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/dax.h>
41 #include <linux/cleancache.h>
42 #include <linux/uaccess.h>
44 #include <linux/kthread.h>
45 #include <linux/freezer.h>
48 #include "ext4_extents.h" /* Needed for trace points definition */
49 #include "ext4_jbd2.h"
55 #define CREATE_TRACE_POINTS
56 #include <trace/events/ext4.h>
58 static struct ext4_lazy_init *ext4_li_info;
59 static struct mutex ext4_li_mtx;
60 static struct ratelimit_state ext4_mount_msg_ratelimit;
62 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
63 unsigned long journal_devnum);
64 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
65 static int ext4_commit_super(struct super_block *sb, int sync);
66 static void ext4_mark_recovery_complete(struct super_block *sb,
67 struct ext4_super_block *es);
68 static void ext4_clear_journal_err(struct super_block *sb,
69 struct ext4_super_block *es);
70 static int ext4_sync_fs(struct super_block *sb, int wait);
71 static int ext4_remount(struct super_block *sb, int *flags, char *data);
72 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
73 static int ext4_unfreeze(struct super_block *sb);
74 static int ext4_freeze(struct super_block *sb);
75 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
76 const char *dev_name, void *data);
77 static inline int ext2_feature_set_ok(struct super_block *sb);
78 static inline int ext3_feature_set_ok(struct super_block *sb);
79 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
80 static void ext4_destroy_lazyinit_thread(void);
81 static void ext4_unregister_li_request(struct super_block *sb);
82 static void ext4_clear_request_list(void);
83 static struct inode *ext4_get_journal_inode(struct super_block *sb,
84 unsigned int journal_inum);
89 * Note the difference between i_mmap_sem (EXT4_I(inode)->i_mmap_sem) and
90 * i_mmap_rwsem (inode->i_mmap_rwsem)!
93 * mmap_sem -> sb_start_pagefault -> i_mmap_sem (r) -> transaction start ->
94 * page lock -> i_data_sem (rw)
96 * buffered write path:
97 * sb_start_write -> i_mutex -> mmap_sem
98 * sb_start_write -> i_mutex -> transaction start -> page lock ->
102 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) ->
103 * i_mmap_rwsem (w) -> page lock
104 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) ->
105 * transaction start -> i_data_sem (rw)
108 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) -> mmap_sem
109 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) ->
110 * transaction start -> i_data_sem (rw)
113 * transaction start -> page lock(s) -> i_data_sem (rw)
116 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
117 static struct file_system_type ext2_fs_type = {
118 .owner = THIS_MODULE,
121 .kill_sb = kill_block_super,
122 .fs_flags = FS_REQUIRES_DEV,
124 MODULE_ALIAS_FS("ext2");
125 MODULE_ALIAS("ext2");
126 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
128 #define IS_EXT2_SB(sb) (0)
132 static struct file_system_type ext3_fs_type = {
133 .owner = THIS_MODULE,
136 .kill_sb = kill_block_super,
137 .fs_flags = FS_REQUIRES_DEV,
139 MODULE_ALIAS_FS("ext3");
140 MODULE_ALIAS("ext3");
141 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
143 static int ext4_verify_csum_type(struct super_block *sb,
144 struct ext4_super_block *es)
146 if (!ext4_has_feature_metadata_csum(sb))
149 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
152 static __le32 ext4_superblock_csum(struct super_block *sb,
153 struct ext4_super_block *es)
155 struct ext4_sb_info *sbi = EXT4_SB(sb);
156 int offset = offsetof(struct ext4_super_block, s_checksum);
159 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
161 return cpu_to_le32(csum);
164 static int ext4_superblock_csum_verify(struct super_block *sb,
165 struct ext4_super_block *es)
167 if (!ext4_has_metadata_csum(sb))
170 return es->s_checksum == ext4_superblock_csum(sb, es);
173 void ext4_superblock_csum_set(struct super_block *sb)
175 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
177 if (!ext4_has_metadata_csum(sb))
180 es->s_checksum = ext4_superblock_csum(sb, es);
183 void *ext4_kvmalloc(size_t size, gfp_t flags)
187 ret = kmalloc(size, flags | __GFP_NOWARN);
189 ret = __vmalloc(size, flags, PAGE_KERNEL);
193 void *ext4_kvzalloc(size_t size, gfp_t flags)
197 ret = kzalloc(size, flags | __GFP_NOWARN);
199 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
203 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
204 struct ext4_group_desc *bg)
206 return le32_to_cpu(bg->bg_block_bitmap_lo) |
207 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
208 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
211 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
212 struct ext4_group_desc *bg)
214 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
215 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
216 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
219 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
220 struct ext4_group_desc *bg)
222 return le32_to_cpu(bg->bg_inode_table_lo) |
223 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
224 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
227 __u32 ext4_free_group_clusters(struct super_block *sb,
228 struct ext4_group_desc *bg)
230 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
231 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
232 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
235 __u32 ext4_free_inodes_count(struct super_block *sb,
236 struct ext4_group_desc *bg)
238 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
239 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
240 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
243 __u32 ext4_used_dirs_count(struct super_block *sb,
244 struct ext4_group_desc *bg)
246 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
247 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
248 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
251 __u32 ext4_itable_unused_count(struct super_block *sb,
252 struct ext4_group_desc *bg)
254 return le16_to_cpu(bg->bg_itable_unused_lo) |
255 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
256 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
259 void ext4_block_bitmap_set(struct super_block *sb,
260 struct ext4_group_desc *bg, ext4_fsblk_t blk)
262 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
263 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
264 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
267 void ext4_inode_bitmap_set(struct super_block *sb,
268 struct ext4_group_desc *bg, ext4_fsblk_t blk)
270 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
271 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
272 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
275 void ext4_inode_table_set(struct super_block *sb,
276 struct ext4_group_desc *bg, ext4_fsblk_t blk)
278 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
279 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
280 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
283 void ext4_free_group_clusters_set(struct super_block *sb,
284 struct ext4_group_desc *bg, __u32 count)
286 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
287 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
288 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
291 void ext4_free_inodes_set(struct super_block *sb,
292 struct ext4_group_desc *bg, __u32 count)
294 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
295 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
296 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
299 void ext4_used_dirs_set(struct super_block *sb,
300 struct ext4_group_desc *bg, __u32 count)
302 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
303 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
304 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
307 void ext4_itable_unused_set(struct super_block *sb,
308 struct ext4_group_desc *bg, __u32 count)
310 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
311 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
312 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
316 static void __save_error_info(struct super_block *sb, const char *func,
319 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
321 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
322 if (bdev_read_only(sb->s_bdev))
324 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
325 es->s_last_error_time = cpu_to_le32(get_seconds());
326 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
327 es->s_last_error_line = cpu_to_le32(line);
328 if (!es->s_first_error_time) {
329 es->s_first_error_time = es->s_last_error_time;
330 strncpy(es->s_first_error_func, func,
331 sizeof(es->s_first_error_func));
332 es->s_first_error_line = cpu_to_le32(line);
333 es->s_first_error_ino = es->s_last_error_ino;
334 es->s_first_error_block = es->s_last_error_block;
337 * Start the daily error reporting function if it hasn't been
340 if (!es->s_error_count)
341 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
342 le32_add_cpu(&es->s_error_count, 1);
345 static void save_error_info(struct super_block *sb, const char *func,
348 __save_error_info(sb, func, line);
349 ext4_commit_super(sb, 1);
353 * The del_gendisk() function uninitializes the disk-specific data
354 * structures, including the bdi structure, without telling anyone
355 * else. Once this happens, any attempt to call mark_buffer_dirty()
356 * (for example, by ext4_commit_super), will cause a kernel OOPS.
357 * This is a kludge to prevent these oops until we can put in a proper
358 * hook in del_gendisk() to inform the VFS and file system layers.
360 static int block_device_ejected(struct super_block *sb)
362 struct inode *bd_inode = sb->s_bdev->bd_inode;
363 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
365 return bdi->dev == NULL;
368 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
370 struct super_block *sb = journal->j_private;
371 struct ext4_sb_info *sbi = EXT4_SB(sb);
372 int error = is_journal_aborted(journal);
373 struct ext4_journal_cb_entry *jce;
375 BUG_ON(txn->t_state == T_FINISHED);
377 ext4_process_freed_data(sb, txn->t_tid);
379 spin_lock(&sbi->s_md_lock);
380 while (!list_empty(&txn->t_private_list)) {
381 jce = list_entry(txn->t_private_list.next,
382 struct ext4_journal_cb_entry, jce_list);
383 list_del_init(&jce->jce_list);
384 spin_unlock(&sbi->s_md_lock);
385 jce->jce_func(sb, jce, error);
386 spin_lock(&sbi->s_md_lock);
388 spin_unlock(&sbi->s_md_lock);
391 /* Deal with the reporting of failure conditions on a filesystem such as
392 * inconsistencies detected or read IO failures.
394 * On ext2, we can store the error state of the filesystem in the
395 * superblock. That is not possible on ext4, because we may have other
396 * write ordering constraints on the superblock which prevent us from
397 * writing it out straight away; and given that the journal is about to
398 * be aborted, we can't rely on the current, or future, transactions to
399 * write out the superblock safely.
401 * We'll just use the jbd2_journal_abort() error code to record an error in
402 * the journal instead. On recovery, the journal will complain about
403 * that error until we've noted it down and cleared it.
406 static void ext4_handle_error(struct super_block *sb)
408 if (sb->s_flags & MS_RDONLY)
411 if (!test_opt(sb, ERRORS_CONT)) {
412 journal_t *journal = EXT4_SB(sb)->s_journal;
414 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
416 jbd2_journal_abort(journal, -EIO);
418 if (test_opt(sb, ERRORS_RO)) {
419 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
421 * Make sure updated value of ->s_mount_flags will be visible
422 * before ->s_flags update
425 sb->s_flags |= MS_RDONLY;
427 if (test_opt(sb, ERRORS_PANIC)) {
428 if (EXT4_SB(sb)->s_journal &&
429 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
431 panic("EXT4-fs (device %s): panic forced after error\n",
436 #define ext4_error_ratelimit(sb) \
437 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
440 void __ext4_error(struct super_block *sb, const char *function,
441 unsigned int line, const char *fmt, ...)
443 struct va_format vaf;
446 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
449 if (ext4_error_ratelimit(sb)) {
454 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
455 sb->s_id, function, line, current->comm, &vaf);
458 save_error_info(sb, function, line);
459 ext4_handle_error(sb);
462 void __ext4_error_inode(struct inode *inode, const char *function,
463 unsigned int line, ext4_fsblk_t block,
464 const char *fmt, ...)
467 struct va_format vaf;
468 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
470 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
473 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
474 es->s_last_error_block = cpu_to_le64(block);
475 if (ext4_error_ratelimit(inode->i_sb)) {
480 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
481 "inode #%lu: block %llu: comm %s: %pV\n",
482 inode->i_sb->s_id, function, line, inode->i_ino,
483 block, current->comm, &vaf);
485 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
486 "inode #%lu: comm %s: %pV\n",
487 inode->i_sb->s_id, function, line, inode->i_ino,
488 current->comm, &vaf);
491 save_error_info(inode->i_sb, function, line);
492 ext4_handle_error(inode->i_sb);
495 void __ext4_error_file(struct file *file, const char *function,
496 unsigned int line, ext4_fsblk_t block,
497 const char *fmt, ...)
500 struct va_format vaf;
501 struct ext4_super_block *es;
502 struct inode *inode = file_inode(file);
503 char pathname[80], *path;
505 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
508 es = EXT4_SB(inode->i_sb)->s_es;
509 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
510 if (ext4_error_ratelimit(inode->i_sb)) {
511 path = file_path(file, pathname, sizeof(pathname));
519 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
520 "block %llu: comm %s: path %s: %pV\n",
521 inode->i_sb->s_id, function, line, inode->i_ino,
522 block, current->comm, path, &vaf);
525 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
526 "comm %s: path %s: %pV\n",
527 inode->i_sb->s_id, function, line, inode->i_ino,
528 current->comm, path, &vaf);
531 save_error_info(inode->i_sb, function, line);
532 ext4_handle_error(inode->i_sb);
535 const char *ext4_decode_error(struct super_block *sb, int errno,
542 errstr = "Corrupt filesystem";
545 errstr = "Filesystem failed CRC";
548 errstr = "IO failure";
551 errstr = "Out of memory";
554 if (!sb || (EXT4_SB(sb)->s_journal &&
555 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
556 errstr = "Journal has aborted";
558 errstr = "Readonly filesystem";
561 /* If the caller passed in an extra buffer for unknown
562 * errors, textualise them now. Else we just return
565 /* Check for truncated error codes... */
566 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
575 /* __ext4_std_error decodes expected errors from journaling functions
576 * automatically and invokes the appropriate error response. */
578 void __ext4_std_error(struct super_block *sb, const char *function,
579 unsigned int line, int errno)
584 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
587 /* Special case: if the error is EROFS, and we're not already
588 * inside a transaction, then there's really no point in logging
590 if (errno == -EROFS && journal_current_handle() == NULL &&
591 (sb->s_flags & MS_RDONLY))
594 if (ext4_error_ratelimit(sb)) {
595 errstr = ext4_decode_error(sb, errno, nbuf);
596 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
597 sb->s_id, function, line, errstr);
600 save_error_info(sb, function, line);
601 ext4_handle_error(sb);
605 * ext4_abort is a much stronger failure handler than ext4_error. The
606 * abort function may be used to deal with unrecoverable failures such
607 * as journal IO errors or ENOMEM at a critical moment in log management.
609 * We unconditionally force the filesystem into an ABORT|READONLY state,
610 * unless the error response on the fs has been set to panic in which
611 * case we take the easy way out and panic immediately.
614 void __ext4_abort(struct super_block *sb, const char *function,
615 unsigned int line, const char *fmt, ...)
617 struct va_format vaf;
620 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
623 save_error_info(sb, function, line);
627 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: %pV\n",
628 sb->s_id, function, line, &vaf);
631 if ((sb->s_flags & MS_RDONLY) == 0) {
632 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
633 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
635 * Make sure updated value of ->s_mount_flags will be visible
636 * before ->s_flags update
639 sb->s_flags |= MS_RDONLY;
640 if (EXT4_SB(sb)->s_journal)
641 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
642 save_error_info(sb, function, line);
644 if (test_opt(sb, ERRORS_PANIC)) {
645 if (EXT4_SB(sb)->s_journal &&
646 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
648 panic("EXT4-fs panic from previous error\n");
652 void __ext4_msg(struct super_block *sb,
653 const char *prefix, const char *fmt, ...)
655 struct va_format vaf;
658 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
664 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
668 #define ext4_warning_ratelimit(sb) \
669 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \
672 void __ext4_warning(struct super_block *sb, const char *function,
673 unsigned int line, const char *fmt, ...)
675 struct va_format vaf;
678 if (!ext4_warning_ratelimit(sb))
684 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
685 sb->s_id, function, line, &vaf);
689 void __ext4_warning_inode(const struct inode *inode, const char *function,
690 unsigned int line, const char *fmt, ...)
692 struct va_format vaf;
695 if (!ext4_warning_ratelimit(inode->i_sb))
701 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
702 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
703 function, line, inode->i_ino, current->comm, &vaf);
707 void __ext4_grp_locked_error(const char *function, unsigned int line,
708 struct super_block *sb, ext4_group_t grp,
709 unsigned long ino, ext4_fsblk_t block,
710 const char *fmt, ...)
714 struct va_format vaf;
716 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
718 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
721 es->s_last_error_ino = cpu_to_le32(ino);
722 es->s_last_error_block = cpu_to_le64(block);
723 __save_error_info(sb, function, line);
725 if (ext4_error_ratelimit(sb)) {
729 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
730 sb->s_id, function, line, grp);
732 printk(KERN_CONT "inode %lu: ", ino);
734 printk(KERN_CONT "block %llu:",
735 (unsigned long long) block);
736 printk(KERN_CONT "%pV\n", &vaf);
740 if (test_opt(sb, ERRORS_CONT)) {
741 ext4_commit_super(sb, 0);
745 ext4_unlock_group(sb, grp);
746 ext4_handle_error(sb);
748 * We only get here in the ERRORS_RO case; relocking the group
749 * may be dangerous, but nothing bad will happen since the
750 * filesystem will have already been marked read/only and the
751 * journal has been aborted. We return 1 as a hint to callers
752 * who might what to use the return value from
753 * ext4_grp_locked_error() to distinguish between the
754 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
755 * aggressively from the ext4 function in question, with a
756 * more appropriate error code.
758 ext4_lock_group(sb, grp);
762 void ext4_update_dynamic_rev(struct super_block *sb)
764 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
766 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
770 "updating to rev %d because of new feature flag, "
771 "running e2fsck is recommended",
774 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
775 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
776 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
777 /* leave es->s_feature_*compat flags alone */
778 /* es->s_uuid will be set by e2fsck if empty */
781 * The rest of the superblock fields should be zero, and if not it
782 * means they are likely already in use, so leave them alone. We
783 * can leave it up to e2fsck to clean up any inconsistencies there.
788 * Open the external journal device
790 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
792 struct block_device *bdev;
793 char b[BDEVNAME_SIZE];
795 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
801 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
802 __bdevname(dev, b), PTR_ERR(bdev));
807 * Release the journal device
809 static void ext4_blkdev_put(struct block_device *bdev)
811 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
814 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
816 struct block_device *bdev;
817 bdev = sbi->journal_bdev;
819 ext4_blkdev_put(bdev);
820 sbi->journal_bdev = NULL;
824 static inline struct inode *orphan_list_entry(struct list_head *l)
826 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
829 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
833 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
834 le32_to_cpu(sbi->s_es->s_last_orphan));
836 printk(KERN_ERR "sb_info orphan list:\n");
837 list_for_each(l, &sbi->s_orphan) {
838 struct inode *inode = orphan_list_entry(l);
840 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
841 inode->i_sb->s_id, inode->i_ino, inode,
842 inode->i_mode, inode->i_nlink,
848 static int ext4_quota_off(struct super_block *sb, int type);
850 static inline void ext4_quota_off_umount(struct super_block *sb)
854 /* Use our quota_off function to clear inode flags etc. */
855 for (type = 0; type < EXT4_MAXQUOTAS; type++)
856 ext4_quota_off(sb, type);
859 static inline void ext4_quota_off_umount(struct super_block *sb)
864 static void ext4_put_super(struct super_block *sb)
866 struct ext4_sb_info *sbi = EXT4_SB(sb);
867 struct ext4_super_block *es = sbi->s_es;
871 ext4_unregister_li_request(sb);
872 ext4_quota_off_umount(sb);
874 flush_workqueue(sbi->rsv_conversion_wq);
875 destroy_workqueue(sbi->rsv_conversion_wq);
877 if (sbi->s_journal) {
878 aborted = is_journal_aborted(sbi->s_journal);
879 err = jbd2_journal_destroy(sbi->s_journal);
880 sbi->s_journal = NULL;
881 if ((err < 0) && !aborted)
882 ext4_abort(sb, "Couldn't clean up the journal");
885 ext4_unregister_sysfs(sb);
886 ext4_es_unregister_shrinker(sbi);
887 del_timer_sync(&sbi->s_err_report);
888 ext4_release_system_zone(sb);
890 ext4_ext_release(sb);
892 if (!(sb->s_flags & MS_RDONLY) && !aborted) {
893 ext4_clear_feature_journal_needs_recovery(sb);
894 es->s_state = cpu_to_le16(sbi->s_mount_state);
896 if (!(sb->s_flags & MS_RDONLY))
897 ext4_commit_super(sb, 1);
899 for (i = 0; i < sbi->s_gdb_count; i++)
900 brelse(sbi->s_group_desc[i]);
901 kvfree(sbi->s_group_desc);
902 kvfree(sbi->s_flex_groups);
903 percpu_counter_destroy(&sbi->s_freeclusters_counter);
904 percpu_counter_destroy(&sbi->s_freeinodes_counter);
905 percpu_counter_destroy(&sbi->s_dirs_counter);
906 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
907 percpu_free_rwsem(&sbi->s_journal_flag_rwsem);
909 for (i = 0; i < EXT4_MAXQUOTAS; i++)
910 kfree(sbi->s_qf_names[i]);
913 /* Debugging code just in case the in-memory inode orphan list
914 * isn't empty. The on-disk one can be non-empty if we've
915 * detected an error and taken the fs readonly, but the
916 * in-memory list had better be clean by this point. */
917 if (!list_empty(&sbi->s_orphan))
918 dump_orphan_list(sb, sbi);
919 J_ASSERT(list_empty(&sbi->s_orphan));
921 sync_blockdev(sb->s_bdev);
922 invalidate_bdev(sb->s_bdev);
923 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
925 * Invalidate the journal device's buffers. We don't want them
926 * floating about in memory - the physical journal device may
927 * hotswapped, and it breaks the `ro-after' testing code.
929 sync_blockdev(sbi->journal_bdev);
930 invalidate_bdev(sbi->journal_bdev);
931 ext4_blkdev_remove(sbi);
933 if (sbi->s_ea_inode_cache) {
934 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
935 sbi->s_ea_inode_cache = NULL;
937 if (sbi->s_ea_block_cache) {
938 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
939 sbi->s_ea_block_cache = NULL;
942 kthread_stop(sbi->s_mmp_tsk);
944 sb->s_fs_info = NULL;
946 * Now that we are completely done shutting down the
947 * superblock, we need to actually destroy the kobject.
949 kobject_put(&sbi->s_kobj);
950 wait_for_completion(&sbi->s_kobj_unregister);
951 if (sbi->s_chksum_driver)
952 crypto_free_shash(sbi->s_chksum_driver);
953 kfree(sbi->s_blockgroup_lock);
957 static struct kmem_cache *ext4_inode_cachep;
960 * Called inside transaction, so use GFP_NOFS
962 static struct inode *ext4_alloc_inode(struct super_block *sb)
964 struct ext4_inode_info *ei;
966 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
970 ei->vfs_inode.i_version = 1;
971 spin_lock_init(&ei->i_raw_lock);
972 INIT_LIST_HEAD(&ei->i_prealloc_list);
973 spin_lock_init(&ei->i_prealloc_lock);
974 ext4_es_init_tree(&ei->i_es_tree);
975 rwlock_init(&ei->i_es_lock);
976 INIT_LIST_HEAD(&ei->i_es_list);
979 ei->i_es_shrink_lblk = 0;
980 ei->i_reserved_data_blocks = 0;
981 ei->i_da_metadata_calc_len = 0;
982 ei->i_da_metadata_calc_last_lblock = 0;
983 spin_lock_init(&(ei->i_block_reservation_lock));
985 ei->i_reserved_quota = 0;
986 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
989 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
990 spin_lock_init(&ei->i_completed_io_lock);
992 ei->i_datasync_tid = 0;
993 atomic_set(&ei->i_unwritten, 0);
994 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
995 return &ei->vfs_inode;
998 static int ext4_drop_inode(struct inode *inode)
1000 int drop = generic_drop_inode(inode);
1002 trace_ext4_drop_inode(inode, drop);
1006 static void ext4_i_callback(struct rcu_head *head)
1008 struct inode *inode = container_of(head, struct inode, i_rcu);
1009 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
1012 static void ext4_destroy_inode(struct inode *inode)
1014 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
1015 ext4_msg(inode->i_sb, KERN_ERR,
1016 "Inode %lu (%p): orphan list check failed!",
1017 inode->i_ino, EXT4_I(inode));
1018 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1019 EXT4_I(inode), sizeof(struct ext4_inode_info),
1023 call_rcu(&inode->i_rcu, ext4_i_callback);
1026 static void init_once(void *foo)
1028 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
1030 INIT_LIST_HEAD(&ei->i_orphan);
1031 init_rwsem(&ei->xattr_sem);
1032 init_rwsem(&ei->i_data_sem);
1033 init_rwsem(&ei->i_mmap_sem);
1034 inode_init_once(&ei->vfs_inode);
1037 static int __init init_inodecache(void)
1039 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
1040 sizeof(struct ext4_inode_info),
1041 0, (SLAB_RECLAIM_ACCOUNT|
1042 SLAB_MEM_SPREAD|SLAB_ACCOUNT),
1044 if (ext4_inode_cachep == NULL)
1049 static void destroy_inodecache(void)
1052 * Make sure all delayed rcu free inodes are flushed before we
1056 kmem_cache_destroy(ext4_inode_cachep);
1059 void ext4_clear_inode(struct inode *inode)
1061 invalidate_inode_buffers(inode);
1064 ext4_discard_preallocations(inode);
1065 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
1066 if (EXT4_I(inode)->jinode) {
1067 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1068 EXT4_I(inode)->jinode);
1069 jbd2_free_inode(EXT4_I(inode)->jinode);
1070 EXT4_I(inode)->jinode = NULL;
1072 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1073 fscrypt_put_encryption_info(inode, NULL);
1077 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1078 u64 ino, u32 generation)
1080 struct inode *inode;
1082 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1083 return ERR_PTR(-ESTALE);
1084 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1085 return ERR_PTR(-ESTALE);
1087 /* iget isn't really right if the inode is currently unallocated!!
1089 * ext4_read_inode will return a bad_inode if the inode had been
1090 * deleted, so we should be safe.
1092 * Currently we don't know the generation for parent directory, so
1093 * a generation of 0 means "accept any"
1095 inode = ext4_iget_normal(sb, ino);
1097 return ERR_CAST(inode);
1098 if (generation && inode->i_generation != generation) {
1100 return ERR_PTR(-ESTALE);
1106 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1107 int fh_len, int fh_type)
1109 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1110 ext4_nfs_get_inode);
1113 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1114 int fh_len, int fh_type)
1116 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1117 ext4_nfs_get_inode);
1121 * Try to release metadata pages (indirect blocks, directories) which are
1122 * mapped via the block device. Since these pages could have journal heads
1123 * which would prevent try_to_free_buffers() from freeing them, we must use
1124 * jbd2 layer's try_to_free_buffers() function to release them.
1126 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1129 journal_t *journal = EXT4_SB(sb)->s_journal;
1131 WARN_ON(PageChecked(page));
1132 if (!page_has_buffers(page))
1135 return jbd2_journal_try_to_free_buffers(journal, page,
1136 wait & ~__GFP_DIRECT_RECLAIM);
1137 return try_to_free_buffers(page);
1140 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1141 static int ext4_get_context(struct inode *inode, void *ctx, size_t len)
1143 return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1144 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len);
1147 static int ext4_set_context(struct inode *inode, const void *ctx, size_t len,
1150 handle_t *handle = fs_data;
1151 int res, res2, credits, retries = 0;
1154 * Encrypting the root directory is not allowed because e2fsck expects
1155 * lost+found to exist and be unencrypted, and encrypting the root
1156 * directory would imply encrypting the lost+found directory as well as
1157 * the filename "lost+found" itself.
1159 if (inode->i_ino == EXT4_ROOT_INO)
1162 res = ext4_convert_inline_data(inode);
1167 * If a journal handle was specified, then the encryption context is
1168 * being set on a new inode via inheritance and is part of a larger
1169 * transaction to create the inode. Otherwise the encryption context is
1170 * being set on an existing inode in its own transaction. Only in the
1171 * latter case should the "retry on ENOSPC" logic be used.
1175 res = ext4_xattr_set_handle(handle, inode,
1176 EXT4_XATTR_INDEX_ENCRYPTION,
1177 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1180 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1181 ext4_clear_inode_state(inode,
1182 EXT4_STATE_MAY_INLINE_DATA);
1184 * Update inode->i_flags - e.g. S_DAX may get disabled
1186 ext4_set_inode_flags(inode);
1191 res = dquot_initialize(inode);
1195 res = ext4_xattr_set_credits(inode, len, false /* is_create */,
1200 handle = ext4_journal_start(inode, EXT4_HT_MISC, credits);
1202 return PTR_ERR(handle);
1204 res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION,
1205 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1208 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1209 /* Update inode->i_flags - e.g. S_DAX may get disabled */
1210 ext4_set_inode_flags(inode);
1211 res = ext4_mark_inode_dirty(handle, inode);
1213 EXT4_ERROR_INODE(inode, "Failed to mark inode dirty");
1215 res2 = ext4_journal_stop(handle);
1217 if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
1224 static bool ext4_dummy_context(struct inode *inode)
1226 return DUMMY_ENCRYPTION_ENABLED(EXT4_SB(inode->i_sb));
1229 static unsigned ext4_max_namelen(struct inode *inode)
1231 return S_ISLNK(inode->i_mode) ? inode->i_sb->s_blocksize :
1235 static const struct fscrypt_operations ext4_cryptops = {
1236 .key_prefix = "ext4:",
1237 .get_context = ext4_get_context,
1238 .set_context = ext4_set_context,
1239 .dummy_context = ext4_dummy_context,
1240 .is_encrypted = ext4_encrypted_inode,
1241 .empty_dir = ext4_empty_dir,
1242 .max_namelen = ext4_max_namelen,
1245 static const struct fscrypt_operations ext4_cryptops = {
1246 .is_encrypted = ext4_encrypted_inode,
1251 static const char * const quotatypes[] = INITQFNAMES;
1252 #define QTYPE2NAME(t) (quotatypes[t])
1254 static int ext4_write_dquot(struct dquot *dquot);
1255 static int ext4_acquire_dquot(struct dquot *dquot);
1256 static int ext4_release_dquot(struct dquot *dquot);
1257 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1258 static int ext4_write_info(struct super_block *sb, int type);
1259 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1260 const struct path *path);
1261 static int ext4_quota_on_mount(struct super_block *sb, int type);
1262 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1263 size_t len, loff_t off);
1264 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1265 const char *data, size_t len, loff_t off);
1266 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1267 unsigned int flags);
1268 static int ext4_enable_quotas(struct super_block *sb);
1269 static int ext4_get_next_id(struct super_block *sb, struct kqid *qid);
1271 static struct dquot **ext4_get_dquots(struct inode *inode)
1273 return EXT4_I(inode)->i_dquot;
1276 static const struct dquot_operations ext4_quota_operations = {
1277 .get_reserved_space = ext4_get_reserved_space,
1278 .write_dquot = ext4_write_dquot,
1279 .acquire_dquot = ext4_acquire_dquot,
1280 .release_dquot = ext4_release_dquot,
1281 .mark_dirty = ext4_mark_dquot_dirty,
1282 .write_info = ext4_write_info,
1283 .alloc_dquot = dquot_alloc,
1284 .destroy_dquot = dquot_destroy,
1285 .get_projid = ext4_get_projid,
1286 .get_inode_usage = ext4_get_inode_usage,
1287 .get_next_id = ext4_get_next_id,
1290 static const struct quotactl_ops ext4_qctl_operations = {
1291 .quota_on = ext4_quota_on,
1292 .quota_off = ext4_quota_off,
1293 .quota_sync = dquot_quota_sync,
1294 .get_state = dquot_get_state,
1295 .set_info = dquot_set_dqinfo,
1296 .get_dqblk = dquot_get_dqblk,
1297 .set_dqblk = dquot_set_dqblk,
1298 .get_nextdqblk = dquot_get_next_dqblk,
1302 static const struct super_operations ext4_sops = {
1303 .alloc_inode = ext4_alloc_inode,
1304 .destroy_inode = ext4_destroy_inode,
1305 .write_inode = ext4_write_inode,
1306 .dirty_inode = ext4_dirty_inode,
1307 .drop_inode = ext4_drop_inode,
1308 .evict_inode = ext4_evict_inode,
1309 .put_super = ext4_put_super,
1310 .sync_fs = ext4_sync_fs,
1311 .freeze_fs = ext4_freeze,
1312 .unfreeze_fs = ext4_unfreeze,
1313 .statfs = ext4_statfs,
1314 .remount_fs = ext4_remount,
1315 .show_options = ext4_show_options,
1317 .quota_read = ext4_quota_read,
1318 .quota_write = ext4_quota_write,
1319 .get_dquots = ext4_get_dquots,
1321 .bdev_try_to_free_page = bdev_try_to_free_page,
1324 static const struct export_operations ext4_export_ops = {
1325 .fh_to_dentry = ext4_fh_to_dentry,
1326 .fh_to_parent = ext4_fh_to_parent,
1327 .get_parent = ext4_get_parent,
1331 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1332 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1333 Opt_nouid32, Opt_debug, Opt_removed,
1334 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1335 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1336 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1337 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1338 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1339 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1340 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1341 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1342 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1343 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version, Opt_dax,
1344 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1345 Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize,
1346 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1347 Opt_inode_readahead_blks, Opt_journal_ioprio,
1348 Opt_dioread_nolock, Opt_dioread_lock,
1349 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1350 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache,
1353 static const match_table_t tokens = {
1354 {Opt_bsd_df, "bsddf"},
1355 {Opt_minix_df, "minixdf"},
1356 {Opt_grpid, "grpid"},
1357 {Opt_grpid, "bsdgroups"},
1358 {Opt_nogrpid, "nogrpid"},
1359 {Opt_nogrpid, "sysvgroups"},
1360 {Opt_resgid, "resgid=%u"},
1361 {Opt_resuid, "resuid=%u"},
1363 {Opt_err_cont, "errors=continue"},
1364 {Opt_err_panic, "errors=panic"},
1365 {Opt_err_ro, "errors=remount-ro"},
1366 {Opt_nouid32, "nouid32"},
1367 {Opt_debug, "debug"},
1368 {Opt_removed, "oldalloc"},
1369 {Opt_removed, "orlov"},
1370 {Opt_user_xattr, "user_xattr"},
1371 {Opt_nouser_xattr, "nouser_xattr"},
1373 {Opt_noacl, "noacl"},
1374 {Opt_noload, "norecovery"},
1375 {Opt_noload, "noload"},
1376 {Opt_removed, "nobh"},
1377 {Opt_removed, "bh"},
1378 {Opt_commit, "commit=%u"},
1379 {Opt_min_batch_time, "min_batch_time=%u"},
1380 {Opt_max_batch_time, "max_batch_time=%u"},
1381 {Opt_journal_dev, "journal_dev=%u"},
1382 {Opt_journal_path, "journal_path=%s"},
1383 {Opt_journal_checksum, "journal_checksum"},
1384 {Opt_nojournal_checksum, "nojournal_checksum"},
1385 {Opt_journal_async_commit, "journal_async_commit"},
1386 {Opt_abort, "abort"},
1387 {Opt_data_journal, "data=journal"},
1388 {Opt_data_ordered, "data=ordered"},
1389 {Opt_data_writeback, "data=writeback"},
1390 {Opt_data_err_abort, "data_err=abort"},
1391 {Opt_data_err_ignore, "data_err=ignore"},
1392 {Opt_offusrjquota, "usrjquota="},
1393 {Opt_usrjquota, "usrjquota=%s"},
1394 {Opt_offgrpjquota, "grpjquota="},
1395 {Opt_grpjquota, "grpjquota=%s"},
1396 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1397 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1398 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1399 {Opt_grpquota, "grpquota"},
1400 {Opt_noquota, "noquota"},
1401 {Opt_quota, "quota"},
1402 {Opt_usrquota, "usrquota"},
1403 {Opt_prjquota, "prjquota"},
1404 {Opt_barrier, "barrier=%u"},
1405 {Opt_barrier, "barrier"},
1406 {Opt_nobarrier, "nobarrier"},
1407 {Opt_i_version, "i_version"},
1409 {Opt_stripe, "stripe=%u"},
1410 {Opt_delalloc, "delalloc"},
1411 {Opt_lazytime, "lazytime"},
1412 {Opt_nolazytime, "nolazytime"},
1413 {Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"},
1414 {Opt_nodelalloc, "nodelalloc"},
1415 {Opt_removed, "mblk_io_submit"},
1416 {Opt_removed, "nomblk_io_submit"},
1417 {Opt_block_validity, "block_validity"},
1418 {Opt_noblock_validity, "noblock_validity"},
1419 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1420 {Opt_journal_ioprio, "journal_ioprio=%u"},
1421 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1422 {Opt_auto_da_alloc, "auto_da_alloc"},
1423 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1424 {Opt_dioread_nolock, "dioread_nolock"},
1425 {Opt_dioread_lock, "dioread_lock"},
1426 {Opt_discard, "discard"},
1427 {Opt_nodiscard, "nodiscard"},
1428 {Opt_init_itable, "init_itable=%u"},
1429 {Opt_init_itable, "init_itable"},
1430 {Opt_noinit_itable, "noinit_itable"},
1431 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1432 {Opt_test_dummy_encryption, "test_dummy_encryption"},
1433 {Opt_nombcache, "nombcache"},
1434 {Opt_nombcache, "no_mbcache"}, /* for backward compatibility */
1435 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1436 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1437 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1438 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1439 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1443 static ext4_fsblk_t get_sb_block(void **data)
1445 ext4_fsblk_t sb_block;
1446 char *options = (char *) *data;
1448 if (!options || strncmp(options, "sb=", 3) != 0)
1449 return 1; /* Default location */
1452 /* TODO: use simple_strtoll with >32bit ext4 */
1453 sb_block = simple_strtoul(options, &options, 0);
1454 if (*options && *options != ',') {
1455 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1459 if (*options == ',')
1461 *data = (void *) options;
1466 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1467 static const char deprecated_msg[] =
1468 "Mount option \"%s\" will be removed by %s\n"
1469 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1472 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1474 struct ext4_sb_info *sbi = EXT4_SB(sb);
1478 if (sb_any_quota_loaded(sb) &&
1479 !sbi->s_qf_names[qtype]) {
1480 ext4_msg(sb, KERN_ERR,
1481 "Cannot change journaled "
1482 "quota options when quota turned on");
1485 if (ext4_has_feature_quota(sb)) {
1486 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1487 "ignored when QUOTA feature is enabled");
1490 qname = match_strdup(args);
1492 ext4_msg(sb, KERN_ERR,
1493 "Not enough memory for storing quotafile name");
1496 if (sbi->s_qf_names[qtype]) {
1497 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1500 ext4_msg(sb, KERN_ERR,
1501 "%s quota file already specified",
1505 if (strchr(qname, '/')) {
1506 ext4_msg(sb, KERN_ERR,
1507 "quotafile must be on filesystem root");
1510 sbi->s_qf_names[qtype] = qname;
1518 static int clear_qf_name(struct super_block *sb, int qtype)
1521 struct ext4_sb_info *sbi = EXT4_SB(sb);
1523 if (sb_any_quota_loaded(sb) &&
1524 sbi->s_qf_names[qtype]) {
1525 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1526 " when quota turned on");
1529 kfree(sbi->s_qf_names[qtype]);
1530 sbi->s_qf_names[qtype] = NULL;
1535 #define MOPT_SET 0x0001
1536 #define MOPT_CLEAR 0x0002
1537 #define MOPT_NOSUPPORT 0x0004
1538 #define MOPT_EXPLICIT 0x0008
1539 #define MOPT_CLEAR_ERR 0x0010
1540 #define MOPT_GTE0 0x0020
1543 #define MOPT_QFMT 0x0040
1545 #define MOPT_Q MOPT_NOSUPPORT
1546 #define MOPT_QFMT MOPT_NOSUPPORT
1548 #define MOPT_DATAJ 0x0080
1549 #define MOPT_NO_EXT2 0x0100
1550 #define MOPT_NO_EXT3 0x0200
1551 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1552 #define MOPT_STRING 0x0400
1554 static const struct mount_opts {
1558 } ext4_mount_opts[] = {
1559 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1560 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1561 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1562 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1563 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1564 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1565 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1566 MOPT_EXT4_ONLY | MOPT_SET},
1567 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1568 MOPT_EXT4_ONLY | MOPT_CLEAR},
1569 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1570 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1571 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1572 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1573 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1574 MOPT_EXT4_ONLY | MOPT_CLEAR},
1575 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1576 MOPT_EXT4_ONLY | MOPT_CLEAR},
1577 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1578 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1579 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1580 EXT4_MOUNT_JOURNAL_CHECKSUM),
1581 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1582 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1583 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1584 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1585 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1586 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1588 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1590 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1591 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1592 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1593 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1594 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1595 {Opt_commit, 0, MOPT_GTE0},
1596 {Opt_max_batch_time, 0, MOPT_GTE0},
1597 {Opt_min_batch_time, 0, MOPT_GTE0},
1598 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1599 {Opt_init_itable, 0, MOPT_GTE0},
1600 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
1601 {Opt_stripe, 0, MOPT_GTE0},
1602 {Opt_resuid, 0, MOPT_GTE0},
1603 {Opt_resgid, 0, MOPT_GTE0},
1604 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1605 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1606 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1607 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1608 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1609 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1610 MOPT_NO_EXT2 | MOPT_DATAJ},
1611 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1612 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1613 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1614 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1615 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1617 {Opt_acl, 0, MOPT_NOSUPPORT},
1618 {Opt_noacl, 0, MOPT_NOSUPPORT},
1620 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1621 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1622 {Opt_debug_want_extra_isize, 0, MOPT_GTE0},
1623 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1624 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1626 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1628 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1630 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1631 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1632 MOPT_CLEAR | MOPT_Q},
1633 {Opt_usrjquota, 0, MOPT_Q},
1634 {Opt_grpjquota, 0, MOPT_Q},
1635 {Opt_offusrjquota, 0, MOPT_Q},
1636 {Opt_offgrpjquota, 0, MOPT_Q},
1637 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1638 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1639 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1640 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1641 {Opt_test_dummy_encryption, 0, MOPT_GTE0},
1642 {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET},
1646 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1647 substring_t *args, unsigned long *journal_devnum,
1648 unsigned int *journal_ioprio, int is_remount)
1650 struct ext4_sb_info *sbi = EXT4_SB(sb);
1651 const struct mount_opts *m;
1657 if (token == Opt_usrjquota)
1658 return set_qf_name(sb, USRQUOTA, &args[0]);
1659 else if (token == Opt_grpjquota)
1660 return set_qf_name(sb, GRPQUOTA, &args[0]);
1661 else if (token == Opt_offusrjquota)
1662 return clear_qf_name(sb, USRQUOTA);
1663 else if (token == Opt_offgrpjquota)
1664 return clear_qf_name(sb, GRPQUOTA);
1668 case Opt_nouser_xattr:
1669 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1672 return 1; /* handled by get_sb_block() */
1674 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1677 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1680 sb->s_flags |= MS_I_VERSION;
1683 sb->s_flags |= MS_LAZYTIME;
1685 case Opt_nolazytime:
1686 sb->s_flags &= ~MS_LAZYTIME;
1690 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1691 if (token == m->token)
1694 if (m->token == Opt_err) {
1695 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1696 "or missing value", opt);
1700 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1701 ext4_msg(sb, KERN_ERR,
1702 "Mount option \"%s\" incompatible with ext2", opt);
1705 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1706 ext4_msg(sb, KERN_ERR,
1707 "Mount option \"%s\" incompatible with ext3", opt);
1711 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
1713 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1715 if (m->flags & MOPT_EXPLICIT) {
1716 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
1717 set_opt2(sb, EXPLICIT_DELALLOC);
1718 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
1719 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
1723 if (m->flags & MOPT_CLEAR_ERR)
1724 clear_opt(sb, ERRORS_MASK);
1725 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1726 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1727 "options when quota turned on");
1731 if (m->flags & MOPT_NOSUPPORT) {
1732 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1733 } else if (token == Opt_commit) {
1735 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1736 sbi->s_commit_interval = HZ * arg;
1737 } else if (token == Opt_debug_want_extra_isize) {
1738 sbi->s_want_extra_isize = arg;
1739 } else if (token == Opt_max_batch_time) {
1740 sbi->s_max_batch_time = arg;
1741 } else if (token == Opt_min_batch_time) {
1742 sbi->s_min_batch_time = arg;
1743 } else if (token == Opt_inode_readahead_blks) {
1744 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1745 ext4_msg(sb, KERN_ERR,
1746 "EXT4-fs: inode_readahead_blks must be "
1747 "0 or a power of 2 smaller than 2^31");
1750 sbi->s_inode_readahead_blks = arg;
1751 } else if (token == Opt_init_itable) {
1752 set_opt(sb, INIT_INODE_TABLE);
1754 arg = EXT4_DEF_LI_WAIT_MULT;
1755 sbi->s_li_wait_mult = arg;
1756 } else if (token == Opt_max_dir_size_kb) {
1757 sbi->s_max_dir_size_kb = arg;
1758 } else if (token == Opt_stripe) {
1759 sbi->s_stripe = arg;
1760 } else if (token == Opt_resuid) {
1761 uid = make_kuid(current_user_ns(), arg);
1762 if (!uid_valid(uid)) {
1763 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1766 sbi->s_resuid = uid;
1767 } else if (token == Opt_resgid) {
1768 gid = make_kgid(current_user_ns(), arg);
1769 if (!gid_valid(gid)) {
1770 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1773 sbi->s_resgid = gid;
1774 } else if (token == Opt_journal_dev) {
1776 ext4_msg(sb, KERN_ERR,
1777 "Cannot specify journal on remount");
1780 *journal_devnum = arg;
1781 } else if (token == Opt_journal_path) {
1783 struct inode *journal_inode;
1788 ext4_msg(sb, KERN_ERR,
1789 "Cannot specify journal on remount");
1792 journal_path = match_strdup(&args[0]);
1793 if (!journal_path) {
1794 ext4_msg(sb, KERN_ERR, "error: could not dup "
1795 "journal device string");
1799 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1801 ext4_msg(sb, KERN_ERR, "error: could not find "
1802 "journal device path: error %d", error);
1803 kfree(journal_path);
1807 journal_inode = d_inode(path.dentry);
1808 if (!S_ISBLK(journal_inode->i_mode)) {
1809 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1810 "is not a block device", journal_path);
1812 kfree(journal_path);
1816 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1818 kfree(journal_path);
1819 } else if (token == Opt_journal_ioprio) {
1821 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1826 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1827 } else if (token == Opt_test_dummy_encryption) {
1828 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1829 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1830 ext4_msg(sb, KERN_WARNING,
1831 "Test dummy encryption mode enabled");
1833 ext4_msg(sb, KERN_WARNING,
1834 "Test dummy encryption mount option ignored");
1836 } else if (m->flags & MOPT_DATAJ) {
1838 if (!sbi->s_journal)
1839 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1840 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1841 ext4_msg(sb, KERN_ERR,
1842 "Cannot change data mode on remount");
1846 clear_opt(sb, DATA_FLAGS);
1847 sbi->s_mount_opt |= m->mount_opt;
1850 } else if (m->flags & MOPT_QFMT) {
1851 if (sb_any_quota_loaded(sb) &&
1852 sbi->s_jquota_fmt != m->mount_opt) {
1853 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1854 "quota options when quota turned on");
1857 if (ext4_has_feature_quota(sb)) {
1858 ext4_msg(sb, KERN_INFO,
1859 "Quota format mount options ignored "
1860 "when QUOTA feature is enabled");
1863 sbi->s_jquota_fmt = m->mount_opt;
1865 } else if (token == Opt_dax) {
1866 #ifdef CONFIG_FS_DAX
1867 ext4_msg(sb, KERN_WARNING,
1868 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
1869 sbi->s_mount_opt |= m->mount_opt;
1871 ext4_msg(sb, KERN_INFO, "dax option not supported");
1874 } else if (token == Opt_data_err_abort) {
1875 sbi->s_mount_opt |= m->mount_opt;
1876 } else if (token == Opt_data_err_ignore) {
1877 sbi->s_mount_opt &= ~m->mount_opt;
1881 if (m->flags & MOPT_CLEAR)
1883 else if (unlikely(!(m->flags & MOPT_SET))) {
1884 ext4_msg(sb, KERN_WARNING,
1885 "buggy handling of option %s", opt);
1890 sbi->s_mount_opt |= m->mount_opt;
1892 sbi->s_mount_opt &= ~m->mount_opt;
1897 static int parse_options(char *options, struct super_block *sb,
1898 unsigned long *journal_devnum,
1899 unsigned int *journal_ioprio,
1902 struct ext4_sb_info *sbi = EXT4_SB(sb);
1904 substring_t args[MAX_OPT_ARGS];
1910 while ((p = strsep(&options, ",")) != NULL) {
1914 * Initialize args struct so we know whether arg was
1915 * found; some options take optional arguments.
1917 args[0].to = args[0].from = NULL;
1918 token = match_token(p, tokens, args);
1919 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1920 journal_ioprio, is_remount) < 0)
1925 * We do the test below only for project quotas. 'usrquota' and
1926 * 'grpquota' mount options are allowed even without quota feature
1927 * to support legacy quotas in quota files.
1929 if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) {
1930 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. "
1931 "Cannot enable project quota enforcement.");
1934 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1935 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1936 clear_opt(sb, USRQUOTA);
1938 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1939 clear_opt(sb, GRPQUOTA);
1941 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1942 ext4_msg(sb, KERN_ERR, "old and new quota "
1947 if (!sbi->s_jquota_fmt) {
1948 ext4_msg(sb, KERN_ERR, "journaled quota format "
1954 if (test_opt(sb, DIOREAD_NOLOCK)) {
1956 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1958 if (blocksize < PAGE_SIZE) {
1959 ext4_msg(sb, KERN_ERR, "can't mount with "
1960 "dioread_nolock if block size != PAGE_SIZE");
1967 static inline void ext4_show_quota_options(struct seq_file *seq,
1968 struct super_block *sb)
1970 #if defined(CONFIG_QUOTA)
1971 struct ext4_sb_info *sbi = EXT4_SB(sb);
1973 if (sbi->s_jquota_fmt) {
1976 switch (sbi->s_jquota_fmt) {
1987 seq_printf(seq, ",jqfmt=%s", fmtname);
1990 if (sbi->s_qf_names[USRQUOTA])
1991 seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]);
1993 if (sbi->s_qf_names[GRPQUOTA])
1994 seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]);
1998 static const char *token2str(int token)
2000 const struct match_token *t;
2002 for (t = tokens; t->token != Opt_err; t++)
2003 if (t->token == token && !strchr(t->pattern, '='))
2010 * - it's set to a non-default value OR
2011 * - if the per-sb default is different from the global default
2013 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
2016 struct ext4_sb_info *sbi = EXT4_SB(sb);
2017 struct ext4_super_block *es = sbi->s_es;
2018 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
2019 const struct mount_opts *m;
2020 char sep = nodefs ? '\n' : ',';
2022 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2023 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2025 if (sbi->s_sb_block != 1)
2026 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
2028 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
2029 int want_set = m->flags & MOPT_SET;
2030 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
2031 (m->flags & MOPT_CLEAR_ERR))
2033 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
2034 continue; /* skip if same as the default */
2036 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
2037 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
2038 continue; /* select Opt_noFoo vs Opt_Foo */
2039 SEQ_OPTS_PRINT("%s", token2str(m->token));
2042 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
2043 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
2044 SEQ_OPTS_PRINT("resuid=%u",
2045 from_kuid_munged(&init_user_ns, sbi->s_resuid));
2046 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
2047 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
2048 SEQ_OPTS_PRINT("resgid=%u",
2049 from_kgid_munged(&init_user_ns, sbi->s_resgid));
2050 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
2051 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
2052 SEQ_OPTS_PUTS("errors=remount-ro");
2053 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
2054 SEQ_OPTS_PUTS("errors=continue");
2055 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
2056 SEQ_OPTS_PUTS("errors=panic");
2057 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
2058 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
2059 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
2060 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
2061 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
2062 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2063 if (sb->s_flags & MS_I_VERSION)
2064 SEQ_OPTS_PUTS("i_version");
2065 if (nodefs || sbi->s_stripe)
2066 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
2067 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
2068 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2069 SEQ_OPTS_PUTS("data=journal");
2070 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2071 SEQ_OPTS_PUTS("data=ordered");
2072 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2073 SEQ_OPTS_PUTS("data=writeback");
2076 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
2077 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2078 sbi->s_inode_readahead_blks);
2080 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
2081 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
2082 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
2083 if (nodefs || sbi->s_max_dir_size_kb)
2084 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2085 if (test_opt(sb, DATA_ERR_ABORT))
2086 SEQ_OPTS_PUTS("data_err=abort");
2088 ext4_show_quota_options(seq, sb);
2092 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2094 return _ext4_show_options(seq, root->d_sb, 0);
2097 int ext4_seq_options_show(struct seq_file *seq, void *offset)
2099 struct super_block *sb = seq->private;
2102 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
2103 rc = _ext4_show_options(seq, sb, 1);
2104 seq_puts(seq, "\n");
2108 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
2111 struct ext4_sb_info *sbi = EXT4_SB(sb);
2114 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
2115 ext4_msg(sb, KERN_ERR, "revision level too high, "
2116 "forcing read-only mode");
2121 if (!(sbi->s_mount_state & EXT4_VALID_FS))
2122 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
2123 "running e2fsck is recommended");
2124 else if (sbi->s_mount_state & EXT4_ERROR_FS)
2125 ext4_msg(sb, KERN_WARNING,
2126 "warning: mounting fs with errors, "
2127 "running e2fsck is recommended");
2128 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
2129 le16_to_cpu(es->s_mnt_count) >=
2130 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
2131 ext4_msg(sb, KERN_WARNING,
2132 "warning: maximal mount count reached, "
2133 "running e2fsck is recommended");
2134 else if (le32_to_cpu(es->s_checkinterval) &&
2135 (le32_to_cpu(es->s_lastcheck) +
2136 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
2137 ext4_msg(sb, KERN_WARNING,
2138 "warning: checktime reached, "
2139 "running e2fsck is recommended");
2140 if (!sbi->s_journal)
2141 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
2142 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
2143 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
2144 le16_add_cpu(&es->s_mnt_count, 1);
2145 es->s_mtime = cpu_to_le32(get_seconds());
2146 ext4_update_dynamic_rev(sb);
2148 ext4_set_feature_journal_needs_recovery(sb);
2150 ext4_commit_super(sb, 1);
2152 if (test_opt(sb, DEBUG))
2153 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
2154 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
2156 sbi->s_groups_count,
2157 EXT4_BLOCKS_PER_GROUP(sb),
2158 EXT4_INODES_PER_GROUP(sb),
2159 sbi->s_mount_opt, sbi->s_mount_opt2);
2161 cleancache_init_fs(sb);
2165 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2167 struct ext4_sb_info *sbi = EXT4_SB(sb);
2168 struct flex_groups *new_groups;
2171 if (!sbi->s_log_groups_per_flex)
2174 size = ext4_flex_group(sbi, ngroup - 1) + 1;
2175 if (size <= sbi->s_flex_groups_allocated)
2178 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
2179 new_groups = kvzalloc(size, GFP_KERNEL);
2181 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
2182 size / (int) sizeof(struct flex_groups));
2186 if (sbi->s_flex_groups) {
2187 memcpy(new_groups, sbi->s_flex_groups,
2188 (sbi->s_flex_groups_allocated *
2189 sizeof(struct flex_groups)));
2190 kvfree(sbi->s_flex_groups);
2192 sbi->s_flex_groups = new_groups;
2193 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
2197 static int ext4_fill_flex_info(struct super_block *sb)
2199 struct ext4_sb_info *sbi = EXT4_SB(sb);
2200 struct ext4_group_desc *gdp = NULL;
2201 ext4_group_t flex_group;
2204 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2205 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
2206 sbi->s_log_groups_per_flex = 0;
2210 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2214 for (i = 0; i < sbi->s_groups_count; i++) {
2215 gdp = ext4_get_group_desc(sb, i, NULL);
2217 flex_group = ext4_flex_group(sbi, i);
2218 atomic_add(ext4_free_inodes_count(sb, gdp),
2219 &sbi->s_flex_groups[flex_group].free_inodes);
2220 atomic64_add(ext4_free_group_clusters(sb, gdp),
2221 &sbi->s_flex_groups[flex_group].free_clusters);
2222 atomic_add(ext4_used_dirs_count(sb, gdp),
2223 &sbi->s_flex_groups[flex_group].used_dirs);
2231 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
2232 struct ext4_group_desc *gdp)
2234 int offset = offsetof(struct ext4_group_desc, bg_checksum);
2236 __le32 le_group = cpu_to_le32(block_group);
2237 struct ext4_sb_info *sbi = EXT4_SB(sb);
2239 if (ext4_has_metadata_csum(sbi->s_sb)) {
2240 /* Use new metadata_csum algorithm */
2242 __u16 dummy_csum = 0;
2244 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2246 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2247 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2248 sizeof(dummy_csum));
2249 offset += sizeof(dummy_csum);
2250 if (offset < sbi->s_desc_size)
2251 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2252 sbi->s_desc_size - offset);
2254 crc = csum32 & 0xFFFF;
2258 /* old crc16 code */
2259 if (!ext4_has_feature_gdt_csum(sb))
2262 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2263 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2264 crc = crc16(crc, (__u8 *)gdp, offset);
2265 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2266 /* for checksum of struct ext4_group_desc do the rest...*/
2267 if (ext4_has_feature_64bit(sb) &&
2268 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2269 crc = crc16(crc, (__u8 *)gdp + offset,
2270 le16_to_cpu(sbi->s_es->s_desc_size) -
2274 return cpu_to_le16(crc);
2277 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2278 struct ext4_group_desc *gdp)
2280 if (ext4_has_group_desc_csum(sb) &&
2281 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
2287 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2288 struct ext4_group_desc *gdp)
2290 if (!ext4_has_group_desc_csum(sb))
2292 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
2295 /* Called at mount-time, super-block is locked */
2296 static int ext4_check_descriptors(struct super_block *sb,
2297 ext4_fsblk_t sb_block,
2298 ext4_group_t *first_not_zeroed)
2300 struct ext4_sb_info *sbi = EXT4_SB(sb);
2301 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2302 ext4_fsblk_t last_block;
2303 ext4_fsblk_t block_bitmap;
2304 ext4_fsblk_t inode_bitmap;
2305 ext4_fsblk_t inode_table;
2306 int flexbg_flag = 0;
2307 ext4_group_t i, grp = sbi->s_groups_count;
2309 if (ext4_has_feature_flex_bg(sb))
2312 ext4_debug("Checking group descriptors");
2314 for (i = 0; i < sbi->s_groups_count; i++) {
2315 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2317 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2318 last_block = ext4_blocks_count(sbi->s_es) - 1;
2320 last_block = first_block +
2321 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2323 if ((grp == sbi->s_groups_count) &&
2324 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2327 block_bitmap = ext4_block_bitmap(sb, gdp);
2328 if (block_bitmap == sb_block) {
2329 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2330 "Block bitmap for group %u overlaps "
2333 if (block_bitmap < first_block || block_bitmap > last_block) {
2334 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2335 "Block bitmap for group %u not in group "
2336 "(block %llu)!", i, block_bitmap);
2339 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2340 if (inode_bitmap == sb_block) {
2341 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2342 "Inode bitmap for group %u overlaps "
2345 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2346 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2347 "Inode bitmap for group %u not in group "
2348 "(block %llu)!", i, inode_bitmap);
2351 inode_table = ext4_inode_table(sb, gdp);
2352 if (inode_table == sb_block) {
2353 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2354 "Inode table for group %u overlaps "
2357 if (inode_table < first_block ||
2358 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2359 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2360 "Inode table for group %u not in group "
2361 "(block %llu)!", i, inode_table);
2364 ext4_lock_group(sb, i);
2365 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2366 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2367 "Checksum for group %u failed (%u!=%u)",
2368 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
2369 gdp)), le16_to_cpu(gdp->bg_checksum));
2370 if (!(sb->s_flags & MS_RDONLY)) {
2371 ext4_unlock_group(sb, i);
2375 ext4_unlock_group(sb, i);
2377 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2379 if (NULL != first_not_zeroed)
2380 *first_not_zeroed = grp;
2384 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2385 * the superblock) which were deleted from all directories, but held open by
2386 * a process at the time of a crash. We walk the list and try to delete these
2387 * inodes at recovery time (only with a read-write filesystem).
2389 * In order to keep the orphan inode chain consistent during traversal (in
2390 * case of crash during recovery), we link each inode into the superblock
2391 * orphan list_head and handle it the same way as an inode deletion during
2392 * normal operation (which journals the operations for us).
2394 * We only do an iget() and an iput() on each inode, which is very safe if we
2395 * accidentally point at an in-use or already deleted inode. The worst that
2396 * can happen in this case is that we get a "bit already cleared" message from
2397 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2398 * e2fsck was run on this filesystem, and it must have already done the orphan
2399 * inode cleanup for us, so we can safely abort without any further action.
2401 static void ext4_orphan_cleanup(struct super_block *sb,
2402 struct ext4_super_block *es)
2404 unsigned int s_flags = sb->s_flags;
2405 int ret, nr_orphans = 0, nr_truncates = 0;
2407 int quota_update = 0;
2410 if (!es->s_last_orphan) {
2411 jbd_debug(4, "no orphan inodes to clean up\n");
2415 if (bdev_read_only(sb->s_bdev)) {
2416 ext4_msg(sb, KERN_ERR, "write access "
2417 "unavailable, skipping orphan cleanup");
2421 /* Check if feature set would not allow a r/w mount */
2422 if (!ext4_feature_set_ok(sb, 0)) {
2423 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2424 "unknown ROCOMPAT features");
2428 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2429 /* don't clear list on RO mount w/ errors */
2430 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
2431 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
2432 "clearing orphan list.\n");
2433 es->s_last_orphan = 0;
2435 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2439 if (s_flags & MS_RDONLY) {
2440 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2441 sb->s_flags &= ~MS_RDONLY;
2444 /* Needed for iput() to work correctly and not trash data */
2445 sb->s_flags |= MS_ACTIVE;
2448 * Turn on quotas which were not enabled for read-only mounts if
2449 * filesystem has quota feature, so that they are updated correctly.
2451 if (ext4_has_feature_quota(sb) && (s_flags & MS_RDONLY)) {
2452 int ret = ext4_enable_quotas(sb);
2457 ext4_msg(sb, KERN_ERR,
2458 "Cannot turn on quotas: error %d", ret);
2461 /* Turn on journaled quotas used for old sytle */
2462 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2463 if (EXT4_SB(sb)->s_qf_names[i]) {
2464 int ret = ext4_quota_on_mount(sb, i);
2469 ext4_msg(sb, KERN_ERR,
2470 "Cannot turn on journaled "
2471 "quota: type %d: error %d", i, ret);
2476 while (es->s_last_orphan) {
2477 struct inode *inode;
2480 * We may have encountered an error during cleanup; if
2481 * so, skip the rest.
2483 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2484 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2485 es->s_last_orphan = 0;
2489 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2490 if (IS_ERR(inode)) {
2491 es->s_last_orphan = 0;
2495 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2496 dquot_initialize(inode);
2497 if (inode->i_nlink) {
2498 if (test_opt(sb, DEBUG))
2499 ext4_msg(sb, KERN_DEBUG,
2500 "%s: truncating inode %lu to %lld bytes",
2501 __func__, inode->i_ino, inode->i_size);
2502 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2503 inode->i_ino, inode->i_size);
2505 truncate_inode_pages(inode->i_mapping, inode->i_size);
2506 ret = ext4_truncate(inode);
2508 ext4_std_error(inode->i_sb, ret);
2509 inode_unlock(inode);
2512 if (test_opt(sb, DEBUG))
2513 ext4_msg(sb, KERN_DEBUG,
2514 "%s: deleting unreferenced inode %lu",
2515 __func__, inode->i_ino);
2516 jbd_debug(2, "deleting unreferenced inode %lu\n",
2520 iput(inode); /* The delete magic happens here! */
2523 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2526 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2527 PLURAL(nr_orphans));
2529 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2530 PLURAL(nr_truncates));
2532 /* Turn off quotas if they were enabled for orphan cleanup */
2534 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2535 if (sb_dqopt(sb)->files[i])
2536 dquot_quota_off(sb, i);
2540 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2544 * Maximal extent format file size.
2545 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2546 * extent format containers, within a sector_t, and within i_blocks
2547 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2548 * so that won't be a limiting factor.
2550 * However there is other limiting factor. We do store extents in the form
2551 * of starting block and length, hence the resulting length of the extent
2552 * covering maximum file size must fit into on-disk format containers as
2553 * well. Given that length is always by 1 unit bigger than max unit (because
2554 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2556 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2558 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2561 loff_t upper_limit = MAX_LFS_FILESIZE;
2563 /* small i_blocks in vfs inode? */
2564 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2566 * CONFIG_LBDAF is not enabled implies the inode
2567 * i_block represent total blocks in 512 bytes
2568 * 32 == size of vfs inode i_blocks * 8
2570 upper_limit = (1LL << 32) - 1;
2572 /* total blocks in file system block size */
2573 upper_limit >>= (blkbits - 9);
2574 upper_limit <<= blkbits;
2578 * 32-bit extent-start container, ee_block. We lower the maxbytes
2579 * by one fs block, so ee_len can cover the extent of maximum file
2582 res = (1LL << 32) - 1;
2585 /* Sanity check against vm- & vfs- imposed limits */
2586 if (res > upper_limit)
2593 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2594 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2595 * We need to be 1 filesystem block less than the 2^48 sector limit.
2597 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2599 loff_t res = EXT4_NDIR_BLOCKS;
2602 /* This is calculated to be the largest file size for a dense, block
2603 * mapped file such that the file's total number of 512-byte sectors,
2604 * including data and all indirect blocks, does not exceed (2^48 - 1).
2606 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2607 * number of 512-byte sectors of the file.
2610 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2612 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2613 * the inode i_block field represents total file blocks in
2614 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2616 upper_limit = (1LL << 32) - 1;
2618 /* total blocks in file system block size */
2619 upper_limit >>= (bits - 9);
2623 * We use 48 bit ext4_inode i_blocks
2624 * With EXT4_HUGE_FILE_FL set the i_blocks
2625 * represent total number of blocks in
2626 * file system block size
2628 upper_limit = (1LL << 48) - 1;
2632 /* indirect blocks */
2634 /* double indirect blocks */
2635 meta_blocks += 1 + (1LL << (bits-2));
2636 /* tripple indirect blocks */
2637 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2639 upper_limit -= meta_blocks;
2640 upper_limit <<= bits;
2642 res += 1LL << (bits-2);
2643 res += 1LL << (2*(bits-2));
2644 res += 1LL << (3*(bits-2));
2646 if (res > upper_limit)
2649 if (res > MAX_LFS_FILESIZE)
2650 res = MAX_LFS_FILESIZE;
2655 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2656 ext4_fsblk_t logical_sb_block, int nr)
2658 struct ext4_sb_info *sbi = EXT4_SB(sb);
2659 ext4_group_t bg, first_meta_bg;
2662 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2664 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
2665 return logical_sb_block + nr + 1;
2666 bg = sbi->s_desc_per_block * nr;
2667 if (ext4_bg_has_super(sb, bg))
2671 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2672 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2673 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2676 if (sb->s_blocksize == 1024 && nr == 0 &&
2677 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2680 return (has_super + ext4_group_first_block_no(sb, bg));
2684 * ext4_get_stripe_size: Get the stripe size.
2685 * @sbi: In memory super block info
2687 * If we have specified it via mount option, then
2688 * use the mount option value. If the value specified at mount time is
2689 * greater than the blocks per group use the super block value.
2690 * If the super block value is greater than blocks per group return 0.
2691 * Allocator needs it be less than blocks per group.
2694 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2696 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2697 unsigned long stripe_width =
2698 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2701 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2702 ret = sbi->s_stripe;
2703 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
2705 else if (stride && stride <= sbi->s_blocks_per_group)
2711 * If the stripe width is 1, this makes no sense and
2712 * we set it to 0 to turn off stripe handling code.
2721 * Check whether this filesystem can be mounted based on
2722 * the features present and the RDONLY/RDWR mount requested.
2723 * Returns 1 if this filesystem can be mounted as requested,
2724 * 0 if it cannot be.
2726 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2728 if (ext4_has_unknown_ext4_incompat_features(sb)) {
2729 ext4_msg(sb, KERN_ERR,
2730 "Couldn't mount because of "
2731 "unsupported optional features (%x)",
2732 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2733 ~EXT4_FEATURE_INCOMPAT_SUPP));
2740 if (ext4_has_feature_readonly(sb)) {
2741 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2742 sb->s_flags |= MS_RDONLY;
2746 /* Check that feature set is OK for a read-write mount */
2747 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
2748 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2749 "unsupported optional features (%x)",
2750 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2751 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2755 * Large file size enabled file system can only be mounted
2756 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2758 if (ext4_has_feature_huge_file(sb)) {
2759 if (sizeof(blkcnt_t) < sizeof(u64)) {
2760 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2761 "cannot be mounted RDWR without "
2766 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
2767 ext4_msg(sb, KERN_ERR,
2768 "Can't support bigalloc feature without "
2769 "extents feature\n");
2773 #ifndef CONFIG_QUOTA
2774 if (ext4_has_feature_quota(sb) && !readonly) {
2775 ext4_msg(sb, KERN_ERR,
2776 "Filesystem with quota feature cannot be mounted RDWR "
2777 "without CONFIG_QUOTA");
2780 if (ext4_has_feature_project(sb) && !readonly) {
2781 ext4_msg(sb, KERN_ERR,
2782 "Filesystem with project quota feature cannot be mounted RDWR "
2783 "without CONFIG_QUOTA");
2786 #endif /* CONFIG_QUOTA */
2791 * This function is called once a day if we have errors logged
2792 * on the file system
2794 static void print_daily_error_info(unsigned long arg)
2796 struct super_block *sb = (struct super_block *) arg;
2797 struct ext4_sb_info *sbi;
2798 struct ext4_super_block *es;
2803 if (es->s_error_count)
2804 /* fsck newer than v1.41.13 is needed to clean this condition. */
2805 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
2806 le32_to_cpu(es->s_error_count));
2807 if (es->s_first_error_time) {
2808 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
2809 sb->s_id, le32_to_cpu(es->s_first_error_time),
2810 (int) sizeof(es->s_first_error_func),
2811 es->s_first_error_func,
2812 le32_to_cpu(es->s_first_error_line));
2813 if (es->s_first_error_ino)
2814 printk(KERN_CONT ": inode %u",
2815 le32_to_cpu(es->s_first_error_ino));
2816 if (es->s_first_error_block)
2817 printk(KERN_CONT ": block %llu", (unsigned long long)
2818 le64_to_cpu(es->s_first_error_block));
2819 printk(KERN_CONT "\n");
2821 if (es->s_last_error_time) {
2822 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
2823 sb->s_id, le32_to_cpu(es->s_last_error_time),
2824 (int) sizeof(es->s_last_error_func),
2825 es->s_last_error_func,
2826 le32_to_cpu(es->s_last_error_line));
2827 if (es->s_last_error_ino)
2828 printk(KERN_CONT ": inode %u",
2829 le32_to_cpu(es->s_last_error_ino));
2830 if (es->s_last_error_block)
2831 printk(KERN_CONT ": block %llu", (unsigned long long)
2832 le64_to_cpu(es->s_last_error_block));
2833 printk(KERN_CONT "\n");
2835 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2838 /* Find next suitable group and run ext4_init_inode_table */
2839 static int ext4_run_li_request(struct ext4_li_request *elr)
2841 struct ext4_group_desc *gdp = NULL;
2842 ext4_group_t group, ngroups;
2843 struct super_block *sb;
2844 unsigned long timeout = 0;
2848 ngroups = EXT4_SB(sb)->s_groups_count;
2850 for (group = elr->lr_next_group; group < ngroups; group++) {
2851 gdp = ext4_get_group_desc(sb, group, NULL);
2857 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2861 if (group >= ngroups)
2866 ret = ext4_init_inode_table(sb, group,
2867 elr->lr_timeout ? 0 : 1);
2868 if (elr->lr_timeout == 0) {
2869 timeout = (jiffies - timeout) *
2870 elr->lr_sbi->s_li_wait_mult;
2871 elr->lr_timeout = timeout;
2873 elr->lr_next_sched = jiffies + elr->lr_timeout;
2874 elr->lr_next_group = group + 1;
2880 * Remove lr_request from the list_request and free the
2881 * request structure. Should be called with li_list_mtx held
2883 static void ext4_remove_li_request(struct ext4_li_request *elr)
2885 struct ext4_sb_info *sbi;
2892 list_del(&elr->lr_request);
2893 sbi->s_li_request = NULL;
2897 static void ext4_unregister_li_request(struct super_block *sb)
2899 mutex_lock(&ext4_li_mtx);
2900 if (!ext4_li_info) {
2901 mutex_unlock(&ext4_li_mtx);
2905 mutex_lock(&ext4_li_info->li_list_mtx);
2906 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2907 mutex_unlock(&ext4_li_info->li_list_mtx);
2908 mutex_unlock(&ext4_li_mtx);
2911 static struct task_struct *ext4_lazyinit_task;
2914 * This is the function where ext4lazyinit thread lives. It walks
2915 * through the request list searching for next scheduled filesystem.
2916 * When such a fs is found, run the lazy initialization request
2917 * (ext4_rn_li_request) and keep track of the time spend in this
2918 * function. Based on that time we compute next schedule time of
2919 * the request. When walking through the list is complete, compute
2920 * next waking time and put itself into sleep.
2922 static int ext4_lazyinit_thread(void *arg)
2924 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2925 struct list_head *pos, *n;
2926 struct ext4_li_request *elr;
2927 unsigned long next_wakeup, cur;
2929 BUG_ON(NULL == eli);
2933 next_wakeup = MAX_JIFFY_OFFSET;
2935 mutex_lock(&eli->li_list_mtx);
2936 if (list_empty(&eli->li_request_list)) {
2937 mutex_unlock(&eli->li_list_mtx);
2940 list_for_each_safe(pos, n, &eli->li_request_list) {
2943 elr = list_entry(pos, struct ext4_li_request,
2946 if (time_before(jiffies, elr->lr_next_sched)) {
2947 if (time_before(elr->lr_next_sched, next_wakeup))
2948 next_wakeup = elr->lr_next_sched;
2951 if (down_read_trylock(&elr->lr_super->s_umount)) {
2952 if (sb_start_write_trylock(elr->lr_super)) {
2955 * We hold sb->s_umount, sb can not
2956 * be removed from the list, it is
2957 * now safe to drop li_list_mtx
2959 mutex_unlock(&eli->li_list_mtx);
2960 err = ext4_run_li_request(elr);
2961 sb_end_write(elr->lr_super);
2962 mutex_lock(&eli->li_list_mtx);
2965 up_read((&elr->lr_super->s_umount));
2967 /* error, remove the lazy_init job */
2969 ext4_remove_li_request(elr);
2973 elr->lr_next_sched = jiffies +
2975 % (EXT4_DEF_LI_MAX_START_DELAY * HZ));
2977 if (time_before(elr->lr_next_sched, next_wakeup))
2978 next_wakeup = elr->lr_next_sched;
2980 mutex_unlock(&eli->li_list_mtx);
2985 if ((time_after_eq(cur, next_wakeup)) ||
2986 (MAX_JIFFY_OFFSET == next_wakeup)) {
2991 schedule_timeout_interruptible(next_wakeup - cur);
2993 if (kthread_should_stop()) {
2994 ext4_clear_request_list();
3001 * It looks like the request list is empty, but we need
3002 * to check it under the li_list_mtx lock, to prevent any
3003 * additions into it, and of course we should lock ext4_li_mtx
3004 * to atomically free the list and ext4_li_info, because at
3005 * this point another ext4 filesystem could be registering
3008 mutex_lock(&ext4_li_mtx);
3009 mutex_lock(&eli->li_list_mtx);
3010 if (!list_empty(&eli->li_request_list)) {
3011 mutex_unlock(&eli->li_list_mtx);
3012 mutex_unlock(&ext4_li_mtx);
3015 mutex_unlock(&eli->li_list_mtx);
3016 kfree(ext4_li_info);
3017 ext4_li_info = NULL;
3018 mutex_unlock(&ext4_li_mtx);
3023 static void ext4_clear_request_list(void)
3025 struct list_head *pos, *n;
3026 struct ext4_li_request *elr;
3028 mutex_lock(&ext4_li_info->li_list_mtx);
3029 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3030 elr = list_entry(pos, struct ext4_li_request,
3032 ext4_remove_li_request(elr);
3034 mutex_unlock(&ext4_li_info->li_list_mtx);
3037 static int ext4_run_lazyinit_thread(void)
3039 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3040 ext4_li_info, "ext4lazyinit");
3041 if (IS_ERR(ext4_lazyinit_task)) {
3042 int err = PTR_ERR(ext4_lazyinit_task);
3043 ext4_clear_request_list();
3044 kfree(ext4_li_info);
3045 ext4_li_info = NULL;
3046 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3047 "initialization thread\n",
3051 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3056 * Check whether it make sense to run itable init. thread or not.
3057 * If there is at least one uninitialized inode table, return
3058 * corresponding group number, else the loop goes through all
3059 * groups and return total number of groups.
3061 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3063 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3064 struct ext4_group_desc *gdp = NULL;
3066 for (group = 0; group < ngroups; group++) {
3067 gdp = ext4_get_group_desc(sb, group, NULL);
3071 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3078 static int ext4_li_info_new(void)
3080 struct ext4_lazy_init *eli = NULL;
3082 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3086 INIT_LIST_HEAD(&eli->li_request_list);
3087 mutex_init(&eli->li_list_mtx);
3089 eli->li_state |= EXT4_LAZYINIT_QUIT;
3096 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3099 struct ext4_sb_info *sbi = EXT4_SB(sb);
3100 struct ext4_li_request *elr;
3102 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3108 elr->lr_next_group = start;
3111 * Randomize first schedule time of the request to
3112 * spread the inode table initialization requests
3115 elr->lr_next_sched = jiffies + (prandom_u32() %
3116 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3120 int ext4_register_li_request(struct super_block *sb,
3121 ext4_group_t first_not_zeroed)
3123 struct ext4_sb_info *sbi = EXT4_SB(sb);
3124 struct ext4_li_request *elr = NULL;
3125 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3128 mutex_lock(&ext4_li_mtx);
3129 if (sbi->s_li_request != NULL) {
3131 * Reset timeout so it can be computed again, because
3132 * s_li_wait_mult might have changed.
3134 sbi->s_li_request->lr_timeout = 0;
3138 if (first_not_zeroed == ngroups ||
3139 (sb->s_flags & MS_RDONLY) ||
3140 !test_opt(sb, INIT_INODE_TABLE))
3143 elr = ext4_li_request_new(sb, first_not_zeroed);
3149 if (NULL == ext4_li_info) {
3150 ret = ext4_li_info_new();
3155 mutex_lock(&ext4_li_info->li_list_mtx);
3156 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3157 mutex_unlock(&ext4_li_info->li_list_mtx);
3159 sbi->s_li_request = elr;
3161 * set elr to NULL here since it has been inserted to
3162 * the request_list and the removal and free of it is
3163 * handled by ext4_clear_request_list from now on.
3167 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3168 ret = ext4_run_lazyinit_thread();
3173 mutex_unlock(&ext4_li_mtx);
3180 * We do not need to lock anything since this is called on
3183 static void ext4_destroy_lazyinit_thread(void)
3186 * If thread exited earlier
3187 * there's nothing to be done.
3189 if (!ext4_li_info || !ext4_lazyinit_task)
3192 kthread_stop(ext4_lazyinit_task);
3195 static int set_journal_csum_feature_set(struct super_block *sb)
3198 int compat, incompat;
3199 struct ext4_sb_info *sbi = EXT4_SB(sb);
3201 if (ext4_has_metadata_csum(sb)) {
3202 /* journal checksum v3 */
3204 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
3206 /* journal checksum v1 */
3207 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3211 jbd2_journal_clear_features(sbi->s_journal,
3212 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3213 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3214 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3215 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3216 ret = jbd2_journal_set_features(sbi->s_journal,
3218 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3220 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3221 ret = jbd2_journal_set_features(sbi->s_journal,
3224 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3225 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3227 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3228 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3235 * Note: calculating the overhead so we can be compatible with
3236 * historical BSD practice is quite difficult in the face of
3237 * clusters/bigalloc. This is because multiple metadata blocks from
3238 * different block group can end up in the same allocation cluster.
3239 * Calculating the exact overhead in the face of clustered allocation
3240 * requires either O(all block bitmaps) in memory or O(number of block
3241 * groups**2) in time. We will still calculate the superblock for
3242 * older file systems --- and if we come across with a bigalloc file
3243 * system with zero in s_overhead_clusters the estimate will be close to
3244 * correct especially for very large cluster sizes --- but for newer
3245 * file systems, it's better to calculate this figure once at mkfs
3246 * time, and store it in the superblock. If the superblock value is
3247 * present (even for non-bigalloc file systems), we will use it.
3249 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3252 struct ext4_sb_info *sbi = EXT4_SB(sb);
3253 struct ext4_group_desc *gdp;
3254 ext4_fsblk_t first_block, last_block, b;
3255 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3256 int s, j, count = 0;
3258 if (!ext4_has_feature_bigalloc(sb))
3259 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3260 sbi->s_itb_per_group + 2);
3262 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3263 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3264 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3265 for (i = 0; i < ngroups; i++) {
3266 gdp = ext4_get_group_desc(sb, i, NULL);
3267 b = ext4_block_bitmap(sb, gdp);
3268 if (b >= first_block && b <= last_block) {
3269 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3272 b = ext4_inode_bitmap(sb, gdp);
3273 if (b >= first_block && b <= last_block) {
3274 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3277 b = ext4_inode_table(sb, gdp);
3278 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3279 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3280 int c = EXT4_B2C(sbi, b - first_block);
3281 ext4_set_bit(c, buf);
3287 if (ext4_bg_has_super(sb, grp)) {
3288 ext4_set_bit(s++, buf);
3291 j = ext4_bg_num_gdb(sb, grp);
3292 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
3293 ext4_error(sb, "Invalid number of block group "
3294 "descriptor blocks: %d", j);
3295 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
3299 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3303 return EXT4_CLUSTERS_PER_GROUP(sb) -
3304 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3308 * Compute the overhead and stash it in sbi->s_overhead
3310 int ext4_calculate_overhead(struct super_block *sb)
3312 struct ext4_sb_info *sbi = EXT4_SB(sb);
3313 struct ext4_super_block *es = sbi->s_es;
3314 struct inode *j_inode;
3315 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
3316 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3317 ext4_fsblk_t overhead = 0;
3318 char *buf = (char *) get_zeroed_page(GFP_NOFS);
3324 * Compute the overhead (FS structures). This is constant
3325 * for a given filesystem unless the number of block groups
3326 * changes so we cache the previous value until it does.
3330 * All of the blocks before first_data_block are overhead
3332 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3335 * Add the overhead found in each block group
3337 for (i = 0; i < ngroups; i++) {
3340 blks = count_overhead(sb, i, buf);
3343 memset(buf, 0, PAGE_SIZE);
3348 * Add the internal journal blocks whether the journal has been
3351 if (sbi->s_journal && !sbi->journal_bdev)
3352 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3353 else if (ext4_has_feature_journal(sb) && !sbi->s_journal) {
3354 j_inode = ext4_get_journal_inode(sb, j_inum);
3356 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
3357 overhead += EXT4_NUM_B2C(sbi, j_blocks);
3360 ext4_msg(sb, KERN_ERR, "can't get journal size");
3363 sbi->s_overhead = overhead;
3365 free_page((unsigned long) buf);
3369 static void ext4_set_resv_clusters(struct super_block *sb)
3371 ext4_fsblk_t resv_clusters;
3372 struct ext4_sb_info *sbi = EXT4_SB(sb);
3375 * There's no need to reserve anything when we aren't using extents.
3376 * The space estimates are exact, there are no unwritten extents,
3377 * hole punching doesn't need new metadata... This is needed especially
3378 * to keep ext2/3 backward compatibility.
3380 if (!ext4_has_feature_extents(sb))
3383 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3384 * This should cover the situations where we can not afford to run
3385 * out of space like for example punch hole, or converting
3386 * unwritten extents in delalloc path. In most cases such
3387 * allocation would require 1, or 2 blocks, higher numbers are
3390 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3391 sbi->s_cluster_bits);
3393 do_div(resv_clusters, 50);
3394 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3396 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
3399 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3401 char *orig_data = kstrdup(data, GFP_KERNEL);
3402 struct buffer_head *bh;
3403 struct ext4_super_block *es = NULL;
3404 struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3406 ext4_fsblk_t sb_block = get_sb_block(&data);
3407 ext4_fsblk_t logical_sb_block;
3408 unsigned long offset = 0;
3409 unsigned long journal_devnum = 0;
3410 unsigned long def_mount_opts;
3414 int blocksize, clustersize;
3415 unsigned int db_count;
3417 int needs_recovery, has_huge_files, has_bigalloc;
3420 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3421 ext4_group_t first_not_zeroed;
3423 if ((data && !orig_data) || !sbi)
3426 sbi->s_blockgroup_lock =
3427 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3428 if (!sbi->s_blockgroup_lock)
3431 sb->s_fs_info = sbi;
3433 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3434 sbi->s_sb_block = sb_block;
3435 if (sb->s_bdev->bd_part)
3436 sbi->s_sectors_written_start =
3437 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3439 /* Cleanup superblock name */
3440 strreplace(sb->s_id, '/', '!');
3442 /* -EINVAL is default */
3444 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3446 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3451 * The ext4 superblock will not be buffer aligned for other than 1kB
3452 * block sizes. We need to calculate the offset from buffer start.
3454 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3455 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3456 offset = do_div(logical_sb_block, blocksize);
3458 logical_sb_block = sb_block;
3461 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
3462 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3466 * Note: s_es must be initialized as soon as possible because
3467 * some ext4 macro-instructions depend on its value
3469 es = (struct ext4_super_block *) (bh->b_data + offset);
3471 sb->s_magic = le16_to_cpu(es->s_magic);
3472 if (sb->s_magic != EXT4_SUPER_MAGIC)
3474 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3476 /* Warn if metadata_csum and gdt_csum are both set. */
3477 if (ext4_has_feature_metadata_csum(sb) &&
3478 ext4_has_feature_gdt_csum(sb))
3479 ext4_warning(sb, "metadata_csum and uninit_bg are "
3480 "redundant flags; please run fsck.");
3482 /* Check for a known checksum algorithm */
3483 if (!ext4_verify_csum_type(sb, es)) {
3484 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3485 "unknown checksum algorithm.");
3490 /* Load the checksum driver */
3491 if (ext4_has_feature_metadata_csum(sb) ||
3492 ext4_has_feature_ea_inode(sb)) {
3493 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3494 if (IS_ERR(sbi->s_chksum_driver)) {
3495 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3496 ret = PTR_ERR(sbi->s_chksum_driver);
3497 sbi->s_chksum_driver = NULL;
3502 /* Check superblock checksum */
3503 if (!ext4_superblock_csum_verify(sb, es)) {
3504 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3505 "invalid superblock checksum. Run e2fsck?");
3511 /* Precompute checksum seed for all metadata */
3512 if (ext4_has_feature_csum_seed(sb))
3513 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
3514 else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb))
3515 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3516 sizeof(es->s_uuid));
3518 /* Set defaults before we parse the mount options */
3519 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3520 set_opt(sb, INIT_INODE_TABLE);
3521 if (def_mount_opts & EXT4_DEFM_DEBUG)
3523 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3525 if (def_mount_opts & EXT4_DEFM_UID16)
3526 set_opt(sb, NO_UID32);
3527 /* xattr user namespace & acls are now defaulted on */
3528 set_opt(sb, XATTR_USER);
3529 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3530 set_opt(sb, POSIX_ACL);
3532 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3533 if (ext4_has_metadata_csum(sb))
3534 set_opt(sb, JOURNAL_CHECKSUM);
3536 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3537 set_opt(sb, JOURNAL_DATA);
3538 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3539 set_opt(sb, ORDERED_DATA);
3540 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3541 set_opt(sb, WRITEBACK_DATA);
3543 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3544 set_opt(sb, ERRORS_PANIC);
3545 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3546 set_opt(sb, ERRORS_CONT);
3548 set_opt(sb, ERRORS_RO);
3549 /* block_validity enabled by default; disable with noblock_validity */
3550 set_opt(sb, BLOCK_VALIDITY);
3551 if (def_mount_opts & EXT4_DEFM_DISCARD)
3552 set_opt(sb, DISCARD);
3554 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3555 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3556 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3557 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3558 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3560 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3561 set_opt(sb, BARRIER);
3564 * enable delayed allocation by default
3565 * Use -o nodelalloc to turn it off
3567 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3568 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3569 set_opt(sb, DELALLOC);
3572 * set default s_li_wait_mult for lazyinit, for the case there is
3573 * no mount option specified.
3575 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3577 if (sbi->s_es->s_mount_opts[0]) {
3578 char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
3579 sizeof(sbi->s_es->s_mount_opts),
3583 if (!parse_options(s_mount_opts, sb, &journal_devnum,
3584 &journal_ioprio, 0)) {
3585 ext4_msg(sb, KERN_WARNING,
3586 "failed to parse options in superblock: %s",
3589 kfree(s_mount_opts);
3591 sbi->s_def_mount_opt = sbi->s_mount_opt;
3592 if (!parse_options((char *) data, sb, &journal_devnum,
3593 &journal_ioprio, 0))
3596 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3597 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3598 "with data=journal disables delayed "
3599 "allocation and O_DIRECT support!\n");
3600 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3601 ext4_msg(sb, KERN_ERR, "can't mount with "
3602 "both data=journal and delalloc");
3605 if (test_opt(sb, DIOREAD_NOLOCK)) {
3606 ext4_msg(sb, KERN_ERR, "can't mount with "
3607 "both data=journal and dioread_nolock");
3610 if (test_opt(sb, DAX)) {
3611 ext4_msg(sb, KERN_ERR, "can't mount with "
3612 "both data=journal and dax");
3615 if (ext4_has_feature_encrypt(sb)) {
3616 ext4_msg(sb, KERN_WARNING,
3617 "encrypted files will use data=ordered "
3618 "instead of data journaling mode");
3620 if (test_opt(sb, DELALLOC))
3621 clear_opt(sb, DELALLOC);
3623 sb->s_iflags |= SB_I_CGROUPWB;
3626 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3627 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3629 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3630 (ext4_has_compat_features(sb) ||
3631 ext4_has_ro_compat_features(sb) ||
3632 ext4_has_incompat_features(sb)))
3633 ext4_msg(sb, KERN_WARNING,
3634 "feature flags set on rev 0 fs, "
3635 "running e2fsck is recommended");
3637 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3638 set_opt2(sb, HURD_COMPAT);
3639 if (ext4_has_feature_64bit(sb)) {
3640 ext4_msg(sb, KERN_ERR,
3641 "The Hurd can't support 64-bit file systems");
3646 * ea_inode feature uses l_i_version field which is not
3647 * available in HURD_COMPAT mode.
3649 if (ext4_has_feature_ea_inode(sb)) {
3650 ext4_msg(sb, KERN_ERR,
3651 "ea_inode feature is not supported for Hurd");
3656 if (IS_EXT2_SB(sb)) {
3657 if (ext2_feature_set_ok(sb))
3658 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3659 "using the ext4 subsystem");
3661 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3662 "to feature incompatibilities");
3667 if (IS_EXT3_SB(sb)) {
3668 if (ext3_feature_set_ok(sb))
3669 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3670 "using the ext4 subsystem");
3672 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3673 "to feature incompatibilities");
3679 * Check feature flags regardless of the revision level, since we
3680 * previously didn't change the revision level when setting the flags,
3681 * so there is a chance incompat flags are set on a rev 0 filesystem.
3683 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3686 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3687 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3688 blocksize > EXT4_MAX_BLOCK_SIZE) {
3689 ext4_msg(sb, KERN_ERR,
3690 "Unsupported filesystem blocksize %d (%d log_block_size)",
3691 blocksize, le32_to_cpu(es->s_log_block_size));
3694 if (le32_to_cpu(es->s_log_block_size) >
3695 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3696 ext4_msg(sb, KERN_ERR,
3697 "Invalid log block size: %u",
3698 le32_to_cpu(es->s_log_block_size));
3702 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
3703 ext4_msg(sb, KERN_ERR,
3704 "Number of reserved GDT blocks insanely large: %d",
3705 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
3709 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
3710 err = bdev_dax_supported(sb, blocksize);
3715 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
3716 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
3717 es->s_encryption_level);
3721 if (sb->s_blocksize != blocksize) {
3722 /* Validate the filesystem blocksize */
3723 if (!sb_set_blocksize(sb, blocksize)) {
3724 ext4_msg(sb, KERN_ERR, "bad block size %d",
3730 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3731 offset = do_div(logical_sb_block, blocksize);
3732 bh = sb_bread_unmovable(sb, logical_sb_block);
3734 ext4_msg(sb, KERN_ERR,
3735 "Can't read superblock on 2nd try");
3738 es = (struct ext4_super_block *)(bh->b_data + offset);
3740 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3741 ext4_msg(sb, KERN_ERR,
3742 "Magic mismatch, very weird!");
3747 has_huge_files = ext4_has_feature_huge_file(sb);
3748 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3750 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3752 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3753 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3754 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3756 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3757 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3758 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3759 (!is_power_of_2(sbi->s_inode_size)) ||
3760 (sbi->s_inode_size > blocksize)) {
3761 ext4_msg(sb, KERN_ERR,
3762 "unsupported inode size: %d",
3766 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3767 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3770 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3771 if (ext4_has_feature_64bit(sb)) {
3772 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3773 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3774 !is_power_of_2(sbi->s_desc_size)) {
3775 ext4_msg(sb, KERN_ERR,
3776 "unsupported descriptor size %lu",
3781 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3783 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3784 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3786 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3787 if (sbi->s_inodes_per_block == 0)
3789 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
3790 sbi->s_inodes_per_group > blocksize * 8) {
3791 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
3792 sbi->s_blocks_per_group);
3795 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3796 sbi->s_inodes_per_block;
3797 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3799 sbi->s_mount_state = le16_to_cpu(es->s_state);
3800 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3801 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3803 for (i = 0; i < 4; i++)
3804 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3805 sbi->s_def_hash_version = es->s_def_hash_version;
3806 if (ext4_has_feature_dir_index(sb)) {
3807 i = le32_to_cpu(es->s_flags);
3808 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3809 sbi->s_hash_unsigned = 3;
3810 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3811 #ifdef __CHAR_UNSIGNED__
3812 if (!(sb->s_flags & MS_RDONLY))
3814 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3815 sbi->s_hash_unsigned = 3;
3817 if (!(sb->s_flags & MS_RDONLY))
3819 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3824 /* Handle clustersize */
3825 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3826 has_bigalloc = ext4_has_feature_bigalloc(sb);
3828 if (clustersize < blocksize) {
3829 ext4_msg(sb, KERN_ERR,
3830 "cluster size (%d) smaller than "
3831 "block size (%d)", clustersize, blocksize);
3834 if (le32_to_cpu(es->s_log_cluster_size) >
3835 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3836 ext4_msg(sb, KERN_ERR,
3837 "Invalid log cluster size: %u",
3838 le32_to_cpu(es->s_log_cluster_size));
3841 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3842 le32_to_cpu(es->s_log_block_size);
3843 sbi->s_clusters_per_group =
3844 le32_to_cpu(es->s_clusters_per_group);
3845 if (sbi->s_clusters_per_group > blocksize * 8) {
3846 ext4_msg(sb, KERN_ERR,
3847 "#clusters per group too big: %lu",
3848 sbi->s_clusters_per_group);
3851 if (sbi->s_blocks_per_group !=
3852 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3853 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3854 "clusters per group (%lu) inconsistent",
3855 sbi->s_blocks_per_group,
3856 sbi->s_clusters_per_group);
3860 if (clustersize != blocksize) {
3861 ext4_warning(sb, "fragment/cluster size (%d) != "
3862 "block size (%d)", clustersize,
3864 clustersize = blocksize;
3866 if (sbi->s_blocks_per_group > blocksize * 8) {
3867 ext4_msg(sb, KERN_ERR,
3868 "#blocks per group too big: %lu",
3869 sbi->s_blocks_per_group);
3872 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3873 sbi->s_cluster_bits = 0;
3875 sbi->s_cluster_ratio = clustersize / blocksize;
3877 /* Do we have standard group size of clustersize * 8 blocks ? */
3878 if (sbi->s_blocks_per_group == clustersize << 3)
3879 set_opt2(sb, STD_GROUP_SIZE);
3882 * Test whether we have more sectors than will fit in sector_t,
3883 * and whether the max offset is addressable by the page cache.
3885 err = generic_check_addressable(sb->s_blocksize_bits,
3886 ext4_blocks_count(es));
3888 ext4_msg(sb, KERN_ERR, "filesystem"
3889 " too large to mount safely on this system");
3890 if (sizeof(sector_t) < 8)
3891 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3895 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3898 /* check blocks count against device size */
3899 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3900 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3901 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3902 "exceeds size of device (%llu blocks)",
3903 ext4_blocks_count(es), blocks_count);
3908 * It makes no sense for the first data block to be beyond the end
3909 * of the filesystem.
3911 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3912 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3913 "block %u is beyond end of filesystem (%llu)",
3914 le32_to_cpu(es->s_first_data_block),
3915 ext4_blocks_count(es));
3918 blocks_count = (ext4_blocks_count(es) -
3919 le32_to_cpu(es->s_first_data_block) +
3920 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3921 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3922 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3923 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3924 "(block count %llu, first data block %u, "
3925 "blocks per group %lu)", sbi->s_groups_count,
3926 ext4_blocks_count(es),
3927 le32_to_cpu(es->s_first_data_block),
3928 EXT4_BLOCKS_PER_GROUP(sb));
3931 sbi->s_groups_count = blocks_count;
3932 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3933 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3934 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3935 EXT4_DESC_PER_BLOCK(sb);
3936 if (ext4_has_feature_meta_bg(sb)) {
3937 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
3938 ext4_msg(sb, KERN_WARNING,
3939 "first meta block group too large: %u "
3940 "(group descriptor block count %u)",
3941 le32_to_cpu(es->s_first_meta_bg), db_count);
3945 sbi->s_group_desc = kvmalloc(db_count *
3946 sizeof(struct buffer_head *),
3948 if (sbi->s_group_desc == NULL) {
3949 ext4_msg(sb, KERN_ERR, "not enough memory");
3954 bgl_lock_init(sbi->s_blockgroup_lock);
3956 /* Pre-read the descriptors into the buffer cache */
3957 for (i = 0; i < db_count; i++) {
3958 block = descriptor_loc(sb, logical_sb_block, i);
3959 sb_breadahead(sb, block);
3962 for (i = 0; i < db_count; i++) {
3963 block = descriptor_loc(sb, logical_sb_block, i);
3964 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
3965 if (!sbi->s_group_desc[i]) {
3966 ext4_msg(sb, KERN_ERR,
3967 "can't read group descriptor %d", i);
3972 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
3973 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3974 ret = -EFSCORRUPTED;
3978 sbi->s_gdb_count = db_count;
3979 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3980 spin_lock_init(&sbi->s_next_gen_lock);
3982 setup_timer(&sbi->s_err_report, print_daily_error_info,
3983 (unsigned long) sb);
3985 /* Register extent status tree shrinker */
3986 if (ext4_es_register_shrinker(sbi))
3989 sbi->s_stripe = ext4_get_stripe_size(sbi);
3990 sbi->s_extent_max_zeroout_kb = 32;
3993 * set up enough so that it can read an inode
3995 sb->s_op = &ext4_sops;
3996 sb->s_export_op = &ext4_export_ops;
3997 sb->s_xattr = ext4_xattr_handlers;
3998 sb->s_cop = &ext4_cryptops;
4000 sb->dq_op = &ext4_quota_operations;
4001 if (ext4_has_feature_quota(sb))
4002 sb->s_qcop = &dquot_quotactl_sysfile_ops;
4004 sb->s_qcop = &ext4_qctl_operations;
4005 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
4007 memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
4009 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
4010 mutex_init(&sbi->s_orphan_lock);
4014 needs_recovery = (es->s_last_orphan != 0 ||
4015 ext4_has_feature_journal_needs_recovery(sb));
4017 if (ext4_has_feature_mmp(sb) && !(sb->s_flags & MS_RDONLY))
4018 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
4019 goto failed_mount3a;
4022 * The first inode we look at is the journal inode. Don't try
4023 * root first: it may be modified in the journal!
4025 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
4026 err = ext4_load_journal(sb, es, journal_devnum);
4028 goto failed_mount3a;
4029 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
4030 ext4_has_feature_journal_needs_recovery(sb)) {
4031 ext4_msg(sb, KERN_ERR, "required journal recovery "
4032 "suppressed and not mounted read-only");
4033 goto failed_mount_wq;
4035 /* Nojournal mode, all journal mount options are illegal */
4036 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
4037 ext4_msg(sb, KERN_ERR, "can't mount with "
4038 "journal_checksum, fs mounted w/o journal");
4039 goto failed_mount_wq;
4041 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4042 ext4_msg(sb, KERN_ERR, "can't mount with "
4043 "journal_async_commit, fs mounted w/o journal");
4044 goto failed_mount_wq;
4046 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
4047 ext4_msg(sb, KERN_ERR, "can't mount with "
4048 "commit=%lu, fs mounted w/o journal",
4049 sbi->s_commit_interval / HZ);
4050 goto failed_mount_wq;
4052 if (EXT4_MOUNT_DATA_FLAGS &
4053 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
4054 ext4_msg(sb, KERN_ERR, "can't mount with "
4055 "data=, fs mounted w/o journal");
4056 goto failed_mount_wq;
4058 sbi->s_def_mount_opt &= EXT4_MOUNT_JOURNAL_CHECKSUM;
4059 clear_opt(sb, JOURNAL_CHECKSUM);
4060 clear_opt(sb, DATA_FLAGS);
4061 sbi->s_journal = NULL;
4066 if (ext4_has_feature_64bit(sb) &&
4067 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4068 JBD2_FEATURE_INCOMPAT_64BIT)) {
4069 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
4070 goto failed_mount_wq;
4073 if (!set_journal_csum_feature_set(sb)) {
4074 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4076 goto failed_mount_wq;
4079 /* We have now updated the journal if required, so we can
4080 * validate the data journaling mode. */
4081 switch (test_opt(sb, DATA_FLAGS)) {
4083 /* No mode set, assume a default based on the journal
4084 * capabilities: ORDERED_DATA if the journal can
4085 * cope, else JOURNAL_DATA
4087 if (jbd2_journal_check_available_features
4088 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
4089 set_opt(sb, ORDERED_DATA);
4091 set_opt(sb, JOURNAL_DATA);
4094 case EXT4_MOUNT_ORDERED_DATA:
4095 case EXT4_MOUNT_WRITEBACK_DATA:
4096 if (!jbd2_journal_check_available_features
4097 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4098 ext4_msg(sb, KERN_ERR, "Journal does not support "
4099 "requested data journaling mode");
4100 goto failed_mount_wq;
4106 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
4107 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4108 ext4_msg(sb, KERN_ERR, "can't mount with "
4109 "journal_async_commit in data=ordered mode");
4110 goto failed_mount_wq;
4113 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4115 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4118 if (!test_opt(sb, NO_MBCACHE)) {
4119 sbi->s_ea_block_cache = ext4_xattr_create_cache();
4120 if (!sbi->s_ea_block_cache) {
4121 ext4_msg(sb, KERN_ERR,
4122 "Failed to create ea_block_cache");
4123 goto failed_mount_wq;
4126 if (ext4_has_feature_ea_inode(sb)) {
4127 sbi->s_ea_inode_cache = ext4_xattr_create_cache();
4128 if (!sbi->s_ea_inode_cache) {
4129 ext4_msg(sb, KERN_ERR,
4130 "Failed to create ea_inode_cache");
4131 goto failed_mount_wq;
4136 if ((DUMMY_ENCRYPTION_ENABLED(sbi) || ext4_has_feature_encrypt(sb)) &&
4137 (blocksize != PAGE_SIZE)) {
4138 ext4_msg(sb, KERN_ERR,
4139 "Unsupported blocksize for fs encryption");
4140 goto failed_mount_wq;
4143 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !(sb->s_flags & MS_RDONLY) &&
4144 !ext4_has_feature_encrypt(sb)) {
4145 ext4_set_feature_encrypt(sb);
4146 ext4_commit_super(sb, 1);
4150 * Get the # of file system overhead blocks from the
4151 * superblock if present.
4153 if (es->s_overhead_clusters)
4154 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4156 err = ext4_calculate_overhead(sb);
4158 goto failed_mount_wq;
4162 * The maximum number of concurrent works can be high and
4163 * concurrency isn't really necessary. Limit it to 1.
4165 EXT4_SB(sb)->rsv_conversion_wq =
4166 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4167 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4168 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
4174 * The jbd2_journal_load will have done any necessary log recovery,
4175 * so we can safely mount the rest of the filesystem now.
4178 root = ext4_iget(sb, EXT4_ROOT_INO);
4180 ext4_msg(sb, KERN_ERR, "get root inode failed");
4181 ret = PTR_ERR(root);
4185 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4186 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4190 sb->s_root = d_make_root(root);
4192 ext4_msg(sb, KERN_ERR, "get root dentry failed");
4197 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
4198 sb->s_flags |= MS_RDONLY;
4200 /* determine the minimum size of new large inodes, if present */
4201 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE &&
4202 sbi->s_want_extra_isize == 0) {
4203 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4204 EXT4_GOOD_OLD_INODE_SIZE;
4205 if (ext4_has_feature_extra_isize(sb)) {
4206 if (sbi->s_want_extra_isize <
4207 le16_to_cpu(es->s_want_extra_isize))
4208 sbi->s_want_extra_isize =
4209 le16_to_cpu(es->s_want_extra_isize);
4210 if (sbi->s_want_extra_isize <
4211 le16_to_cpu(es->s_min_extra_isize))
4212 sbi->s_want_extra_isize =
4213 le16_to_cpu(es->s_min_extra_isize);
4216 /* Check if enough inode space is available */
4217 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4218 sbi->s_inode_size) {
4219 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4220 EXT4_GOOD_OLD_INODE_SIZE;
4221 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4225 ext4_set_resv_clusters(sb);
4227 err = ext4_setup_system_zone(sb);
4229 ext4_msg(sb, KERN_ERR, "failed to initialize system "
4231 goto failed_mount4a;
4235 err = ext4_mb_init(sb);
4237 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4242 block = ext4_count_free_clusters(sb);
4243 ext4_free_blocks_count_set(sbi->s_es,
4244 EXT4_C2B(sbi, block));
4245 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4248 unsigned long freei = ext4_count_free_inodes(sb);
4249 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
4250 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4254 err = percpu_counter_init(&sbi->s_dirs_counter,
4255 ext4_count_dirs(sb), GFP_KERNEL);
4257 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4260 err = percpu_init_rwsem(&sbi->s_journal_flag_rwsem);
4263 ext4_msg(sb, KERN_ERR, "insufficient memory");
4267 if (ext4_has_feature_flex_bg(sb))
4268 if (!ext4_fill_flex_info(sb)) {
4269 ext4_msg(sb, KERN_ERR,
4270 "unable to initialize "
4271 "flex_bg meta info!");
4275 err = ext4_register_li_request(sb, first_not_zeroed);
4279 err = ext4_register_sysfs(sb);
4284 /* Enable quota usage during mount. */
4285 if (ext4_has_feature_quota(sb) && !(sb->s_flags & MS_RDONLY)) {
4286 err = ext4_enable_quotas(sb);
4290 #endif /* CONFIG_QUOTA */
4292 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4293 ext4_orphan_cleanup(sb, es);
4294 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4295 if (needs_recovery) {
4296 ext4_msg(sb, KERN_INFO, "recovery complete");
4297 ext4_mark_recovery_complete(sb, es);
4299 if (EXT4_SB(sb)->s_journal) {
4300 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4301 descr = " journalled data mode";
4302 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4303 descr = " ordered data mode";
4305 descr = " writeback data mode";
4307 descr = "out journal";
4309 if (test_opt(sb, DISCARD)) {
4310 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4311 if (!blk_queue_discard(q))
4312 ext4_msg(sb, KERN_WARNING,
4313 "mounting with \"discard\" option, but "
4314 "the device does not support discard");
4317 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
4318 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4319 "Opts: %.*s%s%s", descr,
4320 (int) sizeof(sbi->s_es->s_mount_opts),
4321 sbi->s_es->s_mount_opts,
4322 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4324 if (es->s_error_count)
4325 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4327 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4328 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4329 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4330 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4337 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4342 ext4_unregister_sysfs(sb);
4345 ext4_unregister_li_request(sb);
4347 ext4_mb_release(sb);
4348 if (sbi->s_flex_groups)
4349 kvfree(sbi->s_flex_groups);
4350 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4351 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4352 percpu_counter_destroy(&sbi->s_dirs_counter);
4353 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4355 ext4_ext_release(sb);
4356 ext4_release_system_zone(sb);
4361 ext4_msg(sb, KERN_ERR, "mount failed");
4362 if (EXT4_SB(sb)->rsv_conversion_wq)
4363 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4365 if (sbi->s_ea_inode_cache) {
4366 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
4367 sbi->s_ea_inode_cache = NULL;
4369 if (sbi->s_ea_block_cache) {
4370 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
4371 sbi->s_ea_block_cache = NULL;
4373 if (sbi->s_journal) {
4374 jbd2_journal_destroy(sbi->s_journal);
4375 sbi->s_journal = NULL;
4378 ext4_es_unregister_shrinker(sbi);
4380 del_timer_sync(&sbi->s_err_report);
4382 kthread_stop(sbi->s_mmp_tsk);
4384 for (i = 0; i < db_count; i++)
4385 brelse(sbi->s_group_desc[i]);
4386 kvfree(sbi->s_group_desc);
4388 if (sbi->s_chksum_driver)
4389 crypto_free_shash(sbi->s_chksum_driver);
4391 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4392 kfree(sbi->s_qf_names[i]);
4394 ext4_blkdev_remove(sbi);
4397 sb->s_fs_info = NULL;
4398 kfree(sbi->s_blockgroup_lock);
4402 return err ? err : ret;
4406 * Setup any per-fs journal parameters now. We'll do this both on
4407 * initial mount, once the journal has been initialised but before we've
4408 * done any recovery; and again on any subsequent remount.
4410 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4412 struct ext4_sb_info *sbi = EXT4_SB(sb);
4414 journal->j_commit_interval = sbi->s_commit_interval;
4415 journal->j_min_batch_time = sbi->s_min_batch_time;
4416 journal->j_max_batch_time = sbi->s_max_batch_time;
4418 write_lock(&journal->j_state_lock);
4419 if (test_opt(sb, BARRIER))
4420 journal->j_flags |= JBD2_BARRIER;
4422 journal->j_flags &= ~JBD2_BARRIER;
4423 if (test_opt(sb, DATA_ERR_ABORT))
4424 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4426 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4427 write_unlock(&journal->j_state_lock);
4430 static struct inode *ext4_get_journal_inode(struct super_block *sb,
4431 unsigned int journal_inum)
4433 struct inode *journal_inode;
4436 * Test for the existence of a valid inode on disk. Bad things
4437 * happen if we iget() an unused inode, as the subsequent iput()
4438 * will try to delete it.
4440 journal_inode = ext4_iget(sb, journal_inum);
4441 if (IS_ERR(journal_inode)) {
4442 ext4_msg(sb, KERN_ERR, "no journal found");
4445 if (!journal_inode->i_nlink) {
4446 make_bad_inode(journal_inode);
4447 iput(journal_inode);
4448 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4452 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4453 journal_inode, journal_inode->i_size);
4454 if (!S_ISREG(journal_inode->i_mode)) {
4455 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4456 iput(journal_inode);
4459 return journal_inode;
4462 static journal_t *ext4_get_journal(struct super_block *sb,
4463 unsigned int journal_inum)
4465 struct inode *journal_inode;
4468 BUG_ON(!ext4_has_feature_journal(sb));
4470 journal_inode = ext4_get_journal_inode(sb, journal_inum);
4474 journal = jbd2_journal_init_inode(journal_inode);
4476 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4477 iput(journal_inode);
4480 journal->j_private = sb;
4481 ext4_init_journal_params(sb, journal);
4485 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4488 struct buffer_head *bh;
4492 int hblock, blocksize;
4493 ext4_fsblk_t sb_block;
4494 unsigned long offset;
4495 struct ext4_super_block *es;
4496 struct block_device *bdev;
4498 BUG_ON(!ext4_has_feature_journal(sb));
4500 bdev = ext4_blkdev_get(j_dev, sb);
4504 blocksize = sb->s_blocksize;
4505 hblock = bdev_logical_block_size(bdev);
4506 if (blocksize < hblock) {
4507 ext4_msg(sb, KERN_ERR,
4508 "blocksize too small for journal device");
4512 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4513 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4514 set_blocksize(bdev, blocksize);
4515 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4516 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4517 "external journal");
4521 es = (struct ext4_super_block *) (bh->b_data + offset);
4522 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4523 !(le32_to_cpu(es->s_feature_incompat) &
4524 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4525 ext4_msg(sb, KERN_ERR, "external journal has "
4531 if ((le32_to_cpu(es->s_feature_ro_compat) &
4532 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4533 es->s_checksum != ext4_superblock_csum(sb, es)) {
4534 ext4_msg(sb, KERN_ERR, "external journal has "
4535 "corrupt superblock");
4540 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4541 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4546 len = ext4_blocks_count(es);
4547 start = sb_block + 1;
4548 brelse(bh); /* we're done with the superblock */
4550 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4551 start, len, blocksize);
4553 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4556 journal->j_private = sb;
4557 ll_rw_block(REQ_OP_READ, REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4558 wait_on_buffer(journal->j_sb_buffer);
4559 if (!buffer_uptodate(journal->j_sb_buffer)) {
4560 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4563 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4564 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4565 "user (unsupported) - %d",
4566 be32_to_cpu(journal->j_superblock->s_nr_users));
4569 EXT4_SB(sb)->journal_bdev = bdev;
4570 ext4_init_journal_params(sb, journal);
4574 jbd2_journal_destroy(journal);
4576 ext4_blkdev_put(bdev);
4580 static int ext4_load_journal(struct super_block *sb,
4581 struct ext4_super_block *es,
4582 unsigned long journal_devnum)
4585 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4588 int really_read_only;
4590 BUG_ON(!ext4_has_feature_journal(sb));
4592 if (journal_devnum &&
4593 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4594 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4595 "numbers have changed");
4596 journal_dev = new_decode_dev(journal_devnum);
4598 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4600 really_read_only = bdev_read_only(sb->s_bdev);
4603 * Are we loading a blank journal or performing recovery after a
4604 * crash? For recovery, we need to check in advance whether we
4605 * can get read-write access to the device.
4607 if (ext4_has_feature_journal_needs_recovery(sb)) {
4608 if (sb->s_flags & MS_RDONLY) {
4609 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4610 "required on readonly filesystem");
4611 if (really_read_only) {
4612 ext4_msg(sb, KERN_ERR, "write access "
4613 "unavailable, cannot proceed");
4616 ext4_msg(sb, KERN_INFO, "write access will "
4617 "be enabled during recovery");
4621 if (journal_inum && journal_dev) {
4622 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4623 "and inode journals!");
4628 if (!(journal = ext4_get_journal(sb, journal_inum)))
4631 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4635 if (!(journal->j_flags & JBD2_BARRIER))
4636 ext4_msg(sb, KERN_INFO, "barriers disabled");
4638 if (!ext4_has_feature_journal_needs_recovery(sb))
4639 err = jbd2_journal_wipe(journal, !really_read_only);
4641 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4643 memcpy(save, ((char *) es) +
4644 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4645 err = jbd2_journal_load(journal);
4647 memcpy(((char *) es) + EXT4_S_ERR_START,
4648 save, EXT4_S_ERR_LEN);
4653 ext4_msg(sb, KERN_ERR, "error loading journal");
4654 jbd2_journal_destroy(journal);
4658 EXT4_SB(sb)->s_journal = journal;
4659 ext4_clear_journal_err(sb, es);
4661 if (!really_read_only && journal_devnum &&
4662 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4663 es->s_journal_dev = cpu_to_le32(journal_devnum);
4665 /* Make sure we flush the recovery flag to disk. */
4666 ext4_commit_super(sb, 1);
4672 static int ext4_commit_super(struct super_block *sb, int sync)
4674 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4675 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4678 if (!sbh || block_device_ejected(sb))
4681 * If the file system is mounted read-only, don't update the
4682 * superblock write time. This avoids updating the superblock
4683 * write time when we are mounting the root file system
4684 * read/only but we need to replay the journal; at that point,
4685 * for people who are east of GMT and who make their clock
4686 * tick in localtime for Windows bug-for-bug compatibility,
4687 * the clock is set in the future, and this will cause e2fsck
4688 * to complain and force a full file system check.
4690 if (!(sb->s_flags & MS_RDONLY))
4691 es->s_wtime = cpu_to_le32(get_seconds());
4692 if (sb->s_bdev->bd_part)
4693 es->s_kbytes_written =
4694 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4695 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4696 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4698 es->s_kbytes_written =
4699 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4700 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4701 ext4_free_blocks_count_set(es,
4702 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4703 &EXT4_SB(sb)->s_freeclusters_counter)));
4704 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4705 es->s_free_inodes_count =
4706 cpu_to_le32(percpu_counter_sum_positive(
4707 &EXT4_SB(sb)->s_freeinodes_counter));
4708 BUFFER_TRACE(sbh, "marking dirty");
4709 ext4_superblock_csum_set(sb);
4712 if (buffer_write_io_error(sbh)) {
4714 * Oh, dear. A previous attempt to write the
4715 * superblock failed. This could happen because the
4716 * USB device was yanked out. Or it could happen to
4717 * be a transient write error and maybe the block will
4718 * be remapped. Nothing we can do but to retry the
4719 * write and hope for the best.
4721 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4722 "superblock detected");
4723 clear_buffer_write_io_error(sbh);
4724 set_buffer_uptodate(sbh);
4726 mark_buffer_dirty(sbh);
4729 error = __sync_dirty_buffer(sbh,
4730 REQ_SYNC | (test_opt(sb, BARRIER) ? REQ_FUA : 0));
4734 error = buffer_write_io_error(sbh);
4736 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4738 clear_buffer_write_io_error(sbh);
4739 set_buffer_uptodate(sbh);
4746 * Have we just finished recovery? If so, and if we are mounting (or
4747 * remounting) the filesystem readonly, then we will end up with a
4748 * consistent fs on disk. Record that fact.
4750 static void ext4_mark_recovery_complete(struct super_block *sb,
4751 struct ext4_super_block *es)
4753 journal_t *journal = EXT4_SB(sb)->s_journal;
4755 if (!ext4_has_feature_journal(sb)) {
4756 BUG_ON(journal != NULL);
4759 jbd2_journal_lock_updates(journal);
4760 if (jbd2_journal_flush(journal) < 0)
4763 if (ext4_has_feature_journal_needs_recovery(sb) &&
4764 sb->s_flags & MS_RDONLY) {
4765 ext4_clear_feature_journal_needs_recovery(sb);
4766 ext4_commit_super(sb, 1);
4770 jbd2_journal_unlock_updates(journal);
4774 * If we are mounting (or read-write remounting) a filesystem whose journal
4775 * has recorded an error from a previous lifetime, move that error to the
4776 * main filesystem now.
4778 static void ext4_clear_journal_err(struct super_block *sb,
4779 struct ext4_super_block *es)
4785 BUG_ON(!ext4_has_feature_journal(sb));
4787 journal = EXT4_SB(sb)->s_journal;
4790 * Now check for any error status which may have been recorded in the
4791 * journal by a prior ext4_error() or ext4_abort()
4794 j_errno = jbd2_journal_errno(journal);
4798 errstr = ext4_decode_error(sb, j_errno, nbuf);
4799 ext4_warning(sb, "Filesystem error recorded "
4800 "from previous mount: %s", errstr);
4801 ext4_warning(sb, "Marking fs in need of filesystem check.");
4803 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4804 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4805 ext4_commit_super(sb, 1);
4807 jbd2_journal_clear_err(journal);
4808 jbd2_journal_update_sb_errno(journal);
4813 * Force the running and committing transactions to commit,
4814 * and wait on the commit.
4816 int ext4_force_commit(struct super_block *sb)
4820 if (sb->s_flags & MS_RDONLY)
4823 journal = EXT4_SB(sb)->s_journal;
4824 return ext4_journal_force_commit(journal);
4827 static int ext4_sync_fs(struct super_block *sb, int wait)
4831 bool needs_barrier = false;
4832 struct ext4_sb_info *sbi = EXT4_SB(sb);
4834 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
4837 trace_ext4_sync_fs(sb, wait);
4838 flush_workqueue(sbi->rsv_conversion_wq);
4840 * Writeback quota in non-journalled quota case - journalled quota has
4843 dquot_writeback_dquots(sb, -1);
4845 * Data writeback is possible w/o journal transaction, so barrier must
4846 * being sent at the end of the function. But we can skip it if
4847 * transaction_commit will do it for us.
4849 if (sbi->s_journal) {
4850 target = jbd2_get_latest_transaction(sbi->s_journal);
4851 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4852 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4853 needs_barrier = true;
4855 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4857 ret = jbd2_log_wait_commit(sbi->s_journal,
4860 } else if (wait && test_opt(sb, BARRIER))
4861 needs_barrier = true;
4862 if (needs_barrier) {
4864 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4873 * LVM calls this function before a (read-only) snapshot is created. This
4874 * gives us a chance to flush the journal completely and mark the fs clean.
4876 * Note that only this function cannot bring a filesystem to be in a clean
4877 * state independently. It relies on upper layer to stop all data & metadata
4880 static int ext4_freeze(struct super_block *sb)
4885 if (sb->s_flags & MS_RDONLY)
4888 journal = EXT4_SB(sb)->s_journal;
4891 /* Now we set up the journal barrier. */
4892 jbd2_journal_lock_updates(journal);
4895 * Don't clear the needs_recovery flag if we failed to
4896 * flush the journal.
4898 error = jbd2_journal_flush(journal);
4902 /* Journal blocked and flushed, clear needs_recovery flag. */
4903 ext4_clear_feature_journal_needs_recovery(sb);
4906 error = ext4_commit_super(sb, 1);
4909 /* we rely on upper layer to stop further updates */
4910 jbd2_journal_unlock_updates(journal);
4915 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4916 * flag here, even though the filesystem is not technically dirty yet.
4918 static int ext4_unfreeze(struct super_block *sb)
4920 if ((sb->s_flags & MS_RDONLY) || ext4_forced_shutdown(EXT4_SB(sb)))
4923 if (EXT4_SB(sb)->s_journal) {
4924 /* Reset the needs_recovery flag before the fs is unlocked. */
4925 ext4_set_feature_journal_needs_recovery(sb);
4928 ext4_commit_super(sb, 1);
4933 * Structure to save mount options for ext4_remount's benefit
4935 struct ext4_mount_options {
4936 unsigned long s_mount_opt;
4937 unsigned long s_mount_opt2;
4940 unsigned long s_commit_interval;
4941 u32 s_min_batch_time, s_max_batch_time;
4944 char *s_qf_names[EXT4_MAXQUOTAS];
4948 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4950 struct ext4_super_block *es;
4951 struct ext4_sb_info *sbi = EXT4_SB(sb);
4952 unsigned long old_sb_flags;
4953 struct ext4_mount_options old_opts;
4954 int enable_quota = 0;
4956 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4961 char *orig_data = kstrdup(data, GFP_KERNEL);
4963 /* Store the original options */
4964 old_sb_flags = sb->s_flags;
4965 old_opts.s_mount_opt = sbi->s_mount_opt;
4966 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4967 old_opts.s_resuid = sbi->s_resuid;
4968 old_opts.s_resgid = sbi->s_resgid;
4969 old_opts.s_commit_interval = sbi->s_commit_interval;
4970 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4971 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4973 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4974 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4975 if (sbi->s_qf_names[i]) {
4976 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4978 if (!old_opts.s_qf_names[i]) {
4979 for (j = 0; j < i; j++)
4980 kfree(old_opts.s_qf_names[j]);
4985 old_opts.s_qf_names[i] = NULL;
4987 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4988 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4990 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4995 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
4996 test_opt(sb, JOURNAL_CHECKSUM)) {
4997 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
4998 "during remount not supported; ignoring");
4999 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
5002 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
5003 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
5004 ext4_msg(sb, KERN_ERR, "can't mount with "
5005 "both data=journal and delalloc");
5009 if (test_opt(sb, DIOREAD_NOLOCK)) {
5010 ext4_msg(sb, KERN_ERR, "can't mount with "
5011 "both data=journal and dioread_nolock");
5015 if (test_opt(sb, DAX)) {
5016 ext4_msg(sb, KERN_ERR, "can't mount with "
5017 "both data=journal and dax");
5021 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
5022 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
5023 ext4_msg(sb, KERN_ERR, "can't mount with "
5024 "journal_async_commit in data=ordered mode");
5030 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) {
5031 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount");
5036 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
5037 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
5038 "dax flag with busy inodes while remounting");
5039 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
5042 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
5043 ext4_abort(sb, "Abort forced by user");
5045 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
5046 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
5050 if (sbi->s_journal) {
5051 ext4_init_journal_params(sb, sbi->s_journal);
5052 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
5055 if (*flags & MS_LAZYTIME)
5056 sb->s_flags |= MS_LAZYTIME;
5058 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
5059 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
5064 if (*flags & MS_RDONLY) {
5065 err = sync_filesystem(sb);
5068 err = dquot_suspend(sb, -1);
5073 * First of all, the unconditional stuff we have to do
5074 * to disable replay of the journal when we next remount
5076 sb->s_flags |= MS_RDONLY;
5079 * OK, test if we are remounting a valid rw partition
5080 * readonly, and if so set the rdonly flag and then
5081 * mark the partition as valid again.
5083 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
5084 (sbi->s_mount_state & EXT4_VALID_FS))
5085 es->s_state = cpu_to_le16(sbi->s_mount_state);
5088 ext4_mark_recovery_complete(sb, es);
5090 /* Make sure we can mount this feature set readwrite */
5091 if (ext4_has_feature_readonly(sb) ||
5092 !ext4_feature_set_ok(sb, 0)) {
5097 * Make sure the group descriptor checksums
5098 * are sane. If they aren't, refuse to remount r/w.
5100 for (g = 0; g < sbi->s_groups_count; g++) {
5101 struct ext4_group_desc *gdp =
5102 ext4_get_group_desc(sb, g, NULL);
5104 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
5105 ext4_msg(sb, KERN_ERR,
5106 "ext4_remount: Checksum for group %u failed (%u!=%u)",
5107 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
5108 le16_to_cpu(gdp->bg_checksum));
5115 * If we have an unprocessed orphan list hanging
5116 * around from a previously readonly bdev mount,
5117 * require a full umount/remount for now.
5119 if (es->s_last_orphan) {
5120 ext4_msg(sb, KERN_WARNING, "Couldn't "
5121 "remount RDWR because of unprocessed "
5122 "orphan inode list. Please "
5123 "umount/remount instead");
5129 * Mounting a RDONLY partition read-write, so reread
5130 * and store the current valid flag. (It may have
5131 * been changed by e2fsck since we originally mounted
5135 ext4_clear_journal_err(sb, es);
5136 sbi->s_mount_state = le16_to_cpu(es->s_state);
5137 if (!ext4_setup_super(sb, es, 0))
5138 sb->s_flags &= ~MS_RDONLY;
5139 if (ext4_has_feature_mmp(sb))
5140 if (ext4_multi_mount_protect(sb,
5141 le64_to_cpu(es->s_mmp_block))) {
5150 * Reinitialize lazy itable initialization thread based on
5153 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
5154 ext4_unregister_li_request(sb);
5156 ext4_group_t first_not_zeroed;
5157 first_not_zeroed = ext4_has_uninit_itable(sb);
5158 ext4_register_li_request(sb, first_not_zeroed);
5161 ext4_setup_system_zone(sb);
5162 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
5163 ext4_commit_super(sb, 1);
5166 /* Release old quota file names */
5167 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5168 kfree(old_opts.s_qf_names[i]);
5170 if (sb_any_quota_suspended(sb))
5171 dquot_resume(sb, -1);
5172 else if (ext4_has_feature_quota(sb)) {
5173 err = ext4_enable_quotas(sb);
5180 *flags = (*flags & ~MS_LAZYTIME) | (sb->s_flags & MS_LAZYTIME);
5181 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5186 sb->s_flags = old_sb_flags;
5187 sbi->s_mount_opt = old_opts.s_mount_opt;
5188 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
5189 sbi->s_resuid = old_opts.s_resuid;
5190 sbi->s_resgid = old_opts.s_resgid;
5191 sbi->s_commit_interval = old_opts.s_commit_interval;
5192 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5193 sbi->s_max_batch_time = old_opts.s_max_batch_time;
5195 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
5196 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
5197 kfree(sbi->s_qf_names[i]);
5198 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
5206 static int ext4_statfs_project(struct super_block *sb,
5207 kprojid_t projid, struct kstatfs *buf)
5210 struct dquot *dquot;
5214 qid = make_kqid_projid(projid);
5215 dquot = dqget(sb, qid);
5217 return PTR_ERR(dquot);
5218 spin_lock(&dquot->dq_dqb_lock);
5220 limit = (dquot->dq_dqb.dqb_bsoftlimit ?
5221 dquot->dq_dqb.dqb_bsoftlimit :
5222 dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits;
5223 if (limit && buf->f_blocks > limit) {
5224 curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits;
5225 buf->f_blocks = limit;
5226 buf->f_bfree = buf->f_bavail =
5227 (buf->f_blocks > curblock) ?
5228 (buf->f_blocks - curblock) : 0;
5231 limit = dquot->dq_dqb.dqb_isoftlimit ?
5232 dquot->dq_dqb.dqb_isoftlimit :
5233 dquot->dq_dqb.dqb_ihardlimit;
5234 if (limit && buf->f_files > limit) {
5235 buf->f_files = limit;
5237 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
5238 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
5241 spin_unlock(&dquot->dq_dqb_lock);
5247 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
5249 struct super_block *sb = dentry->d_sb;
5250 struct ext4_sb_info *sbi = EXT4_SB(sb);
5251 struct ext4_super_block *es = sbi->s_es;
5252 ext4_fsblk_t overhead = 0, resv_blocks;
5255 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
5257 if (!test_opt(sb, MINIX_DF))
5258 overhead = sbi->s_overhead;
5260 buf->f_type = EXT4_SUPER_MAGIC;
5261 buf->f_bsize = sb->s_blocksize;
5262 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
5263 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5264 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
5265 /* prevent underflow in case that few free space is available */
5266 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
5267 buf->f_bavail = buf->f_bfree -
5268 (ext4_r_blocks_count(es) + resv_blocks);
5269 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
5271 buf->f_files = le32_to_cpu(es->s_inodes_count);
5272 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5273 buf->f_namelen = EXT4_NAME_LEN;
5274 fsid = le64_to_cpup((void *)es->s_uuid) ^
5275 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5276 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5277 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
5280 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
5281 sb_has_quota_limits_enabled(sb, PRJQUOTA))
5282 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
5291 * Helper functions so that transaction is started before we acquire dqio_sem
5292 * to keep correct lock ordering of transaction > dqio_sem
5294 static inline struct inode *dquot_to_inode(struct dquot *dquot)
5296 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
5299 static int ext4_write_dquot(struct dquot *dquot)
5303 struct inode *inode;
5305 inode = dquot_to_inode(dquot);
5306 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
5307 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
5309 return PTR_ERR(handle);
5310 ret = dquot_commit(dquot);
5311 err = ext4_journal_stop(handle);
5317 static int ext4_acquire_dquot(struct dquot *dquot)
5322 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5323 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
5325 return PTR_ERR(handle);
5326 ret = dquot_acquire(dquot);
5327 err = ext4_journal_stop(handle);
5333 static int ext4_release_dquot(struct dquot *dquot)
5338 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5339 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
5340 if (IS_ERR(handle)) {
5341 /* Release dquot anyway to avoid endless cycle in dqput() */
5342 dquot_release(dquot);
5343 return PTR_ERR(handle);
5345 ret = dquot_release(dquot);
5346 err = ext4_journal_stop(handle);
5352 static int ext4_mark_dquot_dirty(struct dquot *dquot)
5354 struct super_block *sb = dquot->dq_sb;
5355 struct ext4_sb_info *sbi = EXT4_SB(sb);
5357 /* Are we journaling quotas? */
5358 if (ext4_has_feature_quota(sb) ||
5359 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
5360 dquot_mark_dquot_dirty(dquot);
5361 return ext4_write_dquot(dquot);
5363 return dquot_mark_dquot_dirty(dquot);
5367 static int ext4_write_info(struct super_block *sb, int type)
5372 /* Data block + inode block */
5373 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
5375 return PTR_ERR(handle);
5376 ret = dquot_commit_info(sb, type);
5377 err = ext4_journal_stop(handle);
5384 * Turn on quotas during mount time - we need to find
5385 * the quota file and such...
5387 static int ext4_quota_on_mount(struct super_block *sb, int type)
5389 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5390 EXT4_SB(sb)->s_jquota_fmt, type);
5393 static void lockdep_set_quota_inode(struct inode *inode, int subclass)
5395 struct ext4_inode_info *ei = EXT4_I(inode);
5397 /* The first argument of lockdep_set_subclass has to be
5398 * *exactly* the same as the argument to init_rwsem() --- in
5399 * this case, in init_once() --- or lockdep gets unhappy
5400 * because the name of the lock is set using the
5401 * stringification of the argument to init_rwsem().
5403 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
5404 lockdep_set_subclass(&ei->i_data_sem, subclass);
5408 * Standard function to be called on quota_on
5410 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
5411 const struct path *path)
5415 if (!test_opt(sb, QUOTA))
5418 /* Quotafile not on the same filesystem? */
5419 if (path->dentry->d_sb != sb)
5421 /* Journaling quota? */
5422 if (EXT4_SB(sb)->s_qf_names[type]) {
5423 /* Quotafile not in fs root? */
5424 if (path->dentry->d_parent != sb->s_root)
5425 ext4_msg(sb, KERN_WARNING,
5426 "Quota file not on filesystem root. "
5427 "Journaled quota will not work");
5428 sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY;
5431 * Clear the flag just in case mount options changed since
5434 sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY;
5438 * When we journal data on quota file, we have to flush journal to see
5439 * all updates to the file when we bypass pagecache...
5441 if (EXT4_SB(sb)->s_journal &&
5442 ext4_should_journal_data(d_inode(path->dentry))) {
5444 * We don't need to lock updates but journal_flush() could
5445 * otherwise be livelocked...
5447 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
5448 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
5449 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
5454 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
5455 err = dquot_quota_on(sb, type, format_id, path);
5457 lockdep_set_quota_inode(path->dentry->d_inode,
5460 struct inode *inode = d_inode(path->dentry);
5464 * Set inode flags to prevent userspace from messing with quota
5465 * files. If this fails, we return success anyway since quotas
5466 * are already enabled and this is not a hard failure.
5469 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5472 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL;
5473 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
5474 S_NOATIME | S_IMMUTABLE);
5475 ext4_mark_inode_dirty(handle, inode);
5476 ext4_journal_stop(handle);
5478 inode_unlock(inode);
5483 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5487 struct inode *qf_inode;
5488 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5489 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5490 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5491 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
5494 BUG_ON(!ext4_has_feature_quota(sb));
5496 if (!qf_inums[type])
5499 qf_inode = ext4_iget(sb, qf_inums[type]);
5500 if (IS_ERR(qf_inode)) {
5501 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5502 return PTR_ERR(qf_inode);
5505 /* Don't account quota for quota files to avoid recursion */
5506 qf_inode->i_flags |= S_NOQUOTA;
5507 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
5508 err = dquot_enable(qf_inode, type, format_id, flags);
5511 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
5516 /* Enable usage tracking for all quota types. */
5517 static int ext4_enable_quotas(struct super_block *sb)
5520 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5521 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5522 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5523 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
5525 bool quota_mopt[EXT4_MAXQUOTAS] = {
5526 test_opt(sb, USRQUOTA),
5527 test_opt(sb, GRPQUOTA),
5528 test_opt(sb, PRJQUOTA),
5531 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY;
5532 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
5533 if (qf_inums[type]) {
5534 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5535 DQUOT_USAGE_ENABLED |
5536 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
5538 for (type--; type >= 0; type--)
5539 dquot_quota_off(sb, type);
5542 "Failed to enable quota tracking "
5543 "(type=%d, err=%d). Please run "
5544 "e2fsck to fix.", type, err);
5552 static int ext4_quota_off(struct super_block *sb, int type)
5554 struct inode *inode = sb_dqopt(sb)->files[type];
5558 /* Force all delayed allocation blocks to be allocated.
5559 * Caller already holds s_umount sem */
5560 if (test_opt(sb, DELALLOC))
5561 sync_filesystem(sb);
5563 if (!inode || !igrab(inode))
5566 err = dquot_quota_off(sb, type);
5567 if (err || ext4_has_feature_quota(sb))
5572 * Update modification times of quota files when userspace can
5573 * start looking at them. If we fail, we return success anyway since
5574 * this is not a hard failure and quotas are already disabled.
5576 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5579 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL);
5580 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
5581 inode->i_mtime = inode->i_ctime = current_time(inode);
5582 ext4_mark_inode_dirty(handle, inode);
5583 ext4_journal_stop(handle);
5585 inode_unlock(inode);
5587 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL);
5591 return dquot_quota_off(sb, type);
5594 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5595 * acquiring the locks... As quota files are never truncated and quota code
5596 * itself serializes the operations (and no one else should touch the files)
5597 * we don't have to be afraid of races */
5598 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5599 size_t len, loff_t off)
5601 struct inode *inode = sb_dqopt(sb)->files[type];
5602 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5603 int offset = off & (sb->s_blocksize - 1);
5606 struct buffer_head *bh;
5607 loff_t i_size = i_size_read(inode);
5611 if (off+len > i_size)
5614 while (toread > 0) {
5615 tocopy = sb->s_blocksize - offset < toread ?
5616 sb->s_blocksize - offset : toread;
5617 bh = ext4_bread(NULL, inode, blk, 0);
5620 if (!bh) /* A hole? */
5621 memset(data, 0, tocopy);
5623 memcpy(data, bh->b_data+offset, tocopy);
5633 /* Write to quotafile (we know the transaction is already started and has
5634 * enough credits) */
5635 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5636 const char *data, size_t len, loff_t off)
5638 struct inode *inode = sb_dqopt(sb)->files[type];
5639 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5640 int err, offset = off & (sb->s_blocksize - 1);
5642 struct buffer_head *bh;
5643 handle_t *handle = journal_current_handle();
5645 if (EXT4_SB(sb)->s_journal && !handle) {
5646 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5647 " cancelled because transaction is not started",
5648 (unsigned long long)off, (unsigned long long)len);
5652 * Since we account only one data block in transaction credits,
5653 * then it is impossible to cross a block boundary.
5655 if (sb->s_blocksize - offset < len) {
5656 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5657 " cancelled because not block aligned",
5658 (unsigned long long)off, (unsigned long long)len);
5663 bh = ext4_bread(handle, inode, blk,
5664 EXT4_GET_BLOCKS_CREATE |
5665 EXT4_GET_BLOCKS_METADATA_NOFAIL);
5666 } while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) &&
5667 ext4_should_retry_alloc(inode->i_sb, &retries));
5672 BUFFER_TRACE(bh, "get write access");
5673 err = ext4_journal_get_write_access(handle, bh);
5679 memcpy(bh->b_data+offset, data, len);
5680 flush_dcache_page(bh->b_page);
5682 err = ext4_handle_dirty_metadata(handle, NULL, bh);
5685 if (inode->i_size < off + len) {
5686 i_size_write(inode, off + len);
5687 EXT4_I(inode)->i_disksize = inode->i_size;
5688 ext4_mark_inode_dirty(handle, inode);
5693 static int ext4_get_next_id(struct super_block *sb, struct kqid *qid)
5695 const struct quota_format_ops *ops;
5697 if (!sb_has_quota_loaded(sb, qid->type))
5699 ops = sb_dqopt(sb)->ops[qid->type];
5700 if (!ops || !ops->get_next_id)
5702 return dquot_get_next_id(sb, qid);
5706 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5707 const char *dev_name, void *data)
5709 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5712 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
5713 static inline void register_as_ext2(void)
5715 int err = register_filesystem(&ext2_fs_type);
5718 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5721 static inline void unregister_as_ext2(void)
5723 unregister_filesystem(&ext2_fs_type);
5726 static inline int ext2_feature_set_ok(struct super_block *sb)
5728 if (ext4_has_unknown_ext2_incompat_features(sb))
5730 if (sb->s_flags & MS_RDONLY)
5732 if (ext4_has_unknown_ext2_ro_compat_features(sb))
5737 static inline void register_as_ext2(void) { }
5738 static inline void unregister_as_ext2(void) { }
5739 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5742 static inline void register_as_ext3(void)
5744 int err = register_filesystem(&ext3_fs_type);
5747 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5750 static inline void unregister_as_ext3(void)
5752 unregister_filesystem(&ext3_fs_type);
5755 static inline int ext3_feature_set_ok(struct super_block *sb)
5757 if (ext4_has_unknown_ext3_incompat_features(sb))
5759 if (!ext4_has_feature_journal(sb))
5761 if (sb->s_flags & MS_RDONLY)
5763 if (ext4_has_unknown_ext3_ro_compat_features(sb))
5768 static struct file_system_type ext4_fs_type = {
5769 .owner = THIS_MODULE,
5771 .mount = ext4_mount,
5772 .kill_sb = kill_block_super,
5773 .fs_flags = FS_REQUIRES_DEV,
5775 MODULE_ALIAS_FS("ext4");
5777 /* Shared across all ext4 file systems */
5778 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5780 static int __init ext4_init_fs(void)
5784 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
5785 ext4_li_info = NULL;
5786 mutex_init(&ext4_li_mtx);
5788 /* Build-time check for flags consistency */
5789 ext4_check_flag_values();
5791 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
5792 init_waitqueue_head(&ext4__ioend_wq[i]);
5794 err = ext4_init_es();
5798 err = ext4_init_pageio();
5802 err = ext4_init_system_zone();
5806 err = ext4_init_sysfs();
5810 err = ext4_init_mballoc();
5813 err = init_inodecache();
5818 err = register_filesystem(&ext4_fs_type);
5824 unregister_as_ext2();
5825 unregister_as_ext3();
5826 destroy_inodecache();
5828 ext4_exit_mballoc();
5832 ext4_exit_system_zone();
5841 static void __exit ext4_exit_fs(void)
5843 ext4_destroy_lazyinit_thread();
5844 unregister_as_ext2();
5845 unregister_as_ext3();
5846 unregister_filesystem(&ext4_fs_type);
5847 destroy_inodecache();
5848 ext4_exit_mballoc();
5850 ext4_exit_system_zone();
5855 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5856 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5857 MODULE_LICENSE("GPL");
5858 module_init(ext4_init_fs)
5859 module_exit(ext4_exit_fs)