2 * linux/fs/ext4/super.c
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
11 * linux/fs/minix/inode.c
13 * Copyright (C) 1991, 1992 Linus Torvalds
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
19 #include <linux/module.h>
20 #include <linux/string.h>
22 #include <linux/time.h>
23 #include <linux/vmalloc.h>
24 #include <linux/slab.h>
25 #include <linux/init.h>
26 #include <linux/blkdev.h>
27 #include <linux/backing-dev.h>
28 #include <linux/parser.h>
29 #include <linux/buffer_head.h>
30 #include <linux/exportfs.h>
31 #include <linux/vfs.h>
32 #include <linux/random.h>
33 #include <linux/mount.h>
34 #include <linux/namei.h>
35 #include <linux/quotaops.h>
36 #include <linux/seq_file.h>
37 #include <linux/ctype.h>
38 #include <linux/log2.h>
39 #include <linux/crc16.h>
40 #include <linux/dax.h>
41 #include <linux/cleancache.h>
42 #include <linux/uaccess.h>
44 #include <linux/kthread.h>
45 #include <linux/freezer.h>
48 #include "ext4_extents.h" /* Needed for trace points definition */
49 #include "ext4_jbd2.h"
55 #define CREATE_TRACE_POINTS
56 #include <trace/events/ext4.h>
58 static struct ext4_lazy_init *ext4_li_info;
59 static struct mutex ext4_li_mtx;
60 static struct ratelimit_state ext4_mount_msg_ratelimit;
62 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
63 unsigned long journal_devnum);
64 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
65 static int ext4_commit_super(struct super_block *sb, int sync);
66 static void ext4_mark_recovery_complete(struct super_block *sb,
67 struct ext4_super_block *es);
68 static void ext4_clear_journal_err(struct super_block *sb,
69 struct ext4_super_block *es);
70 static int ext4_sync_fs(struct super_block *sb, int wait);
71 static int ext4_remount(struct super_block *sb, int *flags, char *data);
72 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
73 static int ext4_unfreeze(struct super_block *sb);
74 static int ext4_freeze(struct super_block *sb);
75 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
76 const char *dev_name, void *data);
77 static inline int ext2_feature_set_ok(struct super_block *sb);
78 static inline int ext3_feature_set_ok(struct super_block *sb);
79 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
80 static void ext4_destroy_lazyinit_thread(void);
81 static void ext4_unregister_li_request(struct super_block *sb);
82 static void ext4_clear_request_list(void);
83 static struct inode *ext4_get_journal_inode(struct super_block *sb,
84 unsigned int journal_inum);
89 * Note the difference between i_mmap_sem (EXT4_I(inode)->i_mmap_sem) and
90 * i_mmap_rwsem (inode->i_mmap_rwsem)!
93 * mmap_sem -> sb_start_pagefault -> i_mmap_sem (r) -> transaction start ->
94 * page lock -> i_data_sem (rw)
96 * buffered write path:
97 * sb_start_write -> i_mutex -> mmap_sem
98 * sb_start_write -> i_mutex -> transaction start -> page lock ->
102 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) ->
103 * i_mmap_rwsem (w) -> page lock
104 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) ->
105 * transaction start -> i_data_sem (rw)
108 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) -> mmap_sem
109 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) ->
110 * transaction start -> i_data_sem (rw)
113 * transaction start -> page lock(s) -> i_data_sem (rw)
116 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
117 static struct file_system_type ext2_fs_type = {
118 .owner = THIS_MODULE,
121 .kill_sb = kill_block_super,
122 .fs_flags = FS_REQUIRES_DEV,
124 MODULE_ALIAS_FS("ext2");
125 MODULE_ALIAS("ext2");
126 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
128 #define IS_EXT2_SB(sb) (0)
132 static struct file_system_type ext3_fs_type = {
133 .owner = THIS_MODULE,
136 .kill_sb = kill_block_super,
137 .fs_flags = FS_REQUIRES_DEV,
139 MODULE_ALIAS_FS("ext3");
140 MODULE_ALIAS("ext3");
141 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
143 static int ext4_verify_csum_type(struct super_block *sb,
144 struct ext4_super_block *es)
146 if (!ext4_has_feature_metadata_csum(sb))
149 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
152 static __le32 ext4_superblock_csum(struct super_block *sb,
153 struct ext4_super_block *es)
155 struct ext4_sb_info *sbi = EXT4_SB(sb);
156 int offset = offsetof(struct ext4_super_block, s_checksum);
159 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
161 return cpu_to_le32(csum);
164 static int ext4_superblock_csum_verify(struct super_block *sb,
165 struct ext4_super_block *es)
167 if (!ext4_has_metadata_csum(sb))
170 return es->s_checksum == ext4_superblock_csum(sb, es);
173 void ext4_superblock_csum_set(struct super_block *sb)
175 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
177 if (!ext4_has_metadata_csum(sb))
180 es->s_checksum = ext4_superblock_csum(sb, es);
183 void *ext4_kvmalloc(size_t size, gfp_t flags)
187 ret = kmalloc(size, flags | __GFP_NOWARN);
189 ret = __vmalloc(size, flags, PAGE_KERNEL);
193 void *ext4_kvzalloc(size_t size, gfp_t flags)
197 ret = kzalloc(size, flags | __GFP_NOWARN);
199 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
203 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
204 struct ext4_group_desc *bg)
206 return le32_to_cpu(bg->bg_block_bitmap_lo) |
207 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
208 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
211 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
212 struct ext4_group_desc *bg)
214 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
215 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
216 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
219 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
220 struct ext4_group_desc *bg)
222 return le32_to_cpu(bg->bg_inode_table_lo) |
223 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
224 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
227 __u32 ext4_free_group_clusters(struct super_block *sb,
228 struct ext4_group_desc *bg)
230 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
231 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
232 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
235 __u32 ext4_free_inodes_count(struct super_block *sb,
236 struct ext4_group_desc *bg)
238 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
239 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
240 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
243 __u32 ext4_used_dirs_count(struct super_block *sb,
244 struct ext4_group_desc *bg)
246 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
247 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
248 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
251 __u32 ext4_itable_unused_count(struct super_block *sb,
252 struct ext4_group_desc *bg)
254 return le16_to_cpu(bg->bg_itable_unused_lo) |
255 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
256 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
259 void ext4_block_bitmap_set(struct super_block *sb,
260 struct ext4_group_desc *bg, ext4_fsblk_t blk)
262 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
263 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
264 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
267 void ext4_inode_bitmap_set(struct super_block *sb,
268 struct ext4_group_desc *bg, ext4_fsblk_t blk)
270 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
271 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
272 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
275 void ext4_inode_table_set(struct super_block *sb,
276 struct ext4_group_desc *bg, ext4_fsblk_t blk)
278 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
279 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
280 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
283 void ext4_free_group_clusters_set(struct super_block *sb,
284 struct ext4_group_desc *bg, __u32 count)
286 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
287 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
288 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
291 void ext4_free_inodes_set(struct super_block *sb,
292 struct ext4_group_desc *bg, __u32 count)
294 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
295 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
296 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
299 void ext4_used_dirs_set(struct super_block *sb,
300 struct ext4_group_desc *bg, __u32 count)
302 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
303 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
304 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
307 void ext4_itable_unused_set(struct super_block *sb,
308 struct ext4_group_desc *bg, __u32 count)
310 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
311 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
312 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
316 static void __save_error_info(struct super_block *sb, const char *func,
319 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
321 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
322 if (bdev_read_only(sb->s_bdev))
324 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
325 es->s_last_error_time = cpu_to_le32(get_seconds());
326 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
327 es->s_last_error_line = cpu_to_le32(line);
328 if (!es->s_first_error_time) {
329 es->s_first_error_time = es->s_last_error_time;
330 strncpy(es->s_first_error_func, func,
331 sizeof(es->s_first_error_func));
332 es->s_first_error_line = cpu_to_le32(line);
333 es->s_first_error_ino = es->s_last_error_ino;
334 es->s_first_error_block = es->s_last_error_block;
337 * Start the daily error reporting function if it hasn't been
340 if (!es->s_error_count)
341 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
342 le32_add_cpu(&es->s_error_count, 1);
345 static void save_error_info(struct super_block *sb, const char *func,
348 __save_error_info(sb, func, line);
349 ext4_commit_super(sb, 1);
353 * The del_gendisk() function uninitializes the disk-specific data
354 * structures, including the bdi structure, without telling anyone
355 * else. Once this happens, any attempt to call mark_buffer_dirty()
356 * (for example, by ext4_commit_super), will cause a kernel OOPS.
357 * This is a kludge to prevent these oops until we can put in a proper
358 * hook in del_gendisk() to inform the VFS and file system layers.
360 static int block_device_ejected(struct super_block *sb)
362 struct inode *bd_inode = sb->s_bdev->bd_inode;
363 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
365 return bdi->dev == NULL;
368 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
370 struct super_block *sb = journal->j_private;
371 struct ext4_sb_info *sbi = EXT4_SB(sb);
372 int error = is_journal_aborted(journal);
373 struct ext4_journal_cb_entry *jce;
375 BUG_ON(txn->t_state == T_FINISHED);
377 ext4_process_freed_data(sb, txn->t_tid);
379 spin_lock(&sbi->s_md_lock);
380 while (!list_empty(&txn->t_private_list)) {
381 jce = list_entry(txn->t_private_list.next,
382 struct ext4_journal_cb_entry, jce_list);
383 list_del_init(&jce->jce_list);
384 spin_unlock(&sbi->s_md_lock);
385 jce->jce_func(sb, jce, error);
386 spin_lock(&sbi->s_md_lock);
388 spin_unlock(&sbi->s_md_lock);
391 /* Deal with the reporting of failure conditions on a filesystem such as
392 * inconsistencies detected or read IO failures.
394 * On ext2, we can store the error state of the filesystem in the
395 * superblock. That is not possible on ext4, because we may have other
396 * write ordering constraints on the superblock which prevent us from
397 * writing it out straight away; and given that the journal is about to
398 * be aborted, we can't rely on the current, or future, transactions to
399 * write out the superblock safely.
401 * We'll just use the jbd2_journal_abort() error code to record an error in
402 * the journal instead. On recovery, the journal will complain about
403 * that error until we've noted it down and cleared it.
406 static void ext4_handle_error(struct super_block *sb)
408 if (sb->s_flags & MS_RDONLY)
411 if (!test_opt(sb, ERRORS_CONT)) {
412 journal_t *journal = EXT4_SB(sb)->s_journal;
414 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
416 jbd2_journal_abort(journal, -EIO);
418 if (test_opt(sb, ERRORS_RO)) {
419 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
421 * Make sure updated value of ->s_mount_flags will be visible
422 * before ->s_flags update
425 sb->s_flags |= MS_RDONLY;
427 if (test_opt(sb, ERRORS_PANIC)) {
428 if (EXT4_SB(sb)->s_journal &&
429 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
431 panic("EXT4-fs (device %s): panic forced after error\n",
436 #define ext4_error_ratelimit(sb) \
437 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
440 void __ext4_error(struct super_block *sb, const char *function,
441 unsigned int line, const char *fmt, ...)
443 struct va_format vaf;
446 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
449 if (ext4_error_ratelimit(sb)) {
454 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
455 sb->s_id, function, line, current->comm, &vaf);
458 save_error_info(sb, function, line);
459 ext4_handle_error(sb);
462 void __ext4_error_inode(struct inode *inode, const char *function,
463 unsigned int line, ext4_fsblk_t block,
464 const char *fmt, ...)
467 struct va_format vaf;
468 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
470 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
473 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
474 es->s_last_error_block = cpu_to_le64(block);
475 if (ext4_error_ratelimit(inode->i_sb)) {
480 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
481 "inode #%lu: block %llu: comm %s: %pV\n",
482 inode->i_sb->s_id, function, line, inode->i_ino,
483 block, current->comm, &vaf);
485 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
486 "inode #%lu: comm %s: %pV\n",
487 inode->i_sb->s_id, function, line, inode->i_ino,
488 current->comm, &vaf);
491 save_error_info(inode->i_sb, function, line);
492 ext4_handle_error(inode->i_sb);
495 void __ext4_error_file(struct file *file, const char *function,
496 unsigned int line, ext4_fsblk_t block,
497 const char *fmt, ...)
500 struct va_format vaf;
501 struct ext4_super_block *es;
502 struct inode *inode = file_inode(file);
503 char pathname[80], *path;
505 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
508 es = EXT4_SB(inode->i_sb)->s_es;
509 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
510 if (ext4_error_ratelimit(inode->i_sb)) {
511 path = file_path(file, pathname, sizeof(pathname));
519 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
520 "block %llu: comm %s: path %s: %pV\n",
521 inode->i_sb->s_id, function, line, inode->i_ino,
522 block, current->comm, path, &vaf);
525 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
526 "comm %s: path %s: %pV\n",
527 inode->i_sb->s_id, function, line, inode->i_ino,
528 current->comm, path, &vaf);
531 save_error_info(inode->i_sb, function, line);
532 ext4_handle_error(inode->i_sb);
535 const char *ext4_decode_error(struct super_block *sb, int errno,
542 errstr = "Corrupt filesystem";
545 errstr = "Filesystem failed CRC";
548 errstr = "IO failure";
551 errstr = "Out of memory";
554 if (!sb || (EXT4_SB(sb)->s_journal &&
555 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
556 errstr = "Journal has aborted";
558 errstr = "Readonly filesystem";
561 /* If the caller passed in an extra buffer for unknown
562 * errors, textualise them now. Else we just return
565 /* Check for truncated error codes... */
566 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
575 /* __ext4_std_error decodes expected errors from journaling functions
576 * automatically and invokes the appropriate error response. */
578 void __ext4_std_error(struct super_block *sb, const char *function,
579 unsigned int line, int errno)
584 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
587 /* Special case: if the error is EROFS, and we're not already
588 * inside a transaction, then there's really no point in logging
590 if (errno == -EROFS && journal_current_handle() == NULL &&
591 (sb->s_flags & MS_RDONLY))
594 if (ext4_error_ratelimit(sb)) {
595 errstr = ext4_decode_error(sb, errno, nbuf);
596 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
597 sb->s_id, function, line, errstr);
600 save_error_info(sb, function, line);
601 ext4_handle_error(sb);
605 * ext4_abort is a much stronger failure handler than ext4_error. The
606 * abort function may be used to deal with unrecoverable failures such
607 * as journal IO errors or ENOMEM at a critical moment in log management.
609 * We unconditionally force the filesystem into an ABORT|READONLY state,
610 * unless the error response on the fs has been set to panic in which
611 * case we take the easy way out and panic immediately.
614 void __ext4_abort(struct super_block *sb, const char *function,
615 unsigned int line, const char *fmt, ...)
617 struct va_format vaf;
620 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
623 save_error_info(sb, function, line);
627 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: %pV\n",
628 sb->s_id, function, line, &vaf);
631 if ((sb->s_flags & MS_RDONLY) == 0) {
632 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
633 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
635 * Make sure updated value of ->s_mount_flags will be visible
636 * before ->s_flags update
639 sb->s_flags |= MS_RDONLY;
640 if (EXT4_SB(sb)->s_journal)
641 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
642 save_error_info(sb, function, line);
644 if (test_opt(sb, ERRORS_PANIC)) {
645 if (EXT4_SB(sb)->s_journal &&
646 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
648 panic("EXT4-fs panic from previous error\n");
652 void __ext4_msg(struct super_block *sb,
653 const char *prefix, const char *fmt, ...)
655 struct va_format vaf;
658 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
664 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
668 #define ext4_warning_ratelimit(sb) \
669 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \
672 void __ext4_warning(struct super_block *sb, const char *function,
673 unsigned int line, const char *fmt, ...)
675 struct va_format vaf;
678 if (!ext4_warning_ratelimit(sb))
684 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
685 sb->s_id, function, line, &vaf);
689 void __ext4_warning_inode(const struct inode *inode, const char *function,
690 unsigned int line, const char *fmt, ...)
692 struct va_format vaf;
695 if (!ext4_warning_ratelimit(inode->i_sb))
701 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
702 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
703 function, line, inode->i_ino, current->comm, &vaf);
707 void __ext4_grp_locked_error(const char *function, unsigned int line,
708 struct super_block *sb, ext4_group_t grp,
709 unsigned long ino, ext4_fsblk_t block,
710 const char *fmt, ...)
714 struct va_format vaf;
716 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
718 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
721 es->s_last_error_ino = cpu_to_le32(ino);
722 es->s_last_error_block = cpu_to_le64(block);
723 __save_error_info(sb, function, line);
725 if (ext4_error_ratelimit(sb)) {
729 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
730 sb->s_id, function, line, grp);
732 printk(KERN_CONT "inode %lu: ", ino);
734 printk(KERN_CONT "block %llu:",
735 (unsigned long long) block);
736 printk(KERN_CONT "%pV\n", &vaf);
740 if (test_opt(sb, ERRORS_CONT)) {
741 ext4_commit_super(sb, 0);
745 ext4_unlock_group(sb, grp);
746 ext4_handle_error(sb);
748 * We only get here in the ERRORS_RO case; relocking the group
749 * may be dangerous, but nothing bad will happen since the
750 * filesystem will have already been marked read/only and the
751 * journal has been aborted. We return 1 as a hint to callers
752 * who might what to use the return value from
753 * ext4_grp_locked_error() to distinguish between the
754 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
755 * aggressively from the ext4 function in question, with a
756 * more appropriate error code.
758 ext4_lock_group(sb, grp);
762 void ext4_update_dynamic_rev(struct super_block *sb)
764 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
766 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
770 "updating to rev %d because of new feature flag, "
771 "running e2fsck is recommended",
774 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
775 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
776 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
777 /* leave es->s_feature_*compat flags alone */
778 /* es->s_uuid will be set by e2fsck if empty */
781 * The rest of the superblock fields should be zero, and if not it
782 * means they are likely already in use, so leave them alone. We
783 * can leave it up to e2fsck to clean up any inconsistencies there.
788 * Open the external journal device
790 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
792 struct block_device *bdev;
793 char b[BDEVNAME_SIZE];
795 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
801 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
802 __bdevname(dev, b), PTR_ERR(bdev));
807 * Release the journal device
809 static void ext4_blkdev_put(struct block_device *bdev)
811 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
814 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
816 struct block_device *bdev;
817 bdev = sbi->journal_bdev;
819 ext4_blkdev_put(bdev);
820 sbi->journal_bdev = NULL;
824 static inline struct inode *orphan_list_entry(struct list_head *l)
826 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
829 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
833 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
834 le32_to_cpu(sbi->s_es->s_last_orphan));
836 printk(KERN_ERR "sb_info orphan list:\n");
837 list_for_each(l, &sbi->s_orphan) {
838 struct inode *inode = orphan_list_entry(l);
840 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
841 inode->i_sb->s_id, inode->i_ino, inode,
842 inode->i_mode, inode->i_nlink,
848 static int ext4_quota_off(struct super_block *sb, int type);
850 static inline void ext4_quota_off_umount(struct super_block *sb)
854 /* Use our quota_off function to clear inode flags etc. */
855 for (type = 0; type < EXT4_MAXQUOTAS; type++)
856 ext4_quota_off(sb, type);
859 static inline void ext4_quota_off_umount(struct super_block *sb)
864 static void ext4_put_super(struct super_block *sb)
866 struct ext4_sb_info *sbi = EXT4_SB(sb);
867 struct ext4_super_block *es = sbi->s_es;
871 ext4_unregister_li_request(sb);
872 ext4_quota_off_umount(sb);
874 flush_workqueue(sbi->rsv_conversion_wq);
875 destroy_workqueue(sbi->rsv_conversion_wq);
877 if (sbi->s_journal) {
878 aborted = is_journal_aborted(sbi->s_journal);
879 err = jbd2_journal_destroy(sbi->s_journal);
880 sbi->s_journal = NULL;
881 if ((err < 0) && !aborted)
882 ext4_abort(sb, "Couldn't clean up the journal");
885 ext4_unregister_sysfs(sb);
886 ext4_es_unregister_shrinker(sbi);
887 del_timer_sync(&sbi->s_err_report);
888 ext4_release_system_zone(sb);
890 ext4_ext_release(sb);
892 if (!(sb->s_flags & MS_RDONLY) && !aborted) {
893 ext4_clear_feature_journal_needs_recovery(sb);
894 es->s_state = cpu_to_le16(sbi->s_mount_state);
896 if (!(sb->s_flags & MS_RDONLY))
897 ext4_commit_super(sb, 1);
899 for (i = 0; i < sbi->s_gdb_count; i++)
900 brelse(sbi->s_group_desc[i]);
901 kvfree(sbi->s_group_desc);
902 kvfree(sbi->s_flex_groups);
903 percpu_counter_destroy(&sbi->s_freeclusters_counter);
904 percpu_counter_destroy(&sbi->s_freeinodes_counter);
905 percpu_counter_destroy(&sbi->s_dirs_counter);
906 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
907 percpu_free_rwsem(&sbi->s_journal_flag_rwsem);
909 for (i = 0; i < EXT4_MAXQUOTAS; i++)
910 kfree(sbi->s_qf_names[i]);
913 /* Debugging code just in case the in-memory inode orphan list
914 * isn't empty. The on-disk one can be non-empty if we've
915 * detected an error and taken the fs readonly, but the
916 * in-memory list had better be clean by this point. */
917 if (!list_empty(&sbi->s_orphan))
918 dump_orphan_list(sb, sbi);
919 J_ASSERT(list_empty(&sbi->s_orphan));
921 sync_blockdev(sb->s_bdev);
922 invalidate_bdev(sb->s_bdev);
923 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
925 * Invalidate the journal device's buffers. We don't want them
926 * floating about in memory - the physical journal device may
927 * hotswapped, and it breaks the `ro-after' testing code.
929 sync_blockdev(sbi->journal_bdev);
930 invalidate_bdev(sbi->journal_bdev);
931 ext4_blkdev_remove(sbi);
933 if (sbi->s_ea_inode_cache) {
934 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
935 sbi->s_ea_inode_cache = NULL;
937 if (sbi->s_ea_block_cache) {
938 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
939 sbi->s_ea_block_cache = NULL;
942 kthread_stop(sbi->s_mmp_tsk);
944 sb->s_fs_info = NULL;
946 * Now that we are completely done shutting down the
947 * superblock, we need to actually destroy the kobject.
949 kobject_put(&sbi->s_kobj);
950 wait_for_completion(&sbi->s_kobj_unregister);
951 if (sbi->s_chksum_driver)
952 crypto_free_shash(sbi->s_chksum_driver);
953 kfree(sbi->s_blockgroup_lock);
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 ei->vfs_inode.i_version = 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 res = ext4_convert_inline_data(inode);
1168 * If a journal handle was specified, then the encryption context is
1169 * being set on a new inode via inheritance and is part of a larger
1170 * transaction to create the inode. Otherwise the encryption context is
1171 * being set on an existing inode in its own transaction. Only in the
1172 * latter case should the "retry on ENOSPC" logic be used.
1176 res = ext4_xattr_set_handle(handle, inode,
1177 EXT4_XATTR_INDEX_ENCRYPTION,
1178 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1181 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1182 ext4_clear_inode_state(inode,
1183 EXT4_STATE_MAY_INLINE_DATA);
1185 * Update inode->i_flags - e.g. S_DAX may get disabled
1187 ext4_set_inode_flags(inode);
1192 res = dquot_initialize(inode);
1196 res = ext4_xattr_set_credits(inode, len, false /* is_create */,
1201 handle = ext4_journal_start(inode, EXT4_HT_MISC, credits);
1203 return PTR_ERR(handle);
1205 res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION,
1206 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1209 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1210 /* Update inode->i_flags - e.g. S_DAX may get disabled */
1211 ext4_set_inode_flags(inode);
1212 res = ext4_mark_inode_dirty(handle, inode);
1214 EXT4_ERROR_INODE(inode, "Failed to mark inode dirty");
1216 res2 = ext4_journal_stop(handle);
1218 if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
1225 static bool ext4_dummy_context(struct inode *inode)
1227 return DUMMY_ENCRYPTION_ENABLED(EXT4_SB(inode->i_sb));
1230 static unsigned ext4_max_namelen(struct inode *inode)
1232 return S_ISLNK(inode->i_mode) ? inode->i_sb->s_blocksize :
1236 static const struct fscrypt_operations ext4_cryptops = {
1237 .key_prefix = "ext4:",
1238 .get_context = ext4_get_context,
1239 .set_context = ext4_set_context,
1240 .dummy_context = ext4_dummy_context,
1241 .is_encrypted = ext4_encrypted_inode,
1242 .empty_dir = ext4_empty_dir,
1243 .max_namelen = ext4_max_namelen,
1246 static const struct fscrypt_operations ext4_cryptops = {
1247 .is_encrypted = ext4_encrypted_inode,
1252 static const char * const quotatypes[] = INITQFNAMES;
1253 #define QTYPE2NAME(t) (quotatypes[t])
1255 static int ext4_write_dquot(struct dquot *dquot);
1256 static int ext4_acquire_dquot(struct dquot *dquot);
1257 static int ext4_release_dquot(struct dquot *dquot);
1258 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1259 static int ext4_write_info(struct super_block *sb, int type);
1260 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1261 const struct path *path);
1262 static int ext4_quota_on_mount(struct super_block *sb, int type);
1263 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1264 size_t len, loff_t off);
1265 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1266 const char *data, size_t len, loff_t off);
1267 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1268 unsigned int flags);
1269 static int ext4_enable_quotas(struct super_block *sb);
1270 static int ext4_get_next_id(struct super_block *sb, struct kqid *qid);
1272 static struct dquot **ext4_get_dquots(struct inode *inode)
1274 return EXT4_I(inode)->i_dquot;
1277 static const struct dquot_operations ext4_quota_operations = {
1278 .get_reserved_space = ext4_get_reserved_space,
1279 .write_dquot = ext4_write_dquot,
1280 .acquire_dquot = ext4_acquire_dquot,
1281 .release_dquot = ext4_release_dquot,
1282 .mark_dirty = ext4_mark_dquot_dirty,
1283 .write_info = ext4_write_info,
1284 .alloc_dquot = dquot_alloc,
1285 .destroy_dquot = dquot_destroy,
1286 .get_projid = ext4_get_projid,
1287 .get_inode_usage = ext4_get_inode_usage,
1288 .get_next_id = ext4_get_next_id,
1291 static const struct quotactl_ops ext4_qctl_operations = {
1292 .quota_on = ext4_quota_on,
1293 .quota_off = ext4_quota_off,
1294 .quota_sync = dquot_quota_sync,
1295 .get_state = dquot_get_state,
1296 .set_info = dquot_set_dqinfo,
1297 .get_dqblk = dquot_get_dqblk,
1298 .set_dqblk = dquot_set_dqblk,
1299 .get_nextdqblk = dquot_get_next_dqblk,
1303 static const struct super_operations ext4_sops = {
1304 .alloc_inode = ext4_alloc_inode,
1305 .destroy_inode = ext4_destroy_inode,
1306 .write_inode = ext4_write_inode,
1307 .dirty_inode = ext4_dirty_inode,
1308 .drop_inode = ext4_drop_inode,
1309 .evict_inode = ext4_evict_inode,
1310 .put_super = ext4_put_super,
1311 .sync_fs = ext4_sync_fs,
1312 .freeze_fs = ext4_freeze,
1313 .unfreeze_fs = ext4_unfreeze,
1314 .statfs = ext4_statfs,
1315 .remount_fs = ext4_remount,
1316 .show_options = ext4_show_options,
1318 .quota_read = ext4_quota_read,
1319 .quota_write = ext4_quota_write,
1320 .get_dquots = ext4_get_dquots,
1322 .bdev_try_to_free_page = bdev_try_to_free_page,
1325 static const struct export_operations ext4_export_ops = {
1326 .fh_to_dentry = ext4_fh_to_dentry,
1327 .fh_to_parent = ext4_fh_to_parent,
1328 .get_parent = ext4_get_parent,
1332 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1333 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1334 Opt_nouid32, Opt_debug, Opt_removed,
1335 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1336 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1337 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1338 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1339 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1340 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1341 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1342 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1343 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1344 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version, Opt_dax,
1345 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1346 Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize,
1347 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1348 Opt_inode_readahead_blks, Opt_journal_ioprio,
1349 Opt_dioread_nolock, Opt_dioread_lock,
1350 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1351 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache,
1354 static const match_table_t tokens = {
1355 {Opt_bsd_df, "bsddf"},
1356 {Opt_minix_df, "minixdf"},
1357 {Opt_grpid, "grpid"},
1358 {Opt_grpid, "bsdgroups"},
1359 {Opt_nogrpid, "nogrpid"},
1360 {Opt_nogrpid, "sysvgroups"},
1361 {Opt_resgid, "resgid=%u"},
1362 {Opt_resuid, "resuid=%u"},
1364 {Opt_err_cont, "errors=continue"},
1365 {Opt_err_panic, "errors=panic"},
1366 {Opt_err_ro, "errors=remount-ro"},
1367 {Opt_nouid32, "nouid32"},
1368 {Opt_debug, "debug"},
1369 {Opt_removed, "oldalloc"},
1370 {Opt_removed, "orlov"},
1371 {Opt_user_xattr, "user_xattr"},
1372 {Opt_nouser_xattr, "nouser_xattr"},
1374 {Opt_noacl, "noacl"},
1375 {Opt_noload, "norecovery"},
1376 {Opt_noload, "noload"},
1377 {Opt_removed, "nobh"},
1378 {Opt_removed, "bh"},
1379 {Opt_commit, "commit=%u"},
1380 {Opt_min_batch_time, "min_batch_time=%u"},
1381 {Opt_max_batch_time, "max_batch_time=%u"},
1382 {Opt_journal_dev, "journal_dev=%u"},
1383 {Opt_journal_path, "journal_path=%s"},
1384 {Opt_journal_checksum, "journal_checksum"},
1385 {Opt_nojournal_checksum, "nojournal_checksum"},
1386 {Opt_journal_async_commit, "journal_async_commit"},
1387 {Opt_abort, "abort"},
1388 {Opt_data_journal, "data=journal"},
1389 {Opt_data_ordered, "data=ordered"},
1390 {Opt_data_writeback, "data=writeback"},
1391 {Opt_data_err_abort, "data_err=abort"},
1392 {Opt_data_err_ignore, "data_err=ignore"},
1393 {Opt_offusrjquota, "usrjquota="},
1394 {Opt_usrjquota, "usrjquota=%s"},
1395 {Opt_offgrpjquota, "grpjquota="},
1396 {Opt_grpjquota, "grpjquota=%s"},
1397 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1398 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1399 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1400 {Opt_grpquota, "grpquota"},
1401 {Opt_noquota, "noquota"},
1402 {Opt_quota, "quota"},
1403 {Opt_usrquota, "usrquota"},
1404 {Opt_prjquota, "prjquota"},
1405 {Opt_barrier, "barrier=%u"},
1406 {Opt_barrier, "barrier"},
1407 {Opt_nobarrier, "nobarrier"},
1408 {Opt_i_version, "i_version"},
1410 {Opt_stripe, "stripe=%u"},
1411 {Opt_delalloc, "delalloc"},
1412 {Opt_lazytime, "lazytime"},
1413 {Opt_nolazytime, "nolazytime"},
1414 {Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"},
1415 {Opt_nodelalloc, "nodelalloc"},
1416 {Opt_removed, "mblk_io_submit"},
1417 {Opt_removed, "nomblk_io_submit"},
1418 {Opt_block_validity, "block_validity"},
1419 {Opt_noblock_validity, "noblock_validity"},
1420 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1421 {Opt_journal_ioprio, "journal_ioprio=%u"},
1422 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1423 {Opt_auto_da_alloc, "auto_da_alloc"},
1424 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1425 {Opt_dioread_nolock, "dioread_nolock"},
1426 {Opt_dioread_lock, "dioread_lock"},
1427 {Opt_discard, "discard"},
1428 {Opt_nodiscard, "nodiscard"},
1429 {Opt_init_itable, "init_itable=%u"},
1430 {Opt_init_itable, "init_itable"},
1431 {Opt_noinit_itable, "noinit_itable"},
1432 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1433 {Opt_test_dummy_encryption, "test_dummy_encryption"},
1434 {Opt_nombcache, "nombcache"},
1435 {Opt_nombcache, "no_mbcache"}, /* for backward compatibility */
1436 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1437 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1438 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1439 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1440 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1444 static ext4_fsblk_t get_sb_block(void **data)
1446 ext4_fsblk_t sb_block;
1447 char *options = (char *) *data;
1449 if (!options || strncmp(options, "sb=", 3) != 0)
1450 return 1; /* Default location */
1453 /* TODO: use simple_strtoll with >32bit ext4 */
1454 sb_block = simple_strtoul(options, &options, 0);
1455 if (*options && *options != ',') {
1456 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1460 if (*options == ',')
1462 *data = (void *) options;
1467 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1468 static const char deprecated_msg[] =
1469 "Mount option \"%s\" will be removed by %s\n"
1470 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1473 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1475 struct ext4_sb_info *sbi = EXT4_SB(sb);
1479 if (sb_any_quota_loaded(sb) &&
1480 !sbi->s_qf_names[qtype]) {
1481 ext4_msg(sb, KERN_ERR,
1482 "Cannot change journaled "
1483 "quota options when quota turned on");
1486 if (ext4_has_feature_quota(sb)) {
1487 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1488 "ignored when QUOTA feature is enabled");
1491 qname = match_strdup(args);
1493 ext4_msg(sb, KERN_ERR,
1494 "Not enough memory for storing quotafile name");
1497 if (sbi->s_qf_names[qtype]) {
1498 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1501 ext4_msg(sb, KERN_ERR,
1502 "%s quota file already specified",
1506 if (strchr(qname, '/')) {
1507 ext4_msg(sb, KERN_ERR,
1508 "quotafile must be on filesystem root");
1511 sbi->s_qf_names[qtype] = qname;
1519 static int clear_qf_name(struct super_block *sb, int qtype)
1522 struct ext4_sb_info *sbi = EXT4_SB(sb);
1524 if (sb_any_quota_loaded(sb) &&
1525 sbi->s_qf_names[qtype]) {
1526 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1527 " when quota turned on");
1530 kfree(sbi->s_qf_names[qtype]);
1531 sbi->s_qf_names[qtype] = NULL;
1536 #define MOPT_SET 0x0001
1537 #define MOPT_CLEAR 0x0002
1538 #define MOPT_NOSUPPORT 0x0004
1539 #define MOPT_EXPLICIT 0x0008
1540 #define MOPT_CLEAR_ERR 0x0010
1541 #define MOPT_GTE0 0x0020
1544 #define MOPT_QFMT 0x0040
1546 #define MOPT_Q MOPT_NOSUPPORT
1547 #define MOPT_QFMT MOPT_NOSUPPORT
1549 #define MOPT_DATAJ 0x0080
1550 #define MOPT_NO_EXT2 0x0100
1551 #define MOPT_NO_EXT3 0x0200
1552 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1553 #define MOPT_STRING 0x0400
1555 static const struct mount_opts {
1559 } ext4_mount_opts[] = {
1560 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1561 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1562 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1563 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1564 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1565 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1566 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1567 MOPT_EXT4_ONLY | MOPT_SET},
1568 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1569 MOPT_EXT4_ONLY | MOPT_CLEAR},
1570 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1571 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1572 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1573 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1574 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1575 MOPT_EXT4_ONLY | MOPT_CLEAR},
1576 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1577 MOPT_EXT4_ONLY | MOPT_CLEAR},
1578 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1579 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1580 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1581 EXT4_MOUNT_JOURNAL_CHECKSUM),
1582 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1583 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1584 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1585 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1586 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1587 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1589 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1591 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1592 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1593 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1594 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1595 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1596 {Opt_commit, 0, MOPT_GTE0},
1597 {Opt_max_batch_time, 0, MOPT_GTE0},
1598 {Opt_min_batch_time, 0, MOPT_GTE0},
1599 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1600 {Opt_init_itable, 0, MOPT_GTE0},
1601 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
1602 {Opt_stripe, 0, MOPT_GTE0},
1603 {Opt_resuid, 0, MOPT_GTE0},
1604 {Opt_resgid, 0, MOPT_GTE0},
1605 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1606 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1607 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1608 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1609 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1610 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1611 MOPT_NO_EXT2 | MOPT_DATAJ},
1612 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1613 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1614 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1615 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1616 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1618 {Opt_acl, 0, MOPT_NOSUPPORT},
1619 {Opt_noacl, 0, MOPT_NOSUPPORT},
1621 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1622 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1623 {Opt_debug_want_extra_isize, 0, MOPT_GTE0},
1624 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1625 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1627 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1629 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1631 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1632 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1633 MOPT_CLEAR | MOPT_Q},
1634 {Opt_usrjquota, 0, MOPT_Q},
1635 {Opt_grpjquota, 0, MOPT_Q},
1636 {Opt_offusrjquota, 0, MOPT_Q},
1637 {Opt_offgrpjquota, 0, MOPT_Q},
1638 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1639 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1640 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1641 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1642 {Opt_test_dummy_encryption, 0, MOPT_GTE0},
1643 {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET},
1647 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1648 substring_t *args, unsigned long *journal_devnum,
1649 unsigned int *journal_ioprio, int is_remount)
1651 struct ext4_sb_info *sbi = EXT4_SB(sb);
1652 const struct mount_opts *m;
1658 if (token == Opt_usrjquota)
1659 return set_qf_name(sb, USRQUOTA, &args[0]);
1660 else if (token == Opt_grpjquota)
1661 return set_qf_name(sb, GRPQUOTA, &args[0]);
1662 else if (token == Opt_offusrjquota)
1663 return clear_qf_name(sb, USRQUOTA);
1664 else if (token == Opt_offgrpjquota)
1665 return clear_qf_name(sb, GRPQUOTA);
1669 case Opt_nouser_xattr:
1670 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1673 return 1; /* handled by get_sb_block() */
1675 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1678 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1681 sb->s_flags |= MS_I_VERSION;
1684 sb->s_flags |= MS_LAZYTIME;
1686 case Opt_nolazytime:
1687 sb->s_flags &= ~MS_LAZYTIME;
1691 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1692 if (token == m->token)
1695 if (m->token == Opt_err) {
1696 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1697 "or missing value", opt);
1701 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1702 ext4_msg(sb, KERN_ERR,
1703 "Mount option \"%s\" incompatible with ext2", opt);
1706 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1707 ext4_msg(sb, KERN_ERR,
1708 "Mount option \"%s\" incompatible with ext3", opt);
1712 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
1714 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1716 if (m->flags & MOPT_EXPLICIT) {
1717 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
1718 set_opt2(sb, EXPLICIT_DELALLOC);
1719 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
1720 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
1724 if (m->flags & MOPT_CLEAR_ERR)
1725 clear_opt(sb, ERRORS_MASK);
1726 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1727 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1728 "options when quota turned on");
1732 if (m->flags & MOPT_NOSUPPORT) {
1733 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1734 } else if (token == Opt_commit) {
1736 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1737 sbi->s_commit_interval = HZ * arg;
1738 } else if (token == Opt_debug_want_extra_isize) {
1739 sbi->s_want_extra_isize = arg;
1740 } else if (token == Opt_max_batch_time) {
1741 sbi->s_max_batch_time = arg;
1742 } else if (token == Opt_min_batch_time) {
1743 sbi->s_min_batch_time = arg;
1744 } else if (token == Opt_inode_readahead_blks) {
1745 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1746 ext4_msg(sb, KERN_ERR,
1747 "EXT4-fs: inode_readahead_blks must be "
1748 "0 or a power of 2 smaller than 2^31");
1751 sbi->s_inode_readahead_blks = arg;
1752 } else if (token == Opt_init_itable) {
1753 set_opt(sb, INIT_INODE_TABLE);
1755 arg = EXT4_DEF_LI_WAIT_MULT;
1756 sbi->s_li_wait_mult = arg;
1757 } else if (token == Opt_max_dir_size_kb) {
1758 sbi->s_max_dir_size_kb = arg;
1759 } else if (token == Opt_stripe) {
1760 sbi->s_stripe = arg;
1761 } else if (token == Opt_resuid) {
1762 uid = make_kuid(current_user_ns(), arg);
1763 if (!uid_valid(uid)) {
1764 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1767 sbi->s_resuid = uid;
1768 } else if (token == Opt_resgid) {
1769 gid = make_kgid(current_user_ns(), arg);
1770 if (!gid_valid(gid)) {
1771 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1774 sbi->s_resgid = gid;
1775 } else if (token == Opt_journal_dev) {
1777 ext4_msg(sb, KERN_ERR,
1778 "Cannot specify journal on remount");
1781 *journal_devnum = arg;
1782 } else if (token == Opt_journal_path) {
1784 struct inode *journal_inode;
1789 ext4_msg(sb, KERN_ERR,
1790 "Cannot specify journal on remount");
1793 journal_path = match_strdup(&args[0]);
1794 if (!journal_path) {
1795 ext4_msg(sb, KERN_ERR, "error: could not dup "
1796 "journal device string");
1800 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1802 ext4_msg(sb, KERN_ERR, "error: could not find "
1803 "journal device path: error %d", error);
1804 kfree(journal_path);
1808 journal_inode = d_inode(path.dentry);
1809 if (!S_ISBLK(journal_inode->i_mode)) {
1810 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1811 "is not a block device", journal_path);
1813 kfree(journal_path);
1817 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1819 kfree(journal_path);
1820 } else if (token == Opt_journal_ioprio) {
1822 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1827 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1828 } else if (token == Opt_test_dummy_encryption) {
1829 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1830 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1831 ext4_msg(sb, KERN_WARNING,
1832 "Test dummy encryption mode enabled");
1834 ext4_msg(sb, KERN_WARNING,
1835 "Test dummy encryption mount option ignored");
1837 } else if (m->flags & MOPT_DATAJ) {
1839 if (!sbi->s_journal)
1840 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1841 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1842 ext4_msg(sb, KERN_ERR,
1843 "Cannot change data mode on remount");
1847 clear_opt(sb, DATA_FLAGS);
1848 sbi->s_mount_opt |= m->mount_opt;
1851 } else if (m->flags & MOPT_QFMT) {
1852 if (sb_any_quota_loaded(sb) &&
1853 sbi->s_jquota_fmt != m->mount_opt) {
1854 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1855 "quota options when quota turned on");
1858 if (ext4_has_feature_quota(sb)) {
1859 ext4_msg(sb, KERN_INFO,
1860 "Quota format mount options ignored "
1861 "when QUOTA feature is enabled");
1864 sbi->s_jquota_fmt = m->mount_opt;
1866 } else if (token == Opt_dax) {
1867 #ifdef CONFIG_FS_DAX
1868 ext4_msg(sb, KERN_WARNING,
1869 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
1870 sbi->s_mount_opt |= m->mount_opt;
1872 ext4_msg(sb, KERN_INFO, "dax option not supported");
1875 } else if (token == Opt_data_err_abort) {
1876 sbi->s_mount_opt |= m->mount_opt;
1877 } else if (token == Opt_data_err_ignore) {
1878 sbi->s_mount_opt &= ~m->mount_opt;
1882 if (m->flags & MOPT_CLEAR)
1884 else if (unlikely(!(m->flags & MOPT_SET))) {
1885 ext4_msg(sb, KERN_WARNING,
1886 "buggy handling of option %s", opt);
1891 sbi->s_mount_opt |= m->mount_opt;
1893 sbi->s_mount_opt &= ~m->mount_opt;
1898 static int parse_options(char *options, struct super_block *sb,
1899 unsigned long *journal_devnum,
1900 unsigned int *journal_ioprio,
1903 struct ext4_sb_info *sbi = EXT4_SB(sb);
1905 substring_t args[MAX_OPT_ARGS];
1911 while ((p = strsep(&options, ",")) != NULL) {
1915 * Initialize args struct so we know whether arg was
1916 * found; some options take optional arguments.
1918 args[0].to = args[0].from = NULL;
1919 token = match_token(p, tokens, args);
1920 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1921 journal_ioprio, is_remount) < 0)
1926 * We do the test below only for project quotas. 'usrquota' and
1927 * 'grpquota' mount options are allowed even without quota feature
1928 * to support legacy quotas in quota files.
1930 if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) {
1931 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. "
1932 "Cannot enable project quota enforcement.");
1935 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1936 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1937 clear_opt(sb, USRQUOTA);
1939 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1940 clear_opt(sb, GRPQUOTA);
1942 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1943 ext4_msg(sb, KERN_ERR, "old and new quota "
1948 if (!sbi->s_jquota_fmt) {
1949 ext4_msg(sb, KERN_ERR, "journaled quota format "
1955 if (test_opt(sb, DIOREAD_NOLOCK)) {
1957 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1959 if (blocksize < PAGE_SIZE) {
1960 ext4_msg(sb, KERN_ERR, "can't mount with "
1961 "dioread_nolock if block size != PAGE_SIZE");
1968 static inline void ext4_show_quota_options(struct seq_file *seq,
1969 struct super_block *sb)
1971 #if defined(CONFIG_QUOTA)
1972 struct ext4_sb_info *sbi = EXT4_SB(sb);
1974 if (sbi->s_jquota_fmt) {
1977 switch (sbi->s_jquota_fmt) {
1988 seq_printf(seq, ",jqfmt=%s", fmtname);
1991 if (sbi->s_qf_names[USRQUOTA])
1992 seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]);
1994 if (sbi->s_qf_names[GRPQUOTA])
1995 seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]);
1999 static const char *token2str(int token)
2001 const struct match_token *t;
2003 for (t = tokens; t->token != Opt_err; t++)
2004 if (t->token == token && !strchr(t->pattern, '='))
2011 * - it's set to a non-default value OR
2012 * - if the per-sb default is different from the global default
2014 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
2017 struct ext4_sb_info *sbi = EXT4_SB(sb);
2018 struct ext4_super_block *es = sbi->s_es;
2019 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
2020 const struct mount_opts *m;
2021 char sep = nodefs ? '\n' : ',';
2023 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2024 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2026 if (sbi->s_sb_block != 1)
2027 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
2029 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
2030 int want_set = m->flags & MOPT_SET;
2031 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
2032 (m->flags & MOPT_CLEAR_ERR))
2034 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
2035 continue; /* skip if same as the default */
2037 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
2038 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
2039 continue; /* select Opt_noFoo vs Opt_Foo */
2040 SEQ_OPTS_PRINT("%s", token2str(m->token));
2043 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
2044 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
2045 SEQ_OPTS_PRINT("resuid=%u",
2046 from_kuid_munged(&init_user_ns, sbi->s_resuid));
2047 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
2048 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
2049 SEQ_OPTS_PRINT("resgid=%u",
2050 from_kgid_munged(&init_user_ns, sbi->s_resgid));
2051 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
2052 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
2053 SEQ_OPTS_PUTS("errors=remount-ro");
2054 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
2055 SEQ_OPTS_PUTS("errors=continue");
2056 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
2057 SEQ_OPTS_PUTS("errors=panic");
2058 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
2059 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
2060 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
2061 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
2062 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
2063 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2064 if (sb->s_flags & MS_I_VERSION)
2065 SEQ_OPTS_PUTS("i_version");
2066 if (nodefs || sbi->s_stripe)
2067 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
2068 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
2069 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2070 SEQ_OPTS_PUTS("data=journal");
2071 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2072 SEQ_OPTS_PUTS("data=ordered");
2073 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2074 SEQ_OPTS_PUTS("data=writeback");
2077 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
2078 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2079 sbi->s_inode_readahead_blks);
2081 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
2082 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
2083 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
2084 if (nodefs || sbi->s_max_dir_size_kb)
2085 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2086 if (test_opt(sb, DATA_ERR_ABORT))
2087 SEQ_OPTS_PUTS("data_err=abort");
2089 ext4_show_quota_options(seq, sb);
2093 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2095 return _ext4_show_options(seq, root->d_sb, 0);
2098 int ext4_seq_options_show(struct seq_file *seq, void *offset)
2100 struct super_block *sb = seq->private;
2103 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
2104 rc = _ext4_show_options(seq, sb, 1);
2105 seq_puts(seq, "\n");
2109 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
2112 struct ext4_sb_info *sbi = EXT4_SB(sb);
2115 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
2116 ext4_msg(sb, KERN_ERR, "revision level too high, "
2117 "forcing read-only mode");
2122 if (!(sbi->s_mount_state & EXT4_VALID_FS))
2123 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
2124 "running e2fsck is recommended");
2125 else if (sbi->s_mount_state & EXT4_ERROR_FS)
2126 ext4_msg(sb, KERN_WARNING,
2127 "warning: mounting fs with errors, "
2128 "running e2fsck is recommended");
2129 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
2130 le16_to_cpu(es->s_mnt_count) >=
2131 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
2132 ext4_msg(sb, KERN_WARNING,
2133 "warning: maximal mount count reached, "
2134 "running e2fsck is recommended");
2135 else if (le32_to_cpu(es->s_checkinterval) &&
2136 (le32_to_cpu(es->s_lastcheck) +
2137 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
2138 ext4_msg(sb, KERN_WARNING,
2139 "warning: checktime reached, "
2140 "running e2fsck is recommended");
2141 if (!sbi->s_journal)
2142 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
2143 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
2144 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
2145 le16_add_cpu(&es->s_mnt_count, 1);
2146 es->s_mtime = cpu_to_le32(get_seconds());
2147 ext4_update_dynamic_rev(sb);
2149 ext4_set_feature_journal_needs_recovery(sb);
2151 ext4_commit_super(sb, 1);
2153 if (test_opt(sb, DEBUG))
2154 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
2155 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
2157 sbi->s_groups_count,
2158 EXT4_BLOCKS_PER_GROUP(sb),
2159 EXT4_INODES_PER_GROUP(sb),
2160 sbi->s_mount_opt, sbi->s_mount_opt2);
2162 cleancache_init_fs(sb);
2166 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2168 struct ext4_sb_info *sbi = EXT4_SB(sb);
2169 struct flex_groups *new_groups;
2172 if (!sbi->s_log_groups_per_flex)
2175 size = ext4_flex_group(sbi, ngroup - 1) + 1;
2176 if (size <= sbi->s_flex_groups_allocated)
2179 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
2180 new_groups = kvzalloc(size, GFP_KERNEL);
2182 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
2183 size / (int) sizeof(struct flex_groups));
2187 if (sbi->s_flex_groups) {
2188 memcpy(new_groups, sbi->s_flex_groups,
2189 (sbi->s_flex_groups_allocated *
2190 sizeof(struct flex_groups)));
2191 kvfree(sbi->s_flex_groups);
2193 sbi->s_flex_groups = new_groups;
2194 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
2198 static int ext4_fill_flex_info(struct super_block *sb)
2200 struct ext4_sb_info *sbi = EXT4_SB(sb);
2201 struct ext4_group_desc *gdp = NULL;
2202 ext4_group_t flex_group;
2205 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2206 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
2207 sbi->s_log_groups_per_flex = 0;
2211 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2215 for (i = 0; i < sbi->s_groups_count; i++) {
2216 gdp = ext4_get_group_desc(sb, i, NULL);
2218 flex_group = ext4_flex_group(sbi, i);
2219 atomic_add(ext4_free_inodes_count(sb, gdp),
2220 &sbi->s_flex_groups[flex_group].free_inodes);
2221 atomic64_add(ext4_free_group_clusters(sb, gdp),
2222 &sbi->s_flex_groups[flex_group].free_clusters);
2223 atomic_add(ext4_used_dirs_count(sb, gdp),
2224 &sbi->s_flex_groups[flex_group].used_dirs);
2232 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
2233 struct ext4_group_desc *gdp)
2235 int offset = offsetof(struct ext4_group_desc, bg_checksum);
2237 __le32 le_group = cpu_to_le32(block_group);
2238 struct ext4_sb_info *sbi = EXT4_SB(sb);
2240 if (ext4_has_metadata_csum(sbi->s_sb)) {
2241 /* Use new metadata_csum algorithm */
2243 __u16 dummy_csum = 0;
2245 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2247 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2248 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2249 sizeof(dummy_csum));
2250 offset += sizeof(dummy_csum);
2251 if (offset < sbi->s_desc_size)
2252 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2253 sbi->s_desc_size - offset);
2255 crc = csum32 & 0xFFFF;
2259 /* old crc16 code */
2260 if (!ext4_has_feature_gdt_csum(sb))
2263 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2264 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2265 crc = crc16(crc, (__u8 *)gdp, offset);
2266 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2267 /* for checksum of struct ext4_group_desc do the rest...*/
2268 if (ext4_has_feature_64bit(sb) &&
2269 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2270 crc = crc16(crc, (__u8 *)gdp + offset,
2271 le16_to_cpu(sbi->s_es->s_desc_size) -
2275 return cpu_to_le16(crc);
2278 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2279 struct ext4_group_desc *gdp)
2281 if (ext4_has_group_desc_csum(sb) &&
2282 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
2288 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2289 struct ext4_group_desc *gdp)
2291 if (!ext4_has_group_desc_csum(sb))
2293 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
2296 /* Called at mount-time, super-block is locked */
2297 static int ext4_check_descriptors(struct super_block *sb,
2298 ext4_fsblk_t sb_block,
2299 ext4_group_t *first_not_zeroed)
2301 struct ext4_sb_info *sbi = EXT4_SB(sb);
2302 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2303 ext4_fsblk_t last_block;
2304 ext4_fsblk_t block_bitmap;
2305 ext4_fsblk_t inode_bitmap;
2306 ext4_fsblk_t inode_table;
2307 int flexbg_flag = 0;
2308 ext4_group_t i, grp = sbi->s_groups_count;
2310 if (ext4_has_feature_flex_bg(sb))
2313 ext4_debug("Checking group descriptors");
2315 for (i = 0; i < sbi->s_groups_count; i++) {
2316 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2318 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2319 last_block = ext4_blocks_count(sbi->s_es) - 1;
2321 last_block = first_block +
2322 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2324 if ((grp == sbi->s_groups_count) &&
2325 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2328 block_bitmap = ext4_block_bitmap(sb, gdp);
2329 if (block_bitmap == sb_block) {
2330 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2331 "Block bitmap for group %u overlaps "
2334 if (block_bitmap < first_block || block_bitmap > last_block) {
2335 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2336 "Block bitmap for group %u not in group "
2337 "(block %llu)!", i, block_bitmap);
2340 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2341 if (inode_bitmap == sb_block) {
2342 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2343 "Inode bitmap for group %u overlaps "
2346 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2347 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2348 "Inode bitmap for group %u not in group "
2349 "(block %llu)!", i, inode_bitmap);
2352 inode_table = ext4_inode_table(sb, gdp);
2353 if (inode_table == sb_block) {
2354 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2355 "Inode table for group %u overlaps "
2358 if (inode_table < first_block ||
2359 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2360 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2361 "Inode table for group %u not in group "
2362 "(block %llu)!", i, inode_table);
2365 ext4_lock_group(sb, i);
2366 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2367 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2368 "Checksum for group %u failed (%u!=%u)",
2369 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
2370 gdp)), le16_to_cpu(gdp->bg_checksum));
2371 if (!(sb->s_flags & MS_RDONLY)) {
2372 ext4_unlock_group(sb, i);
2376 ext4_unlock_group(sb, i);
2378 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2380 if (NULL != first_not_zeroed)
2381 *first_not_zeroed = grp;
2385 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2386 * the superblock) which were deleted from all directories, but held open by
2387 * a process at the time of a crash. We walk the list and try to delete these
2388 * inodes at recovery time (only with a read-write filesystem).
2390 * In order to keep the orphan inode chain consistent during traversal (in
2391 * case of crash during recovery), we link each inode into the superblock
2392 * orphan list_head and handle it the same way as an inode deletion during
2393 * normal operation (which journals the operations for us).
2395 * We only do an iget() and an iput() on each inode, which is very safe if we
2396 * accidentally point at an in-use or already deleted inode. The worst that
2397 * can happen in this case is that we get a "bit already cleared" message from
2398 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2399 * e2fsck was run on this filesystem, and it must have already done the orphan
2400 * inode cleanup for us, so we can safely abort without any further action.
2402 static void ext4_orphan_cleanup(struct super_block *sb,
2403 struct ext4_super_block *es)
2405 unsigned int s_flags = sb->s_flags;
2406 int ret, nr_orphans = 0, nr_truncates = 0;
2408 int quota_update = 0;
2411 if (!es->s_last_orphan) {
2412 jbd_debug(4, "no orphan inodes to clean up\n");
2416 if (bdev_read_only(sb->s_bdev)) {
2417 ext4_msg(sb, KERN_ERR, "write access "
2418 "unavailable, skipping orphan cleanup");
2422 /* Check if feature set would not allow a r/w mount */
2423 if (!ext4_feature_set_ok(sb, 0)) {
2424 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2425 "unknown ROCOMPAT features");
2429 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2430 /* don't clear list on RO mount w/ errors */
2431 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
2432 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
2433 "clearing orphan list.\n");
2434 es->s_last_orphan = 0;
2436 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2440 if (s_flags & MS_RDONLY) {
2441 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2442 sb->s_flags &= ~MS_RDONLY;
2445 /* Needed for iput() to work correctly and not trash data */
2446 sb->s_flags |= MS_ACTIVE;
2449 * Turn on quotas which were not enabled for read-only mounts if
2450 * filesystem has quota feature, so that they are updated correctly.
2452 if (ext4_has_feature_quota(sb) && (s_flags & MS_RDONLY)) {
2453 int ret = ext4_enable_quotas(sb);
2458 ext4_msg(sb, KERN_ERR,
2459 "Cannot turn on quotas: error %d", ret);
2462 /* Turn on journaled quotas used for old sytle */
2463 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2464 if (EXT4_SB(sb)->s_qf_names[i]) {
2465 int ret = ext4_quota_on_mount(sb, i);
2470 ext4_msg(sb, KERN_ERR,
2471 "Cannot turn on journaled "
2472 "quota: type %d: error %d", i, ret);
2477 while (es->s_last_orphan) {
2478 struct inode *inode;
2481 * We may have encountered an error during cleanup; if
2482 * so, skip the rest.
2484 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2485 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2486 es->s_last_orphan = 0;
2490 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2491 if (IS_ERR(inode)) {
2492 es->s_last_orphan = 0;
2496 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2497 dquot_initialize(inode);
2498 if (inode->i_nlink) {
2499 if (test_opt(sb, DEBUG))
2500 ext4_msg(sb, KERN_DEBUG,
2501 "%s: truncating inode %lu to %lld bytes",
2502 __func__, inode->i_ino, inode->i_size);
2503 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2504 inode->i_ino, inode->i_size);
2506 truncate_inode_pages(inode->i_mapping, inode->i_size);
2507 ret = ext4_truncate(inode);
2509 ext4_std_error(inode->i_sb, ret);
2510 inode_unlock(inode);
2513 if (test_opt(sb, DEBUG))
2514 ext4_msg(sb, KERN_DEBUG,
2515 "%s: deleting unreferenced inode %lu",
2516 __func__, inode->i_ino);
2517 jbd_debug(2, "deleting unreferenced inode %lu\n",
2521 iput(inode); /* The delete magic happens here! */
2524 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2527 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2528 PLURAL(nr_orphans));
2530 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2531 PLURAL(nr_truncates));
2533 /* Turn off quotas if they were enabled for orphan cleanup */
2535 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2536 if (sb_dqopt(sb)->files[i])
2537 dquot_quota_off(sb, i);
2541 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2545 * Maximal extent format file size.
2546 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2547 * extent format containers, within a sector_t, and within i_blocks
2548 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2549 * so that won't be a limiting factor.
2551 * However there is other limiting factor. We do store extents in the form
2552 * of starting block and length, hence the resulting length of the extent
2553 * covering maximum file size must fit into on-disk format containers as
2554 * well. Given that length is always by 1 unit bigger than max unit (because
2555 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2557 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2559 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2562 loff_t upper_limit = MAX_LFS_FILESIZE;
2564 /* small i_blocks in vfs inode? */
2565 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2567 * CONFIG_LBDAF is not enabled implies the inode
2568 * i_block represent total blocks in 512 bytes
2569 * 32 == size of vfs inode i_blocks * 8
2571 upper_limit = (1LL << 32) - 1;
2573 /* total blocks in file system block size */
2574 upper_limit >>= (blkbits - 9);
2575 upper_limit <<= blkbits;
2579 * 32-bit extent-start container, ee_block. We lower the maxbytes
2580 * by one fs block, so ee_len can cover the extent of maximum file
2583 res = (1LL << 32) - 1;
2586 /* Sanity check against vm- & vfs- imposed limits */
2587 if (res > upper_limit)
2594 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2595 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2596 * We need to be 1 filesystem block less than the 2^48 sector limit.
2598 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2600 loff_t res = EXT4_NDIR_BLOCKS;
2603 /* This is calculated to be the largest file size for a dense, block
2604 * mapped file such that the file's total number of 512-byte sectors,
2605 * including data and all indirect blocks, does not exceed (2^48 - 1).
2607 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2608 * number of 512-byte sectors of the file.
2611 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2613 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2614 * the inode i_block field represents total file blocks in
2615 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2617 upper_limit = (1LL << 32) - 1;
2619 /* total blocks in file system block size */
2620 upper_limit >>= (bits - 9);
2624 * We use 48 bit ext4_inode i_blocks
2625 * With EXT4_HUGE_FILE_FL set the i_blocks
2626 * represent total number of blocks in
2627 * file system block size
2629 upper_limit = (1LL << 48) - 1;
2633 /* indirect blocks */
2635 /* double indirect blocks */
2636 meta_blocks += 1 + (1LL << (bits-2));
2637 /* tripple indirect blocks */
2638 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2640 upper_limit -= meta_blocks;
2641 upper_limit <<= bits;
2643 res += 1LL << (bits-2);
2644 res += 1LL << (2*(bits-2));
2645 res += 1LL << (3*(bits-2));
2647 if (res > upper_limit)
2650 if (res > MAX_LFS_FILESIZE)
2651 res = MAX_LFS_FILESIZE;
2656 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2657 ext4_fsblk_t logical_sb_block, int nr)
2659 struct ext4_sb_info *sbi = EXT4_SB(sb);
2660 ext4_group_t bg, first_meta_bg;
2663 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2665 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
2666 return logical_sb_block + nr + 1;
2667 bg = sbi->s_desc_per_block * nr;
2668 if (ext4_bg_has_super(sb, bg))
2672 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2673 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2674 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2677 if (sb->s_blocksize == 1024 && nr == 0 &&
2678 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2681 return (has_super + ext4_group_first_block_no(sb, bg));
2685 * ext4_get_stripe_size: Get the stripe size.
2686 * @sbi: In memory super block info
2688 * If we have specified it via mount option, then
2689 * use the mount option value. If the value specified at mount time is
2690 * greater than the blocks per group use the super block value.
2691 * If the super block value is greater than blocks per group return 0.
2692 * Allocator needs it be less than blocks per group.
2695 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2697 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2698 unsigned long stripe_width =
2699 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2702 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2703 ret = sbi->s_stripe;
2704 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
2706 else if (stride && stride <= sbi->s_blocks_per_group)
2712 * If the stripe width is 1, this makes no sense and
2713 * we set it to 0 to turn off stripe handling code.
2722 * Check whether this filesystem can be mounted based on
2723 * the features present and the RDONLY/RDWR mount requested.
2724 * Returns 1 if this filesystem can be mounted as requested,
2725 * 0 if it cannot be.
2727 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2729 if (ext4_has_unknown_ext4_incompat_features(sb)) {
2730 ext4_msg(sb, KERN_ERR,
2731 "Couldn't mount because of "
2732 "unsupported optional features (%x)",
2733 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2734 ~EXT4_FEATURE_INCOMPAT_SUPP));
2741 if (ext4_has_feature_readonly(sb)) {
2742 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2743 sb->s_flags |= MS_RDONLY;
2747 /* Check that feature set is OK for a read-write mount */
2748 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
2749 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2750 "unsupported optional features (%x)",
2751 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2752 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2756 * Large file size enabled file system can only be mounted
2757 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2759 if (ext4_has_feature_huge_file(sb)) {
2760 if (sizeof(blkcnt_t) < sizeof(u64)) {
2761 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2762 "cannot be mounted RDWR without "
2767 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
2768 ext4_msg(sb, KERN_ERR,
2769 "Can't support bigalloc feature without "
2770 "extents feature\n");
2774 #ifndef CONFIG_QUOTA
2775 if (ext4_has_feature_quota(sb) && !readonly) {
2776 ext4_msg(sb, KERN_ERR,
2777 "Filesystem with quota feature cannot be mounted RDWR "
2778 "without CONFIG_QUOTA");
2781 if (ext4_has_feature_project(sb) && !readonly) {
2782 ext4_msg(sb, KERN_ERR,
2783 "Filesystem with project quota feature cannot be mounted RDWR "
2784 "without CONFIG_QUOTA");
2787 #endif /* CONFIG_QUOTA */
2792 * This function is called once a day if we have errors logged
2793 * on the file system
2795 static void print_daily_error_info(unsigned long arg)
2797 struct super_block *sb = (struct super_block *) arg;
2798 struct ext4_sb_info *sbi;
2799 struct ext4_super_block *es;
2804 if (es->s_error_count)
2805 /* fsck newer than v1.41.13 is needed to clean this condition. */
2806 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
2807 le32_to_cpu(es->s_error_count));
2808 if (es->s_first_error_time) {
2809 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
2810 sb->s_id, le32_to_cpu(es->s_first_error_time),
2811 (int) sizeof(es->s_first_error_func),
2812 es->s_first_error_func,
2813 le32_to_cpu(es->s_first_error_line));
2814 if (es->s_first_error_ino)
2815 printk(KERN_CONT ": inode %u",
2816 le32_to_cpu(es->s_first_error_ino));
2817 if (es->s_first_error_block)
2818 printk(KERN_CONT ": block %llu", (unsigned long long)
2819 le64_to_cpu(es->s_first_error_block));
2820 printk(KERN_CONT "\n");
2822 if (es->s_last_error_time) {
2823 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
2824 sb->s_id, le32_to_cpu(es->s_last_error_time),
2825 (int) sizeof(es->s_last_error_func),
2826 es->s_last_error_func,
2827 le32_to_cpu(es->s_last_error_line));
2828 if (es->s_last_error_ino)
2829 printk(KERN_CONT ": inode %u",
2830 le32_to_cpu(es->s_last_error_ino));
2831 if (es->s_last_error_block)
2832 printk(KERN_CONT ": block %llu", (unsigned long long)
2833 le64_to_cpu(es->s_last_error_block));
2834 printk(KERN_CONT "\n");
2836 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2839 /* Find next suitable group and run ext4_init_inode_table */
2840 static int ext4_run_li_request(struct ext4_li_request *elr)
2842 struct ext4_group_desc *gdp = NULL;
2843 ext4_group_t group, ngroups;
2844 struct super_block *sb;
2845 unsigned long timeout = 0;
2849 ngroups = EXT4_SB(sb)->s_groups_count;
2851 for (group = elr->lr_next_group; group < ngroups; group++) {
2852 gdp = ext4_get_group_desc(sb, group, NULL);
2858 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2862 if (group >= ngroups)
2867 ret = ext4_init_inode_table(sb, group,
2868 elr->lr_timeout ? 0 : 1);
2869 if (elr->lr_timeout == 0) {
2870 timeout = (jiffies - timeout) *
2871 elr->lr_sbi->s_li_wait_mult;
2872 elr->lr_timeout = timeout;
2874 elr->lr_next_sched = jiffies + elr->lr_timeout;
2875 elr->lr_next_group = group + 1;
2881 * Remove lr_request from the list_request and free the
2882 * request structure. Should be called with li_list_mtx held
2884 static void ext4_remove_li_request(struct ext4_li_request *elr)
2886 struct ext4_sb_info *sbi;
2893 list_del(&elr->lr_request);
2894 sbi->s_li_request = NULL;
2898 static void ext4_unregister_li_request(struct super_block *sb)
2900 mutex_lock(&ext4_li_mtx);
2901 if (!ext4_li_info) {
2902 mutex_unlock(&ext4_li_mtx);
2906 mutex_lock(&ext4_li_info->li_list_mtx);
2907 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2908 mutex_unlock(&ext4_li_info->li_list_mtx);
2909 mutex_unlock(&ext4_li_mtx);
2912 static struct task_struct *ext4_lazyinit_task;
2915 * This is the function where ext4lazyinit thread lives. It walks
2916 * through the request list searching for next scheduled filesystem.
2917 * When such a fs is found, run the lazy initialization request
2918 * (ext4_rn_li_request) and keep track of the time spend in this
2919 * function. Based on that time we compute next schedule time of
2920 * the request. When walking through the list is complete, compute
2921 * next waking time and put itself into sleep.
2923 static int ext4_lazyinit_thread(void *arg)
2925 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2926 struct list_head *pos, *n;
2927 struct ext4_li_request *elr;
2928 unsigned long next_wakeup, cur;
2930 BUG_ON(NULL == eli);
2934 next_wakeup = MAX_JIFFY_OFFSET;
2936 mutex_lock(&eli->li_list_mtx);
2937 if (list_empty(&eli->li_request_list)) {
2938 mutex_unlock(&eli->li_list_mtx);
2941 list_for_each_safe(pos, n, &eli->li_request_list) {
2944 elr = list_entry(pos, struct ext4_li_request,
2947 if (time_before(jiffies, elr->lr_next_sched)) {
2948 if (time_before(elr->lr_next_sched, next_wakeup))
2949 next_wakeup = elr->lr_next_sched;
2952 if (down_read_trylock(&elr->lr_super->s_umount)) {
2953 if (sb_start_write_trylock(elr->lr_super)) {
2956 * We hold sb->s_umount, sb can not
2957 * be removed from the list, it is
2958 * now safe to drop li_list_mtx
2960 mutex_unlock(&eli->li_list_mtx);
2961 err = ext4_run_li_request(elr);
2962 sb_end_write(elr->lr_super);
2963 mutex_lock(&eli->li_list_mtx);
2966 up_read((&elr->lr_super->s_umount));
2968 /* error, remove the lazy_init job */
2970 ext4_remove_li_request(elr);
2974 elr->lr_next_sched = jiffies +
2976 % (EXT4_DEF_LI_MAX_START_DELAY * HZ));
2978 if (time_before(elr->lr_next_sched, next_wakeup))
2979 next_wakeup = elr->lr_next_sched;
2981 mutex_unlock(&eli->li_list_mtx);
2986 if ((time_after_eq(cur, next_wakeup)) ||
2987 (MAX_JIFFY_OFFSET == next_wakeup)) {
2992 schedule_timeout_interruptible(next_wakeup - cur);
2994 if (kthread_should_stop()) {
2995 ext4_clear_request_list();
3002 * It looks like the request list is empty, but we need
3003 * to check it under the li_list_mtx lock, to prevent any
3004 * additions into it, and of course we should lock ext4_li_mtx
3005 * to atomically free the list and ext4_li_info, because at
3006 * this point another ext4 filesystem could be registering
3009 mutex_lock(&ext4_li_mtx);
3010 mutex_lock(&eli->li_list_mtx);
3011 if (!list_empty(&eli->li_request_list)) {
3012 mutex_unlock(&eli->li_list_mtx);
3013 mutex_unlock(&ext4_li_mtx);
3016 mutex_unlock(&eli->li_list_mtx);
3017 kfree(ext4_li_info);
3018 ext4_li_info = NULL;
3019 mutex_unlock(&ext4_li_mtx);
3024 static void ext4_clear_request_list(void)
3026 struct list_head *pos, *n;
3027 struct ext4_li_request *elr;
3029 mutex_lock(&ext4_li_info->li_list_mtx);
3030 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3031 elr = list_entry(pos, struct ext4_li_request,
3033 ext4_remove_li_request(elr);
3035 mutex_unlock(&ext4_li_info->li_list_mtx);
3038 static int ext4_run_lazyinit_thread(void)
3040 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3041 ext4_li_info, "ext4lazyinit");
3042 if (IS_ERR(ext4_lazyinit_task)) {
3043 int err = PTR_ERR(ext4_lazyinit_task);
3044 ext4_clear_request_list();
3045 kfree(ext4_li_info);
3046 ext4_li_info = NULL;
3047 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3048 "initialization thread\n",
3052 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3057 * Check whether it make sense to run itable init. thread or not.
3058 * If there is at least one uninitialized inode table, return
3059 * corresponding group number, else the loop goes through all
3060 * groups and return total number of groups.
3062 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3064 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3065 struct ext4_group_desc *gdp = NULL;
3067 for (group = 0; group < ngroups; group++) {
3068 gdp = ext4_get_group_desc(sb, group, NULL);
3072 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3079 static int ext4_li_info_new(void)
3081 struct ext4_lazy_init *eli = NULL;
3083 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3087 INIT_LIST_HEAD(&eli->li_request_list);
3088 mutex_init(&eli->li_list_mtx);
3090 eli->li_state |= EXT4_LAZYINIT_QUIT;
3097 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3100 struct ext4_sb_info *sbi = EXT4_SB(sb);
3101 struct ext4_li_request *elr;
3103 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3109 elr->lr_next_group = start;
3112 * Randomize first schedule time of the request to
3113 * spread the inode table initialization requests
3116 elr->lr_next_sched = jiffies + (prandom_u32() %
3117 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3121 int ext4_register_li_request(struct super_block *sb,
3122 ext4_group_t first_not_zeroed)
3124 struct ext4_sb_info *sbi = EXT4_SB(sb);
3125 struct ext4_li_request *elr = NULL;
3126 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3129 mutex_lock(&ext4_li_mtx);
3130 if (sbi->s_li_request != NULL) {
3132 * Reset timeout so it can be computed again, because
3133 * s_li_wait_mult might have changed.
3135 sbi->s_li_request->lr_timeout = 0;
3139 if (first_not_zeroed == ngroups ||
3140 (sb->s_flags & MS_RDONLY) ||
3141 !test_opt(sb, INIT_INODE_TABLE))
3144 elr = ext4_li_request_new(sb, first_not_zeroed);
3150 if (NULL == ext4_li_info) {
3151 ret = ext4_li_info_new();
3156 mutex_lock(&ext4_li_info->li_list_mtx);
3157 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3158 mutex_unlock(&ext4_li_info->li_list_mtx);
3160 sbi->s_li_request = elr;
3162 * set elr to NULL here since it has been inserted to
3163 * the request_list and the removal and free of it is
3164 * handled by ext4_clear_request_list from now on.
3168 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3169 ret = ext4_run_lazyinit_thread();
3174 mutex_unlock(&ext4_li_mtx);
3181 * We do not need to lock anything since this is called on
3184 static void ext4_destroy_lazyinit_thread(void)
3187 * If thread exited earlier
3188 * there's nothing to be done.
3190 if (!ext4_li_info || !ext4_lazyinit_task)
3193 kthread_stop(ext4_lazyinit_task);
3196 static int set_journal_csum_feature_set(struct super_block *sb)
3199 int compat, incompat;
3200 struct ext4_sb_info *sbi = EXT4_SB(sb);
3202 if (ext4_has_metadata_csum(sb)) {
3203 /* journal checksum v3 */
3205 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
3207 /* journal checksum v1 */
3208 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3212 jbd2_journal_clear_features(sbi->s_journal,
3213 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3214 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3215 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3216 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3217 ret = jbd2_journal_set_features(sbi->s_journal,
3219 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3221 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3222 ret = jbd2_journal_set_features(sbi->s_journal,
3225 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3226 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3228 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3229 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3236 * Note: calculating the overhead so we can be compatible with
3237 * historical BSD practice is quite difficult in the face of
3238 * clusters/bigalloc. This is because multiple metadata blocks from
3239 * different block group can end up in the same allocation cluster.
3240 * Calculating the exact overhead in the face of clustered allocation
3241 * requires either O(all block bitmaps) in memory or O(number of block
3242 * groups**2) in time. We will still calculate the superblock for
3243 * older file systems --- and if we come across with a bigalloc file
3244 * system with zero in s_overhead_clusters the estimate will be close to
3245 * correct especially for very large cluster sizes --- but for newer
3246 * file systems, it's better to calculate this figure once at mkfs
3247 * time, and store it in the superblock. If the superblock value is
3248 * present (even for non-bigalloc file systems), we will use it.
3250 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3253 struct ext4_sb_info *sbi = EXT4_SB(sb);
3254 struct ext4_group_desc *gdp;
3255 ext4_fsblk_t first_block, last_block, b;
3256 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3257 int s, j, count = 0;
3259 if (!ext4_has_feature_bigalloc(sb))
3260 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3261 sbi->s_itb_per_group + 2);
3263 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3264 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3265 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3266 for (i = 0; i < ngroups; i++) {
3267 gdp = ext4_get_group_desc(sb, i, NULL);
3268 b = ext4_block_bitmap(sb, gdp);
3269 if (b >= first_block && b <= last_block) {
3270 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3273 b = ext4_inode_bitmap(sb, gdp);
3274 if (b >= first_block && b <= last_block) {
3275 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3278 b = ext4_inode_table(sb, gdp);
3279 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3280 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3281 int c = EXT4_B2C(sbi, b - first_block);
3282 ext4_set_bit(c, buf);
3288 if (ext4_bg_has_super(sb, grp)) {
3289 ext4_set_bit(s++, buf);
3292 j = ext4_bg_num_gdb(sb, grp);
3293 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
3294 ext4_error(sb, "Invalid number of block group "
3295 "descriptor blocks: %d", j);
3296 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
3300 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3304 return EXT4_CLUSTERS_PER_GROUP(sb) -
3305 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3309 * Compute the overhead and stash it in sbi->s_overhead
3311 int ext4_calculate_overhead(struct super_block *sb)
3313 struct ext4_sb_info *sbi = EXT4_SB(sb);
3314 struct ext4_super_block *es = sbi->s_es;
3315 struct inode *j_inode;
3316 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
3317 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3318 ext4_fsblk_t overhead = 0;
3319 char *buf = (char *) get_zeroed_page(GFP_NOFS);
3325 * Compute the overhead (FS structures). This is constant
3326 * for a given filesystem unless the number of block groups
3327 * changes so we cache the previous value until it does.
3331 * All of the blocks before first_data_block are overhead
3333 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3336 * Add the overhead found in each block group
3338 for (i = 0; i < ngroups; i++) {
3341 blks = count_overhead(sb, i, buf);
3344 memset(buf, 0, PAGE_SIZE);
3349 * Add the internal journal blocks whether the journal has been
3352 if (sbi->s_journal && !sbi->journal_bdev)
3353 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3354 else if (ext4_has_feature_journal(sb) && !sbi->s_journal) {
3355 j_inode = ext4_get_journal_inode(sb, j_inum);
3357 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
3358 overhead += EXT4_NUM_B2C(sbi, j_blocks);
3361 ext4_msg(sb, KERN_ERR, "can't get journal size");
3364 sbi->s_overhead = overhead;
3366 free_page((unsigned long) buf);
3370 static void ext4_set_resv_clusters(struct super_block *sb)
3372 ext4_fsblk_t resv_clusters;
3373 struct ext4_sb_info *sbi = EXT4_SB(sb);
3376 * There's no need to reserve anything when we aren't using extents.
3377 * The space estimates are exact, there are no unwritten extents,
3378 * hole punching doesn't need new metadata... This is needed especially
3379 * to keep ext2/3 backward compatibility.
3381 if (!ext4_has_feature_extents(sb))
3384 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3385 * This should cover the situations where we can not afford to run
3386 * out of space like for example punch hole, or converting
3387 * unwritten extents in delalloc path. In most cases such
3388 * allocation would require 1, or 2 blocks, higher numbers are
3391 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3392 sbi->s_cluster_bits);
3394 do_div(resv_clusters, 50);
3395 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3397 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
3400 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3402 struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev);
3403 char *orig_data = kstrdup(data, GFP_KERNEL);
3404 struct buffer_head *bh;
3405 struct ext4_super_block *es = NULL;
3406 struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3408 ext4_fsblk_t sb_block = get_sb_block(&data);
3409 ext4_fsblk_t logical_sb_block;
3410 unsigned long offset = 0;
3411 unsigned long journal_devnum = 0;
3412 unsigned long def_mount_opts;
3416 int blocksize, clustersize;
3417 unsigned int db_count;
3419 int needs_recovery, has_huge_files, has_bigalloc;
3422 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3423 ext4_group_t first_not_zeroed;
3425 if ((data && !orig_data) || !sbi)
3428 sbi->s_daxdev = dax_dev;
3429 sbi->s_blockgroup_lock =
3430 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3431 if (!sbi->s_blockgroup_lock)
3434 sb->s_fs_info = sbi;
3436 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3437 sbi->s_sb_block = sb_block;
3438 if (sb->s_bdev->bd_part)
3439 sbi->s_sectors_written_start =
3440 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3442 /* Cleanup superblock name */
3443 strreplace(sb->s_id, '/', '!');
3445 /* -EINVAL is default */
3447 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3449 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3454 * The ext4 superblock will not be buffer aligned for other than 1kB
3455 * block sizes. We need to calculate the offset from buffer start.
3457 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3458 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3459 offset = do_div(logical_sb_block, blocksize);
3461 logical_sb_block = sb_block;
3464 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
3465 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3469 * Note: s_es must be initialized as soon as possible because
3470 * some ext4 macro-instructions depend on its value
3472 es = (struct ext4_super_block *) (bh->b_data + offset);
3474 sb->s_magic = le16_to_cpu(es->s_magic);
3475 if (sb->s_magic != EXT4_SUPER_MAGIC)
3477 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3479 /* Warn if metadata_csum and gdt_csum are both set. */
3480 if (ext4_has_feature_metadata_csum(sb) &&
3481 ext4_has_feature_gdt_csum(sb))
3482 ext4_warning(sb, "metadata_csum and uninit_bg are "
3483 "redundant flags; please run fsck.");
3485 /* Check for a known checksum algorithm */
3486 if (!ext4_verify_csum_type(sb, es)) {
3487 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3488 "unknown checksum algorithm.");
3493 /* Load the checksum driver */
3494 if (ext4_has_feature_metadata_csum(sb) ||
3495 ext4_has_feature_ea_inode(sb)) {
3496 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3497 if (IS_ERR(sbi->s_chksum_driver)) {
3498 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3499 ret = PTR_ERR(sbi->s_chksum_driver);
3500 sbi->s_chksum_driver = NULL;
3505 /* Check superblock checksum */
3506 if (!ext4_superblock_csum_verify(sb, es)) {
3507 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3508 "invalid superblock checksum. Run e2fsck?");
3514 /* Precompute checksum seed for all metadata */
3515 if (ext4_has_feature_csum_seed(sb))
3516 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
3517 else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb))
3518 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3519 sizeof(es->s_uuid));
3521 /* Set defaults before we parse the mount options */
3522 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3523 set_opt(sb, INIT_INODE_TABLE);
3524 if (def_mount_opts & EXT4_DEFM_DEBUG)
3526 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3528 if (def_mount_opts & EXT4_DEFM_UID16)
3529 set_opt(sb, NO_UID32);
3530 /* xattr user namespace & acls are now defaulted on */
3531 set_opt(sb, XATTR_USER);
3532 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3533 set_opt(sb, POSIX_ACL);
3535 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3536 if (ext4_has_metadata_csum(sb))
3537 set_opt(sb, JOURNAL_CHECKSUM);
3539 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3540 set_opt(sb, JOURNAL_DATA);
3541 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3542 set_opt(sb, ORDERED_DATA);
3543 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3544 set_opt(sb, WRITEBACK_DATA);
3546 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3547 set_opt(sb, ERRORS_PANIC);
3548 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3549 set_opt(sb, ERRORS_CONT);
3551 set_opt(sb, ERRORS_RO);
3552 /* block_validity enabled by default; disable with noblock_validity */
3553 set_opt(sb, BLOCK_VALIDITY);
3554 if (def_mount_opts & EXT4_DEFM_DISCARD)
3555 set_opt(sb, DISCARD);
3557 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3558 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3559 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3560 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3561 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3563 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3564 set_opt(sb, BARRIER);
3567 * enable delayed allocation by default
3568 * Use -o nodelalloc to turn it off
3570 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3571 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3572 set_opt(sb, DELALLOC);
3575 * set default s_li_wait_mult for lazyinit, for the case there is
3576 * no mount option specified.
3578 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3580 if (sbi->s_es->s_mount_opts[0]) {
3581 char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
3582 sizeof(sbi->s_es->s_mount_opts),
3586 if (!parse_options(s_mount_opts, sb, &journal_devnum,
3587 &journal_ioprio, 0)) {
3588 ext4_msg(sb, KERN_WARNING,
3589 "failed to parse options in superblock: %s",
3592 kfree(s_mount_opts);
3594 sbi->s_def_mount_opt = sbi->s_mount_opt;
3595 if (!parse_options((char *) data, sb, &journal_devnum,
3596 &journal_ioprio, 0))
3599 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3600 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3601 "with data=journal disables delayed "
3602 "allocation and O_DIRECT support!\n");
3603 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3604 ext4_msg(sb, KERN_ERR, "can't mount with "
3605 "both data=journal and delalloc");
3608 if (test_opt(sb, DIOREAD_NOLOCK)) {
3609 ext4_msg(sb, KERN_ERR, "can't mount with "
3610 "both data=journal and dioread_nolock");
3613 if (test_opt(sb, DAX)) {
3614 ext4_msg(sb, KERN_ERR, "can't mount with "
3615 "both data=journal and dax");
3618 if (ext4_has_feature_encrypt(sb)) {
3619 ext4_msg(sb, KERN_WARNING,
3620 "encrypted files will use data=ordered "
3621 "instead of data journaling mode");
3623 if (test_opt(sb, DELALLOC))
3624 clear_opt(sb, DELALLOC);
3626 sb->s_iflags |= SB_I_CGROUPWB;
3629 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3630 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3632 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3633 (ext4_has_compat_features(sb) ||
3634 ext4_has_ro_compat_features(sb) ||
3635 ext4_has_incompat_features(sb)))
3636 ext4_msg(sb, KERN_WARNING,
3637 "feature flags set on rev 0 fs, "
3638 "running e2fsck is recommended");
3640 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3641 set_opt2(sb, HURD_COMPAT);
3642 if (ext4_has_feature_64bit(sb)) {
3643 ext4_msg(sb, KERN_ERR,
3644 "The Hurd can't support 64-bit file systems");
3649 * ea_inode feature uses l_i_version field which is not
3650 * available in HURD_COMPAT mode.
3652 if (ext4_has_feature_ea_inode(sb)) {
3653 ext4_msg(sb, KERN_ERR,
3654 "ea_inode feature is not supported for Hurd");
3659 if (IS_EXT2_SB(sb)) {
3660 if (ext2_feature_set_ok(sb))
3661 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3662 "using the ext4 subsystem");
3664 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3665 "to feature incompatibilities");
3670 if (IS_EXT3_SB(sb)) {
3671 if (ext3_feature_set_ok(sb))
3672 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3673 "using the ext4 subsystem");
3675 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3676 "to feature incompatibilities");
3682 * Check feature flags regardless of the revision level, since we
3683 * previously didn't change the revision level when setting the flags,
3684 * so there is a chance incompat flags are set on a rev 0 filesystem.
3686 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3689 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3690 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3691 blocksize > EXT4_MAX_BLOCK_SIZE) {
3692 ext4_msg(sb, KERN_ERR,
3693 "Unsupported filesystem blocksize %d (%d log_block_size)",
3694 blocksize, le32_to_cpu(es->s_log_block_size));
3697 if (le32_to_cpu(es->s_log_block_size) >
3698 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3699 ext4_msg(sb, KERN_ERR,
3700 "Invalid log block size: %u",
3701 le32_to_cpu(es->s_log_block_size));
3705 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
3706 ext4_msg(sb, KERN_ERR,
3707 "Number of reserved GDT blocks insanely large: %d",
3708 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
3712 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
3713 err = bdev_dax_supported(sb, blocksize);
3718 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
3719 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
3720 es->s_encryption_level);
3724 if (sb->s_blocksize != blocksize) {
3725 /* Validate the filesystem blocksize */
3726 if (!sb_set_blocksize(sb, blocksize)) {
3727 ext4_msg(sb, KERN_ERR, "bad block size %d",
3733 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3734 offset = do_div(logical_sb_block, blocksize);
3735 bh = sb_bread_unmovable(sb, logical_sb_block);
3737 ext4_msg(sb, KERN_ERR,
3738 "Can't read superblock on 2nd try");
3741 es = (struct ext4_super_block *)(bh->b_data + offset);
3743 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3744 ext4_msg(sb, KERN_ERR,
3745 "Magic mismatch, very weird!");
3750 has_huge_files = ext4_has_feature_huge_file(sb);
3751 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3753 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3755 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3756 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3757 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3759 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3760 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3761 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3762 (!is_power_of_2(sbi->s_inode_size)) ||
3763 (sbi->s_inode_size > blocksize)) {
3764 ext4_msg(sb, KERN_ERR,
3765 "unsupported inode size: %d",
3769 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3770 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3773 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3774 if (ext4_has_feature_64bit(sb)) {
3775 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3776 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3777 !is_power_of_2(sbi->s_desc_size)) {
3778 ext4_msg(sb, KERN_ERR,
3779 "unsupported descriptor size %lu",
3784 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3786 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3787 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3789 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3790 if (sbi->s_inodes_per_block == 0)
3792 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
3793 sbi->s_inodes_per_group > blocksize * 8) {
3794 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
3795 sbi->s_blocks_per_group);
3798 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3799 sbi->s_inodes_per_block;
3800 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3802 sbi->s_mount_state = le16_to_cpu(es->s_state);
3803 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3804 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3806 for (i = 0; i < 4; i++)
3807 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3808 sbi->s_def_hash_version = es->s_def_hash_version;
3809 if (ext4_has_feature_dir_index(sb)) {
3810 i = le32_to_cpu(es->s_flags);
3811 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3812 sbi->s_hash_unsigned = 3;
3813 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3814 #ifdef __CHAR_UNSIGNED__
3815 if (!(sb->s_flags & MS_RDONLY))
3817 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3818 sbi->s_hash_unsigned = 3;
3820 if (!(sb->s_flags & MS_RDONLY))
3822 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3827 /* Handle clustersize */
3828 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3829 has_bigalloc = ext4_has_feature_bigalloc(sb);
3831 if (clustersize < blocksize) {
3832 ext4_msg(sb, KERN_ERR,
3833 "cluster size (%d) smaller than "
3834 "block size (%d)", clustersize, blocksize);
3837 if (le32_to_cpu(es->s_log_cluster_size) >
3838 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3839 ext4_msg(sb, KERN_ERR,
3840 "Invalid log cluster size: %u",
3841 le32_to_cpu(es->s_log_cluster_size));
3844 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3845 le32_to_cpu(es->s_log_block_size);
3846 sbi->s_clusters_per_group =
3847 le32_to_cpu(es->s_clusters_per_group);
3848 if (sbi->s_clusters_per_group > blocksize * 8) {
3849 ext4_msg(sb, KERN_ERR,
3850 "#clusters per group too big: %lu",
3851 sbi->s_clusters_per_group);
3854 if (sbi->s_blocks_per_group !=
3855 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3856 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3857 "clusters per group (%lu) inconsistent",
3858 sbi->s_blocks_per_group,
3859 sbi->s_clusters_per_group);
3863 if (clustersize != blocksize) {
3864 ext4_warning(sb, "fragment/cluster size (%d) != "
3865 "block size (%d)", clustersize,
3867 clustersize = blocksize;
3869 if (sbi->s_blocks_per_group > blocksize * 8) {
3870 ext4_msg(sb, KERN_ERR,
3871 "#blocks per group too big: %lu",
3872 sbi->s_blocks_per_group);
3875 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3876 sbi->s_cluster_bits = 0;
3878 sbi->s_cluster_ratio = clustersize / blocksize;
3880 /* Do we have standard group size of clustersize * 8 blocks ? */
3881 if (sbi->s_blocks_per_group == clustersize << 3)
3882 set_opt2(sb, STD_GROUP_SIZE);
3885 * Test whether we have more sectors than will fit in sector_t,
3886 * and whether the max offset is addressable by the page cache.
3888 err = generic_check_addressable(sb->s_blocksize_bits,
3889 ext4_blocks_count(es));
3891 ext4_msg(sb, KERN_ERR, "filesystem"
3892 " too large to mount safely on this system");
3893 if (sizeof(sector_t) < 8)
3894 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3898 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3901 /* check blocks count against device size */
3902 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3903 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3904 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3905 "exceeds size of device (%llu blocks)",
3906 ext4_blocks_count(es), blocks_count);
3911 * It makes no sense for the first data block to be beyond the end
3912 * of the filesystem.
3914 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3915 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3916 "block %u is beyond end of filesystem (%llu)",
3917 le32_to_cpu(es->s_first_data_block),
3918 ext4_blocks_count(es));
3921 blocks_count = (ext4_blocks_count(es) -
3922 le32_to_cpu(es->s_first_data_block) +
3923 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3924 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3925 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3926 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3927 "(block count %llu, first data block %u, "
3928 "blocks per group %lu)", sbi->s_groups_count,
3929 ext4_blocks_count(es),
3930 le32_to_cpu(es->s_first_data_block),
3931 EXT4_BLOCKS_PER_GROUP(sb));
3934 sbi->s_groups_count = blocks_count;
3935 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3936 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3937 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3938 EXT4_DESC_PER_BLOCK(sb);
3939 if (ext4_has_feature_meta_bg(sb)) {
3940 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
3941 ext4_msg(sb, KERN_WARNING,
3942 "first meta block group too large: %u "
3943 "(group descriptor block count %u)",
3944 le32_to_cpu(es->s_first_meta_bg), db_count);
3948 sbi->s_group_desc = kvmalloc(db_count *
3949 sizeof(struct buffer_head *),
3951 if (sbi->s_group_desc == NULL) {
3952 ext4_msg(sb, KERN_ERR, "not enough memory");
3957 bgl_lock_init(sbi->s_blockgroup_lock);
3959 /* Pre-read the descriptors into the buffer cache */
3960 for (i = 0; i < db_count; i++) {
3961 block = descriptor_loc(sb, logical_sb_block, i);
3962 sb_breadahead(sb, block);
3965 for (i = 0; i < db_count; i++) {
3966 block = descriptor_loc(sb, logical_sb_block, i);
3967 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
3968 if (!sbi->s_group_desc[i]) {
3969 ext4_msg(sb, KERN_ERR,
3970 "can't read group descriptor %d", i);
3975 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
3976 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3977 ret = -EFSCORRUPTED;
3981 sbi->s_gdb_count = db_count;
3982 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3983 spin_lock_init(&sbi->s_next_gen_lock);
3985 setup_timer(&sbi->s_err_report, print_daily_error_info,
3986 (unsigned long) sb);
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 sb->s_cop = &ext4_cryptops;
4003 sb->dq_op = &ext4_quota_operations;
4004 if (ext4_has_feature_quota(sb))
4005 sb->s_qcop = &dquot_quotactl_sysfile_ops;
4007 sb->s_qcop = &ext4_qctl_operations;
4008 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
4010 memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
4012 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
4013 mutex_init(&sbi->s_orphan_lock);
4017 needs_recovery = (es->s_last_orphan != 0 ||
4018 ext4_has_feature_journal_needs_recovery(sb));
4020 if (ext4_has_feature_mmp(sb) && !(sb->s_flags & MS_RDONLY))
4021 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
4022 goto failed_mount3a;
4025 * The first inode we look at is the journal inode. Don't try
4026 * root first: it may be modified in the journal!
4028 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
4029 err = ext4_load_journal(sb, es, journal_devnum);
4031 goto failed_mount3a;
4032 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
4033 ext4_has_feature_journal_needs_recovery(sb)) {
4034 ext4_msg(sb, KERN_ERR, "required journal recovery "
4035 "suppressed and not mounted read-only");
4036 goto failed_mount_wq;
4038 /* Nojournal mode, all journal mount options are illegal */
4039 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
4040 ext4_msg(sb, KERN_ERR, "can't mount with "
4041 "journal_checksum, fs mounted w/o journal");
4042 goto failed_mount_wq;
4044 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4045 ext4_msg(sb, KERN_ERR, "can't mount with "
4046 "journal_async_commit, fs mounted w/o journal");
4047 goto failed_mount_wq;
4049 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
4050 ext4_msg(sb, KERN_ERR, "can't mount with "
4051 "commit=%lu, fs mounted w/o journal",
4052 sbi->s_commit_interval / HZ);
4053 goto failed_mount_wq;
4055 if (EXT4_MOUNT_DATA_FLAGS &
4056 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
4057 ext4_msg(sb, KERN_ERR, "can't mount with "
4058 "data=, fs mounted w/o journal");
4059 goto failed_mount_wq;
4061 sbi->s_def_mount_opt &= EXT4_MOUNT_JOURNAL_CHECKSUM;
4062 clear_opt(sb, JOURNAL_CHECKSUM);
4063 clear_opt(sb, DATA_FLAGS);
4064 sbi->s_journal = NULL;
4069 if (ext4_has_feature_64bit(sb) &&
4070 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4071 JBD2_FEATURE_INCOMPAT_64BIT)) {
4072 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
4073 goto failed_mount_wq;
4076 if (!set_journal_csum_feature_set(sb)) {
4077 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4079 goto failed_mount_wq;
4082 /* We have now updated the journal if required, so we can
4083 * validate the data journaling mode. */
4084 switch (test_opt(sb, DATA_FLAGS)) {
4086 /* No mode set, assume a default based on the journal
4087 * capabilities: ORDERED_DATA if the journal can
4088 * cope, else JOURNAL_DATA
4090 if (jbd2_journal_check_available_features
4091 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
4092 set_opt(sb, ORDERED_DATA);
4094 set_opt(sb, JOURNAL_DATA);
4097 case EXT4_MOUNT_ORDERED_DATA:
4098 case EXT4_MOUNT_WRITEBACK_DATA:
4099 if (!jbd2_journal_check_available_features
4100 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4101 ext4_msg(sb, KERN_ERR, "Journal does not support "
4102 "requested data journaling mode");
4103 goto failed_mount_wq;
4109 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
4110 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4111 ext4_msg(sb, KERN_ERR, "can't mount with "
4112 "journal_async_commit in data=ordered mode");
4113 goto failed_mount_wq;
4116 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4118 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4121 if (!test_opt(sb, NO_MBCACHE)) {
4122 sbi->s_ea_block_cache = ext4_xattr_create_cache();
4123 if (!sbi->s_ea_block_cache) {
4124 ext4_msg(sb, KERN_ERR,
4125 "Failed to create ea_block_cache");
4126 goto failed_mount_wq;
4129 if (ext4_has_feature_ea_inode(sb)) {
4130 sbi->s_ea_inode_cache = ext4_xattr_create_cache();
4131 if (!sbi->s_ea_inode_cache) {
4132 ext4_msg(sb, KERN_ERR,
4133 "Failed to create ea_inode_cache");
4134 goto failed_mount_wq;
4139 if ((DUMMY_ENCRYPTION_ENABLED(sbi) || ext4_has_feature_encrypt(sb)) &&
4140 (blocksize != PAGE_SIZE)) {
4141 ext4_msg(sb, KERN_ERR,
4142 "Unsupported blocksize for fs encryption");
4143 goto failed_mount_wq;
4146 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !(sb->s_flags & MS_RDONLY) &&
4147 !ext4_has_feature_encrypt(sb)) {
4148 ext4_set_feature_encrypt(sb);
4149 ext4_commit_super(sb, 1);
4153 * Get the # of file system overhead blocks from the
4154 * superblock if present.
4156 if (es->s_overhead_clusters)
4157 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4159 err = ext4_calculate_overhead(sb);
4161 goto failed_mount_wq;
4165 * The maximum number of concurrent works can be high and
4166 * concurrency isn't really necessary. Limit it to 1.
4168 EXT4_SB(sb)->rsv_conversion_wq =
4169 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4170 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4171 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
4177 * The jbd2_journal_load will have done any necessary log recovery,
4178 * so we can safely mount the rest of the filesystem now.
4181 root = ext4_iget(sb, EXT4_ROOT_INO);
4183 ext4_msg(sb, KERN_ERR, "get root inode failed");
4184 ret = PTR_ERR(root);
4188 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4189 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4193 sb->s_root = d_make_root(root);
4195 ext4_msg(sb, KERN_ERR, "get root dentry failed");
4200 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
4201 sb->s_flags |= MS_RDONLY;
4203 /* determine the minimum size of new large inodes, if present */
4204 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE &&
4205 sbi->s_want_extra_isize == 0) {
4206 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4207 EXT4_GOOD_OLD_INODE_SIZE;
4208 if (ext4_has_feature_extra_isize(sb)) {
4209 if (sbi->s_want_extra_isize <
4210 le16_to_cpu(es->s_want_extra_isize))
4211 sbi->s_want_extra_isize =
4212 le16_to_cpu(es->s_want_extra_isize);
4213 if (sbi->s_want_extra_isize <
4214 le16_to_cpu(es->s_min_extra_isize))
4215 sbi->s_want_extra_isize =
4216 le16_to_cpu(es->s_min_extra_isize);
4219 /* Check if enough inode space is available */
4220 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4221 sbi->s_inode_size) {
4222 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4223 EXT4_GOOD_OLD_INODE_SIZE;
4224 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4228 ext4_set_resv_clusters(sb);
4230 err = ext4_setup_system_zone(sb);
4232 ext4_msg(sb, KERN_ERR, "failed to initialize system "
4234 goto failed_mount4a;
4238 err = ext4_mb_init(sb);
4240 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4245 block = ext4_count_free_clusters(sb);
4246 ext4_free_blocks_count_set(sbi->s_es,
4247 EXT4_C2B(sbi, block));
4248 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4251 unsigned long freei = ext4_count_free_inodes(sb);
4252 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
4253 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4257 err = percpu_counter_init(&sbi->s_dirs_counter,
4258 ext4_count_dirs(sb), GFP_KERNEL);
4260 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4263 err = percpu_init_rwsem(&sbi->s_journal_flag_rwsem);
4266 ext4_msg(sb, KERN_ERR, "insufficient memory");
4270 if (ext4_has_feature_flex_bg(sb))
4271 if (!ext4_fill_flex_info(sb)) {
4272 ext4_msg(sb, KERN_ERR,
4273 "unable to initialize "
4274 "flex_bg meta info!");
4278 err = ext4_register_li_request(sb, first_not_zeroed);
4282 err = ext4_register_sysfs(sb);
4287 /* Enable quota usage during mount. */
4288 if (ext4_has_feature_quota(sb) && !(sb->s_flags & MS_RDONLY)) {
4289 err = ext4_enable_quotas(sb);
4293 #endif /* CONFIG_QUOTA */
4295 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4296 ext4_orphan_cleanup(sb, es);
4297 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4298 if (needs_recovery) {
4299 ext4_msg(sb, KERN_INFO, "recovery complete");
4300 ext4_mark_recovery_complete(sb, es);
4302 if (EXT4_SB(sb)->s_journal) {
4303 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4304 descr = " journalled data mode";
4305 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4306 descr = " ordered data mode";
4308 descr = " writeback data mode";
4310 descr = "out journal";
4312 if (test_opt(sb, DISCARD)) {
4313 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4314 if (!blk_queue_discard(q))
4315 ext4_msg(sb, KERN_WARNING,
4316 "mounting with \"discard\" option, but "
4317 "the device does not support discard");
4320 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
4321 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4322 "Opts: %.*s%s%s", descr,
4323 (int) sizeof(sbi->s_es->s_mount_opts),
4324 sbi->s_es->s_mount_opts,
4325 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4327 if (es->s_error_count)
4328 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4330 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4331 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4332 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4333 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4340 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4345 ext4_unregister_sysfs(sb);
4348 ext4_unregister_li_request(sb);
4350 ext4_mb_release(sb);
4351 if (sbi->s_flex_groups)
4352 kvfree(sbi->s_flex_groups);
4353 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4354 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4355 percpu_counter_destroy(&sbi->s_dirs_counter);
4356 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4358 ext4_ext_release(sb);
4359 ext4_release_system_zone(sb);
4364 ext4_msg(sb, KERN_ERR, "mount failed");
4365 if (EXT4_SB(sb)->rsv_conversion_wq)
4366 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4368 if (sbi->s_ea_inode_cache) {
4369 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
4370 sbi->s_ea_inode_cache = NULL;
4372 if (sbi->s_ea_block_cache) {
4373 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
4374 sbi->s_ea_block_cache = NULL;
4376 if (sbi->s_journal) {
4377 jbd2_journal_destroy(sbi->s_journal);
4378 sbi->s_journal = NULL;
4381 ext4_es_unregister_shrinker(sbi);
4383 del_timer_sync(&sbi->s_err_report);
4385 kthread_stop(sbi->s_mmp_tsk);
4387 for (i = 0; i < db_count; i++)
4388 brelse(sbi->s_group_desc[i]);
4389 kvfree(sbi->s_group_desc);
4391 if (sbi->s_chksum_driver)
4392 crypto_free_shash(sbi->s_chksum_driver);
4394 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4395 kfree(sbi->s_qf_names[i]);
4397 ext4_blkdev_remove(sbi);
4400 sb->s_fs_info = NULL;
4401 kfree(sbi->s_blockgroup_lock);
4405 fs_put_dax(dax_dev);
4406 return err ? err : ret;
4410 * Setup any per-fs journal parameters now. We'll do this both on
4411 * initial mount, once the journal has been initialised but before we've
4412 * done any recovery; and again on any subsequent remount.
4414 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4416 struct ext4_sb_info *sbi = EXT4_SB(sb);
4418 journal->j_commit_interval = sbi->s_commit_interval;
4419 journal->j_min_batch_time = sbi->s_min_batch_time;
4420 journal->j_max_batch_time = sbi->s_max_batch_time;
4422 write_lock(&journal->j_state_lock);
4423 if (test_opt(sb, BARRIER))
4424 journal->j_flags |= JBD2_BARRIER;
4426 journal->j_flags &= ~JBD2_BARRIER;
4427 if (test_opt(sb, DATA_ERR_ABORT))
4428 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4430 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4431 write_unlock(&journal->j_state_lock);
4434 static struct inode *ext4_get_journal_inode(struct super_block *sb,
4435 unsigned int journal_inum)
4437 struct inode *journal_inode;
4440 * Test for the existence of a valid inode on disk. Bad things
4441 * happen if we iget() an unused inode, as the subsequent iput()
4442 * will try to delete it.
4444 journal_inode = ext4_iget(sb, journal_inum);
4445 if (IS_ERR(journal_inode)) {
4446 ext4_msg(sb, KERN_ERR, "no journal found");
4449 if (!journal_inode->i_nlink) {
4450 make_bad_inode(journal_inode);
4451 iput(journal_inode);
4452 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4456 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4457 journal_inode, journal_inode->i_size);
4458 if (!S_ISREG(journal_inode->i_mode)) {
4459 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4460 iput(journal_inode);
4463 return journal_inode;
4466 static journal_t *ext4_get_journal(struct super_block *sb,
4467 unsigned int journal_inum)
4469 struct inode *journal_inode;
4472 BUG_ON(!ext4_has_feature_journal(sb));
4474 journal_inode = ext4_get_journal_inode(sb, journal_inum);
4478 journal = jbd2_journal_init_inode(journal_inode);
4480 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4481 iput(journal_inode);
4484 journal->j_private = sb;
4485 ext4_init_journal_params(sb, journal);
4489 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4492 struct buffer_head *bh;
4496 int hblock, blocksize;
4497 ext4_fsblk_t sb_block;
4498 unsigned long offset;
4499 struct ext4_super_block *es;
4500 struct block_device *bdev;
4502 BUG_ON(!ext4_has_feature_journal(sb));
4504 bdev = ext4_blkdev_get(j_dev, sb);
4508 blocksize = sb->s_blocksize;
4509 hblock = bdev_logical_block_size(bdev);
4510 if (blocksize < hblock) {
4511 ext4_msg(sb, KERN_ERR,
4512 "blocksize too small for journal device");
4516 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4517 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4518 set_blocksize(bdev, blocksize);
4519 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4520 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4521 "external journal");
4525 es = (struct ext4_super_block *) (bh->b_data + offset);
4526 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4527 !(le32_to_cpu(es->s_feature_incompat) &
4528 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4529 ext4_msg(sb, KERN_ERR, "external journal has "
4535 if ((le32_to_cpu(es->s_feature_ro_compat) &
4536 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4537 es->s_checksum != ext4_superblock_csum(sb, es)) {
4538 ext4_msg(sb, KERN_ERR, "external journal has "
4539 "corrupt superblock");
4544 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4545 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4550 len = ext4_blocks_count(es);
4551 start = sb_block + 1;
4552 brelse(bh); /* we're done with the superblock */
4554 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4555 start, len, blocksize);
4557 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4560 journal->j_private = sb;
4561 ll_rw_block(REQ_OP_READ, REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4562 wait_on_buffer(journal->j_sb_buffer);
4563 if (!buffer_uptodate(journal->j_sb_buffer)) {
4564 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4567 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4568 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4569 "user (unsupported) - %d",
4570 be32_to_cpu(journal->j_superblock->s_nr_users));
4573 EXT4_SB(sb)->journal_bdev = bdev;
4574 ext4_init_journal_params(sb, journal);
4578 jbd2_journal_destroy(journal);
4580 ext4_blkdev_put(bdev);
4584 static int ext4_load_journal(struct super_block *sb,
4585 struct ext4_super_block *es,
4586 unsigned long journal_devnum)
4589 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4592 int really_read_only;
4594 BUG_ON(!ext4_has_feature_journal(sb));
4596 if (journal_devnum &&
4597 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4598 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4599 "numbers have changed");
4600 journal_dev = new_decode_dev(journal_devnum);
4602 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4604 really_read_only = bdev_read_only(sb->s_bdev);
4607 * Are we loading a blank journal or performing recovery after a
4608 * crash? For recovery, we need to check in advance whether we
4609 * can get read-write access to the device.
4611 if (ext4_has_feature_journal_needs_recovery(sb)) {
4612 if (sb->s_flags & MS_RDONLY) {
4613 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4614 "required on readonly filesystem");
4615 if (really_read_only) {
4616 ext4_msg(sb, KERN_ERR, "write access "
4617 "unavailable, cannot proceed");
4620 ext4_msg(sb, KERN_INFO, "write access will "
4621 "be enabled during recovery");
4625 if (journal_inum && journal_dev) {
4626 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4627 "and inode journals!");
4632 if (!(journal = ext4_get_journal(sb, journal_inum)))
4635 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4639 if (!(journal->j_flags & JBD2_BARRIER))
4640 ext4_msg(sb, KERN_INFO, "barriers disabled");
4642 if (!ext4_has_feature_journal_needs_recovery(sb))
4643 err = jbd2_journal_wipe(journal, !really_read_only);
4645 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4647 memcpy(save, ((char *) es) +
4648 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4649 err = jbd2_journal_load(journal);
4651 memcpy(((char *) es) + EXT4_S_ERR_START,
4652 save, EXT4_S_ERR_LEN);
4657 ext4_msg(sb, KERN_ERR, "error loading journal");
4658 jbd2_journal_destroy(journal);
4662 EXT4_SB(sb)->s_journal = journal;
4663 ext4_clear_journal_err(sb, es);
4665 if (!really_read_only && journal_devnum &&
4666 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4667 es->s_journal_dev = cpu_to_le32(journal_devnum);
4669 /* Make sure we flush the recovery flag to disk. */
4670 ext4_commit_super(sb, 1);
4676 static int ext4_commit_super(struct super_block *sb, int sync)
4678 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4679 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4682 if (!sbh || block_device_ejected(sb))
4685 * If the file system is mounted read-only, don't update the
4686 * superblock write time. This avoids updating the superblock
4687 * write time when we are mounting the root file system
4688 * read/only but we need to replay the journal; at that point,
4689 * for people who are east of GMT and who make their clock
4690 * tick in localtime for Windows bug-for-bug compatibility,
4691 * the clock is set in the future, and this will cause e2fsck
4692 * to complain and force a full file system check.
4694 if (!(sb->s_flags & MS_RDONLY))
4695 es->s_wtime = cpu_to_le32(get_seconds());
4696 if (sb->s_bdev->bd_part)
4697 es->s_kbytes_written =
4698 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4699 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4700 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4702 es->s_kbytes_written =
4703 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4704 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4705 ext4_free_blocks_count_set(es,
4706 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4707 &EXT4_SB(sb)->s_freeclusters_counter)));
4708 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4709 es->s_free_inodes_count =
4710 cpu_to_le32(percpu_counter_sum_positive(
4711 &EXT4_SB(sb)->s_freeinodes_counter));
4712 BUFFER_TRACE(sbh, "marking dirty");
4713 ext4_superblock_csum_set(sb);
4716 if (buffer_write_io_error(sbh)) {
4718 * Oh, dear. A previous attempt to write the
4719 * superblock failed. This could happen because the
4720 * USB device was yanked out. Or it could happen to
4721 * be a transient write error and maybe the block will
4722 * be remapped. Nothing we can do but to retry the
4723 * write and hope for the best.
4725 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4726 "superblock detected");
4727 clear_buffer_write_io_error(sbh);
4728 set_buffer_uptodate(sbh);
4730 mark_buffer_dirty(sbh);
4733 error = __sync_dirty_buffer(sbh,
4734 REQ_SYNC | (test_opt(sb, BARRIER) ? REQ_FUA : 0));
4738 error = buffer_write_io_error(sbh);
4740 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4742 clear_buffer_write_io_error(sbh);
4743 set_buffer_uptodate(sbh);
4750 * Have we just finished recovery? If so, and if we are mounting (or
4751 * remounting) the filesystem readonly, then we will end up with a
4752 * consistent fs on disk. Record that fact.
4754 static void ext4_mark_recovery_complete(struct super_block *sb,
4755 struct ext4_super_block *es)
4757 journal_t *journal = EXT4_SB(sb)->s_journal;
4759 if (!ext4_has_feature_journal(sb)) {
4760 BUG_ON(journal != NULL);
4763 jbd2_journal_lock_updates(journal);
4764 if (jbd2_journal_flush(journal) < 0)
4767 if (ext4_has_feature_journal_needs_recovery(sb) &&
4768 sb->s_flags & MS_RDONLY) {
4769 ext4_clear_feature_journal_needs_recovery(sb);
4770 ext4_commit_super(sb, 1);
4774 jbd2_journal_unlock_updates(journal);
4778 * If we are mounting (or read-write remounting) a filesystem whose journal
4779 * has recorded an error from a previous lifetime, move that error to the
4780 * main filesystem now.
4782 static void ext4_clear_journal_err(struct super_block *sb,
4783 struct ext4_super_block *es)
4789 BUG_ON(!ext4_has_feature_journal(sb));
4791 journal = EXT4_SB(sb)->s_journal;
4794 * Now check for any error status which may have been recorded in the
4795 * journal by a prior ext4_error() or ext4_abort()
4798 j_errno = jbd2_journal_errno(journal);
4802 errstr = ext4_decode_error(sb, j_errno, nbuf);
4803 ext4_warning(sb, "Filesystem error recorded "
4804 "from previous mount: %s", errstr);
4805 ext4_warning(sb, "Marking fs in need of filesystem check.");
4807 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4808 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4809 ext4_commit_super(sb, 1);
4811 jbd2_journal_clear_err(journal);
4812 jbd2_journal_update_sb_errno(journal);
4817 * Force the running and committing transactions to commit,
4818 * and wait on the commit.
4820 int ext4_force_commit(struct super_block *sb)
4824 if (sb->s_flags & MS_RDONLY)
4827 journal = EXT4_SB(sb)->s_journal;
4828 return ext4_journal_force_commit(journal);
4831 static int ext4_sync_fs(struct super_block *sb, int wait)
4835 bool needs_barrier = false;
4836 struct ext4_sb_info *sbi = EXT4_SB(sb);
4838 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
4841 trace_ext4_sync_fs(sb, wait);
4842 flush_workqueue(sbi->rsv_conversion_wq);
4844 * Writeback quota in non-journalled quota case - journalled quota has
4847 dquot_writeback_dquots(sb, -1);
4849 * Data writeback is possible w/o journal transaction, so barrier must
4850 * being sent at the end of the function. But we can skip it if
4851 * transaction_commit will do it for us.
4853 if (sbi->s_journal) {
4854 target = jbd2_get_latest_transaction(sbi->s_journal);
4855 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4856 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4857 needs_barrier = true;
4859 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4861 ret = jbd2_log_wait_commit(sbi->s_journal,
4864 } else if (wait && test_opt(sb, BARRIER))
4865 needs_barrier = true;
4866 if (needs_barrier) {
4868 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4877 * LVM calls this function before a (read-only) snapshot is created. This
4878 * gives us a chance to flush the journal completely and mark the fs clean.
4880 * Note that only this function cannot bring a filesystem to be in a clean
4881 * state independently. It relies on upper layer to stop all data & metadata
4884 static int ext4_freeze(struct super_block *sb)
4889 if (sb->s_flags & MS_RDONLY)
4892 journal = EXT4_SB(sb)->s_journal;
4895 /* Now we set up the journal barrier. */
4896 jbd2_journal_lock_updates(journal);
4899 * Don't clear the needs_recovery flag if we failed to
4900 * flush the journal.
4902 error = jbd2_journal_flush(journal);
4906 /* Journal blocked and flushed, clear needs_recovery flag. */
4907 ext4_clear_feature_journal_needs_recovery(sb);
4910 error = ext4_commit_super(sb, 1);
4913 /* we rely on upper layer to stop further updates */
4914 jbd2_journal_unlock_updates(journal);
4919 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4920 * flag here, even though the filesystem is not technically dirty yet.
4922 static int ext4_unfreeze(struct super_block *sb)
4924 if ((sb->s_flags & MS_RDONLY) || ext4_forced_shutdown(EXT4_SB(sb)))
4927 if (EXT4_SB(sb)->s_journal) {
4928 /* Reset the needs_recovery flag before the fs is unlocked. */
4929 ext4_set_feature_journal_needs_recovery(sb);
4932 ext4_commit_super(sb, 1);
4937 * Structure to save mount options for ext4_remount's benefit
4939 struct ext4_mount_options {
4940 unsigned long s_mount_opt;
4941 unsigned long s_mount_opt2;
4944 unsigned long s_commit_interval;
4945 u32 s_min_batch_time, s_max_batch_time;
4948 char *s_qf_names[EXT4_MAXQUOTAS];
4952 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4954 struct ext4_super_block *es;
4955 struct ext4_sb_info *sbi = EXT4_SB(sb);
4956 unsigned long old_sb_flags;
4957 struct ext4_mount_options old_opts;
4958 int enable_quota = 0;
4960 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4965 char *orig_data = kstrdup(data, GFP_KERNEL);
4967 /* Store the original options */
4968 old_sb_flags = sb->s_flags;
4969 old_opts.s_mount_opt = sbi->s_mount_opt;
4970 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4971 old_opts.s_resuid = sbi->s_resuid;
4972 old_opts.s_resgid = sbi->s_resgid;
4973 old_opts.s_commit_interval = sbi->s_commit_interval;
4974 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4975 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4977 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4978 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4979 if (sbi->s_qf_names[i]) {
4980 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4982 if (!old_opts.s_qf_names[i]) {
4983 for (j = 0; j < i; j++)
4984 kfree(old_opts.s_qf_names[j]);
4989 old_opts.s_qf_names[i] = NULL;
4991 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4992 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4994 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4999 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
5000 test_opt(sb, JOURNAL_CHECKSUM)) {
5001 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
5002 "during remount not supported; ignoring");
5003 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
5006 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
5007 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
5008 ext4_msg(sb, KERN_ERR, "can't mount with "
5009 "both data=journal and delalloc");
5013 if (test_opt(sb, DIOREAD_NOLOCK)) {
5014 ext4_msg(sb, KERN_ERR, "can't mount with "
5015 "both data=journal and dioread_nolock");
5019 if (test_opt(sb, DAX)) {
5020 ext4_msg(sb, KERN_ERR, "can't mount with "
5021 "both data=journal and dax");
5025 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
5026 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
5027 ext4_msg(sb, KERN_ERR, "can't mount with "
5028 "journal_async_commit in data=ordered mode");
5034 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) {
5035 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount");
5040 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
5041 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
5042 "dax flag with busy inodes while remounting");
5043 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
5046 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
5047 ext4_abort(sb, "Abort forced by user");
5049 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
5050 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
5054 if (sbi->s_journal) {
5055 ext4_init_journal_params(sb, sbi->s_journal);
5056 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
5059 if (*flags & MS_LAZYTIME)
5060 sb->s_flags |= MS_LAZYTIME;
5062 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
5063 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
5068 if (*flags & MS_RDONLY) {
5069 err = sync_filesystem(sb);
5072 err = dquot_suspend(sb, -1);
5077 * First of all, the unconditional stuff we have to do
5078 * to disable replay of the journal when we next remount
5080 sb->s_flags |= MS_RDONLY;
5083 * OK, test if we are remounting a valid rw partition
5084 * readonly, and if so set the rdonly flag and then
5085 * mark the partition as valid again.
5087 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
5088 (sbi->s_mount_state & EXT4_VALID_FS))
5089 es->s_state = cpu_to_le16(sbi->s_mount_state);
5092 ext4_mark_recovery_complete(sb, es);
5094 /* Make sure we can mount this feature set readwrite */
5095 if (ext4_has_feature_readonly(sb) ||
5096 !ext4_feature_set_ok(sb, 0)) {
5101 * Make sure the group descriptor checksums
5102 * are sane. If they aren't, refuse to remount r/w.
5104 for (g = 0; g < sbi->s_groups_count; g++) {
5105 struct ext4_group_desc *gdp =
5106 ext4_get_group_desc(sb, g, NULL);
5108 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
5109 ext4_msg(sb, KERN_ERR,
5110 "ext4_remount: Checksum for group %u failed (%u!=%u)",
5111 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
5112 le16_to_cpu(gdp->bg_checksum));
5119 * If we have an unprocessed orphan list hanging
5120 * around from a previously readonly bdev mount,
5121 * require a full umount/remount for now.
5123 if (es->s_last_orphan) {
5124 ext4_msg(sb, KERN_WARNING, "Couldn't "
5125 "remount RDWR because of unprocessed "
5126 "orphan inode list. Please "
5127 "umount/remount instead");
5133 * Mounting a RDONLY partition read-write, so reread
5134 * and store the current valid flag. (It may have
5135 * been changed by e2fsck since we originally mounted
5139 ext4_clear_journal_err(sb, es);
5140 sbi->s_mount_state = le16_to_cpu(es->s_state);
5141 if (!ext4_setup_super(sb, es, 0))
5142 sb->s_flags &= ~MS_RDONLY;
5143 if (ext4_has_feature_mmp(sb))
5144 if (ext4_multi_mount_protect(sb,
5145 le64_to_cpu(es->s_mmp_block))) {
5154 * Reinitialize lazy itable initialization thread based on
5157 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
5158 ext4_unregister_li_request(sb);
5160 ext4_group_t first_not_zeroed;
5161 first_not_zeroed = ext4_has_uninit_itable(sb);
5162 ext4_register_li_request(sb, first_not_zeroed);
5165 ext4_setup_system_zone(sb);
5166 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
5167 ext4_commit_super(sb, 1);
5170 /* Release old quota file names */
5171 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5172 kfree(old_opts.s_qf_names[i]);
5174 if (sb_any_quota_suspended(sb))
5175 dquot_resume(sb, -1);
5176 else if (ext4_has_feature_quota(sb)) {
5177 err = ext4_enable_quotas(sb);
5184 *flags = (*flags & ~MS_LAZYTIME) | (sb->s_flags & MS_LAZYTIME);
5185 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5190 sb->s_flags = old_sb_flags;
5191 sbi->s_mount_opt = old_opts.s_mount_opt;
5192 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
5193 sbi->s_resuid = old_opts.s_resuid;
5194 sbi->s_resgid = old_opts.s_resgid;
5195 sbi->s_commit_interval = old_opts.s_commit_interval;
5196 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5197 sbi->s_max_batch_time = old_opts.s_max_batch_time;
5199 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
5200 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
5201 kfree(sbi->s_qf_names[i]);
5202 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
5210 static int ext4_statfs_project(struct super_block *sb,
5211 kprojid_t projid, struct kstatfs *buf)
5214 struct dquot *dquot;
5218 qid = make_kqid_projid(projid);
5219 dquot = dqget(sb, qid);
5221 return PTR_ERR(dquot);
5222 spin_lock(&dquot->dq_dqb_lock);
5224 limit = (dquot->dq_dqb.dqb_bsoftlimit ?
5225 dquot->dq_dqb.dqb_bsoftlimit :
5226 dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits;
5227 if (limit && buf->f_blocks > limit) {
5228 curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits;
5229 buf->f_blocks = limit;
5230 buf->f_bfree = buf->f_bavail =
5231 (buf->f_blocks > curblock) ?
5232 (buf->f_blocks - curblock) : 0;
5235 limit = dquot->dq_dqb.dqb_isoftlimit ?
5236 dquot->dq_dqb.dqb_isoftlimit :
5237 dquot->dq_dqb.dqb_ihardlimit;
5238 if (limit && buf->f_files > limit) {
5239 buf->f_files = limit;
5241 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
5242 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
5245 spin_unlock(&dquot->dq_dqb_lock);
5251 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
5253 struct super_block *sb = dentry->d_sb;
5254 struct ext4_sb_info *sbi = EXT4_SB(sb);
5255 struct ext4_super_block *es = sbi->s_es;
5256 ext4_fsblk_t overhead = 0, resv_blocks;
5259 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
5261 if (!test_opt(sb, MINIX_DF))
5262 overhead = sbi->s_overhead;
5264 buf->f_type = EXT4_SUPER_MAGIC;
5265 buf->f_bsize = sb->s_blocksize;
5266 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
5267 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5268 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
5269 /* prevent underflow in case that few free space is available */
5270 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
5271 buf->f_bavail = buf->f_bfree -
5272 (ext4_r_blocks_count(es) + resv_blocks);
5273 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
5275 buf->f_files = le32_to_cpu(es->s_inodes_count);
5276 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5277 buf->f_namelen = EXT4_NAME_LEN;
5278 fsid = le64_to_cpup((void *)es->s_uuid) ^
5279 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5280 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5281 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
5284 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
5285 sb_has_quota_limits_enabled(sb, PRJQUOTA))
5286 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
5295 * Helper functions so that transaction is started before we acquire dqio_sem
5296 * to keep correct lock ordering of transaction > dqio_sem
5298 static inline struct inode *dquot_to_inode(struct dquot *dquot)
5300 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
5303 static int ext4_write_dquot(struct dquot *dquot)
5307 struct inode *inode;
5309 inode = dquot_to_inode(dquot);
5310 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
5311 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
5313 return PTR_ERR(handle);
5314 ret = dquot_commit(dquot);
5315 err = ext4_journal_stop(handle);
5321 static int ext4_acquire_dquot(struct dquot *dquot)
5326 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5327 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
5329 return PTR_ERR(handle);
5330 ret = dquot_acquire(dquot);
5331 err = ext4_journal_stop(handle);
5337 static int ext4_release_dquot(struct dquot *dquot)
5342 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5343 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
5344 if (IS_ERR(handle)) {
5345 /* Release dquot anyway to avoid endless cycle in dqput() */
5346 dquot_release(dquot);
5347 return PTR_ERR(handle);
5349 ret = dquot_release(dquot);
5350 err = ext4_journal_stop(handle);
5356 static int ext4_mark_dquot_dirty(struct dquot *dquot)
5358 struct super_block *sb = dquot->dq_sb;
5359 struct ext4_sb_info *sbi = EXT4_SB(sb);
5361 /* Are we journaling quotas? */
5362 if (ext4_has_feature_quota(sb) ||
5363 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
5364 dquot_mark_dquot_dirty(dquot);
5365 return ext4_write_dquot(dquot);
5367 return dquot_mark_dquot_dirty(dquot);
5371 static int ext4_write_info(struct super_block *sb, int type)
5376 /* Data block + inode block */
5377 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
5379 return PTR_ERR(handle);
5380 ret = dquot_commit_info(sb, type);
5381 err = ext4_journal_stop(handle);
5388 * Turn on quotas during mount time - we need to find
5389 * the quota file and such...
5391 static int ext4_quota_on_mount(struct super_block *sb, int type)
5393 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5394 EXT4_SB(sb)->s_jquota_fmt, type);
5397 static void lockdep_set_quota_inode(struct inode *inode, int subclass)
5399 struct ext4_inode_info *ei = EXT4_I(inode);
5401 /* The first argument of lockdep_set_subclass has to be
5402 * *exactly* the same as the argument to init_rwsem() --- in
5403 * this case, in init_once() --- or lockdep gets unhappy
5404 * because the name of the lock is set using the
5405 * stringification of the argument to init_rwsem().
5407 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
5408 lockdep_set_subclass(&ei->i_data_sem, subclass);
5412 * Standard function to be called on quota_on
5414 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
5415 const struct path *path)
5419 if (!test_opt(sb, QUOTA))
5422 /* Quotafile not on the same filesystem? */
5423 if (path->dentry->d_sb != sb)
5425 /* Journaling quota? */
5426 if (EXT4_SB(sb)->s_qf_names[type]) {
5427 /* Quotafile not in fs root? */
5428 if (path->dentry->d_parent != sb->s_root)
5429 ext4_msg(sb, KERN_WARNING,
5430 "Quota file not on filesystem root. "
5431 "Journaled quota will not work");
5432 sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY;
5435 * Clear the flag just in case mount options changed since
5438 sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY;
5442 * When we journal data on quota file, we have to flush journal to see
5443 * all updates to the file when we bypass pagecache...
5445 if (EXT4_SB(sb)->s_journal &&
5446 ext4_should_journal_data(d_inode(path->dentry))) {
5448 * We don't need to lock updates but journal_flush() could
5449 * otherwise be livelocked...
5451 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
5452 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
5453 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
5458 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
5459 err = dquot_quota_on(sb, type, format_id, path);
5461 lockdep_set_quota_inode(path->dentry->d_inode,
5464 struct inode *inode = d_inode(path->dentry);
5468 * Set inode flags to prevent userspace from messing with quota
5469 * files. If this fails, we return success anyway since quotas
5470 * are already enabled and this is not a hard failure.
5473 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5476 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL;
5477 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
5478 S_NOATIME | S_IMMUTABLE);
5479 ext4_mark_inode_dirty(handle, inode);
5480 ext4_journal_stop(handle);
5482 inode_unlock(inode);
5487 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5491 struct inode *qf_inode;
5492 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5493 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5494 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5495 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
5498 BUG_ON(!ext4_has_feature_quota(sb));
5500 if (!qf_inums[type])
5503 qf_inode = ext4_iget(sb, qf_inums[type]);
5504 if (IS_ERR(qf_inode)) {
5505 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5506 return PTR_ERR(qf_inode);
5509 /* Don't account quota for quota files to avoid recursion */
5510 qf_inode->i_flags |= S_NOQUOTA;
5511 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
5512 err = dquot_enable(qf_inode, type, format_id, flags);
5515 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
5520 /* Enable usage tracking for all quota types. */
5521 static int ext4_enable_quotas(struct super_block *sb)
5524 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5525 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5526 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5527 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
5529 bool quota_mopt[EXT4_MAXQUOTAS] = {
5530 test_opt(sb, USRQUOTA),
5531 test_opt(sb, GRPQUOTA),
5532 test_opt(sb, PRJQUOTA),
5535 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY;
5536 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
5537 if (qf_inums[type]) {
5538 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5539 DQUOT_USAGE_ENABLED |
5540 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
5542 for (type--; type >= 0; type--)
5543 dquot_quota_off(sb, type);
5546 "Failed to enable quota tracking "
5547 "(type=%d, err=%d). Please run "
5548 "e2fsck to fix.", type, err);
5556 static int ext4_quota_off(struct super_block *sb, int type)
5558 struct inode *inode = sb_dqopt(sb)->files[type];
5562 /* Force all delayed allocation blocks to be allocated.
5563 * Caller already holds s_umount sem */
5564 if (test_opt(sb, DELALLOC))
5565 sync_filesystem(sb);
5567 if (!inode || !igrab(inode))
5570 err = dquot_quota_off(sb, type);
5571 if (err || ext4_has_feature_quota(sb))
5576 * Update modification times of quota files when userspace can
5577 * start looking at them. If we fail, we return success anyway since
5578 * this is not a hard failure and quotas are already disabled.
5580 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5583 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL);
5584 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
5585 inode->i_mtime = inode->i_ctime = current_time(inode);
5586 ext4_mark_inode_dirty(handle, inode);
5587 ext4_journal_stop(handle);
5589 inode_unlock(inode);
5591 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL);
5595 return dquot_quota_off(sb, type);
5598 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5599 * acquiring the locks... As quota files are never truncated and quota code
5600 * itself serializes the operations (and no one else should touch the files)
5601 * we don't have to be afraid of races */
5602 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5603 size_t len, loff_t off)
5605 struct inode *inode = sb_dqopt(sb)->files[type];
5606 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5607 int offset = off & (sb->s_blocksize - 1);
5610 struct buffer_head *bh;
5611 loff_t i_size = i_size_read(inode);
5615 if (off+len > i_size)
5618 while (toread > 0) {
5619 tocopy = sb->s_blocksize - offset < toread ?
5620 sb->s_blocksize - offset : toread;
5621 bh = ext4_bread(NULL, inode, blk, 0);
5624 if (!bh) /* A hole? */
5625 memset(data, 0, tocopy);
5627 memcpy(data, bh->b_data+offset, tocopy);
5637 /* Write to quotafile (we know the transaction is already started and has
5638 * enough credits) */
5639 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5640 const char *data, size_t len, loff_t off)
5642 struct inode *inode = sb_dqopt(sb)->files[type];
5643 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5644 int err, offset = off & (sb->s_blocksize - 1);
5646 struct buffer_head *bh;
5647 handle_t *handle = journal_current_handle();
5649 if (EXT4_SB(sb)->s_journal && !handle) {
5650 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5651 " cancelled because transaction is not started",
5652 (unsigned long long)off, (unsigned long long)len);
5656 * Since we account only one data block in transaction credits,
5657 * then it is impossible to cross a block boundary.
5659 if (sb->s_blocksize - offset < len) {
5660 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5661 " cancelled because not block aligned",
5662 (unsigned long long)off, (unsigned long long)len);
5667 bh = ext4_bread(handle, inode, blk,
5668 EXT4_GET_BLOCKS_CREATE |
5669 EXT4_GET_BLOCKS_METADATA_NOFAIL);
5670 } while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) &&
5671 ext4_should_retry_alloc(inode->i_sb, &retries));
5676 BUFFER_TRACE(bh, "get write access");
5677 err = ext4_journal_get_write_access(handle, bh);
5683 memcpy(bh->b_data+offset, data, len);
5684 flush_dcache_page(bh->b_page);
5686 err = ext4_handle_dirty_metadata(handle, NULL, bh);
5689 if (inode->i_size < off + len) {
5690 i_size_write(inode, off + len);
5691 EXT4_I(inode)->i_disksize = inode->i_size;
5692 ext4_mark_inode_dirty(handle, inode);
5697 static int ext4_get_next_id(struct super_block *sb, struct kqid *qid)
5699 const struct quota_format_ops *ops;
5701 if (!sb_has_quota_loaded(sb, qid->type))
5703 ops = sb_dqopt(sb)->ops[qid->type];
5704 if (!ops || !ops->get_next_id)
5706 return dquot_get_next_id(sb, qid);
5710 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5711 const char *dev_name, void *data)
5713 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5716 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
5717 static inline void register_as_ext2(void)
5719 int err = register_filesystem(&ext2_fs_type);
5722 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5725 static inline void unregister_as_ext2(void)
5727 unregister_filesystem(&ext2_fs_type);
5730 static inline int ext2_feature_set_ok(struct super_block *sb)
5732 if (ext4_has_unknown_ext2_incompat_features(sb))
5734 if (sb->s_flags & MS_RDONLY)
5736 if (ext4_has_unknown_ext2_ro_compat_features(sb))
5741 static inline void register_as_ext2(void) { }
5742 static inline void unregister_as_ext2(void) { }
5743 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5746 static inline void register_as_ext3(void)
5748 int err = register_filesystem(&ext3_fs_type);
5751 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5754 static inline void unregister_as_ext3(void)
5756 unregister_filesystem(&ext3_fs_type);
5759 static inline int ext3_feature_set_ok(struct super_block *sb)
5761 if (ext4_has_unknown_ext3_incompat_features(sb))
5763 if (!ext4_has_feature_journal(sb))
5765 if (sb->s_flags & MS_RDONLY)
5767 if (ext4_has_unknown_ext3_ro_compat_features(sb))
5772 static struct file_system_type ext4_fs_type = {
5773 .owner = THIS_MODULE,
5775 .mount = ext4_mount,
5776 .kill_sb = kill_block_super,
5777 .fs_flags = FS_REQUIRES_DEV,
5779 MODULE_ALIAS_FS("ext4");
5781 /* Shared across all ext4 file systems */
5782 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5784 static int __init ext4_init_fs(void)
5788 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
5789 ext4_li_info = NULL;
5790 mutex_init(&ext4_li_mtx);
5792 /* Build-time check for flags consistency */
5793 ext4_check_flag_values();
5795 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
5796 init_waitqueue_head(&ext4__ioend_wq[i]);
5798 err = ext4_init_es();
5802 err = ext4_init_pageio();
5806 err = ext4_init_system_zone();
5810 err = ext4_init_sysfs();
5814 err = ext4_init_mballoc();
5817 err = init_inodecache();
5822 err = register_filesystem(&ext4_fs_type);
5828 unregister_as_ext2();
5829 unregister_as_ext3();
5830 destroy_inodecache();
5832 ext4_exit_mballoc();
5836 ext4_exit_system_zone();
5845 static void __exit ext4_exit_fs(void)
5847 ext4_destroy_lazyinit_thread();
5848 unregister_as_ext2();
5849 unregister_as_ext3();
5850 unregister_filesystem(&ext4_fs_type);
5851 destroy_inodecache();
5852 ext4_exit_mballoc();
5854 ext4_exit_system_zone();
5859 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5860 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5861 MODULE_LICENSE("GPL");
5862 module_init(ext4_init_fs)
5863 module_exit(ext4_exit_fs)