2 * linux/fs/ext4/super.c
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
11 * linux/fs/minix/inode.c
13 * Copyright (C) 1991, 1992 Linus Torvalds
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
19 #include <linux/module.h>
20 #include <linux/string.h>
22 #include <linux/time.h>
23 #include <linux/vmalloc.h>
24 #include <linux/slab.h>
25 #include <linux/init.h>
26 #include <linux/blkdev.h>
27 #include <linux/backing-dev.h>
28 #include <linux/parser.h>
29 #include <linux/buffer_head.h>
30 #include <linux/exportfs.h>
31 #include <linux/vfs.h>
32 #include <linux/random.h>
33 #include <linux/mount.h>
34 #include <linux/namei.h>
35 #include <linux/quotaops.h>
36 #include <linux/seq_file.h>
37 #include <linux/ctype.h>
38 #include <linux/log2.h>
39 #include <linux/crc16.h>
40 #include <linux/cleancache.h>
41 #include <asm/uaccess.h>
43 #include <linux/kthread.h>
44 #include <linux/freezer.h>
47 #include "ext4_extents.h" /* Needed for trace points definition */
48 #include "ext4_jbd2.h"
53 #define CREATE_TRACE_POINTS
54 #include <trace/events/ext4.h>
56 static struct ext4_lazy_init *ext4_li_info;
57 static struct mutex ext4_li_mtx;
58 static struct ratelimit_state ext4_mount_msg_ratelimit;
60 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
61 unsigned long journal_devnum);
62 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
63 static int ext4_commit_super(struct super_block *sb, int sync);
64 static void ext4_mark_recovery_complete(struct super_block *sb,
65 struct ext4_super_block *es);
66 static void ext4_clear_journal_err(struct super_block *sb,
67 struct ext4_super_block *es);
68 static int ext4_sync_fs(struct super_block *sb, int wait);
69 static int ext4_remount(struct super_block *sb, int *flags, char *data);
70 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
71 static int ext4_unfreeze(struct super_block *sb);
72 static int ext4_freeze(struct super_block *sb);
73 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
74 const char *dev_name, void *data);
75 static inline int ext2_feature_set_ok(struct super_block *sb);
76 static inline int ext3_feature_set_ok(struct super_block *sb);
77 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
78 static void ext4_destroy_lazyinit_thread(void);
79 static void ext4_unregister_li_request(struct super_block *sb);
80 static void ext4_clear_request_list(void);
81 static struct inode *ext4_get_journal_inode(struct super_block *sb,
82 unsigned int journal_inum);
87 * Note the difference between i_mmap_sem (EXT4_I(inode)->i_mmap_sem) and
88 * i_mmap_rwsem (inode->i_mmap_rwsem)!
91 * mmap_sem -> sb_start_pagefault -> i_mmap_sem (r) -> transaction start ->
92 * page lock -> i_data_sem (rw)
94 * buffered write path:
95 * sb_start_write -> i_mutex -> mmap_sem
96 * sb_start_write -> i_mutex -> transaction start -> page lock ->
100 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) ->
101 * i_mmap_rwsem (w) -> page lock
102 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) ->
103 * transaction start -> i_data_sem (rw)
106 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) -> mmap_sem
107 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) ->
108 * transaction start -> i_data_sem (rw)
111 * transaction start -> page lock(s) -> i_data_sem (rw)
114 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
115 static struct file_system_type ext2_fs_type = {
116 .owner = THIS_MODULE,
119 .kill_sb = kill_block_super,
120 .fs_flags = FS_REQUIRES_DEV,
122 MODULE_ALIAS_FS("ext2");
123 MODULE_ALIAS("ext2");
124 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
126 #define IS_EXT2_SB(sb) (0)
130 static struct file_system_type ext3_fs_type = {
131 .owner = THIS_MODULE,
134 .kill_sb = kill_block_super,
135 .fs_flags = FS_REQUIRES_DEV,
137 MODULE_ALIAS_FS("ext3");
138 MODULE_ALIAS("ext3");
139 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
141 static int ext4_verify_csum_type(struct super_block *sb,
142 struct ext4_super_block *es)
144 if (!ext4_has_feature_metadata_csum(sb))
147 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
150 static __le32 ext4_superblock_csum(struct super_block *sb,
151 struct ext4_super_block *es)
153 struct ext4_sb_info *sbi = EXT4_SB(sb);
154 int offset = offsetof(struct ext4_super_block, s_checksum);
157 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
159 return cpu_to_le32(csum);
162 static int ext4_superblock_csum_verify(struct super_block *sb,
163 struct ext4_super_block *es)
165 if (!ext4_has_metadata_csum(sb))
168 return es->s_checksum == ext4_superblock_csum(sb, es);
171 void ext4_superblock_csum_set(struct super_block *sb)
173 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
175 if (!ext4_has_metadata_csum(sb))
178 es->s_checksum = ext4_superblock_csum(sb, es);
181 void *ext4_kvmalloc(size_t size, gfp_t flags)
185 ret = kmalloc(size, flags | __GFP_NOWARN);
187 ret = __vmalloc(size, flags, PAGE_KERNEL);
191 void *ext4_kvzalloc(size_t size, gfp_t flags)
195 ret = kzalloc(size, flags | __GFP_NOWARN);
197 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
201 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
202 struct ext4_group_desc *bg)
204 return le32_to_cpu(bg->bg_block_bitmap_lo) |
205 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
206 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
209 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
210 struct ext4_group_desc *bg)
212 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
213 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
214 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
217 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
218 struct ext4_group_desc *bg)
220 return le32_to_cpu(bg->bg_inode_table_lo) |
221 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
222 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
225 __u32 ext4_free_group_clusters(struct super_block *sb,
226 struct ext4_group_desc *bg)
228 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
229 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
230 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
233 __u32 ext4_free_inodes_count(struct super_block *sb,
234 struct ext4_group_desc *bg)
236 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
237 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
238 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
241 __u32 ext4_used_dirs_count(struct super_block *sb,
242 struct ext4_group_desc *bg)
244 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
245 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
246 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
249 __u32 ext4_itable_unused_count(struct super_block *sb,
250 struct ext4_group_desc *bg)
252 return le16_to_cpu(bg->bg_itable_unused_lo) |
253 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
254 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
257 void ext4_block_bitmap_set(struct super_block *sb,
258 struct ext4_group_desc *bg, ext4_fsblk_t blk)
260 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
261 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
262 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
265 void ext4_inode_bitmap_set(struct super_block *sb,
266 struct ext4_group_desc *bg, ext4_fsblk_t blk)
268 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
269 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
270 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
273 void ext4_inode_table_set(struct super_block *sb,
274 struct ext4_group_desc *bg, ext4_fsblk_t blk)
276 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
277 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
278 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
281 void ext4_free_group_clusters_set(struct super_block *sb,
282 struct ext4_group_desc *bg, __u32 count)
284 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
285 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
286 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
289 void ext4_free_inodes_set(struct super_block *sb,
290 struct ext4_group_desc *bg, __u32 count)
292 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
293 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
294 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
297 void ext4_used_dirs_set(struct super_block *sb,
298 struct ext4_group_desc *bg, __u32 count)
300 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
301 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
302 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
305 void ext4_itable_unused_set(struct super_block *sb,
306 struct ext4_group_desc *bg, __u32 count)
308 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
309 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
310 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
314 static void __save_error_info(struct super_block *sb, const char *func,
317 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
319 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
320 if (bdev_read_only(sb->s_bdev))
322 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
323 es->s_last_error_time = cpu_to_le32(get_seconds());
324 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
325 es->s_last_error_line = cpu_to_le32(line);
326 if (!es->s_first_error_time) {
327 es->s_first_error_time = es->s_last_error_time;
328 strncpy(es->s_first_error_func, func,
329 sizeof(es->s_first_error_func));
330 es->s_first_error_line = cpu_to_le32(line);
331 es->s_first_error_ino = es->s_last_error_ino;
332 es->s_first_error_block = es->s_last_error_block;
335 * Start the daily error reporting function if it hasn't been
338 if (!es->s_error_count)
339 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
340 le32_add_cpu(&es->s_error_count, 1);
343 static void save_error_info(struct super_block *sb, const char *func,
346 __save_error_info(sb, func, line);
347 ext4_commit_super(sb, 1);
351 * The del_gendisk() function uninitializes the disk-specific data
352 * structures, including the bdi structure, without telling anyone
353 * else. Once this happens, any attempt to call mark_buffer_dirty()
354 * (for example, by ext4_commit_super), will cause a kernel OOPS.
355 * This is a kludge to prevent these oops until we can put in a proper
356 * hook in del_gendisk() to inform the VFS and file system layers.
358 static int block_device_ejected(struct super_block *sb)
360 struct inode *bd_inode = sb->s_bdev->bd_inode;
361 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
363 return bdi->dev == NULL;
366 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
368 struct super_block *sb = journal->j_private;
369 struct ext4_sb_info *sbi = EXT4_SB(sb);
370 int error = is_journal_aborted(journal);
371 struct ext4_journal_cb_entry *jce;
373 BUG_ON(txn->t_state == T_FINISHED);
374 spin_lock(&sbi->s_md_lock);
375 while (!list_empty(&txn->t_private_list)) {
376 jce = list_entry(txn->t_private_list.next,
377 struct ext4_journal_cb_entry, jce_list);
378 list_del_init(&jce->jce_list);
379 spin_unlock(&sbi->s_md_lock);
380 jce->jce_func(sb, jce, error);
381 spin_lock(&sbi->s_md_lock);
383 spin_unlock(&sbi->s_md_lock);
386 /* Deal with the reporting of failure conditions on a filesystem such as
387 * inconsistencies detected or read IO failures.
389 * On ext2, we can store the error state of the filesystem in the
390 * superblock. That is not possible on ext4, because we may have other
391 * write ordering constraints on the superblock which prevent us from
392 * writing it out straight away; and given that the journal is about to
393 * be aborted, we can't rely on the current, or future, transactions to
394 * write out the superblock safely.
396 * We'll just use the jbd2_journal_abort() error code to record an error in
397 * the journal instead. On recovery, the journal will complain about
398 * that error until we've noted it down and cleared it.
401 static void ext4_handle_error(struct super_block *sb)
403 if (sb->s_flags & MS_RDONLY)
406 if (!test_opt(sb, ERRORS_CONT)) {
407 journal_t *journal = EXT4_SB(sb)->s_journal;
409 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
411 jbd2_journal_abort(journal, -EIO);
413 if (test_opt(sb, ERRORS_RO)) {
414 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
416 * Make sure updated value of ->s_mount_flags will be visible
417 * before ->s_flags update
420 sb->s_flags |= MS_RDONLY;
422 if (test_opt(sb, ERRORS_PANIC)) {
423 if (EXT4_SB(sb)->s_journal &&
424 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
426 panic("EXT4-fs (device %s): panic forced after error\n",
431 #define ext4_error_ratelimit(sb) \
432 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
435 void __ext4_error(struct super_block *sb, const char *function,
436 unsigned int line, const char *fmt, ...)
438 struct va_format vaf;
441 if (ext4_error_ratelimit(sb)) {
446 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
447 sb->s_id, function, line, current->comm, &vaf);
450 save_error_info(sb, function, line);
451 ext4_handle_error(sb);
454 void __ext4_error_inode(struct inode *inode, const char *function,
455 unsigned int line, ext4_fsblk_t block,
456 const char *fmt, ...)
459 struct va_format vaf;
460 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
462 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
463 es->s_last_error_block = cpu_to_le64(block);
464 if (ext4_error_ratelimit(inode->i_sb)) {
469 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
470 "inode #%lu: block %llu: comm %s: %pV\n",
471 inode->i_sb->s_id, function, line, inode->i_ino,
472 block, current->comm, &vaf);
474 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
475 "inode #%lu: comm %s: %pV\n",
476 inode->i_sb->s_id, function, line, inode->i_ino,
477 current->comm, &vaf);
480 save_error_info(inode->i_sb, function, line);
481 ext4_handle_error(inode->i_sb);
484 void __ext4_error_file(struct file *file, const char *function,
485 unsigned int line, ext4_fsblk_t block,
486 const char *fmt, ...)
489 struct va_format vaf;
490 struct ext4_super_block *es;
491 struct inode *inode = file_inode(file);
492 char pathname[80], *path;
494 es = EXT4_SB(inode->i_sb)->s_es;
495 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
496 if (ext4_error_ratelimit(inode->i_sb)) {
497 path = file_path(file, pathname, sizeof(pathname));
505 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
506 "block %llu: comm %s: path %s: %pV\n",
507 inode->i_sb->s_id, function, line, inode->i_ino,
508 block, current->comm, path, &vaf);
511 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
512 "comm %s: path %s: %pV\n",
513 inode->i_sb->s_id, function, line, inode->i_ino,
514 current->comm, path, &vaf);
517 save_error_info(inode->i_sb, function, line);
518 ext4_handle_error(inode->i_sb);
521 const char *ext4_decode_error(struct super_block *sb, int errno,
528 errstr = "Corrupt filesystem";
531 errstr = "Filesystem failed CRC";
534 errstr = "IO failure";
537 errstr = "Out of memory";
540 if (!sb || (EXT4_SB(sb)->s_journal &&
541 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
542 errstr = "Journal has aborted";
544 errstr = "Readonly filesystem";
547 /* If the caller passed in an extra buffer for unknown
548 * errors, textualise them now. Else we just return
551 /* Check for truncated error codes... */
552 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
561 /* __ext4_std_error decodes expected errors from journaling functions
562 * automatically and invokes the appropriate error response. */
564 void __ext4_std_error(struct super_block *sb, const char *function,
565 unsigned int line, int errno)
570 /* Special case: if the error is EROFS, and we're not already
571 * inside a transaction, then there's really no point in logging
573 if (errno == -EROFS && journal_current_handle() == NULL &&
574 (sb->s_flags & MS_RDONLY))
577 if (ext4_error_ratelimit(sb)) {
578 errstr = ext4_decode_error(sb, errno, nbuf);
579 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
580 sb->s_id, function, line, errstr);
583 save_error_info(sb, function, line);
584 ext4_handle_error(sb);
588 * ext4_abort is a much stronger failure handler than ext4_error. The
589 * abort function may be used to deal with unrecoverable failures such
590 * as journal IO errors or ENOMEM at a critical moment in log management.
592 * We unconditionally force the filesystem into an ABORT|READONLY state,
593 * unless the error response on the fs has been set to panic in which
594 * case we take the easy way out and panic immediately.
597 void __ext4_abort(struct super_block *sb, const char *function,
598 unsigned int line, const char *fmt, ...)
600 struct va_format vaf;
603 save_error_info(sb, function, line);
607 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: %pV\n",
608 sb->s_id, function, line, &vaf);
611 if ((sb->s_flags & MS_RDONLY) == 0) {
612 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
613 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
615 * Make sure updated value of ->s_mount_flags will be visible
616 * before ->s_flags update
619 sb->s_flags |= MS_RDONLY;
620 if (EXT4_SB(sb)->s_journal)
621 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
622 save_error_info(sb, function, line);
624 if (test_opt(sb, ERRORS_PANIC)) {
625 if (EXT4_SB(sb)->s_journal &&
626 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
628 panic("EXT4-fs panic from previous error\n");
632 void __ext4_msg(struct super_block *sb,
633 const char *prefix, const char *fmt, ...)
635 struct va_format vaf;
638 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
644 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
648 #define ext4_warning_ratelimit(sb) \
649 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \
652 void __ext4_warning(struct super_block *sb, const char *function,
653 unsigned int line, const char *fmt, ...)
655 struct va_format vaf;
658 if (!ext4_warning_ratelimit(sb))
664 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
665 sb->s_id, function, line, &vaf);
669 void __ext4_warning_inode(const struct inode *inode, const char *function,
670 unsigned int line, const char *fmt, ...)
672 struct va_format vaf;
675 if (!ext4_warning_ratelimit(inode->i_sb))
681 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
682 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
683 function, line, inode->i_ino, current->comm, &vaf);
687 void __ext4_grp_locked_error(const char *function, unsigned int line,
688 struct super_block *sb, ext4_group_t grp,
689 unsigned long ino, ext4_fsblk_t block,
690 const char *fmt, ...)
694 struct va_format vaf;
696 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
698 es->s_last_error_ino = cpu_to_le32(ino);
699 es->s_last_error_block = cpu_to_le64(block);
700 __save_error_info(sb, function, line);
702 if (ext4_error_ratelimit(sb)) {
706 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
707 sb->s_id, function, line, grp);
709 printk(KERN_CONT "inode %lu: ", ino);
711 printk(KERN_CONT "block %llu:",
712 (unsigned long long) block);
713 printk(KERN_CONT "%pV\n", &vaf);
717 if (test_opt(sb, ERRORS_CONT)) {
718 ext4_commit_super(sb, 0);
722 ext4_unlock_group(sb, grp);
723 ext4_commit_super(sb, 1);
724 ext4_handle_error(sb);
726 * We only get here in the ERRORS_RO case; relocking the group
727 * may be dangerous, but nothing bad will happen since the
728 * filesystem will have already been marked read/only and the
729 * journal has been aborted. We return 1 as a hint to callers
730 * who might what to use the return value from
731 * ext4_grp_locked_error() to distinguish between the
732 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
733 * aggressively from the ext4 function in question, with a
734 * more appropriate error code.
736 ext4_lock_group(sb, grp);
740 void ext4_update_dynamic_rev(struct super_block *sb)
742 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
744 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
748 "updating to rev %d because of new feature flag, "
749 "running e2fsck is recommended",
752 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
753 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
754 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
755 /* leave es->s_feature_*compat flags alone */
756 /* es->s_uuid will be set by e2fsck if empty */
759 * The rest of the superblock fields should be zero, and if not it
760 * means they are likely already in use, so leave them alone. We
761 * can leave it up to e2fsck to clean up any inconsistencies there.
766 * Open the external journal device
768 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
770 struct block_device *bdev;
771 char b[BDEVNAME_SIZE];
773 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
779 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
780 __bdevname(dev, b), PTR_ERR(bdev));
785 * Release the journal device
787 static void ext4_blkdev_put(struct block_device *bdev)
789 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
792 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
794 struct block_device *bdev;
795 bdev = sbi->journal_bdev;
797 ext4_blkdev_put(bdev);
798 sbi->journal_bdev = NULL;
802 static inline struct inode *orphan_list_entry(struct list_head *l)
804 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
807 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
811 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
812 le32_to_cpu(sbi->s_es->s_last_orphan));
814 printk(KERN_ERR "sb_info orphan list:\n");
815 list_for_each(l, &sbi->s_orphan) {
816 struct inode *inode = orphan_list_entry(l);
818 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
819 inode->i_sb->s_id, inode->i_ino, inode,
820 inode->i_mode, inode->i_nlink,
825 static void ext4_put_super(struct super_block *sb)
827 struct ext4_sb_info *sbi = EXT4_SB(sb);
828 struct ext4_super_block *es = sbi->s_es;
832 ext4_unregister_li_request(sb);
833 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
835 flush_workqueue(sbi->rsv_conversion_wq);
836 destroy_workqueue(sbi->rsv_conversion_wq);
838 if (sbi->s_journal) {
839 aborted = is_journal_aborted(sbi->s_journal);
840 err = jbd2_journal_destroy(sbi->s_journal);
841 sbi->s_journal = NULL;
842 if ((err < 0) && !aborted)
843 ext4_abort(sb, "Couldn't clean up the journal");
846 ext4_unregister_sysfs(sb);
847 ext4_es_unregister_shrinker(sbi);
848 del_timer_sync(&sbi->s_err_report);
849 ext4_release_system_zone(sb);
851 ext4_ext_release(sb);
853 if (!(sb->s_flags & MS_RDONLY) && !aborted) {
854 ext4_clear_feature_journal_needs_recovery(sb);
855 es->s_state = cpu_to_le16(sbi->s_mount_state);
857 if (!(sb->s_flags & MS_RDONLY))
858 ext4_commit_super(sb, 1);
860 for (i = 0; i < sbi->s_gdb_count; i++)
861 brelse(sbi->s_group_desc[i]);
862 kvfree(sbi->s_group_desc);
863 kvfree(sbi->s_flex_groups);
864 percpu_counter_destroy(&sbi->s_freeclusters_counter);
865 percpu_counter_destroy(&sbi->s_freeinodes_counter);
866 percpu_counter_destroy(&sbi->s_dirs_counter);
867 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
868 percpu_free_rwsem(&sbi->s_journal_flag_rwsem);
871 for (i = 0; i < EXT4_MAXQUOTAS; i++)
872 kfree(sbi->s_qf_names[i]);
875 /* Debugging code just in case the in-memory inode orphan list
876 * isn't empty. The on-disk one can be non-empty if we've
877 * detected an error and taken the fs readonly, but the
878 * in-memory list had better be clean by this point. */
879 if (!list_empty(&sbi->s_orphan))
880 dump_orphan_list(sb, sbi);
881 J_ASSERT(list_empty(&sbi->s_orphan));
883 sync_blockdev(sb->s_bdev);
884 invalidate_bdev(sb->s_bdev);
885 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
887 * Invalidate the journal device's buffers. We don't want them
888 * floating about in memory - the physical journal device may
889 * hotswapped, and it breaks the `ro-after' testing code.
891 sync_blockdev(sbi->journal_bdev);
892 invalidate_bdev(sbi->journal_bdev);
893 ext4_blkdev_remove(sbi);
895 if (sbi->s_mb_cache) {
896 ext4_xattr_destroy_cache(sbi->s_mb_cache);
897 sbi->s_mb_cache = NULL;
900 kthread_stop(sbi->s_mmp_tsk);
901 sb->s_fs_info = NULL;
903 * Now that we are completely done shutting down the
904 * superblock, we need to actually destroy the kobject.
906 kobject_put(&sbi->s_kobj);
907 wait_for_completion(&sbi->s_kobj_unregister);
908 if (sbi->s_chksum_driver)
909 crypto_free_shash(sbi->s_chksum_driver);
910 kfree(sbi->s_blockgroup_lock);
914 static struct kmem_cache *ext4_inode_cachep;
917 * Called inside transaction, so use GFP_NOFS
919 static struct inode *ext4_alloc_inode(struct super_block *sb)
921 struct ext4_inode_info *ei;
923 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
927 ei->vfs_inode.i_version = 1;
928 spin_lock_init(&ei->i_raw_lock);
929 INIT_LIST_HEAD(&ei->i_prealloc_list);
930 spin_lock_init(&ei->i_prealloc_lock);
931 ext4_es_init_tree(&ei->i_es_tree);
932 rwlock_init(&ei->i_es_lock);
933 INIT_LIST_HEAD(&ei->i_es_list);
936 ei->i_es_shrink_lblk = 0;
937 ei->i_reserved_data_blocks = 0;
938 ei->i_reserved_meta_blocks = 0;
939 ei->i_allocated_meta_blocks = 0;
940 ei->i_da_metadata_calc_len = 0;
941 ei->i_da_metadata_calc_last_lblock = 0;
942 spin_lock_init(&(ei->i_block_reservation_lock));
944 ei->i_reserved_quota = 0;
945 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
948 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
949 spin_lock_init(&ei->i_completed_io_lock);
951 ei->i_datasync_tid = 0;
952 atomic_set(&ei->i_unwritten, 0);
953 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
954 return &ei->vfs_inode;
957 static int ext4_drop_inode(struct inode *inode)
959 int drop = generic_drop_inode(inode);
961 trace_ext4_drop_inode(inode, drop);
965 static void ext4_i_callback(struct rcu_head *head)
967 struct inode *inode = container_of(head, struct inode, i_rcu);
968 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
971 static void ext4_destroy_inode(struct inode *inode)
973 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
974 ext4_msg(inode->i_sb, KERN_ERR,
975 "Inode %lu (%p): orphan list check failed!",
976 inode->i_ino, EXT4_I(inode));
977 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
978 EXT4_I(inode), sizeof(struct ext4_inode_info),
982 call_rcu(&inode->i_rcu, ext4_i_callback);
985 static void init_once(void *foo)
987 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
989 INIT_LIST_HEAD(&ei->i_orphan);
990 init_rwsem(&ei->xattr_sem);
991 init_rwsem(&ei->i_data_sem);
992 init_rwsem(&ei->i_mmap_sem);
993 inode_init_once(&ei->vfs_inode);
996 static int __init init_inodecache(void)
998 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
999 sizeof(struct ext4_inode_info),
1000 0, (SLAB_RECLAIM_ACCOUNT|
1001 SLAB_MEM_SPREAD|SLAB_ACCOUNT),
1003 if (ext4_inode_cachep == NULL)
1008 static void destroy_inodecache(void)
1011 * Make sure all delayed rcu free inodes are flushed before we
1015 kmem_cache_destroy(ext4_inode_cachep);
1018 void ext4_clear_inode(struct inode *inode)
1020 invalidate_inode_buffers(inode);
1023 ext4_discard_preallocations(inode);
1024 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
1025 if (EXT4_I(inode)->jinode) {
1026 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1027 EXT4_I(inode)->jinode);
1028 jbd2_free_inode(EXT4_I(inode)->jinode);
1029 EXT4_I(inode)->jinode = NULL;
1031 fscrypt_put_encryption_info(inode);
1034 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1035 u64 ino, u32 generation)
1037 struct inode *inode;
1039 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1040 return ERR_PTR(-ESTALE);
1041 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1042 return ERR_PTR(-ESTALE);
1044 /* iget isn't really right if the inode is currently unallocated!!
1046 * ext4_read_inode will return a bad_inode if the inode had been
1047 * deleted, so we should be safe.
1049 * Currently we don't know the generation for parent directory, so
1050 * a generation of 0 means "accept any"
1052 inode = ext4_iget_normal(sb, ino);
1054 return ERR_CAST(inode);
1055 if (generation && inode->i_generation != generation) {
1057 return ERR_PTR(-ESTALE);
1063 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1064 int fh_len, int fh_type)
1066 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1067 ext4_nfs_get_inode);
1070 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1071 int fh_len, int fh_type)
1073 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1074 ext4_nfs_get_inode);
1078 * Try to release metadata pages (indirect blocks, directories) which are
1079 * mapped via the block device. Since these pages could have journal heads
1080 * which would prevent try_to_free_buffers() from freeing them, we must use
1081 * jbd2 layer's try_to_free_buffers() function to release them.
1083 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1086 journal_t *journal = EXT4_SB(sb)->s_journal;
1088 WARN_ON(PageChecked(page));
1089 if (!page_has_buffers(page))
1092 return jbd2_journal_try_to_free_buffers(journal, page,
1093 wait & ~__GFP_DIRECT_RECLAIM);
1094 return try_to_free_buffers(page);
1097 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1098 static int ext4_get_context(struct inode *inode, void *ctx, size_t len)
1100 return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1101 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len);
1104 static int ext4_set_context(struct inode *inode, const void *ctx, size_t len,
1107 handle_t *handle = fs_data;
1108 int res, res2, retries = 0;
1110 res = ext4_convert_inline_data(inode);
1115 * If a journal handle was specified, then the encryption context is
1116 * being set on a new inode via inheritance and is part of a larger
1117 * transaction to create the inode. Otherwise the encryption context is
1118 * being set on an existing inode in its own transaction. Only in the
1119 * latter case should the "retry on ENOSPC" logic be used.
1123 res = ext4_xattr_set_handle(handle, inode,
1124 EXT4_XATTR_INDEX_ENCRYPTION,
1125 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1128 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1129 ext4_clear_inode_state(inode,
1130 EXT4_STATE_MAY_INLINE_DATA);
1132 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1133 * S_DAX may be disabled
1135 ext4_set_inode_flags(inode);
1140 res = dquot_initialize(inode);
1144 handle = ext4_journal_start(inode, EXT4_HT_MISC,
1145 ext4_jbd2_credits_xattr(inode));
1147 return PTR_ERR(handle);
1149 res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION,
1150 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1153 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1155 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1156 * S_DAX may be disabled
1158 ext4_set_inode_flags(inode);
1159 res = ext4_mark_inode_dirty(handle, inode);
1161 EXT4_ERROR_INODE(inode, "Failed to mark inode dirty");
1163 res2 = ext4_journal_stop(handle);
1165 if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
1172 static bool ext4_dummy_context(struct inode *inode)
1174 return DUMMY_ENCRYPTION_ENABLED(EXT4_SB(inode->i_sb));
1177 static const struct fscrypt_operations ext4_cryptops = {
1178 .key_prefix = "ext4:",
1179 .get_context = ext4_get_context,
1180 .set_context = ext4_set_context,
1181 .dummy_context = ext4_dummy_context,
1182 .empty_dir = ext4_empty_dir,
1183 .max_namelen = EXT4_NAME_LEN,
1188 static char *quotatypes[] = INITQFNAMES;
1189 #define QTYPE2NAME(t) (quotatypes[t])
1191 static int ext4_write_dquot(struct dquot *dquot);
1192 static int ext4_acquire_dquot(struct dquot *dquot);
1193 static int ext4_release_dquot(struct dquot *dquot);
1194 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1195 static int ext4_write_info(struct super_block *sb, int type);
1196 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1198 static int ext4_quota_off(struct super_block *sb, int type);
1199 static int ext4_quota_on_mount(struct super_block *sb, int type);
1200 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1201 size_t len, loff_t off);
1202 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1203 const char *data, size_t len, loff_t off);
1204 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1205 unsigned int flags);
1206 static int ext4_enable_quotas(struct super_block *sb);
1207 static int ext4_get_next_id(struct super_block *sb, struct kqid *qid);
1209 static struct dquot **ext4_get_dquots(struct inode *inode)
1211 return EXT4_I(inode)->i_dquot;
1214 static const struct dquot_operations ext4_quota_operations = {
1215 .get_reserved_space = ext4_get_reserved_space,
1216 .write_dquot = ext4_write_dquot,
1217 .acquire_dquot = ext4_acquire_dquot,
1218 .release_dquot = ext4_release_dquot,
1219 .mark_dirty = ext4_mark_dquot_dirty,
1220 .write_info = ext4_write_info,
1221 .alloc_dquot = dquot_alloc,
1222 .destroy_dquot = dquot_destroy,
1223 .get_projid = ext4_get_projid,
1224 .get_next_id = ext4_get_next_id,
1227 static const struct quotactl_ops ext4_qctl_operations = {
1228 .quota_on = ext4_quota_on,
1229 .quota_off = ext4_quota_off,
1230 .quota_sync = dquot_quota_sync,
1231 .get_state = dquot_get_state,
1232 .set_info = dquot_set_dqinfo,
1233 .get_dqblk = dquot_get_dqblk,
1234 .set_dqblk = dquot_set_dqblk,
1235 .get_nextdqblk = dquot_get_next_dqblk,
1239 static const struct super_operations ext4_sops = {
1240 .alloc_inode = ext4_alloc_inode,
1241 .destroy_inode = ext4_destroy_inode,
1242 .write_inode = ext4_write_inode,
1243 .dirty_inode = ext4_dirty_inode,
1244 .drop_inode = ext4_drop_inode,
1245 .evict_inode = ext4_evict_inode,
1246 .put_super = ext4_put_super,
1247 .sync_fs = ext4_sync_fs,
1248 .freeze_fs = ext4_freeze,
1249 .unfreeze_fs = ext4_unfreeze,
1250 .statfs = ext4_statfs,
1251 .remount_fs = ext4_remount,
1252 .show_options = ext4_show_options,
1254 .quota_read = ext4_quota_read,
1255 .quota_write = ext4_quota_write,
1256 .get_dquots = ext4_get_dquots,
1258 .bdev_try_to_free_page = bdev_try_to_free_page,
1261 static const struct export_operations ext4_export_ops = {
1262 .fh_to_dentry = ext4_fh_to_dentry,
1263 .fh_to_parent = ext4_fh_to_parent,
1264 .get_parent = ext4_get_parent,
1268 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1269 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1270 Opt_nouid32, Opt_debug, Opt_removed,
1271 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1272 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1273 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1274 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1275 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1276 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1277 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1278 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1279 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1280 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version, Opt_dax,
1281 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1282 Opt_lazytime, Opt_nolazytime,
1283 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1284 Opt_inode_readahead_blks, Opt_journal_ioprio,
1285 Opt_dioread_nolock, Opt_dioread_lock,
1286 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1287 Opt_max_dir_size_kb, Opt_nojournal_checksum,
1290 static const match_table_t tokens = {
1291 {Opt_bsd_df, "bsddf"},
1292 {Opt_minix_df, "minixdf"},
1293 {Opt_grpid, "grpid"},
1294 {Opt_grpid, "bsdgroups"},
1295 {Opt_nogrpid, "nogrpid"},
1296 {Opt_nogrpid, "sysvgroups"},
1297 {Opt_resgid, "resgid=%u"},
1298 {Opt_resuid, "resuid=%u"},
1300 {Opt_err_cont, "errors=continue"},
1301 {Opt_err_panic, "errors=panic"},
1302 {Opt_err_ro, "errors=remount-ro"},
1303 {Opt_nouid32, "nouid32"},
1304 {Opt_debug, "debug"},
1305 {Opt_removed, "oldalloc"},
1306 {Opt_removed, "orlov"},
1307 {Opt_user_xattr, "user_xattr"},
1308 {Opt_nouser_xattr, "nouser_xattr"},
1310 {Opt_noacl, "noacl"},
1311 {Opt_noload, "norecovery"},
1312 {Opt_noload, "noload"},
1313 {Opt_removed, "nobh"},
1314 {Opt_removed, "bh"},
1315 {Opt_commit, "commit=%u"},
1316 {Opt_min_batch_time, "min_batch_time=%u"},
1317 {Opt_max_batch_time, "max_batch_time=%u"},
1318 {Opt_journal_dev, "journal_dev=%u"},
1319 {Opt_journal_path, "journal_path=%s"},
1320 {Opt_journal_checksum, "journal_checksum"},
1321 {Opt_nojournal_checksum, "nojournal_checksum"},
1322 {Opt_journal_async_commit, "journal_async_commit"},
1323 {Opt_abort, "abort"},
1324 {Opt_data_journal, "data=journal"},
1325 {Opt_data_ordered, "data=ordered"},
1326 {Opt_data_writeback, "data=writeback"},
1327 {Opt_data_err_abort, "data_err=abort"},
1328 {Opt_data_err_ignore, "data_err=ignore"},
1329 {Opt_offusrjquota, "usrjquota="},
1330 {Opt_usrjquota, "usrjquota=%s"},
1331 {Opt_offgrpjquota, "grpjquota="},
1332 {Opt_grpjquota, "grpjquota=%s"},
1333 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1334 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1335 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1336 {Opt_grpquota, "grpquota"},
1337 {Opt_noquota, "noquota"},
1338 {Opt_quota, "quota"},
1339 {Opt_usrquota, "usrquota"},
1340 {Opt_prjquota, "prjquota"},
1341 {Opt_barrier, "barrier=%u"},
1342 {Opt_barrier, "barrier"},
1343 {Opt_nobarrier, "nobarrier"},
1344 {Opt_i_version, "i_version"},
1346 {Opt_stripe, "stripe=%u"},
1347 {Opt_delalloc, "delalloc"},
1348 {Opt_lazytime, "lazytime"},
1349 {Opt_nolazytime, "nolazytime"},
1350 {Opt_nodelalloc, "nodelalloc"},
1351 {Opt_removed, "mblk_io_submit"},
1352 {Opt_removed, "nomblk_io_submit"},
1353 {Opt_block_validity, "block_validity"},
1354 {Opt_noblock_validity, "noblock_validity"},
1355 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1356 {Opt_journal_ioprio, "journal_ioprio=%u"},
1357 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1358 {Opt_auto_da_alloc, "auto_da_alloc"},
1359 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1360 {Opt_dioread_nolock, "dioread_nolock"},
1361 {Opt_dioread_lock, "dioread_lock"},
1362 {Opt_discard, "discard"},
1363 {Opt_nodiscard, "nodiscard"},
1364 {Opt_init_itable, "init_itable=%u"},
1365 {Opt_init_itable, "init_itable"},
1366 {Opt_noinit_itable, "noinit_itable"},
1367 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1368 {Opt_test_dummy_encryption, "test_dummy_encryption"},
1369 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1370 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1371 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1372 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1373 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1377 static ext4_fsblk_t get_sb_block(void **data)
1379 ext4_fsblk_t sb_block;
1380 char *options = (char *) *data;
1382 if (!options || strncmp(options, "sb=", 3) != 0)
1383 return 1; /* Default location */
1386 /* TODO: use simple_strtoll with >32bit ext4 */
1387 sb_block = simple_strtoul(options, &options, 0);
1388 if (*options && *options != ',') {
1389 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1393 if (*options == ',')
1395 *data = (void *) options;
1400 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1401 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1402 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1405 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1407 struct ext4_sb_info *sbi = EXT4_SB(sb);
1411 if (sb_any_quota_loaded(sb) &&
1412 !sbi->s_qf_names[qtype]) {
1413 ext4_msg(sb, KERN_ERR,
1414 "Cannot change journaled "
1415 "quota options when quota turned on");
1418 if (ext4_has_feature_quota(sb)) {
1419 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1420 "ignored when QUOTA feature is enabled");
1423 qname = match_strdup(args);
1425 ext4_msg(sb, KERN_ERR,
1426 "Not enough memory for storing quotafile name");
1429 if (sbi->s_qf_names[qtype]) {
1430 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1433 ext4_msg(sb, KERN_ERR,
1434 "%s quota file already specified",
1438 if (strchr(qname, '/')) {
1439 ext4_msg(sb, KERN_ERR,
1440 "quotafile must be on filesystem root");
1443 sbi->s_qf_names[qtype] = qname;
1451 static int clear_qf_name(struct super_block *sb, int qtype)
1454 struct ext4_sb_info *sbi = EXT4_SB(sb);
1456 if (sb_any_quota_loaded(sb) &&
1457 sbi->s_qf_names[qtype]) {
1458 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1459 " when quota turned on");
1462 kfree(sbi->s_qf_names[qtype]);
1463 sbi->s_qf_names[qtype] = NULL;
1468 #define MOPT_SET 0x0001
1469 #define MOPT_CLEAR 0x0002
1470 #define MOPT_NOSUPPORT 0x0004
1471 #define MOPT_EXPLICIT 0x0008
1472 #define MOPT_CLEAR_ERR 0x0010
1473 #define MOPT_GTE0 0x0020
1476 #define MOPT_QFMT 0x0040
1478 #define MOPT_Q MOPT_NOSUPPORT
1479 #define MOPT_QFMT MOPT_NOSUPPORT
1481 #define MOPT_DATAJ 0x0080
1482 #define MOPT_NO_EXT2 0x0100
1483 #define MOPT_NO_EXT3 0x0200
1484 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1485 #define MOPT_STRING 0x0400
1487 static const struct mount_opts {
1491 } ext4_mount_opts[] = {
1492 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1493 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1494 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1495 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1496 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1497 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1498 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1499 MOPT_EXT4_ONLY | MOPT_SET},
1500 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1501 MOPT_EXT4_ONLY | MOPT_CLEAR},
1502 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1503 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1504 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1505 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1506 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1507 MOPT_EXT4_ONLY | MOPT_CLEAR},
1508 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1509 MOPT_EXT4_ONLY | MOPT_CLEAR},
1510 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1511 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1512 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1513 EXT4_MOUNT_JOURNAL_CHECKSUM),
1514 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1515 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1516 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1517 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1518 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1519 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1521 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1523 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1524 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1525 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1526 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1527 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1528 {Opt_commit, 0, MOPT_GTE0},
1529 {Opt_max_batch_time, 0, MOPT_GTE0},
1530 {Opt_min_batch_time, 0, MOPT_GTE0},
1531 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1532 {Opt_init_itable, 0, MOPT_GTE0},
1533 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
1534 {Opt_stripe, 0, MOPT_GTE0},
1535 {Opt_resuid, 0, MOPT_GTE0},
1536 {Opt_resgid, 0, MOPT_GTE0},
1537 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1538 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1539 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1540 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1541 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1542 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1543 MOPT_NO_EXT2 | MOPT_DATAJ},
1544 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1545 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1546 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1547 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1548 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1550 {Opt_acl, 0, MOPT_NOSUPPORT},
1551 {Opt_noacl, 0, MOPT_NOSUPPORT},
1553 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1554 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1555 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1556 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1558 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1560 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1562 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1563 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1564 MOPT_CLEAR | MOPT_Q},
1565 {Opt_usrjquota, 0, MOPT_Q},
1566 {Opt_grpjquota, 0, MOPT_Q},
1567 {Opt_offusrjquota, 0, MOPT_Q},
1568 {Opt_offgrpjquota, 0, MOPT_Q},
1569 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1570 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1571 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1572 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1573 {Opt_test_dummy_encryption, 0, MOPT_GTE0},
1577 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1578 substring_t *args, unsigned long *journal_devnum,
1579 unsigned int *journal_ioprio, int is_remount)
1581 struct ext4_sb_info *sbi = EXT4_SB(sb);
1582 const struct mount_opts *m;
1588 if (token == Opt_usrjquota)
1589 return set_qf_name(sb, USRQUOTA, &args[0]);
1590 else if (token == Opt_grpjquota)
1591 return set_qf_name(sb, GRPQUOTA, &args[0]);
1592 else if (token == Opt_offusrjquota)
1593 return clear_qf_name(sb, USRQUOTA);
1594 else if (token == Opt_offgrpjquota)
1595 return clear_qf_name(sb, GRPQUOTA);
1599 case Opt_nouser_xattr:
1600 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1603 return 1; /* handled by get_sb_block() */
1605 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1608 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1611 sb->s_flags |= MS_I_VERSION;
1614 sb->s_flags |= MS_LAZYTIME;
1616 case Opt_nolazytime:
1617 sb->s_flags &= ~MS_LAZYTIME;
1621 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1622 if (token == m->token)
1625 if (m->token == Opt_err) {
1626 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1627 "or missing value", opt);
1631 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1632 ext4_msg(sb, KERN_ERR,
1633 "Mount option \"%s\" incompatible with ext2", opt);
1636 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1637 ext4_msg(sb, KERN_ERR,
1638 "Mount option \"%s\" incompatible with ext3", opt);
1642 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
1644 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1646 if (m->flags & MOPT_EXPLICIT) {
1647 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
1648 set_opt2(sb, EXPLICIT_DELALLOC);
1649 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
1650 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
1654 if (m->flags & MOPT_CLEAR_ERR)
1655 clear_opt(sb, ERRORS_MASK);
1656 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1657 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1658 "options when quota turned on");
1662 if (m->flags & MOPT_NOSUPPORT) {
1663 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1664 } else if (token == Opt_commit) {
1666 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1667 sbi->s_commit_interval = HZ * arg;
1668 } else if (token == Opt_max_batch_time) {
1669 sbi->s_max_batch_time = arg;
1670 } else if (token == Opt_min_batch_time) {
1671 sbi->s_min_batch_time = arg;
1672 } else if (token == Opt_inode_readahead_blks) {
1673 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1674 ext4_msg(sb, KERN_ERR,
1675 "EXT4-fs: inode_readahead_blks must be "
1676 "0 or a power of 2 smaller than 2^31");
1679 sbi->s_inode_readahead_blks = arg;
1680 } else if (token == Opt_init_itable) {
1681 set_opt(sb, INIT_INODE_TABLE);
1683 arg = EXT4_DEF_LI_WAIT_MULT;
1684 sbi->s_li_wait_mult = arg;
1685 } else if (token == Opt_max_dir_size_kb) {
1686 sbi->s_max_dir_size_kb = arg;
1687 } else if (token == Opt_stripe) {
1688 sbi->s_stripe = arg;
1689 } else if (token == Opt_resuid) {
1690 uid = make_kuid(current_user_ns(), arg);
1691 if (!uid_valid(uid)) {
1692 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1695 sbi->s_resuid = uid;
1696 } else if (token == Opt_resgid) {
1697 gid = make_kgid(current_user_ns(), arg);
1698 if (!gid_valid(gid)) {
1699 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1702 sbi->s_resgid = gid;
1703 } else if (token == Opt_journal_dev) {
1705 ext4_msg(sb, KERN_ERR,
1706 "Cannot specify journal on remount");
1709 *journal_devnum = arg;
1710 } else if (token == Opt_journal_path) {
1712 struct inode *journal_inode;
1717 ext4_msg(sb, KERN_ERR,
1718 "Cannot specify journal on remount");
1721 journal_path = match_strdup(&args[0]);
1722 if (!journal_path) {
1723 ext4_msg(sb, KERN_ERR, "error: could not dup "
1724 "journal device string");
1728 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1730 ext4_msg(sb, KERN_ERR, "error: could not find "
1731 "journal device path: error %d", error);
1732 kfree(journal_path);
1736 journal_inode = d_inode(path.dentry);
1737 if (!S_ISBLK(journal_inode->i_mode)) {
1738 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1739 "is not a block device", journal_path);
1741 kfree(journal_path);
1745 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1747 kfree(journal_path);
1748 } else if (token == Opt_journal_ioprio) {
1750 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1755 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1756 } else if (token == Opt_test_dummy_encryption) {
1757 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1758 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1759 ext4_msg(sb, KERN_WARNING,
1760 "Test dummy encryption mode enabled");
1762 ext4_msg(sb, KERN_WARNING,
1763 "Test dummy encryption mount option ignored");
1765 } else if (m->flags & MOPT_DATAJ) {
1767 if (!sbi->s_journal)
1768 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1769 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1770 ext4_msg(sb, KERN_ERR,
1771 "Cannot change data mode on remount");
1775 clear_opt(sb, DATA_FLAGS);
1776 sbi->s_mount_opt |= m->mount_opt;
1779 } else if (m->flags & MOPT_QFMT) {
1780 if (sb_any_quota_loaded(sb) &&
1781 sbi->s_jquota_fmt != m->mount_opt) {
1782 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1783 "quota options when quota turned on");
1786 if (ext4_has_feature_quota(sb)) {
1787 ext4_msg(sb, KERN_INFO,
1788 "Quota format mount options ignored "
1789 "when QUOTA feature is enabled");
1792 sbi->s_jquota_fmt = m->mount_opt;
1794 } else if (token == Opt_dax) {
1795 #ifdef CONFIG_FS_DAX
1796 ext4_msg(sb, KERN_WARNING,
1797 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
1798 sbi->s_mount_opt |= m->mount_opt;
1800 ext4_msg(sb, KERN_INFO, "dax option not supported");
1803 } else if (token == Opt_data_err_abort) {
1804 sbi->s_mount_opt |= m->mount_opt;
1805 } else if (token == Opt_data_err_ignore) {
1806 sbi->s_mount_opt &= ~m->mount_opt;
1810 if (m->flags & MOPT_CLEAR)
1812 else if (unlikely(!(m->flags & MOPT_SET))) {
1813 ext4_msg(sb, KERN_WARNING,
1814 "buggy handling of option %s", opt);
1819 sbi->s_mount_opt |= m->mount_opt;
1821 sbi->s_mount_opt &= ~m->mount_opt;
1826 static int parse_options(char *options, struct super_block *sb,
1827 unsigned long *journal_devnum,
1828 unsigned int *journal_ioprio,
1831 struct ext4_sb_info *sbi = EXT4_SB(sb);
1833 substring_t args[MAX_OPT_ARGS];
1839 while ((p = strsep(&options, ",")) != NULL) {
1843 * Initialize args struct so we know whether arg was
1844 * found; some options take optional arguments.
1846 args[0].to = args[0].from = NULL;
1847 token = match_token(p, tokens, args);
1848 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1849 journal_ioprio, is_remount) < 0)
1854 * We do the test below only for project quotas. 'usrquota' and
1855 * 'grpquota' mount options are allowed even without quota feature
1856 * to support legacy quotas in quota files.
1858 if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) {
1859 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. "
1860 "Cannot enable project quota enforcement.");
1863 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1864 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1865 clear_opt(sb, USRQUOTA);
1867 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1868 clear_opt(sb, GRPQUOTA);
1870 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1871 ext4_msg(sb, KERN_ERR, "old and new quota "
1876 if (!sbi->s_jquota_fmt) {
1877 ext4_msg(sb, KERN_ERR, "journaled quota format "
1883 if (test_opt(sb, DIOREAD_NOLOCK)) {
1885 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1887 if (blocksize < PAGE_SIZE) {
1888 ext4_msg(sb, KERN_ERR, "can't mount with "
1889 "dioread_nolock if block size != PAGE_SIZE");
1893 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
1894 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
1895 ext4_msg(sb, KERN_ERR, "can't mount with journal_async_commit "
1896 "in data=ordered mode");
1902 static inline void ext4_show_quota_options(struct seq_file *seq,
1903 struct super_block *sb)
1905 #if defined(CONFIG_QUOTA)
1906 struct ext4_sb_info *sbi = EXT4_SB(sb);
1908 if (sbi->s_jquota_fmt) {
1911 switch (sbi->s_jquota_fmt) {
1922 seq_printf(seq, ",jqfmt=%s", fmtname);
1925 if (sbi->s_qf_names[USRQUOTA])
1926 seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]);
1928 if (sbi->s_qf_names[GRPQUOTA])
1929 seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]);
1933 static const char *token2str(int token)
1935 const struct match_token *t;
1937 for (t = tokens; t->token != Opt_err; t++)
1938 if (t->token == token && !strchr(t->pattern, '='))
1945 * - it's set to a non-default value OR
1946 * - if the per-sb default is different from the global default
1948 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1951 struct ext4_sb_info *sbi = EXT4_SB(sb);
1952 struct ext4_super_block *es = sbi->s_es;
1953 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1954 const struct mount_opts *m;
1955 char sep = nodefs ? '\n' : ',';
1957 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1958 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1960 if (sbi->s_sb_block != 1)
1961 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1963 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1964 int want_set = m->flags & MOPT_SET;
1965 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1966 (m->flags & MOPT_CLEAR_ERR))
1968 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
1969 continue; /* skip if same as the default */
1971 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1972 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1973 continue; /* select Opt_noFoo vs Opt_Foo */
1974 SEQ_OPTS_PRINT("%s", token2str(m->token));
1977 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
1978 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1979 SEQ_OPTS_PRINT("resuid=%u",
1980 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1981 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
1982 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1983 SEQ_OPTS_PRINT("resgid=%u",
1984 from_kgid_munged(&init_user_ns, sbi->s_resgid));
1985 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
1986 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1987 SEQ_OPTS_PUTS("errors=remount-ro");
1988 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1989 SEQ_OPTS_PUTS("errors=continue");
1990 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1991 SEQ_OPTS_PUTS("errors=panic");
1992 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
1993 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
1994 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
1995 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
1996 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
1997 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
1998 if (sb->s_flags & MS_I_VERSION)
1999 SEQ_OPTS_PUTS("i_version");
2000 if (nodefs || sbi->s_stripe)
2001 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
2002 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
2003 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2004 SEQ_OPTS_PUTS("data=journal");
2005 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2006 SEQ_OPTS_PUTS("data=ordered");
2007 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2008 SEQ_OPTS_PUTS("data=writeback");
2011 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
2012 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2013 sbi->s_inode_readahead_blks);
2015 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
2016 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
2017 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
2018 if (nodefs || sbi->s_max_dir_size_kb)
2019 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2020 if (test_opt(sb, DATA_ERR_ABORT))
2021 SEQ_OPTS_PUTS("data_err=abort");
2023 ext4_show_quota_options(seq, sb);
2027 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2029 return _ext4_show_options(seq, root->d_sb, 0);
2032 int ext4_seq_options_show(struct seq_file *seq, void *offset)
2034 struct super_block *sb = seq->private;
2037 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
2038 rc = _ext4_show_options(seq, sb, 1);
2039 seq_puts(seq, "\n");
2043 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
2046 struct ext4_sb_info *sbi = EXT4_SB(sb);
2049 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
2050 ext4_msg(sb, KERN_ERR, "revision level too high, "
2051 "forcing read-only mode");
2056 if (!(sbi->s_mount_state & EXT4_VALID_FS))
2057 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
2058 "running e2fsck is recommended");
2059 else if (sbi->s_mount_state & EXT4_ERROR_FS)
2060 ext4_msg(sb, KERN_WARNING,
2061 "warning: mounting fs with errors, "
2062 "running e2fsck is recommended");
2063 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
2064 le16_to_cpu(es->s_mnt_count) >=
2065 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
2066 ext4_msg(sb, KERN_WARNING,
2067 "warning: maximal mount count reached, "
2068 "running e2fsck is recommended");
2069 else if (le32_to_cpu(es->s_checkinterval) &&
2070 (le32_to_cpu(es->s_lastcheck) +
2071 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
2072 ext4_msg(sb, KERN_WARNING,
2073 "warning: checktime reached, "
2074 "running e2fsck is recommended");
2075 if (!sbi->s_journal)
2076 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
2077 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
2078 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
2079 le16_add_cpu(&es->s_mnt_count, 1);
2080 es->s_mtime = cpu_to_le32(get_seconds());
2081 ext4_update_dynamic_rev(sb);
2083 ext4_set_feature_journal_needs_recovery(sb);
2085 ext4_commit_super(sb, 1);
2087 if (test_opt(sb, DEBUG))
2088 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
2089 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
2091 sbi->s_groups_count,
2092 EXT4_BLOCKS_PER_GROUP(sb),
2093 EXT4_INODES_PER_GROUP(sb),
2094 sbi->s_mount_opt, sbi->s_mount_opt2);
2096 cleancache_init_fs(sb);
2100 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2102 struct ext4_sb_info *sbi = EXT4_SB(sb);
2103 struct flex_groups *new_groups;
2106 if (!sbi->s_log_groups_per_flex)
2109 size = ext4_flex_group(sbi, ngroup - 1) + 1;
2110 if (size <= sbi->s_flex_groups_allocated)
2113 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
2114 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
2116 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
2117 size / (int) sizeof(struct flex_groups));
2121 if (sbi->s_flex_groups) {
2122 memcpy(new_groups, sbi->s_flex_groups,
2123 (sbi->s_flex_groups_allocated *
2124 sizeof(struct flex_groups)));
2125 kvfree(sbi->s_flex_groups);
2127 sbi->s_flex_groups = new_groups;
2128 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
2132 static int ext4_fill_flex_info(struct super_block *sb)
2134 struct ext4_sb_info *sbi = EXT4_SB(sb);
2135 struct ext4_group_desc *gdp = NULL;
2136 ext4_group_t flex_group;
2139 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2140 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
2141 sbi->s_log_groups_per_flex = 0;
2145 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2149 for (i = 0; i < sbi->s_groups_count; i++) {
2150 gdp = ext4_get_group_desc(sb, i, NULL);
2152 flex_group = ext4_flex_group(sbi, i);
2153 atomic_add(ext4_free_inodes_count(sb, gdp),
2154 &sbi->s_flex_groups[flex_group].free_inodes);
2155 atomic64_add(ext4_free_group_clusters(sb, gdp),
2156 &sbi->s_flex_groups[flex_group].free_clusters);
2157 atomic_add(ext4_used_dirs_count(sb, gdp),
2158 &sbi->s_flex_groups[flex_group].used_dirs);
2166 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
2167 struct ext4_group_desc *gdp)
2169 int offset = offsetof(struct ext4_group_desc, bg_checksum);
2171 __le32 le_group = cpu_to_le32(block_group);
2172 struct ext4_sb_info *sbi = EXT4_SB(sb);
2174 if (ext4_has_metadata_csum(sbi->s_sb)) {
2175 /* Use new metadata_csum algorithm */
2177 __u16 dummy_csum = 0;
2179 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2181 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2182 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2183 sizeof(dummy_csum));
2184 offset += sizeof(dummy_csum);
2185 if (offset < sbi->s_desc_size)
2186 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2187 sbi->s_desc_size - offset);
2189 crc = csum32 & 0xFFFF;
2193 /* old crc16 code */
2194 if (!ext4_has_feature_gdt_csum(sb))
2197 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2198 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2199 crc = crc16(crc, (__u8 *)gdp, offset);
2200 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2201 /* for checksum of struct ext4_group_desc do the rest...*/
2202 if (ext4_has_feature_64bit(sb) &&
2203 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2204 crc = crc16(crc, (__u8 *)gdp + offset,
2205 le16_to_cpu(sbi->s_es->s_desc_size) -
2209 return cpu_to_le16(crc);
2212 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2213 struct ext4_group_desc *gdp)
2215 if (ext4_has_group_desc_csum(sb) &&
2216 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
2222 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2223 struct ext4_group_desc *gdp)
2225 if (!ext4_has_group_desc_csum(sb))
2227 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
2230 /* Called at mount-time, super-block is locked */
2231 static int ext4_check_descriptors(struct super_block *sb,
2232 ext4_fsblk_t sb_block,
2233 ext4_group_t *first_not_zeroed)
2235 struct ext4_sb_info *sbi = EXT4_SB(sb);
2236 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2237 ext4_fsblk_t last_block;
2238 ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0);
2239 ext4_fsblk_t block_bitmap;
2240 ext4_fsblk_t inode_bitmap;
2241 ext4_fsblk_t inode_table;
2242 int flexbg_flag = 0;
2243 ext4_group_t i, grp = sbi->s_groups_count;
2245 if (ext4_has_feature_flex_bg(sb))
2248 ext4_debug("Checking group descriptors");
2250 for (i = 0; i < sbi->s_groups_count; i++) {
2251 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2253 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2254 last_block = ext4_blocks_count(sbi->s_es) - 1;
2256 last_block = first_block +
2257 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2259 if ((grp == sbi->s_groups_count) &&
2260 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2263 block_bitmap = ext4_block_bitmap(sb, gdp);
2264 if (block_bitmap == sb_block) {
2265 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2266 "Block bitmap for group %u overlaps "
2268 if (!(sb->s_flags & MS_RDONLY))
2271 if (block_bitmap >= sb_block + 1 &&
2272 block_bitmap <= last_bg_block) {
2273 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2274 "Block bitmap for group %u overlaps "
2275 "block group descriptors", i);
2276 if (!(sb->s_flags & MS_RDONLY))
2279 if (block_bitmap < first_block || block_bitmap > last_block) {
2280 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2281 "Block bitmap for group %u not in group "
2282 "(block %llu)!", i, block_bitmap);
2285 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2286 if (inode_bitmap == sb_block) {
2287 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2288 "Inode bitmap for group %u overlaps "
2290 if (!(sb->s_flags & MS_RDONLY))
2293 if (inode_bitmap >= sb_block + 1 &&
2294 inode_bitmap <= last_bg_block) {
2295 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2296 "Inode bitmap for group %u overlaps "
2297 "block group descriptors", i);
2298 if (!(sb->s_flags & MS_RDONLY))
2301 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2302 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2303 "Inode bitmap for group %u not in group "
2304 "(block %llu)!", i, inode_bitmap);
2307 inode_table = ext4_inode_table(sb, gdp);
2308 if (inode_table == sb_block) {
2309 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2310 "Inode table for group %u overlaps "
2312 if (!(sb->s_flags & MS_RDONLY))
2315 if (inode_table >= sb_block + 1 &&
2316 inode_table <= last_bg_block) {
2317 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2318 "Inode table for group %u overlaps "
2319 "block group descriptors", i);
2320 if (!(sb->s_flags & MS_RDONLY))
2323 if (inode_table < first_block ||
2324 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2325 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2326 "Inode table for group %u not in group "
2327 "(block %llu)!", i, inode_table);
2330 ext4_lock_group(sb, i);
2331 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2332 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2333 "Checksum for group %u failed (%u!=%u)",
2334 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
2335 gdp)), le16_to_cpu(gdp->bg_checksum));
2336 if (!(sb->s_flags & MS_RDONLY)) {
2337 ext4_unlock_group(sb, i);
2341 ext4_unlock_group(sb, i);
2343 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2345 if (NULL != first_not_zeroed)
2346 *first_not_zeroed = grp;
2350 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2351 * the superblock) which were deleted from all directories, but held open by
2352 * a process at the time of a crash. We walk the list and try to delete these
2353 * inodes at recovery time (only with a read-write filesystem).
2355 * In order to keep the orphan inode chain consistent during traversal (in
2356 * case of crash during recovery), we link each inode into the superblock
2357 * orphan list_head and handle it the same way as an inode deletion during
2358 * normal operation (which journals the operations for us).
2360 * We only do an iget() and an iput() on each inode, which is very safe if we
2361 * accidentally point at an in-use or already deleted inode. The worst that
2362 * can happen in this case is that we get a "bit already cleared" message from
2363 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2364 * e2fsck was run on this filesystem, and it must have already done the orphan
2365 * inode cleanup for us, so we can safely abort without any further action.
2367 static void ext4_orphan_cleanup(struct super_block *sb,
2368 struct ext4_super_block *es)
2370 unsigned int s_flags = sb->s_flags;
2371 int nr_orphans = 0, nr_truncates = 0;
2373 int quota_update = 0;
2376 if (!es->s_last_orphan) {
2377 jbd_debug(4, "no orphan inodes to clean up\n");
2381 if (bdev_read_only(sb->s_bdev)) {
2382 ext4_msg(sb, KERN_ERR, "write access "
2383 "unavailable, skipping orphan cleanup");
2387 /* Check if feature set would not allow a r/w mount */
2388 if (!ext4_feature_set_ok(sb, 0)) {
2389 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2390 "unknown ROCOMPAT features");
2394 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2395 /* don't clear list on RO mount w/ errors */
2396 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
2397 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
2398 "clearing orphan list.\n");
2399 es->s_last_orphan = 0;
2401 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2405 if (s_flags & MS_RDONLY) {
2406 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2407 sb->s_flags &= ~MS_RDONLY;
2410 /* Needed for iput() to work correctly and not trash data */
2411 sb->s_flags |= MS_ACTIVE;
2414 * Turn on quotas which were not enabled for read-only mounts if
2415 * filesystem has quota feature, so that they are updated correctly.
2417 if (ext4_has_feature_quota(sb) && (s_flags & MS_RDONLY)) {
2418 int ret = ext4_enable_quotas(sb);
2423 ext4_msg(sb, KERN_ERR,
2424 "Cannot turn on quotas: error %d", ret);
2427 /* Turn on journaled quotas used for old sytle */
2428 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2429 if (EXT4_SB(sb)->s_qf_names[i]) {
2430 int ret = ext4_quota_on_mount(sb, i);
2435 ext4_msg(sb, KERN_ERR,
2436 "Cannot turn on journaled "
2437 "quota: type %d: error %d", i, ret);
2442 while (es->s_last_orphan) {
2443 struct inode *inode;
2446 * We may have encountered an error during cleanup; if
2447 * so, skip the rest.
2449 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2450 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2451 es->s_last_orphan = 0;
2455 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2456 if (IS_ERR(inode)) {
2457 es->s_last_orphan = 0;
2461 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2462 dquot_initialize(inode);
2463 if (inode->i_nlink) {
2464 if (test_opt(sb, DEBUG))
2465 ext4_msg(sb, KERN_DEBUG,
2466 "%s: truncating inode %lu to %lld bytes",
2467 __func__, inode->i_ino, inode->i_size);
2468 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2469 inode->i_ino, inode->i_size);
2471 truncate_inode_pages(inode->i_mapping, inode->i_size);
2472 ext4_truncate(inode);
2473 inode_unlock(inode);
2476 if (test_opt(sb, DEBUG))
2477 ext4_msg(sb, KERN_DEBUG,
2478 "%s: deleting unreferenced inode %lu",
2479 __func__, inode->i_ino);
2480 jbd_debug(2, "deleting unreferenced inode %lu\n",
2484 iput(inode); /* The delete magic happens here! */
2487 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2490 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2491 PLURAL(nr_orphans));
2493 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2494 PLURAL(nr_truncates));
2496 /* Turn off quotas if they were enabled for orphan cleanup */
2498 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2499 if (sb_dqopt(sb)->files[i])
2500 dquot_quota_off(sb, i);
2504 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2508 * Maximal extent format file size.
2509 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2510 * extent format containers, within a sector_t, and within i_blocks
2511 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2512 * so that won't be a limiting factor.
2514 * However there is other limiting factor. We do store extents in the form
2515 * of starting block and length, hence the resulting length of the extent
2516 * covering maximum file size must fit into on-disk format containers as
2517 * well. Given that length is always by 1 unit bigger than max unit (because
2518 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2520 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2522 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2525 loff_t upper_limit = MAX_LFS_FILESIZE;
2527 /* small i_blocks in vfs inode? */
2528 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2530 * CONFIG_LBDAF is not enabled implies the inode
2531 * i_block represent total blocks in 512 bytes
2532 * 32 == size of vfs inode i_blocks * 8
2534 upper_limit = (1LL << 32) - 1;
2536 /* total blocks in file system block size */
2537 upper_limit >>= (blkbits - 9);
2538 upper_limit <<= blkbits;
2542 * 32-bit extent-start container, ee_block. We lower the maxbytes
2543 * by one fs block, so ee_len can cover the extent of maximum file
2546 res = (1LL << 32) - 1;
2549 /* Sanity check against vm- & vfs- imposed limits */
2550 if (res > upper_limit)
2557 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2558 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2559 * We need to be 1 filesystem block less than the 2^48 sector limit.
2561 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2563 loff_t res = EXT4_NDIR_BLOCKS;
2566 /* This is calculated to be the largest file size for a dense, block
2567 * mapped file such that the file's total number of 512-byte sectors,
2568 * including data and all indirect blocks, does not exceed (2^48 - 1).
2570 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2571 * number of 512-byte sectors of the file.
2574 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2576 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2577 * the inode i_block field represents total file blocks in
2578 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2580 upper_limit = (1LL << 32) - 1;
2582 /* total blocks in file system block size */
2583 upper_limit >>= (bits - 9);
2587 * We use 48 bit ext4_inode i_blocks
2588 * With EXT4_HUGE_FILE_FL set the i_blocks
2589 * represent total number of blocks in
2590 * file system block size
2592 upper_limit = (1LL << 48) - 1;
2596 /* indirect blocks */
2598 /* double indirect blocks */
2599 meta_blocks += 1 + (1LL << (bits-2));
2600 /* tripple indirect blocks */
2601 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2603 upper_limit -= meta_blocks;
2604 upper_limit <<= bits;
2606 res += 1LL << (bits-2);
2607 res += 1LL << (2*(bits-2));
2608 res += 1LL << (3*(bits-2));
2610 if (res > upper_limit)
2613 if (res > MAX_LFS_FILESIZE)
2614 res = MAX_LFS_FILESIZE;
2619 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2620 ext4_fsblk_t logical_sb_block, int nr)
2622 struct ext4_sb_info *sbi = EXT4_SB(sb);
2623 ext4_group_t bg, first_meta_bg;
2626 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2628 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
2629 return logical_sb_block + nr + 1;
2630 bg = sbi->s_desc_per_block * nr;
2631 if (ext4_bg_has_super(sb, bg))
2635 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2636 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2637 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2640 if (sb->s_blocksize == 1024 && nr == 0 &&
2641 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2644 return (has_super + ext4_group_first_block_no(sb, bg));
2648 * ext4_get_stripe_size: Get the stripe size.
2649 * @sbi: In memory super block info
2651 * If we have specified it via mount option, then
2652 * use the mount option value. If the value specified at mount time is
2653 * greater than the blocks per group use the super block value.
2654 * If the super block value is greater than blocks per group return 0.
2655 * Allocator needs it be less than blocks per group.
2658 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2660 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2661 unsigned long stripe_width =
2662 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2665 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2666 ret = sbi->s_stripe;
2667 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
2669 else if (stride && stride <= sbi->s_blocks_per_group)
2675 * If the stripe width is 1, this makes no sense and
2676 * we set it to 0 to turn off stripe handling code.
2685 * Check whether this filesystem can be mounted based on
2686 * the features present and the RDONLY/RDWR mount requested.
2687 * Returns 1 if this filesystem can be mounted as requested,
2688 * 0 if it cannot be.
2690 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2692 if (ext4_has_unknown_ext4_incompat_features(sb)) {
2693 ext4_msg(sb, KERN_ERR,
2694 "Couldn't mount because of "
2695 "unsupported optional features (%x)",
2696 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2697 ~EXT4_FEATURE_INCOMPAT_SUPP));
2704 if (ext4_has_feature_readonly(sb)) {
2705 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2706 sb->s_flags |= MS_RDONLY;
2710 /* Check that feature set is OK for a read-write mount */
2711 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
2712 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2713 "unsupported optional features (%x)",
2714 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2715 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2719 * Large file size enabled file system can only be mounted
2720 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2722 if (ext4_has_feature_huge_file(sb)) {
2723 if (sizeof(blkcnt_t) < sizeof(u64)) {
2724 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2725 "cannot be mounted RDWR without "
2730 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
2731 ext4_msg(sb, KERN_ERR,
2732 "Can't support bigalloc feature without "
2733 "extents feature\n");
2737 #ifndef CONFIG_QUOTA
2738 if (ext4_has_feature_quota(sb) && !readonly) {
2739 ext4_msg(sb, KERN_ERR,
2740 "Filesystem with quota feature cannot be mounted RDWR "
2741 "without CONFIG_QUOTA");
2744 if (ext4_has_feature_project(sb) && !readonly) {
2745 ext4_msg(sb, KERN_ERR,
2746 "Filesystem with project quota feature cannot be mounted RDWR "
2747 "without CONFIG_QUOTA");
2750 #endif /* CONFIG_QUOTA */
2755 * This function is called once a day if we have errors logged
2756 * on the file system
2758 static void print_daily_error_info(unsigned long arg)
2760 struct super_block *sb = (struct super_block *) arg;
2761 struct ext4_sb_info *sbi;
2762 struct ext4_super_block *es;
2767 if (es->s_error_count)
2768 /* fsck newer than v1.41.13 is needed to clean this condition. */
2769 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
2770 le32_to_cpu(es->s_error_count));
2771 if (es->s_first_error_time) {
2772 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
2773 sb->s_id, le32_to_cpu(es->s_first_error_time),
2774 (int) sizeof(es->s_first_error_func),
2775 es->s_first_error_func,
2776 le32_to_cpu(es->s_first_error_line));
2777 if (es->s_first_error_ino)
2778 printk(KERN_CONT ": inode %u",
2779 le32_to_cpu(es->s_first_error_ino));
2780 if (es->s_first_error_block)
2781 printk(KERN_CONT ": block %llu", (unsigned long long)
2782 le64_to_cpu(es->s_first_error_block));
2783 printk(KERN_CONT "\n");
2785 if (es->s_last_error_time) {
2786 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
2787 sb->s_id, le32_to_cpu(es->s_last_error_time),
2788 (int) sizeof(es->s_last_error_func),
2789 es->s_last_error_func,
2790 le32_to_cpu(es->s_last_error_line));
2791 if (es->s_last_error_ino)
2792 printk(KERN_CONT ": inode %u",
2793 le32_to_cpu(es->s_last_error_ino));
2794 if (es->s_last_error_block)
2795 printk(KERN_CONT ": block %llu", (unsigned long long)
2796 le64_to_cpu(es->s_last_error_block));
2797 printk(KERN_CONT "\n");
2799 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2802 /* Find next suitable group and run ext4_init_inode_table */
2803 static int ext4_run_li_request(struct ext4_li_request *elr)
2805 struct ext4_group_desc *gdp = NULL;
2806 ext4_group_t group, ngroups;
2807 struct super_block *sb;
2808 unsigned long timeout = 0;
2812 ngroups = EXT4_SB(sb)->s_groups_count;
2814 for (group = elr->lr_next_group; group < ngroups; group++) {
2815 gdp = ext4_get_group_desc(sb, group, NULL);
2821 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2825 if (group >= ngroups)
2830 ret = ext4_init_inode_table(sb, group,
2831 elr->lr_timeout ? 0 : 1);
2832 if (elr->lr_timeout == 0) {
2833 timeout = (jiffies - timeout) *
2834 elr->lr_sbi->s_li_wait_mult;
2835 elr->lr_timeout = timeout;
2837 elr->lr_next_sched = jiffies + elr->lr_timeout;
2838 elr->lr_next_group = group + 1;
2844 * Remove lr_request from the list_request and free the
2845 * request structure. Should be called with li_list_mtx held
2847 static void ext4_remove_li_request(struct ext4_li_request *elr)
2849 struct ext4_sb_info *sbi;
2856 list_del(&elr->lr_request);
2857 sbi->s_li_request = NULL;
2861 static void ext4_unregister_li_request(struct super_block *sb)
2863 mutex_lock(&ext4_li_mtx);
2864 if (!ext4_li_info) {
2865 mutex_unlock(&ext4_li_mtx);
2869 mutex_lock(&ext4_li_info->li_list_mtx);
2870 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2871 mutex_unlock(&ext4_li_info->li_list_mtx);
2872 mutex_unlock(&ext4_li_mtx);
2875 static struct task_struct *ext4_lazyinit_task;
2878 * This is the function where ext4lazyinit thread lives. It walks
2879 * through the request list searching for next scheduled filesystem.
2880 * When such a fs is found, run the lazy initialization request
2881 * (ext4_rn_li_request) and keep track of the time spend in this
2882 * function. Based on that time we compute next schedule time of
2883 * the request. When walking through the list is complete, compute
2884 * next waking time and put itself into sleep.
2886 static int ext4_lazyinit_thread(void *arg)
2888 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2889 struct list_head *pos, *n;
2890 struct ext4_li_request *elr;
2891 unsigned long next_wakeup, cur;
2893 BUG_ON(NULL == eli);
2897 next_wakeup = MAX_JIFFY_OFFSET;
2899 mutex_lock(&eli->li_list_mtx);
2900 if (list_empty(&eli->li_request_list)) {
2901 mutex_unlock(&eli->li_list_mtx);
2904 list_for_each_safe(pos, n, &eli->li_request_list) {
2907 elr = list_entry(pos, struct ext4_li_request,
2910 if (time_before(jiffies, elr->lr_next_sched)) {
2911 if (time_before(elr->lr_next_sched, next_wakeup))
2912 next_wakeup = elr->lr_next_sched;
2915 if (down_read_trylock(&elr->lr_super->s_umount)) {
2916 if (sb_start_write_trylock(elr->lr_super)) {
2919 * We hold sb->s_umount, sb can not
2920 * be removed from the list, it is
2921 * now safe to drop li_list_mtx
2923 mutex_unlock(&eli->li_list_mtx);
2924 err = ext4_run_li_request(elr);
2925 sb_end_write(elr->lr_super);
2926 mutex_lock(&eli->li_list_mtx);
2929 up_read((&elr->lr_super->s_umount));
2931 /* error, remove the lazy_init job */
2933 ext4_remove_li_request(elr);
2937 elr->lr_next_sched = jiffies +
2939 % (EXT4_DEF_LI_MAX_START_DELAY * HZ));
2941 if (time_before(elr->lr_next_sched, next_wakeup))
2942 next_wakeup = elr->lr_next_sched;
2944 mutex_unlock(&eli->li_list_mtx);
2949 if ((time_after_eq(cur, next_wakeup)) ||
2950 (MAX_JIFFY_OFFSET == next_wakeup)) {
2955 schedule_timeout_interruptible(next_wakeup - cur);
2957 if (kthread_should_stop()) {
2958 ext4_clear_request_list();
2965 * It looks like the request list is empty, but we need
2966 * to check it under the li_list_mtx lock, to prevent any
2967 * additions into it, and of course we should lock ext4_li_mtx
2968 * to atomically free the list and ext4_li_info, because at
2969 * this point another ext4 filesystem could be registering
2972 mutex_lock(&ext4_li_mtx);
2973 mutex_lock(&eli->li_list_mtx);
2974 if (!list_empty(&eli->li_request_list)) {
2975 mutex_unlock(&eli->li_list_mtx);
2976 mutex_unlock(&ext4_li_mtx);
2979 mutex_unlock(&eli->li_list_mtx);
2980 kfree(ext4_li_info);
2981 ext4_li_info = NULL;
2982 mutex_unlock(&ext4_li_mtx);
2987 static void ext4_clear_request_list(void)
2989 struct list_head *pos, *n;
2990 struct ext4_li_request *elr;
2992 mutex_lock(&ext4_li_info->li_list_mtx);
2993 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2994 elr = list_entry(pos, struct ext4_li_request,
2996 ext4_remove_li_request(elr);
2998 mutex_unlock(&ext4_li_info->li_list_mtx);
3001 static int ext4_run_lazyinit_thread(void)
3003 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3004 ext4_li_info, "ext4lazyinit");
3005 if (IS_ERR(ext4_lazyinit_task)) {
3006 int err = PTR_ERR(ext4_lazyinit_task);
3007 ext4_clear_request_list();
3008 kfree(ext4_li_info);
3009 ext4_li_info = NULL;
3010 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3011 "initialization thread\n",
3015 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3020 * Check whether it make sense to run itable init. thread or not.
3021 * If there is at least one uninitialized inode table, return
3022 * corresponding group number, else the loop goes through all
3023 * groups and return total number of groups.
3025 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3027 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3028 struct ext4_group_desc *gdp = NULL;
3030 if (!ext4_has_group_desc_csum(sb))
3033 for (group = 0; group < ngroups; group++) {
3034 gdp = ext4_get_group_desc(sb, group, NULL);
3038 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3045 static int ext4_li_info_new(void)
3047 struct ext4_lazy_init *eli = NULL;
3049 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3053 INIT_LIST_HEAD(&eli->li_request_list);
3054 mutex_init(&eli->li_list_mtx);
3056 eli->li_state |= EXT4_LAZYINIT_QUIT;
3063 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3066 struct ext4_sb_info *sbi = EXT4_SB(sb);
3067 struct ext4_li_request *elr;
3069 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3075 elr->lr_next_group = start;
3078 * Randomize first schedule time of the request to
3079 * spread the inode table initialization requests
3082 elr->lr_next_sched = jiffies + (prandom_u32() %
3083 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3087 int ext4_register_li_request(struct super_block *sb,
3088 ext4_group_t first_not_zeroed)
3090 struct ext4_sb_info *sbi = EXT4_SB(sb);
3091 struct ext4_li_request *elr = NULL;
3092 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3095 mutex_lock(&ext4_li_mtx);
3096 if (sbi->s_li_request != NULL) {
3098 * Reset timeout so it can be computed again, because
3099 * s_li_wait_mult might have changed.
3101 sbi->s_li_request->lr_timeout = 0;
3105 if (first_not_zeroed == ngroups ||
3106 (sb->s_flags & MS_RDONLY) ||
3107 !test_opt(sb, INIT_INODE_TABLE))
3110 elr = ext4_li_request_new(sb, first_not_zeroed);
3116 if (NULL == ext4_li_info) {
3117 ret = ext4_li_info_new();
3122 mutex_lock(&ext4_li_info->li_list_mtx);
3123 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3124 mutex_unlock(&ext4_li_info->li_list_mtx);
3126 sbi->s_li_request = elr;
3128 * set elr to NULL here since it has been inserted to
3129 * the request_list and the removal and free of it is
3130 * handled by ext4_clear_request_list from now on.
3134 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3135 ret = ext4_run_lazyinit_thread();
3140 mutex_unlock(&ext4_li_mtx);
3147 * We do not need to lock anything since this is called on
3150 static void ext4_destroy_lazyinit_thread(void)
3153 * If thread exited earlier
3154 * there's nothing to be done.
3156 if (!ext4_li_info || !ext4_lazyinit_task)
3159 kthread_stop(ext4_lazyinit_task);
3162 static int set_journal_csum_feature_set(struct super_block *sb)
3165 int compat, incompat;
3166 struct ext4_sb_info *sbi = EXT4_SB(sb);
3168 if (ext4_has_metadata_csum(sb)) {
3169 /* journal checksum v3 */
3171 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
3173 /* journal checksum v1 */
3174 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3178 jbd2_journal_clear_features(sbi->s_journal,
3179 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3180 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3181 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3182 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3183 ret = jbd2_journal_set_features(sbi->s_journal,
3185 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3187 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3188 ret = jbd2_journal_set_features(sbi->s_journal,
3191 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3192 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3194 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3195 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3202 * Note: calculating the overhead so we can be compatible with
3203 * historical BSD practice is quite difficult in the face of
3204 * clusters/bigalloc. This is because multiple metadata blocks from
3205 * different block group can end up in the same allocation cluster.
3206 * Calculating the exact overhead in the face of clustered allocation
3207 * requires either O(all block bitmaps) in memory or O(number of block
3208 * groups**2) in time. We will still calculate the superblock for
3209 * older file systems --- and if we come across with a bigalloc file
3210 * system with zero in s_overhead_clusters the estimate will be close to
3211 * correct especially for very large cluster sizes --- but for newer
3212 * file systems, it's better to calculate this figure once at mkfs
3213 * time, and store it in the superblock. If the superblock value is
3214 * present (even for non-bigalloc file systems), we will use it.
3216 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3219 struct ext4_sb_info *sbi = EXT4_SB(sb);
3220 struct ext4_group_desc *gdp;
3221 ext4_fsblk_t first_block, last_block, b;
3222 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3223 int s, j, count = 0;
3225 if (!ext4_has_feature_bigalloc(sb))
3226 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3227 sbi->s_itb_per_group + 2);
3229 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3230 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3231 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3232 for (i = 0; i < ngroups; i++) {
3233 gdp = ext4_get_group_desc(sb, i, NULL);
3234 b = ext4_block_bitmap(sb, gdp);
3235 if (b >= first_block && b <= last_block) {
3236 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3239 b = ext4_inode_bitmap(sb, gdp);
3240 if (b >= first_block && b <= last_block) {
3241 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3244 b = ext4_inode_table(sb, gdp);
3245 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3246 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3247 int c = EXT4_B2C(sbi, b - first_block);
3248 ext4_set_bit(c, buf);
3254 if (ext4_bg_has_super(sb, grp)) {
3255 ext4_set_bit(s++, buf);
3258 j = ext4_bg_num_gdb(sb, grp);
3259 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
3260 ext4_error(sb, "Invalid number of block group "
3261 "descriptor blocks: %d", j);
3262 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
3266 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3270 return EXT4_CLUSTERS_PER_GROUP(sb) -
3271 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3275 * Compute the overhead and stash it in sbi->s_overhead
3277 int ext4_calculate_overhead(struct super_block *sb)
3279 struct ext4_sb_info *sbi = EXT4_SB(sb);
3280 struct ext4_super_block *es = sbi->s_es;
3281 struct inode *j_inode;
3282 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
3283 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3284 ext4_fsblk_t overhead = 0;
3285 char *buf = (char *) get_zeroed_page(GFP_NOFS);
3291 * Compute the overhead (FS structures). This is constant
3292 * for a given filesystem unless the number of block groups
3293 * changes so we cache the previous value until it does.
3297 * All of the blocks before first_data_block are overhead
3299 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3302 * Add the overhead found in each block group
3304 for (i = 0; i < ngroups; i++) {
3307 blks = count_overhead(sb, i, buf);
3310 memset(buf, 0, PAGE_SIZE);
3315 * Add the internal journal blocks whether the journal has been
3318 if (sbi->s_journal && !sbi->journal_bdev)
3319 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3320 else if (ext4_has_feature_journal(sb) && !sbi->s_journal) {
3321 j_inode = ext4_get_journal_inode(sb, j_inum);
3323 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
3324 overhead += EXT4_NUM_B2C(sbi, j_blocks);
3327 ext4_msg(sb, KERN_ERR, "can't get journal size");
3330 sbi->s_overhead = overhead;
3332 free_page((unsigned long) buf);
3336 static void ext4_set_resv_clusters(struct super_block *sb)
3338 ext4_fsblk_t resv_clusters;
3339 struct ext4_sb_info *sbi = EXT4_SB(sb);
3342 * There's no need to reserve anything when we aren't using extents.
3343 * The space estimates are exact, there are no unwritten extents,
3344 * hole punching doesn't need new metadata... This is needed especially
3345 * to keep ext2/3 backward compatibility.
3347 if (!ext4_has_feature_extents(sb))
3350 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3351 * This should cover the situations where we can not afford to run
3352 * out of space like for example punch hole, or converting
3353 * unwritten extents in delalloc path. In most cases such
3354 * allocation would require 1, or 2 blocks, higher numbers are
3357 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3358 sbi->s_cluster_bits);
3360 do_div(resv_clusters, 50);
3361 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3363 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
3366 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3368 char *orig_data = kstrdup(data, GFP_KERNEL);
3369 struct buffer_head *bh;
3370 struct ext4_super_block *es = NULL;
3371 struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3373 ext4_fsblk_t sb_block = get_sb_block(&data);
3374 ext4_fsblk_t logical_sb_block;
3375 unsigned long offset = 0;
3376 unsigned long journal_devnum = 0;
3377 unsigned long def_mount_opts;
3381 int blocksize, clustersize;
3382 unsigned int db_count;
3384 int needs_recovery, has_huge_files, has_bigalloc;
3387 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3388 ext4_group_t first_not_zeroed;
3390 if ((data && !orig_data) || !sbi)
3393 sbi->s_blockgroup_lock =
3394 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3395 if (!sbi->s_blockgroup_lock)
3398 sb->s_fs_info = sbi;
3400 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3401 sbi->s_sb_block = sb_block;
3402 if (sb->s_bdev->bd_part)
3403 sbi->s_sectors_written_start =
3404 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3406 /* Cleanup superblock name */
3407 strreplace(sb->s_id, '/', '!');
3409 /* -EINVAL is default */
3411 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3413 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3418 * The ext4 superblock will not be buffer aligned for other than 1kB
3419 * block sizes. We need to calculate the offset from buffer start.
3421 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3422 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3423 offset = do_div(logical_sb_block, blocksize);
3425 logical_sb_block = sb_block;
3428 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
3429 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3433 * Note: s_es must be initialized as soon as possible because
3434 * some ext4 macro-instructions depend on its value
3436 es = (struct ext4_super_block *) (bh->b_data + offset);
3438 sb->s_magic = le16_to_cpu(es->s_magic);
3439 if (sb->s_magic != EXT4_SUPER_MAGIC)
3441 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3443 /* Warn if metadata_csum and gdt_csum are both set. */
3444 if (ext4_has_feature_metadata_csum(sb) &&
3445 ext4_has_feature_gdt_csum(sb))
3446 ext4_warning(sb, "metadata_csum and uninit_bg are "
3447 "redundant flags; please run fsck.");
3449 /* Check for a known checksum algorithm */
3450 if (!ext4_verify_csum_type(sb, es)) {
3451 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3452 "unknown checksum algorithm.");
3457 /* Load the checksum driver */
3458 if (ext4_has_feature_metadata_csum(sb)) {
3459 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3460 if (IS_ERR(sbi->s_chksum_driver)) {
3461 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3462 ret = PTR_ERR(sbi->s_chksum_driver);
3463 sbi->s_chksum_driver = NULL;
3468 /* Check superblock checksum */
3469 if (!ext4_superblock_csum_verify(sb, es)) {
3470 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3471 "invalid superblock checksum. Run e2fsck?");
3477 /* Precompute checksum seed for all metadata */
3478 if (ext4_has_feature_csum_seed(sb))
3479 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
3480 else if (ext4_has_metadata_csum(sb))
3481 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3482 sizeof(es->s_uuid));
3484 /* Set defaults before we parse the mount options */
3485 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3486 set_opt(sb, INIT_INODE_TABLE);
3487 if (def_mount_opts & EXT4_DEFM_DEBUG)
3489 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3491 if (def_mount_opts & EXT4_DEFM_UID16)
3492 set_opt(sb, NO_UID32);
3493 /* xattr user namespace & acls are now defaulted on */
3494 set_opt(sb, XATTR_USER);
3495 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3496 set_opt(sb, POSIX_ACL);
3498 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3499 if (ext4_has_metadata_csum(sb))
3500 set_opt(sb, JOURNAL_CHECKSUM);
3502 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3503 set_opt(sb, JOURNAL_DATA);
3504 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3505 set_opt(sb, ORDERED_DATA);
3506 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3507 set_opt(sb, WRITEBACK_DATA);
3509 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3510 set_opt(sb, ERRORS_PANIC);
3511 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3512 set_opt(sb, ERRORS_CONT);
3514 set_opt(sb, ERRORS_RO);
3515 /* block_validity enabled by default; disable with noblock_validity */
3516 set_opt(sb, BLOCK_VALIDITY);
3517 if (def_mount_opts & EXT4_DEFM_DISCARD)
3518 set_opt(sb, DISCARD);
3520 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3521 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3522 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3523 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3524 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3526 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3527 set_opt(sb, BARRIER);
3530 * enable delayed allocation by default
3531 * Use -o nodelalloc to turn it off
3533 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3534 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3535 set_opt(sb, DELALLOC);
3538 * set default s_li_wait_mult for lazyinit, for the case there is
3539 * no mount option specified.
3541 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3543 if (sbi->s_es->s_mount_opts[0]) {
3544 char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
3545 sizeof(sbi->s_es->s_mount_opts),
3549 if (!parse_options(s_mount_opts, sb, &journal_devnum,
3550 &journal_ioprio, 0)) {
3551 ext4_msg(sb, KERN_WARNING,
3552 "failed to parse options in superblock: %s",
3555 kfree(s_mount_opts);
3557 sbi->s_def_mount_opt = sbi->s_mount_opt;
3558 if (!parse_options((char *) data, sb, &journal_devnum,
3559 &journal_ioprio, 0))
3562 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3563 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3564 "with data=journal disables delayed "
3565 "allocation and O_DIRECT support!\n");
3566 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3567 ext4_msg(sb, KERN_ERR, "can't mount with "
3568 "both data=journal and delalloc");
3571 if (test_opt(sb, DIOREAD_NOLOCK)) {
3572 ext4_msg(sb, KERN_ERR, "can't mount with "
3573 "both data=journal and dioread_nolock");
3576 if (test_opt(sb, DAX)) {
3577 ext4_msg(sb, KERN_ERR, "can't mount with "
3578 "both data=journal and dax");
3581 if (ext4_has_feature_encrypt(sb)) {
3582 ext4_msg(sb, KERN_WARNING,
3583 "encrypted files will use data=ordered "
3584 "instead of data journaling mode");
3586 if (test_opt(sb, DELALLOC))
3587 clear_opt(sb, DELALLOC);
3589 sb->s_iflags |= SB_I_CGROUPWB;
3592 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3593 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3595 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3596 (ext4_has_compat_features(sb) ||
3597 ext4_has_ro_compat_features(sb) ||
3598 ext4_has_incompat_features(sb)))
3599 ext4_msg(sb, KERN_WARNING,
3600 "feature flags set on rev 0 fs, "
3601 "running e2fsck is recommended");
3603 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3604 set_opt2(sb, HURD_COMPAT);
3605 if (ext4_has_feature_64bit(sb)) {
3606 ext4_msg(sb, KERN_ERR,
3607 "The Hurd can't support 64-bit file systems");
3612 if (IS_EXT2_SB(sb)) {
3613 if (ext2_feature_set_ok(sb))
3614 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3615 "using the ext4 subsystem");
3617 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3618 "to feature incompatibilities");
3623 if (IS_EXT3_SB(sb)) {
3624 if (ext3_feature_set_ok(sb))
3625 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3626 "using the ext4 subsystem");
3628 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3629 "to feature incompatibilities");
3635 * Check feature flags regardless of the revision level, since we
3636 * previously didn't change the revision level when setting the flags,
3637 * so there is a chance incompat flags are set on a rev 0 filesystem.
3639 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3642 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3643 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3644 blocksize > EXT4_MAX_BLOCK_SIZE) {
3645 ext4_msg(sb, KERN_ERR,
3646 "Unsupported filesystem blocksize %d (%d log_block_size)",
3647 blocksize, le32_to_cpu(es->s_log_block_size));
3650 if (le32_to_cpu(es->s_log_block_size) >
3651 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3652 ext4_msg(sb, KERN_ERR,
3653 "Invalid log block size: %u",
3654 le32_to_cpu(es->s_log_block_size));
3657 if (le32_to_cpu(es->s_log_cluster_size) >
3658 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3659 ext4_msg(sb, KERN_ERR,
3660 "Invalid log cluster size: %u",
3661 le32_to_cpu(es->s_log_cluster_size));
3665 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
3666 ext4_msg(sb, KERN_ERR,
3667 "Number of reserved GDT blocks insanely large: %d",
3668 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
3672 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
3673 err = bdev_dax_supported(sb, blocksize);
3678 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
3679 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
3680 es->s_encryption_level);
3684 if (sb->s_blocksize != blocksize) {
3685 /* Validate the filesystem blocksize */
3686 if (!sb_set_blocksize(sb, blocksize)) {
3687 ext4_msg(sb, KERN_ERR, "bad block size %d",
3693 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3694 offset = do_div(logical_sb_block, blocksize);
3695 bh = sb_bread_unmovable(sb, logical_sb_block);
3697 ext4_msg(sb, KERN_ERR,
3698 "Can't read superblock on 2nd try");
3701 es = (struct ext4_super_block *)(bh->b_data + offset);
3703 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3704 ext4_msg(sb, KERN_ERR,
3705 "Magic mismatch, very weird!");
3710 has_huge_files = ext4_has_feature_huge_file(sb);
3711 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3713 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3715 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3716 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3717 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3719 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3720 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3721 if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) {
3722 ext4_msg(sb, KERN_ERR, "invalid first ino: %u",
3726 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3727 (!is_power_of_2(sbi->s_inode_size)) ||
3728 (sbi->s_inode_size > blocksize)) {
3729 ext4_msg(sb, KERN_ERR,
3730 "unsupported inode size: %d",
3734 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3735 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3738 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3739 if (ext4_has_feature_64bit(sb)) {
3740 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3741 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3742 !is_power_of_2(sbi->s_desc_size)) {
3743 ext4_msg(sb, KERN_ERR,
3744 "unsupported descriptor size %lu",
3749 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3751 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3752 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3754 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3755 if (sbi->s_inodes_per_block == 0)
3757 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
3758 sbi->s_inodes_per_group > blocksize * 8) {
3759 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
3760 sbi->s_blocks_per_group);
3763 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3764 sbi->s_inodes_per_block;
3765 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3767 sbi->s_mount_state = le16_to_cpu(es->s_state);
3768 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3769 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3771 for (i = 0; i < 4; i++)
3772 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3773 sbi->s_def_hash_version = es->s_def_hash_version;
3774 if (ext4_has_feature_dir_index(sb)) {
3775 i = le32_to_cpu(es->s_flags);
3776 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3777 sbi->s_hash_unsigned = 3;
3778 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3779 #ifdef __CHAR_UNSIGNED__
3780 if (!(sb->s_flags & MS_RDONLY))
3782 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3783 sbi->s_hash_unsigned = 3;
3785 if (!(sb->s_flags & MS_RDONLY))
3787 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3792 /* Handle clustersize */
3793 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3794 has_bigalloc = ext4_has_feature_bigalloc(sb);
3796 if (clustersize < blocksize) {
3797 ext4_msg(sb, KERN_ERR,
3798 "cluster size (%d) smaller than "
3799 "block size (%d)", clustersize, blocksize);
3802 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3803 le32_to_cpu(es->s_log_block_size);
3804 sbi->s_clusters_per_group =
3805 le32_to_cpu(es->s_clusters_per_group);
3806 if (sbi->s_clusters_per_group > blocksize * 8) {
3807 ext4_msg(sb, KERN_ERR,
3808 "#clusters per group too big: %lu",
3809 sbi->s_clusters_per_group);
3812 if (sbi->s_blocks_per_group !=
3813 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3814 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3815 "clusters per group (%lu) inconsistent",
3816 sbi->s_blocks_per_group,
3817 sbi->s_clusters_per_group);
3821 if (clustersize != blocksize) {
3822 ext4_msg(sb, KERN_ERR,
3823 "fragment/cluster size (%d) != "
3824 "block size (%d)", clustersize, blocksize);
3827 if (sbi->s_blocks_per_group > blocksize * 8) {
3828 ext4_msg(sb, KERN_ERR,
3829 "#blocks per group too big: %lu",
3830 sbi->s_blocks_per_group);
3833 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3834 sbi->s_cluster_bits = 0;
3836 sbi->s_cluster_ratio = clustersize / blocksize;
3838 /* Do we have standard group size of clustersize * 8 blocks ? */
3839 if (sbi->s_blocks_per_group == clustersize << 3)
3840 set_opt2(sb, STD_GROUP_SIZE);
3843 * Test whether we have more sectors than will fit in sector_t,
3844 * and whether the max offset is addressable by the page cache.
3846 err = generic_check_addressable(sb->s_blocksize_bits,
3847 ext4_blocks_count(es));
3849 ext4_msg(sb, KERN_ERR, "filesystem"
3850 " too large to mount safely on this system");
3851 if (sizeof(sector_t) < 8)
3852 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3856 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3859 /* check blocks count against device size */
3860 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3861 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3862 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3863 "exceeds size of device (%llu blocks)",
3864 ext4_blocks_count(es), blocks_count);
3869 * It makes no sense for the first data block to be beyond the end
3870 * of the filesystem.
3872 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3873 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3874 "block %u is beyond end of filesystem (%llu)",
3875 le32_to_cpu(es->s_first_data_block),
3876 ext4_blocks_count(es));
3879 if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) &&
3880 (sbi->s_cluster_ratio == 1)) {
3881 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3882 "block is 0 with a 1k block and cluster size");
3886 blocks_count = (ext4_blocks_count(es) -
3887 le32_to_cpu(es->s_first_data_block) +
3888 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3889 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3890 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3891 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3892 "(block count %llu, first data block %u, "
3893 "blocks per group %lu)", sbi->s_groups_count,
3894 ext4_blocks_count(es),
3895 le32_to_cpu(es->s_first_data_block),
3896 EXT4_BLOCKS_PER_GROUP(sb));
3899 sbi->s_groups_count = blocks_count;
3900 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3901 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3902 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3903 EXT4_DESC_PER_BLOCK(sb);
3904 if (ext4_has_feature_meta_bg(sb)) {
3905 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
3906 ext4_msg(sb, KERN_WARNING,
3907 "first meta block group too large: %u "
3908 "(group descriptor block count %u)",
3909 le32_to_cpu(es->s_first_meta_bg), db_count);
3913 sbi->s_group_desc = ext4_kvmalloc(db_count *
3914 sizeof(struct buffer_head *),
3916 if (sbi->s_group_desc == NULL) {
3917 ext4_msg(sb, KERN_ERR, "not enough memory");
3921 if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) !=
3922 le32_to_cpu(es->s_inodes_count)) {
3923 ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu",
3924 le32_to_cpu(es->s_inodes_count),
3925 ((u64)sbi->s_groups_count * sbi->s_inodes_per_group));
3930 bgl_lock_init(sbi->s_blockgroup_lock);
3932 for (i = 0; i < db_count; i++) {
3933 block = descriptor_loc(sb, logical_sb_block, i);
3934 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
3935 if (!sbi->s_group_desc[i]) {
3936 ext4_msg(sb, KERN_ERR,
3937 "can't read group descriptor %d", i);
3942 sbi->s_gdb_count = db_count;
3943 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
3944 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3945 ret = -EFSCORRUPTED;
3949 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3950 spin_lock_init(&sbi->s_next_gen_lock);
3952 setup_timer(&sbi->s_err_report, print_daily_error_info,
3953 (unsigned long) sb);
3955 /* Register extent status tree shrinker */
3956 if (ext4_es_register_shrinker(sbi))
3959 sbi->s_stripe = ext4_get_stripe_size(sbi);
3960 sbi->s_extent_max_zeroout_kb = 32;
3963 * set up enough so that it can read an inode
3965 sb->s_op = &ext4_sops;
3966 sb->s_export_op = &ext4_export_ops;
3967 sb->s_xattr = ext4_xattr_handlers;
3968 #ifdef CONFIG_EXT4_FS_ENCRYPTION
3969 sb->s_cop = &ext4_cryptops;
3972 sb->dq_op = &ext4_quota_operations;
3973 if (ext4_has_feature_quota(sb))
3974 sb->s_qcop = &dquot_quotactl_sysfile_ops;
3976 sb->s_qcop = &ext4_qctl_operations;
3977 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
3979 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3981 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3982 mutex_init(&sbi->s_orphan_lock);
3986 needs_recovery = (es->s_last_orphan != 0 ||
3987 ext4_has_feature_journal_needs_recovery(sb));
3989 if (ext4_has_feature_mmp(sb) && !(sb->s_flags & MS_RDONLY))
3990 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3991 goto failed_mount3a;
3994 * The first inode we look at is the journal inode. Don't try
3995 * root first: it may be modified in the journal!
3997 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
3998 err = ext4_load_journal(sb, es, journal_devnum);
4000 goto failed_mount3a;
4001 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
4002 ext4_has_feature_journal_needs_recovery(sb)) {
4003 ext4_msg(sb, KERN_ERR, "required journal recovery "
4004 "suppressed and not mounted read-only");
4005 goto failed_mount_wq;
4007 /* Nojournal mode, all journal mount options are illegal */
4008 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
4009 ext4_msg(sb, KERN_ERR, "can't mount with "
4010 "journal_checksum, fs mounted w/o journal");
4011 goto failed_mount_wq;
4013 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4014 ext4_msg(sb, KERN_ERR, "can't mount with "
4015 "journal_async_commit, fs mounted w/o journal");
4016 goto failed_mount_wq;
4018 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
4019 ext4_msg(sb, KERN_ERR, "can't mount with "
4020 "commit=%lu, fs mounted w/o journal",
4021 sbi->s_commit_interval / HZ);
4022 goto failed_mount_wq;
4024 if (EXT4_MOUNT_DATA_FLAGS &
4025 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
4026 ext4_msg(sb, KERN_ERR, "can't mount with "
4027 "data=, fs mounted w/o journal");
4028 goto failed_mount_wq;
4030 sbi->s_def_mount_opt &= EXT4_MOUNT_JOURNAL_CHECKSUM;
4031 clear_opt(sb, JOURNAL_CHECKSUM);
4032 clear_opt(sb, DATA_FLAGS);
4033 sbi->s_journal = NULL;
4038 if (ext4_has_feature_64bit(sb) &&
4039 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4040 JBD2_FEATURE_INCOMPAT_64BIT)) {
4041 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
4042 goto failed_mount_wq;
4045 if (!set_journal_csum_feature_set(sb)) {
4046 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4048 goto failed_mount_wq;
4051 /* We have now updated the journal if required, so we can
4052 * validate the data journaling mode. */
4053 switch (test_opt(sb, DATA_FLAGS)) {
4055 /* No mode set, assume a default based on the journal
4056 * capabilities: ORDERED_DATA if the journal can
4057 * cope, else JOURNAL_DATA
4059 if (jbd2_journal_check_available_features
4060 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
4061 set_opt(sb, ORDERED_DATA);
4063 set_opt(sb, JOURNAL_DATA);
4066 case EXT4_MOUNT_ORDERED_DATA:
4067 case EXT4_MOUNT_WRITEBACK_DATA:
4068 if (!jbd2_journal_check_available_features
4069 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4070 ext4_msg(sb, KERN_ERR, "Journal does not support "
4071 "requested data journaling mode");
4072 goto failed_mount_wq;
4077 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4079 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4082 sbi->s_mb_cache = ext4_xattr_create_cache();
4083 if (!sbi->s_mb_cache) {
4084 ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
4085 goto failed_mount_wq;
4088 if ((DUMMY_ENCRYPTION_ENABLED(sbi) || ext4_has_feature_encrypt(sb)) &&
4089 (blocksize != PAGE_SIZE)) {
4090 ext4_msg(sb, KERN_ERR,
4091 "Unsupported blocksize for fs encryption");
4092 goto failed_mount_wq;
4095 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !(sb->s_flags & MS_RDONLY) &&
4096 !ext4_has_feature_encrypt(sb)) {
4097 ext4_set_feature_encrypt(sb);
4098 ext4_commit_super(sb, 1);
4102 * Get the # of file system overhead blocks from the
4103 * superblock if present.
4105 if (es->s_overhead_clusters)
4106 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4108 err = ext4_calculate_overhead(sb);
4110 goto failed_mount_wq;
4114 * The maximum number of concurrent works can be high and
4115 * concurrency isn't really necessary. Limit it to 1.
4117 EXT4_SB(sb)->rsv_conversion_wq =
4118 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4119 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4120 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
4126 * The jbd2_journal_load will have done any necessary log recovery,
4127 * so we can safely mount the rest of the filesystem now.
4130 root = ext4_iget(sb, EXT4_ROOT_INO);
4132 ext4_msg(sb, KERN_ERR, "get root inode failed");
4133 ret = PTR_ERR(root);
4137 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4138 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4142 sb->s_root = d_make_root(root);
4144 ext4_msg(sb, KERN_ERR, "get root dentry failed");
4149 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
4150 sb->s_flags |= MS_RDONLY;
4152 /* determine the minimum size of new large inodes, if present */
4153 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4154 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4155 EXT4_GOOD_OLD_INODE_SIZE;
4156 if (ext4_has_feature_extra_isize(sb)) {
4157 if (sbi->s_want_extra_isize <
4158 le16_to_cpu(es->s_want_extra_isize))
4159 sbi->s_want_extra_isize =
4160 le16_to_cpu(es->s_want_extra_isize);
4161 if (sbi->s_want_extra_isize <
4162 le16_to_cpu(es->s_min_extra_isize))
4163 sbi->s_want_extra_isize =
4164 le16_to_cpu(es->s_min_extra_isize);
4167 /* Check if enough inode space is available */
4168 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4169 sbi->s_inode_size) {
4170 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4171 EXT4_GOOD_OLD_INODE_SIZE;
4172 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4176 ext4_set_resv_clusters(sb);
4178 err = ext4_setup_system_zone(sb);
4180 ext4_msg(sb, KERN_ERR, "failed to initialize system "
4182 goto failed_mount4a;
4186 err = ext4_mb_init(sb);
4188 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4193 block = ext4_count_free_clusters(sb);
4194 ext4_free_blocks_count_set(sbi->s_es,
4195 EXT4_C2B(sbi, block));
4196 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4199 unsigned long freei = ext4_count_free_inodes(sb);
4200 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
4201 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4205 err = percpu_counter_init(&sbi->s_dirs_counter,
4206 ext4_count_dirs(sb), GFP_KERNEL);
4208 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4211 err = percpu_init_rwsem(&sbi->s_journal_flag_rwsem);
4214 ext4_msg(sb, KERN_ERR, "insufficient memory");
4218 if (ext4_has_feature_flex_bg(sb))
4219 if (!ext4_fill_flex_info(sb)) {
4220 ext4_msg(sb, KERN_ERR,
4221 "unable to initialize "
4222 "flex_bg meta info!");
4226 err = ext4_register_li_request(sb, first_not_zeroed);
4230 err = ext4_register_sysfs(sb);
4235 /* Enable quota usage during mount. */
4236 if (ext4_has_feature_quota(sb) && !(sb->s_flags & MS_RDONLY)) {
4237 err = ext4_enable_quotas(sb);
4241 #endif /* CONFIG_QUOTA */
4243 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4244 ext4_orphan_cleanup(sb, es);
4245 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4246 if (needs_recovery) {
4247 ext4_msg(sb, KERN_INFO, "recovery complete");
4248 ext4_mark_recovery_complete(sb, es);
4250 if (EXT4_SB(sb)->s_journal) {
4251 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4252 descr = " journalled data mode";
4253 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4254 descr = " ordered data mode";
4256 descr = " writeback data mode";
4258 descr = "out journal";
4260 if (test_opt(sb, DISCARD)) {
4261 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4262 if (!blk_queue_discard(q))
4263 ext4_msg(sb, KERN_WARNING,
4264 "mounting with \"discard\" option, but "
4265 "the device does not support discard");
4268 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
4269 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4270 "Opts: %.*s%s%s", descr,
4271 (int) sizeof(sbi->s_es->s_mount_opts),
4272 sbi->s_es->s_mount_opts,
4273 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4275 if (es->s_error_count)
4276 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4278 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4279 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4280 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4281 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4288 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4293 ext4_unregister_sysfs(sb);
4296 ext4_unregister_li_request(sb);
4298 ext4_mb_release(sb);
4299 if (sbi->s_flex_groups)
4300 kvfree(sbi->s_flex_groups);
4301 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4302 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4303 percpu_counter_destroy(&sbi->s_dirs_counter);
4304 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4306 ext4_ext_release(sb);
4307 ext4_release_system_zone(sb);
4312 ext4_msg(sb, KERN_ERR, "mount failed");
4313 if (EXT4_SB(sb)->rsv_conversion_wq)
4314 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4316 if (sbi->s_mb_cache) {
4317 ext4_xattr_destroy_cache(sbi->s_mb_cache);
4318 sbi->s_mb_cache = NULL;
4320 if (sbi->s_journal) {
4321 jbd2_journal_destroy(sbi->s_journal);
4322 sbi->s_journal = NULL;
4325 ext4_es_unregister_shrinker(sbi);
4327 del_timer_sync(&sbi->s_err_report);
4329 kthread_stop(sbi->s_mmp_tsk);
4331 for (i = 0; i < db_count; i++)
4332 brelse(sbi->s_group_desc[i]);
4333 kvfree(sbi->s_group_desc);
4335 if (sbi->s_chksum_driver)
4336 crypto_free_shash(sbi->s_chksum_driver);
4338 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4339 kfree(sbi->s_qf_names[i]);
4341 ext4_blkdev_remove(sbi);
4344 sb->s_fs_info = NULL;
4345 kfree(sbi->s_blockgroup_lock);
4349 return err ? err : ret;
4353 * Setup any per-fs journal parameters now. We'll do this both on
4354 * initial mount, once the journal has been initialised but before we've
4355 * done any recovery; and again on any subsequent remount.
4357 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4359 struct ext4_sb_info *sbi = EXT4_SB(sb);
4361 journal->j_commit_interval = sbi->s_commit_interval;
4362 journal->j_min_batch_time = sbi->s_min_batch_time;
4363 journal->j_max_batch_time = sbi->s_max_batch_time;
4365 write_lock(&journal->j_state_lock);
4366 if (test_opt(sb, BARRIER))
4367 journal->j_flags |= JBD2_BARRIER;
4369 journal->j_flags &= ~JBD2_BARRIER;
4370 if (test_opt(sb, DATA_ERR_ABORT))
4371 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4373 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4374 write_unlock(&journal->j_state_lock);
4377 static struct inode *ext4_get_journal_inode(struct super_block *sb,
4378 unsigned int journal_inum)
4380 struct inode *journal_inode;
4383 * Test for the existence of a valid inode on disk. Bad things
4384 * happen if we iget() an unused inode, as the subsequent iput()
4385 * will try to delete it.
4387 journal_inode = ext4_iget(sb, journal_inum);
4388 if (IS_ERR(journal_inode)) {
4389 ext4_msg(sb, KERN_ERR, "no journal found");
4392 if (!journal_inode->i_nlink) {
4393 make_bad_inode(journal_inode);
4394 iput(journal_inode);
4395 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4399 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4400 journal_inode, journal_inode->i_size);
4401 if (!S_ISREG(journal_inode->i_mode)) {
4402 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4403 iput(journal_inode);
4406 return journal_inode;
4409 static journal_t *ext4_get_journal(struct super_block *sb,
4410 unsigned int journal_inum)
4412 struct inode *journal_inode;
4415 BUG_ON(!ext4_has_feature_journal(sb));
4417 journal_inode = ext4_get_journal_inode(sb, journal_inum);
4421 journal = jbd2_journal_init_inode(journal_inode);
4423 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4424 iput(journal_inode);
4427 journal->j_private = sb;
4428 ext4_init_journal_params(sb, journal);
4432 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4435 struct buffer_head *bh;
4439 int hblock, blocksize;
4440 ext4_fsblk_t sb_block;
4441 unsigned long offset;
4442 struct ext4_super_block *es;
4443 struct block_device *bdev;
4445 BUG_ON(!ext4_has_feature_journal(sb));
4447 bdev = ext4_blkdev_get(j_dev, sb);
4451 blocksize = sb->s_blocksize;
4452 hblock = bdev_logical_block_size(bdev);
4453 if (blocksize < hblock) {
4454 ext4_msg(sb, KERN_ERR,
4455 "blocksize too small for journal device");
4459 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4460 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4461 set_blocksize(bdev, blocksize);
4462 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4463 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4464 "external journal");
4468 es = (struct ext4_super_block *) (bh->b_data + offset);
4469 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4470 !(le32_to_cpu(es->s_feature_incompat) &
4471 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4472 ext4_msg(sb, KERN_ERR, "external journal has "
4478 if ((le32_to_cpu(es->s_feature_ro_compat) &
4479 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4480 es->s_checksum != ext4_superblock_csum(sb, es)) {
4481 ext4_msg(sb, KERN_ERR, "external journal has "
4482 "corrupt superblock");
4487 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4488 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4493 len = ext4_blocks_count(es);
4494 start = sb_block + 1;
4495 brelse(bh); /* we're done with the superblock */
4497 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4498 start, len, blocksize);
4500 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4503 journal->j_private = sb;
4504 ll_rw_block(REQ_OP_READ, REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4505 wait_on_buffer(journal->j_sb_buffer);
4506 if (!buffer_uptodate(journal->j_sb_buffer)) {
4507 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4510 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4511 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4512 "user (unsupported) - %d",
4513 be32_to_cpu(journal->j_superblock->s_nr_users));
4516 EXT4_SB(sb)->journal_bdev = bdev;
4517 ext4_init_journal_params(sb, journal);
4521 jbd2_journal_destroy(journal);
4523 ext4_blkdev_put(bdev);
4527 static int ext4_load_journal(struct super_block *sb,
4528 struct ext4_super_block *es,
4529 unsigned long journal_devnum)
4532 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4535 int really_read_only;
4537 BUG_ON(!ext4_has_feature_journal(sb));
4539 if (journal_devnum &&
4540 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4541 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4542 "numbers have changed");
4543 journal_dev = new_decode_dev(journal_devnum);
4545 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4547 really_read_only = bdev_read_only(sb->s_bdev);
4550 * Are we loading a blank journal or performing recovery after a
4551 * crash? For recovery, we need to check in advance whether we
4552 * can get read-write access to the device.
4554 if (ext4_has_feature_journal_needs_recovery(sb)) {
4555 if (sb->s_flags & MS_RDONLY) {
4556 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4557 "required on readonly filesystem");
4558 if (really_read_only) {
4559 ext4_msg(sb, KERN_ERR, "write access "
4560 "unavailable, cannot proceed");
4563 ext4_msg(sb, KERN_INFO, "write access will "
4564 "be enabled during recovery");
4568 if (journal_inum && journal_dev) {
4569 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4570 "and inode journals!");
4575 if (!(journal = ext4_get_journal(sb, journal_inum)))
4578 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4582 if (!(journal->j_flags & JBD2_BARRIER))
4583 ext4_msg(sb, KERN_INFO, "barriers disabled");
4585 if (!ext4_has_feature_journal_needs_recovery(sb))
4586 err = jbd2_journal_wipe(journal, !really_read_only);
4588 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4590 memcpy(save, ((char *) es) +
4591 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4592 err = jbd2_journal_load(journal);
4594 memcpy(((char *) es) + EXT4_S_ERR_START,
4595 save, EXT4_S_ERR_LEN);
4600 ext4_msg(sb, KERN_ERR, "error loading journal");
4601 jbd2_journal_destroy(journal);
4605 EXT4_SB(sb)->s_journal = journal;
4606 ext4_clear_journal_err(sb, es);
4608 if (!really_read_only && journal_devnum &&
4609 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4610 es->s_journal_dev = cpu_to_le32(journal_devnum);
4612 /* Make sure we flush the recovery flag to disk. */
4613 ext4_commit_super(sb, 1);
4619 static int ext4_commit_super(struct super_block *sb, int sync)
4621 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4622 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4625 if (!sbh || block_device_ejected(sb))
4629 * The superblock bh should be mapped, but it might not be if the
4630 * device was hot-removed. Not much we can do but fail the I/O.
4632 if (!buffer_mapped(sbh))
4636 * If the file system is mounted read-only, don't update the
4637 * superblock write time. This avoids updating the superblock
4638 * write time when we are mounting the root file system
4639 * read/only but we need to replay the journal; at that point,
4640 * for people who are east of GMT and who make their clock
4641 * tick in localtime for Windows bug-for-bug compatibility,
4642 * the clock is set in the future, and this will cause e2fsck
4643 * to complain and force a full file system check.
4645 if (!(sb->s_flags & MS_RDONLY))
4646 es->s_wtime = cpu_to_le32(get_seconds());
4647 if (sb->s_bdev->bd_part)
4648 es->s_kbytes_written =
4649 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4650 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4651 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4653 es->s_kbytes_written =
4654 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4655 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4656 ext4_free_blocks_count_set(es,
4657 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4658 &EXT4_SB(sb)->s_freeclusters_counter)));
4659 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4660 es->s_free_inodes_count =
4661 cpu_to_le32(percpu_counter_sum_positive(
4662 &EXT4_SB(sb)->s_freeinodes_counter));
4663 BUFFER_TRACE(sbh, "marking dirty");
4664 ext4_superblock_csum_set(sb);
4667 if (buffer_write_io_error(sbh)) {
4669 * Oh, dear. A previous attempt to write the
4670 * superblock failed. This could happen because the
4671 * USB device was yanked out. Or it could happen to
4672 * be a transient write error and maybe the block will
4673 * be remapped. Nothing we can do but to retry the
4674 * write and hope for the best.
4676 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4677 "superblock detected");
4678 clear_buffer_write_io_error(sbh);
4679 set_buffer_uptodate(sbh);
4681 mark_buffer_dirty(sbh);
4684 error = __sync_dirty_buffer(sbh,
4685 test_opt(sb, BARRIER) ? WRITE_FUA : WRITE_SYNC);
4689 error = buffer_write_io_error(sbh);
4691 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4693 clear_buffer_write_io_error(sbh);
4694 set_buffer_uptodate(sbh);
4701 * Have we just finished recovery? If so, and if we are mounting (or
4702 * remounting) the filesystem readonly, then we will end up with a
4703 * consistent fs on disk. Record that fact.
4705 static void ext4_mark_recovery_complete(struct super_block *sb,
4706 struct ext4_super_block *es)
4708 journal_t *journal = EXT4_SB(sb)->s_journal;
4710 if (!ext4_has_feature_journal(sb)) {
4711 BUG_ON(journal != NULL);
4714 jbd2_journal_lock_updates(journal);
4715 if (jbd2_journal_flush(journal) < 0)
4718 if (ext4_has_feature_journal_needs_recovery(sb) &&
4719 sb->s_flags & MS_RDONLY) {
4720 ext4_clear_feature_journal_needs_recovery(sb);
4721 ext4_commit_super(sb, 1);
4725 jbd2_journal_unlock_updates(journal);
4729 * If we are mounting (or read-write remounting) a filesystem whose journal
4730 * has recorded an error from a previous lifetime, move that error to the
4731 * main filesystem now.
4733 static void ext4_clear_journal_err(struct super_block *sb,
4734 struct ext4_super_block *es)
4740 BUG_ON(!ext4_has_feature_journal(sb));
4742 journal = EXT4_SB(sb)->s_journal;
4745 * Now check for any error status which may have been recorded in the
4746 * journal by a prior ext4_error() or ext4_abort()
4749 j_errno = jbd2_journal_errno(journal);
4753 errstr = ext4_decode_error(sb, j_errno, nbuf);
4754 ext4_warning(sb, "Filesystem error recorded "
4755 "from previous mount: %s", errstr);
4756 ext4_warning(sb, "Marking fs in need of filesystem check.");
4758 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4759 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4760 ext4_commit_super(sb, 1);
4762 jbd2_journal_clear_err(journal);
4763 jbd2_journal_update_sb_errno(journal);
4768 * Force the running and committing transactions to commit,
4769 * and wait on the commit.
4771 int ext4_force_commit(struct super_block *sb)
4775 if (sb->s_flags & MS_RDONLY)
4778 journal = EXT4_SB(sb)->s_journal;
4779 return ext4_journal_force_commit(journal);
4782 static int ext4_sync_fs(struct super_block *sb, int wait)
4786 bool needs_barrier = false;
4787 struct ext4_sb_info *sbi = EXT4_SB(sb);
4789 trace_ext4_sync_fs(sb, wait);
4790 flush_workqueue(sbi->rsv_conversion_wq);
4792 * Writeback quota in non-journalled quota case - journalled quota has
4795 dquot_writeback_dquots(sb, -1);
4797 * Data writeback is possible w/o journal transaction, so barrier must
4798 * being sent at the end of the function. But we can skip it if
4799 * transaction_commit will do it for us.
4801 if (sbi->s_journal) {
4802 target = jbd2_get_latest_transaction(sbi->s_journal);
4803 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4804 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4805 needs_barrier = true;
4807 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4809 ret = jbd2_log_wait_commit(sbi->s_journal,
4812 } else if (wait && test_opt(sb, BARRIER))
4813 needs_barrier = true;
4814 if (needs_barrier) {
4816 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4825 * LVM calls this function before a (read-only) snapshot is created. This
4826 * gives us a chance to flush the journal completely and mark the fs clean.
4828 * Note that only this function cannot bring a filesystem to be in a clean
4829 * state independently. It relies on upper layer to stop all data & metadata
4832 static int ext4_freeze(struct super_block *sb)
4837 if (sb->s_flags & MS_RDONLY)
4840 journal = EXT4_SB(sb)->s_journal;
4843 /* Now we set up the journal barrier. */
4844 jbd2_journal_lock_updates(journal);
4847 * Don't clear the needs_recovery flag if we failed to
4848 * flush the journal.
4850 error = jbd2_journal_flush(journal);
4854 /* Journal blocked and flushed, clear needs_recovery flag. */
4855 ext4_clear_feature_journal_needs_recovery(sb);
4858 error = ext4_commit_super(sb, 1);
4861 /* we rely on upper layer to stop further updates */
4862 jbd2_journal_unlock_updates(journal);
4867 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4868 * flag here, even though the filesystem is not technically dirty yet.
4870 static int ext4_unfreeze(struct super_block *sb)
4872 if (sb->s_flags & MS_RDONLY)
4875 if (EXT4_SB(sb)->s_journal) {
4876 /* Reset the needs_recovery flag before the fs is unlocked. */
4877 ext4_set_feature_journal_needs_recovery(sb);
4880 ext4_commit_super(sb, 1);
4885 * Structure to save mount options for ext4_remount's benefit
4887 struct ext4_mount_options {
4888 unsigned long s_mount_opt;
4889 unsigned long s_mount_opt2;
4892 unsigned long s_commit_interval;
4893 u32 s_min_batch_time, s_max_batch_time;
4896 char *s_qf_names[EXT4_MAXQUOTAS];
4900 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4902 struct ext4_super_block *es;
4903 struct ext4_sb_info *sbi = EXT4_SB(sb);
4904 unsigned long old_sb_flags;
4905 struct ext4_mount_options old_opts;
4906 int enable_quota = 0;
4908 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4913 char *orig_data = kstrdup(data, GFP_KERNEL);
4915 /* Store the original options */
4916 old_sb_flags = sb->s_flags;
4917 old_opts.s_mount_opt = sbi->s_mount_opt;
4918 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4919 old_opts.s_resuid = sbi->s_resuid;
4920 old_opts.s_resgid = sbi->s_resgid;
4921 old_opts.s_commit_interval = sbi->s_commit_interval;
4922 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4923 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4925 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4926 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4927 if (sbi->s_qf_names[i]) {
4928 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4930 if (!old_opts.s_qf_names[i]) {
4931 for (j = 0; j < i; j++)
4932 kfree(old_opts.s_qf_names[j]);
4937 old_opts.s_qf_names[i] = NULL;
4939 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4940 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4942 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4947 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
4948 test_opt(sb, JOURNAL_CHECKSUM)) {
4949 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
4950 "during remount not supported; ignoring");
4951 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
4954 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4955 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4956 ext4_msg(sb, KERN_ERR, "can't mount with "
4957 "both data=journal and delalloc");
4961 if (test_opt(sb, DIOREAD_NOLOCK)) {
4962 ext4_msg(sb, KERN_ERR, "can't mount with "
4963 "both data=journal and dioread_nolock");
4967 if (test_opt(sb, DAX)) {
4968 ext4_msg(sb, KERN_ERR, "can't mount with "
4969 "both data=journal and dax");
4975 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
4976 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
4977 "dax flag with busy inodes while remounting");
4978 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
4981 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4982 ext4_abort(sb, "Abort forced by user");
4984 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4985 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4989 if (sbi->s_journal) {
4990 ext4_init_journal_params(sb, sbi->s_journal);
4991 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4994 if (*flags & MS_LAZYTIME)
4995 sb->s_flags |= MS_LAZYTIME;
4997 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4998 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
5003 if (*flags & MS_RDONLY) {
5004 err = sync_filesystem(sb);
5007 err = dquot_suspend(sb, -1);
5012 * First of all, the unconditional stuff we have to do
5013 * to disable replay of the journal when we next remount
5015 sb->s_flags |= MS_RDONLY;
5018 * OK, test if we are remounting a valid rw partition
5019 * readonly, and if so set the rdonly flag and then
5020 * mark the partition as valid again.
5022 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
5023 (sbi->s_mount_state & EXT4_VALID_FS))
5024 es->s_state = cpu_to_le16(sbi->s_mount_state);
5027 ext4_mark_recovery_complete(sb, es);
5029 /* Make sure we can mount this feature set readwrite */
5030 if (ext4_has_feature_readonly(sb) ||
5031 !ext4_feature_set_ok(sb, 0)) {
5036 * Make sure the group descriptor checksums
5037 * are sane. If they aren't, refuse to remount r/w.
5039 for (g = 0; g < sbi->s_groups_count; g++) {
5040 struct ext4_group_desc *gdp =
5041 ext4_get_group_desc(sb, g, NULL);
5043 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
5044 ext4_msg(sb, KERN_ERR,
5045 "ext4_remount: Checksum for group %u failed (%u!=%u)",
5046 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
5047 le16_to_cpu(gdp->bg_checksum));
5054 * If we have an unprocessed orphan list hanging
5055 * around from a previously readonly bdev mount,
5056 * require a full umount/remount for now.
5058 if (es->s_last_orphan) {
5059 ext4_msg(sb, KERN_WARNING, "Couldn't "
5060 "remount RDWR because of unprocessed "
5061 "orphan inode list. Please "
5062 "umount/remount instead");
5068 * Mounting a RDONLY partition read-write, so reread
5069 * and store the current valid flag. (It may have
5070 * been changed by e2fsck since we originally mounted
5074 ext4_clear_journal_err(sb, es);
5075 sbi->s_mount_state = le16_to_cpu(es->s_state);
5076 if (!ext4_setup_super(sb, es, 0))
5077 sb->s_flags &= ~MS_RDONLY;
5078 if (ext4_has_feature_mmp(sb))
5079 if (ext4_multi_mount_protect(sb,
5080 le64_to_cpu(es->s_mmp_block))) {
5089 * Reinitialize lazy itable initialization thread based on
5092 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
5093 ext4_unregister_li_request(sb);
5095 ext4_group_t first_not_zeroed;
5096 first_not_zeroed = ext4_has_uninit_itable(sb);
5097 ext4_register_li_request(sb, first_not_zeroed);
5100 ext4_setup_system_zone(sb);
5101 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
5102 ext4_commit_super(sb, 1);
5105 /* Release old quota file names */
5106 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5107 kfree(old_opts.s_qf_names[i]);
5109 if (sb_any_quota_suspended(sb))
5110 dquot_resume(sb, -1);
5111 else if (ext4_has_feature_quota(sb)) {
5112 err = ext4_enable_quotas(sb);
5119 *flags = (*flags & ~MS_LAZYTIME) | (sb->s_flags & MS_LAZYTIME);
5120 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5125 sb->s_flags = old_sb_flags;
5126 sbi->s_mount_opt = old_opts.s_mount_opt;
5127 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
5128 sbi->s_resuid = old_opts.s_resuid;
5129 sbi->s_resgid = old_opts.s_resgid;
5130 sbi->s_commit_interval = old_opts.s_commit_interval;
5131 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5132 sbi->s_max_batch_time = old_opts.s_max_batch_time;
5134 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
5135 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
5136 kfree(sbi->s_qf_names[i]);
5137 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
5145 static int ext4_statfs_project(struct super_block *sb,
5146 kprojid_t projid, struct kstatfs *buf)
5149 struct dquot *dquot;
5153 qid = make_kqid_projid(projid);
5154 dquot = dqget(sb, qid);
5156 return PTR_ERR(dquot);
5157 spin_lock(&dq_data_lock);
5159 limit = (dquot->dq_dqb.dqb_bsoftlimit ?
5160 dquot->dq_dqb.dqb_bsoftlimit :
5161 dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits;
5162 if (limit && buf->f_blocks > limit) {
5163 curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits;
5164 buf->f_blocks = limit;
5165 buf->f_bfree = buf->f_bavail =
5166 (buf->f_blocks > curblock) ?
5167 (buf->f_blocks - curblock) : 0;
5170 limit = dquot->dq_dqb.dqb_isoftlimit ?
5171 dquot->dq_dqb.dqb_isoftlimit :
5172 dquot->dq_dqb.dqb_ihardlimit;
5173 if (limit && buf->f_files > limit) {
5174 buf->f_files = limit;
5176 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
5177 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
5180 spin_unlock(&dq_data_lock);
5186 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
5188 struct super_block *sb = dentry->d_sb;
5189 struct ext4_sb_info *sbi = EXT4_SB(sb);
5190 struct ext4_super_block *es = sbi->s_es;
5191 ext4_fsblk_t overhead = 0, resv_blocks;
5194 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
5196 if (!test_opt(sb, MINIX_DF))
5197 overhead = sbi->s_overhead;
5199 buf->f_type = EXT4_SUPER_MAGIC;
5200 buf->f_bsize = sb->s_blocksize;
5201 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
5202 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5203 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
5204 /* prevent underflow in case that few free space is available */
5205 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
5206 buf->f_bavail = buf->f_bfree -
5207 (ext4_r_blocks_count(es) + resv_blocks);
5208 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
5210 buf->f_files = le32_to_cpu(es->s_inodes_count);
5211 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5212 buf->f_namelen = EXT4_NAME_LEN;
5213 fsid = le64_to_cpup((void *)es->s_uuid) ^
5214 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5215 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5216 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
5219 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
5220 sb_has_quota_limits_enabled(sb, PRJQUOTA))
5221 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
5226 /* Helper function for writing quotas on sync - we need to start transaction
5227 * before quota file is locked for write. Otherwise the are possible deadlocks:
5228 * Process 1 Process 2
5229 * ext4_create() quota_sync()
5230 * jbd2_journal_start() write_dquot()
5231 * dquot_initialize() down(dqio_mutex)
5232 * down(dqio_mutex) jbd2_journal_start()
5238 static inline struct inode *dquot_to_inode(struct dquot *dquot)
5240 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
5243 static int ext4_write_dquot(struct dquot *dquot)
5247 struct inode *inode;
5249 inode = dquot_to_inode(dquot);
5250 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
5251 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
5253 return PTR_ERR(handle);
5254 ret = dquot_commit(dquot);
5255 err = ext4_journal_stop(handle);
5261 static int ext4_acquire_dquot(struct dquot *dquot)
5266 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5267 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
5269 return PTR_ERR(handle);
5270 ret = dquot_acquire(dquot);
5271 err = ext4_journal_stop(handle);
5277 static int ext4_release_dquot(struct dquot *dquot)
5282 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5283 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
5284 if (IS_ERR(handle)) {
5285 /* Release dquot anyway to avoid endless cycle in dqput() */
5286 dquot_release(dquot);
5287 return PTR_ERR(handle);
5289 ret = dquot_release(dquot);
5290 err = ext4_journal_stop(handle);
5296 static int ext4_mark_dquot_dirty(struct dquot *dquot)
5298 struct super_block *sb = dquot->dq_sb;
5299 struct ext4_sb_info *sbi = EXT4_SB(sb);
5301 /* Are we journaling quotas? */
5302 if (ext4_has_feature_quota(sb) ||
5303 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
5304 dquot_mark_dquot_dirty(dquot);
5305 return ext4_write_dquot(dquot);
5307 return dquot_mark_dquot_dirty(dquot);
5311 static int ext4_write_info(struct super_block *sb, int type)
5316 /* Data block + inode block */
5317 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
5319 return PTR_ERR(handle);
5320 ret = dquot_commit_info(sb, type);
5321 err = ext4_journal_stop(handle);
5328 * Turn on quotas during mount time - we need to find
5329 * the quota file and such...
5331 static int ext4_quota_on_mount(struct super_block *sb, int type)
5333 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5334 EXT4_SB(sb)->s_jquota_fmt, type);
5337 static void lockdep_set_quota_inode(struct inode *inode, int subclass)
5339 struct ext4_inode_info *ei = EXT4_I(inode);
5341 /* The first argument of lockdep_set_subclass has to be
5342 * *exactly* the same as the argument to init_rwsem() --- in
5343 * this case, in init_once() --- or lockdep gets unhappy
5344 * because the name of the lock is set using the
5345 * stringification of the argument to init_rwsem().
5347 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
5348 lockdep_set_subclass(&ei->i_data_sem, subclass);
5352 * Standard function to be called on quota_on
5354 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
5359 if (!test_opt(sb, QUOTA))
5362 /* Quotafile not on the same filesystem? */
5363 if (path->dentry->d_sb != sb)
5365 /* Journaling quota? */
5366 if (EXT4_SB(sb)->s_qf_names[type]) {
5367 /* Quotafile not in fs root? */
5368 if (path->dentry->d_parent != sb->s_root)
5369 ext4_msg(sb, KERN_WARNING,
5370 "Quota file not on filesystem root. "
5371 "Journaled quota will not work");
5375 * When we journal data on quota file, we have to flush journal to see
5376 * all updates to the file when we bypass pagecache...
5378 if (EXT4_SB(sb)->s_journal &&
5379 ext4_should_journal_data(d_inode(path->dentry))) {
5381 * We don't need to lock updates but journal_flush() could
5382 * otherwise be livelocked...
5384 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
5385 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
5386 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
5390 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
5391 err = dquot_quota_on(sb, type, format_id, path);
5393 lockdep_set_quota_inode(path->dentry->d_inode,
5398 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5402 struct inode *qf_inode;
5403 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5404 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5405 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5406 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
5409 BUG_ON(!ext4_has_feature_quota(sb));
5411 if (!qf_inums[type])
5414 qf_inode = ext4_iget(sb, qf_inums[type]);
5415 if (IS_ERR(qf_inode)) {
5416 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5417 return PTR_ERR(qf_inode);
5420 /* Don't account quota for quota files to avoid recursion */
5421 qf_inode->i_flags |= S_NOQUOTA;
5422 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
5423 err = dquot_enable(qf_inode, type, format_id, flags);
5426 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
5431 /* Enable usage tracking for all quota types. */
5432 static int ext4_enable_quotas(struct super_block *sb)
5435 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5436 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5437 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5438 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
5440 bool quota_mopt[EXT4_MAXQUOTAS] = {
5441 test_opt(sb, USRQUOTA),
5442 test_opt(sb, GRPQUOTA),
5443 test_opt(sb, PRJQUOTA),
5446 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5447 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
5448 if (qf_inums[type]) {
5449 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5450 DQUOT_USAGE_ENABLED |
5451 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
5453 for (type--; type >= 0; type--)
5454 dquot_quota_off(sb, type);
5457 "Failed to enable quota tracking "
5458 "(type=%d, err=%d). Please run "
5459 "e2fsck to fix.", type, err);
5467 static int ext4_quota_off(struct super_block *sb, int type)
5469 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);
5480 /* Update modification times of quota files when userspace can
5481 * start looking at them */
5482 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5485 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5486 ext4_mark_inode_dirty(handle, inode);
5487 ext4_journal_stop(handle);
5490 return dquot_quota_off(sb, type);
5493 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5494 * acquiring the locks... As quota files are never truncated and quota code
5495 * itself serializes the operations (and no one else should touch the files)
5496 * we don't have to be afraid of races */
5497 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5498 size_t len, loff_t off)
5500 struct inode *inode = sb_dqopt(sb)->files[type];
5501 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5502 int offset = off & (sb->s_blocksize - 1);
5505 struct buffer_head *bh;
5506 loff_t i_size = i_size_read(inode);
5510 if (off+len > i_size)
5513 while (toread > 0) {
5514 tocopy = sb->s_blocksize - offset < toread ?
5515 sb->s_blocksize - offset : toread;
5516 bh = ext4_bread(NULL, inode, blk, 0);
5519 if (!bh) /* A hole? */
5520 memset(data, 0, tocopy);
5522 memcpy(data, bh->b_data+offset, tocopy);
5532 /* Write to quotafile (we know the transaction is already started and has
5533 * enough credits) */
5534 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5535 const char *data, size_t len, loff_t off)
5537 struct inode *inode = sb_dqopt(sb)->files[type];
5538 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5539 int err, offset = off & (sb->s_blocksize - 1);
5541 struct buffer_head *bh;
5542 handle_t *handle = journal_current_handle();
5544 if (EXT4_SB(sb)->s_journal && !handle) {
5545 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5546 " cancelled because transaction is not started",
5547 (unsigned long long)off, (unsigned long long)len);
5551 * Since we account only one data block in transaction credits,
5552 * then it is impossible to cross a block boundary.
5554 if (sb->s_blocksize - offset < len) {
5555 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5556 " cancelled because not block aligned",
5557 (unsigned long long)off, (unsigned long long)len);
5562 bh = ext4_bread(handle, inode, blk,
5563 EXT4_GET_BLOCKS_CREATE |
5564 EXT4_GET_BLOCKS_METADATA_NOFAIL);
5565 } while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) &&
5566 ext4_should_retry_alloc(inode->i_sb, &retries));
5571 BUFFER_TRACE(bh, "get write access");
5572 err = ext4_journal_get_write_access(handle, bh);
5578 memcpy(bh->b_data+offset, data, len);
5579 flush_dcache_page(bh->b_page);
5581 err = ext4_handle_dirty_metadata(handle, NULL, bh);
5584 if (inode->i_size < off + len) {
5585 i_size_write(inode, off + len);
5586 EXT4_I(inode)->i_disksize = inode->i_size;
5587 ext4_mark_inode_dirty(handle, inode);
5592 static int ext4_get_next_id(struct super_block *sb, struct kqid *qid)
5594 const struct quota_format_ops *ops;
5596 if (!sb_has_quota_loaded(sb, qid->type))
5598 ops = sb_dqopt(sb)->ops[qid->type];
5599 if (!ops || !ops->get_next_id)
5601 return dquot_get_next_id(sb, qid);
5605 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5606 const char *dev_name, void *data)
5608 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5611 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
5612 static inline void register_as_ext2(void)
5614 int err = register_filesystem(&ext2_fs_type);
5617 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5620 static inline void unregister_as_ext2(void)
5622 unregister_filesystem(&ext2_fs_type);
5625 static inline int ext2_feature_set_ok(struct super_block *sb)
5627 if (ext4_has_unknown_ext2_incompat_features(sb))
5629 if (sb->s_flags & MS_RDONLY)
5631 if (ext4_has_unknown_ext2_ro_compat_features(sb))
5636 static inline void register_as_ext2(void) { }
5637 static inline void unregister_as_ext2(void) { }
5638 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5641 static inline void register_as_ext3(void)
5643 int err = register_filesystem(&ext3_fs_type);
5646 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5649 static inline void unregister_as_ext3(void)
5651 unregister_filesystem(&ext3_fs_type);
5654 static inline int ext3_feature_set_ok(struct super_block *sb)
5656 if (ext4_has_unknown_ext3_incompat_features(sb))
5658 if (!ext4_has_feature_journal(sb))
5660 if (sb->s_flags & MS_RDONLY)
5662 if (ext4_has_unknown_ext3_ro_compat_features(sb))
5667 static struct file_system_type ext4_fs_type = {
5668 .owner = THIS_MODULE,
5670 .mount = ext4_mount,
5671 .kill_sb = kill_block_super,
5672 .fs_flags = FS_REQUIRES_DEV,
5674 MODULE_ALIAS_FS("ext4");
5676 /* Shared across all ext4 file systems */
5677 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5679 static int __init ext4_init_fs(void)
5683 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
5684 ext4_li_info = NULL;
5685 mutex_init(&ext4_li_mtx);
5687 /* Build-time check for flags consistency */
5688 ext4_check_flag_values();
5690 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
5691 init_waitqueue_head(&ext4__ioend_wq[i]);
5693 err = ext4_init_es();
5697 err = ext4_init_pageio();
5701 err = ext4_init_system_zone();
5705 err = ext4_init_sysfs();
5709 err = ext4_init_mballoc();
5712 err = init_inodecache();
5717 err = register_filesystem(&ext4_fs_type);
5723 unregister_as_ext2();
5724 unregister_as_ext3();
5725 destroy_inodecache();
5727 ext4_exit_mballoc();
5731 ext4_exit_system_zone();
5740 static void __exit ext4_exit_fs(void)
5742 ext4_destroy_lazyinit_thread();
5743 unregister_as_ext2();
5744 unregister_as_ext3();
5745 unregister_filesystem(&ext4_fs_type);
5746 destroy_inodecache();
5747 ext4_exit_mballoc();
5749 ext4_exit_system_zone();
5754 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5755 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5756 MODULE_LICENSE("GPL");
5757 module_init(ext4_init_fs)
5758 module_exit(ext4_exit_fs)