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>
43 #include <linux/iversion.h>
45 #include <linux/kthread.h>
46 #include <linux/freezer.h>
49 #include "ext4_extents.h" /* Needed for trace points definition */
50 #include "ext4_jbd2.h"
56 #define CREATE_TRACE_POINTS
57 #include <trace/events/ext4.h>
59 static struct ext4_lazy_init *ext4_li_info;
60 static struct mutex ext4_li_mtx;
61 static struct ratelimit_state ext4_mount_msg_ratelimit;
63 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
64 unsigned long journal_devnum);
65 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
66 static int ext4_commit_super(struct super_block *sb, int sync);
67 static void ext4_mark_recovery_complete(struct super_block *sb,
68 struct ext4_super_block *es);
69 static void ext4_clear_journal_err(struct super_block *sb,
70 struct ext4_super_block *es);
71 static int ext4_sync_fs(struct super_block *sb, int wait);
72 static int ext4_remount(struct super_block *sb, int *flags, char *data);
73 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
74 static int ext4_unfreeze(struct super_block *sb);
75 static int ext4_freeze(struct super_block *sb);
76 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
77 const char *dev_name, void *data);
78 static inline int ext2_feature_set_ok(struct super_block *sb);
79 static inline int ext3_feature_set_ok(struct super_block *sb);
80 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
81 static void ext4_destroy_lazyinit_thread(void);
82 static void ext4_unregister_li_request(struct super_block *sb);
83 static void ext4_clear_request_list(void);
84 static struct inode *ext4_get_journal_inode(struct super_block *sb,
85 unsigned int journal_inum);
90 * Note the difference between i_mmap_sem (EXT4_I(inode)->i_mmap_sem) and
91 * i_mmap_rwsem (inode->i_mmap_rwsem)!
94 * mmap_sem -> sb_start_pagefault -> i_mmap_sem (r) -> transaction start ->
95 * page lock -> i_data_sem (rw)
97 * buffered write path:
98 * sb_start_write -> i_mutex -> mmap_sem
99 * sb_start_write -> i_mutex -> transaction start -> page lock ->
103 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) ->
104 * i_mmap_rwsem (w) -> page lock
105 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) ->
106 * transaction start -> i_data_sem (rw)
109 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) -> mmap_sem
110 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) ->
111 * transaction start -> i_data_sem (rw)
114 * transaction start -> page lock(s) -> i_data_sem (rw)
117 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
118 static struct file_system_type ext2_fs_type = {
119 .owner = THIS_MODULE,
122 .kill_sb = kill_block_super,
123 .fs_flags = FS_REQUIRES_DEV,
125 MODULE_ALIAS_FS("ext2");
126 MODULE_ALIAS("ext2");
127 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
129 #define IS_EXT2_SB(sb) (0)
133 static struct file_system_type ext3_fs_type = {
134 .owner = THIS_MODULE,
137 .kill_sb = kill_block_super,
138 .fs_flags = FS_REQUIRES_DEV,
140 MODULE_ALIAS_FS("ext3");
141 MODULE_ALIAS("ext3");
142 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
144 static int ext4_verify_csum_type(struct super_block *sb,
145 struct ext4_super_block *es)
147 if (!ext4_has_feature_metadata_csum(sb))
150 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
153 static __le32 ext4_superblock_csum(struct super_block *sb,
154 struct ext4_super_block *es)
156 struct ext4_sb_info *sbi = EXT4_SB(sb);
157 int offset = offsetof(struct ext4_super_block, s_checksum);
160 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
162 return cpu_to_le32(csum);
165 static int ext4_superblock_csum_verify(struct super_block *sb,
166 struct ext4_super_block *es)
168 if (!ext4_has_metadata_csum(sb))
171 return es->s_checksum == ext4_superblock_csum(sb, es);
174 void ext4_superblock_csum_set(struct super_block *sb)
176 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
178 if (!ext4_has_metadata_csum(sb))
181 es->s_checksum = ext4_superblock_csum(sb, es);
184 void *ext4_kvmalloc(size_t size, gfp_t flags)
188 ret = kmalloc(size, flags | __GFP_NOWARN);
190 ret = __vmalloc(size, flags, PAGE_KERNEL);
194 void *ext4_kvzalloc(size_t size, gfp_t flags)
198 ret = kzalloc(size, flags | __GFP_NOWARN);
200 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
204 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
205 struct ext4_group_desc *bg)
207 return le32_to_cpu(bg->bg_block_bitmap_lo) |
208 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
209 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
212 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
213 struct ext4_group_desc *bg)
215 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
216 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
217 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
220 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
221 struct ext4_group_desc *bg)
223 return le32_to_cpu(bg->bg_inode_table_lo) |
224 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
225 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
228 __u32 ext4_free_group_clusters(struct super_block *sb,
229 struct ext4_group_desc *bg)
231 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
232 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
233 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
236 __u32 ext4_free_inodes_count(struct super_block *sb,
237 struct ext4_group_desc *bg)
239 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
240 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
241 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
244 __u32 ext4_used_dirs_count(struct super_block *sb,
245 struct ext4_group_desc *bg)
247 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
248 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
249 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
252 __u32 ext4_itable_unused_count(struct super_block *sb,
253 struct ext4_group_desc *bg)
255 return le16_to_cpu(bg->bg_itable_unused_lo) |
256 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
257 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
260 void ext4_block_bitmap_set(struct super_block *sb,
261 struct ext4_group_desc *bg, ext4_fsblk_t blk)
263 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
264 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
265 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
268 void ext4_inode_bitmap_set(struct super_block *sb,
269 struct ext4_group_desc *bg, ext4_fsblk_t blk)
271 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
272 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
273 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
276 void ext4_inode_table_set(struct super_block *sb,
277 struct ext4_group_desc *bg, ext4_fsblk_t blk)
279 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
280 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
281 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
284 void ext4_free_group_clusters_set(struct super_block *sb,
285 struct ext4_group_desc *bg, __u32 count)
287 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
288 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
289 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
292 void ext4_free_inodes_set(struct super_block *sb,
293 struct ext4_group_desc *bg, __u32 count)
295 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
296 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
297 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
300 void ext4_used_dirs_set(struct super_block *sb,
301 struct ext4_group_desc *bg, __u32 count)
303 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
304 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
305 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
308 void ext4_itable_unused_set(struct super_block *sb,
309 struct ext4_group_desc *bg, __u32 count)
311 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
312 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
313 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
317 static void __save_error_info(struct super_block *sb, const char *func,
320 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
322 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
323 if (bdev_read_only(sb->s_bdev))
325 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
326 es->s_last_error_time = cpu_to_le32(get_seconds());
327 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
328 es->s_last_error_line = cpu_to_le32(line);
329 if (!es->s_first_error_time) {
330 es->s_first_error_time = es->s_last_error_time;
331 strncpy(es->s_first_error_func, func,
332 sizeof(es->s_first_error_func));
333 es->s_first_error_line = cpu_to_le32(line);
334 es->s_first_error_ino = es->s_last_error_ino;
335 es->s_first_error_block = es->s_last_error_block;
338 * Start the daily error reporting function if it hasn't been
341 if (!es->s_error_count)
342 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
343 le32_add_cpu(&es->s_error_count, 1);
346 static void save_error_info(struct super_block *sb, const char *func,
349 __save_error_info(sb, func, line);
350 ext4_commit_super(sb, 1);
354 * The del_gendisk() function uninitializes the disk-specific data
355 * structures, including the bdi structure, without telling anyone
356 * else. Once this happens, any attempt to call mark_buffer_dirty()
357 * (for example, by ext4_commit_super), will cause a kernel OOPS.
358 * This is a kludge to prevent these oops until we can put in a proper
359 * hook in del_gendisk() to inform the VFS and file system layers.
361 static int block_device_ejected(struct super_block *sb)
363 struct inode *bd_inode = sb->s_bdev->bd_inode;
364 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
366 return bdi->dev == NULL;
369 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
371 struct super_block *sb = journal->j_private;
372 struct ext4_sb_info *sbi = EXT4_SB(sb);
373 int error = is_journal_aborted(journal);
374 struct ext4_journal_cb_entry *jce;
376 BUG_ON(txn->t_state == T_FINISHED);
378 ext4_process_freed_data(sb, txn->t_tid);
380 spin_lock(&sbi->s_md_lock);
381 while (!list_empty(&txn->t_private_list)) {
382 jce = list_entry(txn->t_private_list.next,
383 struct ext4_journal_cb_entry, jce_list);
384 list_del_init(&jce->jce_list);
385 spin_unlock(&sbi->s_md_lock);
386 jce->jce_func(sb, jce, error);
387 spin_lock(&sbi->s_md_lock);
389 spin_unlock(&sbi->s_md_lock);
392 /* Deal with the reporting of failure conditions on a filesystem such as
393 * inconsistencies detected or read IO failures.
395 * On ext2, we can store the error state of the filesystem in the
396 * superblock. That is not possible on ext4, because we may have other
397 * write ordering constraints on the superblock which prevent us from
398 * writing it out straight away; and given that the journal is about to
399 * be aborted, we can't rely on the current, or future, transactions to
400 * write out the superblock safely.
402 * We'll just use the jbd2_journal_abort() error code to record an error in
403 * the journal instead. On recovery, the journal will complain about
404 * that error until we've noted it down and cleared it.
407 static void ext4_handle_error(struct super_block *sb)
412 if (!test_opt(sb, ERRORS_CONT)) {
413 journal_t *journal = EXT4_SB(sb)->s_journal;
415 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
417 jbd2_journal_abort(journal, -EIO);
419 if (test_opt(sb, ERRORS_RO)) {
420 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
422 * Make sure updated value of ->s_mount_flags will be visible
423 * before ->s_flags update
426 sb->s_flags |= SB_RDONLY;
428 if (test_opt(sb, ERRORS_PANIC)) {
429 if (EXT4_SB(sb)->s_journal &&
430 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
432 panic("EXT4-fs (device %s): panic forced after error\n",
437 #define ext4_error_ratelimit(sb) \
438 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
441 void __ext4_error(struct super_block *sb, const char *function,
442 unsigned int line, const char *fmt, ...)
444 struct va_format vaf;
447 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
450 if (ext4_error_ratelimit(sb)) {
455 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
456 sb->s_id, function, line, current->comm, &vaf);
459 save_error_info(sb, function, line);
460 ext4_handle_error(sb);
463 void __ext4_error_inode(struct inode *inode, const char *function,
464 unsigned int line, ext4_fsblk_t block,
465 const char *fmt, ...)
468 struct va_format vaf;
469 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
471 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
474 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
475 es->s_last_error_block = cpu_to_le64(block);
476 if (ext4_error_ratelimit(inode->i_sb)) {
481 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
482 "inode #%lu: block %llu: comm %s: %pV\n",
483 inode->i_sb->s_id, function, line, inode->i_ino,
484 block, current->comm, &vaf);
486 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
487 "inode #%lu: comm %s: %pV\n",
488 inode->i_sb->s_id, function, line, inode->i_ino,
489 current->comm, &vaf);
492 save_error_info(inode->i_sb, function, line);
493 ext4_handle_error(inode->i_sb);
496 void __ext4_error_file(struct file *file, const char *function,
497 unsigned int line, ext4_fsblk_t block,
498 const char *fmt, ...)
501 struct va_format vaf;
502 struct ext4_super_block *es;
503 struct inode *inode = file_inode(file);
504 char pathname[80], *path;
506 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
509 es = EXT4_SB(inode->i_sb)->s_es;
510 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
511 if (ext4_error_ratelimit(inode->i_sb)) {
512 path = file_path(file, pathname, sizeof(pathname));
520 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
521 "block %llu: comm %s: path %s: %pV\n",
522 inode->i_sb->s_id, function, line, inode->i_ino,
523 block, current->comm, path, &vaf);
526 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
527 "comm %s: path %s: %pV\n",
528 inode->i_sb->s_id, function, line, inode->i_ino,
529 current->comm, path, &vaf);
532 save_error_info(inode->i_sb, function, line);
533 ext4_handle_error(inode->i_sb);
536 const char *ext4_decode_error(struct super_block *sb, int errno,
543 errstr = "Corrupt filesystem";
546 errstr = "Filesystem failed CRC";
549 errstr = "IO failure";
552 errstr = "Out of memory";
555 if (!sb || (EXT4_SB(sb)->s_journal &&
556 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
557 errstr = "Journal has aborted";
559 errstr = "Readonly filesystem";
562 /* If the caller passed in an extra buffer for unknown
563 * errors, textualise them now. Else we just return
566 /* Check for truncated error codes... */
567 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
576 /* __ext4_std_error decodes expected errors from journaling functions
577 * automatically and invokes the appropriate error response. */
579 void __ext4_std_error(struct super_block *sb, const char *function,
580 unsigned int line, int errno)
585 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
588 /* Special case: if the error is EROFS, and we're not already
589 * inside a transaction, then there's really no point in logging
591 if (errno == -EROFS && journal_current_handle() == NULL && sb_rdonly(sb))
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_rdonly(sb) == 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 |= SB_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_rdonly(sb) && !aborted) {
893 ext4_clear_feature_journal_needs_recovery(sb);
894 es->s_state = cpu_to_le16(sbi->s_mount_state);
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);
954 fs_put_dax(sbi->s_daxdev);
958 static struct kmem_cache *ext4_inode_cachep;
961 * Called inside transaction, so use GFP_NOFS
963 static struct inode *ext4_alloc_inode(struct super_block *sb)
965 struct ext4_inode_info *ei;
967 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
971 inode_set_iversion(&ei->vfs_inode, 1);
972 spin_lock_init(&ei->i_raw_lock);
973 INIT_LIST_HEAD(&ei->i_prealloc_list);
974 spin_lock_init(&ei->i_prealloc_lock);
975 ext4_es_init_tree(&ei->i_es_tree);
976 rwlock_init(&ei->i_es_lock);
977 INIT_LIST_HEAD(&ei->i_es_list);
980 ei->i_es_shrink_lblk = 0;
981 ei->i_reserved_data_blocks = 0;
982 ei->i_da_metadata_calc_len = 0;
983 ei->i_da_metadata_calc_last_lblock = 0;
984 spin_lock_init(&(ei->i_block_reservation_lock));
986 ei->i_reserved_quota = 0;
987 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
990 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
991 spin_lock_init(&ei->i_completed_io_lock);
993 ei->i_datasync_tid = 0;
994 atomic_set(&ei->i_unwritten, 0);
995 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
996 return &ei->vfs_inode;
999 static int ext4_drop_inode(struct inode *inode)
1001 int drop = generic_drop_inode(inode);
1003 trace_ext4_drop_inode(inode, drop);
1007 static void ext4_i_callback(struct rcu_head *head)
1009 struct inode *inode = container_of(head, struct inode, i_rcu);
1010 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
1013 static void ext4_destroy_inode(struct inode *inode)
1015 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
1016 ext4_msg(inode->i_sb, KERN_ERR,
1017 "Inode %lu (%p): orphan list check failed!",
1018 inode->i_ino, EXT4_I(inode));
1019 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1020 EXT4_I(inode), sizeof(struct ext4_inode_info),
1024 call_rcu(&inode->i_rcu, ext4_i_callback);
1027 static void init_once(void *foo)
1029 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
1031 INIT_LIST_HEAD(&ei->i_orphan);
1032 init_rwsem(&ei->xattr_sem);
1033 init_rwsem(&ei->i_data_sem);
1034 init_rwsem(&ei->i_mmap_sem);
1035 inode_init_once(&ei->vfs_inode);
1038 static int __init init_inodecache(void)
1040 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
1041 sizeof(struct ext4_inode_info),
1042 0, (SLAB_RECLAIM_ACCOUNT|
1043 SLAB_MEM_SPREAD|SLAB_ACCOUNT),
1045 if (ext4_inode_cachep == NULL)
1050 static void destroy_inodecache(void)
1053 * Make sure all delayed rcu free inodes are flushed before we
1057 kmem_cache_destroy(ext4_inode_cachep);
1060 void ext4_clear_inode(struct inode *inode)
1062 invalidate_inode_buffers(inode);
1065 ext4_discard_preallocations(inode);
1066 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
1067 if (EXT4_I(inode)->jinode) {
1068 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1069 EXT4_I(inode)->jinode);
1070 jbd2_free_inode(EXT4_I(inode)->jinode);
1071 EXT4_I(inode)->jinode = NULL;
1073 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1074 fscrypt_put_encryption_info(inode, NULL);
1078 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1079 u64 ino, u32 generation)
1081 struct inode *inode;
1083 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1084 return ERR_PTR(-ESTALE);
1085 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1086 return ERR_PTR(-ESTALE);
1088 /* iget isn't really right if the inode is currently unallocated!!
1090 * ext4_read_inode will return a bad_inode if the inode had been
1091 * deleted, so we should be safe.
1093 * Currently we don't know the generation for parent directory, so
1094 * a generation of 0 means "accept any"
1096 inode = ext4_iget_normal(sb, ino);
1098 return ERR_CAST(inode);
1099 if (generation && inode->i_generation != generation) {
1101 return ERR_PTR(-ESTALE);
1107 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1108 int fh_len, int fh_type)
1110 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1111 ext4_nfs_get_inode);
1114 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1115 int fh_len, int fh_type)
1117 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1118 ext4_nfs_get_inode);
1122 * Try to release metadata pages (indirect blocks, directories) which are
1123 * mapped via the block device. Since these pages could have journal heads
1124 * which would prevent try_to_free_buffers() from freeing them, we must use
1125 * jbd2 layer's try_to_free_buffers() function to release them.
1127 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1130 journal_t *journal = EXT4_SB(sb)->s_journal;
1132 WARN_ON(PageChecked(page));
1133 if (!page_has_buffers(page))
1136 return jbd2_journal_try_to_free_buffers(journal, page,
1137 wait & ~__GFP_DIRECT_RECLAIM);
1138 return try_to_free_buffers(page);
1141 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1142 static int ext4_get_context(struct inode *inode, void *ctx, size_t len)
1144 return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1145 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len);
1148 static int ext4_set_context(struct inode *inode, const void *ctx, size_t len,
1151 handle_t *handle = fs_data;
1152 int res, res2, credits, retries = 0;
1155 * Encrypting the root directory is not allowed because e2fsck expects
1156 * lost+found to exist and be unencrypted, and encrypting the root
1157 * directory would imply encrypting the lost+found directory as well as
1158 * the filename "lost+found" itself.
1160 if (inode->i_ino == EXT4_ROOT_INO)
1163 if (WARN_ON_ONCE(IS_DAX(inode) && i_size_read(inode)))
1166 res = ext4_convert_inline_data(inode);
1171 * If a journal handle was specified, then the encryption context is
1172 * being set on a new inode via inheritance and is part of a larger
1173 * transaction to create the inode. Otherwise the encryption context is
1174 * being set on an existing inode in its own transaction. Only in the
1175 * latter case should the "retry on ENOSPC" logic be used.
1179 res = ext4_xattr_set_handle(handle, inode,
1180 EXT4_XATTR_INDEX_ENCRYPTION,
1181 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1184 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1185 ext4_clear_inode_state(inode,
1186 EXT4_STATE_MAY_INLINE_DATA);
1188 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1189 * S_DAX may be disabled
1191 ext4_set_inode_flags(inode);
1196 res = dquot_initialize(inode);
1200 res = ext4_xattr_set_credits(inode, len, false /* is_create */,
1205 handle = ext4_journal_start(inode, EXT4_HT_MISC, credits);
1207 return PTR_ERR(handle);
1209 res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION,
1210 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1213 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1215 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1216 * S_DAX may be disabled
1218 ext4_set_inode_flags(inode);
1219 res = ext4_mark_inode_dirty(handle, inode);
1221 EXT4_ERROR_INODE(inode, "Failed to mark inode dirty");
1223 res2 = ext4_journal_stop(handle);
1225 if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
1232 static bool ext4_dummy_context(struct inode *inode)
1234 return DUMMY_ENCRYPTION_ENABLED(EXT4_SB(inode->i_sb));
1237 static unsigned ext4_max_namelen(struct inode *inode)
1239 return S_ISLNK(inode->i_mode) ? inode->i_sb->s_blocksize :
1243 static const struct fscrypt_operations ext4_cryptops = {
1244 .key_prefix = "ext4:",
1245 .get_context = ext4_get_context,
1246 .set_context = ext4_set_context,
1247 .dummy_context = ext4_dummy_context,
1248 .empty_dir = ext4_empty_dir,
1249 .max_namelen = ext4_max_namelen,
1254 static const char * const quotatypes[] = INITQFNAMES;
1255 #define QTYPE2NAME(t) (quotatypes[t])
1257 static int ext4_write_dquot(struct dquot *dquot);
1258 static int ext4_acquire_dquot(struct dquot *dquot);
1259 static int ext4_release_dquot(struct dquot *dquot);
1260 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1261 static int ext4_write_info(struct super_block *sb, int type);
1262 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1263 const struct path *path);
1264 static int ext4_quota_on_mount(struct super_block *sb, int type);
1265 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1266 size_t len, loff_t off);
1267 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1268 const char *data, size_t len, loff_t off);
1269 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1270 unsigned int flags);
1271 static int ext4_enable_quotas(struct super_block *sb);
1272 static int ext4_get_next_id(struct super_block *sb, struct kqid *qid);
1274 static struct dquot **ext4_get_dquots(struct inode *inode)
1276 return EXT4_I(inode)->i_dquot;
1279 static const struct dquot_operations ext4_quota_operations = {
1280 .get_reserved_space = ext4_get_reserved_space,
1281 .write_dquot = ext4_write_dquot,
1282 .acquire_dquot = ext4_acquire_dquot,
1283 .release_dquot = ext4_release_dquot,
1284 .mark_dirty = ext4_mark_dquot_dirty,
1285 .write_info = ext4_write_info,
1286 .alloc_dquot = dquot_alloc,
1287 .destroy_dquot = dquot_destroy,
1288 .get_projid = ext4_get_projid,
1289 .get_inode_usage = ext4_get_inode_usage,
1290 .get_next_id = ext4_get_next_id,
1293 static const struct quotactl_ops ext4_qctl_operations = {
1294 .quota_on = ext4_quota_on,
1295 .quota_off = ext4_quota_off,
1296 .quota_sync = dquot_quota_sync,
1297 .get_state = dquot_get_state,
1298 .set_info = dquot_set_dqinfo,
1299 .get_dqblk = dquot_get_dqblk,
1300 .set_dqblk = dquot_set_dqblk,
1301 .get_nextdqblk = dquot_get_next_dqblk,
1305 static const struct super_operations ext4_sops = {
1306 .alloc_inode = ext4_alloc_inode,
1307 .destroy_inode = ext4_destroy_inode,
1308 .write_inode = ext4_write_inode,
1309 .dirty_inode = ext4_dirty_inode,
1310 .drop_inode = ext4_drop_inode,
1311 .evict_inode = ext4_evict_inode,
1312 .put_super = ext4_put_super,
1313 .sync_fs = ext4_sync_fs,
1314 .freeze_fs = ext4_freeze,
1315 .unfreeze_fs = ext4_unfreeze,
1316 .statfs = ext4_statfs,
1317 .remount_fs = ext4_remount,
1318 .show_options = ext4_show_options,
1320 .quota_read = ext4_quota_read,
1321 .quota_write = ext4_quota_write,
1322 .get_dquots = ext4_get_dquots,
1324 .bdev_try_to_free_page = bdev_try_to_free_page,
1327 static const struct export_operations ext4_export_ops = {
1328 .fh_to_dentry = ext4_fh_to_dentry,
1329 .fh_to_parent = ext4_fh_to_parent,
1330 .get_parent = ext4_get_parent,
1334 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1335 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1336 Opt_nouid32, Opt_debug, Opt_removed,
1337 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1338 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1339 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1340 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1341 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1342 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1343 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1344 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1345 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1346 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version, Opt_dax,
1347 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1348 Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize,
1349 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1350 Opt_inode_readahead_blks, Opt_journal_ioprio,
1351 Opt_dioread_nolock, Opt_dioread_lock,
1352 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1353 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache,
1356 static const match_table_t tokens = {
1357 {Opt_bsd_df, "bsddf"},
1358 {Opt_minix_df, "minixdf"},
1359 {Opt_grpid, "grpid"},
1360 {Opt_grpid, "bsdgroups"},
1361 {Opt_nogrpid, "nogrpid"},
1362 {Opt_nogrpid, "sysvgroups"},
1363 {Opt_resgid, "resgid=%u"},
1364 {Opt_resuid, "resuid=%u"},
1366 {Opt_err_cont, "errors=continue"},
1367 {Opt_err_panic, "errors=panic"},
1368 {Opt_err_ro, "errors=remount-ro"},
1369 {Opt_nouid32, "nouid32"},
1370 {Opt_debug, "debug"},
1371 {Opt_removed, "oldalloc"},
1372 {Opt_removed, "orlov"},
1373 {Opt_user_xattr, "user_xattr"},
1374 {Opt_nouser_xattr, "nouser_xattr"},
1376 {Opt_noacl, "noacl"},
1377 {Opt_noload, "norecovery"},
1378 {Opt_noload, "noload"},
1379 {Opt_removed, "nobh"},
1380 {Opt_removed, "bh"},
1381 {Opt_commit, "commit=%u"},
1382 {Opt_min_batch_time, "min_batch_time=%u"},
1383 {Opt_max_batch_time, "max_batch_time=%u"},
1384 {Opt_journal_dev, "journal_dev=%u"},
1385 {Opt_journal_path, "journal_path=%s"},
1386 {Opt_journal_checksum, "journal_checksum"},
1387 {Opt_nojournal_checksum, "nojournal_checksum"},
1388 {Opt_journal_async_commit, "journal_async_commit"},
1389 {Opt_abort, "abort"},
1390 {Opt_data_journal, "data=journal"},
1391 {Opt_data_ordered, "data=ordered"},
1392 {Opt_data_writeback, "data=writeback"},
1393 {Opt_data_err_abort, "data_err=abort"},
1394 {Opt_data_err_ignore, "data_err=ignore"},
1395 {Opt_offusrjquota, "usrjquota="},
1396 {Opt_usrjquota, "usrjquota=%s"},
1397 {Opt_offgrpjquota, "grpjquota="},
1398 {Opt_grpjquota, "grpjquota=%s"},
1399 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1400 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1401 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1402 {Opt_grpquota, "grpquota"},
1403 {Opt_noquota, "noquota"},
1404 {Opt_quota, "quota"},
1405 {Opt_usrquota, "usrquota"},
1406 {Opt_prjquota, "prjquota"},
1407 {Opt_barrier, "barrier=%u"},
1408 {Opt_barrier, "barrier"},
1409 {Opt_nobarrier, "nobarrier"},
1410 {Opt_i_version, "i_version"},
1412 {Opt_stripe, "stripe=%u"},
1413 {Opt_delalloc, "delalloc"},
1414 {Opt_lazytime, "lazytime"},
1415 {Opt_nolazytime, "nolazytime"},
1416 {Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"},
1417 {Opt_nodelalloc, "nodelalloc"},
1418 {Opt_removed, "mblk_io_submit"},
1419 {Opt_removed, "nomblk_io_submit"},
1420 {Opt_block_validity, "block_validity"},
1421 {Opt_noblock_validity, "noblock_validity"},
1422 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1423 {Opt_journal_ioprio, "journal_ioprio=%u"},
1424 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1425 {Opt_auto_da_alloc, "auto_da_alloc"},
1426 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1427 {Opt_dioread_nolock, "dioread_nolock"},
1428 {Opt_dioread_lock, "dioread_lock"},
1429 {Opt_discard, "discard"},
1430 {Opt_nodiscard, "nodiscard"},
1431 {Opt_init_itable, "init_itable=%u"},
1432 {Opt_init_itable, "init_itable"},
1433 {Opt_noinit_itable, "noinit_itable"},
1434 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1435 {Opt_test_dummy_encryption, "test_dummy_encryption"},
1436 {Opt_nombcache, "nombcache"},
1437 {Opt_nombcache, "no_mbcache"}, /* for backward compatibility */
1438 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1439 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1440 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1441 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1442 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1446 static ext4_fsblk_t get_sb_block(void **data)
1448 ext4_fsblk_t sb_block;
1449 char *options = (char *) *data;
1451 if (!options || strncmp(options, "sb=", 3) != 0)
1452 return 1; /* Default location */
1455 /* TODO: use simple_strtoll with >32bit ext4 */
1456 sb_block = simple_strtoul(options, &options, 0);
1457 if (*options && *options != ',') {
1458 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1462 if (*options == ',')
1464 *data = (void *) options;
1469 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1470 static const char deprecated_msg[] =
1471 "Mount option \"%s\" will be removed by %s\n"
1472 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1475 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1477 struct ext4_sb_info *sbi = EXT4_SB(sb);
1481 if (sb_any_quota_loaded(sb) &&
1482 !sbi->s_qf_names[qtype]) {
1483 ext4_msg(sb, KERN_ERR,
1484 "Cannot change journaled "
1485 "quota options when quota turned on");
1488 if (ext4_has_feature_quota(sb)) {
1489 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1490 "ignored when QUOTA feature is enabled");
1493 qname = match_strdup(args);
1495 ext4_msg(sb, KERN_ERR,
1496 "Not enough memory for storing quotafile name");
1499 if (sbi->s_qf_names[qtype]) {
1500 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1503 ext4_msg(sb, KERN_ERR,
1504 "%s quota file already specified",
1508 if (strchr(qname, '/')) {
1509 ext4_msg(sb, KERN_ERR,
1510 "quotafile must be on filesystem root");
1513 sbi->s_qf_names[qtype] = qname;
1521 static int clear_qf_name(struct super_block *sb, int qtype)
1524 struct ext4_sb_info *sbi = EXT4_SB(sb);
1526 if (sb_any_quota_loaded(sb) &&
1527 sbi->s_qf_names[qtype]) {
1528 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1529 " when quota turned on");
1532 kfree(sbi->s_qf_names[qtype]);
1533 sbi->s_qf_names[qtype] = NULL;
1538 #define MOPT_SET 0x0001
1539 #define MOPT_CLEAR 0x0002
1540 #define MOPT_NOSUPPORT 0x0004
1541 #define MOPT_EXPLICIT 0x0008
1542 #define MOPT_CLEAR_ERR 0x0010
1543 #define MOPT_GTE0 0x0020
1546 #define MOPT_QFMT 0x0040
1548 #define MOPT_Q MOPT_NOSUPPORT
1549 #define MOPT_QFMT MOPT_NOSUPPORT
1551 #define MOPT_DATAJ 0x0080
1552 #define MOPT_NO_EXT2 0x0100
1553 #define MOPT_NO_EXT3 0x0200
1554 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1555 #define MOPT_STRING 0x0400
1557 static const struct mount_opts {
1561 } ext4_mount_opts[] = {
1562 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1563 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1564 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1565 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1566 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1567 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1568 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1569 MOPT_EXT4_ONLY | MOPT_SET},
1570 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1571 MOPT_EXT4_ONLY | MOPT_CLEAR},
1572 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1573 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1574 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1575 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1576 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1577 MOPT_EXT4_ONLY | MOPT_CLEAR},
1578 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1579 MOPT_EXT4_ONLY | MOPT_CLEAR},
1580 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1581 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1582 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1583 EXT4_MOUNT_JOURNAL_CHECKSUM),
1584 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1585 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1586 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1587 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1588 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1589 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1591 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1593 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1594 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1595 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1596 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1597 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1598 {Opt_commit, 0, MOPT_GTE0},
1599 {Opt_max_batch_time, 0, MOPT_GTE0},
1600 {Opt_min_batch_time, 0, MOPT_GTE0},
1601 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1602 {Opt_init_itable, 0, MOPT_GTE0},
1603 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
1604 {Opt_stripe, 0, MOPT_GTE0},
1605 {Opt_resuid, 0, MOPT_GTE0},
1606 {Opt_resgid, 0, MOPT_GTE0},
1607 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1608 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1609 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1610 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1611 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1612 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1613 MOPT_NO_EXT2 | MOPT_DATAJ},
1614 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1615 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1616 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1617 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1618 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1620 {Opt_acl, 0, MOPT_NOSUPPORT},
1621 {Opt_noacl, 0, MOPT_NOSUPPORT},
1623 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1624 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1625 {Opt_debug_want_extra_isize, 0, MOPT_GTE0},
1626 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1627 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1629 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1631 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1633 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1634 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1635 MOPT_CLEAR | MOPT_Q},
1636 {Opt_usrjquota, 0, MOPT_Q},
1637 {Opt_grpjquota, 0, MOPT_Q},
1638 {Opt_offusrjquota, 0, MOPT_Q},
1639 {Opt_offgrpjquota, 0, MOPT_Q},
1640 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1641 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1642 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1643 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1644 {Opt_test_dummy_encryption, 0, MOPT_GTE0},
1645 {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET},
1649 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1650 substring_t *args, unsigned long *journal_devnum,
1651 unsigned int *journal_ioprio, int is_remount)
1653 struct ext4_sb_info *sbi = EXT4_SB(sb);
1654 const struct mount_opts *m;
1660 if (token == Opt_usrjquota)
1661 return set_qf_name(sb, USRQUOTA, &args[0]);
1662 else if (token == Opt_grpjquota)
1663 return set_qf_name(sb, GRPQUOTA, &args[0]);
1664 else if (token == Opt_offusrjquota)
1665 return clear_qf_name(sb, USRQUOTA);
1666 else if (token == Opt_offgrpjquota)
1667 return clear_qf_name(sb, GRPQUOTA);
1671 case Opt_nouser_xattr:
1672 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1675 return 1; /* handled by get_sb_block() */
1677 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1680 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1683 sb->s_flags |= SB_I_VERSION;
1686 sb->s_flags |= SB_LAZYTIME;
1688 case Opt_nolazytime:
1689 sb->s_flags &= ~SB_LAZYTIME;
1693 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1694 if (token == m->token)
1697 if (m->token == Opt_err) {
1698 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1699 "or missing value", opt);
1703 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1704 ext4_msg(sb, KERN_ERR,
1705 "Mount option \"%s\" incompatible with ext2", opt);
1708 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1709 ext4_msg(sb, KERN_ERR,
1710 "Mount option \"%s\" incompatible with ext3", opt);
1714 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
1716 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1718 if (m->flags & MOPT_EXPLICIT) {
1719 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
1720 set_opt2(sb, EXPLICIT_DELALLOC);
1721 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
1722 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
1726 if (m->flags & MOPT_CLEAR_ERR)
1727 clear_opt(sb, ERRORS_MASK);
1728 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1729 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1730 "options when quota turned on");
1734 if (m->flags & MOPT_NOSUPPORT) {
1735 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1736 } else if (token == Opt_commit) {
1738 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1739 sbi->s_commit_interval = HZ * arg;
1740 } else if (token == Opt_debug_want_extra_isize) {
1741 sbi->s_want_extra_isize = arg;
1742 } else if (token == Opt_max_batch_time) {
1743 sbi->s_max_batch_time = arg;
1744 } else if (token == Opt_min_batch_time) {
1745 sbi->s_min_batch_time = arg;
1746 } else if (token == Opt_inode_readahead_blks) {
1747 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1748 ext4_msg(sb, KERN_ERR,
1749 "EXT4-fs: inode_readahead_blks must be "
1750 "0 or a power of 2 smaller than 2^31");
1753 sbi->s_inode_readahead_blks = arg;
1754 } else if (token == Opt_init_itable) {
1755 set_opt(sb, INIT_INODE_TABLE);
1757 arg = EXT4_DEF_LI_WAIT_MULT;
1758 sbi->s_li_wait_mult = arg;
1759 } else if (token == Opt_max_dir_size_kb) {
1760 sbi->s_max_dir_size_kb = arg;
1761 } else if (token == Opt_stripe) {
1762 sbi->s_stripe = arg;
1763 } else if (token == Opt_resuid) {
1764 uid = make_kuid(current_user_ns(), arg);
1765 if (!uid_valid(uid)) {
1766 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1769 sbi->s_resuid = uid;
1770 } else if (token == Opt_resgid) {
1771 gid = make_kgid(current_user_ns(), arg);
1772 if (!gid_valid(gid)) {
1773 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1776 sbi->s_resgid = gid;
1777 } else if (token == Opt_journal_dev) {
1779 ext4_msg(sb, KERN_ERR,
1780 "Cannot specify journal on remount");
1783 *journal_devnum = arg;
1784 } else if (token == Opt_journal_path) {
1786 struct inode *journal_inode;
1791 ext4_msg(sb, KERN_ERR,
1792 "Cannot specify journal on remount");
1795 journal_path = match_strdup(&args[0]);
1796 if (!journal_path) {
1797 ext4_msg(sb, KERN_ERR, "error: could not dup "
1798 "journal device string");
1802 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1804 ext4_msg(sb, KERN_ERR, "error: could not find "
1805 "journal device path: error %d", error);
1806 kfree(journal_path);
1810 journal_inode = d_inode(path.dentry);
1811 if (!S_ISBLK(journal_inode->i_mode)) {
1812 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1813 "is not a block device", journal_path);
1815 kfree(journal_path);
1819 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1821 kfree(journal_path);
1822 } else if (token == Opt_journal_ioprio) {
1824 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1829 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1830 } else if (token == Opt_test_dummy_encryption) {
1831 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1832 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1833 ext4_msg(sb, KERN_WARNING,
1834 "Test dummy encryption mode enabled");
1836 ext4_msg(sb, KERN_WARNING,
1837 "Test dummy encryption mount option ignored");
1839 } else if (m->flags & MOPT_DATAJ) {
1841 if (!sbi->s_journal)
1842 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1843 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1844 ext4_msg(sb, KERN_ERR,
1845 "Cannot change data mode on remount");
1849 clear_opt(sb, DATA_FLAGS);
1850 sbi->s_mount_opt |= m->mount_opt;
1853 } else if (m->flags & MOPT_QFMT) {
1854 if (sb_any_quota_loaded(sb) &&
1855 sbi->s_jquota_fmt != m->mount_opt) {
1856 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1857 "quota options when quota turned on");
1860 if (ext4_has_feature_quota(sb)) {
1861 ext4_msg(sb, KERN_INFO,
1862 "Quota format mount options ignored "
1863 "when QUOTA feature is enabled");
1866 sbi->s_jquota_fmt = m->mount_opt;
1868 } else if (token == Opt_dax) {
1869 #ifdef CONFIG_FS_DAX
1870 ext4_msg(sb, KERN_WARNING,
1871 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
1872 sbi->s_mount_opt |= m->mount_opt;
1874 ext4_msg(sb, KERN_INFO, "dax option not supported");
1877 } else if (token == Opt_data_err_abort) {
1878 sbi->s_mount_opt |= m->mount_opt;
1879 } else if (token == Opt_data_err_ignore) {
1880 sbi->s_mount_opt &= ~m->mount_opt;
1884 if (m->flags & MOPT_CLEAR)
1886 else if (unlikely(!(m->flags & MOPT_SET))) {
1887 ext4_msg(sb, KERN_WARNING,
1888 "buggy handling of option %s", opt);
1893 sbi->s_mount_opt |= m->mount_opt;
1895 sbi->s_mount_opt &= ~m->mount_opt;
1900 static int parse_options(char *options, struct super_block *sb,
1901 unsigned long *journal_devnum,
1902 unsigned int *journal_ioprio,
1905 struct ext4_sb_info *sbi = EXT4_SB(sb);
1907 substring_t args[MAX_OPT_ARGS];
1913 while ((p = strsep(&options, ",")) != NULL) {
1917 * Initialize args struct so we know whether arg was
1918 * found; some options take optional arguments.
1920 args[0].to = args[0].from = NULL;
1921 token = match_token(p, tokens, args);
1922 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1923 journal_ioprio, is_remount) < 0)
1928 * We do the test below only for project quotas. 'usrquota' and
1929 * 'grpquota' mount options are allowed even without quota feature
1930 * to support legacy quotas in quota files.
1932 if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) {
1933 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. "
1934 "Cannot enable project quota enforcement.");
1937 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1938 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1939 clear_opt(sb, USRQUOTA);
1941 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1942 clear_opt(sb, GRPQUOTA);
1944 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1945 ext4_msg(sb, KERN_ERR, "old and new quota "
1950 if (!sbi->s_jquota_fmt) {
1951 ext4_msg(sb, KERN_ERR, "journaled quota format "
1957 if (test_opt(sb, DIOREAD_NOLOCK)) {
1959 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1961 if (blocksize < PAGE_SIZE) {
1962 ext4_msg(sb, KERN_ERR, "can't mount with "
1963 "dioread_nolock if block size != PAGE_SIZE");
1970 static inline void ext4_show_quota_options(struct seq_file *seq,
1971 struct super_block *sb)
1973 #if defined(CONFIG_QUOTA)
1974 struct ext4_sb_info *sbi = EXT4_SB(sb);
1976 if (sbi->s_jquota_fmt) {
1979 switch (sbi->s_jquota_fmt) {
1990 seq_printf(seq, ",jqfmt=%s", fmtname);
1993 if (sbi->s_qf_names[USRQUOTA])
1994 seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]);
1996 if (sbi->s_qf_names[GRPQUOTA])
1997 seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]);
2001 static const char *token2str(int token)
2003 const struct match_token *t;
2005 for (t = tokens; t->token != Opt_err; t++)
2006 if (t->token == token && !strchr(t->pattern, '='))
2013 * - it's set to a non-default value OR
2014 * - if the per-sb default is different from the global default
2016 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
2019 struct ext4_sb_info *sbi = EXT4_SB(sb);
2020 struct ext4_super_block *es = sbi->s_es;
2021 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
2022 const struct mount_opts *m;
2023 char sep = nodefs ? '\n' : ',';
2025 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2026 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2028 if (sbi->s_sb_block != 1)
2029 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
2031 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
2032 int want_set = m->flags & MOPT_SET;
2033 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
2034 (m->flags & MOPT_CLEAR_ERR))
2036 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
2037 continue; /* skip if same as the default */
2039 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
2040 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
2041 continue; /* select Opt_noFoo vs Opt_Foo */
2042 SEQ_OPTS_PRINT("%s", token2str(m->token));
2045 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
2046 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
2047 SEQ_OPTS_PRINT("resuid=%u",
2048 from_kuid_munged(&init_user_ns, sbi->s_resuid));
2049 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
2050 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
2051 SEQ_OPTS_PRINT("resgid=%u",
2052 from_kgid_munged(&init_user_ns, sbi->s_resgid));
2053 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
2054 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
2055 SEQ_OPTS_PUTS("errors=remount-ro");
2056 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
2057 SEQ_OPTS_PUTS("errors=continue");
2058 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
2059 SEQ_OPTS_PUTS("errors=panic");
2060 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
2061 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
2062 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
2063 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
2064 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
2065 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2066 if (sb->s_flags & SB_I_VERSION)
2067 SEQ_OPTS_PUTS("i_version");
2068 if (nodefs || sbi->s_stripe)
2069 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
2070 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
2071 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2072 SEQ_OPTS_PUTS("data=journal");
2073 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2074 SEQ_OPTS_PUTS("data=ordered");
2075 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2076 SEQ_OPTS_PUTS("data=writeback");
2079 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
2080 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2081 sbi->s_inode_readahead_blks);
2083 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
2084 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
2085 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
2086 if (nodefs || sbi->s_max_dir_size_kb)
2087 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2088 if (test_opt(sb, DATA_ERR_ABORT))
2089 SEQ_OPTS_PUTS("data_err=abort");
2091 ext4_show_quota_options(seq, sb);
2095 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2097 return _ext4_show_options(seq, root->d_sb, 0);
2100 int ext4_seq_options_show(struct seq_file *seq, void *offset)
2102 struct super_block *sb = seq->private;
2105 seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw");
2106 rc = _ext4_show_options(seq, sb, 1);
2107 seq_puts(seq, "\n");
2111 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
2114 struct ext4_sb_info *sbi = EXT4_SB(sb);
2117 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
2118 ext4_msg(sb, KERN_ERR, "revision level too high, "
2119 "forcing read-only mode");
2124 if (!(sbi->s_mount_state & EXT4_VALID_FS))
2125 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
2126 "running e2fsck is recommended");
2127 else if (sbi->s_mount_state & EXT4_ERROR_FS)
2128 ext4_msg(sb, KERN_WARNING,
2129 "warning: mounting fs with errors, "
2130 "running e2fsck is recommended");
2131 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
2132 le16_to_cpu(es->s_mnt_count) >=
2133 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
2134 ext4_msg(sb, KERN_WARNING,
2135 "warning: maximal mount count reached, "
2136 "running e2fsck is recommended");
2137 else if (le32_to_cpu(es->s_checkinterval) &&
2138 (le32_to_cpu(es->s_lastcheck) +
2139 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
2140 ext4_msg(sb, KERN_WARNING,
2141 "warning: checktime reached, "
2142 "running e2fsck is recommended");
2143 if (!sbi->s_journal)
2144 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
2145 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
2146 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
2147 le16_add_cpu(&es->s_mnt_count, 1);
2148 es->s_mtime = cpu_to_le32(get_seconds());
2149 ext4_update_dynamic_rev(sb);
2151 ext4_set_feature_journal_needs_recovery(sb);
2153 ext4_commit_super(sb, 1);
2155 if (test_opt(sb, DEBUG))
2156 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
2157 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
2159 sbi->s_groups_count,
2160 EXT4_BLOCKS_PER_GROUP(sb),
2161 EXT4_INODES_PER_GROUP(sb),
2162 sbi->s_mount_opt, sbi->s_mount_opt2);
2164 cleancache_init_fs(sb);
2168 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2170 struct ext4_sb_info *sbi = EXT4_SB(sb);
2171 struct flex_groups *new_groups;
2174 if (!sbi->s_log_groups_per_flex)
2177 size = ext4_flex_group(sbi, ngroup - 1) + 1;
2178 if (size <= sbi->s_flex_groups_allocated)
2181 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
2182 new_groups = kvzalloc(size, GFP_KERNEL);
2184 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
2185 size / (int) sizeof(struct flex_groups));
2189 if (sbi->s_flex_groups) {
2190 memcpy(new_groups, sbi->s_flex_groups,
2191 (sbi->s_flex_groups_allocated *
2192 sizeof(struct flex_groups)));
2193 kvfree(sbi->s_flex_groups);
2195 sbi->s_flex_groups = new_groups;
2196 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
2200 static int ext4_fill_flex_info(struct super_block *sb)
2202 struct ext4_sb_info *sbi = EXT4_SB(sb);
2203 struct ext4_group_desc *gdp = NULL;
2204 ext4_group_t flex_group;
2207 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2208 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
2209 sbi->s_log_groups_per_flex = 0;
2213 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2217 for (i = 0; i < sbi->s_groups_count; i++) {
2218 gdp = ext4_get_group_desc(sb, i, NULL);
2220 flex_group = ext4_flex_group(sbi, i);
2221 atomic_add(ext4_free_inodes_count(sb, gdp),
2222 &sbi->s_flex_groups[flex_group].free_inodes);
2223 atomic64_add(ext4_free_group_clusters(sb, gdp),
2224 &sbi->s_flex_groups[flex_group].free_clusters);
2225 atomic_add(ext4_used_dirs_count(sb, gdp),
2226 &sbi->s_flex_groups[flex_group].used_dirs);
2234 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
2235 struct ext4_group_desc *gdp)
2237 int offset = offsetof(struct ext4_group_desc, bg_checksum);
2239 __le32 le_group = cpu_to_le32(block_group);
2240 struct ext4_sb_info *sbi = EXT4_SB(sb);
2242 if (ext4_has_metadata_csum(sbi->s_sb)) {
2243 /* Use new metadata_csum algorithm */
2245 __u16 dummy_csum = 0;
2247 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2249 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2250 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2251 sizeof(dummy_csum));
2252 offset += sizeof(dummy_csum);
2253 if (offset < sbi->s_desc_size)
2254 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2255 sbi->s_desc_size - offset);
2257 crc = csum32 & 0xFFFF;
2261 /* old crc16 code */
2262 if (!ext4_has_feature_gdt_csum(sb))
2265 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2266 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2267 crc = crc16(crc, (__u8 *)gdp, offset);
2268 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2269 /* for checksum of struct ext4_group_desc do the rest...*/
2270 if (ext4_has_feature_64bit(sb) &&
2271 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2272 crc = crc16(crc, (__u8 *)gdp + offset,
2273 le16_to_cpu(sbi->s_es->s_desc_size) -
2277 return cpu_to_le16(crc);
2280 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2281 struct ext4_group_desc *gdp)
2283 if (ext4_has_group_desc_csum(sb) &&
2284 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
2290 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2291 struct ext4_group_desc *gdp)
2293 if (!ext4_has_group_desc_csum(sb))
2295 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
2298 /* Called at mount-time, super-block is locked */
2299 static int ext4_check_descriptors(struct super_block *sb,
2300 ext4_fsblk_t sb_block,
2301 ext4_group_t *first_not_zeroed)
2303 struct ext4_sb_info *sbi = EXT4_SB(sb);
2304 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2305 ext4_fsblk_t last_block;
2306 ext4_fsblk_t block_bitmap;
2307 ext4_fsblk_t inode_bitmap;
2308 ext4_fsblk_t inode_table;
2309 int flexbg_flag = 0;
2310 ext4_group_t i, grp = sbi->s_groups_count;
2312 if (ext4_has_feature_flex_bg(sb))
2315 ext4_debug("Checking group descriptors");
2317 for (i = 0; i < sbi->s_groups_count; i++) {
2318 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2320 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2321 last_block = ext4_blocks_count(sbi->s_es) - 1;
2323 last_block = first_block +
2324 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2326 if ((grp == sbi->s_groups_count) &&
2327 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2330 block_bitmap = ext4_block_bitmap(sb, gdp);
2331 if (block_bitmap == sb_block) {
2332 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2333 "Block bitmap for group %u overlaps "
2336 if (block_bitmap < first_block || block_bitmap > last_block) {
2337 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2338 "Block bitmap for group %u not in group "
2339 "(block %llu)!", i, block_bitmap);
2342 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2343 if (inode_bitmap == sb_block) {
2344 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2345 "Inode bitmap for group %u overlaps "
2348 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2349 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2350 "Inode bitmap for group %u not in group "
2351 "(block %llu)!", i, inode_bitmap);
2354 inode_table = ext4_inode_table(sb, gdp);
2355 if (inode_table == sb_block) {
2356 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2357 "Inode table for group %u overlaps "
2360 if (inode_table < first_block ||
2361 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2362 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2363 "Inode table for group %u not in group "
2364 "(block %llu)!", i, inode_table);
2367 ext4_lock_group(sb, i);
2368 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2369 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2370 "Checksum for group %u failed (%u!=%u)",
2371 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
2372 gdp)), le16_to_cpu(gdp->bg_checksum));
2373 if (!sb_rdonly(sb)) {
2374 ext4_unlock_group(sb, i);
2378 ext4_unlock_group(sb, i);
2380 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2382 if (NULL != first_not_zeroed)
2383 *first_not_zeroed = grp;
2387 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2388 * the superblock) which were deleted from all directories, but held open by
2389 * a process at the time of a crash. We walk the list and try to delete these
2390 * inodes at recovery time (only with a read-write filesystem).
2392 * In order to keep the orphan inode chain consistent during traversal (in
2393 * case of crash during recovery), we link each inode into the superblock
2394 * orphan list_head and handle it the same way as an inode deletion during
2395 * normal operation (which journals the operations for us).
2397 * We only do an iget() and an iput() on each inode, which is very safe if we
2398 * accidentally point at an in-use or already deleted inode. The worst that
2399 * can happen in this case is that we get a "bit already cleared" message from
2400 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2401 * e2fsck was run on this filesystem, and it must have already done the orphan
2402 * inode cleanup for us, so we can safely abort without any further action.
2404 static void ext4_orphan_cleanup(struct super_block *sb,
2405 struct ext4_super_block *es)
2407 unsigned int s_flags = sb->s_flags;
2408 int ret, nr_orphans = 0, nr_truncates = 0;
2410 int quota_update = 0;
2413 if (!es->s_last_orphan) {
2414 jbd_debug(4, "no orphan inodes to clean up\n");
2418 if (bdev_read_only(sb->s_bdev)) {
2419 ext4_msg(sb, KERN_ERR, "write access "
2420 "unavailable, skipping orphan cleanup");
2424 /* Check if feature set would not allow a r/w mount */
2425 if (!ext4_feature_set_ok(sb, 0)) {
2426 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2427 "unknown ROCOMPAT features");
2431 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2432 /* don't clear list on RO mount w/ errors */
2433 if (es->s_last_orphan && !(s_flags & SB_RDONLY)) {
2434 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
2435 "clearing orphan list.\n");
2436 es->s_last_orphan = 0;
2438 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2442 if (s_flags & SB_RDONLY) {
2443 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2444 sb->s_flags &= ~SB_RDONLY;
2447 /* Needed for iput() to work correctly and not trash data */
2448 sb->s_flags |= SB_ACTIVE;
2451 * Turn on quotas which were not enabled for read-only mounts if
2452 * filesystem has quota feature, so that they are updated correctly.
2454 if (ext4_has_feature_quota(sb) && (s_flags & SB_RDONLY)) {
2455 int ret = ext4_enable_quotas(sb);
2460 ext4_msg(sb, KERN_ERR,
2461 "Cannot turn on quotas: error %d", ret);
2464 /* Turn on journaled quotas used for old sytle */
2465 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2466 if (EXT4_SB(sb)->s_qf_names[i]) {
2467 int ret = ext4_quota_on_mount(sb, i);
2472 ext4_msg(sb, KERN_ERR,
2473 "Cannot turn on journaled "
2474 "quota: type %d: error %d", i, ret);
2479 while (es->s_last_orphan) {
2480 struct inode *inode;
2483 * We may have encountered an error during cleanup; if
2484 * so, skip the rest.
2486 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2487 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2488 es->s_last_orphan = 0;
2492 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2493 if (IS_ERR(inode)) {
2494 es->s_last_orphan = 0;
2498 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2499 dquot_initialize(inode);
2500 if (inode->i_nlink) {
2501 if (test_opt(sb, DEBUG))
2502 ext4_msg(sb, KERN_DEBUG,
2503 "%s: truncating inode %lu to %lld bytes",
2504 __func__, inode->i_ino, inode->i_size);
2505 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2506 inode->i_ino, inode->i_size);
2508 truncate_inode_pages(inode->i_mapping, inode->i_size);
2509 ret = ext4_truncate(inode);
2511 ext4_std_error(inode->i_sb, ret);
2512 inode_unlock(inode);
2515 if (test_opt(sb, DEBUG))
2516 ext4_msg(sb, KERN_DEBUG,
2517 "%s: deleting unreferenced inode %lu",
2518 __func__, inode->i_ino);
2519 jbd_debug(2, "deleting unreferenced inode %lu\n",
2523 iput(inode); /* The delete magic happens here! */
2526 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2529 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2530 PLURAL(nr_orphans));
2532 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2533 PLURAL(nr_truncates));
2535 /* Turn off quotas if they were enabled for orphan cleanup */
2537 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2538 if (sb_dqopt(sb)->files[i])
2539 dquot_quota_off(sb, i);
2543 sb->s_flags = s_flags; /* Restore SB_RDONLY status */
2547 * Maximal extent format file size.
2548 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2549 * extent format containers, within a sector_t, and within i_blocks
2550 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2551 * so that won't be a limiting factor.
2553 * However there is other limiting factor. We do store extents in the form
2554 * of starting block and length, hence the resulting length of the extent
2555 * covering maximum file size must fit into on-disk format containers as
2556 * well. Given that length is always by 1 unit bigger than max unit (because
2557 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2559 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2561 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2564 loff_t upper_limit = MAX_LFS_FILESIZE;
2566 /* small i_blocks in vfs inode? */
2567 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2569 * CONFIG_LBDAF is not enabled implies the inode
2570 * i_block represent total blocks in 512 bytes
2571 * 32 == size of vfs inode i_blocks * 8
2573 upper_limit = (1LL << 32) - 1;
2575 /* total blocks in file system block size */
2576 upper_limit >>= (blkbits - 9);
2577 upper_limit <<= blkbits;
2581 * 32-bit extent-start container, ee_block. We lower the maxbytes
2582 * by one fs block, so ee_len can cover the extent of maximum file
2585 res = (1LL << 32) - 1;
2588 /* Sanity check against vm- & vfs- imposed limits */
2589 if (res > upper_limit)
2596 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2597 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2598 * We need to be 1 filesystem block less than the 2^48 sector limit.
2600 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2602 loff_t res = EXT4_NDIR_BLOCKS;
2605 /* This is calculated to be the largest file size for a dense, block
2606 * mapped file such that the file's total number of 512-byte sectors,
2607 * including data and all indirect blocks, does not exceed (2^48 - 1).
2609 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2610 * number of 512-byte sectors of the file.
2613 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2615 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2616 * the inode i_block field represents total file blocks in
2617 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2619 upper_limit = (1LL << 32) - 1;
2621 /* total blocks in file system block size */
2622 upper_limit >>= (bits - 9);
2626 * We use 48 bit ext4_inode i_blocks
2627 * With EXT4_HUGE_FILE_FL set the i_blocks
2628 * represent total number of blocks in
2629 * file system block size
2631 upper_limit = (1LL << 48) - 1;
2635 /* indirect blocks */
2637 /* double indirect blocks */
2638 meta_blocks += 1 + (1LL << (bits-2));
2639 /* tripple indirect blocks */
2640 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2642 upper_limit -= meta_blocks;
2643 upper_limit <<= bits;
2645 res += 1LL << (bits-2);
2646 res += 1LL << (2*(bits-2));
2647 res += 1LL << (3*(bits-2));
2649 if (res > upper_limit)
2652 if (res > MAX_LFS_FILESIZE)
2653 res = MAX_LFS_FILESIZE;
2658 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2659 ext4_fsblk_t logical_sb_block, int nr)
2661 struct ext4_sb_info *sbi = EXT4_SB(sb);
2662 ext4_group_t bg, first_meta_bg;
2665 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2667 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
2668 return logical_sb_block + nr + 1;
2669 bg = sbi->s_desc_per_block * nr;
2670 if (ext4_bg_has_super(sb, bg))
2674 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2675 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2676 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2679 if (sb->s_blocksize == 1024 && nr == 0 &&
2680 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2683 return (has_super + ext4_group_first_block_no(sb, bg));
2687 * ext4_get_stripe_size: Get the stripe size.
2688 * @sbi: In memory super block info
2690 * If we have specified it via mount option, then
2691 * use the mount option value. If the value specified at mount time is
2692 * greater than the blocks per group use the super block value.
2693 * If the super block value is greater than blocks per group return 0.
2694 * Allocator needs it be less than blocks per group.
2697 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2699 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2700 unsigned long stripe_width =
2701 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2704 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2705 ret = sbi->s_stripe;
2706 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
2708 else if (stride && stride <= sbi->s_blocks_per_group)
2714 * If the stripe width is 1, this makes no sense and
2715 * we set it to 0 to turn off stripe handling code.
2724 * Check whether this filesystem can be mounted based on
2725 * the features present and the RDONLY/RDWR mount requested.
2726 * Returns 1 if this filesystem can be mounted as requested,
2727 * 0 if it cannot be.
2729 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2731 if (ext4_has_unknown_ext4_incompat_features(sb)) {
2732 ext4_msg(sb, KERN_ERR,
2733 "Couldn't mount because of "
2734 "unsupported optional features (%x)",
2735 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2736 ~EXT4_FEATURE_INCOMPAT_SUPP));
2743 if (ext4_has_feature_readonly(sb)) {
2744 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2745 sb->s_flags |= SB_RDONLY;
2749 /* Check that feature set is OK for a read-write mount */
2750 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
2751 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2752 "unsupported optional features (%x)",
2753 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2754 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2758 * Large file size enabled file system can only be mounted
2759 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2761 if (ext4_has_feature_huge_file(sb)) {
2762 if (sizeof(blkcnt_t) < sizeof(u64)) {
2763 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2764 "cannot be mounted RDWR without "
2769 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
2770 ext4_msg(sb, KERN_ERR,
2771 "Can't support bigalloc feature without "
2772 "extents feature\n");
2776 #ifndef CONFIG_QUOTA
2777 if (ext4_has_feature_quota(sb) && !readonly) {
2778 ext4_msg(sb, KERN_ERR,
2779 "Filesystem with quota feature cannot be mounted RDWR "
2780 "without CONFIG_QUOTA");
2783 if (ext4_has_feature_project(sb) && !readonly) {
2784 ext4_msg(sb, KERN_ERR,
2785 "Filesystem with project quota feature cannot be mounted RDWR "
2786 "without CONFIG_QUOTA");
2789 #endif /* CONFIG_QUOTA */
2794 * This function is called once a day if we have errors logged
2795 * on the file system
2797 static void print_daily_error_info(struct timer_list *t)
2799 struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report);
2800 struct super_block *sb = sbi->s_sb;
2801 struct ext4_super_block *es = sbi->s_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 || sb_rdonly(sb) ||
3139 !test_opt(sb, INIT_INODE_TABLE))
3142 elr = ext4_li_request_new(sb, first_not_zeroed);
3148 if (NULL == ext4_li_info) {
3149 ret = ext4_li_info_new();
3154 mutex_lock(&ext4_li_info->li_list_mtx);
3155 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3156 mutex_unlock(&ext4_li_info->li_list_mtx);
3158 sbi->s_li_request = elr;
3160 * set elr to NULL here since it has been inserted to
3161 * the request_list and the removal and free of it is
3162 * handled by ext4_clear_request_list from now on.
3166 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3167 ret = ext4_run_lazyinit_thread();
3172 mutex_unlock(&ext4_li_mtx);
3179 * We do not need to lock anything since this is called on
3182 static void ext4_destroy_lazyinit_thread(void)
3185 * If thread exited earlier
3186 * there's nothing to be done.
3188 if (!ext4_li_info || !ext4_lazyinit_task)
3191 kthread_stop(ext4_lazyinit_task);
3194 static int set_journal_csum_feature_set(struct super_block *sb)
3197 int compat, incompat;
3198 struct ext4_sb_info *sbi = EXT4_SB(sb);
3200 if (ext4_has_metadata_csum(sb)) {
3201 /* journal checksum v3 */
3203 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
3205 /* journal checksum v1 */
3206 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3210 jbd2_journal_clear_features(sbi->s_journal,
3211 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3212 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3213 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3214 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3215 ret = jbd2_journal_set_features(sbi->s_journal,
3217 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3219 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3220 ret = jbd2_journal_set_features(sbi->s_journal,
3223 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3224 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3226 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3227 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3234 * Note: calculating the overhead so we can be compatible with
3235 * historical BSD practice is quite difficult in the face of
3236 * clusters/bigalloc. This is because multiple metadata blocks from
3237 * different block group can end up in the same allocation cluster.
3238 * Calculating the exact overhead in the face of clustered allocation
3239 * requires either O(all block bitmaps) in memory or O(number of block
3240 * groups**2) in time. We will still calculate the superblock for
3241 * older file systems --- and if we come across with a bigalloc file
3242 * system with zero in s_overhead_clusters the estimate will be close to
3243 * correct especially for very large cluster sizes --- but for newer
3244 * file systems, it's better to calculate this figure once at mkfs
3245 * time, and store it in the superblock. If the superblock value is
3246 * present (even for non-bigalloc file systems), we will use it.
3248 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3251 struct ext4_sb_info *sbi = EXT4_SB(sb);
3252 struct ext4_group_desc *gdp;
3253 ext4_fsblk_t first_block, last_block, b;
3254 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3255 int s, j, count = 0;
3257 if (!ext4_has_feature_bigalloc(sb))
3258 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3259 sbi->s_itb_per_group + 2);
3261 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3262 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3263 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3264 for (i = 0; i < ngroups; i++) {
3265 gdp = ext4_get_group_desc(sb, i, NULL);
3266 b = ext4_block_bitmap(sb, gdp);
3267 if (b >= first_block && b <= last_block) {
3268 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3271 b = ext4_inode_bitmap(sb, gdp);
3272 if (b >= first_block && b <= last_block) {
3273 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3276 b = ext4_inode_table(sb, gdp);
3277 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3278 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3279 int c = EXT4_B2C(sbi, b - first_block);
3280 ext4_set_bit(c, buf);
3286 if (ext4_bg_has_super(sb, grp)) {
3287 ext4_set_bit(s++, buf);
3290 j = ext4_bg_num_gdb(sb, grp);
3291 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
3292 ext4_error(sb, "Invalid number of block group "
3293 "descriptor blocks: %d", j);
3294 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
3298 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3302 return EXT4_CLUSTERS_PER_GROUP(sb) -
3303 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3307 * Compute the overhead and stash it in sbi->s_overhead
3309 int ext4_calculate_overhead(struct super_block *sb)
3311 struct ext4_sb_info *sbi = EXT4_SB(sb);
3312 struct ext4_super_block *es = sbi->s_es;
3313 struct inode *j_inode;
3314 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
3315 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3316 ext4_fsblk_t overhead = 0;
3317 char *buf = (char *) get_zeroed_page(GFP_NOFS);
3323 * Compute the overhead (FS structures). This is constant
3324 * for a given filesystem unless the number of block groups
3325 * changes so we cache the previous value until it does.
3329 * All of the blocks before first_data_block are overhead
3331 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3334 * Add the overhead found in each block group
3336 for (i = 0; i < ngroups; i++) {
3339 blks = count_overhead(sb, i, buf);
3342 memset(buf, 0, PAGE_SIZE);
3347 * Add the internal journal blocks whether the journal has been
3350 if (sbi->s_journal && !sbi->journal_bdev)
3351 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3352 else if (ext4_has_feature_journal(sb) && !sbi->s_journal) {
3353 j_inode = ext4_get_journal_inode(sb, j_inum);
3355 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
3356 overhead += EXT4_NUM_B2C(sbi, j_blocks);
3359 ext4_msg(sb, KERN_ERR, "can't get journal size");
3362 sbi->s_overhead = overhead;
3364 free_page((unsigned long) buf);
3368 static void ext4_set_resv_clusters(struct super_block *sb)
3370 ext4_fsblk_t resv_clusters;
3371 struct ext4_sb_info *sbi = EXT4_SB(sb);
3374 * There's no need to reserve anything when we aren't using extents.
3375 * The space estimates are exact, there are no unwritten extents,
3376 * hole punching doesn't need new metadata... This is needed especially
3377 * to keep ext2/3 backward compatibility.
3379 if (!ext4_has_feature_extents(sb))
3382 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3383 * This should cover the situations where we can not afford to run
3384 * out of space like for example punch hole, or converting
3385 * unwritten extents in delalloc path. In most cases such
3386 * allocation would require 1, or 2 blocks, higher numbers are
3389 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3390 sbi->s_cluster_bits);
3392 do_div(resv_clusters, 50);
3393 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3395 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
3398 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3400 struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev);
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_daxdev = dax_dev;
3427 sbi->s_blockgroup_lock =
3428 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3429 if (!sbi->s_blockgroup_lock)
3432 sb->s_fs_info = sbi;
3434 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3435 sbi->s_sb_block = sb_block;
3436 if (sb->s_bdev->bd_part)
3437 sbi->s_sectors_written_start =
3438 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3440 /* Cleanup superblock name */
3441 strreplace(sb->s_id, '/', '!');
3443 /* -EINVAL is default */
3445 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3447 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3452 * The ext4 superblock will not be buffer aligned for other than 1kB
3453 * block sizes. We need to calculate the offset from buffer start.
3455 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3456 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3457 offset = do_div(logical_sb_block, blocksize);
3459 logical_sb_block = sb_block;
3462 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
3463 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3467 * Note: s_es must be initialized as soon as possible because
3468 * some ext4 macro-instructions depend on its value
3470 es = (struct ext4_super_block *) (bh->b_data + offset);
3472 sb->s_magic = le16_to_cpu(es->s_magic);
3473 if (sb->s_magic != EXT4_SUPER_MAGIC)
3475 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3477 /* Warn if metadata_csum and gdt_csum are both set. */
3478 if (ext4_has_feature_metadata_csum(sb) &&
3479 ext4_has_feature_gdt_csum(sb))
3480 ext4_warning(sb, "metadata_csum and uninit_bg are "
3481 "redundant flags; please run fsck.");
3483 /* Check for a known checksum algorithm */
3484 if (!ext4_verify_csum_type(sb, es)) {
3485 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3486 "unknown checksum algorithm.");
3491 /* Load the checksum driver */
3492 if (ext4_has_feature_metadata_csum(sb) ||
3493 ext4_has_feature_ea_inode(sb)) {
3494 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3495 if (IS_ERR(sbi->s_chksum_driver)) {
3496 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3497 ret = PTR_ERR(sbi->s_chksum_driver);
3498 sbi->s_chksum_driver = NULL;
3503 /* Check superblock checksum */
3504 if (!ext4_superblock_csum_verify(sb, es)) {
3505 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3506 "invalid superblock checksum. Run e2fsck?");
3512 /* Precompute checksum seed for all metadata */
3513 if (ext4_has_feature_csum_seed(sb))
3514 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
3515 else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb))
3516 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3517 sizeof(es->s_uuid));
3519 /* Set defaults before we parse the mount options */
3520 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3521 set_opt(sb, INIT_INODE_TABLE);
3522 if (def_mount_opts & EXT4_DEFM_DEBUG)
3524 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3526 if (def_mount_opts & EXT4_DEFM_UID16)
3527 set_opt(sb, NO_UID32);
3528 /* xattr user namespace & acls are now defaulted on */
3529 set_opt(sb, XATTR_USER);
3530 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3531 set_opt(sb, POSIX_ACL);
3533 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3534 if (ext4_has_metadata_csum(sb))
3535 set_opt(sb, JOURNAL_CHECKSUM);
3537 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3538 set_opt(sb, JOURNAL_DATA);
3539 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3540 set_opt(sb, ORDERED_DATA);
3541 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3542 set_opt(sb, WRITEBACK_DATA);
3544 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3545 set_opt(sb, ERRORS_PANIC);
3546 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3547 set_opt(sb, ERRORS_CONT);
3549 set_opt(sb, ERRORS_RO);
3550 /* block_validity enabled by default; disable with noblock_validity */
3551 set_opt(sb, BLOCK_VALIDITY);
3552 if (def_mount_opts & EXT4_DEFM_DISCARD)
3553 set_opt(sb, DISCARD);
3555 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3556 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3557 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3558 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3559 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3561 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3562 set_opt(sb, BARRIER);
3565 * enable delayed allocation by default
3566 * Use -o nodelalloc to turn it off
3568 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3569 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3570 set_opt(sb, DELALLOC);
3573 * set default s_li_wait_mult for lazyinit, for the case there is
3574 * no mount option specified.
3576 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3578 if (sbi->s_es->s_mount_opts[0]) {
3579 char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
3580 sizeof(sbi->s_es->s_mount_opts),
3584 if (!parse_options(s_mount_opts, sb, &journal_devnum,
3585 &journal_ioprio, 0)) {
3586 ext4_msg(sb, KERN_WARNING,
3587 "failed to parse options in superblock: %s",
3590 kfree(s_mount_opts);
3592 sbi->s_def_mount_opt = sbi->s_mount_opt;
3593 if (!parse_options((char *) data, sb, &journal_devnum,
3594 &journal_ioprio, 0))
3597 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3598 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3599 "with data=journal disables delayed "
3600 "allocation and O_DIRECT support!\n");
3601 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3602 ext4_msg(sb, KERN_ERR, "can't mount with "
3603 "both data=journal and delalloc");
3606 if (test_opt(sb, DIOREAD_NOLOCK)) {
3607 ext4_msg(sb, KERN_ERR, "can't mount with "
3608 "both data=journal and dioread_nolock");
3611 if (test_opt(sb, DAX)) {
3612 ext4_msg(sb, KERN_ERR, "can't mount with "
3613 "both data=journal and dax");
3616 if (ext4_has_feature_encrypt(sb)) {
3617 ext4_msg(sb, KERN_WARNING,
3618 "encrypted files will use data=ordered "
3619 "instead of data journaling mode");
3621 if (test_opt(sb, DELALLOC))
3622 clear_opt(sb, DELALLOC);
3624 sb->s_iflags |= SB_I_CGROUPWB;
3627 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
3628 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
3630 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3631 (ext4_has_compat_features(sb) ||
3632 ext4_has_ro_compat_features(sb) ||
3633 ext4_has_incompat_features(sb)))
3634 ext4_msg(sb, KERN_WARNING,
3635 "feature flags set on rev 0 fs, "
3636 "running e2fsck is recommended");
3638 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3639 set_opt2(sb, HURD_COMPAT);
3640 if (ext4_has_feature_64bit(sb)) {
3641 ext4_msg(sb, KERN_ERR,
3642 "The Hurd can't support 64-bit file systems");
3647 * ea_inode feature uses l_i_version field which is not
3648 * available in HURD_COMPAT mode.
3650 if (ext4_has_feature_ea_inode(sb)) {
3651 ext4_msg(sb, KERN_ERR,
3652 "ea_inode feature is not supported for Hurd");
3657 if (IS_EXT2_SB(sb)) {
3658 if (ext2_feature_set_ok(sb))
3659 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3660 "using the ext4 subsystem");
3662 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3663 "to feature incompatibilities");
3668 if (IS_EXT3_SB(sb)) {
3669 if (ext3_feature_set_ok(sb))
3670 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3671 "using the ext4 subsystem");
3673 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3674 "to feature incompatibilities");
3680 * Check feature flags regardless of the revision level, since we
3681 * previously didn't change the revision level when setting the flags,
3682 * so there is a chance incompat flags are set on a rev 0 filesystem.
3684 if (!ext4_feature_set_ok(sb, (sb_rdonly(sb))))
3687 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3688 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3689 blocksize > EXT4_MAX_BLOCK_SIZE) {
3690 ext4_msg(sb, KERN_ERR,
3691 "Unsupported filesystem blocksize %d (%d log_block_size)",
3692 blocksize, le32_to_cpu(es->s_log_block_size));
3695 if (le32_to_cpu(es->s_log_block_size) >
3696 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3697 ext4_msg(sb, KERN_ERR,
3698 "Invalid log block size: %u",
3699 le32_to_cpu(es->s_log_block_size));
3703 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
3704 ext4_msg(sb, KERN_ERR,
3705 "Number of reserved GDT blocks insanely large: %d",
3706 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
3710 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
3711 if (ext4_has_feature_inline_data(sb)) {
3712 ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem"
3713 " that may contain inline data");
3716 err = bdev_dax_supported(sb, blocksize);
3721 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
3722 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
3723 es->s_encryption_level);
3727 if (sb->s_blocksize != blocksize) {
3728 /* Validate the filesystem blocksize */
3729 if (!sb_set_blocksize(sb, blocksize)) {
3730 ext4_msg(sb, KERN_ERR, "bad block size %d",
3736 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3737 offset = do_div(logical_sb_block, blocksize);
3738 bh = sb_bread_unmovable(sb, logical_sb_block);
3740 ext4_msg(sb, KERN_ERR,
3741 "Can't read superblock on 2nd try");
3744 es = (struct ext4_super_block *)(bh->b_data + offset);
3746 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3747 ext4_msg(sb, KERN_ERR,
3748 "Magic mismatch, very weird!");
3753 has_huge_files = ext4_has_feature_huge_file(sb);
3754 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3756 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3758 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3759 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3760 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3762 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3763 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3764 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3765 (!is_power_of_2(sbi->s_inode_size)) ||
3766 (sbi->s_inode_size > blocksize)) {
3767 ext4_msg(sb, KERN_ERR,
3768 "unsupported inode size: %d",
3772 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3773 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3776 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3777 if (ext4_has_feature_64bit(sb)) {
3778 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3779 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3780 !is_power_of_2(sbi->s_desc_size)) {
3781 ext4_msg(sb, KERN_ERR,
3782 "unsupported descriptor size %lu",
3787 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3789 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3790 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3792 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3793 if (sbi->s_inodes_per_block == 0)
3795 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
3796 sbi->s_inodes_per_group > blocksize * 8) {
3797 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
3798 sbi->s_blocks_per_group);
3801 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3802 sbi->s_inodes_per_block;
3803 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3805 sbi->s_mount_state = le16_to_cpu(es->s_state);
3806 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3807 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3809 for (i = 0; i < 4; i++)
3810 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3811 sbi->s_def_hash_version = es->s_def_hash_version;
3812 if (ext4_has_feature_dir_index(sb)) {
3813 i = le32_to_cpu(es->s_flags);
3814 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3815 sbi->s_hash_unsigned = 3;
3816 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3817 #ifdef __CHAR_UNSIGNED__
3820 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3821 sbi->s_hash_unsigned = 3;
3825 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3830 /* Handle clustersize */
3831 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3832 has_bigalloc = ext4_has_feature_bigalloc(sb);
3834 if (clustersize < blocksize) {
3835 ext4_msg(sb, KERN_ERR,
3836 "cluster size (%d) smaller than "
3837 "block size (%d)", clustersize, blocksize);
3840 if (le32_to_cpu(es->s_log_cluster_size) >
3841 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3842 ext4_msg(sb, KERN_ERR,
3843 "Invalid log cluster size: %u",
3844 le32_to_cpu(es->s_log_cluster_size));
3847 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3848 le32_to_cpu(es->s_log_block_size);
3849 sbi->s_clusters_per_group =
3850 le32_to_cpu(es->s_clusters_per_group);
3851 if (sbi->s_clusters_per_group > blocksize * 8) {
3852 ext4_msg(sb, KERN_ERR,
3853 "#clusters per group too big: %lu",
3854 sbi->s_clusters_per_group);
3857 if (sbi->s_blocks_per_group !=
3858 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3859 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3860 "clusters per group (%lu) inconsistent",
3861 sbi->s_blocks_per_group,
3862 sbi->s_clusters_per_group);
3866 if (clustersize != blocksize) {
3867 ext4_warning(sb, "fragment/cluster size (%d) != "
3868 "block size (%d)", clustersize,
3870 clustersize = blocksize;
3872 if (sbi->s_blocks_per_group > blocksize * 8) {
3873 ext4_msg(sb, KERN_ERR,
3874 "#blocks per group too big: %lu",
3875 sbi->s_blocks_per_group);
3878 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3879 sbi->s_cluster_bits = 0;
3881 sbi->s_cluster_ratio = clustersize / blocksize;
3883 /* Do we have standard group size of clustersize * 8 blocks ? */
3884 if (sbi->s_blocks_per_group == clustersize << 3)
3885 set_opt2(sb, STD_GROUP_SIZE);
3888 * Test whether we have more sectors than will fit in sector_t,
3889 * and whether the max offset is addressable by the page cache.
3891 err = generic_check_addressable(sb->s_blocksize_bits,
3892 ext4_blocks_count(es));
3894 ext4_msg(sb, KERN_ERR, "filesystem"
3895 " too large to mount safely on this system");
3896 if (sizeof(sector_t) < 8)
3897 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3901 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3904 /* check blocks count against device size */
3905 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3906 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3907 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3908 "exceeds size of device (%llu blocks)",
3909 ext4_blocks_count(es), blocks_count);
3914 * It makes no sense for the first data block to be beyond the end
3915 * of the filesystem.
3917 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3918 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3919 "block %u is beyond end of filesystem (%llu)",
3920 le32_to_cpu(es->s_first_data_block),
3921 ext4_blocks_count(es));
3924 blocks_count = (ext4_blocks_count(es) -
3925 le32_to_cpu(es->s_first_data_block) +
3926 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3927 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3928 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3929 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3930 "(block count %llu, first data block %u, "
3931 "blocks per group %lu)", sbi->s_groups_count,
3932 ext4_blocks_count(es),
3933 le32_to_cpu(es->s_first_data_block),
3934 EXT4_BLOCKS_PER_GROUP(sb));
3937 sbi->s_groups_count = blocks_count;
3938 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3939 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3940 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3941 EXT4_DESC_PER_BLOCK(sb);
3942 if (ext4_has_feature_meta_bg(sb)) {
3943 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
3944 ext4_msg(sb, KERN_WARNING,
3945 "first meta block group too large: %u "
3946 "(group descriptor block count %u)",
3947 le32_to_cpu(es->s_first_meta_bg), db_count);
3951 sbi->s_group_desc = kvmalloc(db_count *
3952 sizeof(struct buffer_head *),
3954 if (sbi->s_group_desc == NULL) {
3955 ext4_msg(sb, KERN_ERR, "not enough memory");
3960 bgl_lock_init(sbi->s_blockgroup_lock);
3962 /* Pre-read the descriptors into the buffer cache */
3963 for (i = 0; i < db_count; i++) {
3964 block = descriptor_loc(sb, logical_sb_block, i);
3965 sb_breadahead(sb, block);
3968 for (i = 0; i < db_count; i++) {
3969 block = descriptor_loc(sb, logical_sb_block, i);
3970 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
3971 if (!sbi->s_group_desc[i]) {
3972 ext4_msg(sb, KERN_ERR,
3973 "can't read group descriptor %d", i);
3978 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
3979 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3980 ret = -EFSCORRUPTED;
3984 sbi->s_gdb_count = db_count;
3986 timer_setup(&sbi->s_err_report, print_daily_error_info, 0);
3988 /* Register extent status tree shrinker */
3989 if (ext4_es_register_shrinker(sbi))
3992 sbi->s_stripe = ext4_get_stripe_size(sbi);
3993 sbi->s_extent_max_zeroout_kb = 32;
3996 * set up enough so that it can read an inode
3998 sb->s_op = &ext4_sops;
3999 sb->s_export_op = &ext4_export_ops;
4000 sb->s_xattr = ext4_xattr_handlers;
4001 #ifdef CONFIG_EXT4_FS_ENCRYPTION
4002 sb->s_cop = &ext4_cryptops;
4005 sb->dq_op = &ext4_quota_operations;
4006 if (ext4_has_feature_quota(sb))
4007 sb->s_qcop = &dquot_quotactl_sysfile_ops;
4009 sb->s_qcop = &ext4_qctl_operations;
4010 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
4012 memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
4014 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
4015 mutex_init(&sbi->s_orphan_lock);
4019 needs_recovery = (es->s_last_orphan != 0 ||
4020 ext4_has_feature_journal_needs_recovery(sb));
4022 if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb))
4023 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
4024 goto failed_mount3a;
4027 * The first inode we look at is the journal inode. Don't try
4028 * root first: it may be modified in the journal!
4030 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
4031 err = ext4_load_journal(sb, es, journal_devnum);
4033 goto failed_mount3a;
4034 } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) &&
4035 ext4_has_feature_journal_needs_recovery(sb)) {
4036 ext4_msg(sb, KERN_ERR, "required journal recovery "
4037 "suppressed and not mounted read-only");
4038 goto failed_mount_wq;
4040 /* Nojournal mode, all journal mount options are illegal */
4041 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
4042 ext4_msg(sb, KERN_ERR, "can't mount with "
4043 "journal_checksum, fs mounted w/o journal");
4044 goto failed_mount_wq;
4046 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4047 ext4_msg(sb, KERN_ERR, "can't mount with "
4048 "journal_async_commit, fs mounted w/o journal");
4049 goto failed_mount_wq;
4051 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
4052 ext4_msg(sb, KERN_ERR, "can't mount with "
4053 "commit=%lu, fs mounted w/o journal",
4054 sbi->s_commit_interval / HZ);
4055 goto failed_mount_wq;
4057 if (EXT4_MOUNT_DATA_FLAGS &
4058 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
4059 ext4_msg(sb, KERN_ERR, "can't mount with "
4060 "data=, fs mounted w/o journal");
4061 goto failed_mount_wq;
4063 sbi->s_def_mount_opt &= EXT4_MOUNT_JOURNAL_CHECKSUM;
4064 clear_opt(sb, JOURNAL_CHECKSUM);
4065 clear_opt(sb, DATA_FLAGS);
4066 sbi->s_journal = NULL;
4071 if (ext4_has_feature_64bit(sb) &&
4072 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4073 JBD2_FEATURE_INCOMPAT_64BIT)) {
4074 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
4075 goto failed_mount_wq;
4078 if (!set_journal_csum_feature_set(sb)) {
4079 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4081 goto failed_mount_wq;
4084 /* We have now updated the journal if required, so we can
4085 * validate the data journaling mode. */
4086 switch (test_opt(sb, DATA_FLAGS)) {
4088 /* No mode set, assume a default based on the journal
4089 * capabilities: ORDERED_DATA if the journal can
4090 * cope, else JOURNAL_DATA
4092 if (jbd2_journal_check_available_features
4093 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
4094 set_opt(sb, ORDERED_DATA);
4096 set_opt(sb, JOURNAL_DATA);
4099 case EXT4_MOUNT_ORDERED_DATA:
4100 case EXT4_MOUNT_WRITEBACK_DATA:
4101 if (!jbd2_journal_check_available_features
4102 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4103 ext4_msg(sb, KERN_ERR, "Journal does not support "
4104 "requested data journaling mode");
4105 goto failed_mount_wq;
4111 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
4112 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4113 ext4_msg(sb, KERN_ERR, "can't mount with "
4114 "journal_async_commit in data=ordered mode");
4115 goto failed_mount_wq;
4118 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4120 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4123 if (!test_opt(sb, NO_MBCACHE)) {
4124 sbi->s_ea_block_cache = ext4_xattr_create_cache();
4125 if (!sbi->s_ea_block_cache) {
4126 ext4_msg(sb, KERN_ERR,
4127 "Failed to create ea_block_cache");
4128 goto failed_mount_wq;
4131 if (ext4_has_feature_ea_inode(sb)) {
4132 sbi->s_ea_inode_cache = ext4_xattr_create_cache();
4133 if (!sbi->s_ea_inode_cache) {
4134 ext4_msg(sb, KERN_ERR,
4135 "Failed to create ea_inode_cache");
4136 goto failed_mount_wq;
4141 if ((DUMMY_ENCRYPTION_ENABLED(sbi) || ext4_has_feature_encrypt(sb)) &&
4142 (blocksize != PAGE_SIZE)) {
4143 ext4_msg(sb, KERN_ERR,
4144 "Unsupported blocksize for fs encryption");
4145 goto failed_mount_wq;
4148 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !sb_rdonly(sb) &&
4149 !ext4_has_feature_encrypt(sb)) {
4150 ext4_set_feature_encrypt(sb);
4151 ext4_commit_super(sb, 1);
4155 * Get the # of file system overhead blocks from the
4156 * superblock if present.
4158 if (es->s_overhead_clusters)
4159 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4161 err = ext4_calculate_overhead(sb);
4163 goto failed_mount_wq;
4167 * The maximum number of concurrent works can be high and
4168 * concurrency isn't really necessary. Limit it to 1.
4170 EXT4_SB(sb)->rsv_conversion_wq =
4171 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4172 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4173 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
4179 * The jbd2_journal_load will have done any necessary log recovery,
4180 * so we can safely mount the rest of the filesystem now.
4183 root = ext4_iget(sb, EXT4_ROOT_INO);
4185 ext4_msg(sb, KERN_ERR, "get root inode failed");
4186 ret = PTR_ERR(root);
4190 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4191 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4195 sb->s_root = d_make_root(root);
4197 ext4_msg(sb, KERN_ERR, "get root dentry failed");
4202 if (ext4_setup_super(sb, es, sb_rdonly(sb)))
4203 sb->s_flags |= SB_RDONLY;
4205 /* determine the minimum size of new large inodes, if present */
4206 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE &&
4207 sbi->s_want_extra_isize == 0) {
4208 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4209 EXT4_GOOD_OLD_INODE_SIZE;
4210 if (ext4_has_feature_extra_isize(sb)) {
4211 if (sbi->s_want_extra_isize <
4212 le16_to_cpu(es->s_want_extra_isize))
4213 sbi->s_want_extra_isize =
4214 le16_to_cpu(es->s_want_extra_isize);
4215 if (sbi->s_want_extra_isize <
4216 le16_to_cpu(es->s_min_extra_isize))
4217 sbi->s_want_extra_isize =
4218 le16_to_cpu(es->s_min_extra_isize);
4221 /* Check if enough inode space is available */
4222 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4223 sbi->s_inode_size) {
4224 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4225 EXT4_GOOD_OLD_INODE_SIZE;
4226 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4230 ext4_set_resv_clusters(sb);
4232 err = ext4_setup_system_zone(sb);
4234 ext4_msg(sb, KERN_ERR, "failed to initialize system "
4236 goto failed_mount4a;
4240 err = ext4_mb_init(sb);
4242 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4247 block = ext4_count_free_clusters(sb);
4248 ext4_free_blocks_count_set(sbi->s_es,
4249 EXT4_C2B(sbi, block));
4250 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4253 unsigned long freei = ext4_count_free_inodes(sb);
4254 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
4255 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4259 err = percpu_counter_init(&sbi->s_dirs_counter,
4260 ext4_count_dirs(sb), GFP_KERNEL);
4262 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4265 err = percpu_init_rwsem(&sbi->s_journal_flag_rwsem);
4268 ext4_msg(sb, KERN_ERR, "insufficient memory");
4272 if (ext4_has_feature_flex_bg(sb))
4273 if (!ext4_fill_flex_info(sb)) {
4274 ext4_msg(sb, KERN_ERR,
4275 "unable to initialize "
4276 "flex_bg meta info!");
4280 err = ext4_register_li_request(sb, first_not_zeroed);
4284 err = ext4_register_sysfs(sb);
4289 /* Enable quota usage during mount. */
4290 if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) {
4291 err = ext4_enable_quotas(sb);
4295 #endif /* CONFIG_QUOTA */
4297 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4298 ext4_orphan_cleanup(sb, es);
4299 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4300 if (needs_recovery) {
4301 ext4_msg(sb, KERN_INFO, "recovery complete");
4302 ext4_mark_recovery_complete(sb, es);
4304 if (EXT4_SB(sb)->s_journal) {
4305 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4306 descr = " journalled data mode";
4307 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4308 descr = " ordered data mode";
4310 descr = " writeback data mode";
4312 descr = "out journal";
4314 if (test_opt(sb, DISCARD)) {
4315 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4316 if (!blk_queue_discard(q))
4317 ext4_msg(sb, KERN_WARNING,
4318 "mounting with \"discard\" option, but "
4319 "the device does not support discard");
4322 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
4323 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4324 "Opts: %.*s%s%s", descr,
4325 (int) sizeof(sbi->s_es->s_mount_opts),
4326 sbi->s_es->s_mount_opts,
4327 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4329 if (es->s_error_count)
4330 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4332 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4333 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4334 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4335 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4342 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4347 ext4_unregister_sysfs(sb);
4350 ext4_unregister_li_request(sb);
4352 ext4_mb_release(sb);
4353 if (sbi->s_flex_groups)
4354 kvfree(sbi->s_flex_groups);
4355 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4356 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4357 percpu_counter_destroy(&sbi->s_dirs_counter);
4358 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4360 ext4_ext_release(sb);
4361 ext4_release_system_zone(sb);
4366 ext4_msg(sb, KERN_ERR, "mount failed");
4367 if (EXT4_SB(sb)->rsv_conversion_wq)
4368 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4370 if (sbi->s_ea_inode_cache) {
4371 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
4372 sbi->s_ea_inode_cache = NULL;
4374 if (sbi->s_ea_block_cache) {
4375 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
4376 sbi->s_ea_block_cache = NULL;
4378 if (sbi->s_journal) {
4379 jbd2_journal_destroy(sbi->s_journal);
4380 sbi->s_journal = NULL;
4383 ext4_es_unregister_shrinker(sbi);
4385 del_timer_sync(&sbi->s_err_report);
4387 kthread_stop(sbi->s_mmp_tsk);
4389 for (i = 0; i < db_count; i++)
4390 brelse(sbi->s_group_desc[i]);
4391 kvfree(sbi->s_group_desc);
4393 if (sbi->s_chksum_driver)
4394 crypto_free_shash(sbi->s_chksum_driver);
4396 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4397 kfree(sbi->s_qf_names[i]);
4399 ext4_blkdev_remove(sbi);
4402 sb->s_fs_info = NULL;
4403 kfree(sbi->s_blockgroup_lock);
4407 fs_put_dax(dax_dev);
4408 return err ? err : ret;
4412 * Setup any per-fs journal parameters now. We'll do this both on
4413 * initial mount, once the journal has been initialised but before we've
4414 * done any recovery; and again on any subsequent remount.
4416 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4418 struct ext4_sb_info *sbi = EXT4_SB(sb);
4420 journal->j_commit_interval = sbi->s_commit_interval;
4421 journal->j_min_batch_time = sbi->s_min_batch_time;
4422 journal->j_max_batch_time = sbi->s_max_batch_time;
4424 write_lock(&journal->j_state_lock);
4425 if (test_opt(sb, BARRIER))
4426 journal->j_flags |= JBD2_BARRIER;
4428 journal->j_flags &= ~JBD2_BARRIER;
4429 if (test_opt(sb, DATA_ERR_ABORT))
4430 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4432 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4433 write_unlock(&journal->j_state_lock);
4436 static struct inode *ext4_get_journal_inode(struct super_block *sb,
4437 unsigned int journal_inum)
4439 struct inode *journal_inode;
4442 * Test for the existence of a valid inode on disk. Bad things
4443 * happen if we iget() an unused inode, as the subsequent iput()
4444 * will try to delete it.
4446 journal_inode = ext4_iget(sb, journal_inum);
4447 if (IS_ERR(journal_inode)) {
4448 ext4_msg(sb, KERN_ERR, "no journal found");
4451 if (!journal_inode->i_nlink) {
4452 make_bad_inode(journal_inode);
4453 iput(journal_inode);
4454 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4458 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4459 journal_inode, journal_inode->i_size);
4460 if (!S_ISREG(journal_inode->i_mode)) {
4461 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4462 iput(journal_inode);
4465 return journal_inode;
4468 static journal_t *ext4_get_journal(struct super_block *sb,
4469 unsigned int journal_inum)
4471 struct inode *journal_inode;
4474 BUG_ON(!ext4_has_feature_journal(sb));
4476 journal_inode = ext4_get_journal_inode(sb, journal_inum);
4480 journal = jbd2_journal_init_inode(journal_inode);
4482 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4483 iput(journal_inode);
4486 journal->j_private = sb;
4487 ext4_init_journal_params(sb, journal);
4491 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4494 struct buffer_head *bh;
4498 int hblock, blocksize;
4499 ext4_fsblk_t sb_block;
4500 unsigned long offset;
4501 struct ext4_super_block *es;
4502 struct block_device *bdev;
4504 BUG_ON(!ext4_has_feature_journal(sb));
4506 bdev = ext4_blkdev_get(j_dev, sb);
4510 blocksize = sb->s_blocksize;
4511 hblock = bdev_logical_block_size(bdev);
4512 if (blocksize < hblock) {
4513 ext4_msg(sb, KERN_ERR,
4514 "blocksize too small for journal device");
4518 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4519 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4520 set_blocksize(bdev, blocksize);
4521 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4522 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4523 "external journal");
4527 es = (struct ext4_super_block *) (bh->b_data + offset);
4528 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4529 !(le32_to_cpu(es->s_feature_incompat) &
4530 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4531 ext4_msg(sb, KERN_ERR, "external journal has "
4537 if ((le32_to_cpu(es->s_feature_ro_compat) &
4538 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4539 es->s_checksum != ext4_superblock_csum(sb, es)) {
4540 ext4_msg(sb, KERN_ERR, "external journal has "
4541 "corrupt superblock");
4546 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4547 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4552 len = ext4_blocks_count(es);
4553 start = sb_block + 1;
4554 brelse(bh); /* we're done with the superblock */
4556 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4557 start, len, blocksize);
4559 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4562 journal->j_private = sb;
4563 ll_rw_block(REQ_OP_READ, REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4564 wait_on_buffer(journal->j_sb_buffer);
4565 if (!buffer_uptodate(journal->j_sb_buffer)) {
4566 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4569 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4570 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4571 "user (unsupported) - %d",
4572 be32_to_cpu(journal->j_superblock->s_nr_users));
4575 EXT4_SB(sb)->journal_bdev = bdev;
4576 ext4_init_journal_params(sb, journal);
4580 jbd2_journal_destroy(journal);
4582 ext4_blkdev_put(bdev);
4586 static int ext4_load_journal(struct super_block *sb,
4587 struct ext4_super_block *es,
4588 unsigned long journal_devnum)
4591 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4594 int really_read_only;
4596 BUG_ON(!ext4_has_feature_journal(sb));
4598 if (journal_devnum &&
4599 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4600 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4601 "numbers have changed");
4602 journal_dev = new_decode_dev(journal_devnum);
4604 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4606 really_read_only = bdev_read_only(sb->s_bdev);
4609 * Are we loading a blank journal or performing recovery after a
4610 * crash? For recovery, we need to check in advance whether we
4611 * can get read-write access to the device.
4613 if (ext4_has_feature_journal_needs_recovery(sb)) {
4614 if (sb_rdonly(sb)) {
4615 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4616 "required on readonly filesystem");
4617 if (really_read_only) {
4618 ext4_msg(sb, KERN_ERR, "write access "
4619 "unavailable, cannot proceed "
4620 "(try mounting with noload)");
4623 ext4_msg(sb, KERN_INFO, "write access will "
4624 "be enabled during recovery");
4628 if (journal_inum && journal_dev) {
4629 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4630 "and inode journals!");
4635 if (!(journal = ext4_get_journal(sb, journal_inum)))
4638 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4642 if (!(journal->j_flags & JBD2_BARRIER))
4643 ext4_msg(sb, KERN_INFO, "barriers disabled");
4645 if (!ext4_has_feature_journal_needs_recovery(sb))
4646 err = jbd2_journal_wipe(journal, !really_read_only);
4648 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4650 memcpy(save, ((char *) es) +
4651 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4652 err = jbd2_journal_load(journal);
4654 memcpy(((char *) es) + EXT4_S_ERR_START,
4655 save, EXT4_S_ERR_LEN);
4660 ext4_msg(sb, KERN_ERR, "error loading journal");
4661 jbd2_journal_destroy(journal);
4665 EXT4_SB(sb)->s_journal = journal;
4666 ext4_clear_journal_err(sb, es);
4668 if (!really_read_only && journal_devnum &&
4669 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4670 es->s_journal_dev = cpu_to_le32(journal_devnum);
4672 /* Make sure we flush the recovery flag to disk. */
4673 ext4_commit_super(sb, 1);
4679 static int ext4_commit_super(struct super_block *sb, int sync)
4681 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4682 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4685 if (!sbh || block_device_ejected(sb))
4688 * If the file system is mounted read-only, don't update the
4689 * superblock write time. This avoids updating the superblock
4690 * write time when we are mounting the root file system
4691 * read/only but we need to replay the journal; at that point,
4692 * for people who are east of GMT and who make their clock
4693 * tick in localtime for Windows bug-for-bug compatibility,
4694 * the clock is set in the future, and this will cause e2fsck
4695 * to complain and force a full file system check.
4697 if (!(sb->s_flags & SB_RDONLY))
4698 es->s_wtime = cpu_to_le32(get_seconds());
4699 if (sb->s_bdev->bd_part)
4700 es->s_kbytes_written =
4701 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4702 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4703 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4705 es->s_kbytes_written =
4706 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4707 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4708 ext4_free_blocks_count_set(es,
4709 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4710 &EXT4_SB(sb)->s_freeclusters_counter)));
4711 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4712 es->s_free_inodes_count =
4713 cpu_to_le32(percpu_counter_sum_positive(
4714 &EXT4_SB(sb)->s_freeinodes_counter));
4715 BUFFER_TRACE(sbh, "marking dirty");
4716 ext4_superblock_csum_set(sb);
4719 if (buffer_write_io_error(sbh)) {
4721 * Oh, dear. A previous attempt to write the
4722 * superblock failed. This could happen because the
4723 * USB device was yanked out. Or it could happen to
4724 * be a transient write error and maybe the block will
4725 * be remapped. Nothing we can do but to retry the
4726 * write and hope for the best.
4728 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4729 "superblock detected");
4730 clear_buffer_write_io_error(sbh);
4731 set_buffer_uptodate(sbh);
4733 mark_buffer_dirty(sbh);
4736 error = __sync_dirty_buffer(sbh,
4737 REQ_SYNC | (test_opt(sb, BARRIER) ? REQ_FUA : 0));
4741 error = buffer_write_io_error(sbh);
4743 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4745 clear_buffer_write_io_error(sbh);
4746 set_buffer_uptodate(sbh);
4753 * Have we just finished recovery? If so, and if we are mounting (or
4754 * remounting) the filesystem readonly, then we will end up with a
4755 * consistent fs on disk. Record that fact.
4757 static void ext4_mark_recovery_complete(struct super_block *sb,
4758 struct ext4_super_block *es)
4760 journal_t *journal = EXT4_SB(sb)->s_journal;
4762 if (!ext4_has_feature_journal(sb)) {
4763 BUG_ON(journal != NULL);
4766 jbd2_journal_lock_updates(journal);
4767 if (jbd2_journal_flush(journal) < 0)
4770 if (ext4_has_feature_journal_needs_recovery(sb) && sb_rdonly(sb)) {
4771 ext4_clear_feature_journal_needs_recovery(sb);
4772 ext4_commit_super(sb, 1);
4776 jbd2_journal_unlock_updates(journal);
4780 * If we are mounting (or read-write remounting) a filesystem whose journal
4781 * has recorded an error from a previous lifetime, move that error to the
4782 * main filesystem now.
4784 static void ext4_clear_journal_err(struct super_block *sb,
4785 struct ext4_super_block *es)
4791 BUG_ON(!ext4_has_feature_journal(sb));
4793 journal = EXT4_SB(sb)->s_journal;
4796 * Now check for any error status which may have been recorded in the
4797 * journal by a prior ext4_error() or ext4_abort()
4800 j_errno = jbd2_journal_errno(journal);
4804 errstr = ext4_decode_error(sb, j_errno, nbuf);
4805 ext4_warning(sb, "Filesystem error recorded "
4806 "from previous mount: %s", errstr);
4807 ext4_warning(sb, "Marking fs in need of filesystem check.");
4809 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4810 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4811 ext4_commit_super(sb, 1);
4813 jbd2_journal_clear_err(journal);
4814 jbd2_journal_update_sb_errno(journal);
4819 * Force the running and committing transactions to commit,
4820 * and wait on the commit.
4822 int ext4_force_commit(struct super_block *sb)
4829 journal = EXT4_SB(sb)->s_journal;
4830 return ext4_journal_force_commit(journal);
4833 static int ext4_sync_fs(struct super_block *sb, int wait)
4837 bool needs_barrier = false;
4838 struct ext4_sb_info *sbi = EXT4_SB(sb);
4840 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
4843 trace_ext4_sync_fs(sb, wait);
4844 flush_workqueue(sbi->rsv_conversion_wq);
4846 * Writeback quota in non-journalled quota case - journalled quota has
4849 dquot_writeback_dquots(sb, -1);
4851 * Data writeback is possible w/o journal transaction, so barrier must
4852 * being sent at the end of the function. But we can skip it if
4853 * transaction_commit will do it for us.
4855 if (sbi->s_journal) {
4856 target = jbd2_get_latest_transaction(sbi->s_journal);
4857 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4858 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4859 needs_barrier = true;
4861 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4863 ret = jbd2_log_wait_commit(sbi->s_journal,
4866 } else if (wait && test_opt(sb, BARRIER))
4867 needs_barrier = true;
4868 if (needs_barrier) {
4870 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4879 * LVM calls this function before a (read-only) snapshot is created. This
4880 * gives us a chance to flush the journal completely and mark the fs clean.
4882 * Note that only this function cannot bring a filesystem to be in a clean
4883 * state independently. It relies on upper layer to stop all data & metadata
4886 static int ext4_freeze(struct super_block *sb)
4894 journal = EXT4_SB(sb)->s_journal;
4897 /* Now we set up the journal barrier. */
4898 jbd2_journal_lock_updates(journal);
4901 * Don't clear the needs_recovery flag if we failed to
4902 * flush the journal.
4904 error = jbd2_journal_flush(journal);
4908 /* Journal blocked and flushed, clear needs_recovery flag. */
4909 ext4_clear_feature_journal_needs_recovery(sb);
4912 error = ext4_commit_super(sb, 1);
4915 /* we rely on upper layer to stop further updates */
4916 jbd2_journal_unlock_updates(journal);
4921 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4922 * flag here, even though the filesystem is not technically dirty yet.
4924 static int ext4_unfreeze(struct super_block *sb)
4926 if (sb_rdonly(sb) || ext4_forced_shutdown(EXT4_SB(sb)))
4929 if (EXT4_SB(sb)->s_journal) {
4930 /* Reset the needs_recovery flag before the fs is unlocked. */
4931 ext4_set_feature_journal_needs_recovery(sb);
4934 ext4_commit_super(sb, 1);
4939 * Structure to save mount options for ext4_remount's benefit
4941 struct ext4_mount_options {
4942 unsigned long s_mount_opt;
4943 unsigned long s_mount_opt2;
4946 unsigned long s_commit_interval;
4947 u32 s_min_batch_time, s_max_batch_time;
4950 char *s_qf_names[EXT4_MAXQUOTAS];
4954 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4956 struct ext4_super_block *es;
4957 struct ext4_sb_info *sbi = EXT4_SB(sb);
4958 unsigned long old_sb_flags;
4959 struct ext4_mount_options old_opts;
4960 int enable_quota = 0;
4962 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4967 char *orig_data = kstrdup(data, GFP_KERNEL);
4969 /* Store the original options */
4970 old_sb_flags = sb->s_flags;
4971 old_opts.s_mount_opt = sbi->s_mount_opt;
4972 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4973 old_opts.s_resuid = sbi->s_resuid;
4974 old_opts.s_resgid = sbi->s_resgid;
4975 old_opts.s_commit_interval = sbi->s_commit_interval;
4976 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4977 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4979 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4980 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4981 if (sbi->s_qf_names[i]) {
4982 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4984 if (!old_opts.s_qf_names[i]) {
4985 for (j = 0; j < i; j++)
4986 kfree(old_opts.s_qf_names[j]);
4991 old_opts.s_qf_names[i] = NULL;
4993 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4994 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4996 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
5001 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
5002 test_opt(sb, JOURNAL_CHECKSUM)) {
5003 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
5004 "during remount not supported; ignoring");
5005 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
5008 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
5009 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
5010 ext4_msg(sb, KERN_ERR, "can't mount with "
5011 "both data=journal and delalloc");
5015 if (test_opt(sb, DIOREAD_NOLOCK)) {
5016 ext4_msg(sb, KERN_ERR, "can't mount with "
5017 "both data=journal and dioread_nolock");
5021 if (test_opt(sb, DAX)) {
5022 ext4_msg(sb, KERN_ERR, "can't mount with "
5023 "both data=journal and dax");
5027 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
5028 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
5029 ext4_msg(sb, KERN_ERR, "can't mount with "
5030 "journal_async_commit in data=ordered mode");
5036 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) {
5037 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount");
5042 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
5043 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
5044 "dax flag with busy inodes while remounting");
5045 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
5048 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
5049 ext4_abort(sb, "Abort forced by user");
5051 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
5052 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
5056 if (sbi->s_journal) {
5057 ext4_init_journal_params(sb, sbi->s_journal);
5058 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
5061 if (*flags & SB_LAZYTIME)
5062 sb->s_flags |= SB_LAZYTIME;
5064 if ((bool)(*flags & SB_RDONLY) != sb_rdonly(sb)) {
5065 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
5070 if (*flags & SB_RDONLY) {
5071 err = sync_filesystem(sb);
5074 err = dquot_suspend(sb, -1);
5079 * First of all, the unconditional stuff we have to do
5080 * to disable replay of the journal when we next remount
5082 sb->s_flags |= SB_RDONLY;
5085 * OK, test if we are remounting a valid rw partition
5086 * readonly, and if so set the rdonly flag and then
5087 * mark the partition as valid again.
5089 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
5090 (sbi->s_mount_state & EXT4_VALID_FS))
5091 es->s_state = cpu_to_le16(sbi->s_mount_state);
5094 ext4_mark_recovery_complete(sb, es);
5096 /* Make sure we can mount this feature set readwrite */
5097 if (ext4_has_feature_readonly(sb) ||
5098 !ext4_feature_set_ok(sb, 0)) {
5103 * Make sure the group descriptor checksums
5104 * are sane. If they aren't, refuse to remount r/w.
5106 for (g = 0; g < sbi->s_groups_count; g++) {
5107 struct ext4_group_desc *gdp =
5108 ext4_get_group_desc(sb, g, NULL);
5110 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
5111 ext4_msg(sb, KERN_ERR,
5112 "ext4_remount: Checksum for group %u failed (%u!=%u)",
5113 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
5114 le16_to_cpu(gdp->bg_checksum));
5121 * If we have an unprocessed orphan list hanging
5122 * around from a previously readonly bdev mount,
5123 * require a full umount/remount for now.
5125 if (es->s_last_orphan) {
5126 ext4_msg(sb, KERN_WARNING, "Couldn't "
5127 "remount RDWR because of unprocessed "
5128 "orphan inode list. Please "
5129 "umount/remount instead");
5135 * Mounting a RDONLY partition read-write, so reread
5136 * and store the current valid flag. (It may have
5137 * been changed by e2fsck since we originally mounted
5141 ext4_clear_journal_err(sb, es);
5142 sbi->s_mount_state = le16_to_cpu(es->s_state);
5143 if (!ext4_setup_super(sb, es, 0))
5144 sb->s_flags &= ~SB_RDONLY;
5145 if (ext4_has_feature_mmp(sb))
5146 if (ext4_multi_mount_protect(sb,
5147 le64_to_cpu(es->s_mmp_block))) {
5156 * Reinitialize lazy itable initialization thread based on
5159 if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE))
5160 ext4_unregister_li_request(sb);
5162 ext4_group_t first_not_zeroed;
5163 first_not_zeroed = ext4_has_uninit_itable(sb);
5164 ext4_register_li_request(sb, first_not_zeroed);
5167 ext4_setup_system_zone(sb);
5168 if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY))
5169 ext4_commit_super(sb, 1);
5172 /* Release old quota file names */
5173 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5174 kfree(old_opts.s_qf_names[i]);
5176 if (sb_any_quota_suspended(sb))
5177 dquot_resume(sb, -1);
5178 else if (ext4_has_feature_quota(sb)) {
5179 err = ext4_enable_quotas(sb);
5186 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
5187 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5192 sb->s_flags = old_sb_flags;
5193 sbi->s_mount_opt = old_opts.s_mount_opt;
5194 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
5195 sbi->s_resuid = old_opts.s_resuid;
5196 sbi->s_resgid = old_opts.s_resgid;
5197 sbi->s_commit_interval = old_opts.s_commit_interval;
5198 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5199 sbi->s_max_batch_time = old_opts.s_max_batch_time;
5201 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
5202 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
5203 kfree(sbi->s_qf_names[i]);
5204 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
5212 static int ext4_statfs_project(struct super_block *sb,
5213 kprojid_t projid, struct kstatfs *buf)
5216 struct dquot *dquot;
5220 qid = make_kqid_projid(projid);
5221 dquot = dqget(sb, qid);
5223 return PTR_ERR(dquot);
5224 spin_lock(&dquot->dq_dqb_lock);
5226 limit = (dquot->dq_dqb.dqb_bsoftlimit ?
5227 dquot->dq_dqb.dqb_bsoftlimit :
5228 dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits;
5229 if (limit && buf->f_blocks > limit) {
5230 curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits;
5231 buf->f_blocks = limit;
5232 buf->f_bfree = buf->f_bavail =
5233 (buf->f_blocks > curblock) ?
5234 (buf->f_blocks - curblock) : 0;
5237 limit = dquot->dq_dqb.dqb_isoftlimit ?
5238 dquot->dq_dqb.dqb_isoftlimit :
5239 dquot->dq_dqb.dqb_ihardlimit;
5240 if (limit && buf->f_files > limit) {
5241 buf->f_files = limit;
5243 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
5244 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
5247 spin_unlock(&dquot->dq_dqb_lock);
5253 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
5255 struct super_block *sb = dentry->d_sb;
5256 struct ext4_sb_info *sbi = EXT4_SB(sb);
5257 struct ext4_super_block *es = sbi->s_es;
5258 ext4_fsblk_t overhead = 0, resv_blocks;
5261 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
5263 if (!test_opt(sb, MINIX_DF))
5264 overhead = sbi->s_overhead;
5266 buf->f_type = EXT4_SUPER_MAGIC;
5267 buf->f_bsize = sb->s_blocksize;
5268 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
5269 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5270 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
5271 /* prevent underflow in case that few free space is available */
5272 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
5273 buf->f_bavail = buf->f_bfree -
5274 (ext4_r_blocks_count(es) + resv_blocks);
5275 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
5277 buf->f_files = le32_to_cpu(es->s_inodes_count);
5278 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5279 buf->f_namelen = EXT4_NAME_LEN;
5280 fsid = le64_to_cpup((void *)es->s_uuid) ^
5281 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5282 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5283 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
5286 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
5287 sb_has_quota_limits_enabled(sb, PRJQUOTA))
5288 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
5297 * Helper functions so that transaction is started before we acquire dqio_sem
5298 * to keep correct lock ordering of transaction > dqio_sem
5300 static inline struct inode *dquot_to_inode(struct dquot *dquot)
5302 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
5305 static int ext4_write_dquot(struct dquot *dquot)
5309 struct inode *inode;
5311 inode = dquot_to_inode(dquot);
5312 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
5313 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
5315 return PTR_ERR(handle);
5316 ret = dquot_commit(dquot);
5317 err = ext4_journal_stop(handle);
5323 static int ext4_acquire_dquot(struct dquot *dquot)
5328 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5329 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
5331 return PTR_ERR(handle);
5332 ret = dquot_acquire(dquot);
5333 err = ext4_journal_stop(handle);
5339 static int ext4_release_dquot(struct dquot *dquot)
5344 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5345 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
5346 if (IS_ERR(handle)) {
5347 /* Release dquot anyway to avoid endless cycle in dqput() */
5348 dquot_release(dquot);
5349 return PTR_ERR(handle);
5351 ret = dquot_release(dquot);
5352 err = ext4_journal_stop(handle);
5358 static int ext4_mark_dquot_dirty(struct dquot *dquot)
5360 struct super_block *sb = dquot->dq_sb;
5361 struct ext4_sb_info *sbi = EXT4_SB(sb);
5363 /* Are we journaling quotas? */
5364 if (ext4_has_feature_quota(sb) ||
5365 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
5366 dquot_mark_dquot_dirty(dquot);
5367 return ext4_write_dquot(dquot);
5369 return dquot_mark_dquot_dirty(dquot);
5373 static int ext4_write_info(struct super_block *sb, int type)
5378 /* Data block + inode block */
5379 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
5381 return PTR_ERR(handle);
5382 ret = dquot_commit_info(sb, type);
5383 err = ext4_journal_stop(handle);
5390 * Turn on quotas during mount time - we need to find
5391 * the quota file and such...
5393 static int ext4_quota_on_mount(struct super_block *sb, int type)
5395 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5396 EXT4_SB(sb)->s_jquota_fmt, type);
5399 static void lockdep_set_quota_inode(struct inode *inode, int subclass)
5401 struct ext4_inode_info *ei = EXT4_I(inode);
5403 /* The first argument of lockdep_set_subclass has to be
5404 * *exactly* the same as the argument to init_rwsem() --- in
5405 * this case, in init_once() --- or lockdep gets unhappy
5406 * because the name of the lock is set using the
5407 * stringification of the argument to init_rwsem().
5409 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
5410 lockdep_set_subclass(&ei->i_data_sem, subclass);
5414 * Standard function to be called on quota_on
5416 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
5417 const struct path *path)
5421 if (!test_opt(sb, QUOTA))
5424 /* Quotafile not on the same filesystem? */
5425 if (path->dentry->d_sb != sb)
5427 /* Journaling quota? */
5428 if (EXT4_SB(sb)->s_qf_names[type]) {
5429 /* Quotafile not in fs root? */
5430 if (path->dentry->d_parent != sb->s_root)
5431 ext4_msg(sb, KERN_WARNING,
5432 "Quota file not on filesystem root. "
5433 "Journaled quota will not work");
5434 sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY;
5437 * Clear the flag just in case mount options changed since
5440 sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY;
5444 * When we journal data on quota file, we have to flush journal to see
5445 * all updates to the file when we bypass pagecache...
5447 if (EXT4_SB(sb)->s_journal &&
5448 ext4_should_journal_data(d_inode(path->dentry))) {
5450 * We don't need to lock updates but journal_flush() could
5451 * otherwise be livelocked...
5453 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
5454 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
5455 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
5460 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
5461 err = dquot_quota_on(sb, type, format_id, path);
5463 lockdep_set_quota_inode(path->dentry->d_inode,
5466 struct inode *inode = d_inode(path->dentry);
5470 * Set inode flags to prevent userspace from messing with quota
5471 * files. If this fails, we return success anyway since quotas
5472 * are already enabled and this is not a hard failure.
5475 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5478 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL;
5479 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
5480 S_NOATIME | S_IMMUTABLE);
5481 ext4_mark_inode_dirty(handle, inode);
5482 ext4_journal_stop(handle);
5484 inode_unlock(inode);
5489 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5493 struct inode *qf_inode;
5494 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5495 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5496 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5497 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
5500 BUG_ON(!ext4_has_feature_quota(sb));
5502 if (!qf_inums[type])
5505 qf_inode = ext4_iget(sb, qf_inums[type]);
5506 if (IS_ERR(qf_inode)) {
5507 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5508 return PTR_ERR(qf_inode);
5511 /* Don't account quota for quota files to avoid recursion */
5512 qf_inode->i_flags |= S_NOQUOTA;
5513 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
5514 err = dquot_enable(qf_inode, type, format_id, flags);
5517 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
5522 /* Enable usage tracking for all quota types. */
5523 static int ext4_enable_quotas(struct super_block *sb)
5526 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5527 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5528 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5529 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
5531 bool quota_mopt[EXT4_MAXQUOTAS] = {
5532 test_opt(sb, USRQUOTA),
5533 test_opt(sb, GRPQUOTA),
5534 test_opt(sb, PRJQUOTA),
5537 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY;
5538 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
5539 if (qf_inums[type]) {
5540 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5541 DQUOT_USAGE_ENABLED |
5542 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
5544 for (type--; type >= 0; type--)
5545 dquot_quota_off(sb, type);
5548 "Failed to enable quota tracking "
5549 "(type=%d, err=%d). Please run "
5550 "e2fsck to fix.", type, err);
5558 static int ext4_quota_off(struct super_block *sb, int type)
5560 struct inode *inode = sb_dqopt(sb)->files[type];
5564 /* Force all delayed allocation blocks to be allocated.
5565 * Caller already holds s_umount sem */
5566 if (test_opt(sb, DELALLOC))
5567 sync_filesystem(sb);
5569 if (!inode || !igrab(inode))
5572 err = dquot_quota_off(sb, type);
5573 if (err || ext4_has_feature_quota(sb))
5578 * Update modification times of quota files when userspace can
5579 * start looking at them. If we fail, we return success anyway since
5580 * this is not a hard failure and quotas are already disabled.
5582 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5585 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL);
5586 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
5587 inode->i_mtime = inode->i_ctime = current_time(inode);
5588 ext4_mark_inode_dirty(handle, inode);
5589 ext4_journal_stop(handle);
5591 inode_unlock(inode);
5593 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL);
5597 return dquot_quota_off(sb, type);
5600 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5601 * acquiring the locks... As quota files are never truncated and quota code
5602 * itself serializes the operations (and no one else should touch the files)
5603 * we don't have to be afraid of races */
5604 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5605 size_t len, loff_t off)
5607 struct inode *inode = sb_dqopt(sb)->files[type];
5608 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5609 int offset = off & (sb->s_blocksize - 1);
5612 struct buffer_head *bh;
5613 loff_t i_size = i_size_read(inode);
5617 if (off+len > i_size)
5620 while (toread > 0) {
5621 tocopy = sb->s_blocksize - offset < toread ?
5622 sb->s_blocksize - offset : toread;
5623 bh = ext4_bread(NULL, inode, blk, 0);
5626 if (!bh) /* A hole? */
5627 memset(data, 0, tocopy);
5629 memcpy(data, bh->b_data+offset, tocopy);
5639 /* Write to quotafile (we know the transaction is already started and has
5640 * enough credits) */
5641 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5642 const char *data, size_t len, loff_t off)
5644 struct inode *inode = sb_dqopt(sb)->files[type];
5645 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5646 int err, offset = off & (sb->s_blocksize - 1);
5648 struct buffer_head *bh;
5649 handle_t *handle = journal_current_handle();
5651 if (EXT4_SB(sb)->s_journal && !handle) {
5652 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5653 " cancelled because transaction is not started",
5654 (unsigned long long)off, (unsigned long long)len);
5658 * Since we account only one data block in transaction credits,
5659 * then it is impossible to cross a block boundary.
5661 if (sb->s_blocksize - offset < len) {
5662 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5663 " cancelled because not block aligned",
5664 (unsigned long long)off, (unsigned long long)len);
5669 bh = ext4_bread(handle, inode, blk,
5670 EXT4_GET_BLOCKS_CREATE |
5671 EXT4_GET_BLOCKS_METADATA_NOFAIL);
5672 } while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) &&
5673 ext4_should_retry_alloc(inode->i_sb, &retries));
5678 BUFFER_TRACE(bh, "get write access");
5679 err = ext4_journal_get_write_access(handle, bh);
5685 memcpy(bh->b_data+offset, data, len);
5686 flush_dcache_page(bh->b_page);
5688 err = ext4_handle_dirty_metadata(handle, NULL, bh);
5691 if (inode->i_size < off + len) {
5692 i_size_write(inode, off + len);
5693 EXT4_I(inode)->i_disksize = inode->i_size;
5694 ext4_mark_inode_dirty(handle, inode);
5699 static int ext4_get_next_id(struct super_block *sb, struct kqid *qid)
5701 const struct quota_format_ops *ops;
5703 if (!sb_has_quota_loaded(sb, qid->type))
5705 ops = sb_dqopt(sb)->ops[qid->type];
5706 if (!ops || !ops->get_next_id)
5708 return dquot_get_next_id(sb, qid);
5712 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5713 const char *dev_name, void *data)
5715 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5718 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
5719 static inline void register_as_ext2(void)
5721 int err = register_filesystem(&ext2_fs_type);
5724 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5727 static inline void unregister_as_ext2(void)
5729 unregister_filesystem(&ext2_fs_type);
5732 static inline int ext2_feature_set_ok(struct super_block *sb)
5734 if (ext4_has_unknown_ext2_incompat_features(sb))
5738 if (ext4_has_unknown_ext2_ro_compat_features(sb))
5743 static inline void register_as_ext2(void) { }
5744 static inline void unregister_as_ext2(void) { }
5745 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5748 static inline void register_as_ext3(void)
5750 int err = register_filesystem(&ext3_fs_type);
5753 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5756 static inline void unregister_as_ext3(void)
5758 unregister_filesystem(&ext3_fs_type);
5761 static inline int ext3_feature_set_ok(struct super_block *sb)
5763 if (ext4_has_unknown_ext3_incompat_features(sb))
5765 if (!ext4_has_feature_journal(sb))
5769 if (ext4_has_unknown_ext3_ro_compat_features(sb))
5774 static struct file_system_type ext4_fs_type = {
5775 .owner = THIS_MODULE,
5777 .mount = ext4_mount,
5778 .kill_sb = kill_block_super,
5779 .fs_flags = FS_REQUIRES_DEV,
5781 MODULE_ALIAS_FS("ext4");
5783 /* Shared across all ext4 file systems */
5784 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5786 static int __init ext4_init_fs(void)
5790 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
5791 ext4_li_info = NULL;
5792 mutex_init(&ext4_li_mtx);
5794 /* Build-time check for flags consistency */
5795 ext4_check_flag_values();
5797 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
5798 init_waitqueue_head(&ext4__ioend_wq[i]);
5800 err = ext4_init_es();
5804 err = ext4_init_pageio();
5808 err = ext4_init_system_zone();
5812 err = ext4_init_sysfs();
5816 err = ext4_init_mballoc();
5819 err = init_inodecache();
5824 err = register_filesystem(&ext4_fs_type);
5830 unregister_as_ext2();
5831 unregister_as_ext3();
5832 destroy_inodecache();
5834 ext4_exit_mballoc();
5838 ext4_exit_system_zone();
5847 static void __exit ext4_exit_fs(void)
5849 ext4_destroy_lazyinit_thread();
5850 unregister_as_ext2();
5851 unregister_as_ext3();
5852 unregister_filesystem(&ext4_fs_type);
5853 destroy_inodecache();
5854 ext4_exit_mballoc();
5856 ext4_exit_system_zone();
5861 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5862 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5863 MODULE_LICENSE("GPL");
5864 module_init(ext4_init_fs)
5865 module_exit(ext4_exit_fs)