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>
45 #include <linux/unicode.h>
46 #include <linux/part_stat.h>
47 #include <linux/kthread.h>
48 #include <linux/freezer.h>
51 #include "ext4_extents.h" /* Needed for trace points definition */
52 #include "ext4_jbd2.h"
58 #define CREATE_TRACE_POINTS
59 #include <trace/events/ext4.h>
61 static struct ext4_lazy_init *ext4_li_info;
62 static struct mutex ext4_li_mtx;
63 static struct ratelimit_state ext4_mount_msg_ratelimit;
65 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
66 unsigned long journal_devnum);
67 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
68 static int ext4_commit_super(struct super_block *sb, int sync);
69 static int ext4_mark_recovery_complete(struct super_block *sb,
70 struct ext4_super_block *es);
71 static int ext4_clear_journal_err(struct super_block *sb,
72 struct ext4_super_block *es);
73 static int ext4_sync_fs(struct super_block *sb, int wait);
74 static int ext4_remount(struct super_block *sb, int *flags, char *data);
75 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
76 static int ext4_unfreeze(struct super_block *sb);
77 static int ext4_freeze(struct super_block *sb);
78 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
79 const char *dev_name, void *data);
80 static inline int ext2_feature_set_ok(struct super_block *sb);
81 static inline int ext3_feature_set_ok(struct super_block *sb);
82 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
83 static void ext4_destroy_lazyinit_thread(void);
84 static void ext4_unregister_li_request(struct super_block *sb);
85 static void ext4_clear_request_list(void);
86 static struct inode *ext4_get_journal_inode(struct super_block *sb,
87 unsigned int journal_inum);
92 * Note the difference between i_mmap_sem (EXT4_I(inode)->i_mmap_sem) and
93 * i_mmap_rwsem (inode->i_mmap_rwsem)!
96 * mmap_lock -> sb_start_pagefault -> i_mmap_sem (r) -> transaction start ->
97 * page lock -> i_data_sem (rw)
99 * buffered write path:
100 * sb_start_write -> i_mutex -> mmap_lock
101 * sb_start_write -> i_mutex -> transaction start -> page lock ->
105 * sb_start_write -> i_mutex -> i_mmap_sem (w) -> i_mmap_rwsem (w) -> page lock
106 * sb_start_write -> i_mutex -> i_mmap_sem (w) -> transaction start ->
110 * sb_start_write -> i_mutex -> mmap_lock
111 * sb_start_write -> i_mutex -> transaction start -> i_data_sem (rw)
114 * transaction start -> page lock(s) -> i_data_sem (rw)
117 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
118 static struct file_system_type ext2_fs_type = {
119 .owner = THIS_MODULE,
122 .kill_sb = kill_block_super,
123 .fs_flags = FS_REQUIRES_DEV,
125 MODULE_ALIAS_FS("ext2");
126 MODULE_ALIAS("ext2");
127 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
129 #define IS_EXT2_SB(sb) (0)
133 static struct file_system_type ext3_fs_type = {
134 .owner = THIS_MODULE,
137 .kill_sb = kill_block_super,
138 .fs_flags = FS_REQUIRES_DEV,
140 MODULE_ALIAS_FS("ext3");
141 MODULE_ALIAS("ext3");
142 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
145 static inline void __ext4_read_bh(struct buffer_head *bh, int op_flags,
149 * buffer's verified bit is no longer valid after reading from
150 * disk again due to write out error, clear it to make sure we
151 * recheck the buffer contents.
153 clear_buffer_verified(bh);
155 bh->b_end_io = end_io ? end_io : end_buffer_read_sync;
157 submit_bh(REQ_OP_READ, op_flags, bh);
160 void ext4_read_bh_nowait(struct buffer_head *bh, int op_flags,
163 BUG_ON(!buffer_locked(bh));
165 if (ext4_buffer_uptodate(bh)) {
169 __ext4_read_bh(bh, op_flags, end_io);
172 int ext4_read_bh(struct buffer_head *bh, int op_flags, bh_end_io_t *end_io)
174 BUG_ON(!buffer_locked(bh));
176 if (ext4_buffer_uptodate(bh)) {
181 __ext4_read_bh(bh, op_flags, end_io);
184 if (buffer_uptodate(bh))
189 int ext4_read_bh_lock(struct buffer_head *bh, int op_flags, bool wait)
191 if (trylock_buffer(bh)) {
193 return ext4_read_bh(bh, op_flags, NULL);
194 ext4_read_bh_nowait(bh, op_flags, NULL);
199 if (buffer_uptodate(bh))
207 * This works like __bread_gfp() except it uses ERR_PTR for error
208 * returns. Currently with sb_bread it's impossible to distinguish
209 * between ENOMEM and EIO situations (since both result in a NULL
212 static struct buffer_head *__ext4_sb_bread_gfp(struct super_block *sb,
213 sector_t block, int op_flags,
216 struct buffer_head *bh;
219 bh = sb_getblk_gfp(sb, block, gfp);
221 return ERR_PTR(-ENOMEM);
222 if (ext4_buffer_uptodate(bh))
225 ret = ext4_read_bh_lock(bh, REQ_META | op_flags, true);
233 struct buffer_head *ext4_sb_bread(struct super_block *sb, sector_t block,
236 return __ext4_sb_bread_gfp(sb, block, op_flags, __GFP_MOVABLE);
239 struct buffer_head *ext4_sb_bread_unmovable(struct super_block *sb,
242 return __ext4_sb_bread_gfp(sb, block, 0, 0);
245 void ext4_sb_breadahead_unmovable(struct super_block *sb, sector_t block)
247 struct buffer_head *bh = sb_getblk_gfp(sb, block, 0);
250 ext4_read_bh_lock(bh, REQ_RAHEAD, false);
255 static int ext4_verify_csum_type(struct super_block *sb,
256 struct ext4_super_block *es)
258 if (!ext4_has_feature_metadata_csum(sb))
261 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
264 static __le32 ext4_superblock_csum(struct super_block *sb,
265 struct ext4_super_block *es)
267 struct ext4_sb_info *sbi = EXT4_SB(sb);
268 int offset = offsetof(struct ext4_super_block, s_checksum);
271 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
273 return cpu_to_le32(csum);
276 static int ext4_superblock_csum_verify(struct super_block *sb,
277 struct ext4_super_block *es)
279 if (!ext4_has_metadata_csum(sb))
282 return es->s_checksum == ext4_superblock_csum(sb, es);
285 void ext4_superblock_csum_set(struct super_block *sb)
287 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
289 if (!ext4_has_metadata_csum(sb))
293 * Locking the superblock prevents the scenario
295 * 1) a first thread pauses during checksum calculation.
296 * 2) a second thread updates the superblock, recalculates
297 * the checksum, and updates s_checksum
298 * 3) the first thread resumes and finishes its checksum calculation
299 * and updates s_checksum with a potentially stale or torn value.
301 lock_buffer(EXT4_SB(sb)->s_sbh);
302 es->s_checksum = ext4_superblock_csum(sb, es);
303 unlock_buffer(EXT4_SB(sb)->s_sbh);
306 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
307 struct ext4_group_desc *bg)
309 return le32_to_cpu(bg->bg_block_bitmap_lo) |
310 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
311 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
314 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
315 struct ext4_group_desc *bg)
317 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
318 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
319 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
322 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
323 struct ext4_group_desc *bg)
325 return le32_to_cpu(bg->bg_inode_table_lo) |
326 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
327 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
330 __u32 ext4_free_group_clusters(struct super_block *sb,
331 struct ext4_group_desc *bg)
333 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
334 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
335 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
338 __u32 ext4_free_inodes_count(struct super_block *sb,
339 struct ext4_group_desc *bg)
341 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
342 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
343 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
346 __u32 ext4_used_dirs_count(struct super_block *sb,
347 struct ext4_group_desc *bg)
349 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
350 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
351 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
354 __u32 ext4_itable_unused_count(struct super_block *sb,
355 struct ext4_group_desc *bg)
357 return le16_to_cpu(bg->bg_itable_unused_lo) |
358 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
359 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
362 void ext4_block_bitmap_set(struct super_block *sb,
363 struct ext4_group_desc *bg, ext4_fsblk_t blk)
365 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
366 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
367 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
370 void ext4_inode_bitmap_set(struct super_block *sb,
371 struct ext4_group_desc *bg, ext4_fsblk_t blk)
373 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
374 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
375 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
378 void ext4_inode_table_set(struct super_block *sb,
379 struct ext4_group_desc *bg, ext4_fsblk_t blk)
381 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
382 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
383 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
386 void ext4_free_group_clusters_set(struct super_block *sb,
387 struct ext4_group_desc *bg, __u32 count)
389 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
390 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
391 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
394 void ext4_free_inodes_set(struct super_block *sb,
395 struct ext4_group_desc *bg, __u32 count)
397 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
398 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
399 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
402 void ext4_used_dirs_set(struct super_block *sb,
403 struct ext4_group_desc *bg, __u32 count)
405 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
406 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
407 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
410 void ext4_itable_unused_set(struct super_block *sb,
411 struct ext4_group_desc *bg, __u32 count)
413 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
414 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
415 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
418 static void __ext4_update_tstamp(__le32 *lo, __u8 *hi)
420 time64_t now = ktime_get_real_seconds();
422 now = clamp_val(now, 0, (1ull << 40) - 1);
424 *lo = cpu_to_le32(lower_32_bits(now));
425 *hi = upper_32_bits(now);
428 static time64_t __ext4_get_tstamp(__le32 *lo, __u8 *hi)
430 return ((time64_t)(*hi) << 32) + le32_to_cpu(*lo);
432 #define ext4_update_tstamp(es, tstamp) \
433 __ext4_update_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi)
434 #define ext4_get_tstamp(es, tstamp) \
435 __ext4_get_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi)
437 static void __save_error_info(struct super_block *sb, int error,
438 __u32 ino, __u64 block,
439 const char *func, unsigned int line)
441 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
444 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
445 if (bdev_read_only(sb->s_bdev))
447 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
448 ext4_update_tstamp(es, s_last_error_time);
449 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
450 es->s_last_error_line = cpu_to_le32(line);
451 es->s_last_error_ino = cpu_to_le32(ino);
452 es->s_last_error_block = cpu_to_le64(block);
458 err = EXT4_ERR_ENOMEM;
461 err = EXT4_ERR_EFSBADCRC;
465 err = EXT4_ERR_EFSCORRUPTED;
468 err = EXT4_ERR_ENOSPC;
471 err = EXT4_ERR_ENOKEY;
474 err = EXT4_ERR_EROFS;
477 err = EXT4_ERR_EFBIG;
480 err = EXT4_ERR_EEXIST;
483 err = EXT4_ERR_ERANGE;
486 err = EXT4_ERR_EOVERFLOW;
489 err = EXT4_ERR_EBUSY;
492 err = EXT4_ERR_ENOTDIR;
495 err = EXT4_ERR_ENOTEMPTY;
498 err = EXT4_ERR_ESHUTDOWN;
501 err = EXT4_ERR_EFAULT;
504 err = EXT4_ERR_UNKNOWN;
506 es->s_last_error_errcode = err;
507 if (!es->s_first_error_time) {
508 es->s_first_error_time = es->s_last_error_time;
509 es->s_first_error_time_hi = es->s_last_error_time_hi;
510 strncpy(es->s_first_error_func, func,
511 sizeof(es->s_first_error_func));
512 es->s_first_error_line = cpu_to_le32(line);
513 es->s_first_error_ino = es->s_last_error_ino;
514 es->s_first_error_block = es->s_last_error_block;
515 es->s_first_error_errcode = es->s_last_error_errcode;
518 * Start the daily error reporting function if it hasn't been
521 if (!es->s_error_count)
522 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
523 le32_add_cpu(&es->s_error_count, 1);
526 static void save_error_info(struct super_block *sb, int error,
527 __u32 ino, __u64 block,
528 const char *func, unsigned int line)
530 __save_error_info(sb, error, ino, block, func, line);
531 if (!bdev_read_only(sb->s_bdev))
532 ext4_commit_super(sb, 1);
536 * The del_gendisk() function uninitializes the disk-specific data
537 * structures, including the bdi structure, without telling anyone
538 * else. Once this happens, any attempt to call mark_buffer_dirty()
539 * (for example, by ext4_commit_super), will cause a kernel OOPS.
540 * This is a kludge to prevent these oops until we can put in a proper
541 * hook in del_gendisk() to inform the VFS and file system layers.
543 static int block_device_ejected(struct super_block *sb)
545 struct inode *bd_inode = sb->s_bdev->bd_inode;
546 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
548 return bdi->dev == NULL;
551 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
553 struct super_block *sb = journal->j_private;
554 struct ext4_sb_info *sbi = EXT4_SB(sb);
555 int error = is_journal_aborted(journal);
556 struct ext4_journal_cb_entry *jce;
558 BUG_ON(txn->t_state == T_FINISHED);
560 ext4_process_freed_data(sb, txn->t_tid);
562 spin_lock(&sbi->s_md_lock);
563 while (!list_empty(&txn->t_private_list)) {
564 jce = list_entry(txn->t_private_list.next,
565 struct ext4_journal_cb_entry, jce_list);
566 list_del_init(&jce->jce_list);
567 spin_unlock(&sbi->s_md_lock);
568 jce->jce_func(sb, jce, error);
569 spin_lock(&sbi->s_md_lock);
571 spin_unlock(&sbi->s_md_lock);
574 static bool system_going_down(void)
576 return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF
577 || system_state == SYSTEM_RESTART;
580 /* Deal with the reporting of failure conditions on a filesystem such as
581 * inconsistencies detected or read IO failures.
583 * On ext2, we can store the error state of the filesystem in the
584 * superblock. That is not possible on ext4, because we may have other
585 * write ordering constraints on the superblock which prevent us from
586 * writing it out straight away; and given that the journal is about to
587 * be aborted, we can't rely on the current, or future, transactions to
588 * write out the superblock safely.
590 * We'll just use the jbd2_journal_abort() error code to record an error in
591 * the journal instead. On recovery, the journal will complain about
592 * that error until we've noted it down and cleared it.
595 static void ext4_handle_error(struct super_block *sb)
597 if (test_opt(sb, WARN_ON_ERROR))
603 if (!test_opt(sb, ERRORS_CONT)) {
604 journal_t *journal = EXT4_SB(sb)->s_journal;
606 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
608 jbd2_journal_abort(journal, -EIO);
611 * We force ERRORS_RO behavior when system is rebooting. Otherwise we
612 * could panic during 'reboot -f' as the underlying device got already
615 if (test_opt(sb, ERRORS_RO) || system_going_down()) {
616 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
618 * Make sure updated value of ->s_mount_flags will be visible
619 * before ->s_flags update
622 sb->s_flags |= SB_RDONLY;
623 } else if (test_opt(sb, ERRORS_PANIC)) {
624 panic("EXT4-fs (device %s): panic forced after error\n",
629 #define ext4_error_ratelimit(sb) \
630 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
633 void __ext4_error(struct super_block *sb, const char *function,
634 unsigned int line, int error, __u64 block,
635 const char *fmt, ...)
637 struct va_format vaf;
640 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
643 trace_ext4_error(sb, function, line);
644 if (ext4_error_ratelimit(sb)) {
649 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
650 sb->s_id, function, line, current->comm, &vaf);
653 save_error_info(sb, error, 0, block, function, line);
654 ext4_handle_error(sb);
657 void __ext4_error_inode(struct inode *inode, const char *function,
658 unsigned int line, ext4_fsblk_t block, int error,
659 const char *fmt, ...)
662 struct va_format vaf;
664 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
667 trace_ext4_error(inode->i_sb, function, line);
668 if (ext4_error_ratelimit(inode->i_sb)) {
673 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
674 "inode #%lu: block %llu: comm %s: %pV\n",
675 inode->i_sb->s_id, function, line, inode->i_ino,
676 block, current->comm, &vaf);
678 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
679 "inode #%lu: comm %s: %pV\n",
680 inode->i_sb->s_id, function, line, inode->i_ino,
681 current->comm, &vaf);
684 save_error_info(inode->i_sb, error, inode->i_ino, block,
686 ext4_handle_error(inode->i_sb);
689 void __ext4_error_file(struct file *file, const char *function,
690 unsigned int line, ext4_fsblk_t block,
691 const char *fmt, ...)
694 struct va_format vaf;
695 struct inode *inode = file_inode(file);
696 char pathname[80], *path;
698 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
701 trace_ext4_error(inode->i_sb, function, line);
702 if (ext4_error_ratelimit(inode->i_sb)) {
703 path = file_path(file, pathname, sizeof(pathname));
711 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
712 "block %llu: comm %s: path %s: %pV\n",
713 inode->i_sb->s_id, function, line, inode->i_ino,
714 block, current->comm, path, &vaf);
717 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
718 "comm %s: path %s: %pV\n",
719 inode->i_sb->s_id, function, line, inode->i_ino,
720 current->comm, path, &vaf);
723 save_error_info(inode->i_sb, EFSCORRUPTED, inode->i_ino, block,
725 ext4_handle_error(inode->i_sb);
728 const char *ext4_decode_error(struct super_block *sb, int errno,
735 errstr = "Corrupt filesystem";
738 errstr = "Filesystem failed CRC";
741 errstr = "IO failure";
744 errstr = "Out of memory";
747 if (!sb || (EXT4_SB(sb)->s_journal &&
748 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
749 errstr = "Journal has aborted";
751 errstr = "Readonly filesystem";
754 /* If the caller passed in an extra buffer for unknown
755 * errors, textualise them now. Else we just return
758 /* Check for truncated error codes... */
759 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
768 /* __ext4_std_error decodes expected errors from journaling functions
769 * automatically and invokes the appropriate error response. */
771 void __ext4_std_error(struct super_block *sb, const char *function,
772 unsigned int line, int errno)
777 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
780 /* Special case: if the error is EROFS, and we're not already
781 * inside a transaction, then there's really no point in logging
783 if (errno == -EROFS && journal_current_handle() == NULL && sb_rdonly(sb))
786 if (ext4_error_ratelimit(sb)) {
787 errstr = ext4_decode_error(sb, errno, nbuf);
788 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
789 sb->s_id, function, line, errstr);
792 save_error_info(sb, -errno, 0, 0, function, line);
793 ext4_handle_error(sb);
797 * ext4_abort is a much stronger failure handler than ext4_error. The
798 * abort function may be used to deal with unrecoverable failures such
799 * as journal IO errors or ENOMEM at a critical moment in log management.
801 * We unconditionally force the filesystem into an ABORT|READONLY state,
802 * unless the error response on the fs has been set to panic in which
803 * case we take the easy way out and panic immediately.
806 void __ext4_abort(struct super_block *sb, const char *function,
807 unsigned int line, int error, const char *fmt, ...)
809 struct va_format vaf;
812 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
815 save_error_info(sb, error, 0, 0, function, line);
819 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: %pV\n",
820 sb->s_id, function, line, &vaf);
823 if (sb_rdonly(sb) == 0) {
824 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
825 if (EXT4_SB(sb)->s_journal)
826 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
828 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
830 * Make sure updated value of ->s_mount_flags will be visible
831 * before ->s_flags update
834 sb->s_flags |= SB_RDONLY;
836 if (test_opt(sb, ERRORS_PANIC) && !system_going_down())
837 panic("EXT4-fs panic from previous error\n");
840 void __ext4_msg(struct super_block *sb,
841 const char *prefix, const char *fmt, ...)
843 struct va_format vaf;
846 atomic_inc(&EXT4_SB(sb)->s_msg_count);
847 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
853 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
857 static int ext4_warning_ratelimit(struct super_block *sb)
859 atomic_inc(&EXT4_SB(sb)->s_warning_count);
860 return ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state),
864 void __ext4_warning(struct super_block *sb, const char *function,
865 unsigned int line, const char *fmt, ...)
867 struct va_format vaf;
870 if (!ext4_warning_ratelimit(sb))
876 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
877 sb->s_id, function, line, &vaf);
881 void __ext4_warning_inode(const struct inode *inode, const char *function,
882 unsigned int line, const char *fmt, ...)
884 struct va_format vaf;
887 if (!ext4_warning_ratelimit(inode->i_sb))
893 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
894 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
895 function, line, inode->i_ino, current->comm, &vaf);
899 void __ext4_grp_locked_error(const char *function, unsigned int line,
900 struct super_block *sb, ext4_group_t grp,
901 unsigned long ino, ext4_fsblk_t block,
902 const char *fmt, ...)
906 struct va_format vaf;
909 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
912 trace_ext4_error(sb, function, line);
913 __save_error_info(sb, EFSCORRUPTED, ino, block, function, line);
915 if (ext4_error_ratelimit(sb)) {
919 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
920 sb->s_id, function, line, grp);
922 printk(KERN_CONT "inode %lu: ", ino);
924 printk(KERN_CONT "block %llu:",
925 (unsigned long long) block);
926 printk(KERN_CONT "%pV\n", &vaf);
930 if (test_opt(sb, WARN_ON_ERROR))
933 if (test_opt(sb, ERRORS_CONT)) {
934 ext4_commit_super(sb, 0);
938 ext4_unlock_group(sb, grp);
939 ext4_commit_super(sb, 1);
940 ext4_handle_error(sb);
942 * We only get here in the ERRORS_RO case; relocking the group
943 * may be dangerous, but nothing bad will happen since the
944 * filesystem will have already been marked read/only and the
945 * journal has been aborted. We return 1 as a hint to callers
946 * who might what to use the return value from
947 * ext4_grp_locked_error() to distinguish between the
948 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
949 * aggressively from the ext4 function in question, with a
950 * more appropriate error code.
952 ext4_lock_group(sb, grp);
956 void ext4_mark_group_bitmap_corrupted(struct super_block *sb,
960 struct ext4_sb_info *sbi = EXT4_SB(sb);
961 struct ext4_group_info *grp = ext4_get_group_info(sb, group);
962 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, group, NULL);
965 if (flags & EXT4_GROUP_INFO_BBITMAP_CORRUPT) {
966 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT,
969 percpu_counter_sub(&sbi->s_freeclusters_counter,
973 if (flags & EXT4_GROUP_INFO_IBITMAP_CORRUPT) {
974 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT,
979 count = ext4_free_inodes_count(sb, gdp);
980 percpu_counter_sub(&sbi->s_freeinodes_counter,
986 void ext4_update_dynamic_rev(struct super_block *sb)
988 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
990 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
994 "updating to rev %d because of new feature flag, "
995 "running e2fsck is recommended",
998 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
999 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
1000 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
1001 /* leave es->s_feature_*compat flags alone */
1002 /* es->s_uuid will be set by e2fsck if empty */
1005 * The rest of the superblock fields should be zero, and if not it
1006 * means they are likely already in use, so leave them alone. We
1007 * can leave it up to e2fsck to clean up any inconsistencies there.
1012 * Open the external journal device
1014 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
1016 struct block_device *bdev;
1018 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
1024 ext4_msg(sb, KERN_ERR,
1025 "failed to open journal device unknown-block(%u,%u) %ld",
1026 MAJOR(dev), MINOR(dev), PTR_ERR(bdev));
1031 * Release the journal device
1033 static void ext4_blkdev_put(struct block_device *bdev)
1035 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
1038 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
1040 struct block_device *bdev;
1041 bdev = sbi->s_journal_bdev;
1043 ext4_blkdev_put(bdev);
1044 sbi->s_journal_bdev = NULL;
1048 static inline struct inode *orphan_list_entry(struct list_head *l)
1050 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
1053 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
1055 struct list_head *l;
1057 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
1058 le32_to_cpu(sbi->s_es->s_last_orphan));
1060 printk(KERN_ERR "sb_info orphan list:\n");
1061 list_for_each(l, &sbi->s_orphan) {
1062 struct inode *inode = orphan_list_entry(l);
1064 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
1065 inode->i_sb->s_id, inode->i_ino, inode,
1066 inode->i_mode, inode->i_nlink,
1067 NEXT_ORPHAN(inode));
1072 static int ext4_quota_off(struct super_block *sb, int type);
1074 static inline void ext4_quota_off_umount(struct super_block *sb)
1078 /* Use our quota_off function to clear inode flags etc. */
1079 for (type = 0; type < EXT4_MAXQUOTAS; type++)
1080 ext4_quota_off(sb, type);
1084 * This is a helper function which is used in the mount/remount
1085 * codepaths (which holds s_umount) to fetch the quota file name.
1087 static inline char *get_qf_name(struct super_block *sb,
1088 struct ext4_sb_info *sbi,
1091 return rcu_dereference_protected(sbi->s_qf_names[type],
1092 lockdep_is_held(&sb->s_umount));
1095 static inline void ext4_quota_off_umount(struct super_block *sb)
1100 static void ext4_put_super(struct super_block *sb)
1102 struct ext4_sb_info *sbi = EXT4_SB(sb);
1103 struct ext4_super_block *es = sbi->s_es;
1104 struct buffer_head **group_desc;
1105 struct flex_groups **flex_groups;
1109 ext4_unregister_li_request(sb);
1110 ext4_quota_off_umount(sb);
1112 destroy_workqueue(sbi->rsv_conversion_wq);
1115 * Unregister sysfs before destroying jbd2 journal.
1116 * Since we could still access attr_journal_task attribute via sysfs
1117 * path which could have sbi->s_journal->j_task as NULL
1119 ext4_unregister_sysfs(sb);
1121 if (sbi->s_journal) {
1122 aborted = is_journal_aborted(sbi->s_journal);
1123 err = jbd2_journal_destroy(sbi->s_journal);
1124 sbi->s_journal = NULL;
1125 if ((err < 0) && !aborted) {
1126 ext4_abort(sb, -err, "Couldn't clean up the journal");
1130 ext4_es_unregister_shrinker(sbi);
1131 del_timer_sync(&sbi->s_err_report);
1132 ext4_release_system_zone(sb);
1133 ext4_mb_release(sb);
1134 ext4_ext_release(sb);
1136 if (!sb_rdonly(sb) && !aborted) {
1137 ext4_clear_feature_journal_needs_recovery(sb);
1138 es->s_state = cpu_to_le16(sbi->s_mount_state);
1141 ext4_commit_super(sb, 1);
1144 group_desc = rcu_dereference(sbi->s_group_desc);
1145 for (i = 0; i < sbi->s_gdb_count; i++)
1146 brelse(group_desc[i]);
1148 flex_groups = rcu_dereference(sbi->s_flex_groups);
1150 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
1151 kvfree(flex_groups[i]);
1152 kvfree(flex_groups);
1155 percpu_counter_destroy(&sbi->s_freeclusters_counter);
1156 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1157 percpu_counter_destroy(&sbi->s_dirs_counter);
1158 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
1159 percpu_free_rwsem(&sbi->s_writepages_rwsem);
1161 for (i = 0; i < EXT4_MAXQUOTAS; i++)
1162 kfree(get_qf_name(sb, sbi, i));
1165 /* Debugging code just in case the in-memory inode orphan list
1166 * isn't empty. The on-disk one can be non-empty if we've
1167 * detected an error and taken the fs readonly, but the
1168 * in-memory list had better be clean by this point. */
1169 if (!list_empty(&sbi->s_orphan))
1170 dump_orphan_list(sb, sbi);
1171 J_ASSERT(list_empty(&sbi->s_orphan));
1173 sync_blockdev(sb->s_bdev);
1174 invalidate_bdev(sb->s_bdev);
1175 if (sbi->s_journal_bdev && sbi->s_journal_bdev != sb->s_bdev) {
1177 * Invalidate the journal device's buffers. We don't want them
1178 * floating about in memory - the physical journal device may
1179 * hotswapped, and it breaks the `ro-after' testing code.
1181 sync_blockdev(sbi->s_journal_bdev);
1182 invalidate_bdev(sbi->s_journal_bdev);
1183 ext4_blkdev_remove(sbi);
1186 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
1187 sbi->s_ea_inode_cache = NULL;
1189 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
1190 sbi->s_ea_block_cache = NULL;
1193 kthread_stop(sbi->s_mmp_tsk);
1195 sb->s_fs_info = NULL;
1197 * Now that we are completely done shutting down the
1198 * superblock, we need to actually destroy the kobject.
1200 kobject_put(&sbi->s_kobj);
1201 wait_for_completion(&sbi->s_kobj_unregister);
1202 if (sbi->s_chksum_driver)
1203 crypto_free_shash(sbi->s_chksum_driver);
1204 kfree(sbi->s_blockgroup_lock);
1205 fs_put_dax(sbi->s_daxdev);
1206 fscrypt_free_dummy_context(&sbi->s_dummy_enc_ctx);
1207 #ifdef CONFIG_UNICODE
1208 utf8_unload(sbi->s_encoding);
1213 static struct kmem_cache *ext4_inode_cachep;
1216 * Called inside transaction, so use GFP_NOFS
1218 static struct inode *ext4_alloc_inode(struct super_block *sb)
1220 struct ext4_inode_info *ei;
1222 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
1226 inode_set_iversion(&ei->vfs_inode, 1);
1227 spin_lock_init(&ei->i_raw_lock);
1228 INIT_LIST_HEAD(&ei->i_prealloc_list);
1229 atomic_set(&ei->i_prealloc_active, 0);
1230 spin_lock_init(&ei->i_prealloc_lock);
1231 ext4_es_init_tree(&ei->i_es_tree);
1232 rwlock_init(&ei->i_es_lock);
1233 INIT_LIST_HEAD(&ei->i_es_list);
1234 ei->i_es_all_nr = 0;
1235 ei->i_es_shk_nr = 0;
1236 ei->i_es_shrink_lblk = 0;
1237 ei->i_reserved_data_blocks = 0;
1238 spin_lock_init(&(ei->i_block_reservation_lock));
1239 ext4_init_pending_tree(&ei->i_pending_tree);
1241 ei->i_reserved_quota = 0;
1242 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
1245 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
1246 spin_lock_init(&ei->i_completed_io_lock);
1248 ei->i_datasync_tid = 0;
1249 atomic_set(&ei->i_unwritten, 0);
1250 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
1251 return &ei->vfs_inode;
1254 static int ext4_drop_inode(struct inode *inode)
1256 int drop = generic_drop_inode(inode);
1259 drop = fscrypt_drop_inode(inode);
1261 trace_ext4_drop_inode(inode, drop);
1265 static void ext4_free_in_core_inode(struct inode *inode)
1267 fscrypt_free_inode(inode);
1268 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
1271 static void ext4_destroy_inode(struct inode *inode)
1273 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
1274 ext4_msg(inode->i_sb, KERN_ERR,
1275 "Inode %lu (%p): orphan list check failed!",
1276 inode->i_ino, EXT4_I(inode));
1277 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1278 EXT4_I(inode), sizeof(struct ext4_inode_info),
1284 static void init_once(void *foo)
1286 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
1288 INIT_LIST_HEAD(&ei->i_orphan);
1289 init_rwsem(&ei->xattr_sem);
1290 init_rwsem(&ei->i_data_sem);
1291 init_rwsem(&ei->i_mmap_sem);
1292 inode_init_once(&ei->vfs_inode);
1295 static int __init init_inodecache(void)
1297 ext4_inode_cachep = kmem_cache_create_usercopy("ext4_inode_cache",
1298 sizeof(struct ext4_inode_info), 0,
1299 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD|
1301 offsetof(struct ext4_inode_info, i_data),
1302 sizeof_field(struct ext4_inode_info, i_data),
1304 if (ext4_inode_cachep == NULL)
1309 static void destroy_inodecache(void)
1312 * Make sure all delayed rcu free inodes are flushed before we
1316 kmem_cache_destroy(ext4_inode_cachep);
1319 void ext4_clear_inode(struct inode *inode)
1321 invalidate_inode_buffers(inode);
1323 ext4_discard_preallocations(inode, 0);
1324 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
1326 if (EXT4_I(inode)->jinode) {
1327 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1328 EXT4_I(inode)->jinode);
1329 jbd2_free_inode(EXT4_I(inode)->jinode);
1330 EXT4_I(inode)->jinode = NULL;
1332 fscrypt_put_encryption_info(inode);
1333 fsverity_cleanup_inode(inode);
1336 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1337 u64 ino, u32 generation)
1339 struct inode *inode;
1342 * Currently we don't know the generation for parent directory, so
1343 * a generation of 0 means "accept any"
1345 inode = ext4_iget(sb, ino, EXT4_IGET_HANDLE);
1347 return ERR_CAST(inode);
1348 if (generation && inode->i_generation != generation) {
1350 return ERR_PTR(-ESTALE);
1356 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1357 int fh_len, int fh_type)
1359 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1360 ext4_nfs_get_inode);
1363 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1364 int fh_len, int fh_type)
1366 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1367 ext4_nfs_get_inode);
1370 static int ext4_nfs_commit_metadata(struct inode *inode)
1372 struct writeback_control wbc = {
1373 .sync_mode = WB_SYNC_ALL
1376 trace_ext4_nfs_commit_metadata(inode);
1377 return ext4_write_inode(inode, &wbc);
1381 * Try to release metadata pages (indirect blocks, directories) which are
1382 * mapped via the block device. Since these pages could have journal heads
1383 * which would prevent try_to_free_buffers() from freeing them, we must use
1384 * jbd2 layer's try_to_free_buffers() function to release them.
1386 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1389 journal_t *journal = EXT4_SB(sb)->s_journal;
1391 WARN_ON(PageChecked(page));
1392 if (!page_has_buffers(page))
1395 return jbd2_journal_try_to_free_buffers(journal, page);
1397 return try_to_free_buffers(page);
1400 #ifdef CONFIG_FS_ENCRYPTION
1401 static int ext4_get_context(struct inode *inode, void *ctx, size_t len)
1403 return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1404 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len);
1407 static int ext4_set_context(struct inode *inode, const void *ctx, size_t len,
1410 handle_t *handle = fs_data;
1411 int res, res2, credits, retries = 0;
1414 * Encrypting the root directory is not allowed because e2fsck expects
1415 * lost+found to exist and be unencrypted, and encrypting the root
1416 * directory would imply encrypting the lost+found directory as well as
1417 * the filename "lost+found" itself.
1419 if (inode->i_ino == EXT4_ROOT_INO)
1422 if (WARN_ON_ONCE(IS_DAX(inode) && i_size_read(inode)))
1425 if (ext4_test_inode_flag(inode, EXT4_INODE_DAX))
1428 res = ext4_convert_inline_data(inode);
1433 * If a journal handle was specified, then the encryption context is
1434 * being set on a new inode via inheritance and is part of a larger
1435 * transaction to create the inode. Otherwise the encryption context is
1436 * being set on an existing inode in its own transaction. Only in the
1437 * latter case should the "retry on ENOSPC" logic be used.
1441 res = ext4_xattr_set_handle(handle, inode,
1442 EXT4_XATTR_INDEX_ENCRYPTION,
1443 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1446 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1447 ext4_clear_inode_state(inode,
1448 EXT4_STATE_MAY_INLINE_DATA);
1450 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1451 * S_DAX may be disabled
1453 ext4_set_inode_flags(inode, false);
1458 res = dquot_initialize(inode);
1462 res = ext4_xattr_set_credits(inode, len, false /* is_create */,
1467 handle = ext4_journal_start(inode, EXT4_HT_MISC, credits);
1469 return PTR_ERR(handle);
1471 res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION,
1472 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1475 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1477 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1478 * S_DAX may be disabled
1480 ext4_set_inode_flags(inode, false);
1481 res = ext4_mark_inode_dirty(handle, inode);
1483 EXT4_ERROR_INODE(inode, "Failed to mark inode dirty");
1485 res2 = ext4_journal_stop(handle);
1487 if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
1494 static const union fscrypt_context *
1495 ext4_get_dummy_context(struct super_block *sb)
1497 return EXT4_SB(sb)->s_dummy_enc_ctx.ctx;
1500 static bool ext4_has_stable_inodes(struct super_block *sb)
1502 return ext4_has_feature_stable_inodes(sb);
1505 static void ext4_get_ino_and_lblk_bits(struct super_block *sb,
1506 int *ino_bits_ret, int *lblk_bits_ret)
1508 *ino_bits_ret = 8 * sizeof(EXT4_SB(sb)->s_es->s_inodes_count);
1509 *lblk_bits_ret = 8 * sizeof(ext4_lblk_t);
1512 static const struct fscrypt_operations ext4_cryptops = {
1513 .key_prefix = "ext4:",
1514 .get_context = ext4_get_context,
1515 .set_context = ext4_set_context,
1516 .get_dummy_context = ext4_get_dummy_context,
1517 .empty_dir = ext4_empty_dir,
1518 .max_namelen = EXT4_NAME_LEN,
1519 .has_stable_inodes = ext4_has_stable_inodes,
1520 .get_ino_and_lblk_bits = ext4_get_ino_and_lblk_bits,
1525 static const char * const quotatypes[] = INITQFNAMES;
1526 #define QTYPE2NAME(t) (quotatypes[t])
1528 static int ext4_write_dquot(struct dquot *dquot);
1529 static int ext4_acquire_dquot(struct dquot *dquot);
1530 static int ext4_release_dquot(struct dquot *dquot);
1531 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1532 static int ext4_write_info(struct super_block *sb, int type);
1533 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1534 const struct path *path);
1535 static int ext4_quota_on_mount(struct super_block *sb, int type);
1536 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1537 size_t len, loff_t off);
1538 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1539 const char *data, size_t len, loff_t off);
1540 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1541 unsigned int flags);
1542 static int ext4_enable_quotas(struct super_block *sb);
1544 static struct dquot **ext4_get_dquots(struct inode *inode)
1546 return EXT4_I(inode)->i_dquot;
1549 static const struct dquot_operations ext4_quota_operations = {
1550 .get_reserved_space = ext4_get_reserved_space,
1551 .write_dquot = ext4_write_dquot,
1552 .acquire_dquot = ext4_acquire_dquot,
1553 .release_dquot = ext4_release_dquot,
1554 .mark_dirty = ext4_mark_dquot_dirty,
1555 .write_info = ext4_write_info,
1556 .alloc_dquot = dquot_alloc,
1557 .destroy_dquot = dquot_destroy,
1558 .get_projid = ext4_get_projid,
1559 .get_inode_usage = ext4_get_inode_usage,
1560 .get_next_id = dquot_get_next_id,
1563 static const struct quotactl_ops ext4_qctl_operations = {
1564 .quota_on = ext4_quota_on,
1565 .quota_off = ext4_quota_off,
1566 .quota_sync = dquot_quota_sync,
1567 .get_state = dquot_get_state,
1568 .set_info = dquot_set_dqinfo,
1569 .get_dqblk = dquot_get_dqblk,
1570 .set_dqblk = dquot_set_dqblk,
1571 .get_nextdqblk = dquot_get_next_dqblk,
1575 static const struct super_operations ext4_sops = {
1576 .alloc_inode = ext4_alloc_inode,
1577 .free_inode = ext4_free_in_core_inode,
1578 .destroy_inode = ext4_destroy_inode,
1579 .write_inode = ext4_write_inode,
1580 .dirty_inode = ext4_dirty_inode,
1581 .drop_inode = ext4_drop_inode,
1582 .evict_inode = ext4_evict_inode,
1583 .put_super = ext4_put_super,
1584 .sync_fs = ext4_sync_fs,
1585 .freeze_fs = ext4_freeze,
1586 .unfreeze_fs = ext4_unfreeze,
1587 .statfs = ext4_statfs,
1588 .remount_fs = ext4_remount,
1589 .show_options = ext4_show_options,
1591 .quota_read = ext4_quota_read,
1592 .quota_write = ext4_quota_write,
1593 .get_dquots = ext4_get_dquots,
1595 .bdev_try_to_free_page = bdev_try_to_free_page,
1598 static const struct export_operations ext4_export_ops = {
1599 .fh_to_dentry = ext4_fh_to_dentry,
1600 .fh_to_parent = ext4_fh_to_parent,
1601 .get_parent = ext4_get_parent,
1602 .commit_metadata = ext4_nfs_commit_metadata,
1606 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1607 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1608 Opt_nouid32, Opt_debug, Opt_removed,
1609 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1610 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1611 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1612 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1613 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1614 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1616 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1617 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1618 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1619 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version,
1620 Opt_dax, Opt_dax_always, Opt_dax_inode, Opt_dax_never,
1621 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_warn_on_error,
1622 Opt_nowarn_on_error, Opt_mblk_io_submit,
1623 Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize,
1624 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1625 Opt_inode_readahead_blks, Opt_journal_ioprio,
1626 Opt_dioread_nolock, Opt_dioread_lock,
1627 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1628 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache,
1629 Opt_prefetch_block_bitmaps,
1632 static const match_table_t tokens = {
1633 {Opt_bsd_df, "bsddf"},
1634 {Opt_minix_df, "minixdf"},
1635 {Opt_grpid, "grpid"},
1636 {Opt_grpid, "bsdgroups"},
1637 {Opt_nogrpid, "nogrpid"},
1638 {Opt_nogrpid, "sysvgroups"},
1639 {Opt_resgid, "resgid=%u"},
1640 {Opt_resuid, "resuid=%u"},
1642 {Opt_err_cont, "errors=continue"},
1643 {Opt_err_panic, "errors=panic"},
1644 {Opt_err_ro, "errors=remount-ro"},
1645 {Opt_nouid32, "nouid32"},
1646 {Opt_debug, "debug"},
1647 {Opt_removed, "oldalloc"},
1648 {Opt_removed, "orlov"},
1649 {Opt_user_xattr, "user_xattr"},
1650 {Opt_nouser_xattr, "nouser_xattr"},
1652 {Opt_noacl, "noacl"},
1653 {Opt_noload, "norecovery"},
1654 {Opt_noload, "noload"},
1655 {Opt_removed, "nobh"},
1656 {Opt_removed, "bh"},
1657 {Opt_commit, "commit=%u"},
1658 {Opt_min_batch_time, "min_batch_time=%u"},
1659 {Opt_max_batch_time, "max_batch_time=%u"},
1660 {Opt_journal_dev, "journal_dev=%u"},
1661 {Opt_journal_path, "journal_path=%s"},
1662 {Opt_journal_checksum, "journal_checksum"},
1663 {Opt_nojournal_checksum, "nojournal_checksum"},
1664 {Opt_journal_async_commit, "journal_async_commit"},
1665 {Opt_abort, "abort"},
1666 {Opt_data_journal, "data=journal"},
1667 {Opt_data_ordered, "data=ordered"},
1668 {Opt_data_writeback, "data=writeback"},
1669 {Opt_data_err_abort, "data_err=abort"},
1670 {Opt_data_err_ignore, "data_err=ignore"},
1671 {Opt_offusrjquota, "usrjquota="},
1672 {Opt_usrjquota, "usrjquota=%s"},
1673 {Opt_offgrpjquota, "grpjquota="},
1674 {Opt_grpjquota, "grpjquota=%s"},
1675 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1676 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1677 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1678 {Opt_grpquota, "grpquota"},
1679 {Opt_noquota, "noquota"},
1680 {Opt_quota, "quota"},
1681 {Opt_usrquota, "usrquota"},
1682 {Opt_prjquota, "prjquota"},
1683 {Opt_barrier, "barrier=%u"},
1684 {Opt_barrier, "barrier"},
1685 {Opt_nobarrier, "nobarrier"},
1686 {Opt_i_version, "i_version"},
1688 {Opt_dax_always, "dax=always"},
1689 {Opt_dax_inode, "dax=inode"},
1690 {Opt_dax_never, "dax=never"},
1691 {Opt_stripe, "stripe=%u"},
1692 {Opt_delalloc, "delalloc"},
1693 {Opt_warn_on_error, "warn_on_error"},
1694 {Opt_nowarn_on_error, "nowarn_on_error"},
1695 {Opt_lazytime, "lazytime"},
1696 {Opt_nolazytime, "nolazytime"},
1697 {Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"},
1698 {Opt_nodelalloc, "nodelalloc"},
1699 {Opt_removed, "mblk_io_submit"},
1700 {Opt_removed, "nomblk_io_submit"},
1701 {Opt_block_validity, "block_validity"},
1702 {Opt_noblock_validity, "noblock_validity"},
1703 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1704 {Opt_journal_ioprio, "journal_ioprio=%u"},
1705 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1706 {Opt_auto_da_alloc, "auto_da_alloc"},
1707 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1708 {Opt_dioread_nolock, "dioread_nolock"},
1709 {Opt_dioread_lock, "nodioread_nolock"},
1710 {Opt_dioread_lock, "dioread_lock"},
1711 {Opt_discard, "discard"},
1712 {Opt_nodiscard, "nodiscard"},
1713 {Opt_init_itable, "init_itable=%u"},
1714 {Opt_init_itable, "init_itable"},
1715 {Opt_noinit_itable, "noinit_itable"},
1716 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1717 {Opt_test_dummy_encryption, "test_dummy_encryption=%s"},
1718 {Opt_test_dummy_encryption, "test_dummy_encryption"},
1719 {Opt_inlinecrypt, "inlinecrypt"},
1720 {Opt_nombcache, "nombcache"},
1721 {Opt_nombcache, "no_mbcache"}, /* for backward compatibility */
1722 {Opt_prefetch_block_bitmaps, "prefetch_block_bitmaps"},
1723 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1724 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1725 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1726 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1727 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1731 static ext4_fsblk_t get_sb_block(void **data)
1733 ext4_fsblk_t sb_block;
1734 char *options = (char *) *data;
1736 if (!options || strncmp(options, "sb=", 3) != 0)
1737 return 1; /* Default location */
1740 /* TODO: use simple_strtoll with >32bit ext4 */
1741 sb_block = simple_strtoul(options, &options, 0);
1742 if (*options && *options != ',') {
1743 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1747 if (*options == ',')
1749 *data = (void *) options;
1754 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1755 static const char deprecated_msg[] =
1756 "Mount option \"%s\" will be removed by %s\n"
1757 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1760 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1762 struct ext4_sb_info *sbi = EXT4_SB(sb);
1763 char *qname, *old_qname = get_qf_name(sb, sbi, qtype);
1766 if (sb_any_quota_loaded(sb) && !old_qname) {
1767 ext4_msg(sb, KERN_ERR,
1768 "Cannot change journaled "
1769 "quota options when quota turned on");
1772 if (ext4_has_feature_quota(sb)) {
1773 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1774 "ignored when QUOTA feature is enabled");
1777 qname = match_strdup(args);
1779 ext4_msg(sb, KERN_ERR,
1780 "Not enough memory for storing quotafile name");
1784 if (strcmp(old_qname, qname) == 0)
1787 ext4_msg(sb, KERN_ERR,
1788 "%s quota file already specified",
1792 if (strchr(qname, '/')) {
1793 ext4_msg(sb, KERN_ERR,
1794 "quotafile must be on filesystem root");
1797 rcu_assign_pointer(sbi->s_qf_names[qtype], qname);
1805 static int clear_qf_name(struct super_block *sb, int qtype)
1808 struct ext4_sb_info *sbi = EXT4_SB(sb);
1809 char *old_qname = get_qf_name(sb, sbi, qtype);
1811 if (sb_any_quota_loaded(sb) && old_qname) {
1812 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1813 " when quota turned on");
1816 rcu_assign_pointer(sbi->s_qf_names[qtype], NULL);
1823 #define MOPT_SET 0x0001
1824 #define MOPT_CLEAR 0x0002
1825 #define MOPT_NOSUPPORT 0x0004
1826 #define MOPT_EXPLICIT 0x0008
1827 #define MOPT_CLEAR_ERR 0x0010
1828 #define MOPT_GTE0 0x0020
1831 #define MOPT_QFMT 0x0040
1833 #define MOPT_Q MOPT_NOSUPPORT
1834 #define MOPT_QFMT MOPT_NOSUPPORT
1836 #define MOPT_DATAJ 0x0080
1837 #define MOPT_NO_EXT2 0x0100
1838 #define MOPT_NO_EXT3 0x0200
1839 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1840 #define MOPT_STRING 0x0400
1841 #define MOPT_SKIP 0x0800
1843 static const struct mount_opts {
1847 } ext4_mount_opts[] = {
1848 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1849 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1850 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1851 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1852 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1853 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1854 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1855 MOPT_EXT4_ONLY | MOPT_SET},
1856 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1857 MOPT_EXT4_ONLY | MOPT_CLEAR},
1858 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1859 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1860 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1861 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1862 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1863 MOPT_EXT4_ONLY | MOPT_CLEAR},
1864 {Opt_warn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_SET},
1865 {Opt_nowarn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_CLEAR},
1866 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1867 MOPT_EXT4_ONLY | MOPT_CLEAR},
1868 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1869 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1870 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1871 EXT4_MOUNT_JOURNAL_CHECKSUM),
1872 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1873 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1874 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1875 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1876 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1877 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1879 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1881 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1882 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1883 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1884 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1885 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1886 {Opt_commit, 0, MOPT_GTE0},
1887 {Opt_max_batch_time, 0, MOPT_GTE0},
1888 {Opt_min_batch_time, 0, MOPT_GTE0},
1889 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1890 {Opt_init_itable, 0, MOPT_GTE0},
1891 {Opt_dax, EXT4_MOUNT_DAX_ALWAYS, MOPT_SET | MOPT_SKIP},
1892 {Opt_dax_always, EXT4_MOUNT_DAX_ALWAYS,
1893 MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP},
1894 {Opt_dax_inode, EXT4_MOUNT2_DAX_INODE,
1895 MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP},
1896 {Opt_dax_never, EXT4_MOUNT2_DAX_NEVER,
1897 MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP},
1898 {Opt_stripe, 0, MOPT_GTE0},
1899 {Opt_resuid, 0, MOPT_GTE0},
1900 {Opt_resgid, 0, MOPT_GTE0},
1901 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1902 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1903 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1904 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1905 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1906 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1907 MOPT_NO_EXT2 | MOPT_DATAJ},
1908 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1909 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1910 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1911 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1912 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1914 {Opt_acl, 0, MOPT_NOSUPPORT},
1915 {Opt_noacl, 0, MOPT_NOSUPPORT},
1917 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1918 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1919 {Opt_debug_want_extra_isize, 0, MOPT_GTE0},
1920 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1921 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1923 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1925 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1927 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1928 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1929 MOPT_CLEAR | MOPT_Q},
1930 {Opt_usrjquota, 0, MOPT_Q},
1931 {Opt_grpjquota, 0, MOPT_Q},
1932 {Opt_offusrjquota, 0, MOPT_Q},
1933 {Opt_offgrpjquota, 0, MOPT_Q},
1934 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1935 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1936 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1937 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1938 {Opt_test_dummy_encryption, 0, MOPT_STRING},
1939 {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET},
1940 {Opt_prefetch_block_bitmaps, EXT4_MOUNT_PREFETCH_BLOCK_BITMAPS,
1945 #ifdef CONFIG_UNICODE
1946 static const struct ext4_sb_encodings {
1950 } ext4_sb_encoding_map[] = {
1951 {EXT4_ENC_UTF8_12_1, "utf8", "12.1.0"},
1954 static int ext4_sb_read_encoding(const struct ext4_super_block *es,
1955 const struct ext4_sb_encodings **encoding,
1958 __u16 magic = le16_to_cpu(es->s_encoding);
1961 for (i = 0; i < ARRAY_SIZE(ext4_sb_encoding_map); i++)
1962 if (magic == ext4_sb_encoding_map[i].magic)
1965 if (i >= ARRAY_SIZE(ext4_sb_encoding_map))
1968 *encoding = &ext4_sb_encoding_map[i];
1969 *flags = le16_to_cpu(es->s_encoding_flags);
1975 static int ext4_set_test_dummy_encryption(struct super_block *sb,
1977 const substring_t *arg,
1980 #ifdef CONFIG_FS_ENCRYPTION
1981 struct ext4_sb_info *sbi = EXT4_SB(sb);
1985 * This mount option is just for testing, and it's not worthwhile to
1986 * implement the extra complexity (e.g. RCU protection) that would be
1987 * needed to allow it to be set or changed during remount. We do allow
1988 * it to be specified during remount, but only if there is no change.
1990 if (is_remount && !sbi->s_dummy_enc_ctx.ctx) {
1991 ext4_msg(sb, KERN_WARNING,
1992 "Can't set test_dummy_encryption on remount");
1995 err = fscrypt_set_test_dummy_encryption(sb, arg, &sbi->s_dummy_enc_ctx);
1998 ext4_msg(sb, KERN_WARNING,
1999 "Can't change test_dummy_encryption on remount");
2000 else if (err == -EINVAL)
2001 ext4_msg(sb, KERN_WARNING,
2002 "Value of option \"%s\" is unrecognized", opt);
2004 ext4_msg(sb, KERN_WARNING,
2005 "Error processing option \"%s\" [%d]",
2009 ext4_msg(sb, KERN_WARNING, "Test dummy encryption mode enabled");
2011 ext4_msg(sb, KERN_WARNING,
2012 "Test dummy encryption mount option ignored");
2017 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
2018 substring_t *args, unsigned long *journal_devnum,
2019 unsigned int *journal_ioprio, int is_remount)
2021 struct ext4_sb_info *sbi = EXT4_SB(sb);
2022 const struct mount_opts *m;
2028 if (token == Opt_usrjquota)
2029 return set_qf_name(sb, USRQUOTA, &args[0]);
2030 else if (token == Opt_grpjquota)
2031 return set_qf_name(sb, GRPQUOTA, &args[0]);
2032 else if (token == Opt_offusrjquota)
2033 return clear_qf_name(sb, USRQUOTA);
2034 else if (token == Opt_offgrpjquota)
2035 return clear_qf_name(sb, GRPQUOTA);
2039 case Opt_nouser_xattr:
2040 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
2043 return 1; /* handled by get_sb_block() */
2045 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
2048 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
2051 sb->s_flags |= SB_I_VERSION;
2054 sb->s_flags |= SB_LAZYTIME;
2056 case Opt_nolazytime:
2057 sb->s_flags &= ~SB_LAZYTIME;
2059 case Opt_inlinecrypt:
2060 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT
2061 sb->s_flags |= SB_INLINECRYPT;
2063 ext4_msg(sb, KERN_ERR, "inline encryption not supported");
2068 for (m = ext4_mount_opts; m->token != Opt_err; m++)
2069 if (token == m->token)
2072 if (m->token == Opt_err) {
2073 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
2074 "or missing value", opt);
2078 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
2079 ext4_msg(sb, KERN_ERR,
2080 "Mount option \"%s\" incompatible with ext2", opt);
2083 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
2084 ext4_msg(sb, KERN_ERR,
2085 "Mount option \"%s\" incompatible with ext3", opt);
2089 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
2091 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
2093 if (m->flags & MOPT_EXPLICIT) {
2094 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
2095 set_opt2(sb, EXPLICIT_DELALLOC);
2096 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
2097 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
2101 if (m->flags & MOPT_CLEAR_ERR)
2102 clear_opt(sb, ERRORS_MASK);
2103 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
2104 ext4_msg(sb, KERN_ERR, "Cannot change quota "
2105 "options when quota turned on");
2109 if (m->flags & MOPT_NOSUPPORT) {
2110 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
2111 } else if (token == Opt_commit) {
2113 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
2114 else if (arg > INT_MAX / HZ) {
2115 ext4_msg(sb, KERN_ERR,
2116 "Invalid commit interval %d, "
2117 "must be smaller than %d",
2121 sbi->s_commit_interval = HZ * arg;
2122 } else if (token == Opt_debug_want_extra_isize) {
2125 (arg > (sbi->s_inode_size - EXT4_GOOD_OLD_INODE_SIZE))) {
2126 ext4_msg(sb, KERN_ERR,
2127 "Invalid want_extra_isize %d", arg);
2130 sbi->s_want_extra_isize = arg;
2131 } else if (token == Opt_max_batch_time) {
2132 sbi->s_max_batch_time = arg;
2133 } else if (token == Opt_min_batch_time) {
2134 sbi->s_min_batch_time = arg;
2135 } else if (token == Opt_inode_readahead_blks) {
2136 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
2137 ext4_msg(sb, KERN_ERR,
2138 "EXT4-fs: inode_readahead_blks must be "
2139 "0 or a power of 2 smaller than 2^31");
2142 sbi->s_inode_readahead_blks = arg;
2143 } else if (token == Opt_init_itable) {
2144 set_opt(sb, INIT_INODE_TABLE);
2146 arg = EXT4_DEF_LI_WAIT_MULT;
2147 sbi->s_li_wait_mult = arg;
2148 } else if (token == Opt_max_dir_size_kb) {
2149 sbi->s_max_dir_size_kb = arg;
2150 } else if (token == Opt_stripe) {
2151 sbi->s_stripe = arg;
2152 } else if (token == Opt_resuid) {
2153 uid = make_kuid(current_user_ns(), arg);
2154 if (!uid_valid(uid)) {
2155 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
2158 sbi->s_resuid = uid;
2159 } else if (token == Opt_resgid) {
2160 gid = make_kgid(current_user_ns(), arg);
2161 if (!gid_valid(gid)) {
2162 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
2165 sbi->s_resgid = gid;
2166 } else if (token == Opt_journal_dev) {
2168 ext4_msg(sb, KERN_ERR,
2169 "Cannot specify journal on remount");
2172 *journal_devnum = arg;
2173 } else if (token == Opt_journal_path) {
2175 struct inode *journal_inode;
2180 ext4_msg(sb, KERN_ERR,
2181 "Cannot specify journal on remount");
2184 journal_path = match_strdup(&args[0]);
2185 if (!journal_path) {
2186 ext4_msg(sb, KERN_ERR, "error: could not dup "
2187 "journal device string");
2191 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
2193 ext4_msg(sb, KERN_ERR, "error: could not find "
2194 "journal device path: error %d", error);
2195 kfree(journal_path);
2199 journal_inode = d_inode(path.dentry);
2200 if (!S_ISBLK(journal_inode->i_mode)) {
2201 ext4_msg(sb, KERN_ERR, "error: journal path %s "
2202 "is not a block device", journal_path);
2204 kfree(journal_path);
2208 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
2210 kfree(journal_path);
2211 } else if (token == Opt_journal_ioprio) {
2213 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
2218 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
2219 } else if (token == Opt_test_dummy_encryption) {
2220 return ext4_set_test_dummy_encryption(sb, opt, &args[0],
2222 } else if (m->flags & MOPT_DATAJ) {
2224 if (!sbi->s_journal)
2225 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
2226 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
2227 ext4_msg(sb, KERN_ERR,
2228 "Cannot change data mode on remount");
2232 clear_opt(sb, DATA_FLAGS);
2233 sbi->s_mount_opt |= m->mount_opt;
2236 } else if (m->flags & MOPT_QFMT) {
2237 if (sb_any_quota_loaded(sb) &&
2238 sbi->s_jquota_fmt != m->mount_opt) {
2239 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
2240 "quota options when quota turned on");
2243 if (ext4_has_feature_quota(sb)) {
2244 ext4_msg(sb, KERN_INFO,
2245 "Quota format mount options ignored "
2246 "when QUOTA feature is enabled");
2249 sbi->s_jquota_fmt = m->mount_opt;
2251 } else if (token == Opt_dax || token == Opt_dax_always ||
2252 token == Opt_dax_inode || token == Opt_dax_never) {
2253 #ifdef CONFIG_FS_DAX
2256 case Opt_dax_always:
2258 (!(sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2259 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER))) {
2260 fail_dax_change_remount:
2261 ext4_msg(sb, KERN_ERR, "can't change "
2262 "dax mount option while remounting");
2266 (test_opt(sb, DATA_FLAGS) ==
2267 EXT4_MOUNT_JOURNAL_DATA)) {
2268 ext4_msg(sb, KERN_ERR, "can't mount with "
2269 "both data=journal and dax");
2272 ext4_msg(sb, KERN_WARNING,
2273 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
2274 sbi->s_mount_opt |= EXT4_MOUNT_DAX_ALWAYS;
2275 sbi->s_mount_opt2 &= ~EXT4_MOUNT2_DAX_NEVER;
2279 (!(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2280 (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS)))
2281 goto fail_dax_change_remount;
2282 sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_NEVER;
2283 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS;
2287 ((sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2288 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2289 !(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_INODE)))
2290 goto fail_dax_change_remount;
2291 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS;
2292 sbi->s_mount_opt2 &= ~EXT4_MOUNT2_DAX_NEVER;
2293 /* Strictly for printing options */
2294 sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_INODE;
2298 ext4_msg(sb, KERN_INFO, "dax option not supported");
2299 sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_NEVER;
2300 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS;
2303 } else if (token == Opt_data_err_abort) {
2304 sbi->s_mount_opt |= m->mount_opt;
2305 } else if (token == Opt_data_err_ignore) {
2306 sbi->s_mount_opt &= ~m->mount_opt;
2310 if (m->flags & MOPT_CLEAR)
2312 else if (unlikely(!(m->flags & MOPT_SET))) {
2313 ext4_msg(sb, KERN_WARNING,
2314 "buggy handling of option %s", opt);
2319 sbi->s_mount_opt |= m->mount_opt;
2321 sbi->s_mount_opt &= ~m->mount_opt;
2326 static int parse_options(char *options, struct super_block *sb,
2327 unsigned long *journal_devnum,
2328 unsigned int *journal_ioprio,
2331 struct ext4_sb_info __maybe_unused *sbi = EXT4_SB(sb);
2332 char *p, __maybe_unused *usr_qf_name, __maybe_unused *grp_qf_name;
2333 substring_t args[MAX_OPT_ARGS];
2339 while ((p = strsep(&options, ",")) != NULL) {
2343 * Initialize args struct so we know whether arg was
2344 * found; some options take optional arguments.
2346 args[0].to = args[0].from = NULL;
2347 token = match_token(p, tokens, args);
2348 if (handle_mount_opt(sb, p, token, args, journal_devnum,
2349 journal_ioprio, is_remount) < 0)
2354 * We do the test below only for project quotas. 'usrquota' and
2355 * 'grpquota' mount options are allowed even without quota feature
2356 * to support legacy quotas in quota files.
2358 if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) {
2359 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. "
2360 "Cannot enable project quota enforcement.");
2363 usr_qf_name = get_qf_name(sb, sbi, USRQUOTA);
2364 grp_qf_name = get_qf_name(sb, sbi, GRPQUOTA);
2365 if (usr_qf_name || grp_qf_name) {
2366 if (test_opt(sb, USRQUOTA) && usr_qf_name)
2367 clear_opt(sb, USRQUOTA);
2369 if (test_opt(sb, GRPQUOTA) && grp_qf_name)
2370 clear_opt(sb, GRPQUOTA);
2372 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
2373 ext4_msg(sb, KERN_ERR, "old and new quota "
2378 if (!sbi->s_jquota_fmt) {
2379 ext4_msg(sb, KERN_ERR, "journaled quota format "
2385 if (test_opt(sb, DIOREAD_NOLOCK)) {
2387 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
2388 if (blocksize < PAGE_SIZE)
2389 ext4_msg(sb, KERN_WARNING, "Warning: mounting with an "
2390 "experimental mount option 'dioread_nolock' "
2391 "for blocksize < PAGE_SIZE");
2396 static inline void ext4_show_quota_options(struct seq_file *seq,
2397 struct super_block *sb)
2399 #if defined(CONFIG_QUOTA)
2400 struct ext4_sb_info *sbi = EXT4_SB(sb);
2401 char *usr_qf_name, *grp_qf_name;
2403 if (sbi->s_jquota_fmt) {
2406 switch (sbi->s_jquota_fmt) {
2417 seq_printf(seq, ",jqfmt=%s", fmtname);
2421 usr_qf_name = rcu_dereference(sbi->s_qf_names[USRQUOTA]);
2422 grp_qf_name = rcu_dereference(sbi->s_qf_names[GRPQUOTA]);
2424 seq_show_option(seq, "usrjquota", usr_qf_name);
2426 seq_show_option(seq, "grpjquota", grp_qf_name);
2431 static const char *token2str(int token)
2433 const struct match_token *t;
2435 for (t = tokens; t->token != Opt_err; t++)
2436 if (t->token == token && !strchr(t->pattern, '='))
2443 * - it's set to a non-default value OR
2444 * - if the per-sb default is different from the global default
2446 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
2449 struct ext4_sb_info *sbi = EXT4_SB(sb);
2450 struct ext4_super_block *es = sbi->s_es;
2451 int def_errors, def_mount_opt = sbi->s_def_mount_opt;
2452 const struct mount_opts *m;
2453 char sep = nodefs ? '\n' : ',';
2455 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2456 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2458 if (sbi->s_sb_block != 1)
2459 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
2461 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
2462 int want_set = m->flags & MOPT_SET;
2463 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
2464 (m->flags & MOPT_CLEAR_ERR) || m->flags & MOPT_SKIP)
2466 if (!nodefs && !(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
2467 continue; /* skip if same as the default */
2469 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
2470 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
2471 continue; /* select Opt_noFoo vs Opt_Foo */
2472 SEQ_OPTS_PRINT("%s", token2str(m->token));
2475 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
2476 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
2477 SEQ_OPTS_PRINT("resuid=%u",
2478 from_kuid_munged(&init_user_ns, sbi->s_resuid));
2479 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
2480 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
2481 SEQ_OPTS_PRINT("resgid=%u",
2482 from_kgid_munged(&init_user_ns, sbi->s_resgid));
2483 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
2484 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
2485 SEQ_OPTS_PUTS("errors=remount-ro");
2486 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
2487 SEQ_OPTS_PUTS("errors=continue");
2488 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
2489 SEQ_OPTS_PUTS("errors=panic");
2490 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
2491 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
2492 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
2493 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
2494 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
2495 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2496 if (sb->s_flags & SB_I_VERSION)
2497 SEQ_OPTS_PUTS("i_version");
2498 if (nodefs || sbi->s_stripe)
2499 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
2500 if (nodefs || EXT4_MOUNT_DATA_FLAGS &
2501 (sbi->s_mount_opt ^ def_mount_opt)) {
2502 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2503 SEQ_OPTS_PUTS("data=journal");
2504 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2505 SEQ_OPTS_PUTS("data=ordered");
2506 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2507 SEQ_OPTS_PUTS("data=writeback");
2510 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
2511 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2512 sbi->s_inode_readahead_blks);
2514 if (test_opt(sb, INIT_INODE_TABLE) && (nodefs ||
2515 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
2516 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
2517 if (nodefs || sbi->s_max_dir_size_kb)
2518 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2519 if (test_opt(sb, DATA_ERR_ABORT))
2520 SEQ_OPTS_PUTS("data_err=abort");
2522 fscrypt_show_test_dummy_encryption(seq, sep, sb);
2524 if (sb->s_flags & SB_INLINECRYPT)
2525 SEQ_OPTS_PUTS("inlinecrypt");
2527 if (test_opt(sb, DAX_ALWAYS)) {
2529 SEQ_OPTS_PUTS("dax");
2531 SEQ_OPTS_PUTS("dax=always");
2532 } else if (test_opt2(sb, DAX_NEVER)) {
2533 SEQ_OPTS_PUTS("dax=never");
2534 } else if (test_opt2(sb, DAX_INODE)) {
2535 SEQ_OPTS_PUTS("dax=inode");
2538 ext4_show_quota_options(seq, sb);
2542 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2544 return _ext4_show_options(seq, root->d_sb, 0);
2547 int ext4_seq_options_show(struct seq_file *seq, void *offset)
2549 struct super_block *sb = seq->private;
2552 seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw");
2553 rc = _ext4_show_options(seq, sb, 1);
2554 seq_puts(seq, "\n");
2558 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
2561 struct ext4_sb_info *sbi = EXT4_SB(sb);
2564 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
2565 ext4_msg(sb, KERN_ERR, "revision level too high, "
2566 "forcing read-only mode");
2572 if (!(sbi->s_mount_state & EXT4_VALID_FS))
2573 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
2574 "running e2fsck is recommended");
2575 else if (sbi->s_mount_state & EXT4_ERROR_FS)
2576 ext4_msg(sb, KERN_WARNING,
2577 "warning: mounting fs with errors, "
2578 "running e2fsck is recommended");
2579 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
2580 le16_to_cpu(es->s_mnt_count) >=
2581 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
2582 ext4_msg(sb, KERN_WARNING,
2583 "warning: maximal mount count reached, "
2584 "running e2fsck is recommended");
2585 else if (le32_to_cpu(es->s_checkinterval) &&
2586 (ext4_get_tstamp(es, s_lastcheck) +
2587 le32_to_cpu(es->s_checkinterval) <= ktime_get_real_seconds()))
2588 ext4_msg(sb, KERN_WARNING,
2589 "warning: checktime reached, "
2590 "running e2fsck is recommended");
2591 if (!sbi->s_journal)
2592 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
2593 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
2594 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
2595 le16_add_cpu(&es->s_mnt_count, 1);
2596 ext4_update_tstamp(es, s_mtime);
2598 ext4_set_feature_journal_needs_recovery(sb);
2600 err = ext4_commit_super(sb, 1);
2602 if (test_opt(sb, DEBUG))
2603 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
2604 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
2606 sbi->s_groups_count,
2607 EXT4_BLOCKS_PER_GROUP(sb),
2608 EXT4_INODES_PER_GROUP(sb),
2609 sbi->s_mount_opt, sbi->s_mount_opt2);
2611 cleancache_init_fs(sb);
2615 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2617 struct ext4_sb_info *sbi = EXT4_SB(sb);
2618 struct flex_groups **old_groups, **new_groups;
2621 if (!sbi->s_log_groups_per_flex)
2624 size = ext4_flex_group(sbi, ngroup - 1) + 1;
2625 if (size <= sbi->s_flex_groups_allocated)
2628 new_groups = kvzalloc(roundup_pow_of_two(size *
2629 sizeof(*sbi->s_flex_groups)), GFP_KERNEL);
2631 ext4_msg(sb, KERN_ERR,
2632 "not enough memory for %d flex group pointers", size);
2635 for (i = sbi->s_flex_groups_allocated; i < size; i++) {
2636 new_groups[i] = kvzalloc(roundup_pow_of_two(
2637 sizeof(struct flex_groups)),
2639 if (!new_groups[i]) {
2640 for (j = sbi->s_flex_groups_allocated; j < i; j++)
2641 kvfree(new_groups[j]);
2643 ext4_msg(sb, KERN_ERR,
2644 "not enough memory for %d flex groups", size);
2649 old_groups = rcu_dereference(sbi->s_flex_groups);
2651 memcpy(new_groups, old_groups,
2652 (sbi->s_flex_groups_allocated *
2653 sizeof(struct flex_groups *)));
2655 rcu_assign_pointer(sbi->s_flex_groups, new_groups);
2656 sbi->s_flex_groups_allocated = size;
2658 ext4_kvfree_array_rcu(old_groups);
2662 static int ext4_fill_flex_info(struct super_block *sb)
2664 struct ext4_sb_info *sbi = EXT4_SB(sb);
2665 struct ext4_group_desc *gdp = NULL;
2666 struct flex_groups *fg;
2667 ext4_group_t flex_group;
2670 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2671 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
2672 sbi->s_log_groups_per_flex = 0;
2676 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2680 for (i = 0; i < sbi->s_groups_count; i++) {
2681 gdp = ext4_get_group_desc(sb, i, NULL);
2683 flex_group = ext4_flex_group(sbi, i);
2684 fg = sbi_array_rcu_deref(sbi, s_flex_groups, flex_group);
2685 atomic_add(ext4_free_inodes_count(sb, gdp), &fg->free_inodes);
2686 atomic64_add(ext4_free_group_clusters(sb, gdp),
2687 &fg->free_clusters);
2688 atomic_add(ext4_used_dirs_count(sb, gdp), &fg->used_dirs);
2696 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
2697 struct ext4_group_desc *gdp)
2699 int offset = offsetof(struct ext4_group_desc, bg_checksum);
2701 __le32 le_group = cpu_to_le32(block_group);
2702 struct ext4_sb_info *sbi = EXT4_SB(sb);
2704 if (ext4_has_metadata_csum(sbi->s_sb)) {
2705 /* Use new metadata_csum algorithm */
2707 __u16 dummy_csum = 0;
2709 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2711 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2712 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2713 sizeof(dummy_csum));
2714 offset += sizeof(dummy_csum);
2715 if (offset < sbi->s_desc_size)
2716 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2717 sbi->s_desc_size - offset);
2719 crc = csum32 & 0xFFFF;
2723 /* old crc16 code */
2724 if (!ext4_has_feature_gdt_csum(sb))
2727 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2728 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2729 crc = crc16(crc, (__u8 *)gdp, offset);
2730 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2731 /* for checksum of struct ext4_group_desc do the rest...*/
2732 if (ext4_has_feature_64bit(sb) &&
2733 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2734 crc = crc16(crc, (__u8 *)gdp + offset,
2735 le16_to_cpu(sbi->s_es->s_desc_size) -
2739 return cpu_to_le16(crc);
2742 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2743 struct ext4_group_desc *gdp)
2745 if (ext4_has_group_desc_csum(sb) &&
2746 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
2752 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2753 struct ext4_group_desc *gdp)
2755 if (!ext4_has_group_desc_csum(sb))
2757 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
2760 /* Called at mount-time, super-block is locked */
2761 static int ext4_check_descriptors(struct super_block *sb,
2762 ext4_fsblk_t sb_block,
2763 ext4_group_t *first_not_zeroed)
2765 struct ext4_sb_info *sbi = EXT4_SB(sb);
2766 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2767 ext4_fsblk_t last_block;
2768 ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0);
2769 ext4_fsblk_t block_bitmap;
2770 ext4_fsblk_t inode_bitmap;
2771 ext4_fsblk_t inode_table;
2772 int flexbg_flag = 0;
2773 ext4_group_t i, grp = sbi->s_groups_count;
2775 if (ext4_has_feature_flex_bg(sb))
2778 ext4_debug("Checking group descriptors");
2780 for (i = 0; i < sbi->s_groups_count; i++) {
2781 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2783 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2784 last_block = ext4_blocks_count(sbi->s_es) - 1;
2786 last_block = first_block +
2787 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2789 if ((grp == sbi->s_groups_count) &&
2790 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2793 block_bitmap = ext4_block_bitmap(sb, gdp);
2794 if (block_bitmap == sb_block) {
2795 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2796 "Block bitmap for group %u overlaps "
2801 if (block_bitmap >= sb_block + 1 &&
2802 block_bitmap <= last_bg_block) {
2803 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2804 "Block bitmap for group %u overlaps "
2805 "block group descriptors", i);
2809 if (block_bitmap < first_block || block_bitmap > last_block) {
2810 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2811 "Block bitmap for group %u not in group "
2812 "(block %llu)!", i, block_bitmap);
2815 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2816 if (inode_bitmap == sb_block) {
2817 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2818 "Inode bitmap for group %u overlaps "
2823 if (inode_bitmap >= sb_block + 1 &&
2824 inode_bitmap <= last_bg_block) {
2825 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2826 "Inode bitmap for group %u overlaps "
2827 "block group descriptors", i);
2831 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2832 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2833 "Inode bitmap for group %u not in group "
2834 "(block %llu)!", i, inode_bitmap);
2837 inode_table = ext4_inode_table(sb, gdp);
2838 if (inode_table == sb_block) {
2839 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2840 "Inode table for group %u overlaps "
2845 if (inode_table >= sb_block + 1 &&
2846 inode_table <= last_bg_block) {
2847 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2848 "Inode table for group %u overlaps "
2849 "block group descriptors", i);
2853 if (inode_table < first_block ||
2854 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2855 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2856 "Inode table for group %u not in group "
2857 "(block %llu)!", i, inode_table);
2860 ext4_lock_group(sb, i);
2861 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2862 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2863 "Checksum for group %u failed (%u!=%u)",
2864 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
2865 gdp)), le16_to_cpu(gdp->bg_checksum));
2866 if (!sb_rdonly(sb)) {
2867 ext4_unlock_group(sb, i);
2871 ext4_unlock_group(sb, i);
2873 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2875 if (NULL != first_not_zeroed)
2876 *first_not_zeroed = grp;
2880 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2881 * the superblock) which were deleted from all directories, but held open by
2882 * a process at the time of a crash. We walk the list and try to delete these
2883 * inodes at recovery time (only with a read-write filesystem).
2885 * In order to keep the orphan inode chain consistent during traversal (in
2886 * case of crash during recovery), we link each inode into the superblock
2887 * orphan list_head and handle it the same way as an inode deletion during
2888 * normal operation (which journals the operations for us).
2890 * We only do an iget() and an iput() on each inode, which is very safe if we
2891 * accidentally point at an in-use or already deleted inode. The worst that
2892 * can happen in this case is that we get a "bit already cleared" message from
2893 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2894 * e2fsck was run on this filesystem, and it must have already done the orphan
2895 * inode cleanup for us, so we can safely abort without any further action.
2897 static void ext4_orphan_cleanup(struct super_block *sb,
2898 struct ext4_super_block *es)
2900 unsigned int s_flags = sb->s_flags;
2901 int ret, nr_orphans = 0, nr_truncates = 0;
2903 int quota_update = 0;
2906 if (!es->s_last_orphan) {
2907 jbd_debug(4, "no orphan inodes to clean up\n");
2911 if (bdev_read_only(sb->s_bdev)) {
2912 ext4_msg(sb, KERN_ERR, "write access "
2913 "unavailable, skipping orphan cleanup");
2917 /* Check if feature set would not allow a r/w mount */
2918 if (!ext4_feature_set_ok(sb, 0)) {
2919 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2920 "unknown ROCOMPAT features");
2924 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2925 /* don't clear list on RO mount w/ errors */
2926 if (es->s_last_orphan && !(s_flags & SB_RDONLY)) {
2927 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
2928 "clearing orphan list.\n");
2929 es->s_last_orphan = 0;
2931 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2935 if (s_flags & SB_RDONLY) {
2936 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2937 sb->s_flags &= ~SB_RDONLY;
2940 /* Needed for iput() to work correctly and not trash data */
2941 sb->s_flags |= SB_ACTIVE;
2944 * Turn on quotas which were not enabled for read-only mounts if
2945 * filesystem has quota feature, so that they are updated correctly.
2947 if (ext4_has_feature_quota(sb) && (s_flags & SB_RDONLY)) {
2948 int ret = ext4_enable_quotas(sb);
2953 ext4_msg(sb, KERN_ERR,
2954 "Cannot turn on quotas: error %d", ret);
2957 /* Turn on journaled quotas used for old sytle */
2958 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2959 if (EXT4_SB(sb)->s_qf_names[i]) {
2960 int ret = ext4_quota_on_mount(sb, i);
2965 ext4_msg(sb, KERN_ERR,
2966 "Cannot turn on journaled "
2967 "quota: type %d: error %d", i, ret);
2972 while (es->s_last_orphan) {
2973 struct inode *inode;
2976 * We may have encountered an error during cleanup; if
2977 * so, skip the rest.
2979 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2980 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2981 es->s_last_orphan = 0;
2985 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2986 if (IS_ERR(inode)) {
2987 es->s_last_orphan = 0;
2991 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2992 dquot_initialize(inode);
2993 if (inode->i_nlink) {
2994 if (test_opt(sb, DEBUG))
2995 ext4_msg(sb, KERN_DEBUG,
2996 "%s: truncating inode %lu to %lld bytes",
2997 __func__, inode->i_ino, inode->i_size);
2998 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2999 inode->i_ino, inode->i_size);
3001 truncate_inode_pages(inode->i_mapping, inode->i_size);
3002 ret = ext4_truncate(inode);
3004 ext4_std_error(inode->i_sb, ret);
3005 inode_unlock(inode);
3008 if (test_opt(sb, DEBUG))
3009 ext4_msg(sb, KERN_DEBUG,
3010 "%s: deleting unreferenced inode %lu",
3011 __func__, inode->i_ino);
3012 jbd_debug(2, "deleting unreferenced inode %lu\n",
3016 iput(inode); /* The delete magic happens here! */
3019 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
3022 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
3023 PLURAL(nr_orphans));
3025 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
3026 PLURAL(nr_truncates));
3028 /* Turn off quotas if they were enabled for orphan cleanup */
3030 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
3031 if (sb_dqopt(sb)->files[i])
3032 dquot_quota_off(sb, i);
3036 sb->s_flags = s_flags; /* Restore SB_RDONLY status */
3040 * Maximal extent format file size.
3041 * Resulting logical blkno at s_maxbytes must fit in our on-disk
3042 * extent format containers, within a sector_t, and within i_blocks
3043 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
3044 * so that won't be a limiting factor.
3046 * However there is other limiting factor. We do store extents in the form
3047 * of starting block and length, hence the resulting length of the extent
3048 * covering maximum file size must fit into on-disk format containers as
3049 * well. Given that length is always by 1 unit bigger than max unit (because
3050 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
3052 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
3054 static loff_t ext4_max_size(int blkbits, int has_huge_files)
3057 loff_t upper_limit = MAX_LFS_FILESIZE;
3059 BUILD_BUG_ON(sizeof(blkcnt_t) < sizeof(u64));
3061 if (!has_huge_files) {
3062 upper_limit = (1LL << 32) - 1;
3064 /* total blocks in file system block size */
3065 upper_limit >>= (blkbits - 9);
3066 upper_limit <<= blkbits;
3070 * 32-bit extent-start container, ee_block. We lower the maxbytes
3071 * by one fs block, so ee_len can cover the extent of maximum file
3074 res = (1LL << 32) - 1;
3077 /* Sanity check against vm- & vfs- imposed limits */
3078 if (res > upper_limit)
3085 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
3086 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
3087 * We need to be 1 filesystem block less than the 2^48 sector limit.
3089 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
3091 loff_t res = EXT4_NDIR_BLOCKS;
3094 /* This is calculated to be the largest file size for a dense, block
3095 * mapped file such that the file's total number of 512-byte sectors,
3096 * including data and all indirect blocks, does not exceed (2^48 - 1).
3098 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
3099 * number of 512-byte sectors of the file.
3102 if (!has_huge_files) {
3104 * !has_huge_files or implies that the inode i_block field
3105 * represents total file blocks in 2^32 512-byte sectors ==
3106 * size of vfs inode i_blocks * 8
3108 upper_limit = (1LL << 32) - 1;
3110 /* total blocks in file system block size */
3111 upper_limit >>= (bits - 9);
3115 * We use 48 bit ext4_inode i_blocks
3116 * With EXT4_HUGE_FILE_FL set the i_blocks
3117 * represent total number of blocks in
3118 * file system block size
3120 upper_limit = (1LL << 48) - 1;
3124 /* indirect blocks */
3126 /* double indirect blocks */
3127 meta_blocks += 1 + (1LL << (bits-2));
3128 /* tripple indirect blocks */
3129 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
3131 upper_limit -= meta_blocks;
3132 upper_limit <<= bits;
3134 res += 1LL << (bits-2);
3135 res += 1LL << (2*(bits-2));
3136 res += 1LL << (3*(bits-2));
3138 if (res > upper_limit)
3141 if (res > MAX_LFS_FILESIZE)
3142 res = MAX_LFS_FILESIZE;
3147 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
3148 ext4_fsblk_t logical_sb_block, int nr)
3150 struct ext4_sb_info *sbi = EXT4_SB(sb);
3151 ext4_group_t bg, first_meta_bg;
3154 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
3156 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
3157 return logical_sb_block + nr + 1;
3158 bg = sbi->s_desc_per_block * nr;
3159 if (ext4_bg_has_super(sb, bg))
3163 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
3164 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
3165 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
3168 if (sb->s_blocksize == 1024 && nr == 0 &&
3169 le32_to_cpu(sbi->s_es->s_first_data_block) == 0)
3172 return (has_super + ext4_group_first_block_no(sb, bg));
3176 * ext4_get_stripe_size: Get the stripe size.
3177 * @sbi: In memory super block info
3179 * If we have specified it via mount option, then
3180 * use the mount option value. If the value specified at mount time is
3181 * greater than the blocks per group use the super block value.
3182 * If the super block value is greater than blocks per group return 0.
3183 * Allocator needs it be less than blocks per group.
3186 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
3188 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
3189 unsigned long stripe_width =
3190 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3193 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3194 ret = sbi->s_stripe;
3195 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
3197 else if (stride && stride <= sbi->s_blocks_per_group)
3203 * If the stripe width is 1, this makes no sense and
3204 * we set it to 0 to turn off stripe handling code.
3213 * Check whether this filesystem can be mounted based on
3214 * the features present and the RDONLY/RDWR mount requested.
3215 * Returns 1 if this filesystem can be mounted as requested,
3216 * 0 if it cannot be.
3218 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
3220 if (ext4_has_unknown_ext4_incompat_features(sb)) {
3221 ext4_msg(sb, KERN_ERR,
3222 "Couldn't mount because of "
3223 "unsupported optional features (%x)",
3224 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
3225 ~EXT4_FEATURE_INCOMPAT_SUPP));
3229 #ifndef CONFIG_UNICODE
3230 if (ext4_has_feature_casefold(sb)) {
3231 ext4_msg(sb, KERN_ERR,
3232 "Filesystem with casefold feature cannot be "
3233 "mounted without CONFIG_UNICODE");
3241 if (ext4_has_feature_readonly(sb)) {
3242 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
3243 sb->s_flags |= SB_RDONLY;
3247 /* Check that feature set is OK for a read-write mount */
3248 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
3249 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
3250 "unsupported optional features (%x)",
3251 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
3252 ~EXT4_FEATURE_RO_COMPAT_SUPP));
3255 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
3256 ext4_msg(sb, KERN_ERR,
3257 "Can't support bigalloc feature without "
3258 "extents feature\n");
3262 #if !IS_ENABLED(CONFIG_QUOTA) || !IS_ENABLED(CONFIG_QFMT_V2)
3263 if (!readonly && (ext4_has_feature_quota(sb) ||
3264 ext4_has_feature_project(sb))) {
3265 ext4_msg(sb, KERN_ERR,
3266 "The kernel was not built with CONFIG_QUOTA and CONFIG_QFMT_V2");
3269 #endif /* CONFIG_QUOTA */
3274 * This function is called once a day if we have errors logged
3275 * on the file system
3277 static void print_daily_error_info(struct timer_list *t)
3279 struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report);
3280 struct super_block *sb = sbi->s_sb;
3281 struct ext4_super_block *es = sbi->s_es;
3283 if (es->s_error_count)
3284 /* fsck newer than v1.41.13 is needed to clean this condition. */
3285 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
3286 le32_to_cpu(es->s_error_count));
3287 if (es->s_first_error_time) {
3288 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %llu: %.*s:%d",
3290 ext4_get_tstamp(es, s_first_error_time),
3291 (int) sizeof(es->s_first_error_func),
3292 es->s_first_error_func,
3293 le32_to_cpu(es->s_first_error_line));
3294 if (es->s_first_error_ino)
3295 printk(KERN_CONT ": inode %u",
3296 le32_to_cpu(es->s_first_error_ino));
3297 if (es->s_first_error_block)
3298 printk(KERN_CONT ": block %llu", (unsigned long long)
3299 le64_to_cpu(es->s_first_error_block));
3300 printk(KERN_CONT "\n");
3302 if (es->s_last_error_time) {
3303 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %llu: %.*s:%d",
3305 ext4_get_tstamp(es, s_last_error_time),
3306 (int) sizeof(es->s_last_error_func),
3307 es->s_last_error_func,
3308 le32_to_cpu(es->s_last_error_line));
3309 if (es->s_last_error_ino)
3310 printk(KERN_CONT ": inode %u",
3311 le32_to_cpu(es->s_last_error_ino));
3312 if (es->s_last_error_block)
3313 printk(KERN_CONT ": block %llu", (unsigned long long)
3314 le64_to_cpu(es->s_last_error_block));
3315 printk(KERN_CONT "\n");
3317 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
3320 /* Find next suitable group and run ext4_init_inode_table */
3321 static int ext4_run_li_request(struct ext4_li_request *elr)
3323 struct ext4_group_desc *gdp = NULL;
3324 struct super_block *sb = elr->lr_super;
3325 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3326 ext4_group_t group = elr->lr_next_group;
3327 unsigned long timeout = 0;
3328 unsigned int prefetch_ios = 0;
3331 if (elr->lr_mode == EXT4_LI_MODE_PREFETCH_BBITMAP) {
3332 elr->lr_next_group = ext4_mb_prefetch(sb, group,
3333 EXT4_SB(sb)->s_mb_prefetch, &prefetch_ios);
3335 ext4_mb_prefetch_fini(sb, elr->lr_next_group,
3337 trace_ext4_prefetch_bitmaps(sb, group, elr->lr_next_group,
3339 if (group >= elr->lr_next_group) {
3341 if (elr->lr_first_not_zeroed != ngroups &&
3342 !sb_rdonly(sb) && test_opt(sb, INIT_INODE_TABLE)) {
3343 elr->lr_next_group = elr->lr_first_not_zeroed;
3344 elr->lr_mode = EXT4_LI_MODE_ITABLE;
3351 for (; group < ngroups; group++) {
3352 gdp = ext4_get_group_desc(sb, group, NULL);
3358 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3362 if (group >= ngroups)
3367 ret = ext4_init_inode_table(sb, group,
3368 elr->lr_timeout ? 0 : 1);
3369 trace_ext4_lazy_itable_init(sb, group);
3370 if (elr->lr_timeout == 0) {
3371 timeout = (jiffies - timeout) *
3372 EXT4_SB(elr->lr_super)->s_li_wait_mult;
3373 elr->lr_timeout = timeout;
3375 elr->lr_next_sched = jiffies + elr->lr_timeout;
3376 elr->lr_next_group = group + 1;
3382 * Remove lr_request from the list_request and free the
3383 * request structure. Should be called with li_list_mtx held
3385 static void ext4_remove_li_request(struct ext4_li_request *elr)
3390 list_del(&elr->lr_request);
3391 EXT4_SB(elr->lr_super)->s_li_request = NULL;
3395 static void ext4_unregister_li_request(struct super_block *sb)
3397 mutex_lock(&ext4_li_mtx);
3398 if (!ext4_li_info) {
3399 mutex_unlock(&ext4_li_mtx);
3403 mutex_lock(&ext4_li_info->li_list_mtx);
3404 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
3405 mutex_unlock(&ext4_li_info->li_list_mtx);
3406 mutex_unlock(&ext4_li_mtx);
3409 static struct task_struct *ext4_lazyinit_task;
3412 * This is the function where ext4lazyinit thread lives. It walks
3413 * through the request list searching for next scheduled filesystem.
3414 * When such a fs is found, run the lazy initialization request
3415 * (ext4_rn_li_request) and keep track of the time spend in this
3416 * function. Based on that time we compute next schedule time of
3417 * the request. When walking through the list is complete, compute
3418 * next waking time and put itself into sleep.
3420 static int ext4_lazyinit_thread(void *arg)
3422 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
3423 struct list_head *pos, *n;
3424 struct ext4_li_request *elr;
3425 unsigned long next_wakeup, cur;
3427 BUG_ON(NULL == eli);
3431 next_wakeup = MAX_JIFFY_OFFSET;
3433 mutex_lock(&eli->li_list_mtx);
3434 if (list_empty(&eli->li_request_list)) {
3435 mutex_unlock(&eli->li_list_mtx);
3438 list_for_each_safe(pos, n, &eli->li_request_list) {
3441 elr = list_entry(pos, struct ext4_li_request,
3444 if (time_before(jiffies, elr->lr_next_sched)) {
3445 if (time_before(elr->lr_next_sched, next_wakeup))
3446 next_wakeup = elr->lr_next_sched;
3449 if (down_read_trylock(&elr->lr_super->s_umount)) {
3450 if (sb_start_write_trylock(elr->lr_super)) {
3453 * We hold sb->s_umount, sb can not
3454 * be removed from the list, it is
3455 * now safe to drop li_list_mtx
3457 mutex_unlock(&eli->li_list_mtx);
3458 err = ext4_run_li_request(elr);
3459 sb_end_write(elr->lr_super);
3460 mutex_lock(&eli->li_list_mtx);
3463 up_read((&elr->lr_super->s_umount));
3465 /* error, remove the lazy_init job */
3467 ext4_remove_li_request(elr);
3471 elr->lr_next_sched = jiffies +
3473 % (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3475 if (time_before(elr->lr_next_sched, next_wakeup))
3476 next_wakeup = elr->lr_next_sched;
3478 mutex_unlock(&eli->li_list_mtx);
3483 if ((time_after_eq(cur, next_wakeup)) ||
3484 (MAX_JIFFY_OFFSET == next_wakeup)) {
3489 schedule_timeout_interruptible(next_wakeup - cur);
3491 if (kthread_should_stop()) {
3492 ext4_clear_request_list();
3499 * It looks like the request list is empty, but we need
3500 * to check it under the li_list_mtx lock, to prevent any
3501 * additions into it, and of course we should lock ext4_li_mtx
3502 * to atomically free the list and ext4_li_info, because at
3503 * this point another ext4 filesystem could be registering
3506 mutex_lock(&ext4_li_mtx);
3507 mutex_lock(&eli->li_list_mtx);
3508 if (!list_empty(&eli->li_request_list)) {
3509 mutex_unlock(&eli->li_list_mtx);
3510 mutex_unlock(&ext4_li_mtx);
3513 mutex_unlock(&eli->li_list_mtx);
3514 kfree(ext4_li_info);
3515 ext4_li_info = NULL;
3516 mutex_unlock(&ext4_li_mtx);
3521 static void ext4_clear_request_list(void)
3523 struct list_head *pos, *n;
3524 struct ext4_li_request *elr;
3526 mutex_lock(&ext4_li_info->li_list_mtx);
3527 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3528 elr = list_entry(pos, struct ext4_li_request,
3530 ext4_remove_li_request(elr);
3532 mutex_unlock(&ext4_li_info->li_list_mtx);
3535 static int ext4_run_lazyinit_thread(void)
3537 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3538 ext4_li_info, "ext4lazyinit");
3539 if (IS_ERR(ext4_lazyinit_task)) {
3540 int err = PTR_ERR(ext4_lazyinit_task);
3541 ext4_clear_request_list();
3542 kfree(ext4_li_info);
3543 ext4_li_info = NULL;
3544 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3545 "initialization thread\n",
3549 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3554 * Check whether it make sense to run itable init. thread or not.
3555 * If there is at least one uninitialized inode table, return
3556 * corresponding group number, else the loop goes through all
3557 * groups and return total number of groups.
3559 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3561 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3562 struct ext4_group_desc *gdp = NULL;
3564 if (!ext4_has_group_desc_csum(sb))
3567 for (group = 0; group < ngroups; group++) {
3568 gdp = ext4_get_group_desc(sb, group, NULL);
3572 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3579 static int ext4_li_info_new(void)
3581 struct ext4_lazy_init *eli = NULL;
3583 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3587 INIT_LIST_HEAD(&eli->li_request_list);
3588 mutex_init(&eli->li_list_mtx);
3590 eli->li_state |= EXT4_LAZYINIT_QUIT;
3597 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3600 struct ext4_li_request *elr;
3602 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3607 elr->lr_first_not_zeroed = start;
3608 if (test_opt(sb, PREFETCH_BLOCK_BITMAPS))
3609 elr->lr_mode = EXT4_LI_MODE_PREFETCH_BBITMAP;
3611 elr->lr_mode = EXT4_LI_MODE_ITABLE;
3612 elr->lr_next_group = start;
3616 * Randomize first schedule time of the request to
3617 * spread the inode table initialization requests
3620 elr->lr_next_sched = jiffies + (prandom_u32() %
3621 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3625 int ext4_register_li_request(struct super_block *sb,
3626 ext4_group_t first_not_zeroed)
3628 struct ext4_sb_info *sbi = EXT4_SB(sb);
3629 struct ext4_li_request *elr = NULL;
3630 ext4_group_t ngroups = sbi->s_groups_count;
3633 mutex_lock(&ext4_li_mtx);
3634 if (sbi->s_li_request != NULL) {
3636 * Reset timeout so it can be computed again, because
3637 * s_li_wait_mult might have changed.
3639 sbi->s_li_request->lr_timeout = 0;
3643 if (!test_opt(sb, PREFETCH_BLOCK_BITMAPS) &&
3644 (first_not_zeroed == ngroups || sb_rdonly(sb) ||
3645 !test_opt(sb, INIT_INODE_TABLE)))
3648 elr = ext4_li_request_new(sb, first_not_zeroed);
3654 if (NULL == ext4_li_info) {
3655 ret = ext4_li_info_new();
3660 mutex_lock(&ext4_li_info->li_list_mtx);
3661 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3662 mutex_unlock(&ext4_li_info->li_list_mtx);
3664 sbi->s_li_request = elr;
3666 * set elr to NULL here since it has been inserted to
3667 * the request_list and the removal and free of it is
3668 * handled by ext4_clear_request_list from now on.
3672 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3673 ret = ext4_run_lazyinit_thread();
3678 mutex_unlock(&ext4_li_mtx);
3685 * We do not need to lock anything since this is called on
3688 static void ext4_destroy_lazyinit_thread(void)
3691 * If thread exited earlier
3692 * there's nothing to be done.
3694 if (!ext4_li_info || !ext4_lazyinit_task)
3697 kthread_stop(ext4_lazyinit_task);
3700 static int set_journal_csum_feature_set(struct super_block *sb)
3703 int compat, incompat;
3704 struct ext4_sb_info *sbi = EXT4_SB(sb);
3706 if (ext4_has_metadata_csum(sb)) {
3707 /* journal checksum v3 */
3709 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
3711 /* journal checksum v1 */
3712 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3716 jbd2_journal_clear_features(sbi->s_journal,
3717 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3718 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3719 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3720 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3721 ret = jbd2_journal_set_features(sbi->s_journal,
3723 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3725 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3726 ret = jbd2_journal_set_features(sbi->s_journal,
3729 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3730 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3732 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3733 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3740 * Note: calculating the overhead so we can be compatible with
3741 * historical BSD practice is quite difficult in the face of
3742 * clusters/bigalloc. This is because multiple metadata blocks from
3743 * different block group can end up in the same allocation cluster.
3744 * Calculating the exact overhead in the face of clustered allocation
3745 * requires either O(all block bitmaps) in memory or O(number of block
3746 * groups**2) in time. We will still calculate the superblock for
3747 * older file systems --- and if we come across with a bigalloc file
3748 * system with zero in s_overhead_clusters the estimate will be close to
3749 * correct especially for very large cluster sizes --- but for newer
3750 * file systems, it's better to calculate this figure once at mkfs
3751 * time, and store it in the superblock. If the superblock value is
3752 * present (even for non-bigalloc file systems), we will use it.
3754 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3757 struct ext4_sb_info *sbi = EXT4_SB(sb);
3758 struct ext4_group_desc *gdp;
3759 ext4_fsblk_t first_block, last_block, b;
3760 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3761 int s, j, count = 0;
3763 if (!ext4_has_feature_bigalloc(sb))
3764 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3765 sbi->s_itb_per_group + 2);
3767 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3768 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3769 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3770 for (i = 0; i < ngroups; i++) {
3771 gdp = ext4_get_group_desc(sb, i, NULL);
3772 b = ext4_block_bitmap(sb, gdp);
3773 if (b >= first_block && b <= last_block) {
3774 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3777 b = ext4_inode_bitmap(sb, gdp);
3778 if (b >= first_block && b <= last_block) {
3779 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3782 b = ext4_inode_table(sb, gdp);
3783 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3784 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3785 int c = EXT4_B2C(sbi, b - first_block);
3786 ext4_set_bit(c, buf);
3792 if (ext4_bg_has_super(sb, grp)) {
3793 ext4_set_bit(s++, buf);
3796 j = ext4_bg_num_gdb(sb, grp);
3797 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
3798 ext4_error(sb, "Invalid number of block group "
3799 "descriptor blocks: %d", j);
3800 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
3804 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3808 return EXT4_CLUSTERS_PER_GROUP(sb) -
3809 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3813 * Compute the overhead and stash it in sbi->s_overhead
3815 int ext4_calculate_overhead(struct super_block *sb)
3817 struct ext4_sb_info *sbi = EXT4_SB(sb);
3818 struct ext4_super_block *es = sbi->s_es;
3819 struct inode *j_inode;
3820 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
3821 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3822 ext4_fsblk_t overhead = 0;
3823 char *buf = (char *) get_zeroed_page(GFP_NOFS);
3829 * Compute the overhead (FS structures). This is constant
3830 * for a given filesystem unless the number of block groups
3831 * changes so we cache the previous value until it does.
3835 * All of the blocks before first_data_block are overhead
3837 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3840 * Add the overhead found in each block group
3842 for (i = 0; i < ngroups; i++) {
3845 blks = count_overhead(sb, i, buf);
3848 memset(buf, 0, PAGE_SIZE);
3853 * Add the internal journal blocks whether the journal has been
3856 if (sbi->s_journal && !sbi->s_journal_bdev)
3857 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3858 else if (ext4_has_feature_journal(sb) && !sbi->s_journal && j_inum) {
3859 /* j_inum for internal journal is non-zero */
3860 j_inode = ext4_get_journal_inode(sb, j_inum);
3862 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
3863 overhead += EXT4_NUM_B2C(sbi, j_blocks);
3866 ext4_msg(sb, KERN_ERR, "can't get journal size");
3869 sbi->s_overhead = overhead;
3871 free_page((unsigned long) buf);
3875 static void ext4_set_resv_clusters(struct super_block *sb)
3877 ext4_fsblk_t resv_clusters;
3878 struct ext4_sb_info *sbi = EXT4_SB(sb);
3881 * There's no need to reserve anything when we aren't using extents.
3882 * The space estimates are exact, there are no unwritten extents,
3883 * hole punching doesn't need new metadata... This is needed especially
3884 * to keep ext2/3 backward compatibility.
3886 if (!ext4_has_feature_extents(sb))
3889 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3890 * This should cover the situations where we can not afford to run
3891 * out of space like for example punch hole, or converting
3892 * unwritten extents in delalloc path. In most cases such
3893 * allocation would require 1, or 2 blocks, higher numbers are
3896 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3897 sbi->s_cluster_bits);
3899 do_div(resv_clusters, 50);
3900 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3902 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
3905 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3907 struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev);
3908 char *orig_data = kstrdup(data, GFP_KERNEL);
3909 struct buffer_head *bh, **group_desc;
3910 struct ext4_super_block *es = NULL;
3911 struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3912 struct flex_groups **flex_groups;
3914 ext4_fsblk_t sb_block = get_sb_block(&data);
3915 ext4_fsblk_t logical_sb_block;
3916 unsigned long offset = 0;
3917 unsigned long journal_devnum = 0;
3918 unsigned long def_mount_opts;
3922 int blocksize, clustersize;
3923 unsigned int db_count;
3925 int needs_recovery, has_huge_files;
3928 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3929 ext4_group_t first_not_zeroed;
3931 if ((data && !orig_data) || !sbi)
3934 sbi->s_daxdev = dax_dev;
3935 sbi->s_blockgroup_lock =
3936 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3937 if (!sbi->s_blockgroup_lock)
3940 sb->s_fs_info = sbi;
3942 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3943 sbi->s_sb_block = sb_block;
3944 if (sb->s_bdev->bd_part)
3945 sbi->s_sectors_written_start =
3946 part_stat_read(sb->s_bdev->bd_part, sectors[STAT_WRITE]);
3948 /* Cleanup superblock name */
3949 strreplace(sb->s_id, '/', '!');
3951 /* -EINVAL is default */
3953 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3955 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3960 * The ext4 superblock will not be buffer aligned for other than 1kB
3961 * block sizes. We need to calculate the offset from buffer start.
3963 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3964 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3965 offset = do_div(logical_sb_block, blocksize);
3967 logical_sb_block = sb_block;
3970 bh = ext4_sb_bread_unmovable(sb, logical_sb_block);
3972 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3978 * Note: s_es must be initialized as soon as possible because
3979 * some ext4 macro-instructions depend on its value
3981 es = (struct ext4_super_block *) (bh->b_data + offset);
3983 sb->s_magic = le16_to_cpu(es->s_magic);
3984 if (sb->s_magic != EXT4_SUPER_MAGIC)
3986 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3988 /* Warn if metadata_csum and gdt_csum are both set. */
3989 if (ext4_has_feature_metadata_csum(sb) &&
3990 ext4_has_feature_gdt_csum(sb))
3991 ext4_warning(sb, "metadata_csum and uninit_bg are "
3992 "redundant flags; please run fsck.");
3994 /* Check for a known checksum algorithm */
3995 if (!ext4_verify_csum_type(sb, es)) {
3996 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3997 "unknown checksum algorithm.");
4002 /* Load the checksum driver */
4003 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
4004 if (IS_ERR(sbi->s_chksum_driver)) {
4005 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
4006 ret = PTR_ERR(sbi->s_chksum_driver);
4007 sbi->s_chksum_driver = NULL;
4011 /* Check superblock checksum */
4012 if (!ext4_superblock_csum_verify(sb, es)) {
4013 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
4014 "invalid superblock checksum. Run e2fsck?");
4020 /* Precompute checksum seed for all metadata */
4021 if (ext4_has_feature_csum_seed(sb))
4022 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
4023 else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb))
4024 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
4025 sizeof(es->s_uuid));
4027 /* Set defaults before we parse the mount options */
4028 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
4029 set_opt(sb, INIT_INODE_TABLE);
4030 if (def_mount_opts & EXT4_DEFM_DEBUG)
4032 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
4034 if (def_mount_opts & EXT4_DEFM_UID16)
4035 set_opt(sb, NO_UID32);
4036 /* xattr user namespace & acls are now defaulted on */
4037 set_opt(sb, XATTR_USER);
4038 #ifdef CONFIG_EXT4_FS_POSIX_ACL
4039 set_opt(sb, POSIX_ACL);
4041 /* don't forget to enable journal_csum when metadata_csum is enabled. */
4042 if (ext4_has_metadata_csum(sb))
4043 set_opt(sb, JOURNAL_CHECKSUM);
4045 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
4046 set_opt(sb, JOURNAL_DATA);
4047 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
4048 set_opt(sb, ORDERED_DATA);
4049 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
4050 set_opt(sb, WRITEBACK_DATA);
4052 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
4053 set_opt(sb, ERRORS_PANIC);
4054 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
4055 set_opt(sb, ERRORS_CONT);
4057 set_opt(sb, ERRORS_RO);
4058 /* block_validity enabled by default; disable with noblock_validity */
4059 set_opt(sb, BLOCK_VALIDITY);
4060 if (def_mount_opts & EXT4_DEFM_DISCARD)
4061 set_opt(sb, DISCARD);
4063 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
4064 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
4065 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
4066 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
4067 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
4069 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
4070 set_opt(sb, BARRIER);
4073 * enable delayed allocation by default
4074 * Use -o nodelalloc to turn it off
4076 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
4077 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
4078 set_opt(sb, DELALLOC);
4081 * set default s_li_wait_mult for lazyinit, for the case there is
4082 * no mount option specified.
4084 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
4086 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
4088 if (blocksize == PAGE_SIZE)
4089 set_opt(sb, DIOREAD_NOLOCK);
4091 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
4092 blocksize > EXT4_MAX_BLOCK_SIZE) {
4093 ext4_msg(sb, KERN_ERR,
4094 "Unsupported filesystem blocksize %d (%d log_block_size)",
4095 blocksize, le32_to_cpu(es->s_log_block_size));
4099 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
4100 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
4101 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
4103 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
4104 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
4105 if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) {
4106 ext4_msg(sb, KERN_ERR, "invalid first ino: %u",
4110 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
4111 (!is_power_of_2(sbi->s_inode_size)) ||
4112 (sbi->s_inode_size > blocksize)) {
4113 ext4_msg(sb, KERN_ERR,
4114 "unsupported inode size: %d",
4116 ext4_msg(sb, KERN_ERR, "blocksize: %d", blocksize);
4120 * i_atime_extra is the last extra field available for
4121 * [acm]times in struct ext4_inode. Checking for that
4122 * field should suffice to ensure we have extra space
4125 if (sbi->s_inode_size >= offsetof(struct ext4_inode, i_atime_extra) +
4126 sizeof(((struct ext4_inode *)0)->i_atime_extra)) {
4127 sb->s_time_gran = 1;
4128 sb->s_time_max = EXT4_EXTRA_TIMESTAMP_MAX;
4130 sb->s_time_gran = NSEC_PER_SEC;
4131 sb->s_time_max = EXT4_NON_EXTRA_TIMESTAMP_MAX;
4133 sb->s_time_min = EXT4_TIMESTAMP_MIN;
4135 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4136 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4137 EXT4_GOOD_OLD_INODE_SIZE;
4138 if (ext4_has_feature_extra_isize(sb)) {
4139 unsigned v, max = (sbi->s_inode_size -
4140 EXT4_GOOD_OLD_INODE_SIZE);
4142 v = le16_to_cpu(es->s_want_extra_isize);
4144 ext4_msg(sb, KERN_ERR,
4145 "bad s_want_extra_isize: %d", v);
4148 if (sbi->s_want_extra_isize < v)
4149 sbi->s_want_extra_isize = v;
4151 v = le16_to_cpu(es->s_min_extra_isize);
4153 ext4_msg(sb, KERN_ERR,
4154 "bad s_min_extra_isize: %d", v);
4157 if (sbi->s_want_extra_isize < v)
4158 sbi->s_want_extra_isize = v;
4162 if (sbi->s_es->s_mount_opts[0]) {
4163 char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
4164 sizeof(sbi->s_es->s_mount_opts),
4168 if (!parse_options(s_mount_opts, sb, &journal_devnum,
4169 &journal_ioprio, 0)) {
4170 ext4_msg(sb, KERN_WARNING,
4171 "failed to parse options in superblock: %s",
4174 kfree(s_mount_opts);
4176 sbi->s_def_mount_opt = sbi->s_mount_opt;
4177 if (!parse_options((char *) data, sb, &journal_devnum,
4178 &journal_ioprio, 0))
4181 #ifdef CONFIG_UNICODE
4182 if (ext4_has_feature_casefold(sb) && !sbi->s_encoding) {
4183 const struct ext4_sb_encodings *encoding_info;
4184 struct unicode_map *encoding;
4185 __u16 encoding_flags;
4187 if (ext4_has_feature_encrypt(sb)) {
4188 ext4_msg(sb, KERN_ERR,
4189 "Can't mount with encoding and encryption");
4193 if (ext4_sb_read_encoding(es, &encoding_info,
4195 ext4_msg(sb, KERN_ERR,
4196 "Encoding requested by superblock is unknown");
4200 encoding = utf8_load(encoding_info->version);
4201 if (IS_ERR(encoding)) {
4202 ext4_msg(sb, KERN_ERR,
4203 "can't mount with superblock charset: %s-%s "
4204 "not supported by the kernel. flags: 0x%x.",
4205 encoding_info->name, encoding_info->version,
4209 ext4_msg(sb, KERN_INFO,"Using encoding defined by superblock: "
4210 "%s-%s with flags 0x%hx", encoding_info->name,
4211 encoding_info->version?:"\b", encoding_flags);
4213 sbi->s_encoding = encoding;
4214 sbi->s_encoding_flags = encoding_flags;
4218 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4219 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting with data=journal disables delayed allocation, dioread_nolock, and O_DIRECT support!\n");
4220 /* can't mount with both data=journal and dioread_nolock. */
4221 clear_opt(sb, DIOREAD_NOLOCK);
4222 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4223 ext4_msg(sb, KERN_ERR, "can't mount with "
4224 "both data=journal and delalloc");
4227 if (test_opt(sb, DAX_ALWAYS)) {
4228 ext4_msg(sb, KERN_ERR, "can't mount with "
4229 "both data=journal and dax");
4232 if (ext4_has_feature_encrypt(sb)) {
4233 ext4_msg(sb, KERN_WARNING,
4234 "encrypted files will use data=ordered "
4235 "instead of data journaling mode");
4237 if (test_opt(sb, DELALLOC))
4238 clear_opt(sb, DELALLOC);
4240 sb->s_iflags |= SB_I_CGROUPWB;
4243 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
4244 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
4246 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
4247 (ext4_has_compat_features(sb) ||
4248 ext4_has_ro_compat_features(sb) ||
4249 ext4_has_incompat_features(sb)))
4250 ext4_msg(sb, KERN_WARNING,
4251 "feature flags set on rev 0 fs, "
4252 "running e2fsck is recommended");
4254 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
4255 set_opt2(sb, HURD_COMPAT);
4256 if (ext4_has_feature_64bit(sb)) {
4257 ext4_msg(sb, KERN_ERR,
4258 "The Hurd can't support 64-bit file systems");
4263 * ea_inode feature uses l_i_version field which is not
4264 * available in HURD_COMPAT mode.
4266 if (ext4_has_feature_ea_inode(sb)) {
4267 ext4_msg(sb, KERN_ERR,
4268 "ea_inode feature is not supported for Hurd");
4273 if (IS_EXT2_SB(sb)) {
4274 if (ext2_feature_set_ok(sb))
4275 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
4276 "using the ext4 subsystem");
4279 * If we're probing be silent, if this looks like
4280 * it's actually an ext[34] filesystem.
4282 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4284 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
4285 "to feature incompatibilities");
4290 if (IS_EXT3_SB(sb)) {
4291 if (ext3_feature_set_ok(sb))
4292 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
4293 "using the ext4 subsystem");
4296 * If we're probing be silent, if this looks like
4297 * it's actually an ext4 filesystem.
4299 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4301 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
4302 "to feature incompatibilities");
4308 * Check feature flags regardless of the revision level, since we
4309 * previously didn't change the revision level when setting the flags,
4310 * so there is a chance incompat flags are set on a rev 0 filesystem.
4312 if (!ext4_feature_set_ok(sb, (sb_rdonly(sb))))
4315 if (le32_to_cpu(es->s_log_block_size) >
4316 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
4317 ext4_msg(sb, KERN_ERR,
4318 "Invalid log block size: %u",
4319 le32_to_cpu(es->s_log_block_size));
4322 if (le32_to_cpu(es->s_log_cluster_size) >
4323 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
4324 ext4_msg(sb, KERN_ERR,
4325 "Invalid log cluster size: %u",
4326 le32_to_cpu(es->s_log_cluster_size));
4330 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
4331 ext4_msg(sb, KERN_ERR,
4332 "Number of reserved GDT blocks insanely large: %d",
4333 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
4337 if (bdev_dax_supported(sb->s_bdev, blocksize))
4338 set_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags);
4340 if (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) {
4341 if (ext4_has_feature_inline_data(sb)) {
4342 ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem"
4343 " that may contain inline data");
4346 if (!test_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags)) {
4347 ext4_msg(sb, KERN_ERR,
4348 "DAX unsupported by block device.");
4353 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
4354 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
4355 es->s_encryption_level);
4359 if (sb->s_blocksize != blocksize) {
4360 /* Validate the filesystem blocksize */
4361 if (!sb_set_blocksize(sb, blocksize)) {
4362 ext4_msg(sb, KERN_ERR, "bad block size %d",
4368 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
4369 offset = do_div(logical_sb_block, blocksize);
4370 bh = ext4_sb_bread_unmovable(sb, logical_sb_block);
4372 ext4_msg(sb, KERN_ERR,
4373 "Can't read superblock on 2nd try");
4378 es = (struct ext4_super_block *)(bh->b_data + offset);
4380 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
4381 ext4_msg(sb, KERN_ERR,
4382 "Magic mismatch, very weird!");
4387 has_huge_files = ext4_has_feature_huge_file(sb);
4388 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
4390 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
4392 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
4393 if (ext4_has_feature_64bit(sb)) {
4394 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
4395 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
4396 !is_power_of_2(sbi->s_desc_size)) {
4397 ext4_msg(sb, KERN_ERR,
4398 "unsupported descriptor size %lu",
4403 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
4405 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
4406 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
4408 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
4409 if (sbi->s_inodes_per_block == 0)
4411 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
4412 sbi->s_inodes_per_group > blocksize * 8) {
4413 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
4414 sbi->s_inodes_per_group);
4417 sbi->s_itb_per_group = sbi->s_inodes_per_group /
4418 sbi->s_inodes_per_block;
4419 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
4421 sbi->s_mount_state = le16_to_cpu(es->s_state);
4422 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
4423 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
4425 for (i = 0; i < 4; i++)
4426 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
4427 sbi->s_def_hash_version = es->s_def_hash_version;
4428 if (ext4_has_feature_dir_index(sb)) {
4429 i = le32_to_cpu(es->s_flags);
4430 if (i & EXT2_FLAGS_UNSIGNED_HASH)
4431 sbi->s_hash_unsigned = 3;
4432 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
4433 #ifdef __CHAR_UNSIGNED__
4436 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
4437 sbi->s_hash_unsigned = 3;
4441 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
4446 /* Handle clustersize */
4447 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
4448 if (ext4_has_feature_bigalloc(sb)) {
4449 if (clustersize < blocksize) {
4450 ext4_msg(sb, KERN_ERR,
4451 "cluster size (%d) smaller than "
4452 "block size (%d)", clustersize, blocksize);
4455 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
4456 le32_to_cpu(es->s_log_block_size);
4457 sbi->s_clusters_per_group =
4458 le32_to_cpu(es->s_clusters_per_group);
4459 if (sbi->s_clusters_per_group > blocksize * 8) {
4460 ext4_msg(sb, KERN_ERR,
4461 "#clusters per group too big: %lu",
4462 sbi->s_clusters_per_group);
4465 if (sbi->s_blocks_per_group !=
4466 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
4467 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
4468 "clusters per group (%lu) inconsistent",
4469 sbi->s_blocks_per_group,
4470 sbi->s_clusters_per_group);
4474 if (clustersize != blocksize) {
4475 ext4_msg(sb, KERN_ERR,
4476 "fragment/cluster size (%d) != "
4477 "block size (%d)", clustersize, blocksize);
4480 if (sbi->s_blocks_per_group > blocksize * 8) {
4481 ext4_msg(sb, KERN_ERR,
4482 "#blocks per group too big: %lu",
4483 sbi->s_blocks_per_group);
4486 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
4487 sbi->s_cluster_bits = 0;
4489 sbi->s_cluster_ratio = clustersize / blocksize;
4491 /* Do we have standard group size of clustersize * 8 blocks ? */
4492 if (sbi->s_blocks_per_group == clustersize << 3)
4493 set_opt2(sb, STD_GROUP_SIZE);
4496 * Test whether we have more sectors than will fit in sector_t,
4497 * and whether the max offset is addressable by the page cache.
4499 err = generic_check_addressable(sb->s_blocksize_bits,
4500 ext4_blocks_count(es));
4502 ext4_msg(sb, KERN_ERR, "filesystem"
4503 " too large to mount safely on this system");
4507 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
4510 /* check blocks count against device size */
4511 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
4512 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
4513 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
4514 "exceeds size of device (%llu blocks)",
4515 ext4_blocks_count(es), blocks_count);
4520 * It makes no sense for the first data block to be beyond the end
4521 * of the filesystem.
4523 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
4524 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4525 "block %u is beyond end of filesystem (%llu)",
4526 le32_to_cpu(es->s_first_data_block),
4527 ext4_blocks_count(es));
4530 if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) &&
4531 (sbi->s_cluster_ratio == 1)) {
4532 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4533 "block is 0 with a 1k block and cluster size");
4537 blocks_count = (ext4_blocks_count(es) -
4538 le32_to_cpu(es->s_first_data_block) +
4539 EXT4_BLOCKS_PER_GROUP(sb) - 1);
4540 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4541 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
4542 ext4_msg(sb, KERN_WARNING, "groups count too large: %llu "
4543 "(block count %llu, first data block %u, "
4544 "blocks per group %lu)", blocks_count,
4545 ext4_blocks_count(es),
4546 le32_to_cpu(es->s_first_data_block),
4547 EXT4_BLOCKS_PER_GROUP(sb));
4550 sbi->s_groups_count = blocks_count;
4551 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
4552 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
4553 if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) !=
4554 le32_to_cpu(es->s_inodes_count)) {
4555 ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu",
4556 le32_to_cpu(es->s_inodes_count),
4557 ((u64)sbi->s_groups_count * sbi->s_inodes_per_group));
4561 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
4562 EXT4_DESC_PER_BLOCK(sb);
4563 if (ext4_has_feature_meta_bg(sb)) {
4564 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
4565 ext4_msg(sb, KERN_WARNING,
4566 "first meta block group too large: %u "
4567 "(group descriptor block count %u)",
4568 le32_to_cpu(es->s_first_meta_bg), db_count);
4572 rcu_assign_pointer(sbi->s_group_desc,
4573 kvmalloc_array(db_count,
4574 sizeof(struct buffer_head *),
4576 if (sbi->s_group_desc == NULL) {
4577 ext4_msg(sb, KERN_ERR, "not enough memory");
4582 bgl_lock_init(sbi->s_blockgroup_lock);
4584 /* Pre-read the descriptors into the buffer cache */
4585 for (i = 0; i < db_count; i++) {
4586 block = descriptor_loc(sb, logical_sb_block, i);
4587 ext4_sb_breadahead_unmovable(sb, block);
4590 for (i = 0; i < db_count; i++) {
4591 struct buffer_head *bh;
4593 block = descriptor_loc(sb, logical_sb_block, i);
4594 bh = ext4_sb_bread_unmovable(sb, block);
4596 ext4_msg(sb, KERN_ERR,
4597 "can't read group descriptor %d", i);
4604 rcu_dereference(sbi->s_group_desc)[i] = bh;
4607 sbi->s_gdb_count = db_count;
4608 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
4609 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
4610 ret = -EFSCORRUPTED;
4614 timer_setup(&sbi->s_err_report, print_daily_error_info, 0);
4616 /* Register extent status tree shrinker */
4617 if (ext4_es_register_shrinker(sbi))
4620 sbi->s_stripe = ext4_get_stripe_size(sbi);
4621 sbi->s_extent_max_zeroout_kb = 32;
4624 * set up enough so that it can read an inode
4626 sb->s_op = &ext4_sops;
4627 sb->s_export_op = &ext4_export_ops;
4628 sb->s_xattr = ext4_xattr_handlers;
4629 #ifdef CONFIG_FS_ENCRYPTION
4630 sb->s_cop = &ext4_cryptops;
4632 #ifdef CONFIG_FS_VERITY
4633 sb->s_vop = &ext4_verityops;
4636 sb->dq_op = &ext4_quota_operations;
4637 if (ext4_has_feature_quota(sb))
4638 sb->s_qcop = &dquot_quotactl_sysfile_ops;
4640 sb->s_qcop = &ext4_qctl_operations;
4641 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
4643 memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
4645 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
4646 mutex_init(&sbi->s_orphan_lock);
4650 needs_recovery = (es->s_last_orphan != 0 ||
4651 ext4_has_feature_journal_needs_recovery(sb));
4653 if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb))
4654 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
4655 goto failed_mount3a;
4658 * The first inode we look at is the journal inode. Don't try
4659 * root first: it may be modified in the journal!
4661 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
4662 err = ext4_load_journal(sb, es, journal_devnum);
4664 goto failed_mount3a;
4665 } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) &&
4666 ext4_has_feature_journal_needs_recovery(sb)) {
4667 ext4_msg(sb, KERN_ERR, "required journal recovery "
4668 "suppressed and not mounted read-only");
4669 goto failed_mount_wq;
4671 /* Nojournal mode, all journal mount options are illegal */
4672 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
4673 ext4_msg(sb, KERN_ERR, "can't mount with "
4674 "journal_checksum, fs mounted w/o journal");
4675 goto failed_mount_wq;
4677 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4678 ext4_msg(sb, KERN_ERR, "can't mount with "
4679 "journal_async_commit, fs mounted w/o journal");
4680 goto failed_mount_wq;
4682 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
4683 ext4_msg(sb, KERN_ERR, "can't mount with "
4684 "commit=%lu, fs mounted w/o journal",
4685 sbi->s_commit_interval / HZ);
4686 goto failed_mount_wq;
4688 if (EXT4_MOUNT_DATA_FLAGS &
4689 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
4690 ext4_msg(sb, KERN_ERR, "can't mount with "
4691 "data=, fs mounted w/o journal");
4692 goto failed_mount_wq;
4694 sbi->s_def_mount_opt &= ~EXT4_MOUNT_JOURNAL_CHECKSUM;
4695 clear_opt(sb, JOURNAL_CHECKSUM);
4696 clear_opt(sb, DATA_FLAGS);
4697 sbi->s_journal = NULL;
4702 if (ext4_has_feature_64bit(sb) &&
4703 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4704 JBD2_FEATURE_INCOMPAT_64BIT)) {
4705 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
4706 goto failed_mount_wq;
4709 if (!set_journal_csum_feature_set(sb)) {
4710 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4712 goto failed_mount_wq;
4715 /* We have now updated the journal if required, so we can
4716 * validate the data journaling mode. */
4717 switch (test_opt(sb, DATA_FLAGS)) {
4719 /* No mode set, assume a default based on the journal
4720 * capabilities: ORDERED_DATA if the journal can
4721 * cope, else JOURNAL_DATA
4723 if (jbd2_journal_check_available_features
4724 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4725 set_opt(sb, ORDERED_DATA);
4726 sbi->s_def_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
4728 set_opt(sb, JOURNAL_DATA);
4729 sbi->s_def_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
4733 case EXT4_MOUNT_ORDERED_DATA:
4734 case EXT4_MOUNT_WRITEBACK_DATA:
4735 if (!jbd2_journal_check_available_features
4736 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4737 ext4_msg(sb, KERN_ERR, "Journal does not support "
4738 "requested data journaling mode");
4739 goto failed_mount_wq;
4745 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
4746 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4747 ext4_msg(sb, KERN_ERR, "can't mount with "
4748 "journal_async_commit in data=ordered mode");
4749 goto failed_mount_wq;
4752 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4754 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4755 sbi->s_journal->j_submit_inode_data_buffers =
4756 jbd2_journal_submit_inode_data_buffers;
4757 sbi->s_journal->j_finish_inode_data_buffers =
4758 jbd2_journal_finish_inode_data_buffers;
4761 if (!test_opt(sb, NO_MBCACHE)) {
4762 sbi->s_ea_block_cache = ext4_xattr_create_cache();
4763 if (!sbi->s_ea_block_cache) {
4764 ext4_msg(sb, KERN_ERR,
4765 "Failed to create ea_block_cache");
4766 goto failed_mount_wq;
4769 if (ext4_has_feature_ea_inode(sb)) {
4770 sbi->s_ea_inode_cache = ext4_xattr_create_cache();
4771 if (!sbi->s_ea_inode_cache) {
4772 ext4_msg(sb, KERN_ERR,
4773 "Failed to create ea_inode_cache");
4774 goto failed_mount_wq;
4779 if (ext4_has_feature_verity(sb) && blocksize != PAGE_SIZE) {
4780 ext4_msg(sb, KERN_ERR, "Unsupported blocksize for fs-verity");
4781 goto failed_mount_wq;
4784 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !sb_rdonly(sb) &&
4785 !ext4_has_feature_encrypt(sb)) {
4786 ext4_set_feature_encrypt(sb);
4787 ext4_commit_super(sb, 1);
4791 * Get the # of file system overhead blocks from the
4792 * superblock if present.
4794 if (es->s_overhead_clusters)
4795 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4797 err = ext4_calculate_overhead(sb);
4799 goto failed_mount_wq;
4803 * The maximum number of concurrent works can be high and
4804 * concurrency isn't really necessary. Limit it to 1.
4806 EXT4_SB(sb)->rsv_conversion_wq =
4807 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4808 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4809 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
4815 * The jbd2_journal_load will have done any necessary log recovery,
4816 * so we can safely mount the rest of the filesystem now.
4819 root = ext4_iget(sb, EXT4_ROOT_INO, EXT4_IGET_SPECIAL);
4821 ext4_msg(sb, KERN_ERR, "get root inode failed");
4822 ret = PTR_ERR(root);
4826 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4827 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4832 #ifdef CONFIG_UNICODE
4833 if (sbi->s_encoding)
4834 sb->s_d_op = &ext4_dentry_ops;
4837 sb->s_root = d_make_root(root);
4839 ext4_msg(sb, KERN_ERR, "get root dentry failed");
4844 ret = ext4_setup_super(sb, es, sb_rdonly(sb));
4845 if (ret == -EROFS) {
4846 sb->s_flags |= SB_RDONLY;
4849 goto failed_mount4a;
4851 ext4_set_resv_clusters(sb);
4853 if (test_opt(sb, BLOCK_VALIDITY)) {
4854 err = ext4_setup_system_zone(sb);
4856 ext4_msg(sb, KERN_ERR, "failed to initialize system "
4858 goto failed_mount4a;
4863 err = ext4_mb_init(sb);
4865 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4870 block = ext4_count_free_clusters(sb);
4871 ext4_free_blocks_count_set(sbi->s_es,
4872 EXT4_C2B(sbi, block));
4873 ext4_superblock_csum_set(sb);
4874 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4877 unsigned long freei = ext4_count_free_inodes(sb);
4878 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
4879 ext4_superblock_csum_set(sb);
4880 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4884 err = percpu_counter_init(&sbi->s_dirs_counter,
4885 ext4_count_dirs(sb), GFP_KERNEL);
4887 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4890 err = percpu_init_rwsem(&sbi->s_writepages_rwsem);
4893 ext4_msg(sb, KERN_ERR, "insufficient memory");
4897 if (ext4_has_feature_flex_bg(sb))
4898 if (!ext4_fill_flex_info(sb)) {
4899 ext4_msg(sb, KERN_ERR,
4900 "unable to initialize "
4901 "flex_bg meta info!");
4905 err = ext4_register_li_request(sb, first_not_zeroed);
4909 err = ext4_register_sysfs(sb);
4914 /* Enable quota usage during mount. */
4915 if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) {
4916 err = ext4_enable_quotas(sb);
4920 #endif /* CONFIG_QUOTA */
4923 * Save the original bdev mapping's wb_err value which could be
4924 * used to detect the metadata async write error.
4926 spin_lock_init(&sbi->s_bdev_wb_lock);
4927 errseq_check_and_advance(&sb->s_bdev->bd_inode->i_mapping->wb_err,
4928 &sbi->s_bdev_wb_err);
4929 sb->s_bdev->bd_super = sb;
4930 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4931 ext4_orphan_cleanup(sb, es);
4932 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4933 if (needs_recovery) {
4934 ext4_msg(sb, KERN_INFO, "recovery complete");
4935 err = ext4_mark_recovery_complete(sb, es);
4939 if (EXT4_SB(sb)->s_journal) {
4940 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4941 descr = " journalled data mode";
4942 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4943 descr = " ordered data mode";
4945 descr = " writeback data mode";
4947 descr = "out journal";
4949 if (test_opt(sb, DISCARD)) {
4950 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4951 if (!blk_queue_discard(q))
4952 ext4_msg(sb, KERN_WARNING,
4953 "mounting with \"discard\" option, but "
4954 "the device does not support discard");
4957 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
4958 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4959 "Opts: %.*s%s%s", descr,
4960 (int) sizeof(sbi->s_es->s_mount_opts),
4961 sbi->s_es->s_mount_opts,
4962 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4964 if (es->s_error_count)
4965 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4967 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4968 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4969 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4970 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4971 atomic_set(&sbi->s_warning_count, 0);
4972 atomic_set(&sbi->s_msg_count, 0);
4979 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4983 ext4_unregister_sysfs(sb);
4984 kobject_put(&sbi->s_kobj);
4986 ext4_unregister_li_request(sb);
4988 ext4_mb_release(sb);
4990 flex_groups = rcu_dereference(sbi->s_flex_groups);
4992 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
4993 kvfree(flex_groups[i]);
4994 kvfree(flex_groups);
4997 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4998 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4999 percpu_counter_destroy(&sbi->s_dirs_counter);
5000 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
5001 percpu_free_rwsem(&sbi->s_writepages_rwsem);
5003 ext4_ext_release(sb);
5004 ext4_release_system_zone(sb);
5009 ext4_msg(sb, KERN_ERR, "mount failed");
5010 if (EXT4_SB(sb)->rsv_conversion_wq)
5011 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
5013 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
5014 sbi->s_ea_inode_cache = NULL;
5016 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
5017 sbi->s_ea_block_cache = NULL;
5019 if (sbi->s_journal) {
5020 jbd2_journal_destroy(sbi->s_journal);
5021 sbi->s_journal = NULL;
5024 ext4_es_unregister_shrinker(sbi);
5026 del_timer_sync(&sbi->s_err_report);
5028 kthread_stop(sbi->s_mmp_tsk);
5031 group_desc = rcu_dereference(sbi->s_group_desc);
5032 for (i = 0; i < db_count; i++)
5033 brelse(group_desc[i]);
5037 if (sbi->s_chksum_driver)
5038 crypto_free_shash(sbi->s_chksum_driver);
5040 #ifdef CONFIG_UNICODE
5041 utf8_unload(sbi->s_encoding);
5045 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5046 kfree(get_qf_name(sb, sbi, i));
5048 fscrypt_free_dummy_context(&sbi->s_dummy_enc_ctx);
5049 ext4_blkdev_remove(sbi);
5052 sb->s_fs_info = NULL;
5053 kfree(sbi->s_blockgroup_lock);
5057 fs_put_dax(dax_dev);
5058 return err ? err : ret;
5062 * Setup any per-fs journal parameters now. We'll do this both on
5063 * initial mount, once the journal has been initialised but before we've
5064 * done any recovery; and again on any subsequent remount.
5066 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
5068 struct ext4_sb_info *sbi = EXT4_SB(sb);
5070 journal->j_commit_interval = sbi->s_commit_interval;
5071 journal->j_min_batch_time = sbi->s_min_batch_time;
5072 journal->j_max_batch_time = sbi->s_max_batch_time;
5074 write_lock(&journal->j_state_lock);
5075 if (test_opt(sb, BARRIER))
5076 journal->j_flags |= JBD2_BARRIER;
5078 journal->j_flags &= ~JBD2_BARRIER;
5079 if (test_opt(sb, DATA_ERR_ABORT))
5080 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
5082 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
5083 write_unlock(&journal->j_state_lock);
5086 static struct inode *ext4_get_journal_inode(struct super_block *sb,
5087 unsigned int journal_inum)
5089 struct inode *journal_inode;
5092 * Test for the existence of a valid inode on disk. Bad things
5093 * happen if we iget() an unused inode, as the subsequent iput()
5094 * will try to delete it.
5096 journal_inode = ext4_iget(sb, journal_inum, EXT4_IGET_SPECIAL);
5097 if (IS_ERR(journal_inode)) {
5098 ext4_msg(sb, KERN_ERR, "no journal found");
5101 if (!journal_inode->i_nlink) {
5102 make_bad_inode(journal_inode);
5103 iput(journal_inode);
5104 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
5108 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
5109 journal_inode, journal_inode->i_size);
5110 if (!S_ISREG(journal_inode->i_mode)) {
5111 ext4_msg(sb, KERN_ERR, "invalid journal inode");
5112 iput(journal_inode);
5115 return journal_inode;
5118 static journal_t *ext4_get_journal(struct super_block *sb,
5119 unsigned int journal_inum)
5121 struct inode *journal_inode;
5124 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5127 journal_inode = ext4_get_journal_inode(sb, journal_inum);
5131 journal = jbd2_journal_init_inode(journal_inode);
5133 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
5134 iput(journal_inode);
5137 journal->j_private = sb;
5138 ext4_init_journal_params(sb, journal);
5142 static journal_t *ext4_get_dev_journal(struct super_block *sb,
5145 struct buffer_head *bh;
5149 int hblock, blocksize;
5150 ext4_fsblk_t sb_block;
5151 unsigned long offset;
5152 struct ext4_super_block *es;
5153 struct block_device *bdev;
5155 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5158 bdev = ext4_blkdev_get(j_dev, sb);
5162 blocksize = sb->s_blocksize;
5163 hblock = bdev_logical_block_size(bdev);
5164 if (blocksize < hblock) {
5165 ext4_msg(sb, KERN_ERR,
5166 "blocksize too small for journal device");
5170 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
5171 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
5172 set_blocksize(bdev, blocksize);
5173 if (!(bh = __bread(bdev, sb_block, blocksize))) {
5174 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
5175 "external journal");
5179 es = (struct ext4_super_block *) (bh->b_data + offset);
5180 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
5181 !(le32_to_cpu(es->s_feature_incompat) &
5182 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
5183 ext4_msg(sb, KERN_ERR, "external journal has "
5189 if ((le32_to_cpu(es->s_feature_ro_compat) &
5190 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
5191 es->s_checksum != ext4_superblock_csum(sb, es)) {
5192 ext4_msg(sb, KERN_ERR, "external journal has "
5193 "corrupt superblock");
5198 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
5199 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
5204 len = ext4_blocks_count(es);
5205 start = sb_block + 1;
5206 brelse(bh); /* we're done with the superblock */
5208 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
5209 start, len, blocksize);
5211 ext4_msg(sb, KERN_ERR, "failed to create device journal");
5214 journal->j_private = sb;
5215 if (ext4_read_bh_lock(journal->j_sb_buffer, REQ_META | REQ_PRIO, true)) {
5216 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
5219 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
5220 ext4_msg(sb, KERN_ERR, "External journal has more than one "
5221 "user (unsupported) - %d",
5222 be32_to_cpu(journal->j_superblock->s_nr_users));
5225 EXT4_SB(sb)->s_journal_bdev = bdev;
5226 ext4_init_journal_params(sb, journal);
5230 jbd2_journal_destroy(journal);
5232 ext4_blkdev_put(bdev);
5236 static int ext4_load_journal(struct super_block *sb,
5237 struct ext4_super_block *es,
5238 unsigned long journal_devnum)
5241 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
5244 int really_read_only;
5247 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5248 return -EFSCORRUPTED;
5250 if (journal_devnum &&
5251 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
5252 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
5253 "numbers have changed");
5254 journal_dev = new_decode_dev(journal_devnum);
5256 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
5258 if (journal_inum && journal_dev) {
5259 ext4_msg(sb, KERN_ERR,
5260 "filesystem has both journal inode and journal device!");
5265 journal = ext4_get_journal(sb, journal_inum);
5269 journal = ext4_get_dev_journal(sb, journal_dev);
5274 journal_dev_ro = bdev_read_only(journal->j_dev);
5275 really_read_only = bdev_read_only(sb->s_bdev) | journal_dev_ro;
5277 if (journal_dev_ro && !sb_rdonly(sb)) {
5278 ext4_msg(sb, KERN_ERR,
5279 "journal device read-only, try mounting with '-o ro'");
5285 * Are we loading a blank journal or performing recovery after a
5286 * crash? For recovery, we need to check in advance whether we
5287 * can get read-write access to the device.
5289 if (ext4_has_feature_journal_needs_recovery(sb)) {
5290 if (sb_rdonly(sb)) {
5291 ext4_msg(sb, KERN_INFO, "INFO: recovery "
5292 "required on readonly filesystem");
5293 if (really_read_only) {
5294 ext4_msg(sb, KERN_ERR, "write access "
5295 "unavailable, cannot proceed "
5296 "(try mounting with noload)");
5300 ext4_msg(sb, KERN_INFO, "write access will "
5301 "be enabled during recovery");
5305 if (!(journal->j_flags & JBD2_BARRIER))
5306 ext4_msg(sb, KERN_INFO, "barriers disabled");
5308 if (!ext4_has_feature_journal_needs_recovery(sb))
5309 err = jbd2_journal_wipe(journal, !really_read_only);
5311 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
5313 memcpy(save, ((char *) es) +
5314 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
5315 err = jbd2_journal_load(journal);
5317 memcpy(((char *) es) + EXT4_S_ERR_START,
5318 save, EXT4_S_ERR_LEN);
5323 ext4_msg(sb, KERN_ERR, "error loading journal");
5327 EXT4_SB(sb)->s_journal = journal;
5328 err = ext4_clear_journal_err(sb, es);
5330 EXT4_SB(sb)->s_journal = NULL;
5331 jbd2_journal_destroy(journal);
5335 if (!really_read_only && journal_devnum &&
5336 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
5337 es->s_journal_dev = cpu_to_le32(journal_devnum);
5339 /* Make sure we flush the recovery flag to disk. */
5340 ext4_commit_super(sb, 1);
5346 jbd2_journal_destroy(journal);
5350 static int ext4_commit_super(struct super_block *sb, int sync)
5352 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
5353 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
5356 if (!sbh || block_device_ejected(sb))
5360 * If the file system is mounted read-only, don't update the
5361 * superblock write time. This avoids updating the superblock
5362 * write time when we are mounting the root file system
5363 * read/only but we need to replay the journal; at that point,
5364 * for people who are east of GMT and who make their clock
5365 * tick in localtime for Windows bug-for-bug compatibility,
5366 * the clock is set in the future, and this will cause e2fsck
5367 * to complain and force a full file system check.
5369 if (!(sb->s_flags & SB_RDONLY))
5370 ext4_update_tstamp(es, s_wtime);
5371 if (sb->s_bdev->bd_part)
5372 es->s_kbytes_written =
5373 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
5374 ((part_stat_read(sb->s_bdev->bd_part,
5375 sectors[STAT_WRITE]) -
5376 EXT4_SB(sb)->s_sectors_written_start) >> 1));
5378 es->s_kbytes_written =
5379 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
5380 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
5381 ext4_free_blocks_count_set(es,
5382 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
5383 &EXT4_SB(sb)->s_freeclusters_counter)));
5384 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
5385 es->s_free_inodes_count =
5386 cpu_to_le32(percpu_counter_sum_positive(
5387 &EXT4_SB(sb)->s_freeinodes_counter));
5388 BUFFER_TRACE(sbh, "marking dirty");
5389 ext4_superblock_csum_set(sb);
5392 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
5394 * Oh, dear. A previous attempt to write the
5395 * superblock failed. This could happen because the
5396 * USB device was yanked out. Or it could happen to
5397 * be a transient write error and maybe the block will
5398 * be remapped. Nothing we can do but to retry the
5399 * write and hope for the best.
5401 ext4_msg(sb, KERN_ERR, "previous I/O error to "
5402 "superblock detected");
5403 clear_buffer_write_io_error(sbh);
5404 set_buffer_uptodate(sbh);
5406 mark_buffer_dirty(sbh);
5409 error = __sync_dirty_buffer(sbh,
5410 REQ_SYNC | (test_opt(sb, BARRIER) ? REQ_FUA : 0));
5411 if (buffer_write_io_error(sbh)) {
5412 ext4_msg(sb, KERN_ERR, "I/O error while writing "
5414 clear_buffer_write_io_error(sbh);
5415 set_buffer_uptodate(sbh);
5422 * Have we just finished recovery? If so, and if we are mounting (or
5423 * remounting) the filesystem readonly, then we will end up with a
5424 * consistent fs on disk. Record that fact.
5426 static int ext4_mark_recovery_complete(struct super_block *sb,
5427 struct ext4_super_block *es)
5430 journal_t *journal = EXT4_SB(sb)->s_journal;
5432 if (!ext4_has_feature_journal(sb)) {
5433 if (journal != NULL) {
5434 ext4_error(sb, "Journal got removed while the fs was "
5436 return -EFSCORRUPTED;
5440 jbd2_journal_lock_updates(journal);
5441 err = jbd2_journal_flush(journal);
5445 if (ext4_has_feature_journal_needs_recovery(sb) && sb_rdonly(sb)) {
5446 ext4_clear_feature_journal_needs_recovery(sb);
5447 ext4_commit_super(sb, 1);
5450 jbd2_journal_unlock_updates(journal);
5455 * If we are mounting (or read-write remounting) a filesystem whose journal
5456 * has recorded an error from a previous lifetime, move that error to the
5457 * main filesystem now.
5459 static int ext4_clear_journal_err(struct super_block *sb,
5460 struct ext4_super_block *es)
5466 if (!ext4_has_feature_journal(sb)) {
5467 ext4_error(sb, "Journal got removed while the fs was mounted!");
5468 return -EFSCORRUPTED;
5471 journal = EXT4_SB(sb)->s_journal;
5474 * Now check for any error status which may have been recorded in the
5475 * journal by a prior ext4_error() or ext4_abort()
5478 j_errno = jbd2_journal_errno(journal);
5482 errstr = ext4_decode_error(sb, j_errno, nbuf);
5483 ext4_warning(sb, "Filesystem error recorded "
5484 "from previous mount: %s", errstr);
5485 ext4_warning(sb, "Marking fs in need of filesystem check.");
5487 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
5488 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
5489 ext4_commit_super(sb, 1);
5491 jbd2_journal_clear_err(journal);
5492 jbd2_journal_update_sb_errno(journal);
5498 * Force the running and committing transactions to commit,
5499 * and wait on the commit.
5501 int ext4_force_commit(struct super_block *sb)
5508 journal = EXT4_SB(sb)->s_journal;
5509 return ext4_journal_force_commit(journal);
5512 static int ext4_sync_fs(struct super_block *sb, int wait)
5516 bool needs_barrier = false;
5517 struct ext4_sb_info *sbi = EXT4_SB(sb);
5519 if (unlikely(ext4_forced_shutdown(sbi)))
5522 trace_ext4_sync_fs(sb, wait);
5523 flush_workqueue(sbi->rsv_conversion_wq);
5525 * Writeback quota in non-journalled quota case - journalled quota has
5528 dquot_writeback_dquots(sb, -1);
5530 * Data writeback is possible w/o journal transaction, so barrier must
5531 * being sent at the end of the function. But we can skip it if
5532 * transaction_commit will do it for us.
5534 if (sbi->s_journal) {
5535 target = jbd2_get_latest_transaction(sbi->s_journal);
5536 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
5537 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
5538 needs_barrier = true;
5540 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
5542 ret = jbd2_log_wait_commit(sbi->s_journal,
5545 } else if (wait && test_opt(sb, BARRIER))
5546 needs_barrier = true;
5547 if (needs_barrier) {
5549 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL);
5558 * LVM calls this function before a (read-only) snapshot is created. This
5559 * gives us a chance to flush the journal completely and mark the fs clean.
5561 * Note that only this function cannot bring a filesystem to be in a clean
5562 * state independently. It relies on upper layer to stop all data & metadata
5565 static int ext4_freeze(struct super_block *sb)
5573 journal = EXT4_SB(sb)->s_journal;
5576 /* Now we set up the journal barrier. */
5577 jbd2_journal_lock_updates(journal);
5580 * Don't clear the needs_recovery flag if we failed to
5581 * flush the journal.
5583 error = jbd2_journal_flush(journal);
5587 /* Journal blocked and flushed, clear needs_recovery flag. */
5588 ext4_clear_feature_journal_needs_recovery(sb);
5591 error = ext4_commit_super(sb, 1);
5594 /* we rely on upper layer to stop further updates */
5595 jbd2_journal_unlock_updates(journal);
5600 * Called by LVM after the snapshot is done. We need to reset the RECOVER
5601 * flag here, even though the filesystem is not technically dirty yet.
5603 static int ext4_unfreeze(struct super_block *sb)
5605 if (sb_rdonly(sb) || ext4_forced_shutdown(EXT4_SB(sb)))
5608 if (EXT4_SB(sb)->s_journal) {
5609 /* Reset the needs_recovery flag before the fs is unlocked. */
5610 ext4_set_feature_journal_needs_recovery(sb);
5613 ext4_commit_super(sb, 1);
5618 * Structure to save mount options for ext4_remount's benefit
5620 struct ext4_mount_options {
5621 unsigned long s_mount_opt;
5622 unsigned long s_mount_opt2;
5625 unsigned long s_commit_interval;
5626 u32 s_min_batch_time, s_max_batch_time;
5629 char *s_qf_names[EXT4_MAXQUOTAS];
5633 static int ext4_remount(struct super_block *sb, int *flags, char *data)
5635 struct ext4_super_block *es;
5636 struct ext4_sb_info *sbi = EXT4_SB(sb);
5637 unsigned long old_sb_flags, vfs_flags;
5638 struct ext4_mount_options old_opts;
5639 int enable_quota = 0;
5641 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
5645 char *to_free[EXT4_MAXQUOTAS];
5647 char *orig_data = kstrdup(data, GFP_KERNEL);
5649 if (data && !orig_data)
5652 /* Store the original options */
5653 old_sb_flags = sb->s_flags;
5654 old_opts.s_mount_opt = sbi->s_mount_opt;
5655 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
5656 old_opts.s_resuid = sbi->s_resuid;
5657 old_opts.s_resgid = sbi->s_resgid;
5658 old_opts.s_commit_interval = sbi->s_commit_interval;
5659 old_opts.s_min_batch_time = sbi->s_min_batch_time;
5660 old_opts.s_max_batch_time = sbi->s_max_batch_time;
5662 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
5663 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5664 if (sbi->s_qf_names[i]) {
5665 char *qf_name = get_qf_name(sb, sbi, i);
5667 old_opts.s_qf_names[i] = kstrdup(qf_name, GFP_KERNEL);
5668 if (!old_opts.s_qf_names[i]) {
5669 for (j = 0; j < i; j++)
5670 kfree(old_opts.s_qf_names[j]);
5675 old_opts.s_qf_names[i] = NULL;
5677 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
5678 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
5681 * Some options can be enabled by ext4 and/or by VFS mount flag
5682 * either way we need to make sure it matches in both *flags and
5683 * s_flags. Copy those selected flags from *flags to s_flags
5685 vfs_flags = SB_LAZYTIME | SB_I_VERSION;
5686 sb->s_flags = (sb->s_flags & ~vfs_flags) | (*flags & vfs_flags);
5688 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
5693 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
5694 test_opt(sb, JOURNAL_CHECKSUM)) {
5695 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
5696 "during remount not supported; ignoring");
5697 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
5700 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
5701 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
5702 ext4_msg(sb, KERN_ERR, "can't mount with "
5703 "both data=journal and delalloc");
5707 if (test_opt(sb, DIOREAD_NOLOCK)) {
5708 ext4_msg(sb, KERN_ERR, "can't mount with "
5709 "both data=journal and dioread_nolock");
5713 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
5714 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
5715 ext4_msg(sb, KERN_ERR, "can't mount with "
5716 "journal_async_commit in data=ordered mode");
5722 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) {
5723 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount");
5728 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
5729 ext4_abort(sb, EXT4_ERR_ESHUTDOWN, "Abort forced by user");
5731 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
5732 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
5736 if (sbi->s_journal) {
5737 ext4_init_journal_params(sb, sbi->s_journal);
5738 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
5741 if ((bool)(*flags & SB_RDONLY) != sb_rdonly(sb)) {
5742 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
5747 if (*flags & SB_RDONLY) {
5748 err = sync_filesystem(sb);
5751 err = dquot_suspend(sb, -1);
5756 * First of all, the unconditional stuff we have to do
5757 * to disable replay of the journal when we next remount
5759 sb->s_flags |= SB_RDONLY;
5762 * OK, test if we are remounting a valid rw partition
5763 * readonly, and if so set the rdonly flag and then
5764 * mark the partition as valid again.
5766 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
5767 (sbi->s_mount_state & EXT4_VALID_FS))
5768 es->s_state = cpu_to_le16(sbi->s_mount_state);
5770 if (sbi->s_journal) {
5772 * We let remount-ro finish even if marking fs
5773 * as clean failed...
5775 ext4_mark_recovery_complete(sb, es);
5778 kthread_stop(sbi->s_mmp_tsk);
5780 /* Make sure we can mount this feature set readwrite */
5781 if (ext4_has_feature_readonly(sb) ||
5782 !ext4_feature_set_ok(sb, 0)) {
5787 * Make sure the group descriptor checksums
5788 * are sane. If they aren't, refuse to remount r/w.
5790 for (g = 0; g < sbi->s_groups_count; g++) {
5791 struct ext4_group_desc *gdp =
5792 ext4_get_group_desc(sb, g, NULL);
5794 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
5795 ext4_msg(sb, KERN_ERR,
5796 "ext4_remount: Checksum for group %u failed (%u!=%u)",
5797 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
5798 le16_to_cpu(gdp->bg_checksum));
5805 * If we have an unprocessed orphan list hanging
5806 * around from a previously readonly bdev mount,
5807 * require a full umount/remount for now.
5809 if (es->s_last_orphan) {
5810 ext4_msg(sb, KERN_WARNING, "Couldn't "
5811 "remount RDWR because of unprocessed "
5812 "orphan inode list. Please "
5813 "umount/remount instead");
5819 * Mounting a RDONLY partition read-write, so reread
5820 * and store the current valid flag. (It may have
5821 * been changed by e2fsck since we originally mounted
5824 if (sbi->s_journal) {
5825 err = ext4_clear_journal_err(sb, es);
5829 sbi->s_mount_state = le16_to_cpu(es->s_state);
5831 err = ext4_setup_super(sb, es, 0);
5835 sb->s_flags &= ~SB_RDONLY;
5836 if (ext4_has_feature_mmp(sb))
5837 if (ext4_multi_mount_protect(sb,
5838 le64_to_cpu(es->s_mmp_block))) {
5847 * Reinitialize lazy itable initialization thread based on
5850 if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE))
5851 ext4_unregister_li_request(sb);
5853 ext4_group_t first_not_zeroed;
5854 first_not_zeroed = ext4_has_uninit_itable(sb);
5855 ext4_register_li_request(sb, first_not_zeroed);
5859 * Handle creation of system zone data early because it can fail.
5860 * Releasing of existing data is done when we are sure remount will
5863 if (test_opt(sb, BLOCK_VALIDITY) && !sbi->s_system_blks) {
5864 err = ext4_setup_system_zone(sb);
5869 if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY)) {
5870 err = ext4_commit_super(sb, 1);
5876 /* Release old quota file names */
5877 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5878 kfree(old_opts.s_qf_names[i]);
5880 if (sb_any_quota_suspended(sb))
5881 dquot_resume(sb, -1);
5882 else if (ext4_has_feature_quota(sb)) {
5883 err = ext4_enable_quotas(sb);
5889 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks)
5890 ext4_release_system_zone(sb);
5893 * Some options can be enabled by ext4 and/or by VFS mount flag
5894 * either way we need to make sure it matches in both *flags and
5895 * s_flags. Copy those selected flags from s_flags to *flags
5897 *flags = (*flags & ~vfs_flags) | (sb->s_flags & vfs_flags);
5899 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5904 sb->s_flags = old_sb_flags;
5905 sbi->s_mount_opt = old_opts.s_mount_opt;
5906 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
5907 sbi->s_resuid = old_opts.s_resuid;
5908 sbi->s_resgid = old_opts.s_resgid;
5909 sbi->s_commit_interval = old_opts.s_commit_interval;
5910 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5911 sbi->s_max_batch_time = old_opts.s_max_batch_time;
5912 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks)
5913 ext4_release_system_zone(sb);
5915 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
5916 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
5917 to_free[i] = get_qf_name(sb, sbi, i);
5918 rcu_assign_pointer(sbi->s_qf_names[i], old_opts.s_qf_names[i]);
5921 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5929 static int ext4_statfs_project(struct super_block *sb,
5930 kprojid_t projid, struct kstatfs *buf)
5933 struct dquot *dquot;
5937 qid = make_kqid_projid(projid);
5938 dquot = dqget(sb, qid);
5940 return PTR_ERR(dquot);
5941 spin_lock(&dquot->dq_dqb_lock);
5943 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
5944 dquot->dq_dqb.dqb_bhardlimit);
5945 limit >>= sb->s_blocksize_bits;
5947 if (limit && buf->f_blocks > limit) {
5948 curblock = (dquot->dq_dqb.dqb_curspace +
5949 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
5950 buf->f_blocks = limit;
5951 buf->f_bfree = buf->f_bavail =
5952 (buf->f_blocks > curblock) ?
5953 (buf->f_blocks - curblock) : 0;
5956 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
5957 dquot->dq_dqb.dqb_ihardlimit);
5958 if (limit && buf->f_files > limit) {
5959 buf->f_files = limit;
5961 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
5962 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
5965 spin_unlock(&dquot->dq_dqb_lock);
5971 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
5973 struct super_block *sb = dentry->d_sb;
5974 struct ext4_sb_info *sbi = EXT4_SB(sb);
5975 struct ext4_super_block *es = sbi->s_es;
5976 ext4_fsblk_t overhead = 0, resv_blocks;
5979 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
5981 if (!test_opt(sb, MINIX_DF))
5982 overhead = sbi->s_overhead;
5984 buf->f_type = EXT4_SUPER_MAGIC;
5985 buf->f_bsize = sb->s_blocksize;
5986 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
5987 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5988 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
5989 /* prevent underflow in case that few free space is available */
5990 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
5991 buf->f_bavail = buf->f_bfree -
5992 (ext4_r_blocks_count(es) + resv_blocks);
5993 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
5995 buf->f_files = le32_to_cpu(es->s_inodes_count);
5996 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5997 buf->f_namelen = EXT4_NAME_LEN;
5998 fsid = le64_to_cpup((void *)es->s_uuid) ^
5999 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
6000 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
6001 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
6004 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
6005 sb_has_quota_limits_enabled(sb, PRJQUOTA))
6006 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
6015 * Helper functions so that transaction is started before we acquire dqio_sem
6016 * to keep correct lock ordering of transaction > dqio_sem
6018 static inline struct inode *dquot_to_inode(struct dquot *dquot)
6020 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
6023 static int ext4_write_dquot(struct dquot *dquot)
6027 struct inode *inode;
6029 inode = dquot_to_inode(dquot);
6030 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
6031 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
6033 return PTR_ERR(handle);
6034 ret = dquot_commit(dquot);
6035 err = ext4_journal_stop(handle);
6041 static int ext4_acquire_dquot(struct dquot *dquot)
6046 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
6047 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
6049 return PTR_ERR(handle);
6050 ret = dquot_acquire(dquot);
6051 err = ext4_journal_stop(handle);
6057 static int ext4_release_dquot(struct dquot *dquot)
6062 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
6063 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
6064 if (IS_ERR(handle)) {
6065 /* Release dquot anyway to avoid endless cycle in dqput() */
6066 dquot_release(dquot);
6067 return PTR_ERR(handle);
6069 ret = dquot_release(dquot);
6070 err = ext4_journal_stop(handle);
6076 static int ext4_mark_dquot_dirty(struct dquot *dquot)
6078 struct super_block *sb = dquot->dq_sb;
6079 struct ext4_sb_info *sbi = EXT4_SB(sb);
6081 /* Are we journaling quotas? */
6082 if (ext4_has_feature_quota(sb) ||
6083 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
6084 dquot_mark_dquot_dirty(dquot);
6085 return ext4_write_dquot(dquot);
6087 return dquot_mark_dquot_dirty(dquot);
6091 static int ext4_write_info(struct super_block *sb, int type)
6096 /* Data block + inode block */
6097 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
6099 return PTR_ERR(handle);
6100 ret = dquot_commit_info(sb, type);
6101 err = ext4_journal_stop(handle);
6108 * Turn on quotas during mount time - we need to find
6109 * the quota file and such...
6111 static int ext4_quota_on_mount(struct super_block *sb, int type)
6113 return dquot_quota_on_mount(sb, get_qf_name(sb, EXT4_SB(sb), type),
6114 EXT4_SB(sb)->s_jquota_fmt, type);
6117 static void lockdep_set_quota_inode(struct inode *inode, int subclass)
6119 struct ext4_inode_info *ei = EXT4_I(inode);
6121 /* The first argument of lockdep_set_subclass has to be
6122 * *exactly* the same as the argument to init_rwsem() --- in
6123 * this case, in init_once() --- or lockdep gets unhappy
6124 * because the name of the lock is set using the
6125 * stringification of the argument to init_rwsem().
6127 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
6128 lockdep_set_subclass(&ei->i_data_sem, subclass);
6132 * Standard function to be called on quota_on
6134 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
6135 const struct path *path)
6139 if (!test_opt(sb, QUOTA))
6142 /* Quotafile not on the same filesystem? */
6143 if (path->dentry->d_sb != sb)
6145 /* Journaling quota? */
6146 if (EXT4_SB(sb)->s_qf_names[type]) {
6147 /* Quotafile not in fs root? */
6148 if (path->dentry->d_parent != sb->s_root)
6149 ext4_msg(sb, KERN_WARNING,
6150 "Quota file not on filesystem root. "
6151 "Journaled quota will not work");
6152 sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY;
6155 * Clear the flag just in case mount options changed since
6158 sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY;
6162 * When we journal data on quota file, we have to flush journal to see
6163 * all updates to the file when we bypass pagecache...
6165 if (EXT4_SB(sb)->s_journal &&
6166 ext4_should_journal_data(d_inode(path->dentry))) {
6168 * We don't need to lock updates but journal_flush() could
6169 * otherwise be livelocked...
6171 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
6172 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
6173 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
6178 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
6179 err = dquot_quota_on(sb, type, format_id, path);
6181 lockdep_set_quota_inode(path->dentry->d_inode,
6184 struct inode *inode = d_inode(path->dentry);
6188 * Set inode flags to prevent userspace from messing with quota
6189 * files. If this fails, we return success anyway since quotas
6190 * are already enabled and this is not a hard failure.
6193 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
6196 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL;
6197 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
6198 S_NOATIME | S_IMMUTABLE);
6199 err = ext4_mark_inode_dirty(handle, inode);
6200 ext4_journal_stop(handle);
6202 inode_unlock(inode);
6207 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
6211 struct inode *qf_inode;
6212 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
6213 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
6214 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
6215 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
6218 BUG_ON(!ext4_has_feature_quota(sb));
6220 if (!qf_inums[type])
6223 qf_inode = ext4_iget(sb, qf_inums[type], EXT4_IGET_SPECIAL);
6224 if (IS_ERR(qf_inode)) {
6225 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
6226 return PTR_ERR(qf_inode);
6229 /* Don't account quota for quota files to avoid recursion */
6230 qf_inode->i_flags |= S_NOQUOTA;
6231 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
6232 err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
6234 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
6240 /* Enable usage tracking for all quota types. */
6241 static int ext4_enable_quotas(struct super_block *sb)
6244 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
6245 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
6246 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
6247 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
6249 bool quota_mopt[EXT4_MAXQUOTAS] = {
6250 test_opt(sb, USRQUOTA),
6251 test_opt(sb, GRPQUOTA),
6252 test_opt(sb, PRJQUOTA),
6255 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY;
6256 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
6257 if (qf_inums[type]) {
6258 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
6259 DQUOT_USAGE_ENABLED |
6260 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
6263 "Failed to enable quota tracking "
6264 "(type=%d, err=%d). Please run "
6265 "e2fsck to fix.", type, err);
6266 for (type--; type >= 0; type--)
6267 dquot_quota_off(sb, type);
6276 static int ext4_quota_off(struct super_block *sb, int type)
6278 struct inode *inode = sb_dqopt(sb)->files[type];
6282 /* Force all delayed allocation blocks to be allocated.
6283 * Caller already holds s_umount sem */
6284 if (test_opt(sb, DELALLOC))
6285 sync_filesystem(sb);
6287 if (!inode || !igrab(inode))
6290 err = dquot_quota_off(sb, type);
6291 if (err || ext4_has_feature_quota(sb))
6296 * Update modification times of quota files when userspace can
6297 * start looking at them. If we fail, we return success anyway since
6298 * this is not a hard failure and quotas are already disabled.
6300 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
6301 if (IS_ERR(handle)) {
6302 err = PTR_ERR(handle);
6305 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL);
6306 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
6307 inode->i_mtime = inode->i_ctime = current_time(inode);
6308 err = ext4_mark_inode_dirty(handle, inode);
6309 ext4_journal_stop(handle);
6311 inode_unlock(inode);
6313 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL);
6317 return dquot_quota_off(sb, type);
6320 /* Read data from quotafile - avoid pagecache and such because we cannot afford
6321 * acquiring the locks... As quota files are never truncated and quota code
6322 * itself serializes the operations (and no one else should touch the files)
6323 * we don't have to be afraid of races */
6324 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
6325 size_t len, loff_t off)
6327 struct inode *inode = sb_dqopt(sb)->files[type];
6328 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
6329 int offset = off & (sb->s_blocksize - 1);
6332 struct buffer_head *bh;
6333 loff_t i_size = i_size_read(inode);
6337 if (off+len > i_size)
6340 while (toread > 0) {
6341 tocopy = sb->s_blocksize - offset < toread ?
6342 sb->s_blocksize - offset : toread;
6343 bh = ext4_bread(NULL, inode, blk, 0);
6346 if (!bh) /* A hole? */
6347 memset(data, 0, tocopy);
6349 memcpy(data, bh->b_data+offset, tocopy);
6359 /* Write to quotafile (we know the transaction is already started and has
6360 * enough credits) */
6361 static ssize_t ext4_quota_write(struct super_block *sb, int type,
6362 const char *data, size_t len, loff_t off)
6364 struct inode *inode = sb_dqopt(sb)->files[type];
6365 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
6366 int err = 0, err2 = 0, offset = off & (sb->s_blocksize - 1);
6368 struct buffer_head *bh;
6369 handle_t *handle = journal_current_handle();
6371 if (EXT4_SB(sb)->s_journal && !handle) {
6372 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
6373 " cancelled because transaction is not started",
6374 (unsigned long long)off, (unsigned long long)len);
6378 * Since we account only one data block in transaction credits,
6379 * then it is impossible to cross a block boundary.
6381 if (sb->s_blocksize - offset < len) {
6382 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
6383 " cancelled because not block aligned",
6384 (unsigned long long)off, (unsigned long long)len);
6389 bh = ext4_bread(handle, inode, blk,
6390 EXT4_GET_BLOCKS_CREATE |
6391 EXT4_GET_BLOCKS_METADATA_NOFAIL);
6392 } while (PTR_ERR(bh) == -ENOSPC &&
6393 ext4_should_retry_alloc(inode->i_sb, &retries));
6398 BUFFER_TRACE(bh, "get write access");
6399 err = ext4_journal_get_write_access(handle, bh);
6405 memcpy(bh->b_data+offset, data, len);
6406 flush_dcache_page(bh->b_page);
6408 err = ext4_handle_dirty_metadata(handle, NULL, bh);
6411 if (inode->i_size < off + len) {
6412 i_size_write(inode, off + len);
6413 EXT4_I(inode)->i_disksize = inode->i_size;
6414 err2 = ext4_mark_inode_dirty(handle, inode);
6415 if (unlikely(err2 && !err))
6418 return err ? err : len;
6422 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
6423 const char *dev_name, void *data)
6425 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
6428 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
6429 static inline void register_as_ext2(void)
6431 int err = register_filesystem(&ext2_fs_type);
6434 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
6437 static inline void unregister_as_ext2(void)
6439 unregister_filesystem(&ext2_fs_type);
6442 static inline int ext2_feature_set_ok(struct super_block *sb)
6444 if (ext4_has_unknown_ext2_incompat_features(sb))
6448 if (ext4_has_unknown_ext2_ro_compat_features(sb))
6453 static inline void register_as_ext2(void) { }
6454 static inline void unregister_as_ext2(void) { }
6455 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
6458 static inline void register_as_ext3(void)
6460 int err = register_filesystem(&ext3_fs_type);
6463 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
6466 static inline void unregister_as_ext3(void)
6468 unregister_filesystem(&ext3_fs_type);
6471 static inline int ext3_feature_set_ok(struct super_block *sb)
6473 if (ext4_has_unknown_ext3_incompat_features(sb))
6475 if (!ext4_has_feature_journal(sb))
6479 if (ext4_has_unknown_ext3_ro_compat_features(sb))
6484 static struct file_system_type ext4_fs_type = {
6485 .owner = THIS_MODULE,
6487 .mount = ext4_mount,
6488 .kill_sb = kill_block_super,
6489 .fs_flags = FS_REQUIRES_DEV,
6491 MODULE_ALIAS_FS("ext4");
6493 /* Shared across all ext4 file systems */
6494 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
6496 static int __init ext4_init_fs(void)
6500 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
6501 ext4_li_info = NULL;
6502 mutex_init(&ext4_li_mtx);
6504 /* Build-time check for flags consistency */
6505 ext4_check_flag_values();
6507 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
6508 init_waitqueue_head(&ext4__ioend_wq[i]);
6510 err = ext4_init_es();
6514 err = ext4_init_pending();
6518 err = ext4_init_post_read_processing();
6522 err = ext4_init_pageio();
6526 err = ext4_init_system_zone();
6530 err = ext4_init_sysfs();
6534 err = ext4_init_mballoc();
6537 err = init_inodecache();
6542 err = register_filesystem(&ext4_fs_type);
6548 unregister_as_ext2();
6549 unregister_as_ext3();
6550 destroy_inodecache();
6552 ext4_exit_mballoc();
6556 ext4_exit_system_zone();
6560 ext4_exit_post_read_processing();
6562 ext4_exit_pending();
6569 static void __exit ext4_exit_fs(void)
6571 ext4_destroy_lazyinit_thread();
6572 unregister_as_ext2();
6573 unregister_as_ext3();
6574 unregister_filesystem(&ext4_fs_type);
6575 destroy_inodecache();
6576 ext4_exit_mballoc();
6578 ext4_exit_system_zone();
6580 ext4_exit_post_read_processing();
6582 ext4_exit_pending();
6585 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
6586 MODULE_DESCRIPTION("Fourth Extended Filesystem");
6587 MODULE_LICENSE("GPL");
6588 MODULE_SOFTDEP("pre: crc32c");
6589 module_init(ext4_init_fs)
6590 module_exit(ext4_exit_fs)