext4_handle_dirty_metadata(handle, NULL, bh);
5598 if (inode->i_size < off + len) {
5599 i_size_write(inode, off + len);
5600 EXT4_I(inode)->i_disksize = inode->i_size;
5601 ext4_mark_inode_dirty(handle, inode);
5606 static int ext4_get_next_id(struct super_block *sb, struct kqid *qid)
5608 const struct quota_format_ops *ops;
5610 if (!sb_has_quota_loaded(sb, qid->type))
5612 ops = sb_dqopt(sb)->ops[qid->type];
5613 if (!ops || !ops->get_next_id)
5615 return dquot_get_next_id(sb, qid);
5619 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5620 const char *dev_name, void *data)
5622 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5625 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
5626 static inline void register_as_ext2(void)
5628 int err = register_filesystem(&ext2_fs_type);
5631 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5634 static inline void unregister_as_ext2(void)
5636 unregister_filesystem(&ext2_fs_type);
5639 static inline int ext2_feature_set_ok(struct super_block *sb)
5641 if (ext4_has_unknown_ext2_incompat_features(sb))
5643 if (sb->s_flags & MS_RDONLY)
5645 if (ext4_has_unknown_ext2_ro_compat_features(sb))
5650 static inline void register_as_ext2(void) { }
5651 static inline void unregister_as_ext2(void) { }
5652 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5655 static inline void register_as_ext3(void)
5657 int err = register_filesystem(&ext3_fs_type);
5660 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5663 static inline void unregister_as_ext3(void)
5665 unregister_filesystem(&ext3_fs_type);
5668 static inline int ext3_feature_set_ok(struct super_block *sb)
5670 if (ext4_has_unknown_ext3_incompat_features(sb))
5672 if (!ext4_has_feature_journal(sb))
5674 if (sb->s_flags & MS_RDONLY)
5676 if (ext4_has_unknown_ext3_ro_compat_features(sb))
5681 static struct file_system_type ext4_fs_type = {
5682 .owner = THIS_MODULE,
5684 .mount = ext4_mount,
5685 .kill_sb = kill_block_super,
5686 .fs_flags = FS_REQUIRES_DEV,
5688 MODULE_ALIAS_FS("ext4");
5690 /* Shared across all ext4 file systems */
5691 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5693 static int __init ext4_init_fs(void)
5697 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
5698 ext4_li_info = NULL;
5699 mutex_init(&ext4_li_mtx);
5701 /* Build-time check for flags consistency */
5702 ext4_check_flag_values();
5704 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
5705 init_waitqueue_head(&ext4__ioend_wq[i]);
5707 err = ext4_init_es();
5711 err = ext4_init_pageio();
5715 err = ext4_init_system_zone();
5719 err = ext4_init_sysfs();
5723 err = ext4_init_mballoc();
5726 err = init_inodecache();
5731 err = register_filesystem(&ext4_fs_type);
5737 unregister_as_ext2();
5738 unregister_as_ext3();
5739 destroy_inodecache();
5741 ext4_exit_mballoc();
5745 ext4_exit_system_zone();
5754 static void __exit ext4_exit_fs(void)
5756 ext4_destroy_lazyinit_thread();
5757 unregister_as_ext2();
5758 unregister_as_ext3();
5759 unregister_filesystem(&ext4_fs_type);
5760 destroy_inodecache();
5761 ext4_exit_mballoc();
5763 ext4_exit_system_zone();
5768 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5769 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5770 MODULE_LICENSE("GPL");
5771 module_init(ext4_init_fs)
5772 module_exit(ext4_exit_fs)