1 // SPDX-License-Identifier: GPL-2.0
3 * linux/fs/ext4/super.c
5 * Copyright (C) 1992, 1993, 1994, 1995
6 * Remy Card (card@masi.ibp.fr)
7 * Laboratoire MASI - Institut Blaise Pascal
8 * Universite Pierre et Marie Curie (Paris VI)
12 * linux/fs/minix/inode.c
14 * Copyright (C) 1991, 1992 Linus Torvalds
16 * Big-endian to little-endian byte-swapping/bitmaps by
17 * David S. Miller (davem@caip.rutgers.edu), 1995
20 #include <linux/module.h>
21 #include <linux/string.h>
23 #include <linux/time.h>
24 #include <linux/vmalloc.h>
25 #include <linux/slab.h>
26 #include <linux/init.h>
27 #include <linux/blkdev.h>
28 #include <linux/backing-dev.h>
29 #include <linux/parser.h>
30 #include <linux/buffer_head.h>
31 #include <linux/exportfs.h>
32 #include <linux/vfs.h>
33 #include <linux/random.h>
34 #include <linux/mount.h>
35 #include <linux/namei.h>
36 #include <linux/quotaops.h>
37 #include <linux/seq_file.h>
38 #include <linux/ctype.h>
39 #include <linux/log2.h>
40 #include <linux/crc16.h>
41 #include <linux/dax.h>
42 #include <linux/cleancache.h>
43 #include <linux/uaccess.h>
44 #include <linux/iversion.h>
46 #include <linux/kthread.h>
47 #include <linux/freezer.h>
50 #include "ext4_extents.h" /* Needed for trace points definition */
51 #include "ext4_jbd2.h"
57 #define CREATE_TRACE_POINTS
58 #include <trace/events/ext4.h>
60 static struct ext4_lazy_init *ext4_li_info;
61 static struct mutex ext4_li_mtx;
62 static struct ratelimit_state ext4_mount_msg_ratelimit;
64 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
65 unsigned long journal_devnum);
66 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
67 static int ext4_commit_super(struct super_block *sb, int sync);
68 static void ext4_mark_recovery_complete(struct super_block *sb,
69 struct ext4_super_block *es);
70 static void ext4_clear_journal_err(struct super_block *sb,
71 struct ext4_super_block *es);
72 static int ext4_sync_fs(struct super_block *sb, int wait);
73 static int ext4_remount(struct super_block *sb, int *flags, char *data);
74 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
75 static int ext4_unfreeze(struct super_block *sb);
76 static int ext4_freeze(struct super_block *sb);
77 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
78 const char *dev_name, void *data);
79 static inline int ext2_feature_set_ok(struct super_block *sb);
80 static inline int ext3_feature_set_ok(struct super_block *sb);
81 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
82 static void ext4_destroy_lazyinit_thread(void);
83 static void ext4_unregister_li_request(struct super_block *sb);
84 static void ext4_clear_request_list(void);
85 static struct inode *ext4_get_journal_inode(struct super_block *sb,
86 unsigned int journal_inum);
91 * Note the difference between i_mmap_sem (EXT4_I(inode)->i_mmap_sem) and
92 * i_mmap_rwsem (inode->i_mmap_rwsem)!
95 * mmap_sem -> sb_start_pagefault -> i_mmap_sem (r) -> transaction start ->
96 * page lock -> i_data_sem (rw)
98 * buffered write path:
99 * sb_start_write -> i_mutex -> mmap_sem
100 * sb_start_write -> i_mutex -> transaction start -> page lock ->
104 * sb_start_write -> i_mutex -> i_mmap_sem (w) -> i_mmap_rwsem (w) -> page lock
105 * sb_start_write -> i_mutex -> i_mmap_sem (w) -> transaction start ->
109 * sb_start_write -> i_mutex -> mmap_sem
110 * sb_start_write -> i_mutex -> 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 (test_opt(sb, WARN_ON_ERROR))
414 if (!test_opt(sb, ERRORS_CONT)) {
415 journal_t *journal = EXT4_SB(sb)->s_journal;
417 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
419 jbd2_journal_abort(journal, -EIO);
421 if (test_opt(sb, ERRORS_RO)) {
422 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
424 * Make sure updated value of ->s_mount_flags will be visible
425 * before ->s_flags update
428 sb->s_flags |= SB_RDONLY;
430 if (test_opt(sb, ERRORS_PANIC)) {
431 if (EXT4_SB(sb)->s_journal &&
432 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
434 panic("EXT4-fs (device %s): panic forced after error\n",
439 #define ext4_error_ratelimit(sb) \
440 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
443 void __ext4_error(struct super_block *sb, const char *function,
444 unsigned int line, const char *fmt, ...)
446 struct va_format vaf;
449 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
452 trace_ext4_error(sb, function, line);
453 if (ext4_error_ratelimit(sb)) {
458 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
459 sb->s_id, function, line, current->comm, &vaf);
462 save_error_info(sb, function, line);
463 ext4_handle_error(sb);
466 void __ext4_error_inode(struct inode *inode, const char *function,
467 unsigned int line, ext4_fsblk_t block,
468 const char *fmt, ...)
471 struct va_format vaf;
472 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
474 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
477 trace_ext4_error(inode->i_sb, function, line);
478 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
479 es->s_last_error_block = cpu_to_le64(block);
480 if (ext4_error_ratelimit(inode->i_sb)) {
485 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
486 "inode #%lu: block %llu: comm %s: %pV\n",
487 inode->i_sb->s_id, function, line, inode->i_ino,
488 block, current->comm, &vaf);
490 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
491 "inode #%lu: comm %s: %pV\n",
492 inode->i_sb->s_id, function, line, inode->i_ino,
493 current->comm, &vaf);
496 save_error_info(inode->i_sb, function, line);
497 ext4_handle_error(inode->i_sb);
500 void __ext4_error_file(struct file *file, const char *function,
501 unsigned int line, ext4_fsblk_t block,
502 const char *fmt, ...)
505 struct va_format vaf;
506 struct ext4_super_block *es;
507 struct inode *inode = file_inode(file);
508 char pathname[80], *path;
510 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
513 trace_ext4_error(inode->i_sb, function, line);
514 es = EXT4_SB(inode->i_sb)->s_es;
515 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
516 if (ext4_error_ratelimit(inode->i_sb)) {
517 path = file_path(file, pathname, sizeof(pathname));
525 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
526 "block %llu: comm %s: path %s: %pV\n",
527 inode->i_sb->s_id, function, line, inode->i_ino,
528 block, current->comm, path, &vaf);
531 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
532 "comm %s: path %s: %pV\n",
533 inode->i_sb->s_id, function, line, inode->i_ino,
534 current->comm, path, &vaf);
537 save_error_info(inode->i_sb, function, line);
538 ext4_handle_error(inode->i_sb);
541 const char *ext4_decode_error(struct super_block *sb, int errno,
548 errstr = "Corrupt filesystem";
551 errstr = "Filesystem failed CRC";
554 errstr = "IO failure";
557 errstr = "Out of memory";
560 if (!sb || (EXT4_SB(sb)->s_journal &&
561 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
562 errstr = "Journal has aborted";
564 errstr = "Readonly filesystem";
567 /* If the caller passed in an extra buffer for unknown
568 * errors, textualise them now. Else we just return
571 /* Check for truncated error codes... */
572 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
581 /* __ext4_std_error decodes expected errors from journaling functions
582 * automatically and invokes the appropriate error response. */
584 void __ext4_std_error(struct super_block *sb, const char *function,
585 unsigned int line, int errno)
590 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
593 /* Special case: if the error is EROFS, and we're not already
594 * inside a transaction, then there's really no point in logging
596 if (errno == -EROFS && journal_current_handle() == NULL && sb_rdonly(sb))
599 if (ext4_error_ratelimit(sb)) {
600 errstr = ext4_decode_error(sb, errno, nbuf);
601 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
602 sb->s_id, function, line, errstr);
605 save_error_info(sb, function, line);
606 ext4_handle_error(sb);
610 * ext4_abort is a much stronger failure handler than ext4_error. The
611 * abort function may be used to deal with unrecoverable failures such
612 * as journal IO errors or ENOMEM at a critical moment in log management.
614 * We unconditionally force the filesystem into an ABORT|READONLY state,
615 * unless the error response on the fs has been set to panic in which
616 * case we take the easy way out and panic immediately.
619 void __ext4_abort(struct super_block *sb, const char *function,
620 unsigned int line, const char *fmt, ...)
622 struct va_format vaf;
625 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
628 save_error_info(sb, function, line);
632 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: %pV\n",
633 sb->s_id, function, line, &vaf);
636 if (sb_rdonly(sb) == 0) {
637 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
638 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
640 * Make sure updated value of ->s_mount_flags will be visible
641 * before ->s_flags update
644 sb->s_flags |= SB_RDONLY;
645 if (EXT4_SB(sb)->s_journal)
646 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
647 save_error_info(sb, function, line);
649 if (test_opt(sb, ERRORS_PANIC)) {
650 if (EXT4_SB(sb)->s_journal &&
651 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
653 panic("EXT4-fs panic from previous error\n");
657 void __ext4_msg(struct super_block *sb,
658 const char *prefix, const char *fmt, ...)
660 struct va_format vaf;
663 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
669 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
673 #define ext4_warning_ratelimit(sb) \
674 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \
677 void __ext4_warning(struct super_block *sb, const char *function,
678 unsigned int line, const char *fmt, ...)
680 struct va_format vaf;
683 if (!ext4_warning_ratelimit(sb))
689 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
690 sb->s_id, function, line, &vaf);
694 void __ext4_warning_inode(const struct inode *inode, const char *function,
695 unsigned int line, const char *fmt, ...)
697 struct va_format vaf;
700 if (!ext4_warning_ratelimit(inode->i_sb))
706 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
707 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
708 function, line, inode->i_ino, current->comm, &vaf);
712 void __ext4_grp_locked_error(const char *function, unsigned int line,
713 struct super_block *sb, ext4_group_t grp,
714 unsigned long ino, ext4_fsblk_t block,
715 const char *fmt, ...)
719 struct va_format vaf;
721 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
723 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
726 trace_ext4_error(sb, function, line);
727 es->s_last_error_ino = cpu_to_le32(ino);
728 es->s_last_error_block = cpu_to_le64(block);
729 __save_error_info(sb, function, line);
731 if (ext4_error_ratelimit(sb)) {
735 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
736 sb->s_id, function, line, grp);
738 printk(KERN_CONT "inode %lu: ", ino);
740 printk(KERN_CONT "block %llu:",
741 (unsigned long long) block);
742 printk(KERN_CONT "%pV\n", &vaf);
746 if (test_opt(sb, WARN_ON_ERROR))
749 if (test_opt(sb, ERRORS_CONT)) {
750 ext4_commit_super(sb, 0);
754 ext4_unlock_group(sb, grp);
755 ext4_commit_super(sb, 1);
756 ext4_handle_error(sb);
758 * We only get here in the ERRORS_RO case; relocking the group
759 * may be dangerous, but nothing bad will happen since the
760 * filesystem will have already been marked read/only and the
761 * journal has been aborted. We return 1 as a hint to callers
762 * who might what to use the return value from
763 * ext4_grp_locked_error() to distinguish between the
764 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
765 * aggressively from the ext4 function in question, with a
766 * more appropriate error code.
768 ext4_lock_group(sb, grp);
772 void ext4_mark_group_bitmap_corrupted(struct super_block *sb,
776 struct ext4_sb_info *sbi = EXT4_SB(sb);
777 struct ext4_group_info *grp = ext4_get_group_info(sb, group);
778 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, group, NULL);
780 if ((flags & EXT4_GROUP_INFO_BBITMAP_CORRUPT) &&
781 !EXT4_MB_GRP_BBITMAP_CORRUPT(grp)) {
782 percpu_counter_sub(&sbi->s_freeclusters_counter,
784 set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT,
788 if ((flags & EXT4_GROUP_INFO_IBITMAP_CORRUPT) &&
789 !EXT4_MB_GRP_IBITMAP_CORRUPT(grp)) {
793 count = ext4_free_inodes_count(sb, gdp);
794 percpu_counter_sub(&sbi->s_freeinodes_counter,
797 set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT,
802 void ext4_update_dynamic_rev(struct super_block *sb)
804 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
806 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
810 "updating to rev %d because of new feature flag, "
811 "running e2fsck is recommended",
814 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
815 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
816 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
817 /* leave es->s_feature_*compat flags alone */
818 /* es->s_uuid will be set by e2fsck if empty */
821 * The rest of the superblock fields should be zero, and if not it
822 * means they are likely already in use, so leave them alone. We
823 * can leave it up to e2fsck to clean up any inconsistencies there.
828 * Open the external journal device
830 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
832 struct block_device *bdev;
833 char b[BDEVNAME_SIZE];
835 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
841 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
842 __bdevname(dev, b), PTR_ERR(bdev));
847 * Release the journal device
849 static void ext4_blkdev_put(struct block_device *bdev)
851 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
854 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
856 struct block_device *bdev;
857 bdev = sbi->journal_bdev;
859 ext4_blkdev_put(bdev);
860 sbi->journal_bdev = NULL;
864 static inline struct inode *orphan_list_entry(struct list_head *l)
866 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
869 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
873 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
874 le32_to_cpu(sbi->s_es->s_last_orphan));
876 printk(KERN_ERR "sb_info orphan list:\n");
877 list_for_each(l, &sbi->s_orphan) {
878 struct inode *inode = orphan_list_entry(l);
880 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
881 inode->i_sb->s_id, inode->i_ino, inode,
882 inode->i_mode, inode->i_nlink,
888 static int ext4_quota_off(struct super_block *sb, int type);
890 static inline void ext4_quota_off_umount(struct super_block *sb)
894 /* Use our quota_off function to clear inode flags etc. */
895 for (type = 0; type < EXT4_MAXQUOTAS; type++)
896 ext4_quota_off(sb, type);
899 static inline void ext4_quota_off_umount(struct super_block *sb)
904 static void ext4_put_super(struct super_block *sb)
906 struct ext4_sb_info *sbi = EXT4_SB(sb);
907 struct ext4_super_block *es = sbi->s_es;
911 ext4_unregister_li_request(sb);
912 ext4_quota_off_umount(sb);
914 destroy_workqueue(sbi->rsv_conversion_wq);
916 if (sbi->s_journal) {
917 aborted = is_journal_aborted(sbi->s_journal);
918 err = jbd2_journal_destroy(sbi->s_journal);
919 sbi->s_journal = NULL;
920 if ((err < 0) && !aborted)
921 ext4_abort(sb, "Couldn't clean up the journal");
924 ext4_unregister_sysfs(sb);
925 ext4_es_unregister_shrinker(sbi);
926 del_timer_sync(&sbi->s_err_report);
927 ext4_release_system_zone(sb);
929 ext4_ext_release(sb);
931 if (!sb_rdonly(sb) && !aborted) {
932 ext4_clear_feature_journal_needs_recovery(sb);
933 es->s_state = cpu_to_le16(sbi->s_mount_state);
936 ext4_commit_super(sb, 1);
938 for (i = 0; i < sbi->s_gdb_count; i++)
939 brelse(sbi->s_group_desc[i]);
940 kvfree(sbi->s_group_desc);
941 kvfree(sbi->s_flex_groups);
942 percpu_counter_destroy(&sbi->s_freeclusters_counter);
943 percpu_counter_destroy(&sbi->s_freeinodes_counter);
944 percpu_counter_destroy(&sbi->s_dirs_counter);
945 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
946 percpu_free_rwsem(&sbi->s_journal_flag_rwsem);
948 for (i = 0; i < EXT4_MAXQUOTAS; i++)
949 kfree(sbi->s_qf_names[i]);
952 /* Debugging code just in case the in-memory inode orphan list
953 * isn't empty. The on-disk one can be non-empty if we've
954 * detected an error and taken the fs readonly, but the
955 * in-memory list had better be clean by this point. */
956 if (!list_empty(&sbi->s_orphan))
957 dump_orphan_list(sb, sbi);
958 J_ASSERT(list_empty(&sbi->s_orphan));
960 sync_blockdev(sb->s_bdev);
961 invalidate_bdev(sb->s_bdev);
962 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
964 * Invalidate the journal device's buffers. We don't want them
965 * floating about in memory - the physical journal device may
966 * hotswapped, and it breaks the `ro-after' testing code.
968 sync_blockdev(sbi->journal_bdev);
969 invalidate_bdev(sbi->journal_bdev);
970 ext4_blkdev_remove(sbi);
972 if (sbi->s_ea_inode_cache) {
973 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
974 sbi->s_ea_inode_cache = NULL;
976 if (sbi->s_ea_block_cache) {
977 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
978 sbi->s_ea_block_cache = NULL;
981 kthread_stop(sbi->s_mmp_tsk);
983 sb->s_fs_info = NULL;
985 * Now that we are completely done shutting down the
986 * superblock, we need to actually destroy the kobject.
988 kobject_put(&sbi->s_kobj);
989 wait_for_completion(&sbi->s_kobj_unregister);
990 if (sbi->s_chksum_driver)
991 crypto_free_shash(sbi->s_chksum_driver);
992 kfree(sbi->s_blockgroup_lock);
993 fs_put_dax(sbi->s_daxdev);
997 static struct kmem_cache *ext4_inode_cachep;
1000 * Called inside transaction, so use GFP_NOFS
1002 static struct inode *ext4_alloc_inode(struct super_block *sb)
1004 struct ext4_inode_info *ei;
1006 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
1010 inode_set_iversion(&ei->vfs_inode, 1);
1011 spin_lock_init(&ei->i_raw_lock);
1012 INIT_LIST_HEAD(&ei->i_prealloc_list);
1013 spin_lock_init(&ei->i_prealloc_lock);
1014 ext4_es_init_tree(&ei->i_es_tree);
1015 rwlock_init(&ei->i_es_lock);
1016 INIT_LIST_HEAD(&ei->i_es_list);
1017 ei->i_es_all_nr = 0;
1018 ei->i_es_shk_nr = 0;
1019 ei->i_es_shrink_lblk = 0;
1020 ei->i_reserved_data_blocks = 0;
1021 ei->i_da_metadata_calc_len = 0;
1022 ei->i_da_metadata_calc_last_lblock = 0;
1023 spin_lock_init(&(ei->i_block_reservation_lock));
1025 ei->i_reserved_quota = 0;
1026 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
1029 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
1030 spin_lock_init(&ei->i_completed_io_lock);
1032 ei->i_datasync_tid = 0;
1033 atomic_set(&ei->i_unwritten, 0);
1034 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
1035 return &ei->vfs_inode;
1038 static int ext4_drop_inode(struct inode *inode)
1040 int drop = generic_drop_inode(inode);
1042 trace_ext4_drop_inode(inode, drop);
1046 static void ext4_i_callback(struct rcu_head *head)
1048 struct inode *inode = container_of(head, struct inode, i_rcu);
1049 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
1052 static void ext4_destroy_inode(struct inode *inode)
1054 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
1055 ext4_msg(inode->i_sb, KERN_ERR,
1056 "Inode %lu (%p): orphan list check failed!",
1057 inode->i_ino, EXT4_I(inode));
1058 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1059 EXT4_I(inode), sizeof(struct ext4_inode_info),
1063 call_rcu(&inode->i_rcu, ext4_i_callback);
1066 static void init_once(void *foo)
1068 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
1070 INIT_LIST_HEAD(&ei->i_orphan);
1071 init_rwsem(&ei->xattr_sem);
1072 init_rwsem(&ei->i_data_sem);
1073 init_rwsem(&ei->i_mmap_sem);
1074 inode_init_once(&ei->vfs_inode);
1077 static int __init init_inodecache(void)
1079 ext4_inode_cachep = kmem_cache_create_usercopy("ext4_inode_cache",
1080 sizeof(struct ext4_inode_info), 0,
1081 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD|
1083 offsetof(struct ext4_inode_info, i_data),
1084 sizeof_field(struct ext4_inode_info, i_data),
1086 if (ext4_inode_cachep == NULL)
1091 static void destroy_inodecache(void)
1094 * Make sure all delayed rcu free inodes are flushed before we
1098 kmem_cache_destroy(ext4_inode_cachep);
1101 void ext4_clear_inode(struct inode *inode)
1103 invalidate_inode_buffers(inode);
1106 ext4_discard_preallocations(inode);
1107 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
1108 if (EXT4_I(inode)->jinode) {
1109 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1110 EXT4_I(inode)->jinode);
1111 jbd2_free_inode(EXT4_I(inode)->jinode);
1112 EXT4_I(inode)->jinode = NULL;
1114 fscrypt_put_encryption_info(inode);
1117 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1118 u64 ino, u32 generation)
1120 struct inode *inode;
1122 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1123 return ERR_PTR(-ESTALE);
1124 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1125 return ERR_PTR(-ESTALE);
1127 /* iget isn't really right if the inode is currently unallocated!!
1129 * ext4_read_inode will return a bad_inode if the inode had been
1130 * deleted, so we should be safe.
1132 * Currently we don't know the generation for parent directory, so
1133 * a generation of 0 means "accept any"
1135 inode = ext4_iget_normal(sb, ino);
1137 return ERR_CAST(inode);
1138 if (generation && inode->i_generation != generation) {
1140 return ERR_PTR(-ESTALE);
1146 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1147 int fh_len, int fh_type)
1149 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1150 ext4_nfs_get_inode);
1153 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1154 int fh_len, int fh_type)
1156 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1157 ext4_nfs_get_inode);
1161 * Try to release metadata pages (indirect blocks, directories) which are
1162 * mapped via the block device. Since these pages could have journal heads
1163 * which would prevent try_to_free_buffers() from freeing them, we must use
1164 * jbd2 layer's try_to_free_buffers() function to release them.
1166 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1169 journal_t *journal = EXT4_SB(sb)->s_journal;
1171 WARN_ON(PageChecked(page));
1172 if (!page_has_buffers(page))
1175 return jbd2_journal_try_to_free_buffers(journal, page,
1176 wait & ~__GFP_DIRECT_RECLAIM);
1177 return try_to_free_buffers(page);
1180 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1181 static int ext4_get_context(struct inode *inode, void *ctx, size_t len)
1183 return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1184 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len);
1187 static int ext4_set_context(struct inode *inode, const void *ctx, size_t len,
1190 handle_t *handle = fs_data;
1191 int res, res2, credits, retries = 0;
1194 * Encrypting the root directory is not allowed because e2fsck expects
1195 * lost+found to exist and be unencrypted, and encrypting the root
1196 * directory would imply encrypting the lost+found directory as well as
1197 * the filename "lost+found" itself.
1199 if (inode->i_ino == EXT4_ROOT_INO)
1202 if (WARN_ON_ONCE(IS_DAX(inode) && i_size_read(inode)))
1205 res = ext4_convert_inline_data(inode);
1210 * If a journal handle was specified, then the encryption context is
1211 * being set on a new inode via inheritance and is part of a larger
1212 * transaction to create the inode. Otherwise the encryption context is
1213 * being set on an existing inode in its own transaction. Only in the
1214 * latter case should the "retry on ENOSPC" logic be used.
1218 res = ext4_xattr_set_handle(handle, inode,
1219 EXT4_XATTR_INDEX_ENCRYPTION,
1220 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1223 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1224 ext4_clear_inode_state(inode,
1225 EXT4_STATE_MAY_INLINE_DATA);
1227 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1228 * S_DAX may be disabled
1230 ext4_set_inode_flags(inode);
1235 res = dquot_initialize(inode);
1239 res = ext4_xattr_set_credits(inode, len, false /* is_create */,
1244 handle = ext4_journal_start(inode, EXT4_HT_MISC, credits);
1246 return PTR_ERR(handle);
1248 res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION,
1249 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1252 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1254 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1255 * S_DAX may be disabled
1257 ext4_set_inode_flags(inode);
1258 res = ext4_mark_inode_dirty(handle, inode);
1260 EXT4_ERROR_INODE(inode, "Failed to mark inode dirty");
1262 res2 = ext4_journal_stop(handle);
1264 if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
1271 static bool ext4_dummy_context(struct inode *inode)
1273 return DUMMY_ENCRYPTION_ENABLED(EXT4_SB(inode->i_sb));
1276 static unsigned ext4_max_namelen(struct inode *inode)
1278 return S_ISLNK(inode->i_mode) ? inode->i_sb->s_blocksize :
1282 static const struct fscrypt_operations ext4_cryptops = {
1283 .key_prefix = "ext4:",
1284 .get_context = ext4_get_context,
1285 .set_context = ext4_set_context,
1286 .dummy_context = ext4_dummy_context,
1287 .empty_dir = ext4_empty_dir,
1288 .max_namelen = ext4_max_namelen,
1293 static const char * const quotatypes[] = INITQFNAMES;
1294 #define QTYPE2NAME(t) (quotatypes[t])
1296 static int ext4_write_dquot(struct dquot *dquot);
1297 static int ext4_acquire_dquot(struct dquot *dquot);
1298 static int ext4_release_dquot(struct dquot *dquot);
1299 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1300 static int ext4_write_info(struct super_block *sb, int type);
1301 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1302 const struct path *path);
1303 static int ext4_quota_on_mount(struct super_block *sb, int type);
1304 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1305 size_t len, loff_t off);
1306 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1307 const char *data, size_t len, loff_t off);
1308 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1309 unsigned int flags);
1310 static int ext4_enable_quotas(struct super_block *sb);
1311 static int ext4_get_next_id(struct super_block *sb, struct kqid *qid);
1313 static struct dquot **ext4_get_dquots(struct inode *inode)
1315 return EXT4_I(inode)->i_dquot;
1318 static const struct dquot_operations ext4_quota_operations = {
1319 .get_reserved_space = ext4_get_reserved_space,
1320 .write_dquot = ext4_write_dquot,
1321 .acquire_dquot = ext4_acquire_dquot,
1322 .release_dquot = ext4_release_dquot,
1323 .mark_dirty = ext4_mark_dquot_dirty,
1324 .write_info = ext4_write_info,
1325 .alloc_dquot = dquot_alloc,
1326 .destroy_dquot = dquot_destroy,
1327 .get_projid = ext4_get_projid,
1328 .get_inode_usage = ext4_get_inode_usage,
1329 .get_next_id = ext4_get_next_id,
1332 static const struct quotactl_ops ext4_qctl_operations = {
1333 .quota_on = ext4_quota_on,
1334 .quota_off = ext4_quota_off,
1335 .quota_sync = dquot_quota_sync,
1336 .get_state = dquot_get_state,
1337 .set_info = dquot_set_dqinfo,
1338 .get_dqblk = dquot_get_dqblk,
1339 .set_dqblk = dquot_set_dqblk,
1340 .get_nextdqblk = dquot_get_next_dqblk,
1344 static const struct super_operations ext4_sops = {
1345 .alloc_inode = ext4_alloc_inode,
1346 .destroy_inode = ext4_destroy_inode,
1347 .write_inode = ext4_write_inode,
1348 .dirty_inode = ext4_dirty_inode,
1349 .drop_inode = ext4_drop_inode,
1350 .evict_inode = ext4_evict_inode,
1351 .put_super = ext4_put_super,
1352 .sync_fs = ext4_sync_fs,
1353 .freeze_fs = ext4_freeze,
1354 .unfreeze_fs = ext4_unfreeze,
1355 .statfs = ext4_statfs,
1356 .remount_fs = ext4_remount,
1357 .show_options = ext4_show_options,
1359 .quota_read = ext4_quota_read,
1360 .quota_write = ext4_quota_write,
1361 .get_dquots = ext4_get_dquots,
1363 .bdev_try_to_free_page = bdev_try_to_free_page,
1366 static const struct export_operations ext4_export_ops = {
1367 .fh_to_dentry = ext4_fh_to_dentry,
1368 .fh_to_parent = ext4_fh_to_parent,
1369 .get_parent = ext4_get_parent,
1373 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1374 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1375 Opt_nouid32, Opt_debug, Opt_removed,
1376 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1377 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1378 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1379 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1380 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1381 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1382 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1383 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1384 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1385 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version, Opt_dax,
1386 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_warn_on_error,
1387 Opt_nowarn_on_error, Opt_mblk_io_submit,
1388 Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize,
1389 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1390 Opt_inode_readahead_blks, Opt_journal_ioprio,
1391 Opt_dioread_nolock, Opt_dioread_lock,
1392 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1393 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache,
1396 static const match_table_t tokens = {
1397 {Opt_bsd_df, "bsddf"},
1398 {Opt_minix_df, "minixdf"},
1399 {Opt_grpid, "grpid"},
1400 {Opt_grpid, "bsdgroups"},
1401 {Opt_nogrpid, "nogrpid"},
1402 {Opt_nogrpid, "sysvgroups"},
1403 {Opt_resgid, "resgid=%u"},
1404 {Opt_resuid, "resuid=%u"},
1406 {Opt_err_cont, "errors=continue"},
1407 {Opt_err_panic, "errors=panic"},
1408 {Opt_err_ro, "errors=remount-ro"},
1409 {Opt_nouid32, "nouid32"},
1410 {Opt_debug, "debug"},
1411 {Opt_removed, "oldalloc"},
1412 {Opt_removed, "orlov"},
1413 {Opt_user_xattr, "user_xattr"},
1414 {Opt_nouser_xattr, "nouser_xattr"},
1416 {Opt_noacl, "noacl"},
1417 {Opt_noload, "norecovery"},
1418 {Opt_noload, "noload"},
1419 {Opt_removed, "nobh"},
1420 {Opt_removed, "bh"},
1421 {Opt_commit, "commit=%u"},
1422 {Opt_min_batch_time, "min_batch_time=%u"},
1423 {Opt_max_batch_time, "max_batch_time=%u"},
1424 {Opt_journal_dev, "journal_dev=%u"},
1425 {Opt_journal_path, "journal_path=%s"},
1426 {Opt_journal_checksum, "journal_checksum"},
1427 {Opt_nojournal_checksum, "nojournal_checksum"},
1428 {Opt_journal_async_commit, "journal_async_commit"},
1429 {Opt_abort, "abort"},
1430 {Opt_data_journal, "data=journal"},
1431 {Opt_data_ordered, "data=ordered"},
1432 {Opt_data_writeback, "data=writeback"},
1433 {Opt_data_err_abort, "data_err=abort"},
1434 {Opt_data_err_ignore, "data_err=ignore"},
1435 {Opt_offusrjquota, "usrjquota="},
1436 {Opt_usrjquota, "usrjquota=%s"},
1437 {Opt_offgrpjquota, "grpjquota="},
1438 {Opt_grpjquota, "grpjquota=%s"},
1439 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1440 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1441 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1442 {Opt_grpquota, "grpquota"},
1443 {Opt_noquota, "noquota"},
1444 {Opt_quota, "quota"},
1445 {Opt_usrquota, "usrquota"},
1446 {Opt_prjquota, "prjquota"},
1447 {Opt_barrier, "barrier=%u"},
1448 {Opt_barrier, "barrier"},
1449 {Opt_nobarrier, "nobarrier"},
1450 {Opt_i_version, "i_version"},
1452 {Opt_stripe, "stripe=%u"},
1453 {Opt_delalloc, "delalloc"},
1454 {Opt_warn_on_error, "warn_on_error"},
1455 {Opt_nowarn_on_error, "nowarn_on_error"},
1456 {Opt_lazytime, "lazytime"},
1457 {Opt_nolazytime, "nolazytime"},
1458 {Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"},
1459 {Opt_nodelalloc, "nodelalloc"},
1460 {Opt_removed, "mblk_io_submit"},
1461 {Opt_removed, "nomblk_io_submit"},
1462 {Opt_block_validity, "block_validity"},
1463 {Opt_noblock_validity, "noblock_validity"},
1464 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1465 {Opt_journal_ioprio, "journal_ioprio=%u"},
1466 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1467 {Opt_auto_da_alloc, "auto_da_alloc"},
1468 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1469 {Opt_dioread_nolock, "dioread_nolock"},
1470 {Opt_dioread_lock, "dioread_lock"},
1471 {Opt_discard, "discard"},
1472 {Opt_nodiscard, "nodiscard"},
1473 {Opt_init_itable, "init_itable=%u"},
1474 {Opt_init_itable, "init_itable"},
1475 {Opt_noinit_itable, "noinit_itable"},
1476 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1477 {Opt_test_dummy_encryption, "test_dummy_encryption"},
1478 {Opt_nombcache, "nombcache"},
1479 {Opt_nombcache, "no_mbcache"}, /* for backward compatibility */
1480 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1481 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1482 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1483 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1484 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1488 static ext4_fsblk_t get_sb_block(void **data)
1490 ext4_fsblk_t sb_block;
1491 char *options = (char *) *data;
1493 if (!options || strncmp(options, "sb=", 3) != 0)
1494 return 1; /* Default location */
1497 /* TODO: use simple_strtoll with >32bit ext4 */
1498 sb_block = simple_strtoul(options, &options, 0);
1499 if (*options && *options != ',') {
1500 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1504 if (*options == ',')
1506 *data = (void *) options;
1511 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1512 static const char deprecated_msg[] =
1513 "Mount option \"%s\" will be removed by %s\n"
1514 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1517 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1519 struct ext4_sb_info *sbi = EXT4_SB(sb);
1523 if (sb_any_quota_loaded(sb) &&
1524 !sbi->s_qf_names[qtype]) {
1525 ext4_msg(sb, KERN_ERR,
1526 "Cannot change journaled "
1527 "quota options when quota turned on");
1530 if (ext4_has_feature_quota(sb)) {
1531 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1532 "ignored when QUOTA feature is enabled");
1535 qname = match_strdup(args);
1537 ext4_msg(sb, KERN_ERR,
1538 "Not enough memory for storing quotafile name");
1541 if (sbi->s_qf_names[qtype]) {
1542 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1545 ext4_msg(sb, KERN_ERR,
1546 "%s quota file already specified",
1550 if (strchr(qname, '/')) {
1551 ext4_msg(sb, KERN_ERR,
1552 "quotafile must be on filesystem root");
1555 sbi->s_qf_names[qtype] = qname;
1563 static int clear_qf_name(struct super_block *sb, int qtype)
1566 struct ext4_sb_info *sbi = EXT4_SB(sb);
1568 if (sb_any_quota_loaded(sb) &&
1569 sbi->s_qf_names[qtype]) {
1570 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1571 " when quota turned on");
1574 kfree(sbi->s_qf_names[qtype]);
1575 sbi->s_qf_names[qtype] = NULL;
1580 #define MOPT_SET 0x0001
1581 #define MOPT_CLEAR 0x0002
1582 #define MOPT_NOSUPPORT 0x0004
1583 #define MOPT_EXPLICIT 0x0008
1584 #define MOPT_CLEAR_ERR 0x0010
1585 #define MOPT_GTE0 0x0020
1588 #define MOPT_QFMT 0x0040
1590 #define MOPT_Q MOPT_NOSUPPORT
1591 #define MOPT_QFMT MOPT_NOSUPPORT
1593 #define MOPT_DATAJ 0x0080
1594 #define MOPT_NO_EXT2 0x0100
1595 #define MOPT_NO_EXT3 0x0200
1596 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1597 #define MOPT_STRING 0x0400
1599 static const struct mount_opts {
1603 } ext4_mount_opts[] = {
1604 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1605 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1606 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1607 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1608 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1609 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1610 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1611 MOPT_EXT4_ONLY | MOPT_SET},
1612 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1613 MOPT_EXT4_ONLY | MOPT_CLEAR},
1614 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1615 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1616 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1617 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1618 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1619 MOPT_EXT4_ONLY | MOPT_CLEAR},
1620 {Opt_warn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_SET},
1621 {Opt_nowarn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_CLEAR},
1622 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1623 MOPT_EXT4_ONLY | MOPT_CLEAR},
1624 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1625 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1626 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1627 EXT4_MOUNT_JOURNAL_CHECKSUM),
1628 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1629 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1630 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1631 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1632 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1633 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1635 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1637 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1638 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1639 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1640 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1641 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1642 {Opt_commit, 0, MOPT_GTE0},
1643 {Opt_max_batch_time, 0, MOPT_GTE0},
1644 {Opt_min_batch_time, 0, MOPT_GTE0},
1645 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1646 {Opt_init_itable, 0, MOPT_GTE0},
1647 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
1648 {Opt_stripe, 0, MOPT_GTE0},
1649 {Opt_resuid, 0, MOPT_GTE0},
1650 {Opt_resgid, 0, MOPT_GTE0},
1651 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1652 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1653 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1654 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1655 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1656 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1657 MOPT_NO_EXT2 | MOPT_DATAJ},
1658 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1659 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1660 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1661 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1662 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1664 {Opt_acl, 0, MOPT_NOSUPPORT},
1665 {Opt_noacl, 0, MOPT_NOSUPPORT},
1667 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1668 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1669 {Opt_debug_want_extra_isize, 0, MOPT_GTE0},
1670 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1671 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1673 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1675 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1677 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1678 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1679 MOPT_CLEAR | MOPT_Q},
1680 {Opt_usrjquota, 0, MOPT_Q},
1681 {Opt_grpjquota, 0, MOPT_Q},
1682 {Opt_offusrjquota, 0, MOPT_Q},
1683 {Opt_offgrpjquota, 0, MOPT_Q},
1684 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1685 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1686 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1687 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1688 {Opt_test_dummy_encryption, 0, MOPT_GTE0},
1689 {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET},
1693 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1694 substring_t *args, unsigned long *journal_devnum,
1695 unsigned int *journal_ioprio, int is_remount)
1697 struct ext4_sb_info *sbi = EXT4_SB(sb);
1698 const struct mount_opts *m;
1704 if (token == Opt_usrjquota)
1705 return set_qf_name(sb, USRQUOTA, &args[0]);
1706 else if (token == Opt_grpjquota)
1707 return set_qf_name(sb, GRPQUOTA, &args[0]);
1708 else if (token == Opt_offusrjquota)
1709 return clear_qf_name(sb, USRQUOTA);
1710 else if (token == Opt_offgrpjquota)
1711 return clear_qf_name(sb, GRPQUOTA);
1715 case Opt_nouser_xattr:
1716 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1719 return 1; /* handled by get_sb_block() */
1721 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1724 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1727 sb->s_flags |= SB_I_VERSION;
1730 sb->s_flags |= SB_LAZYTIME;
1732 case Opt_nolazytime:
1733 sb->s_flags &= ~SB_LAZYTIME;
1737 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1738 if (token == m->token)
1741 if (m->token == Opt_err) {
1742 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1743 "or missing value", opt);
1747 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1748 ext4_msg(sb, KERN_ERR,
1749 "Mount option \"%s\" incompatible with ext2", opt);
1752 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1753 ext4_msg(sb, KERN_ERR,
1754 "Mount option \"%s\" incompatible with ext3", opt);
1758 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
1760 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1762 if (m->flags & MOPT_EXPLICIT) {
1763 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
1764 set_opt2(sb, EXPLICIT_DELALLOC);
1765 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
1766 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
1770 if (m->flags & MOPT_CLEAR_ERR)
1771 clear_opt(sb, ERRORS_MASK);
1772 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1773 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1774 "options when quota turned on");
1778 if (m->flags & MOPT_NOSUPPORT) {
1779 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1780 } else if (token == Opt_commit) {
1782 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1783 sbi->s_commit_interval = HZ * arg;
1784 } else if (token == Opt_debug_want_extra_isize) {
1785 sbi->s_want_extra_isize = arg;
1786 } else if (token == Opt_max_batch_time) {
1787 sbi->s_max_batch_time = arg;
1788 } else if (token == Opt_min_batch_time) {
1789 sbi->s_min_batch_time = arg;
1790 } else if (token == Opt_inode_readahead_blks) {
1791 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1792 ext4_msg(sb, KERN_ERR,
1793 "EXT4-fs: inode_readahead_blks must be "
1794 "0 or a power of 2 smaller than 2^31");
1797 sbi->s_inode_readahead_blks = arg;
1798 } else if (token == Opt_init_itable) {
1799 set_opt(sb, INIT_INODE_TABLE);
1801 arg = EXT4_DEF_LI_WAIT_MULT;
1802 sbi->s_li_wait_mult = arg;
1803 } else if (token == Opt_max_dir_size_kb) {
1804 sbi->s_max_dir_size_kb = arg;
1805 } else if (token == Opt_stripe) {
1806 sbi->s_stripe = arg;
1807 } else if (token == Opt_resuid) {
1808 uid = make_kuid(current_user_ns(), arg);
1809 if (!uid_valid(uid)) {
1810 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1813 sbi->s_resuid = uid;
1814 } else if (token == Opt_resgid) {
1815 gid = make_kgid(current_user_ns(), arg);
1816 if (!gid_valid(gid)) {
1817 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1820 sbi->s_resgid = gid;
1821 } else if (token == Opt_journal_dev) {
1823 ext4_msg(sb, KERN_ERR,
1824 "Cannot specify journal on remount");
1827 *journal_devnum = arg;
1828 } else if (token == Opt_journal_path) {
1830 struct inode *journal_inode;
1835 ext4_msg(sb, KERN_ERR,
1836 "Cannot specify journal on remount");
1839 journal_path = match_strdup(&args[0]);
1840 if (!journal_path) {
1841 ext4_msg(sb, KERN_ERR, "error: could not dup "
1842 "journal device string");
1846 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1848 ext4_msg(sb, KERN_ERR, "error: could not find "
1849 "journal device path: error %d", error);
1850 kfree(journal_path);
1854 journal_inode = d_inode(path.dentry);
1855 if (!S_ISBLK(journal_inode->i_mode)) {
1856 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1857 "is not a block device", journal_path);
1859 kfree(journal_path);
1863 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1865 kfree(journal_path);
1866 } else if (token == Opt_journal_ioprio) {
1868 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1873 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1874 } else if (token == Opt_test_dummy_encryption) {
1875 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1876 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1877 ext4_msg(sb, KERN_WARNING,
1878 "Test dummy encryption mode enabled");
1880 ext4_msg(sb, KERN_WARNING,
1881 "Test dummy encryption mount option ignored");
1883 } else if (m->flags & MOPT_DATAJ) {
1885 if (!sbi->s_journal)
1886 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1887 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1888 ext4_msg(sb, KERN_ERR,
1889 "Cannot change data mode on remount");
1893 clear_opt(sb, DATA_FLAGS);
1894 sbi->s_mount_opt |= m->mount_opt;
1897 } else if (m->flags & MOPT_QFMT) {
1898 if (sb_any_quota_loaded(sb) &&
1899 sbi->s_jquota_fmt != m->mount_opt) {
1900 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1901 "quota options when quota turned on");
1904 if (ext4_has_feature_quota(sb)) {
1905 ext4_msg(sb, KERN_INFO,
1906 "Quota format mount options ignored "
1907 "when QUOTA feature is enabled");
1910 sbi->s_jquota_fmt = m->mount_opt;
1912 } else if (token == Opt_dax) {
1913 #ifdef CONFIG_FS_DAX
1914 ext4_msg(sb, KERN_WARNING,
1915 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
1916 sbi->s_mount_opt |= m->mount_opt;
1918 ext4_msg(sb, KERN_INFO, "dax option not supported");
1921 } else if (token == Opt_data_err_abort) {
1922 sbi->s_mount_opt |= m->mount_opt;
1923 } else if (token == Opt_data_err_ignore) {
1924 sbi->s_mount_opt &= ~m->mount_opt;
1928 if (m->flags & MOPT_CLEAR)
1930 else if (unlikely(!(m->flags & MOPT_SET))) {
1931 ext4_msg(sb, KERN_WARNING,
1932 "buggy handling of option %s", opt);
1937 sbi->s_mount_opt |= m->mount_opt;
1939 sbi->s_mount_opt &= ~m->mount_opt;
1944 static int parse_options(char *options, struct super_block *sb,
1945 unsigned long *journal_devnum,
1946 unsigned int *journal_ioprio,
1949 struct ext4_sb_info *sbi = EXT4_SB(sb);
1951 substring_t args[MAX_OPT_ARGS];
1957 while ((p = strsep(&options, ",")) != NULL) {
1961 * Initialize args struct so we know whether arg was
1962 * found; some options take optional arguments.
1964 args[0].to = args[0].from = NULL;
1965 token = match_token(p, tokens, args);
1966 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1967 journal_ioprio, is_remount) < 0)
1972 * We do the test below only for project quotas. 'usrquota' and
1973 * 'grpquota' mount options are allowed even without quota feature
1974 * to support legacy quotas in quota files.
1976 if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) {
1977 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. "
1978 "Cannot enable project quota enforcement.");
1981 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1982 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1983 clear_opt(sb, USRQUOTA);
1985 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1986 clear_opt(sb, GRPQUOTA);
1988 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1989 ext4_msg(sb, KERN_ERR, "old and new quota "
1994 if (!sbi->s_jquota_fmt) {
1995 ext4_msg(sb, KERN_ERR, "journaled quota format "
2001 if (test_opt(sb, DIOREAD_NOLOCK)) {
2003 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
2005 if (blocksize < PAGE_SIZE) {
2006 ext4_msg(sb, KERN_ERR, "can't mount with "
2007 "dioread_nolock if block size != PAGE_SIZE");
2014 static inline void ext4_show_quota_options(struct seq_file *seq,
2015 struct super_block *sb)
2017 #if defined(CONFIG_QUOTA)
2018 struct ext4_sb_info *sbi = EXT4_SB(sb);
2020 if (sbi->s_jquota_fmt) {
2023 switch (sbi->s_jquota_fmt) {
2034 seq_printf(seq, ",jqfmt=%s", fmtname);
2037 if (sbi->s_qf_names[USRQUOTA])
2038 seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]);
2040 if (sbi->s_qf_names[GRPQUOTA])
2041 seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]);
2045 static const char *token2str(int token)
2047 const struct match_token *t;
2049 for (t = tokens; t->token != Opt_err; t++)
2050 if (t->token == token && !strchr(t->pattern, '='))
2057 * - it's set to a non-default value OR
2058 * - if the per-sb default is different from the global default
2060 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
2063 struct ext4_sb_info *sbi = EXT4_SB(sb);
2064 struct ext4_super_block *es = sbi->s_es;
2065 int def_errors, def_mount_opt = sbi->s_def_mount_opt;
2066 const struct mount_opts *m;
2067 char sep = nodefs ? '\n' : ',';
2069 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2070 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2072 if (sbi->s_sb_block != 1)
2073 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
2075 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
2076 int want_set = m->flags & MOPT_SET;
2077 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
2078 (m->flags & MOPT_CLEAR_ERR))
2080 if (!nodefs && !(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
2081 continue; /* skip if same as the default */
2083 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
2084 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
2085 continue; /* select Opt_noFoo vs Opt_Foo */
2086 SEQ_OPTS_PRINT("%s", token2str(m->token));
2089 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
2090 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
2091 SEQ_OPTS_PRINT("resuid=%u",
2092 from_kuid_munged(&init_user_ns, sbi->s_resuid));
2093 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
2094 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
2095 SEQ_OPTS_PRINT("resgid=%u",
2096 from_kgid_munged(&init_user_ns, sbi->s_resgid));
2097 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
2098 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
2099 SEQ_OPTS_PUTS("errors=remount-ro");
2100 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
2101 SEQ_OPTS_PUTS("errors=continue");
2102 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
2103 SEQ_OPTS_PUTS("errors=panic");
2104 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
2105 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
2106 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
2107 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
2108 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
2109 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2110 if (sb->s_flags & SB_I_VERSION)
2111 SEQ_OPTS_PUTS("i_version");
2112 if (nodefs || sbi->s_stripe)
2113 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
2114 if (nodefs || EXT4_MOUNT_DATA_FLAGS &
2115 (sbi->s_mount_opt ^ def_mount_opt)) {
2116 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2117 SEQ_OPTS_PUTS("data=journal");
2118 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2119 SEQ_OPTS_PUTS("data=ordered");
2120 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2121 SEQ_OPTS_PUTS("data=writeback");
2124 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
2125 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2126 sbi->s_inode_readahead_blks);
2128 if (test_opt(sb, INIT_INODE_TABLE) && (nodefs ||
2129 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
2130 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
2131 if (nodefs || sbi->s_max_dir_size_kb)
2132 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2133 if (test_opt(sb, DATA_ERR_ABORT))
2134 SEQ_OPTS_PUTS("data_err=abort");
2136 ext4_show_quota_options(seq, sb);
2140 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2142 return _ext4_show_options(seq, root->d_sb, 0);
2145 int ext4_seq_options_show(struct seq_file *seq, void *offset)
2147 struct super_block *sb = seq->private;
2150 seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw");
2151 rc = _ext4_show_options(seq, sb, 1);
2152 seq_puts(seq, "\n");
2156 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
2159 struct ext4_sb_info *sbi = EXT4_SB(sb);
2162 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
2163 ext4_msg(sb, KERN_ERR, "revision level too high, "
2164 "forcing read-only mode");
2169 if (!(sbi->s_mount_state & EXT4_VALID_FS))
2170 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
2171 "running e2fsck is recommended");
2172 else if (sbi->s_mount_state & EXT4_ERROR_FS)
2173 ext4_msg(sb, KERN_WARNING,
2174 "warning: mounting fs with errors, "
2175 "running e2fsck is recommended");
2176 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
2177 le16_to_cpu(es->s_mnt_count) >=
2178 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
2179 ext4_msg(sb, KERN_WARNING,
2180 "warning: maximal mount count reached, "
2181 "running e2fsck is recommended");
2182 else if (le32_to_cpu(es->s_checkinterval) &&
2183 (le32_to_cpu(es->s_lastcheck) +
2184 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
2185 ext4_msg(sb, KERN_WARNING,
2186 "warning: checktime reached, "
2187 "running e2fsck is recommended");
2188 if (!sbi->s_journal)
2189 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
2190 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
2191 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
2192 le16_add_cpu(&es->s_mnt_count, 1);
2193 es->s_mtime = cpu_to_le32(get_seconds());
2194 ext4_update_dynamic_rev(sb);
2196 ext4_set_feature_journal_needs_recovery(sb);
2198 err = ext4_commit_super(sb, 1);
2200 if (test_opt(sb, DEBUG))
2201 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
2202 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
2204 sbi->s_groups_count,
2205 EXT4_BLOCKS_PER_GROUP(sb),
2206 EXT4_INODES_PER_GROUP(sb),
2207 sbi->s_mount_opt, sbi->s_mount_opt2);
2209 cleancache_init_fs(sb);
2213 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2215 struct ext4_sb_info *sbi = EXT4_SB(sb);
2216 struct flex_groups *new_groups;
2219 if (!sbi->s_log_groups_per_flex)
2222 size = ext4_flex_group(sbi, ngroup - 1) + 1;
2223 if (size <= sbi->s_flex_groups_allocated)
2226 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
2227 new_groups = kvzalloc(size, GFP_KERNEL);
2229 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
2230 size / (int) sizeof(struct flex_groups));
2234 if (sbi->s_flex_groups) {
2235 memcpy(new_groups, sbi->s_flex_groups,
2236 (sbi->s_flex_groups_allocated *
2237 sizeof(struct flex_groups)));
2238 kvfree(sbi->s_flex_groups);
2240 sbi->s_flex_groups = new_groups;
2241 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
2245 static int ext4_fill_flex_info(struct super_block *sb)
2247 struct ext4_sb_info *sbi = EXT4_SB(sb);
2248 struct ext4_group_desc *gdp = NULL;
2249 ext4_group_t flex_group;
2252 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2253 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
2254 sbi->s_log_groups_per_flex = 0;
2258 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2262 for (i = 0; i < sbi->s_groups_count; i++) {
2263 gdp = ext4_get_group_desc(sb, i, NULL);
2265 flex_group = ext4_flex_group(sbi, i);
2266 atomic_add(ext4_free_inodes_count(sb, gdp),
2267 &sbi->s_flex_groups[flex_group].free_inodes);
2268 atomic64_add(ext4_free_group_clusters(sb, gdp),
2269 &sbi->s_flex_groups[flex_group].free_clusters);
2270 atomic_add(ext4_used_dirs_count(sb, gdp),
2271 &sbi->s_flex_groups[flex_group].used_dirs);
2279 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
2280 struct ext4_group_desc *gdp)
2282 int offset = offsetof(struct ext4_group_desc, bg_checksum);
2284 __le32 le_group = cpu_to_le32(block_group);
2285 struct ext4_sb_info *sbi = EXT4_SB(sb);
2287 if (ext4_has_metadata_csum(sbi->s_sb)) {
2288 /* Use new metadata_csum algorithm */
2290 __u16 dummy_csum = 0;
2292 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2294 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2295 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2296 sizeof(dummy_csum));
2297 offset += sizeof(dummy_csum);
2298 if (offset < sbi->s_desc_size)
2299 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2300 sbi->s_desc_size - offset);
2302 crc = csum32 & 0xFFFF;
2306 /* old crc16 code */
2307 if (!ext4_has_feature_gdt_csum(sb))
2310 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2311 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2312 crc = crc16(crc, (__u8 *)gdp, offset);
2313 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2314 /* for checksum of struct ext4_group_desc do the rest...*/
2315 if (ext4_has_feature_64bit(sb) &&
2316 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2317 crc = crc16(crc, (__u8 *)gdp + offset,
2318 le16_to_cpu(sbi->s_es->s_desc_size) -
2322 return cpu_to_le16(crc);
2325 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2326 struct ext4_group_desc *gdp)
2328 if (ext4_has_group_desc_csum(sb) &&
2329 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
2335 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2336 struct ext4_group_desc *gdp)
2338 if (!ext4_has_group_desc_csum(sb))
2340 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
2343 /* Called at mount-time, super-block is locked */
2344 static int ext4_check_descriptors(struct super_block *sb,
2345 ext4_fsblk_t sb_block,
2346 ext4_group_t *first_not_zeroed)
2348 struct ext4_sb_info *sbi = EXT4_SB(sb);
2349 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2350 ext4_fsblk_t last_block;
2351 ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0) + 1;
2352 ext4_fsblk_t block_bitmap;
2353 ext4_fsblk_t inode_bitmap;
2354 ext4_fsblk_t inode_table;
2355 int flexbg_flag = 0;
2356 ext4_group_t i, grp = sbi->s_groups_count;
2358 if (ext4_has_feature_flex_bg(sb))
2361 ext4_debug("Checking group descriptors");
2363 for (i = 0; i < sbi->s_groups_count; i++) {
2364 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2366 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2367 last_block = ext4_blocks_count(sbi->s_es) - 1;
2369 last_block = first_block +
2370 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2372 if ((grp == sbi->s_groups_count) &&
2373 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2376 block_bitmap = ext4_block_bitmap(sb, gdp);
2377 if (block_bitmap == sb_block) {
2378 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2379 "Block bitmap for group %u overlaps "
2384 if (block_bitmap >= sb_block + 1 &&
2385 block_bitmap <= last_bg_block) {
2386 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2387 "Block bitmap for group %u overlaps "
2388 "block group descriptors", i);
2392 if (block_bitmap < first_block || block_bitmap > last_block) {
2393 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2394 "Block bitmap for group %u not in group "
2395 "(block %llu)!", i, block_bitmap);
2398 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2399 if (inode_bitmap == sb_block) {
2400 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2401 "Inode bitmap for group %u overlaps "
2406 if (inode_bitmap >= sb_block + 1 &&
2407 inode_bitmap <= last_bg_block) {
2408 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2409 "Inode bitmap for group %u overlaps "
2410 "block group descriptors", i);
2414 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2415 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2416 "Inode bitmap for group %u not in group "
2417 "(block %llu)!", i, inode_bitmap);
2420 inode_table = ext4_inode_table(sb, gdp);
2421 if (inode_table == sb_block) {
2422 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2423 "Inode table for group %u overlaps "
2428 if (inode_table >= sb_block + 1 &&
2429 inode_table <= last_bg_block) {
2430 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2431 "Inode table for group %u overlaps "
2432 "block group descriptors", i);
2436 if (inode_table < first_block ||
2437 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2438 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2439 "Inode table for group %u not in group "
2440 "(block %llu)!", i, inode_table);
2443 ext4_lock_group(sb, i);
2444 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2445 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2446 "Checksum for group %u failed (%u!=%u)",
2447 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
2448 gdp)), le16_to_cpu(gdp->bg_checksum));
2449 if (!sb_rdonly(sb)) {
2450 ext4_unlock_group(sb, i);
2454 ext4_unlock_group(sb, i);
2456 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2458 if (NULL != first_not_zeroed)
2459 *first_not_zeroed = grp;
2463 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2464 * the superblock) which were deleted from all directories, but held open by
2465 * a process at the time of a crash. We walk the list and try to delete these
2466 * inodes at recovery time (only with a read-write filesystem).
2468 * In order to keep the orphan inode chain consistent during traversal (in
2469 * case of crash during recovery), we link each inode into the superblock
2470 * orphan list_head and handle it the same way as an inode deletion during
2471 * normal operation (which journals the operations for us).
2473 * We only do an iget() and an iput() on each inode, which is very safe if we
2474 * accidentally point at an in-use or already deleted inode. The worst that
2475 * can happen in this case is that we get a "bit already cleared" message from
2476 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2477 * e2fsck was run on this filesystem, and it must have already done the orphan
2478 * inode cleanup for us, so we can safely abort without any further action.
2480 static void ext4_orphan_cleanup(struct super_block *sb,
2481 struct ext4_super_block *es)
2483 unsigned int s_flags = sb->s_flags;
2484 int ret, nr_orphans = 0, nr_truncates = 0;
2486 int quota_update = 0;
2489 if (!es->s_last_orphan) {
2490 jbd_debug(4, "no orphan inodes to clean up\n");
2494 if (bdev_read_only(sb->s_bdev)) {
2495 ext4_msg(sb, KERN_ERR, "write access "
2496 "unavailable, skipping orphan cleanup");
2500 /* Check if feature set would not allow a r/w mount */
2501 if (!ext4_feature_set_ok(sb, 0)) {
2502 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2503 "unknown ROCOMPAT features");
2507 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2508 /* don't clear list on RO mount w/ errors */
2509 if (es->s_last_orphan && !(s_flags & SB_RDONLY)) {
2510 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
2511 "clearing orphan list.\n");
2512 es->s_last_orphan = 0;
2514 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2518 if (s_flags & SB_RDONLY) {
2519 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2520 sb->s_flags &= ~SB_RDONLY;
2523 /* Needed for iput() to work correctly and not trash data */
2524 sb->s_flags |= SB_ACTIVE;
2527 * Turn on quotas which were not enabled for read-only mounts if
2528 * filesystem has quota feature, so that they are updated correctly.
2530 if (ext4_has_feature_quota(sb) && (s_flags & SB_RDONLY)) {
2531 int ret = ext4_enable_quotas(sb);
2536 ext4_msg(sb, KERN_ERR,
2537 "Cannot turn on quotas: error %d", ret);
2540 /* Turn on journaled quotas used for old sytle */
2541 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2542 if (EXT4_SB(sb)->s_qf_names[i]) {
2543 int ret = ext4_quota_on_mount(sb, i);
2548 ext4_msg(sb, KERN_ERR,
2549 "Cannot turn on journaled "
2550 "quota: type %d: error %d", i, ret);
2555 while (es->s_last_orphan) {
2556 struct inode *inode;
2559 * We may have encountered an error during cleanup; if
2560 * so, skip the rest.
2562 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2563 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2564 es->s_last_orphan = 0;
2568 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2569 if (IS_ERR(inode)) {
2570 es->s_last_orphan = 0;
2574 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2575 dquot_initialize(inode);
2576 if (inode->i_nlink) {
2577 if (test_opt(sb, DEBUG))
2578 ext4_msg(sb, KERN_DEBUG,
2579 "%s: truncating inode %lu to %lld bytes",
2580 __func__, inode->i_ino, inode->i_size);
2581 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2582 inode->i_ino, inode->i_size);
2584 truncate_inode_pages(inode->i_mapping, inode->i_size);
2585 ret = ext4_truncate(inode);
2587 ext4_std_error(inode->i_sb, ret);
2588 inode_unlock(inode);
2591 if (test_opt(sb, DEBUG))
2592 ext4_msg(sb, KERN_DEBUG,
2593 "%s: deleting unreferenced inode %lu",
2594 __func__, inode->i_ino);
2595 jbd_debug(2, "deleting unreferenced inode %lu\n",
2599 iput(inode); /* The delete magic happens here! */
2602 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2605 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2606 PLURAL(nr_orphans));
2608 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2609 PLURAL(nr_truncates));
2611 /* Turn off quotas if they were enabled for orphan cleanup */
2613 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2614 if (sb_dqopt(sb)->files[i])
2615 dquot_quota_off(sb, i);
2619 sb->s_flags = s_flags; /* Restore SB_RDONLY status */
2623 * Maximal extent format file size.
2624 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2625 * extent format containers, within a sector_t, and within i_blocks
2626 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2627 * so that won't be a limiting factor.
2629 * However there is other limiting factor. We do store extents in the form
2630 * of starting block and length, hence the resulting length of the extent
2631 * covering maximum file size must fit into on-disk format containers as
2632 * well. Given that length is always by 1 unit bigger than max unit (because
2633 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2635 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2637 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2640 loff_t upper_limit = MAX_LFS_FILESIZE;
2642 /* small i_blocks in vfs inode? */
2643 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2645 * CONFIG_LBDAF is not enabled implies the inode
2646 * i_block represent total blocks in 512 bytes
2647 * 32 == size of vfs inode i_blocks * 8
2649 upper_limit = (1LL << 32) - 1;
2651 /* total blocks in file system block size */
2652 upper_limit >>= (blkbits - 9);
2653 upper_limit <<= blkbits;
2657 * 32-bit extent-start container, ee_block. We lower the maxbytes
2658 * by one fs block, so ee_len can cover the extent of maximum file
2661 res = (1LL << 32) - 1;
2664 /* Sanity check against vm- & vfs- imposed limits */
2665 if (res > upper_limit)
2672 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2673 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2674 * We need to be 1 filesystem block less than the 2^48 sector limit.
2676 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2678 loff_t res = EXT4_NDIR_BLOCKS;
2681 /* This is calculated to be the largest file size for a dense, block
2682 * mapped file such that the file's total number of 512-byte sectors,
2683 * including data and all indirect blocks, does not exceed (2^48 - 1).
2685 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2686 * number of 512-byte sectors of the file.
2689 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2691 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2692 * the inode i_block field represents total file blocks in
2693 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2695 upper_limit = (1LL << 32) - 1;
2697 /* total blocks in file system block size */
2698 upper_limit >>= (bits - 9);
2702 * We use 48 bit ext4_inode i_blocks
2703 * With EXT4_HUGE_FILE_FL set the i_blocks
2704 * represent total number of blocks in
2705 * file system block size
2707 upper_limit = (1LL << 48) - 1;
2711 /* indirect blocks */
2713 /* double indirect blocks */
2714 meta_blocks += 1 + (1LL << (bits-2));
2715 /* tripple indirect blocks */
2716 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2718 upper_limit -= meta_blocks;
2719 upper_limit <<= bits;
2721 res += 1LL << (bits-2);
2722 res += 1LL << (2*(bits-2));
2723 res += 1LL << (3*(bits-2));
2725 if (res > upper_limit)
2728 if (res > MAX_LFS_FILESIZE)
2729 res = MAX_LFS_FILESIZE;
2734 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2735 ext4_fsblk_t logical_sb_block, int nr)
2737 struct ext4_sb_info *sbi = EXT4_SB(sb);
2738 ext4_group_t bg, first_meta_bg;
2741 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2743 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
2744 return logical_sb_block + nr + 1;
2745 bg = sbi->s_desc_per_block * nr;
2746 if (ext4_bg_has_super(sb, bg))
2750 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2751 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2752 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2755 if (sb->s_blocksize == 1024 && nr == 0 &&
2756 le32_to_cpu(sbi->s_es->s_first_data_block) == 0)
2759 return (has_super + ext4_group_first_block_no(sb, bg));
2763 * ext4_get_stripe_size: Get the stripe size.
2764 * @sbi: In memory super block info
2766 * If we have specified it via mount option, then
2767 * use the mount option value. If the value specified at mount time is
2768 * greater than the blocks per group use the super block value.
2769 * If the super block value is greater than blocks per group return 0.
2770 * Allocator needs it be less than blocks per group.
2773 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2775 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2776 unsigned long stripe_width =
2777 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2780 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2781 ret = sbi->s_stripe;
2782 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
2784 else if (stride && stride <= sbi->s_blocks_per_group)
2790 * If the stripe width is 1, this makes no sense and
2791 * we set it to 0 to turn off stripe handling code.
2800 * Check whether this filesystem can be mounted based on
2801 * the features present and the RDONLY/RDWR mount requested.
2802 * Returns 1 if this filesystem can be mounted as requested,
2803 * 0 if it cannot be.
2805 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2807 if (ext4_has_unknown_ext4_incompat_features(sb)) {
2808 ext4_msg(sb, KERN_ERR,
2809 "Couldn't mount because of "
2810 "unsupported optional features (%x)",
2811 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2812 ~EXT4_FEATURE_INCOMPAT_SUPP));
2819 if (ext4_has_feature_readonly(sb)) {
2820 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2821 sb->s_flags |= SB_RDONLY;
2825 /* Check that feature set is OK for a read-write mount */
2826 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
2827 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2828 "unsupported optional features (%x)",
2829 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2830 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2834 * Large file size enabled file system can only be mounted
2835 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2837 if (ext4_has_feature_huge_file(sb)) {
2838 if (sizeof(blkcnt_t) < sizeof(u64)) {
2839 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2840 "cannot be mounted RDWR without "
2845 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
2846 ext4_msg(sb, KERN_ERR,
2847 "Can't support bigalloc feature without "
2848 "extents feature\n");
2852 #ifndef CONFIG_QUOTA
2853 if (ext4_has_feature_quota(sb) && !readonly) {
2854 ext4_msg(sb, KERN_ERR,
2855 "Filesystem with quota feature cannot be mounted RDWR "
2856 "without CONFIG_QUOTA");
2859 if (ext4_has_feature_project(sb) && !readonly) {
2860 ext4_msg(sb, KERN_ERR,
2861 "Filesystem with project quota feature cannot be mounted RDWR "
2862 "without CONFIG_QUOTA");
2865 #endif /* CONFIG_QUOTA */
2870 * This function is called once a day if we have errors logged
2871 * on the file system
2873 static void print_daily_error_info(struct timer_list *t)
2875 struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report);
2876 struct super_block *sb = sbi->s_sb;
2877 struct ext4_super_block *es = sbi->s_es;
2879 if (es->s_error_count)
2880 /* fsck newer than v1.41.13 is needed to clean this condition. */
2881 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
2882 le32_to_cpu(es->s_error_count));
2883 if (es->s_first_error_time) {
2884 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
2885 sb->s_id, le32_to_cpu(es->s_first_error_time),
2886 (int) sizeof(es->s_first_error_func),
2887 es->s_first_error_func,
2888 le32_to_cpu(es->s_first_error_line));
2889 if (es->s_first_error_ino)
2890 printk(KERN_CONT ": inode %u",
2891 le32_to_cpu(es->s_first_error_ino));
2892 if (es->s_first_error_block)
2893 printk(KERN_CONT ": block %llu", (unsigned long long)
2894 le64_to_cpu(es->s_first_error_block));
2895 printk(KERN_CONT "\n");
2897 if (es->s_last_error_time) {
2898 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
2899 sb->s_id, le32_to_cpu(es->s_last_error_time),
2900 (int) sizeof(es->s_last_error_func),
2901 es->s_last_error_func,
2902 le32_to_cpu(es->s_last_error_line));
2903 if (es->s_last_error_ino)
2904 printk(KERN_CONT ": inode %u",
2905 le32_to_cpu(es->s_last_error_ino));
2906 if (es->s_last_error_block)
2907 printk(KERN_CONT ": block %llu", (unsigned long long)
2908 le64_to_cpu(es->s_last_error_block));
2909 printk(KERN_CONT "\n");
2911 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2914 /* Find next suitable group and run ext4_init_inode_table */
2915 static int ext4_run_li_request(struct ext4_li_request *elr)
2917 struct ext4_group_desc *gdp = NULL;
2918 ext4_group_t group, ngroups;
2919 struct super_block *sb;
2920 unsigned long timeout = 0;
2924 ngroups = EXT4_SB(sb)->s_groups_count;
2926 for (group = elr->lr_next_group; group < ngroups; group++) {
2927 gdp = ext4_get_group_desc(sb, group, NULL);
2933 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2937 if (group >= ngroups)
2942 ret = ext4_init_inode_table(sb, group,
2943 elr->lr_timeout ? 0 : 1);
2944 if (elr->lr_timeout == 0) {
2945 timeout = (jiffies - timeout) *
2946 elr->lr_sbi->s_li_wait_mult;
2947 elr->lr_timeout = timeout;
2949 elr->lr_next_sched = jiffies + elr->lr_timeout;
2950 elr->lr_next_group = group + 1;
2956 * Remove lr_request from the list_request and free the
2957 * request structure. Should be called with li_list_mtx held
2959 static void ext4_remove_li_request(struct ext4_li_request *elr)
2961 struct ext4_sb_info *sbi;
2968 list_del(&elr->lr_request);
2969 sbi->s_li_request = NULL;
2973 static void ext4_unregister_li_request(struct super_block *sb)
2975 mutex_lock(&ext4_li_mtx);
2976 if (!ext4_li_info) {
2977 mutex_unlock(&ext4_li_mtx);
2981 mutex_lock(&ext4_li_info->li_list_mtx);
2982 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2983 mutex_unlock(&ext4_li_info->li_list_mtx);
2984 mutex_unlock(&ext4_li_mtx);
2987 static struct task_struct *ext4_lazyinit_task;
2990 * This is the function where ext4lazyinit thread lives. It walks
2991 * through the request list searching for next scheduled filesystem.
2992 * When such a fs is found, run the lazy initialization request
2993 * (ext4_rn_li_request) and keep track of the time spend in this
2994 * function. Based on that time we compute next schedule time of
2995 * the request. When walking through the list is complete, compute
2996 * next waking time and put itself into sleep.
2998 static int ext4_lazyinit_thread(void *arg)
3000 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
3001 struct list_head *pos, *n;
3002 struct ext4_li_request *elr;
3003 unsigned long next_wakeup, cur;
3005 BUG_ON(NULL == eli);
3009 next_wakeup = MAX_JIFFY_OFFSET;
3011 mutex_lock(&eli->li_list_mtx);
3012 if (list_empty(&eli->li_request_list)) {
3013 mutex_unlock(&eli->li_list_mtx);
3016 list_for_each_safe(pos, n, &eli->li_request_list) {
3019 elr = list_entry(pos, struct ext4_li_request,
3022 if (time_before(jiffies, elr->lr_next_sched)) {
3023 if (time_before(elr->lr_next_sched, next_wakeup))
3024 next_wakeup = elr->lr_next_sched;
3027 if (down_read_trylock(&elr->lr_super->s_umount)) {
3028 if (sb_start_write_trylock(elr->lr_super)) {
3031 * We hold sb->s_umount, sb can not
3032 * be removed from the list, it is
3033 * now safe to drop li_list_mtx
3035 mutex_unlock(&eli->li_list_mtx);
3036 err = ext4_run_li_request(elr);
3037 sb_end_write(elr->lr_super);
3038 mutex_lock(&eli->li_list_mtx);
3041 up_read((&elr->lr_super->s_umount));
3043 /* error, remove the lazy_init job */
3045 ext4_remove_li_request(elr);
3049 elr->lr_next_sched = jiffies +
3051 % (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3053 if (time_before(elr->lr_next_sched, next_wakeup))
3054 next_wakeup = elr->lr_next_sched;
3056 mutex_unlock(&eli->li_list_mtx);
3061 if ((time_after_eq(cur, next_wakeup)) ||
3062 (MAX_JIFFY_OFFSET == next_wakeup)) {
3067 schedule_timeout_interruptible(next_wakeup - cur);
3069 if (kthread_should_stop()) {
3070 ext4_clear_request_list();
3077 * It looks like the request list is empty, but we need
3078 * to check it under the li_list_mtx lock, to prevent any
3079 * additions into it, and of course we should lock ext4_li_mtx
3080 * to atomically free the list and ext4_li_info, because at
3081 * this point another ext4 filesystem could be registering
3084 mutex_lock(&ext4_li_mtx);
3085 mutex_lock(&eli->li_list_mtx);
3086 if (!list_empty(&eli->li_request_list)) {
3087 mutex_unlock(&eli->li_list_mtx);
3088 mutex_unlock(&ext4_li_mtx);
3091 mutex_unlock(&eli->li_list_mtx);
3092 kfree(ext4_li_info);
3093 ext4_li_info = NULL;
3094 mutex_unlock(&ext4_li_mtx);
3099 static void ext4_clear_request_list(void)
3101 struct list_head *pos, *n;
3102 struct ext4_li_request *elr;
3104 mutex_lock(&ext4_li_info->li_list_mtx);
3105 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3106 elr = list_entry(pos, struct ext4_li_request,
3108 ext4_remove_li_request(elr);
3110 mutex_unlock(&ext4_li_info->li_list_mtx);
3113 static int ext4_run_lazyinit_thread(void)
3115 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3116 ext4_li_info, "ext4lazyinit");
3117 if (IS_ERR(ext4_lazyinit_task)) {
3118 int err = PTR_ERR(ext4_lazyinit_task);
3119 ext4_clear_request_list();
3120 kfree(ext4_li_info);
3121 ext4_li_info = NULL;
3122 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3123 "initialization thread\n",
3127 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3132 * Check whether it make sense to run itable init. thread or not.
3133 * If there is at least one uninitialized inode table, return
3134 * corresponding group number, else the loop goes through all
3135 * groups and return total number of groups.
3137 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3139 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3140 struct ext4_group_desc *gdp = NULL;
3142 for (group = 0; group < ngroups; group++) {
3143 gdp = ext4_get_group_desc(sb, group, NULL);
3147 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3154 static int ext4_li_info_new(void)
3156 struct ext4_lazy_init *eli = NULL;
3158 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3162 INIT_LIST_HEAD(&eli->li_request_list);
3163 mutex_init(&eli->li_list_mtx);
3165 eli->li_state |= EXT4_LAZYINIT_QUIT;
3172 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3175 struct ext4_sb_info *sbi = EXT4_SB(sb);
3176 struct ext4_li_request *elr;
3178 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3184 elr->lr_next_group = start;
3187 * Randomize first schedule time of the request to
3188 * spread the inode table initialization requests
3191 elr->lr_next_sched = jiffies + (prandom_u32() %
3192 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3196 int ext4_register_li_request(struct super_block *sb,
3197 ext4_group_t first_not_zeroed)
3199 struct ext4_sb_info *sbi = EXT4_SB(sb);
3200 struct ext4_li_request *elr = NULL;
3201 ext4_group_t ngroups = sbi->s_groups_count;
3204 mutex_lock(&ext4_li_mtx);
3205 if (sbi->s_li_request != NULL) {
3207 * Reset timeout so it can be computed again, because
3208 * s_li_wait_mult might have changed.
3210 sbi->s_li_request->lr_timeout = 0;
3214 if (first_not_zeroed == ngroups || sb_rdonly(sb) ||
3215 !test_opt(sb, INIT_INODE_TABLE))
3218 elr = ext4_li_request_new(sb, first_not_zeroed);
3224 if (NULL == ext4_li_info) {
3225 ret = ext4_li_info_new();
3230 mutex_lock(&ext4_li_info->li_list_mtx);
3231 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3232 mutex_unlock(&ext4_li_info->li_list_mtx);
3234 sbi->s_li_request = elr;
3236 * set elr to NULL here since it has been inserted to
3237 * the request_list and the removal and free of it is
3238 * handled by ext4_clear_request_list from now on.
3242 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3243 ret = ext4_run_lazyinit_thread();
3248 mutex_unlock(&ext4_li_mtx);
3255 * We do not need to lock anything since this is called on
3258 static void ext4_destroy_lazyinit_thread(void)
3261 * If thread exited earlier
3262 * there's nothing to be done.
3264 if (!ext4_li_info || !ext4_lazyinit_task)
3267 kthread_stop(ext4_lazyinit_task);
3270 static int set_journal_csum_feature_set(struct super_block *sb)
3273 int compat, incompat;
3274 struct ext4_sb_info *sbi = EXT4_SB(sb);
3276 if (ext4_has_metadata_csum(sb)) {
3277 /* journal checksum v3 */
3279 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
3281 /* journal checksum v1 */
3282 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3286 jbd2_journal_clear_features(sbi->s_journal,
3287 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3288 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3289 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3290 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3291 ret = jbd2_journal_set_features(sbi->s_journal,
3293 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3295 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3296 ret = jbd2_journal_set_features(sbi->s_journal,
3299 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3300 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3302 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3303 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3310 * Note: calculating the overhead so we can be compatible with
3311 * historical BSD practice is quite difficult in the face of
3312 * clusters/bigalloc. This is because multiple metadata blocks from
3313 * different block group can end up in the same allocation cluster.
3314 * Calculating the exact overhead in the face of clustered allocation
3315 * requires either O(all block bitmaps) in memory or O(number of block
3316 * groups**2) in time. We will still calculate the superblock for
3317 * older file systems --- and if we come across with a bigalloc file
3318 * system with zero in s_overhead_clusters the estimate will be close to
3319 * correct especially for very large cluster sizes --- but for newer
3320 * file systems, it's better to calculate this figure once at mkfs
3321 * time, and store it in the superblock. If the superblock value is
3322 * present (even for non-bigalloc file systems), we will use it.
3324 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3327 struct ext4_sb_info *sbi = EXT4_SB(sb);
3328 struct ext4_group_desc *gdp;
3329 ext4_fsblk_t first_block, last_block, b;
3330 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3331 int s, j, count = 0;
3333 if (!ext4_has_feature_bigalloc(sb))
3334 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3335 sbi->s_itb_per_group + 2);
3337 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3338 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3339 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3340 for (i = 0; i < ngroups; i++) {
3341 gdp = ext4_get_group_desc(sb, i, NULL);
3342 b = ext4_block_bitmap(sb, gdp);
3343 if (b >= first_block && b <= last_block) {
3344 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3347 b = ext4_inode_bitmap(sb, gdp);
3348 if (b >= first_block && b <= last_block) {
3349 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3352 b = ext4_inode_table(sb, gdp);
3353 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3354 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3355 int c = EXT4_B2C(sbi, b - first_block);
3356 ext4_set_bit(c, buf);
3362 if (ext4_bg_has_super(sb, grp)) {
3363 ext4_set_bit(s++, buf);
3366 j = ext4_bg_num_gdb(sb, grp);
3367 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
3368 ext4_error(sb, "Invalid number of block group "
3369 "descriptor blocks: %d", j);
3370 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
3374 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3378 return EXT4_CLUSTERS_PER_GROUP(sb) -
3379 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3383 * Compute the overhead and stash it in sbi->s_overhead
3385 int ext4_calculate_overhead(struct super_block *sb)
3387 struct ext4_sb_info *sbi = EXT4_SB(sb);
3388 struct ext4_super_block *es = sbi->s_es;
3389 struct inode *j_inode;
3390 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
3391 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3392 ext4_fsblk_t overhead = 0;
3393 char *buf = (char *) get_zeroed_page(GFP_NOFS);
3399 * Compute the overhead (FS structures). This is constant
3400 * for a given filesystem unless the number of block groups
3401 * changes so we cache the previous value until it does.
3405 * All of the blocks before first_data_block are overhead
3407 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3410 * Add the overhead found in each block group
3412 for (i = 0; i < ngroups; i++) {
3415 blks = count_overhead(sb, i, buf);
3418 memset(buf, 0, PAGE_SIZE);
3423 * Add the internal journal blocks whether the journal has been
3426 if (sbi->s_journal && !sbi->journal_bdev)
3427 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3428 else if (ext4_has_feature_journal(sb) && !sbi->s_journal) {
3429 j_inode = ext4_get_journal_inode(sb, j_inum);
3431 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
3432 overhead += EXT4_NUM_B2C(sbi, j_blocks);
3435 ext4_msg(sb, KERN_ERR, "can't get journal size");
3438 sbi->s_overhead = overhead;
3440 free_page((unsigned long) buf);
3444 static void ext4_set_resv_clusters(struct super_block *sb)
3446 ext4_fsblk_t resv_clusters;
3447 struct ext4_sb_info *sbi = EXT4_SB(sb);
3450 * There's no need to reserve anything when we aren't using extents.
3451 * The space estimates are exact, there are no unwritten extents,
3452 * hole punching doesn't need new metadata... This is needed especially
3453 * to keep ext2/3 backward compatibility.
3455 if (!ext4_has_feature_extents(sb))
3458 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3459 * This should cover the situations where we can not afford to run
3460 * out of space like for example punch hole, or converting
3461 * unwritten extents in delalloc path. In most cases such
3462 * allocation would require 1, or 2 blocks, higher numbers are
3465 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3466 sbi->s_cluster_bits);
3468 do_div(resv_clusters, 50);
3469 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3471 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
3474 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3476 struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev);
3477 char *orig_data = kstrdup(data, GFP_KERNEL);
3478 struct buffer_head *bh;
3479 struct ext4_super_block *es = NULL;
3480 struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3482 ext4_fsblk_t sb_block = get_sb_block(&data);
3483 ext4_fsblk_t logical_sb_block;
3484 unsigned long offset = 0;
3485 unsigned long journal_devnum = 0;
3486 unsigned long def_mount_opts;
3490 int blocksize, clustersize;
3491 unsigned int db_count;
3493 int needs_recovery, has_huge_files, has_bigalloc;
3496 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3497 ext4_group_t first_not_zeroed;
3499 if ((data && !orig_data) || !sbi)
3502 sbi->s_daxdev = dax_dev;
3503 sbi->s_blockgroup_lock =
3504 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3505 if (!sbi->s_blockgroup_lock)
3508 sb->s_fs_info = sbi;
3510 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3511 sbi->s_sb_block = sb_block;
3512 if (sb->s_bdev->bd_part)
3513 sbi->s_sectors_written_start =
3514 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3516 /* Cleanup superblock name */
3517 strreplace(sb->s_id, '/', '!');
3519 /* -EINVAL is default */
3521 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3523 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3528 * The ext4 superblock will not be buffer aligned for other than 1kB
3529 * block sizes. We need to calculate the offset from buffer start.
3531 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3532 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3533 offset = do_div(logical_sb_block, blocksize);
3535 logical_sb_block = sb_block;
3538 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
3539 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3543 * Note: s_es must be initialized as soon as possible because
3544 * some ext4 macro-instructions depend on its value
3546 es = (struct ext4_super_block *) (bh->b_data + offset);
3548 sb->s_magic = le16_to_cpu(es->s_magic);
3549 if (sb->s_magic != EXT4_SUPER_MAGIC)
3551 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3553 /* Warn if metadata_csum and gdt_csum are both set. */
3554 if (ext4_has_feature_metadata_csum(sb) &&
3555 ext4_has_feature_gdt_csum(sb))
3556 ext4_warning(sb, "metadata_csum and uninit_bg are "
3557 "redundant flags; please run fsck.");
3559 /* Check for a known checksum algorithm */
3560 if (!ext4_verify_csum_type(sb, es)) {
3561 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3562 "unknown checksum algorithm.");
3567 /* Load the checksum driver */
3568 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3569 if (IS_ERR(sbi->s_chksum_driver)) {
3570 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3571 ret = PTR_ERR(sbi->s_chksum_driver);
3572 sbi->s_chksum_driver = NULL;
3576 /* Check superblock checksum */
3577 if (!ext4_superblock_csum_verify(sb, es)) {
3578 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3579 "invalid superblock checksum. Run e2fsck?");
3585 /* Precompute checksum seed for all metadata */
3586 if (ext4_has_feature_csum_seed(sb))
3587 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
3588 else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb))
3589 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3590 sizeof(es->s_uuid));
3592 /* Set defaults before we parse the mount options */
3593 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3594 set_opt(sb, INIT_INODE_TABLE);
3595 if (def_mount_opts & EXT4_DEFM_DEBUG)
3597 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3599 if (def_mount_opts & EXT4_DEFM_UID16)
3600 set_opt(sb, NO_UID32);
3601 /* xattr user namespace & acls are now defaulted on */
3602 set_opt(sb, XATTR_USER);
3603 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3604 set_opt(sb, POSIX_ACL);
3606 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3607 if (ext4_has_metadata_csum(sb))
3608 set_opt(sb, JOURNAL_CHECKSUM);
3610 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3611 set_opt(sb, JOURNAL_DATA);
3612 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3613 set_opt(sb, ORDERED_DATA);
3614 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3615 set_opt(sb, WRITEBACK_DATA);
3617 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3618 set_opt(sb, ERRORS_PANIC);
3619 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3620 set_opt(sb, ERRORS_CONT);
3622 set_opt(sb, ERRORS_RO);
3623 /* block_validity enabled by default; disable with noblock_validity */
3624 set_opt(sb, BLOCK_VALIDITY);
3625 if (def_mount_opts & EXT4_DEFM_DISCARD)
3626 set_opt(sb, DISCARD);
3628 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3629 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3630 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3631 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3632 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3634 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3635 set_opt(sb, BARRIER);
3638 * enable delayed allocation by default
3639 * Use -o nodelalloc to turn it off
3641 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3642 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3643 set_opt(sb, DELALLOC);
3646 * set default s_li_wait_mult for lazyinit, for the case there is
3647 * no mount option specified.
3649 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3651 if (sbi->s_es->s_mount_opts[0]) {
3652 char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
3653 sizeof(sbi->s_es->s_mount_opts),
3657 if (!parse_options(s_mount_opts, sb, &journal_devnum,
3658 &journal_ioprio, 0)) {
3659 ext4_msg(sb, KERN_WARNING,
3660 "failed to parse options in superblock: %s",
3663 kfree(s_mount_opts);
3665 sbi->s_def_mount_opt = sbi->s_mount_opt;
3666 if (!parse_options((char *) data, sb, &journal_devnum,
3667 &journal_ioprio, 0))
3670 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3671 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3672 "with data=journal disables delayed "
3673 "allocation and O_DIRECT support!\n");
3674 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3675 ext4_msg(sb, KERN_ERR, "can't mount with "
3676 "both data=journal and delalloc");
3679 if (test_opt(sb, DIOREAD_NOLOCK)) {
3680 ext4_msg(sb, KERN_ERR, "can't mount with "
3681 "both data=journal and dioread_nolock");
3684 if (test_opt(sb, DAX)) {
3685 ext4_msg(sb, KERN_ERR, "can't mount with "
3686 "both data=journal and dax");
3689 if (ext4_has_feature_encrypt(sb)) {
3690 ext4_msg(sb, KERN_WARNING,
3691 "encrypted files will use data=ordered "
3692 "instead of data journaling mode");
3694 if (test_opt(sb, DELALLOC))
3695 clear_opt(sb, DELALLOC);
3697 sb->s_iflags |= SB_I_CGROUPWB;
3700 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
3701 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
3703 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3704 (ext4_has_compat_features(sb) ||
3705 ext4_has_ro_compat_features(sb) ||
3706 ext4_has_incompat_features(sb)))
3707 ext4_msg(sb, KERN_WARNING,
3708 "feature flags set on rev 0 fs, "
3709 "running e2fsck is recommended");
3711 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3712 set_opt2(sb, HURD_COMPAT);
3713 if (ext4_has_feature_64bit(sb)) {
3714 ext4_msg(sb, KERN_ERR,
3715 "The Hurd can't support 64-bit file systems");
3720 * ea_inode feature uses l_i_version field which is not
3721 * available in HURD_COMPAT mode.
3723 if (ext4_has_feature_ea_inode(sb)) {
3724 ext4_msg(sb, KERN_ERR,
3725 "ea_inode feature is not supported for Hurd");
3730 if (IS_EXT2_SB(sb)) {
3731 if (ext2_feature_set_ok(sb))
3732 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3733 "using the ext4 subsystem");
3736 * If we're probing be silent, if this looks like
3737 * it's actually an ext[34] filesystem.
3739 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
3741 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3742 "to feature incompatibilities");
3747 if (IS_EXT3_SB(sb)) {
3748 if (ext3_feature_set_ok(sb))
3749 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3750 "using the ext4 subsystem");
3753 * If we're probing be silent, if this looks like
3754 * it's actually an ext4 filesystem.
3756 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
3758 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3759 "to feature incompatibilities");
3765 * Check feature flags regardless of the revision level, since we
3766 * previously didn't change the revision level when setting the flags,
3767 * so there is a chance incompat flags are set on a rev 0 filesystem.
3769 if (!ext4_feature_set_ok(sb, (sb_rdonly(sb))))
3772 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3773 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3774 blocksize > EXT4_MAX_BLOCK_SIZE) {
3775 ext4_msg(sb, KERN_ERR,
3776 "Unsupported filesystem blocksize %d (%d log_block_size)",
3777 blocksize, le32_to_cpu(es->s_log_block_size));
3780 if (le32_to_cpu(es->s_log_block_size) >
3781 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3782 ext4_msg(sb, KERN_ERR,
3783 "Invalid log block size: %u",
3784 le32_to_cpu(es->s_log_block_size));
3788 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
3789 ext4_msg(sb, KERN_ERR,
3790 "Number of reserved GDT blocks insanely large: %d",
3791 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
3795 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
3796 if (ext4_has_feature_inline_data(sb)) {
3797 ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem"
3798 " that may contain inline data");
3799 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX;
3801 err = bdev_dax_supported(sb, blocksize);
3803 ext4_msg(sb, KERN_ERR,
3804 "DAX unsupported by block device. Turning off DAX.");
3805 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX;
3809 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
3810 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
3811 es->s_encryption_level);
3815 if (sb->s_blocksize != blocksize) {
3816 /* Validate the filesystem blocksize */
3817 if (!sb_set_blocksize(sb, blocksize)) {
3818 ext4_msg(sb, KERN_ERR, "bad block size %d",
3824 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3825 offset = do_div(logical_sb_block, blocksize);
3826 bh = sb_bread_unmovable(sb, logical_sb_block);
3828 ext4_msg(sb, KERN_ERR,
3829 "Can't read superblock on 2nd try");
3832 es = (struct ext4_super_block *)(bh->b_data + offset);
3834 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3835 ext4_msg(sb, KERN_ERR,
3836 "Magic mismatch, very weird!");
3841 has_huge_files = ext4_has_feature_huge_file(sb);
3842 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3844 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3846 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3847 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3848 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3850 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3851 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3852 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3853 (!is_power_of_2(sbi->s_inode_size)) ||
3854 (sbi->s_inode_size > blocksize)) {
3855 ext4_msg(sb, KERN_ERR,
3856 "unsupported inode size: %d",
3860 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3861 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3864 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3865 if (ext4_has_feature_64bit(sb)) {
3866 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3867 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3868 !is_power_of_2(sbi->s_desc_size)) {
3869 ext4_msg(sb, KERN_ERR,
3870 "unsupported descriptor size %lu",
3875 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3877 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3878 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3880 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3881 if (sbi->s_inodes_per_block == 0)
3883 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
3884 sbi->s_inodes_per_group > blocksize * 8) {
3885 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
3886 sbi->s_blocks_per_group);
3889 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3890 sbi->s_inodes_per_block;
3891 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3893 sbi->s_mount_state = le16_to_cpu(es->s_state);
3894 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3895 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3897 for (i = 0; i < 4; i++)
3898 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3899 sbi->s_def_hash_version = es->s_def_hash_version;
3900 if (ext4_has_feature_dir_index(sb)) {
3901 i = le32_to_cpu(es->s_flags);
3902 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3903 sbi->s_hash_unsigned = 3;
3904 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3905 #ifdef __CHAR_UNSIGNED__
3908 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3909 sbi->s_hash_unsigned = 3;
3913 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3918 /* Handle clustersize */
3919 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3920 has_bigalloc = ext4_has_feature_bigalloc(sb);
3922 if (clustersize < blocksize) {
3923 ext4_msg(sb, KERN_ERR,
3924 "cluster size (%d) smaller than "
3925 "block size (%d)", clustersize, blocksize);
3928 if (le32_to_cpu(es->s_log_cluster_size) >
3929 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3930 ext4_msg(sb, KERN_ERR,
3931 "Invalid log cluster size: %u",
3932 le32_to_cpu(es->s_log_cluster_size));
3935 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3936 le32_to_cpu(es->s_log_block_size);
3937 sbi->s_clusters_per_group =
3938 le32_to_cpu(es->s_clusters_per_group);
3939 if (sbi->s_clusters_per_group > blocksize * 8) {
3940 ext4_msg(sb, KERN_ERR,
3941 "#clusters per group too big: %lu",
3942 sbi->s_clusters_per_group);
3945 if (sbi->s_blocks_per_group !=
3946 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3947 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3948 "clusters per group (%lu) inconsistent",
3949 sbi->s_blocks_per_group,
3950 sbi->s_clusters_per_group);
3954 if (clustersize != blocksize) {
3955 ext4_warning(sb, "fragment/cluster size (%d) != "
3956 "block size (%d)", clustersize,
3958 clustersize = blocksize;
3960 if (sbi->s_blocks_per_group > blocksize * 8) {
3961 ext4_msg(sb, KERN_ERR,
3962 "#blocks per group too big: %lu",
3963 sbi->s_blocks_per_group);
3966 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3967 sbi->s_cluster_bits = 0;
3969 sbi->s_cluster_ratio = clustersize / blocksize;
3971 /* Do we have standard group size of clustersize * 8 blocks ? */
3972 if (sbi->s_blocks_per_group == clustersize << 3)
3973 set_opt2(sb, STD_GROUP_SIZE);
3976 * Test whether we have more sectors than will fit in sector_t,
3977 * and whether the max offset is addressable by the page cache.
3979 err = generic_check_addressable(sb->s_blocksize_bits,
3980 ext4_blocks_count(es));
3982 ext4_msg(sb, KERN_ERR, "filesystem"
3983 " too large to mount safely on this system");
3984 if (sizeof(sector_t) < 8)
3985 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3989 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3992 /* check blocks count against device size */
3993 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3994 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3995 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3996 "exceeds size of device (%llu blocks)",
3997 ext4_blocks_count(es), blocks_count);
4002 * It makes no sense for the first data block to be beyond the end
4003 * of the filesystem.
4005 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
4006 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4007 "block %u is beyond end of filesystem (%llu)",
4008 le32_to_cpu(es->s_first_data_block),
4009 ext4_blocks_count(es));
4012 blocks_count = (ext4_blocks_count(es) -
4013 le32_to_cpu(es->s_first_data_block) +
4014 EXT4_BLOCKS_PER_GROUP(sb) - 1);
4015 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4016 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
4017 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4018 "(block count %llu, first data block %u, "
4019 "blocks per group %lu)", sbi->s_groups_count,
4020 ext4_blocks_count(es),
4021 le32_to_cpu(es->s_first_data_block),
4022 EXT4_BLOCKS_PER_GROUP(sb));
4025 sbi->s_groups_count = blocks_count;
4026 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
4027 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
4028 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
4029 EXT4_DESC_PER_BLOCK(sb);
4030 if (ext4_has_feature_meta_bg(sb)) {
4031 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
4032 ext4_msg(sb, KERN_WARNING,
4033 "first meta block group too large: %u "
4034 "(group descriptor block count %u)",
4035 le32_to_cpu(es->s_first_meta_bg), db_count);
4039 sbi->s_group_desc = kvmalloc(db_count *
4040 sizeof(struct buffer_head *),
4042 if (sbi->s_group_desc == NULL) {
4043 ext4_msg(sb, KERN_ERR, "not enough memory");
4048 bgl_lock_init(sbi->s_blockgroup_lock);
4050 /* Pre-read the descriptors into the buffer cache */
4051 for (i = 0; i < db_count; i++) {
4052 block = descriptor_loc(sb, logical_sb_block, i);
4053 sb_breadahead(sb, block);
4056 for (i = 0; i < db_count; i++) {
4057 block = descriptor_loc(sb, logical_sb_block, i);
4058 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
4059 if (!sbi->s_group_desc[i]) {
4060 ext4_msg(sb, KERN_ERR,
4061 "can't read group descriptor %d", i);
4066 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
4067 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
4068 ret = -EFSCORRUPTED;
4072 sbi->s_gdb_count = db_count;
4074 timer_setup(&sbi->s_err_report, print_daily_error_info, 0);
4076 /* Register extent status tree shrinker */
4077 if (ext4_es_register_shrinker(sbi))
4080 sbi->s_stripe = ext4_get_stripe_size(sbi);
4081 sbi->s_extent_max_zeroout_kb = 32;
4084 * set up enough so that it can read an inode
4086 sb->s_op = &ext4_sops;
4087 sb->s_export_op = &ext4_export_ops;
4088 sb->s_xattr = ext4_xattr_handlers;
4089 #ifdef CONFIG_EXT4_FS_ENCRYPTION
4090 sb->s_cop = &ext4_cryptops;
4093 sb->dq_op = &ext4_quota_operations;
4094 if (ext4_has_feature_quota(sb))
4095 sb->s_qcop = &dquot_quotactl_sysfile_ops;
4097 sb->s_qcop = &ext4_qctl_operations;
4098 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
4100 memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
4102 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
4103 mutex_init(&sbi->s_orphan_lock);
4107 needs_recovery = (es->s_last_orphan != 0 ||
4108 ext4_has_feature_journal_needs_recovery(sb));
4110 if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb))
4111 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
4112 goto failed_mount3a;
4115 * The first inode we look at is the journal inode. Don't try
4116 * root first: it may be modified in the journal!
4118 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
4119 err = ext4_load_journal(sb, es, journal_devnum);
4121 goto failed_mount3a;
4122 } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) &&
4123 ext4_has_feature_journal_needs_recovery(sb)) {
4124 ext4_msg(sb, KERN_ERR, "required journal recovery "
4125 "suppressed and not mounted read-only");
4126 goto failed_mount_wq;
4128 /* Nojournal mode, all journal mount options are illegal */
4129 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
4130 ext4_msg(sb, KERN_ERR, "can't mount with "
4131 "journal_checksum, fs mounted w/o journal");
4132 goto failed_mount_wq;
4134 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4135 ext4_msg(sb, KERN_ERR, "can't mount with "
4136 "journal_async_commit, fs mounted w/o journal");
4137 goto failed_mount_wq;
4139 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
4140 ext4_msg(sb, KERN_ERR, "can't mount with "
4141 "commit=%lu, fs mounted w/o journal",
4142 sbi->s_commit_interval / HZ);
4143 goto failed_mount_wq;
4145 if (EXT4_MOUNT_DATA_FLAGS &
4146 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
4147 ext4_msg(sb, KERN_ERR, "can't mount with "
4148 "data=, fs mounted w/o journal");
4149 goto failed_mount_wq;
4151 sbi->s_def_mount_opt &= EXT4_MOUNT_JOURNAL_CHECKSUM;
4152 clear_opt(sb, JOURNAL_CHECKSUM);
4153 clear_opt(sb, DATA_FLAGS);
4154 sbi->s_journal = NULL;
4159 if (ext4_has_feature_64bit(sb) &&
4160 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4161 JBD2_FEATURE_INCOMPAT_64BIT)) {
4162 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
4163 goto failed_mount_wq;
4166 if (!set_journal_csum_feature_set(sb)) {
4167 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4169 goto failed_mount_wq;
4172 /* We have now updated the journal if required, so we can
4173 * validate the data journaling mode. */
4174 switch (test_opt(sb, DATA_FLAGS)) {
4176 /* No mode set, assume a default based on the journal
4177 * capabilities: ORDERED_DATA if the journal can
4178 * cope, else JOURNAL_DATA
4180 if (jbd2_journal_check_available_features
4181 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4182 set_opt(sb, ORDERED_DATA);
4183 sbi->s_def_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
4185 set_opt(sb, JOURNAL_DATA);
4186 sbi->s_def_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
4190 case EXT4_MOUNT_ORDERED_DATA:
4191 case EXT4_MOUNT_WRITEBACK_DATA:
4192 if (!jbd2_journal_check_available_features
4193 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4194 ext4_msg(sb, KERN_ERR, "Journal does not support "
4195 "requested data journaling mode");
4196 goto failed_mount_wq;
4202 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
4203 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4204 ext4_msg(sb, KERN_ERR, "can't mount with "
4205 "journal_async_commit in data=ordered mode");
4206 goto failed_mount_wq;
4209 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4211 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4214 if (!test_opt(sb, NO_MBCACHE)) {
4215 sbi->s_ea_block_cache = ext4_xattr_create_cache();
4216 if (!sbi->s_ea_block_cache) {
4217 ext4_msg(sb, KERN_ERR,
4218 "Failed to create ea_block_cache");
4219 goto failed_mount_wq;
4222 if (ext4_has_feature_ea_inode(sb)) {
4223 sbi->s_ea_inode_cache = ext4_xattr_create_cache();
4224 if (!sbi->s_ea_inode_cache) {
4225 ext4_msg(sb, KERN_ERR,
4226 "Failed to create ea_inode_cache");
4227 goto failed_mount_wq;
4232 if ((DUMMY_ENCRYPTION_ENABLED(sbi) || ext4_has_feature_encrypt(sb)) &&
4233 (blocksize != PAGE_SIZE)) {
4234 ext4_msg(sb, KERN_ERR,
4235 "Unsupported blocksize for fs encryption");
4236 goto failed_mount_wq;
4239 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !sb_rdonly(sb) &&
4240 !ext4_has_feature_encrypt(sb)) {
4241 ext4_set_feature_encrypt(sb);
4242 ext4_commit_super(sb, 1);
4246 * Get the # of file system overhead blocks from the
4247 * superblock if present.
4249 if (es->s_overhead_clusters)
4250 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4252 err = ext4_calculate_overhead(sb);
4254 goto failed_mount_wq;
4258 * The maximum number of concurrent works can be high and
4259 * concurrency isn't really necessary. Limit it to 1.
4261 EXT4_SB(sb)->rsv_conversion_wq =
4262 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4263 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4264 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
4270 * The jbd2_journal_load will have done any necessary log recovery,
4271 * so we can safely mount the rest of the filesystem now.
4274 root = ext4_iget(sb, EXT4_ROOT_INO);
4276 ext4_msg(sb, KERN_ERR, "get root inode failed");
4277 ret = PTR_ERR(root);
4281 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4282 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4286 sb->s_root = d_make_root(root);
4288 ext4_msg(sb, KERN_ERR, "get root dentry failed");
4293 ret = ext4_setup_super(sb, es, sb_rdonly(sb));
4294 if (ret == -EROFS) {
4295 sb->s_flags |= SB_RDONLY;
4298 goto failed_mount4a;
4300 /* determine the minimum size of new large inodes, if present */
4301 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE &&
4302 sbi->s_want_extra_isize == 0) {
4303 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4304 EXT4_GOOD_OLD_INODE_SIZE;
4305 if (ext4_has_feature_extra_isize(sb)) {
4306 if (sbi->s_want_extra_isize <
4307 le16_to_cpu(es->s_want_extra_isize))
4308 sbi->s_want_extra_isize =
4309 le16_to_cpu(es->s_want_extra_isize);
4310 if (sbi->s_want_extra_isize <
4311 le16_to_cpu(es->s_min_extra_isize))
4312 sbi->s_want_extra_isize =
4313 le16_to_cpu(es->s_min_extra_isize);
4316 /* Check if enough inode space is available */
4317 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4318 sbi->s_inode_size) {
4319 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4320 EXT4_GOOD_OLD_INODE_SIZE;
4321 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4325 ext4_set_resv_clusters(sb);
4327 err = ext4_setup_system_zone(sb);
4329 ext4_msg(sb, KERN_ERR, "failed to initialize system "
4331 goto failed_mount4a;
4335 err = ext4_mb_init(sb);
4337 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4342 block = ext4_count_free_clusters(sb);
4343 ext4_free_blocks_count_set(sbi->s_es,
4344 EXT4_C2B(sbi, block));
4345 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4348 unsigned long freei = ext4_count_free_inodes(sb);
4349 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
4350 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4354 err = percpu_counter_init(&sbi->s_dirs_counter,
4355 ext4_count_dirs(sb), GFP_KERNEL);
4357 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4360 err = percpu_init_rwsem(&sbi->s_journal_flag_rwsem);
4363 ext4_msg(sb, KERN_ERR, "insufficient memory");
4367 if (ext4_has_feature_flex_bg(sb))
4368 if (!ext4_fill_flex_info(sb)) {
4369 ext4_msg(sb, KERN_ERR,
4370 "unable to initialize "
4371 "flex_bg meta info!");
4375 err = ext4_register_li_request(sb, first_not_zeroed);
4379 err = ext4_register_sysfs(sb);
4384 /* Enable quota usage during mount. */
4385 if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) {
4386 err = ext4_enable_quotas(sb);
4390 #endif /* CONFIG_QUOTA */
4392 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4393 ext4_orphan_cleanup(sb, es);
4394 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4395 if (needs_recovery) {
4396 ext4_msg(sb, KERN_INFO, "recovery complete");
4397 ext4_mark_recovery_complete(sb, es);
4399 if (EXT4_SB(sb)->s_journal) {
4400 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4401 descr = " journalled data mode";
4402 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4403 descr = " ordered data mode";
4405 descr = " writeback data mode";
4407 descr = "out journal";
4409 if (test_opt(sb, DISCARD)) {
4410 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4411 if (!blk_queue_discard(q))
4412 ext4_msg(sb, KERN_WARNING,
4413 "mounting with \"discard\" option, but "
4414 "the device does not support discard");
4417 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
4418 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4419 "Opts: %.*s%s%s", descr,
4420 (int) sizeof(sbi->s_es->s_mount_opts),
4421 sbi->s_es->s_mount_opts,
4422 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4424 if (es->s_error_count)
4425 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4427 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4428 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4429 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4430 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4437 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4442 ext4_unregister_sysfs(sb);
4445 ext4_unregister_li_request(sb);
4447 ext4_mb_release(sb);
4448 if (sbi->s_flex_groups)
4449 kvfree(sbi->s_flex_groups);
4450 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4451 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4452 percpu_counter_destroy(&sbi->s_dirs_counter);
4453 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4455 ext4_ext_release(sb);
4456 ext4_release_system_zone(sb);
4461 ext4_msg(sb, KERN_ERR, "mount failed");
4462 if (EXT4_SB(sb)->rsv_conversion_wq)
4463 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4465 if (sbi->s_ea_inode_cache) {
4466 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
4467 sbi->s_ea_inode_cache = NULL;
4469 if (sbi->s_ea_block_cache) {
4470 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
4471 sbi->s_ea_block_cache = NULL;
4473 if (sbi->s_journal) {
4474 jbd2_journal_destroy(sbi->s_journal);
4475 sbi->s_journal = NULL;
4478 ext4_es_unregister_shrinker(sbi);
4480 del_timer_sync(&sbi->s_err_report);
4482 kthread_stop(sbi->s_mmp_tsk);
4484 for (i = 0; i < db_count; i++)
4485 brelse(sbi->s_group_desc[i]);
4486 kvfree(sbi->s_group_desc);
4488 if (sbi->s_chksum_driver)
4489 crypto_free_shash(sbi->s_chksum_driver);
4491 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4492 kfree(sbi->s_qf_names[i]);
4494 ext4_blkdev_remove(sbi);
4497 sb->s_fs_info = NULL;
4498 kfree(sbi->s_blockgroup_lock);
4502 fs_put_dax(dax_dev);
4503 return err ? err : ret;
4507 * Setup any per-fs journal parameters now. We'll do this both on
4508 * initial mount, once the journal has been initialised but before we've
4509 * done any recovery; and again on any subsequent remount.
4511 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4513 struct ext4_sb_info *sbi = EXT4_SB(sb);
4515 journal->j_commit_interval = sbi->s_commit_interval;
4516 journal->j_min_batch_time = sbi->s_min_batch_time;
4517 journal->j_max_batch_time = sbi->s_max_batch_time;
4519 write_lock(&journal->j_state_lock);
4520 if (test_opt(sb, BARRIER))
4521 journal->j_flags |= JBD2_BARRIER;
4523 journal->j_flags &= ~JBD2_BARRIER;
4524 if (test_opt(sb, DATA_ERR_ABORT))
4525 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4527 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4528 write_unlock(&journal->j_state_lock);
4531 static struct inode *ext4_get_journal_inode(struct super_block *sb,
4532 unsigned int journal_inum)
4534 struct inode *journal_inode;
4537 * Test for the existence of a valid inode on disk. Bad things
4538 * happen if we iget() an unused inode, as the subsequent iput()
4539 * will try to delete it.
4541 journal_inode = ext4_iget(sb, journal_inum);
4542 if (IS_ERR(journal_inode)) {
4543 ext4_msg(sb, KERN_ERR, "no journal found");
4546 if (!journal_inode->i_nlink) {
4547 make_bad_inode(journal_inode);
4548 iput(journal_inode);
4549 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4553 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4554 journal_inode, journal_inode->i_size);
4555 if (!S_ISREG(journal_inode->i_mode)) {
4556 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4557 iput(journal_inode);
4560 return journal_inode;
4563 static journal_t *ext4_get_journal(struct super_block *sb,
4564 unsigned int journal_inum)
4566 struct inode *journal_inode;
4569 BUG_ON(!ext4_has_feature_journal(sb));
4571 journal_inode = ext4_get_journal_inode(sb, journal_inum);
4575 journal = jbd2_journal_init_inode(journal_inode);
4577 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4578 iput(journal_inode);
4581 journal->j_private = sb;
4582 ext4_init_journal_params(sb, journal);
4586 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4589 struct buffer_head *bh;
4593 int hblock, blocksize;
4594 ext4_fsblk_t sb_block;
4595 unsigned long offset;
4596 struct ext4_super_block *es;
4597 struct block_device *bdev;
4599 BUG_ON(!ext4_has_feature_journal(sb));
4601 bdev = ext4_blkdev_get(j_dev, sb);
4605 blocksize = sb->s_blocksize;
4606 hblock = bdev_logical_block_size(bdev);
4607 if (blocksize < hblock) {
4608 ext4_msg(sb, KERN_ERR,
4609 "blocksize too small for journal device");
4613 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4614 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4615 set_blocksize(bdev, blocksize);
4616 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4617 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4618 "external journal");
4622 es = (struct ext4_super_block *) (bh->b_data + offset);
4623 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4624 !(le32_to_cpu(es->s_feature_incompat) &
4625 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4626 ext4_msg(sb, KERN_ERR, "external journal has "
4632 if ((le32_to_cpu(es->s_feature_ro_compat) &
4633 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4634 es->s_checksum != ext4_superblock_csum(sb, es)) {
4635 ext4_msg(sb, KERN_ERR, "external journal has "
4636 "corrupt superblock");
4641 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4642 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4647 len = ext4_blocks_count(es);
4648 start = sb_block + 1;
4649 brelse(bh); /* we're done with the superblock */
4651 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4652 start, len, blocksize);
4654 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4657 journal->j_private = sb;
4658 ll_rw_block(REQ_OP_READ, REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4659 wait_on_buffer(journal->j_sb_buffer);
4660 if (!buffer_uptodate(journal->j_sb_buffer)) {
4661 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4664 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4665 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4666 "user (unsupported) - %d",
4667 be32_to_cpu(journal->j_superblock->s_nr_users));
4670 EXT4_SB(sb)->journal_bdev = bdev;
4671 ext4_init_journal_params(sb, journal);
4675 jbd2_journal_destroy(journal);
4677 ext4_blkdev_put(bdev);
4681 static int ext4_load_journal(struct super_block *sb,
4682 struct ext4_super_block *es,
4683 unsigned long journal_devnum)
4686 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4689 int really_read_only;
4691 BUG_ON(!ext4_has_feature_journal(sb));
4693 if (journal_devnum &&
4694 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4695 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4696 "numbers have changed");
4697 journal_dev = new_decode_dev(journal_devnum);
4699 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4701 really_read_only = bdev_read_only(sb->s_bdev);
4704 * Are we loading a blank journal or performing recovery after a
4705 * crash? For recovery, we need to check in advance whether we
4706 * can get read-write access to the device.
4708 if (ext4_has_feature_journal_needs_recovery(sb)) {
4709 if (sb_rdonly(sb)) {
4710 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4711 "required on readonly filesystem");
4712 if (really_read_only) {
4713 ext4_msg(sb, KERN_ERR, "write access "
4714 "unavailable, cannot proceed "
4715 "(try mounting with noload)");
4718 ext4_msg(sb, KERN_INFO, "write access will "
4719 "be enabled during recovery");
4723 if (journal_inum && journal_dev) {
4724 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4725 "and inode journals!");
4730 if (!(journal = ext4_get_journal(sb, journal_inum)))
4733 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4737 if (!(journal->j_flags & JBD2_BARRIER))
4738 ext4_msg(sb, KERN_INFO, "barriers disabled");
4740 if (!ext4_has_feature_journal_needs_recovery(sb))
4741 err = jbd2_journal_wipe(journal, !really_read_only);
4743 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4745 memcpy(save, ((char *) es) +
4746 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4747 err = jbd2_journal_load(journal);
4749 memcpy(((char *) es) + EXT4_S_ERR_START,
4750 save, EXT4_S_ERR_LEN);
4755 ext4_msg(sb, KERN_ERR, "error loading journal");
4756 jbd2_journal_destroy(journal);
4760 EXT4_SB(sb)->s_journal = journal;
4761 ext4_clear_journal_err(sb, es);
4763 if (!really_read_only && journal_devnum &&
4764 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4765 es->s_journal_dev = cpu_to_le32(journal_devnum);
4767 /* Make sure we flush the recovery flag to disk. */
4768 ext4_commit_super(sb, 1);
4774 static int ext4_commit_super(struct super_block *sb, int sync)
4776 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4777 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4780 if (!sbh || block_device_ejected(sb))
4783 * If the file system is mounted read-only, don't update the
4784 * superblock write time. This avoids updating the superblock
4785 * write time when we are mounting the root file system
4786 * read/only but we need to replay the journal; at that point,
4787 * for people who are east of GMT and who make their clock
4788 * tick in localtime for Windows bug-for-bug compatibility,
4789 * the clock is set in the future, and this will cause e2fsck
4790 * to complain and force a full file system check.
4792 if (!(sb->s_flags & SB_RDONLY))
4793 es->s_wtime = cpu_to_le32(get_seconds());
4794 if (sb->s_bdev->bd_part)
4795 es->s_kbytes_written =
4796 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4797 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4798 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4800 es->s_kbytes_written =
4801 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4802 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4803 ext4_free_blocks_count_set(es,
4804 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4805 &EXT4_SB(sb)->s_freeclusters_counter)));
4806 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4807 es->s_free_inodes_count =
4808 cpu_to_le32(percpu_counter_sum_positive(
4809 &EXT4_SB(sb)->s_freeinodes_counter));
4810 BUFFER_TRACE(sbh, "marking dirty");
4811 ext4_superblock_csum_set(sb);
4814 if (buffer_write_io_error(sbh)) {
4816 * Oh, dear. A previous attempt to write the
4817 * superblock failed. This could happen because the
4818 * USB device was yanked out. Or it could happen to
4819 * be a transient write error and maybe the block will
4820 * be remapped. Nothing we can do but to retry the
4821 * write and hope for the best.
4823 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4824 "superblock detected");
4825 clear_buffer_write_io_error(sbh);
4826 set_buffer_uptodate(sbh);
4828 mark_buffer_dirty(sbh);
4831 error = __sync_dirty_buffer(sbh,
4832 REQ_SYNC | (test_opt(sb, BARRIER) ? REQ_FUA : 0));
4833 if (buffer_write_io_error(sbh)) {
4834 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4836 clear_buffer_write_io_error(sbh);
4837 set_buffer_uptodate(sbh);
4844 * Have we just finished recovery? If so, and if we are mounting (or
4845 * remounting) the filesystem readonly, then we will end up with a
4846 * consistent fs on disk. Record that fact.
4848 static void ext4_mark_recovery_complete(struct super_block *sb,
4849 struct ext4_super_block *es)
4851 journal_t *journal = EXT4_SB(sb)->s_journal;
4853 if (!ext4_has_feature_journal(sb)) {
4854 BUG_ON(journal != NULL);
4857 jbd2_journal_lock_updates(journal);
4858 if (jbd2_journal_flush(journal) < 0)
4861 if (ext4_has_feature_journal_needs_recovery(sb) && sb_rdonly(sb)) {
4862 ext4_clear_feature_journal_needs_recovery(sb);
4863 ext4_commit_super(sb, 1);
4867 jbd2_journal_unlock_updates(journal);
4871 * If we are mounting (or read-write remounting) a filesystem whose journal
4872 * has recorded an error from a previous lifetime, move that error to the
4873 * main filesystem now.
4875 static void ext4_clear_journal_err(struct super_block *sb,
4876 struct ext4_super_block *es)
4882 BUG_ON(!ext4_has_feature_journal(sb));
4884 journal = EXT4_SB(sb)->s_journal;
4887 * Now check for any error status which may have been recorded in the
4888 * journal by a prior ext4_error() or ext4_abort()
4891 j_errno = jbd2_journal_errno(journal);
4895 errstr = ext4_decode_error(sb, j_errno, nbuf);
4896 ext4_warning(sb, "Filesystem error recorded "
4897 "from previous mount: %s", errstr);
4898 ext4_warning(sb, "Marking fs in need of filesystem check.");
4900 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4901 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4902 ext4_commit_super(sb, 1);
4904 jbd2_journal_clear_err(journal);
4905 jbd2_journal_update_sb_errno(journal);
4910 * Force the running and committing transactions to commit,
4911 * and wait on the commit.
4913 int ext4_force_commit(struct super_block *sb)
4920 journal = EXT4_SB(sb)->s_journal;
4921 return ext4_journal_force_commit(journal);
4924 static int ext4_sync_fs(struct super_block *sb, int wait)
4928 bool needs_barrier = false;
4929 struct ext4_sb_info *sbi = EXT4_SB(sb);
4931 if (unlikely(ext4_forced_shutdown(sbi)))
4934 trace_ext4_sync_fs(sb, wait);
4935 flush_workqueue(sbi->rsv_conversion_wq);
4937 * Writeback quota in non-journalled quota case - journalled quota has
4940 dquot_writeback_dquots(sb, -1);
4942 * Data writeback is possible w/o journal transaction, so barrier must
4943 * being sent at the end of the function. But we can skip it if
4944 * transaction_commit will do it for us.
4946 if (sbi->s_journal) {
4947 target = jbd2_get_latest_transaction(sbi->s_journal);
4948 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4949 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4950 needs_barrier = true;
4952 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4954 ret = jbd2_log_wait_commit(sbi->s_journal,
4957 } else if (wait && test_opt(sb, BARRIER))
4958 needs_barrier = true;
4959 if (needs_barrier) {
4961 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4970 * LVM calls this function before a (read-only) snapshot is created. This
4971 * gives us a chance to flush the journal completely and mark the fs clean.
4973 * Note that only this function cannot bring a filesystem to be in a clean
4974 * state independently. It relies on upper layer to stop all data & metadata
4977 static int ext4_freeze(struct super_block *sb)
4985 journal = EXT4_SB(sb)->s_journal;
4988 /* Now we set up the journal barrier. */
4989 jbd2_journal_lock_updates(journal);
4992 * Don't clear the needs_recovery flag if we failed to
4993 * flush the journal.
4995 error = jbd2_journal_flush(journal);
4999 /* Journal blocked and flushed, clear needs_recovery flag. */
5000 ext4_clear_feature_journal_needs_recovery(sb);
5003 error = ext4_commit_super(sb, 1);
5006 /* we rely on upper layer to stop further updates */
5007 jbd2_journal_unlock_updates(journal);
5012 * Called by LVM after the snapshot is done. We need to reset the RECOVER
5013 * flag here, even though the filesystem is not technically dirty yet.
5015 static int ext4_unfreeze(struct super_block *sb)
5017 if (sb_rdonly(sb) || ext4_forced_shutdown(EXT4_SB(sb)))
5020 if (EXT4_SB(sb)->s_journal) {
5021 /* Reset the needs_recovery flag before the fs is unlocked. */
5022 ext4_set_feature_journal_needs_recovery(sb);
5025 ext4_commit_super(sb, 1);
5030 * Structure to save mount options for ext4_remount's benefit
5032 struct ext4_mount_options {
5033 unsigned long s_mount_opt;
5034 unsigned long s_mount_opt2;
5037 unsigned long s_commit_interval;
5038 u32 s_min_batch_time, s_max_batch_time;
5041 char *s_qf_names[EXT4_MAXQUOTAS];
5045 static int ext4_remount(struct super_block *sb, int *flags, char *data)
5047 struct ext4_super_block *es;
5048 struct ext4_sb_info *sbi = EXT4_SB(sb);
5049 unsigned long old_sb_flags;
5050 struct ext4_mount_options old_opts;
5051 int enable_quota = 0;
5053 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
5058 char *orig_data = kstrdup(data, GFP_KERNEL);
5060 /* Store the original options */
5061 old_sb_flags = sb->s_flags;
5062 old_opts.s_mount_opt = sbi->s_mount_opt;
5063 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
5064 old_opts.s_resuid = sbi->s_resuid;
5065 old_opts.s_resgid = sbi->s_resgid;
5066 old_opts.s_commit_interval = sbi->s_commit_interval;
5067 old_opts.s_min_batch_time = sbi->s_min_batch_time;
5068 old_opts.s_max_batch_time = sbi->s_max_batch_time;
5070 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
5071 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5072 if (sbi->s_qf_names[i]) {
5073 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
5075 if (!old_opts.s_qf_names[i]) {
5076 for (j = 0; j < i; j++)
5077 kfree(old_opts.s_qf_names[j]);
5082 old_opts.s_qf_names[i] = NULL;
5084 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
5085 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
5087 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
5092 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
5093 test_opt(sb, JOURNAL_CHECKSUM)) {
5094 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
5095 "during remount not supported; ignoring");
5096 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
5099 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
5100 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
5101 ext4_msg(sb, KERN_ERR, "can't mount with "
5102 "both data=journal and delalloc");
5106 if (test_opt(sb, DIOREAD_NOLOCK)) {
5107 ext4_msg(sb, KERN_ERR, "can't mount with "
5108 "both data=journal and dioread_nolock");
5112 if (test_opt(sb, DAX)) {
5113 ext4_msg(sb, KERN_ERR, "can't mount with "
5114 "both data=journal and dax");
5118 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
5119 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
5120 ext4_msg(sb, KERN_ERR, "can't mount with "
5121 "journal_async_commit in data=ordered mode");
5127 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) {
5128 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount");
5133 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
5134 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
5135 "dax flag with busy inodes while remounting");
5136 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
5139 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
5140 ext4_abort(sb, "Abort forced by user");
5142 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
5143 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
5147 if (sbi->s_journal) {
5148 ext4_init_journal_params(sb, sbi->s_journal);
5149 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
5152 if (*flags & SB_LAZYTIME)
5153 sb->s_flags |= SB_LAZYTIME;
5155 if ((bool)(*flags & SB_RDONLY) != sb_rdonly(sb)) {
5156 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
5161 if (*flags & SB_RDONLY) {
5162 err = sync_filesystem(sb);
5165 err = dquot_suspend(sb, -1);
5170 * First of all, the unconditional stuff we have to do
5171 * to disable replay of the journal when we next remount
5173 sb->s_flags |= SB_RDONLY;
5176 * OK, test if we are remounting a valid rw partition
5177 * readonly, and if so set the rdonly flag and then
5178 * mark the partition as valid again.
5180 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
5181 (sbi->s_mount_state & EXT4_VALID_FS))
5182 es->s_state = cpu_to_le16(sbi->s_mount_state);
5185 ext4_mark_recovery_complete(sb, es);
5187 /* Make sure we can mount this feature set readwrite */
5188 if (ext4_has_feature_readonly(sb) ||
5189 !ext4_feature_set_ok(sb, 0)) {
5194 * Make sure the group descriptor checksums
5195 * are sane. If they aren't, refuse to remount r/w.
5197 for (g = 0; g < sbi->s_groups_count; g++) {
5198 struct ext4_group_desc *gdp =
5199 ext4_get_group_desc(sb, g, NULL);
5201 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
5202 ext4_msg(sb, KERN_ERR,
5203 "ext4_remount: Checksum for group %u failed (%u!=%u)",
5204 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
5205 le16_to_cpu(gdp->bg_checksum));
5212 * If we have an unprocessed orphan list hanging
5213 * around from a previously readonly bdev mount,
5214 * require a full umount/remount for now.
5216 if (es->s_last_orphan) {
5217 ext4_msg(sb, KERN_WARNING, "Couldn't "
5218 "remount RDWR because of unprocessed "
5219 "orphan inode list. Please "
5220 "umount/remount instead");
5226 * Mounting a RDONLY partition read-write, so reread
5227 * and store the current valid flag. (It may have
5228 * been changed by e2fsck since we originally mounted
5232 ext4_clear_journal_err(sb, es);
5233 sbi->s_mount_state = le16_to_cpu(es->s_state);
5235 err = ext4_setup_super(sb, es, 0);
5239 sb->s_flags &= ~SB_RDONLY;
5240 if (ext4_has_feature_mmp(sb))
5241 if (ext4_multi_mount_protect(sb,
5242 le64_to_cpu(es->s_mmp_block))) {
5251 * Reinitialize lazy itable initialization thread based on
5254 if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE))
5255 ext4_unregister_li_request(sb);
5257 ext4_group_t first_not_zeroed;
5258 first_not_zeroed = ext4_has_uninit_itable(sb);
5259 ext4_register_li_request(sb, first_not_zeroed);
5262 ext4_setup_system_zone(sb);
5263 if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY)) {
5264 err = ext4_commit_super(sb, 1);
5270 /* Release old quota file names */
5271 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5272 kfree(old_opts.s_qf_names[i]);
5274 if (sb_any_quota_suspended(sb))
5275 dquot_resume(sb, -1);
5276 else if (ext4_has_feature_quota(sb)) {
5277 err = ext4_enable_quotas(sb);
5284 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
5285 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5290 sb->s_flags = old_sb_flags;
5291 sbi->s_mount_opt = old_opts.s_mount_opt;
5292 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
5293 sbi->s_resuid = old_opts.s_resuid;
5294 sbi->s_resgid = old_opts.s_resgid;
5295 sbi->s_commit_interval = old_opts.s_commit_interval;
5296 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5297 sbi->s_max_batch_time = old_opts.s_max_batch_time;
5299 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
5300 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
5301 kfree(sbi->s_qf_names[i]);
5302 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
5310 static int ext4_statfs_project(struct super_block *sb,
5311 kprojid_t projid, struct kstatfs *buf)
5314 struct dquot *dquot;
5318 qid = make_kqid_projid(projid);
5319 dquot = dqget(sb, qid);
5321 return PTR_ERR(dquot);
5322 spin_lock(&dquot->dq_dqb_lock);
5324 limit = (dquot->dq_dqb.dqb_bsoftlimit ?
5325 dquot->dq_dqb.dqb_bsoftlimit :
5326 dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits;
5327 if (limit && buf->f_blocks > limit) {
5328 curblock = (dquot->dq_dqb.dqb_curspace +
5329 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
5330 buf->f_blocks = limit;
5331 buf->f_bfree = buf->f_bavail =
5332 (buf->f_blocks > curblock) ?
5333 (buf->f_blocks - curblock) : 0;
5336 limit = dquot->dq_dqb.dqb_isoftlimit ?
5337 dquot->dq_dqb.dqb_isoftlimit :
5338 dquot->dq_dqb.dqb_ihardlimit;
5339 if (limit && buf->f_files > limit) {
5340 buf->f_files = limit;
5342 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
5343 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
5346 spin_unlock(&dquot->dq_dqb_lock);
5352 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
5354 struct super_block *sb = dentry->d_sb;
5355 struct ext4_sb_info *sbi = EXT4_SB(sb);
5356 struct ext4_super_block *es = sbi->s_es;
5357 ext4_fsblk_t overhead = 0, resv_blocks;
5360 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
5362 if (!test_opt(sb, MINIX_DF))
5363 overhead = sbi->s_overhead;
5365 buf->f_type = EXT4_SUPER_MAGIC;
5366 buf->f_bsize = sb->s_blocksize;
5367 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
5368 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5369 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
5370 /* prevent underflow in case that few free space is available */
5371 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
5372 buf->f_bavail = buf->f_bfree -
5373 (ext4_r_blocks_count(es) + resv_blocks);
5374 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
5376 buf->f_files = le32_to_cpu(es->s_inodes_count);
5377 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5378 buf->f_namelen = EXT4_NAME_LEN;
5379 fsid = le64_to_cpup((void *)es->s_uuid) ^
5380 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5381 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5382 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
5385 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
5386 sb_has_quota_limits_enabled(sb, PRJQUOTA))
5387 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
5396 * Helper functions so that transaction is started before we acquire dqio_sem
5397 * to keep correct lock ordering of transaction > dqio_sem
5399 static inline struct inode *dquot_to_inode(struct dquot *dquot)
5401 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
5404 static int ext4_write_dquot(struct dquot *dquot)
5408 struct inode *inode;
5410 inode = dquot_to_inode(dquot);
5411 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
5412 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
5414 return PTR_ERR(handle);
5415 ret = dquot_commit(dquot);
5416 err = ext4_journal_stop(handle);
5422 static int ext4_acquire_dquot(struct dquot *dquot)
5427 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5428 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
5430 return PTR_ERR(handle);
5431 ret = dquot_acquire(dquot);
5432 err = ext4_journal_stop(handle);
5438 static int ext4_release_dquot(struct dquot *dquot)
5443 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5444 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
5445 if (IS_ERR(handle)) {
5446 /* Release dquot anyway to avoid endless cycle in dqput() */
5447 dquot_release(dquot);
5448 return PTR_ERR(handle);
5450 ret = dquot_release(dquot);
5451 err = ext4_journal_stop(handle);
5457 static int ext4_mark_dquot_dirty(struct dquot *dquot)
5459 struct super_block *sb = dquot->dq_sb;
5460 struct ext4_sb_info *sbi = EXT4_SB(sb);
5462 /* Are we journaling quotas? */
5463 if (ext4_has_feature_quota(sb) ||
5464 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
5465 dquot_mark_dquot_dirty(dquot);
5466 return ext4_write_dquot(dquot);
5468 return dquot_mark_dquot_dirty(dquot);
5472 static int ext4_write_info(struct super_block *sb, int type)
5477 /* Data block + inode block */
5478 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
5480 return PTR_ERR(handle);
5481 ret = dquot_commit_info(sb, type);
5482 err = ext4_journal_stop(handle);
5489 * Turn on quotas during mount time - we need to find
5490 * the quota file and such...
5492 static int ext4_quota_on_mount(struct super_block *sb, int type)
5494 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5495 EXT4_SB(sb)->s_jquota_fmt, type);
5498 static void lockdep_set_quota_inode(struct inode *inode, int subclass)
5500 struct ext4_inode_info *ei = EXT4_I(inode);
5502 /* The first argument of lockdep_set_subclass has to be
5503 * *exactly* the same as the argument to init_rwsem() --- in
5504 * this case, in init_once() --- or lockdep gets unhappy
5505 * because the name of the lock is set using the
5506 * stringification of the argument to init_rwsem().
5508 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
5509 lockdep_set_subclass(&ei->i_data_sem, subclass);
5513 * Standard function to be called on quota_on
5515 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
5516 const struct path *path)
5520 if (!test_opt(sb, QUOTA))
5523 /* Quotafile not on the same filesystem? */
5524 if (path->dentry->d_sb != sb)
5526 /* Journaling quota? */
5527 if (EXT4_SB(sb)->s_qf_names[type]) {
5528 /* Quotafile not in fs root? */
5529 if (path->dentry->d_parent != sb->s_root)
5530 ext4_msg(sb, KERN_WARNING,
5531 "Quota file not on filesystem root. "
5532 "Journaled quota will not work");
5533 sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY;
5536 * Clear the flag just in case mount options changed since
5539 sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY;
5543 * When we journal data on quota file, we have to flush journal to see
5544 * all updates to the file when we bypass pagecache...
5546 if (EXT4_SB(sb)->s_journal &&
5547 ext4_should_journal_data(d_inode(path->dentry))) {
5549 * We don't need to lock updates but journal_flush() could
5550 * otherwise be livelocked...
5552 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
5553 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
5554 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
5559 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
5560 err = dquot_quota_on(sb, type, format_id, path);
5562 lockdep_set_quota_inode(path->dentry->d_inode,
5565 struct inode *inode = d_inode(path->dentry);
5569 * Set inode flags to prevent userspace from messing with quota
5570 * files. If this fails, we return success anyway since quotas
5571 * are already enabled and this is not a hard failure.
5574 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5577 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL;
5578 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
5579 S_NOATIME | S_IMMUTABLE);
5580 ext4_mark_inode_dirty(handle, inode);
5581 ext4_journal_stop(handle);
5583 inode_unlock(inode);
5588 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5592 struct inode *qf_inode;
5593 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5594 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5595 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5596 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
5599 BUG_ON(!ext4_has_feature_quota(sb));
5601 if (!qf_inums[type])
5604 qf_inode = ext4_iget(sb, qf_inums[type]);
5605 if (IS_ERR(qf_inode)) {
5606 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5607 return PTR_ERR(qf_inode);
5610 /* Don't account quota for quota files to avoid recursion */
5611 qf_inode->i_flags |= S_NOQUOTA;
5612 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
5613 err = dquot_enable(qf_inode, type, format_id, flags);
5616 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
5621 /* Enable usage tracking for all quota types. */
5622 static int ext4_enable_quotas(struct super_block *sb)
5625 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5626 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5627 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5628 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
5630 bool quota_mopt[EXT4_MAXQUOTAS] = {
5631 test_opt(sb, USRQUOTA),
5632 test_opt(sb, GRPQUOTA),
5633 test_opt(sb, PRJQUOTA),
5636 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY;
5637 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
5638 if (qf_inums[type]) {
5639 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5640 DQUOT_USAGE_ENABLED |
5641 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
5643 for (type--; type >= 0; type--)
5644 dquot_quota_off(sb, type);
5647 "Failed to enable quota tracking "
5648 "(type=%d, err=%d). Please run "
5649 "e2fsck to fix.", type, err);
5657 static int ext4_quota_off(struct super_block *sb, int type)
5659 struct inode *inode = sb_dqopt(sb)->files[type];
5663 /* Force all delayed allocation blocks to be allocated.
5664 * Caller already holds s_umount sem */
5665 if (test_opt(sb, DELALLOC))
5666 sync_filesystem(sb);
5668 if (!inode || !igrab(inode))
5671 err = dquot_quota_off(sb, type);
5672 if (err || ext4_has_feature_quota(sb))
5677 * Update modification times of quota files when userspace can
5678 * start looking at them. If we fail, we return success anyway since
5679 * this is not a hard failure and quotas are already disabled.
5681 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5684 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL);
5685 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
5686 inode->i_mtime = inode->i_ctime = current_time(inode);
5687 ext4_mark_inode_dirty(handle, inode);
5688 ext4_journal_stop(handle);
5690 inode_unlock(inode);
5692 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL);
5696 return dquot_quota_off(sb, type);
5699 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5700 * acquiring the locks... As quota files are never truncated and quota code
5701 * itself serializes the operations (and no one else should touch the files)
5702 * we don't have to be afraid of races */
5703 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5704 size_t len, loff_t off)
5706 struct inode *inode = sb_dqopt(sb)->files[type];
5707 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5708 int offset = off & (sb->s_blocksize - 1);
5711 struct buffer_head *bh;
5712 loff_t i_size = i_size_read(inode);
5716 if (off+len > i_size)
5719 while (toread > 0) {
5720 tocopy = sb->s_blocksize - offset < toread ?
5721 sb->s_blocksize - offset : toread;
5722 bh = ext4_bread(NULL, inode, blk, 0);
5725 if (!bh) /* A hole? */
5726 memset(data, 0, tocopy);
5728 memcpy(data, bh->b_data+offset, tocopy);
5738 /* Write to quotafile (we know the transaction is already started and has
5739 * enough credits) */
5740 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5741 const char *data, size_t len, loff_t off)
5743 struct inode *inode = sb_dqopt(sb)->files[type];
5744 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5745 int err, offset = off & (sb->s_blocksize - 1);
5747 struct buffer_head *bh;
5748 handle_t *handle = journal_current_handle();
5750 if (EXT4_SB(sb)->s_journal && !handle) {
5751 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5752 " cancelled because transaction is not started",
5753 (unsigned long long)off, (unsigned long long)len);
5757 * Since we account only one data block in transaction credits,
5758 * then it is impossible to cross a block boundary.
5760 if (sb->s_blocksize - offset < len) {
5761 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5762 " cancelled because not block aligned",
5763 (unsigned long long)off, (unsigned long long)len);
5768 bh = ext4_bread(handle, inode, blk,
5769 EXT4_GET_BLOCKS_CREATE |
5770 EXT4_GET_BLOCKS_METADATA_NOFAIL);
5771 } while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) &&
5772 ext4_should_retry_alloc(inode->i_sb, &retries));
5777 BUFFER_TRACE(bh, "get write access");
5778 err = ext4_journal_get_write_access(handle, bh);
5784 memcpy(bh->b_data+offset, data, len);
5785 flush_dcache_page(bh->b_page);
5787 err = ext4_handle_dirty_metadata(handle, NULL, bh);
5790 if (inode->i_size < off + len) {
5791 i_size_write(inode, off + len);
5792 EXT4_I(inode)->i_disksize = inode->i_size;
5793 ext4_mark_inode_dirty(handle, inode);
5798 static int ext4_get_next_id(struct super_block *sb, struct kqid *qid)
5800 const struct quota_format_ops *ops;
5802 if (!sb_has_quota_loaded(sb, qid->type))
5804 ops = sb_dqopt(sb)->ops[qid->type];
5805 if (!ops || !ops->get_next_id)
5807 return dquot_get_next_id(sb, qid);
5811 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5812 const char *dev_name, void *data)
5814 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5817 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
5818 static inline void register_as_ext2(void)
5820 int err = register_filesystem(&ext2_fs_type);
5823 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5826 static inline void unregister_as_ext2(void)
5828 unregister_filesystem(&ext2_fs_type);
5831 static inline int ext2_feature_set_ok(struct super_block *sb)
5833 if (ext4_has_unknown_ext2_incompat_features(sb))
5837 if (ext4_has_unknown_ext2_ro_compat_features(sb))
5842 static inline void register_as_ext2(void) { }
5843 static inline void unregister_as_ext2(void) { }
5844 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5847 static inline void register_as_ext3(void)
5849 int err = register_filesystem(&ext3_fs_type);
5852 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5855 static inline void unregister_as_ext3(void)
5857 unregister_filesystem(&ext3_fs_type);
5860 static inline int ext3_feature_set_ok(struct super_block *sb)
5862 if (ext4_has_unknown_ext3_incompat_features(sb))
5864 if (!ext4_has_feature_journal(sb))
5868 if (ext4_has_unknown_ext3_ro_compat_features(sb))
5873 static struct file_system_type ext4_fs_type = {
5874 .owner = THIS_MODULE,
5876 .mount = ext4_mount,
5877 .kill_sb = kill_block_super,
5878 .fs_flags = FS_REQUIRES_DEV,
5880 MODULE_ALIAS_FS("ext4");
5882 /* Shared across all ext4 file systems */
5883 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5885 static int __init ext4_init_fs(void)
5889 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
5890 ext4_li_info = NULL;
5891 mutex_init(&ext4_li_mtx);
5893 /* Build-time check for flags consistency */
5894 ext4_check_flag_values();
5896 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
5897 init_waitqueue_head(&ext4__ioend_wq[i]);
5899 err = ext4_init_es();
5903 err = ext4_init_pageio();
5907 err = ext4_init_system_zone();
5911 err = ext4_init_sysfs();
5915 err = ext4_init_mballoc();
5918 err = init_inodecache();
5923 err = register_filesystem(&ext4_fs_type);
5929 unregister_as_ext2();
5930 unregister_as_ext3();
5931 destroy_inodecache();
5933 ext4_exit_mballoc();
5937 ext4_exit_system_zone();
5946 static void __exit ext4_exit_fs(void)
5948 ext4_destroy_lazyinit_thread();
5949 unregister_as_ext2();
5950 unregister_as_ext3();
5951 unregister_filesystem(&ext4_fs_type);
5952 destroy_inodecache();
5953 ext4_exit_mballoc();
5955 ext4_exit_system_zone();
5960 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5961 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5962 MODULE_LICENSE("GPL");
5963 MODULE_SOFTDEP("pre: crc32c");
5964 module_init(ext4_init_fs)
5965 module_exit(ext4_exit_fs)