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 * This works like sb_bread() except it uses ERR_PTR for error
146 * returns. Currently with sb_bread it's impossible to distinguish
147 * between ENOMEM and EIO situations (since both result in a NULL
151 ext4_sb_bread(struct super_block *sb, sector_t block, int op_flags)
153 struct buffer_head *bh = sb_getblk(sb, block);
156 return ERR_PTR(-ENOMEM);
157 if (ext4_buffer_uptodate(bh))
159 ll_rw_block(REQ_OP_READ, REQ_META | op_flags, 1, &bh);
161 if (buffer_uptodate(bh))
164 return ERR_PTR(-EIO);
167 static int ext4_verify_csum_type(struct super_block *sb,
168 struct ext4_super_block *es)
170 if (!ext4_has_feature_metadata_csum(sb))
173 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
176 static __le32 ext4_superblock_csum(struct super_block *sb,
177 struct ext4_super_block *es)
179 struct ext4_sb_info *sbi = EXT4_SB(sb);
180 int offset = offsetof(struct ext4_super_block, s_checksum);
183 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
185 return cpu_to_le32(csum);
188 static int ext4_superblock_csum_verify(struct super_block *sb,
189 struct ext4_super_block *es)
191 if (!ext4_has_metadata_csum(sb))
194 return es->s_checksum == ext4_superblock_csum(sb, es);
197 void ext4_superblock_csum_set(struct super_block *sb)
199 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
201 if (!ext4_has_metadata_csum(sb))
204 es->s_checksum = ext4_superblock_csum(sb, es);
207 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
208 struct ext4_group_desc *bg)
210 return le32_to_cpu(bg->bg_block_bitmap_lo) |
211 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
212 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
215 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
216 struct ext4_group_desc *bg)
218 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
219 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
220 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
223 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
224 struct ext4_group_desc *bg)
226 return le32_to_cpu(bg->bg_inode_table_lo) |
227 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
228 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
231 __u32 ext4_free_group_clusters(struct super_block *sb,
232 struct ext4_group_desc *bg)
234 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
235 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
236 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
239 __u32 ext4_free_inodes_count(struct super_block *sb,
240 struct ext4_group_desc *bg)
242 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
243 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
244 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
247 __u32 ext4_used_dirs_count(struct super_block *sb,
248 struct ext4_group_desc *bg)
250 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
251 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
252 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
255 __u32 ext4_itable_unused_count(struct super_block *sb,
256 struct ext4_group_desc *bg)
258 return le16_to_cpu(bg->bg_itable_unused_lo) |
259 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
260 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
263 void ext4_block_bitmap_set(struct super_block *sb,
264 struct ext4_group_desc *bg, ext4_fsblk_t blk)
266 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
267 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
268 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
271 void ext4_inode_bitmap_set(struct super_block *sb,
272 struct ext4_group_desc *bg, ext4_fsblk_t blk)
274 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
275 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
276 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
279 void ext4_inode_table_set(struct super_block *sb,
280 struct ext4_group_desc *bg, ext4_fsblk_t blk)
282 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
283 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
284 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
287 void ext4_free_group_clusters_set(struct super_block *sb,
288 struct ext4_group_desc *bg, __u32 count)
290 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
291 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
292 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
295 void ext4_free_inodes_set(struct super_block *sb,
296 struct ext4_group_desc *bg, __u32 count)
298 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
299 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
300 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
303 void ext4_used_dirs_set(struct super_block *sb,
304 struct ext4_group_desc *bg, __u32 count)
306 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
307 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
308 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
311 void ext4_itable_unused_set(struct super_block *sb,
312 struct ext4_group_desc *bg, __u32 count)
314 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
315 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
316 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
319 static void __ext4_update_tstamp(__le32 *lo, __u8 *hi)
321 time64_t now = ktime_get_real_seconds();
323 now = clamp_val(now, 0, (1ull << 40) - 1);
325 *lo = cpu_to_le32(lower_32_bits(now));
326 *hi = upper_32_bits(now);
329 static time64_t __ext4_get_tstamp(__le32 *lo, __u8 *hi)
331 return ((time64_t)(*hi) << 32) + le32_to_cpu(*lo);
333 #define ext4_update_tstamp(es, tstamp) \
334 __ext4_update_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi)
335 #define ext4_get_tstamp(es, tstamp) \
336 __ext4_get_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi)
338 static void __save_error_info(struct super_block *sb, int error,
339 __u32 ino, __u64 block,
340 const char *func, unsigned int line)
342 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
345 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
346 if (bdev_read_only(sb->s_bdev))
348 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
349 ext4_update_tstamp(es, s_last_error_time);
350 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
351 es->s_last_error_line = cpu_to_le32(line);
352 es->s_last_error_ino = cpu_to_le32(ino);
353 es->s_last_error_block = cpu_to_le64(block);
359 err = EXT4_ERR_ENOMEM;
362 err = EXT4_ERR_EFSBADCRC;
366 err = EXT4_ERR_EFSCORRUPTED;
369 err = EXT4_ERR_ENOSPC;
372 err = EXT4_ERR_ENOKEY;
375 err = EXT4_ERR_EROFS;
378 err = EXT4_ERR_EFBIG;
381 err = EXT4_ERR_EEXIST;
384 err = EXT4_ERR_ERANGE;
387 err = EXT4_ERR_EOVERFLOW;
390 err = EXT4_ERR_EBUSY;
393 err = EXT4_ERR_ENOTDIR;
396 err = EXT4_ERR_ENOTEMPTY;
399 err = EXT4_ERR_ESHUTDOWN;
402 err = EXT4_ERR_EFAULT;
405 err = EXT4_ERR_UNKNOWN;
407 es->s_last_error_errcode = err;
408 if (!es->s_first_error_time) {
409 es->s_first_error_time = es->s_last_error_time;
410 es->s_first_error_time_hi = es->s_last_error_time_hi;
411 strncpy(es->s_first_error_func, func,
412 sizeof(es->s_first_error_func));
413 es->s_first_error_line = cpu_to_le32(line);
414 es->s_first_error_ino = es->s_last_error_ino;
415 es->s_first_error_block = es->s_last_error_block;
416 es->s_first_error_errcode = es->s_last_error_errcode;
419 * Start the daily error reporting function if it hasn't been
422 if (!es->s_error_count)
423 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
424 le32_add_cpu(&es->s_error_count, 1);
427 static void save_error_info(struct super_block *sb, int error,
428 __u32 ino, __u64 block,
429 const char *func, unsigned int line)
431 __save_error_info(sb, error, ino, block, func, line);
432 if (!bdev_read_only(sb->s_bdev))
433 ext4_commit_super(sb, 1);
437 * The del_gendisk() function uninitializes the disk-specific data
438 * structures, including the bdi structure, without telling anyone
439 * else. Once this happens, any attempt to call mark_buffer_dirty()
440 * (for example, by ext4_commit_super), will cause a kernel OOPS.
441 * This is a kludge to prevent these oops until we can put in a proper
442 * hook in del_gendisk() to inform the VFS and file system layers.
444 static int block_device_ejected(struct super_block *sb)
446 struct inode *bd_inode = sb->s_bdev->bd_inode;
447 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
449 return bdi->dev == NULL;
452 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
454 struct super_block *sb = journal->j_private;
455 struct ext4_sb_info *sbi = EXT4_SB(sb);
456 int error = is_journal_aborted(journal);
457 struct ext4_journal_cb_entry *jce;
459 BUG_ON(txn->t_state == T_FINISHED);
461 ext4_process_freed_data(sb, txn->t_tid);
463 spin_lock(&sbi->s_md_lock);
464 while (!list_empty(&txn->t_private_list)) {
465 jce = list_entry(txn->t_private_list.next,
466 struct ext4_journal_cb_entry, jce_list);
467 list_del_init(&jce->jce_list);
468 spin_unlock(&sbi->s_md_lock);
469 jce->jce_func(sb, jce, error);
470 spin_lock(&sbi->s_md_lock);
472 spin_unlock(&sbi->s_md_lock);
475 static bool system_going_down(void)
477 return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF
478 || system_state == SYSTEM_RESTART;
481 /* Deal with the reporting of failure conditions on a filesystem such as
482 * inconsistencies detected or read IO failures.
484 * On ext2, we can store the error state of the filesystem in the
485 * superblock. That is not possible on ext4, because we may have other
486 * write ordering constraints on the superblock which prevent us from
487 * writing it out straight away; and given that the journal is about to
488 * be aborted, we can't rely on the current, or future, transactions to
489 * write out the superblock safely.
491 * We'll just use the jbd2_journal_abort() error code to record an error in
492 * the journal instead. On recovery, the journal will complain about
493 * that error until we've noted it down and cleared it.
496 static void ext4_handle_error(struct super_block *sb)
498 if (test_opt(sb, WARN_ON_ERROR))
504 if (!test_opt(sb, ERRORS_CONT)) {
505 journal_t *journal = EXT4_SB(sb)->s_journal;
507 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
509 jbd2_journal_abort(journal, -EIO);
512 * We force ERRORS_RO behavior when system is rebooting. Otherwise we
513 * could panic during 'reboot -f' as the underlying device got already
516 if (test_opt(sb, ERRORS_RO) || system_going_down()) {
517 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
519 * Make sure updated value of ->s_mount_flags will be visible
520 * before ->s_flags update
523 sb->s_flags |= SB_RDONLY;
524 } else if (test_opt(sb, ERRORS_PANIC)) {
525 panic("EXT4-fs (device %s): panic forced after error\n",
530 #define ext4_error_ratelimit(sb) \
531 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
534 void __ext4_error(struct super_block *sb, const char *function,
535 unsigned int line, int error, __u64 block,
536 const char *fmt, ...)
538 struct va_format vaf;
541 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
544 trace_ext4_error(sb, function, line);
545 if (ext4_error_ratelimit(sb)) {
550 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
551 sb->s_id, function, line, current->comm, &vaf);
554 save_error_info(sb, error, 0, block, function, line);
555 ext4_handle_error(sb);
558 void __ext4_error_inode(struct inode *inode, const char *function,
559 unsigned int line, ext4_fsblk_t block, int error,
560 const char *fmt, ...)
563 struct va_format vaf;
565 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
568 trace_ext4_error(inode->i_sb, function, line);
569 if (ext4_error_ratelimit(inode->i_sb)) {
574 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
575 "inode #%lu: block %llu: comm %s: %pV\n",
576 inode->i_sb->s_id, function, line, inode->i_ino,
577 block, current->comm, &vaf);
579 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
580 "inode #%lu: comm %s: %pV\n",
581 inode->i_sb->s_id, function, line, inode->i_ino,
582 current->comm, &vaf);
585 save_error_info(inode->i_sb, error, inode->i_ino, block,
587 ext4_handle_error(inode->i_sb);
590 void __ext4_error_file(struct file *file, const char *function,
591 unsigned int line, ext4_fsblk_t block,
592 const char *fmt, ...)
595 struct va_format vaf;
596 struct inode *inode = file_inode(file);
597 char pathname[80], *path;
599 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
602 trace_ext4_error(inode->i_sb, function, line);
603 if (ext4_error_ratelimit(inode->i_sb)) {
604 path = file_path(file, pathname, sizeof(pathname));
612 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
613 "block %llu: comm %s: path %s: %pV\n",
614 inode->i_sb->s_id, function, line, inode->i_ino,
615 block, current->comm, path, &vaf);
618 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
619 "comm %s: path %s: %pV\n",
620 inode->i_sb->s_id, function, line, inode->i_ino,
621 current->comm, path, &vaf);
624 save_error_info(inode->i_sb, EFSCORRUPTED, inode->i_ino, block,
626 ext4_handle_error(inode->i_sb);
629 const char *ext4_decode_error(struct super_block *sb, int errno,
636 errstr = "Corrupt filesystem";
639 errstr = "Filesystem failed CRC";
642 errstr = "IO failure";
645 errstr = "Out of memory";
648 if (!sb || (EXT4_SB(sb)->s_journal &&
649 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
650 errstr = "Journal has aborted";
652 errstr = "Readonly filesystem";
655 /* If the caller passed in an extra buffer for unknown
656 * errors, textualise them now. Else we just return
659 /* Check for truncated error codes... */
660 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
669 /* __ext4_std_error decodes expected errors from journaling functions
670 * automatically and invokes the appropriate error response. */
672 void __ext4_std_error(struct super_block *sb, const char *function,
673 unsigned int line, int errno)
678 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
681 /* Special case: if the error is EROFS, and we're not already
682 * inside a transaction, then there's really no point in logging
684 if (errno == -EROFS && journal_current_handle() == NULL && sb_rdonly(sb))
687 if (ext4_error_ratelimit(sb)) {
688 errstr = ext4_decode_error(sb, errno, nbuf);
689 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
690 sb->s_id, function, line, errstr);
693 save_error_info(sb, -errno, 0, 0, function, line);
694 ext4_handle_error(sb);
698 * ext4_abort is a much stronger failure handler than ext4_error. The
699 * abort function may be used to deal with unrecoverable failures such
700 * as journal IO errors or ENOMEM at a critical moment in log management.
702 * We unconditionally force the filesystem into an ABORT|READONLY state,
703 * unless the error response on the fs has been set to panic in which
704 * case we take the easy way out and panic immediately.
707 void __ext4_abort(struct super_block *sb, const char *function,
708 unsigned int line, int error, const char *fmt, ...)
710 struct va_format vaf;
713 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
716 save_error_info(sb, error, 0, 0, function, line);
720 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: %pV\n",
721 sb->s_id, function, line, &vaf);
724 if (sb_rdonly(sb) == 0) {
725 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
726 if (EXT4_SB(sb)->s_journal)
727 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
729 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
731 * Make sure updated value of ->s_mount_flags will be visible
732 * before ->s_flags update
735 sb->s_flags |= SB_RDONLY;
737 if (test_opt(sb, ERRORS_PANIC) && !system_going_down())
738 panic("EXT4-fs panic from previous error\n");
741 void __ext4_msg(struct super_block *sb,
742 const char *prefix, const char *fmt, ...)
744 struct va_format vaf;
747 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
753 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
757 #define ext4_warning_ratelimit(sb) \
758 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \
761 void __ext4_warning(struct super_block *sb, const char *function,
762 unsigned int line, const char *fmt, ...)
764 struct va_format vaf;
767 if (!ext4_warning_ratelimit(sb))
773 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
774 sb->s_id, function, line, &vaf);
778 void __ext4_warning_inode(const struct inode *inode, const char *function,
779 unsigned int line, const char *fmt, ...)
781 struct va_format vaf;
784 if (!ext4_warning_ratelimit(inode->i_sb))
790 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
791 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
792 function, line, inode->i_ino, current->comm, &vaf);
796 void __ext4_grp_locked_error(const char *function, unsigned int line,
797 struct super_block *sb, ext4_group_t grp,
798 unsigned long ino, ext4_fsblk_t block,
799 const char *fmt, ...)
803 struct va_format vaf;
806 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
809 trace_ext4_error(sb, function, line);
810 __save_error_info(sb, EFSCORRUPTED, ino, block, function, line);
812 if (ext4_error_ratelimit(sb)) {
816 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
817 sb->s_id, function, line, grp);
819 printk(KERN_CONT "inode %lu: ", ino);
821 printk(KERN_CONT "block %llu:",
822 (unsigned long long) block);
823 printk(KERN_CONT "%pV\n", &vaf);
827 if (test_opt(sb, WARN_ON_ERROR))
830 if (test_opt(sb, ERRORS_CONT)) {
831 ext4_commit_super(sb, 0);
835 ext4_unlock_group(sb, grp);
836 ext4_commit_super(sb, 1);
837 ext4_handle_error(sb);
839 * We only get here in the ERRORS_RO case; relocking the group
840 * may be dangerous, but nothing bad will happen since the
841 * filesystem will have already been marked read/only and the
842 * journal has been aborted. We return 1 as a hint to callers
843 * who might what to use the return value from
844 * ext4_grp_locked_error() to distinguish between the
845 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
846 * aggressively from the ext4 function in question, with a
847 * more appropriate error code.
849 ext4_lock_group(sb, grp);
853 void ext4_mark_group_bitmap_corrupted(struct super_block *sb,
857 struct ext4_sb_info *sbi = EXT4_SB(sb);
858 struct ext4_group_info *grp = ext4_get_group_info(sb, group);
859 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, group, NULL);
862 if (flags & EXT4_GROUP_INFO_BBITMAP_CORRUPT) {
863 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT,
866 percpu_counter_sub(&sbi->s_freeclusters_counter,
870 if (flags & EXT4_GROUP_INFO_IBITMAP_CORRUPT) {
871 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT,
876 count = ext4_free_inodes_count(sb, gdp);
877 percpu_counter_sub(&sbi->s_freeinodes_counter,
883 void ext4_update_dynamic_rev(struct super_block *sb)
885 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
887 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
891 "updating to rev %d because of new feature flag, "
892 "running e2fsck is recommended",
895 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
896 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
897 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
898 /* leave es->s_feature_*compat flags alone */
899 /* es->s_uuid will be set by e2fsck if empty */
902 * The rest of the superblock fields should be zero, and if not it
903 * means they are likely already in use, so leave them alone. We
904 * can leave it up to e2fsck to clean up any inconsistencies there.
909 * Open the external journal device
911 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
913 struct block_device *bdev;
915 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
921 ext4_msg(sb, KERN_ERR,
922 "failed to open journal device unknown-block(%u,%u) %ld",
923 MAJOR(dev), MINOR(dev), PTR_ERR(bdev));
928 * Release the journal device
930 static void ext4_blkdev_put(struct block_device *bdev)
932 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
935 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
937 struct block_device *bdev;
938 bdev = sbi->journal_bdev;
940 ext4_blkdev_put(bdev);
941 sbi->journal_bdev = NULL;
945 static inline struct inode *orphan_list_entry(struct list_head *l)
947 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
950 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
954 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
955 le32_to_cpu(sbi->s_es->s_last_orphan));
957 printk(KERN_ERR "sb_info orphan list:\n");
958 list_for_each(l, &sbi->s_orphan) {
959 struct inode *inode = orphan_list_entry(l);
961 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
962 inode->i_sb->s_id, inode->i_ino, inode,
963 inode->i_mode, inode->i_nlink,
969 static int ext4_quota_off(struct super_block *sb, int type);
971 static inline void ext4_quota_off_umount(struct super_block *sb)
975 /* Use our quota_off function to clear inode flags etc. */
976 for (type = 0; type < EXT4_MAXQUOTAS; type++)
977 ext4_quota_off(sb, type);
981 * This is a helper function which is used in the mount/remount
982 * codepaths (which holds s_umount) to fetch the quota file name.
984 static inline char *get_qf_name(struct super_block *sb,
985 struct ext4_sb_info *sbi,
988 return rcu_dereference_protected(sbi->s_qf_names[type],
989 lockdep_is_held(&sb->s_umount));
992 static inline void ext4_quota_off_umount(struct super_block *sb)
997 static void ext4_put_super(struct super_block *sb)
999 struct ext4_sb_info *sbi = EXT4_SB(sb);
1000 struct ext4_super_block *es = sbi->s_es;
1001 struct buffer_head **group_desc;
1002 struct flex_groups **flex_groups;
1006 ext4_unregister_li_request(sb);
1007 ext4_quota_off_umount(sb);
1009 destroy_workqueue(sbi->rsv_conversion_wq);
1012 * Unregister sysfs before destroying jbd2 journal.
1013 * Since we could still access attr_journal_task attribute via sysfs
1014 * path which could have sbi->s_journal->j_task as NULL
1016 ext4_unregister_sysfs(sb);
1018 if (sbi->s_journal) {
1019 aborted = is_journal_aborted(sbi->s_journal);
1020 err = jbd2_journal_destroy(sbi->s_journal);
1021 sbi->s_journal = NULL;
1022 if ((err < 0) && !aborted) {
1023 ext4_abort(sb, -err, "Couldn't clean up the journal");
1027 ext4_es_unregister_shrinker(sbi);
1028 del_timer_sync(&sbi->s_err_report);
1029 ext4_release_system_zone(sb);
1030 ext4_mb_release(sb);
1031 ext4_ext_release(sb);
1033 if (!sb_rdonly(sb) && !aborted) {
1034 ext4_clear_feature_journal_needs_recovery(sb);
1035 es->s_state = cpu_to_le16(sbi->s_mount_state);
1038 ext4_commit_super(sb, 1);
1041 group_desc = rcu_dereference(sbi->s_group_desc);
1042 for (i = 0; i < sbi->s_gdb_count; i++)
1043 brelse(group_desc[i]);
1045 flex_groups = rcu_dereference(sbi->s_flex_groups);
1047 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
1048 kvfree(flex_groups[i]);
1049 kvfree(flex_groups);
1052 percpu_counter_destroy(&sbi->s_freeclusters_counter);
1053 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1054 percpu_counter_destroy(&sbi->s_dirs_counter);
1055 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
1056 percpu_free_rwsem(&sbi->s_writepages_rwsem);
1058 for (i = 0; i < EXT4_MAXQUOTAS; i++)
1059 kfree(get_qf_name(sb, sbi, i));
1062 /* Debugging code just in case the in-memory inode orphan list
1063 * isn't empty. The on-disk one can be non-empty if we've
1064 * detected an error and taken the fs readonly, but the
1065 * in-memory list had better be clean by this point. */
1066 if (!list_empty(&sbi->s_orphan))
1067 dump_orphan_list(sb, sbi);
1068 J_ASSERT(list_empty(&sbi->s_orphan));
1070 sync_blockdev(sb->s_bdev);
1071 invalidate_bdev(sb->s_bdev);
1072 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
1074 * Invalidate the journal device's buffers. We don't want them
1075 * floating about in memory - the physical journal device may
1076 * hotswapped, and it breaks the `ro-after' testing code.
1078 sync_blockdev(sbi->journal_bdev);
1079 invalidate_bdev(sbi->journal_bdev);
1080 ext4_blkdev_remove(sbi);
1083 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
1084 sbi->s_ea_inode_cache = NULL;
1086 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
1087 sbi->s_ea_block_cache = NULL;
1090 kthread_stop(sbi->s_mmp_tsk);
1092 sb->s_fs_info = NULL;
1094 * Now that we are completely done shutting down the
1095 * superblock, we need to actually destroy the kobject.
1097 kobject_put(&sbi->s_kobj);
1098 wait_for_completion(&sbi->s_kobj_unregister);
1099 if (sbi->s_chksum_driver)
1100 crypto_free_shash(sbi->s_chksum_driver);
1101 kfree(sbi->s_blockgroup_lock);
1102 fs_put_dax(sbi->s_daxdev);
1103 fscrypt_free_dummy_context(&sbi->s_dummy_enc_ctx);
1104 #ifdef CONFIG_UNICODE
1105 utf8_unload(sbi->s_encoding);
1110 static struct kmem_cache *ext4_inode_cachep;
1113 * Called inside transaction, so use GFP_NOFS
1115 static struct inode *ext4_alloc_inode(struct super_block *sb)
1117 struct ext4_inode_info *ei;
1119 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
1123 inode_set_iversion(&ei->vfs_inode, 1);
1124 spin_lock_init(&ei->i_raw_lock);
1125 INIT_LIST_HEAD(&ei->i_prealloc_list);
1126 spin_lock_init(&ei->i_prealloc_lock);
1127 ext4_es_init_tree(&ei->i_es_tree);
1128 rwlock_init(&ei->i_es_lock);
1129 INIT_LIST_HEAD(&ei->i_es_list);
1130 ei->i_es_all_nr = 0;
1131 ei->i_es_shk_nr = 0;
1132 ei->i_es_shrink_lblk = 0;
1133 ei->i_reserved_data_blocks = 0;
1134 spin_lock_init(&(ei->i_block_reservation_lock));
1135 ext4_init_pending_tree(&ei->i_pending_tree);
1137 ei->i_reserved_quota = 0;
1138 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
1141 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
1142 spin_lock_init(&ei->i_completed_io_lock);
1144 ei->i_datasync_tid = 0;
1145 atomic_set(&ei->i_unwritten, 0);
1146 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
1147 return &ei->vfs_inode;
1150 static int ext4_drop_inode(struct inode *inode)
1152 int drop = generic_drop_inode(inode);
1155 drop = fscrypt_drop_inode(inode);
1157 trace_ext4_drop_inode(inode, drop);
1161 static void ext4_free_in_core_inode(struct inode *inode)
1163 fscrypt_free_inode(inode);
1164 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
1167 static void ext4_destroy_inode(struct inode *inode)
1169 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
1170 ext4_msg(inode->i_sb, KERN_ERR,
1171 "Inode %lu (%p): orphan list check failed!",
1172 inode->i_ino, EXT4_I(inode));
1173 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1174 EXT4_I(inode), sizeof(struct ext4_inode_info),
1180 static void init_once(void *foo)
1182 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
1184 INIT_LIST_HEAD(&ei->i_orphan);
1185 init_rwsem(&ei->xattr_sem);
1186 init_rwsem(&ei->i_data_sem);
1187 init_rwsem(&ei->i_mmap_sem);
1188 inode_init_once(&ei->vfs_inode);
1191 static int __init init_inodecache(void)
1193 ext4_inode_cachep = kmem_cache_create_usercopy("ext4_inode_cache",
1194 sizeof(struct ext4_inode_info), 0,
1195 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD|
1197 offsetof(struct ext4_inode_info, i_data),
1198 sizeof_field(struct ext4_inode_info, i_data),
1200 if (ext4_inode_cachep == NULL)
1205 static void destroy_inodecache(void)
1208 * Make sure all delayed rcu free inodes are flushed before we
1212 kmem_cache_destroy(ext4_inode_cachep);
1215 void ext4_clear_inode(struct inode *inode)
1217 invalidate_inode_buffers(inode);
1219 ext4_discard_preallocations(inode);
1220 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
1222 if (EXT4_I(inode)->jinode) {
1223 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1224 EXT4_I(inode)->jinode);
1225 jbd2_free_inode(EXT4_I(inode)->jinode);
1226 EXT4_I(inode)->jinode = NULL;
1228 fscrypt_put_encryption_info(inode);
1229 fsverity_cleanup_inode(inode);
1232 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1233 u64 ino, u32 generation)
1235 struct inode *inode;
1238 * Currently we don't know the generation for parent directory, so
1239 * a generation of 0 means "accept any"
1241 inode = ext4_iget(sb, ino, EXT4_IGET_HANDLE);
1243 return ERR_CAST(inode);
1244 if (generation && inode->i_generation != generation) {
1246 return ERR_PTR(-ESTALE);
1252 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1253 int fh_len, int fh_type)
1255 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1256 ext4_nfs_get_inode);
1259 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1260 int fh_len, int fh_type)
1262 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1263 ext4_nfs_get_inode);
1266 static int ext4_nfs_commit_metadata(struct inode *inode)
1268 struct writeback_control wbc = {
1269 .sync_mode = WB_SYNC_ALL
1272 trace_ext4_nfs_commit_metadata(inode);
1273 return ext4_write_inode(inode, &wbc);
1277 * Try to release metadata pages (indirect blocks, directories) which are
1278 * mapped via the block device. Since these pages could have journal heads
1279 * which would prevent try_to_free_buffers() from freeing them, we must use
1280 * jbd2 layer's try_to_free_buffers() function to release them.
1282 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1285 journal_t *journal = EXT4_SB(sb)->s_journal;
1287 WARN_ON(PageChecked(page));
1288 if (!page_has_buffers(page))
1291 return jbd2_journal_try_to_free_buffers(journal, page);
1293 return try_to_free_buffers(page);
1296 #ifdef CONFIG_FS_ENCRYPTION
1297 static int ext4_get_context(struct inode *inode, void *ctx, size_t len)
1299 return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1300 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len);
1303 static int ext4_set_context(struct inode *inode, const void *ctx, size_t len,
1306 handle_t *handle = fs_data;
1307 int res, res2, credits, retries = 0;
1310 * Encrypting the root directory is not allowed because e2fsck expects
1311 * lost+found to exist and be unencrypted, and encrypting the root
1312 * directory would imply encrypting the lost+found directory as well as
1313 * the filename "lost+found" itself.
1315 if (inode->i_ino == EXT4_ROOT_INO)
1318 if (WARN_ON_ONCE(IS_DAX(inode) && i_size_read(inode)))
1321 if (ext4_test_inode_flag(inode, EXT4_INODE_DAX))
1324 res = ext4_convert_inline_data(inode);
1329 * If a journal handle was specified, then the encryption context is
1330 * being set on a new inode via inheritance and is part of a larger
1331 * transaction to create the inode. Otherwise the encryption context is
1332 * being set on an existing inode in its own transaction. Only in the
1333 * latter case should the "retry on ENOSPC" logic be used.
1337 res = ext4_xattr_set_handle(handle, inode,
1338 EXT4_XATTR_INDEX_ENCRYPTION,
1339 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1342 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1343 ext4_clear_inode_state(inode,
1344 EXT4_STATE_MAY_INLINE_DATA);
1346 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1347 * S_DAX may be disabled
1349 ext4_set_inode_flags(inode, false);
1354 res = dquot_initialize(inode);
1358 res = ext4_xattr_set_credits(inode, len, false /* is_create */,
1363 handle = ext4_journal_start(inode, EXT4_HT_MISC, credits);
1365 return PTR_ERR(handle);
1367 res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION,
1368 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1371 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1373 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1374 * S_DAX may be disabled
1376 ext4_set_inode_flags(inode, false);
1377 res = ext4_mark_inode_dirty(handle, inode);
1379 EXT4_ERROR_INODE(inode, "Failed to mark inode dirty");
1381 res2 = ext4_journal_stop(handle);
1383 if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
1390 static const union fscrypt_context *
1391 ext4_get_dummy_context(struct super_block *sb)
1393 return EXT4_SB(sb)->s_dummy_enc_ctx.ctx;
1396 static bool ext4_has_stable_inodes(struct super_block *sb)
1398 return ext4_has_feature_stable_inodes(sb);
1401 static void ext4_get_ino_and_lblk_bits(struct super_block *sb,
1402 int *ino_bits_ret, int *lblk_bits_ret)
1404 *ino_bits_ret = 8 * sizeof(EXT4_SB(sb)->s_es->s_inodes_count);
1405 *lblk_bits_ret = 8 * sizeof(ext4_lblk_t);
1408 static const struct fscrypt_operations ext4_cryptops = {
1409 .key_prefix = "ext4:",
1410 .get_context = ext4_get_context,
1411 .set_context = ext4_set_context,
1412 .get_dummy_context = ext4_get_dummy_context,
1413 .empty_dir = ext4_empty_dir,
1414 .max_namelen = EXT4_NAME_LEN,
1415 .has_stable_inodes = ext4_has_stable_inodes,
1416 .get_ino_and_lblk_bits = ext4_get_ino_and_lblk_bits,
1421 static const char * const quotatypes[] = INITQFNAMES;
1422 #define QTYPE2NAME(t) (quotatypes[t])
1424 static int ext4_write_dquot(struct dquot *dquot);
1425 static int ext4_acquire_dquot(struct dquot *dquot);
1426 static int ext4_release_dquot(struct dquot *dquot);
1427 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1428 static int ext4_write_info(struct super_block *sb, int type);
1429 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1430 const struct path *path);
1431 static int ext4_quota_on_mount(struct super_block *sb, int type);
1432 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1433 size_t len, loff_t off);
1434 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1435 const char *data, size_t len, loff_t off);
1436 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1437 unsigned int flags);
1438 static int ext4_enable_quotas(struct super_block *sb);
1440 static struct dquot **ext4_get_dquots(struct inode *inode)
1442 return EXT4_I(inode)->i_dquot;
1445 static const struct dquot_operations ext4_quota_operations = {
1446 .get_reserved_space = ext4_get_reserved_space,
1447 .write_dquot = ext4_write_dquot,
1448 .acquire_dquot = ext4_acquire_dquot,
1449 .release_dquot = ext4_release_dquot,
1450 .mark_dirty = ext4_mark_dquot_dirty,
1451 .write_info = ext4_write_info,
1452 .alloc_dquot = dquot_alloc,
1453 .destroy_dquot = dquot_destroy,
1454 .get_projid = ext4_get_projid,
1455 .get_inode_usage = ext4_get_inode_usage,
1456 .get_next_id = dquot_get_next_id,
1459 static const struct quotactl_ops ext4_qctl_operations = {
1460 .quota_on = ext4_quota_on,
1461 .quota_off = ext4_quota_off,
1462 .quota_sync = dquot_quota_sync,
1463 .get_state = dquot_get_state,
1464 .set_info = dquot_set_dqinfo,
1465 .get_dqblk = dquot_get_dqblk,
1466 .set_dqblk = dquot_set_dqblk,
1467 .get_nextdqblk = dquot_get_next_dqblk,
1471 static const struct super_operations ext4_sops = {
1472 .alloc_inode = ext4_alloc_inode,
1473 .free_inode = ext4_free_in_core_inode,
1474 .destroy_inode = ext4_destroy_inode,
1475 .write_inode = ext4_write_inode,
1476 .dirty_inode = ext4_dirty_inode,
1477 .drop_inode = ext4_drop_inode,
1478 .evict_inode = ext4_evict_inode,
1479 .put_super = ext4_put_super,
1480 .sync_fs = ext4_sync_fs,
1481 .freeze_fs = ext4_freeze,
1482 .unfreeze_fs = ext4_unfreeze,
1483 .statfs = ext4_statfs,
1484 .remount_fs = ext4_remount,
1485 .show_options = ext4_show_options,
1487 .quota_read = ext4_quota_read,
1488 .quota_write = ext4_quota_write,
1489 .get_dquots = ext4_get_dquots,
1491 .bdev_try_to_free_page = bdev_try_to_free_page,
1494 static const struct export_operations ext4_export_ops = {
1495 .fh_to_dentry = ext4_fh_to_dentry,
1496 .fh_to_parent = ext4_fh_to_parent,
1497 .get_parent = ext4_get_parent,
1498 .commit_metadata = ext4_nfs_commit_metadata,
1502 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1503 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1504 Opt_nouid32, Opt_debug, Opt_removed,
1505 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1506 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1507 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1508 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1509 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1510 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1511 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1512 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1513 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1514 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version,
1515 Opt_dax, Opt_dax_always, Opt_dax_inode, Opt_dax_never,
1516 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_warn_on_error,
1517 Opt_nowarn_on_error, Opt_mblk_io_submit,
1518 Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize,
1519 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1520 Opt_inode_readahead_blks, Opt_journal_ioprio,
1521 Opt_dioread_nolock, Opt_dioread_lock,
1522 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1523 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache,
1524 Opt_prefetch_block_bitmaps,
1527 static const match_table_t tokens = {
1528 {Opt_bsd_df, "bsddf"},
1529 {Opt_minix_df, "minixdf"},
1530 {Opt_grpid, "grpid"},
1531 {Opt_grpid, "bsdgroups"},
1532 {Opt_nogrpid, "nogrpid"},
1533 {Opt_nogrpid, "sysvgroups"},
1534 {Opt_resgid, "resgid=%u"},
1535 {Opt_resuid, "resuid=%u"},
1537 {Opt_err_cont, "errors=continue"},
1538 {Opt_err_panic, "errors=panic"},
1539 {Opt_err_ro, "errors=remount-ro"},
1540 {Opt_nouid32, "nouid32"},
1541 {Opt_debug, "debug"},
1542 {Opt_removed, "oldalloc"},
1543 {Opt_removed, "orlov"},
1544 {Opt_user_xattr, "user_xattr"},
1545 {Opt_nouser_xattr, "nouser_xattr"},
1547 {Opt_noacl, "noacl"},
1548 {Opt_noload, "norecovery"},
1549 {Opt_noload, "noload"},
1550 {Opt_removed, "nobh"},
1551 {Opt_removed, "bh"},
1552 {Opt_commit, "commit=%u"},
1553 {Opt_min_batch_time, "min_batch_time=%u"},
1554 {Opt_max_batch_time, "max_batch_time=%u"},
1555 {Opt_journal_dev, "journal_dev=%u"},
1556 {Opt_journal_path, "journal_path=%s"},
1557 {Opt_journal_checksum, "journal_checksum"},
1558 {Opt_nojournal_checksum, "nojournal_checksum"},
1559 {Opt_journal_async_commit, "journal_async_commit"},
1560 {Opt_abort, "abort"},
1561 {Opt_data_journal, "data=journal"},
1562 {Opt_data_ordered, "data=ordered"},
1563 {Opt_data_writeback, "data=writeback"},
1564 {Opt_data_err_abort, "data_err=abort"},
1565 {Opt_data_err_ignore, "data_err=ignore"},
1566 {Opt_offusrjquota, "usrjquota="},
1567 {Opt_usrjquota, "usrjquota=%s"},
1568 {Opt_offgrpjquota, "grpjquota="},
1569 {Opt_grpjquota, "grpjquota=%s"},
1570 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1571 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1572 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1573 {Opt_grpquota, "grpquota"},
1574 {Opt_noquota, "noquota"},
1575 {Opt_quota, "quota"},
1576 {Opt_usrquota, "usrquota"},
1577 {Opt_prjquota, "prjquota"},
1578 {Opt_barrier, "barrier=%u"},
1579 {Opt_barrier, "barrier"},
1580 {Opt_nobarrier, "nobarrier"},
1581 {Opt_i_version, "i_version"},
1583 {Opt_dax_always, "dax=always"},
1584 {Opt_dax_inode, "dax=inode"},
1585 {Opt_dax_never, "dax=never"},
1586 {Opt_stripe, "stripe=%u"},
1587 {Opt_delalloc, "delalloc"},
1588 {Opt_warn_on_error, "warn_on_error"},
1589 {Opt_nowarn_on_error, "nowarn_on_error"},
1590 {Opt_lazytime, "lazytime"},
1591 {Opt_nolazytime, "nolazytime"},
1592 {Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"},
1593 {Opt_nodelalloc, "nodelalloc"},
1594 {Opt_removed, "mblk_io_submit"},
1595 {Opt_removed, "nomblk_io_submit"},
1596 {Opt_block_validity, "block_validity"},
1597 {Opt_noblock_validity, "noblock_validity"},
1598 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1599 {Opt_journal_ioprio, "journal_ioprio=%u"},
1600 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1601 {Opt_auto_da_alloc, "auto_da_alloc"},
1602 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1603 {Opt_dioread_nolock, "dioread_nolock"},
1604 {Opt_dioread_lock, "nodioread_nolock"},
1605 {Opt_dioread_lock, "dioread_lock"},
1606 {Opt_discard, "discard"},
1607 {Opt_nodiscard, "nodiscard"},
1608 {Opt_init_itable, "init_itable=%u"},
1609 {Opt_init_itable, "init_itable"},
1610 {Opt_noinit_itable, "noinit_itable"},
1611 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1612 {Opt_test_dummy_encryption, "test_dummy_encryption=%s"},
1613 {Opt_test_dummy_encryption, "test_dummy_encryption"},
1614 {Opt_nombcache, "nombcache"},
1615 {Opt_nombcache, "no_mbcache"}, /* for backward compatibility */
1616 {Opt_prefetch_block_bitmaps, "prefetch_block_bitmaps"},
1617 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1618 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1619 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1620 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1621 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1625 static ext4_fsblk_t get_sb_block(void **data)
1627 ext4_fsblk_t sb_block;
1628 char *options = (char *) *data;
1630 if (!options || strncmp(options, "sb=", 3) != 0)
1631 return 1; /* Default location */
1634 /* TODO: use simple_strtoll with >32bit ext4 */
1635 sb_block = simple_strtoul(options, &options, 0);
1636 if (*options && *options != ',') {
1637 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1641 if (*options == ',')
1643 *data = (void *) options;
1648 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1649 static const char deprecated_msg[] =
1650 "Mount option \"%s\" will be removed by %s\n"
1651 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1654 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1656 struct ext4_sb_info *sbi = EXT4_SB(sb);
1657 char *qname, *old_qname = get_qf_name(sb, sbi, qtype);
1660 if (sb_any_quota_loaded(sb) && !old_qname) {
1661 ext4_msg(sb, KERN_ERR,
1662 "Cannot change journaled "
1663 "quota options when quota turned on");
1666 if (ext4_has_feature_quota(sb)) {
1667 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1668 "ignored when QUOTA feature is enabled");
1671 qname = match_strdup(args);
1673 ext4_msg(sb, KERN_ERR,
1674 "Not enough memory for storing quotafile name");
1678 if (strcmp(old_qname, qname) == 0)
1681 ext4_msg(sb, KERN_ERR,
1682 "%s quota file already specified",
1686 if (strchr(qname, '/')) {
1687 ext4_msg(sb, KERN_ERR,
1688 "quotafile must be on filesystem root");
1691 rcu_assign_pointer(sbi->s_qf_names[qtype], qname);
1699 static int clear_qf_name(struct super_block *sb, int qtype)
1702 struct ext4_sb_info *sbi = EXT4_SB(sb);
1703 char *old_qname = get_qf_name(sb, sbi, qtype);
1705 if (sb_any_quota_loaded(sb) && old_qname) {
1706 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1707 " when quota turned on");
1710 rcu_assign_pointer(sbi->s_qf_names[qtype], NULL);
1717 #define MOPT_SET 0x0001
1718 #define MOPT_CLEAR 0x0002
1719 #define MOPT_NOSUPPORT 0x0004
1720 #define MOPT_EXPLICIT 0x0008
1721 #define MOPT_CLEAR_ERR 0x0010
1722 #define MOPT_GTE0 0x0020
1725 #define MOPT_QFMT 0x0040
1727 #define MOPT_Q MOPT_NOSUPPORT
1728 #define MOPT_QFMT MOPT_NOSUPPORT
1730 #define MOPT_DATAJ 0x0080
1731 #define MOPT_NO_EXT2 0x0100
1732 #define MOPT_NO_EXT3 0x0200
1733 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1734 #define MOPT_STRING 0x0400
1735 #define MOPT_SKIP 0x0800
1737 static const struct mount_opts {
1741 } ext4_mount_opts[] = {
1742 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1743 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1744 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1745 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1746 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1747 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1748 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1749 MOPT_EXT4_ONLY | MOPT_SET},
1750 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1751 MOPT_EXT4_ONLY | MOPT_CLEAR},
1752 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1753 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1754 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1755 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1756 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1757 MOPT_EXT4_ONLY | MOPT_CLEAR},
1758 {Opt_warn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_SET},
1759 {Opt_nowarn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_CLEAR},
1760 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1761 MOPT_EXT4_ONLY | MOPT_CLEAR},
1762 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1763 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1764 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1765 EXT4_MOUNT_JOURNAL_CHECKSUM),
1766 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1767 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1768 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1769 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1770 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1771 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1773 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1775 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1776 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1777 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1778 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1779 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1780 {Opt_commit, 0, MOPT_GTE0},
1781 {Opt_max_batch_time, 0, MOPT_GTE0},
1782 {Opt_min_batch_time, 0, MOPT_GTE0},
1783 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1784 {Opt_init_itable, 0, MOPT_GTE0},
1785 {Opt_dax, EXT4_MOUNT_DAX_ALWAYS, MOPT_SET | MOPT_SKIP},
1786 {Opt_dax_always, EXT4_MOUNT_DAX_ALWAYS,
1787 MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP},
1788 {Opt_dax_inode, EXT4_MOUNT2_DAX_INODE,
1789 MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP},
1790 {Opt_dax_never, EXT4_MOUNT2_DAX_NEVER,
1791 MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP},
1792 {Opt_stripe, 0, MOPT_GTE0},
1793 {Opt_resuid, 0, MOPT_GTE0},
1794 {Opt_resgid, 0, MOPT_GTE0},
1795 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1796 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1797 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1798 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1799 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1800 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1801 MOPT_NO_EXT2 | MOPT_DATAJ},
1802 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1803 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1804 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1805 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1806 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1808 {Opt_acl, 0, MOPT_NOSUPPORT},
1809 {Opt_noacl, 0, MOPT_NOSUPPORT},
1811 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1812 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1813 {Opt_debug_want_extra_isize, 0, MOPT_GTE0},
1814 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1815 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1817 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1819 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1821 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1822 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1823 MOPT_CLEAR | MOPT_Q},
1824 {Opt_usrjquota, 0, MOPT_Q},
1825 {Opt_grpjquota, 0, MOPT_Q},
1826 {Opt_offusrjquota, 0, MOPT_Q},
1827 {Opt_offgrpjquota, 0, MOPT_Q},
1828 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1829 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1830 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1831 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1832 {Opt_test_dummy_encryption, 0, MOPT_STRING},
1833 {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET},
1834 {Opt_prefetch_block_bitmaps, EXT4_MOUNT_PREFETCH_BLOCK_BITMAPS,
1839 #ifdef CONFIG_UNICODE
1840 static const struct ext4_sb_encodings {
1844 } ext4_sb_encoding_map[] = {
1845 {EXT4_ENC_UTF8_12_1, "utf8", "12.1.0"},
1848 static int ext4_sb_read_encoding(const struct ext4_super_block *es,
1849 const struct ext4_sb_encodings **encoding,
1852 __u16 magic = le16_to_cpu(es->s_encoding);
1855 for (i = 0; i < ARRAY_SIZE(ext4_sb_encoding_map); i++)
1856 if (magic == ext4_sb_encoding_map[i].magic)
1859 if (i >= ARRAY_SIZE(ext4_sb_encoding_map))
1862 *encoding = &ext4_sb_encoding_map[i];
1863 *flags = le16_to_cpu(es->s_encoding_flags);
1869 static int ext4_set_test_dummy_encryption(struct super_block *sb,
1871 const substring_t *arg,
1874 #ifdef CONFIG_FS_ENCRYPTION
1875 struct ext4_sb_info *sbi = EXT4_SB(sb);
1879 * This mount option is just for testing, and it's not worthwhile to
1880 * implement the extra complexity (e.g. RCU protection) that would be
1881 * needed to allow it to be set or changed during remount. We do allow
1882 * it to be specified during remount, but only if there is no change.
1884 if (is_remount && !sbi->s_dummy_enc_ctx.ctx) {
1885 ext4_msg(sb, KERN_WARNING,
1886 "Can't set test_dummy_encryption on remount");
1889 err = fscrypt_set_test_dummy_encryption(sb, arg, &sbi->s_dummy_enc_ctx);
1892 ext4_msg(sb, KERN_WARNING,
1893 "Can't change test_dummy_encryption on remount");
1894 else if (err == -EINVAL)
1895 ext4_msg(sb, KERN_WARNING,
1896 "Value of option \"%s\" is unrecognized", opt);
1898 ext4_msg(sb, KERN_WARNING,
1899 "Error processing option \"%s\" [%d]",
1903 ext4_msg(sb, KERN_WARNING, "Test dummy encryption mode enabled");
1905 ext4_msg(sb, KERN_WARNING,
1906 "Test dummy encryption mount option ignored");
1911 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1912 substring_t *args, unsigned long *journal_devnum,
1913 unsigned int *journal_ioprio, int is_remount)
1915 struct ext4_sb_info *sbi = EXT4_SB(sb);
1916 const struct mount_opts *m;
1922 if (token == Opt_usrjquota)
1923 return set_qf_name(sb, USRQUOTA, &args[0]);
1924 else if (token == Opt_grpjquota)
1925 return set_qf_name(sb, GRPQUOTA, &args[0]);
1926 else if (token == Opt_offusrjquota)
1927 return clear_qf_name(sb, USRQUOTA);
1928 else if (token == Opt_offgrpjquota)
1929 return clear_qf_name(sb, GRPQUOTA);
1933 case Opt_nouser_xattr:
1934 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1937 return 1; /* handled by get_sb_block() */
1939 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1942 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1945 sb->s_flags |= SB_I_VERSION;
1948 sb->s_flags |= SB_LAZYTIME;
1950 case Opt_nolazytime:
1951 sb->s_flags &= ~SB_LAZYTIME;
1955 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1956 if (token == m->token)
1959 if (m->token == Opt_err) {
1960 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1961 "or missing value", opt);
1965 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1966 ext4_msg(sb, KERN_ERR,
1967 "Mount option \"%s\" incompatible with ext2", opt);
1970 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1971 ext4_msg(sb, KERN_ERR,
1972 "Mount option \"%s\" incompatible with ext3", opt);
1976 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
1978 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1980 if (m->flags & MOPT_EXPLICIT) {
1981 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
1982 set_opt2(sb, EXPLICIT_DELALLOC);
1983 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
1984 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
1988 if (m->flags & MOPT_CLEAR_ERR)
1989 clear_opt(sb, ERRORS_MASK);
1990 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1991 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1992 "options when quota turned on");
1996 if (m->flags & MOPT_NOSUPPORT) {
1997 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1998 } else if (token == Opt_commit) {
2000 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
2001 else if (arg > INT_MAX / HZ) {
2002 ext4_msg(sb, KERN_ERR,
2003 "Invalid commit interval %d, "
2004 "must be smaller than %d",
2008 sbi->s_commit_interval = HZ * arg;
2009 } else if (token == Opt_debug_want_extra_isize) {
2012 (arg > (sbi->s_inode_size - EXT4_GOOD_OLD_INODE_SIZE))) {
2013 ext4_msg(sb, KERN_ERR,
2014 "Invalid want_extra_isize %d", arg);
2017 sbi->s_want_extra_isize = arg;
2018 } else if (token == Opt_max_batch_time) {
2019 sbi->s_max_batch_time = arg;
2020 } else if (token == Opt_min_batch_time) {
2021 sbi->s_min_batch_time = arg;
2022 } else if (token == Opt_inode_readahead_blks) {
2023 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
2024 ext4_msg(sb, KERN_ERR,
2025 "EXT4-fs: inode_readahead_blks must be "
2026 "0 or a power of 2 smaller than 2^31");
2029 sbi->s_inode_readahead_blks = arg;
2030 } else if (token == Opt_init_itable) {
2031 set_opt(sb, INIT_INODE_TABLE);
2033 arg = EXT4_DEF_LI_WAIT_MULT;
2034 sbi->s_li_wait_mult = arg;
2035 } else if (token == Opt_max_dir_size_kb) {
2036 sbi->s_max_dir_size_kb = arg;
2037 } else if (token == Opt_stripe) {
2038 sbi->s_stripe = arg;
2039 } else if (token == Opt_resuid) {
2040 uid = make_kuid(current_user_ns(), arg);
2041 if (!uid_valid(uid)) {
2042 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
2045 sbi->s_resuid = uid;
2046 } else if (token == Opt_resgid) {
2047 gid = make_kgid(current_user_ns(), arg);
2048 if (!gid_valid(gid)) {
2049 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
2052 sbi->s_resgid = gid;
2053 } else if (token == Opt_journal_dev) {
2055 ext4_msg(sb, KERN_ERR,
2056 "Cannot specify journal on remount");
2059 *journal_devnum = arg;
2060 } else if (token == Opt_journal_path) {
2062 struct inode *journal_inode;
2067 ext4_msg(sb, KERN_ERR,
2068 "Cannot specify journal on remount");
2071 journal_path = match_strdup(&args[0]);
2072 if (!journal_path) {
2073 ext4_msg(sb, KERN_ERR, "error: could not dup "
2074 "journal device string");
2078 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
2080 ext4_msg(sb, KERN_ERR, "error: could not find "
2081 "journal device path: error %d", error);
2082 kfree(journal_path);
2086 journal_inode = d_inode(path.dentry);
2087 if (!S_ISBLK(journal_inode->i_mode)) {
2088 ext4_msg(sb, KERN_ERR, "error: journal path %s "
2089 "is not a block device", journal_path);
2091 kfree(journal_path);
2095 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
2097 kfree(journal_path);
2098 } else if (token == Opt_journal_ioprio) {
2100 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
2105 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
2106 } else if (token == Opt_test_dummy_encryption) {
2107 return ext4_set_test_dummy_encryption(sb, opt, &args[0],
2109 } else if (m->flags & MOPT_DATAJ) {
2111 if (!sbi->s_journal)
2112 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
2113 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
2114 ext4_msg(sb, KERN_ERR,
2115 "Cannot change data mode on remount");
2119 clear_opt(sb, DATA_FLAGS);
2120 sbi->s_mount_opt |= m->mount_opt;
2123 } else if (m->flags & MOPT_QFMT) {
2124 if (sb_any_quota_loaded(sb) &&
2125 sbi->s_jquota_fmt != m->mount_opt) {
2126 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
2127 "quota options when quota turned on");
2130 if (ext4_has_feature_quota(sb)) {
2131 ext4_msg(sb, KERN_INFO,
2132 "Quota format mount options ignored "
2133 "when QUOTA feature is enabled");
2136 sbi->s_jquota_fmt = m->mount_opt;
2138 } else if (token == Opt_dax || token == Opt_dax_always ||
2139 token == Opt_dax_inode || token == Opt_dax_never) {
2140 #ifdef CONFIG_FS_DAX
2143 case Opt_dax_always:
2145 (!(sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2146 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER))) {
2147 fail_dax_change_remount:
2148 ext4_msg(sb, KERN_ERR, "can't change "
2149 "dax mount option while remounting");
2153 (test_opt(sb, DATA_FLAGS) ==
2154 EXT4_MOUNT_JOURNAL_DATA)) {
2155 ext4_msg(sb, KERN_ERR, "can't mount with "
2156 "both data=journal and dax");
2159 ext4_msg(sb, KERN_WARNING,
2160 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
2161 sbi->s_mount_opt |= EXT4_MOUNT_DAX_ALWAYS;
2162 sbi->s_mount_opt2 &= ~EXT4_MOUNT2_DAX_NEVER;
2166 (!(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2167 (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS)))
2168 goto fail_dax_change_remount;
2169 sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_NEVER;
2170 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS;
2174 ((sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2175 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2176 !(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_INODE)))
2177 goto fail_dax_change_remount;
2178 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS;
2179 sbi->s_mount_opt2 &= ~EXT4_MOUNT2_DAX_NEVER;
2180 /* Strictly for printing options */
2181 sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_INODE;
2185 ext4_msg(sb, KERN_INFO, "dax option not supported");
2186 sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_NEVER;
2187 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS;
2190 } else if (token == Opt_data_err_abort) {
2191 sbi->s_mount_opt |= m->mount_opt;
2192 } else if (token == Opt_data_err_ignore) {
2193 sbi->s_mount_opt &= ~m->mount_opt;
2197 if (m->flags & MOPT_CLEAR)
2199 else if (unlikely(!(m->flags & MOPT_SET))) {
2200 ext4_msg(sb, KERN_WARNING,
2201 "buggy handling of option %s", opt);
2206 sbi->s_mount_opt |= m->mount_opt;
2208 sbi->s_mount_opt &= ~m->mount_opt;
2213 static int parse_options(char *options, struct super_block *sb,
2214 unsigned long *journal_devnum,
2215 unsigned int *journal_ioprio,
2218 struct ext4_sb_info __maybe_unused *sbi = EXT4_SB(sb);
2219 char *p, __maybe_unused *usr_qf_name, __maybe_unused *grp_qf_name;
2220 substring_t args[MAX_OPT_ARGS];
2226 while ((p = strsep(&options, ",")) != NULL) {
2230 * Initialize args struct so we know whether arg was
2231 * found; some options take optional arguments.
2233 args[0].to = args[0].from = NULL;
2234 token = match_token(p, tokens, args);
2235 if (handle_mount_opt(sb, p, token, args, journal_devnum,
2236 journal_ioprio, is_remount) < 0)
2241 * We do the test below only for project quotas. 'usrquota' and
2242 * 'grpquota' mount options are allowed even without quota feature
2243 * to support legacy quotas in quota files.
2245 if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) {
2246 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. "
2247 "Cannot enable project quota enforcement.");
2250 usr_qf_name = get_qf_name(sb, sbi, USRQUOTA);
2251 grp_qf_name = get_qf_name(sb, sbi, GRPQUOTA);
2252 if (usr_qf_name || grp_qf_name) {
2253 if (test_opt(sb, USRQUOTA) && usr_qf_name)
2254 clear_opt(sb, USRQUOTA);
2256 if (test_opt(sb, GRPQUOTA) && grp_qf_name)
2257 clear_opt(sb, GRPQUOTA);
2259 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
2260 ext4_msg(sb, KERN_ERR, "old and new quota "
2265 if (!sbi->s_jquota_fmt) {
2266 ext4_msg(sb, KERN_ERR, "journaled quota format "
2272 if (test_opt(sb, DIOREAD_NOLOCK)) {
2274 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
2275 if (blocksize < PAGE_SIZE)
2276 ext4_msg(sb, KERN_WARNING, "Warning: mounting with an "
2277 "experimental mount option 'dioread_nolock' "
2278 "for blocksize < PAGE_SIZE");
2283 static inline void ext4_show_quota_options(struct seq_file *seq,
2284 struct super_block *sb)
2286 #if defined(CONFIG_QUOTA)
2287 struct ext4_sb_info *sbi = EXT4_SB(sb);
2288 char *usr_qf_name, *grp_qf_name;
2290 if (sbi->s_jquota_fmt) {
2293 switch (sbi->s_jquota_fmt) {
2304 seq_printf(seq, ",jqfmt=%s", fmtname);
2308 usr_qf_name = rcu_dereference(sbi->s_qf_names[USRQUOTA]);
2309 grp_qf_name = rcu_dereference(sbi->s_qf_names[GRPQUOTA]);
2311 seq_show_option(seq, "usrjquota", usr_qf_name);
2313 seq_show_option(seq, "grpjquota", grp_qf_name);
2318 static const char *token2str(int token)
2320 const struct match_token *t;
2322 for (t = tokens; t->token != Opt_err; t++)
2323 if (t->token == token && !strchr(t->pattern, '='))
2330 * - it's set to a non-default value OR
2331 * - if the per-sb default is different from the global default
2333 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
2336 struct ext4_sb_info *sbi = EXT4_SB(sb);
2337 struct ext4_super_block *es = sbi->s_es;
2338 int def_errors, def_mount_opt = sbi->s_def_mount_opt;
2339 const struct mount_opts *m;
2340 char sep = nodefs ? '\n' : ',';
2342 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2343 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2345 if (sbi->s_sb_block != 1)
2346 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
2348 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
2349 int want_set = m->flags & MOPT_SET;
2350 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
2351 (m->flags & MOPT_CLEAR_ERR) || m->flags & MOPT_SKIP)
2353 if (!nodefs && !(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
2354 continue; /* skip if same as the default */
2356 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
2357 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
2358 continue; /* select Opt_noFoo vs Opt_Foo */
2359 SEQ_OPTS_PRINT("%s", token2str(m->token));
2362 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
2363 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
2364 SEQ_OPTS_PRINT("resuid=%u",
2365 from_kuid_munged(&init_user_ns, sbi->s_resuid));
2366 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
2367 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
2368 SEQ_OPTS_PRINT("resgid=%u",
2369 from_kgid_munged(&init_user_ns, sbi->s_resgid));
2370 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
2371 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
2372 SEQ_OPTS_PUTS("errors=remount-ro");
2373 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
2374 SEQ_OPTS_PUTS("errors=continue");
2375 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
2376 SEQ_OPTS_PUTS("errors=panic");
2377 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
2378 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
2379 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
2380 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
2381 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
2382 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2383 if (sb->s_flags & SB_I_VERSION)
2384 SEQ_OPTS_PUTS("i_version");
2385 if (nodefs || sbi->s_stripe)
2386 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
2387 if (nodefs || EXT4_MOUNT_DATA_FLAGS &
2388 (sbi->s_mount_opt ^ def_mount_opt)) {
2389 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2390 SEQ_OPTS_PUTS("data=journal");
2391 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2392 SEQ_OPTS_PUTS("data=ordered");
2393 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2394 SEQ_OPTS_PUTS("data=writeback");
2397 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
2398 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2399 sbi->s_inode_readahead_blks);
2401 if (test_opt(sb, INIT_INODE_TABLE) && (nodefs ||
2402 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
2403 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
2404 if (nodefs || sbi->s_max_dir_size_kb)
2405 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2406 if (test_opt(sb, DATA_ERR_ABORT))
2407 SEQ_OPTS_PUTS("data_err=abort");
2409 fscrypt_show_test_dummy_encryption(seq, sep, sb);
2411 if (test_opt(sb, DAX_ALWAYS)) {
2413 SEQ_OPTS_PUTS("dax");
2415 SEQ_OPTS_PUTS("dax=always");
2416 } else if (test_opt2(sb, DAX_NEVER)) {
2417 SEQ_OPTS_PUTS("dax=never");
2418 } else if (test_opt2(sb, DAX_INODE)) {
2419 SEQ_OPTS_PUTS("dax=inode");
2422 ext4_show_quota_options(seq, sb);
2426 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2428 return _ext4_show_options(seq, root->d_sb, 0);
2431 int ext4_seq_options_show(struct seq_file *seq, void *offset)
2433 struct super_block *sb = seq->private;
2436 seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw");
2437 rc = _ext4_show_options(seq, sb, 1);
2438 seq_puts(seq, "\n");
2442 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
2445 struct ext4_sb_info *sbi = EXT4_SB(sb);
2448 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
2449 ext4_msg(sb, KERN_ERR, "revision level too high, "
2450 "forcing read-only mode");
2456 if (!(sbi->s_mount_state & EXT4_VALID_FS))
2457 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
2458 "running e2fsck is recommended");
2459 else if (sbi->s_mount_state & EXT4_ERROR_FS)
2460 ext4_msg(sb, KERN_WARNING,
2461 "warning: mounting fs with errors, "
2462 "running e2fsck is recommended");
2463 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
2464 le16_to_cpu(es->s_mnt_count) >=
2465 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
2466 ext4_msg(sb, KERN_WARNING,
2467 "warning: maximal mount count reached, "
2468 "running e2fsck is recommended");
2469 else if (le32_to_cpu(es->s_checkinterval) &&
2470 (ext4_get_tstamp(es, s_lastcheck) +
2471 le32_to_cpu(es->s_checkinterval) <= ktime_get_real_seconds()))
2472 ext4_msg(sb, KERN_WARNING,
2473 "warning: checktime reached, "
2474 "running e2fsck is recommended");
2475 if (!sbi->s_journal)
2476 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
2477 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
2478 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
2479 le16_add_cpu(&es->s_mnt_count, 1);
2480 ext4_update_tstamp(es, s_mtime);
2482 ext4_set_feature_journal_needs_recovery(sb);
2484 err = ext4_commit_super(sb, 1);
2486 if (test_opt(sb, DEBUG))
2487 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
2488 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
2490 sbi->s_groups_count,
2491 EXT4_BLOCKS_PER_GROUP(sb),
2492 EXT4_INODES_PER_GROUP(sb),
2493 sbi->s_mount_opt, sbi->s_mount_opt2);
2495 cleancache_init_fs(sb);
2499 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2501 struct ext4_sb_info *sbi = EXT4_SB(sb);
2502 struct flex_groups **old_groups, **new_groups;
2505 if (!sbi->s_log_groups_per_flex)
2508 size = ext4_flex_group(sbi, ngroup - 1) + 1;
2509 if (size <= sbi->s_flex_groups_allocated)
2512 new_groups = kvzalloc(roundup_pow_of_two(size *
2513 sizeof(*sbi->s_flex_groups)), GFP_KERNEL);
2515 ext4_msg(sb, KERN_ERR,
2516 "not enough memory for %d flex group pointers", size);
2519 for (i = sbi->s_flex_groups_allocated; i < size; i++) {
2520 new_groups[i] = kvzalloc(roundup_pow_of_two(
2521 sizeof(struct flex_groups)),
2523 if (!new_groups[i]) {
2524 for (j = sbi->s_flex_groups_allocated; j < i; j++)
2525 kvfree(new_groups[j]);
2527 ext4_msg(sb, KERN_ERR,
2528 "not enough memory for %d flex groups", size);
2533 old_groups = rcu_dereference(sbi->s_flex_groups);
2535 memcpy(new_groups, old_groups,
2536 (sbi->s_flex_groups_allocated *
2537 sizeof(struct flex_groups *)));
2539 rcu_assign_pointer(sbi->s_flex_groups, new_groups);
2540 sbi->s_flex_groups_allocated = size;
2542 ext4_kvfree_array_rcu(old_groups);
2546 static int ext4_fill_flex_info(struct super_block *sb)
2548 struct ext4_sb_info *sbi = EXT4_SB(sb);
2549 struct ext4_group_desc *gdp = NULL;
2550 struct flex_groups *fg;
2551 ext4_group_t flex_group;
2554 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2555 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
2556 sbi->s_log_groups_per_flex = 0;
2560 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2564 for (i = 0; i < sbi->s_groups_count; i++) {
2565 gdp = ext4_get_group_desc(sb, i, NULL);
2567 flex_group = ext4_flex_group(sbi, i);
2568 fg = sbi_array_rcu_deref(sbi, s_flex_groups, flex_group);
2569 atomic_add(ext4_free_inodes_count(sb, gdp), &fg->free_inodes);
2570 atomic64_add(ext4_free_group_clusters(sb, gdp),
2571 &fg->free_clusters);
2572 atomic_add(ext4_used_dirs_count(sb, gdp), &fg->used_dirs);
2580 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
2581 struct ext4_group_desc *gdp)
2583 int offset = offsetof(struct ext4_group_desc, bg_checksum);
2585 __le32 le_group = cpu_to_le32(block_group);
2586 struct ext4_sb_info *sbi = EXT4_SB(sb);
2588 if (ext4_has_metadata_csum(sbi->s_sb)) {
2589 /* Use new metadata_csum algorithm */
2591 __u16 dummy_csum = 0;
2593 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2595 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2596 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2597 sizeof(dummy_csum));
2598 offset += sizeof(dummy_csum);
2599 if (offset < sbi->s_desc_size)
2600 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2601 sbi->s_desc_size - offset);
2603 crc = csum32 & 0xFFFF;
2607 /* old crc16 code */
2608 if (!ext4_has_feature_gdt_csum(sb))
2611 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2612 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2613 crc = crc16(crc, (__u8 *)gdp, offset);
2614 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2615 /* for checksum of struct ext4_group_desc do the rest...*/
2616 if (ext4_has_feature_64bit(sb) &&
2617 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2618 crc = crc16(crc, (__u8 *)gdp + offset,
2619 le16_to_cpu(sbi->s_es->s_desc_size) -
2623 return cpu_to_le16(crc);
2626 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2627 struct ext4_group_desc *gdp)
2629 if (ext4_has_group_desc_csum(sb) &&
2630 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
2636 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2637 struct ext4_group_desc *gdp)
2639 if (!ext4_has_group_desc_csum(sb))
2641 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
2644 /* Called at mount-time, super-block is locked */
2645 static int ext4_check_descriptors(struct super_block *sb,
2646 ext4_fsblk_t sb_block,
2647 ext4_group_t *first_not_zeroed)
2649 struct ext4_sb_info *sbi = EXT4_SB(sb);
2650 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2651 ext4_fsblk_t last_block;
2652 ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0);
2653 ext4_fsblk_t block_bitmap;
2654 ext4_fsblk_t inode_bitmap;
2655 ext4_fsblk_t inode_table;
2656 int flexbg_flag = 0;
2657 ext4_group_t i, grp = sbi->s_groups_count;
2659 if (ext4_has_feature_flex_bg(sb))
2662 ext4_debug("Checking group descriptors");
2664 for (i = 0; i < sbi->s_groups_count; i++) {
2665 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2667 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2668 last_block = ext4_blocks_count(sbi->s_es) - 1;
2670 last_block = first_block +
2671 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2673 if ((grp == sbi->s_groups_count) &&
2674 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2677 block_bitmap = ext4_block_bitmap(sb, gdp);
2678 if (block_bitmap == sb_block) {
2679 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2680 "Block bitmap for group %u overlaps "
2685 if (block_bitmap >= sb_block + 1 &&
2686 block_bitmap <= last_bg_block) {
2687 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2688 "Block bitmap for group %u overlaps "
2689 "block group descriptors", i);
2693 if (block_bitmap < first_block || block_bitmap > last_block) {
2694 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2695 "Block bitmap for group %u not in group "
2696 "(block %llu)!", i, block_bitmap);
2699 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2700 if (inode_bitmap == sb_block) {
2701 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2702 "Inode bitmap for group %u overlaps "
2707 if (inode_bitmap >= sb_block + 1 &&
2708 inode_bitmap <= last_bg_block) {
2709 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2710 "Inode bitmap for group %u overlaps "
2711 "block group descriptors", i);
2715 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2716 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2717 "Inode bitmap for group %u not in group "
2718 "(block %llu)!", i, inode_bitmap);
2721 inode_table = ext4_inode_table(sb, gdp);
2722 if (inode_table == sb_block) {
2723 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2724 "Inode table for group %u overlaps "
2729 if (inode_table >= sb_block + 1 &&
2730 inode_table <= last_bg_block) {
2731 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2732 "Inode table for group %u overlaps "
2733 "block group descriptors", i);
2737 if (inode_table < first_block ||
2738 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2739 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2740 "Inode table for group %u not in group "
2741 "(block %llu)!", i, inode_table);
2744 ext4_lock_group(sb, i);
2745 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2746 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2747 "Checksum for group %u failed (%u!=%u)",
2748 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
2749 gdp)), le16_to_cpu(gdp->bg_checksum));
2750 if (!sb_rdonly(sb)) {
2751 ext4_unlock_group(sb, i);
2755 ext4_unlock_group(sb, i);
2757 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2759 if (NULL != first_not_zeroed)
2760 *first_not_zeroed = grp;
2764 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2765 * the superblock) which were deleted from all directories, but held open by
2766 * a process at the time of a crash. We walk the list and try to delete these
2767 * inodes at recovery time (only with a read-write filesystem).
2769 * In order to keep the orphan inode chain consistent during traversal (in
2770 * case of crash during recovery), we link each inode into the superblock
2771 * orphan list_head and handle it the same way as an inode deletion during
2772 * normal operation (which journals the operations for us).
2774 * We only do an iget() and an iput() on each inode, which is very safe if we
2775 * accidentally point at an in-use or already deleted inode. The worst that
2776 * can happen in this case is that we get a "bit already cleared" message from
2777 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2778 * e2fsck was run on this filesystem, and it must have already done the orphan
2779 * inode cleanup for us, so we can safely abort without any further action.
2781 static void ext4_orphan_cleanup(struct super_block *sb,
2782 struct ext4_super_block *es)
2784 unsigned int s_flags = sb->s_flags;
2785 int ret, nr_orphans = 0, nr_truncates = 0;
2787 int quota_update = 0;
2790 if (!es->s_last_orphan) {
2791 jbd_debug(4, "no orphan inodes to clean up\n");
2795 if (bdev_read_only(sb->s_bdev)) {
2796 ext4_msg(sb, KERN_ERR, "write access "
2797 "unavailable, skipping orphan cleanup");
2801 /* Check if feature set would not allow a r/w mount */
2802 if (!ext4_feature_set_ok(sb, 0)) {
2803 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2804 "unknown ROCOMPAT features");
2808 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2809 /* don't clear list on RO mount w/ errors */
2810 if (es->s_last_orphan && !(s_flags & SB_RDONLY)) {
2811 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
2812 "clearing orphan list.\n");
2813 es->s_last_orphan = 0;
2815 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2819 if (s_flags & SB_RDONLY) {
2820 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2821 sb->s_flags &= ~SB_RDONLY;
2824 /* Needed for iput() to work correctly and not trash data */
2825 sb->s_flags |= SB_ACTIVE;
2828 * Turn on quotas which were not enabled for read-only mounts if
2829 * filesystem has quota feature, so that they are updated correctly.
2831 if (ext4_has_feature_quota(sb) && (s_flags & SB_RDONLY)) {
2832 int ret = ext4_enable_quotas(sb);
2837 ext4_msg(sb, KERN_ERR,
2838 "Cannot turn on quotas: error %d", ret);
2841 /* Turn on journaled quotas used for old sytle */
2842 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2843 if (EXT4_SB(sb)->s_qf_names[i]) {
2844 int ret = ext4_quota_on_mount(sb, i);
2849 ext4_msg(sb, KERN_ERR,
2850 "Cannot turn on journaled "
2851 "quota: type %d: error %d", i, ret);
2856 while (es->s_last_orphan) {
2857 struct inode *inode;
2860 * We may have encountered an error during cleanup; if
2861 * so, skip the rest.
2863 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2864 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2865 es->s_last_orphan = 0;
2869 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2870 if (IS_ERR(inode)) {
2871 es->s_last_orphan = 0;
2875 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2876 dquot_initialize(inode);
2877 if (inode->i_nlink) {
2878 if (test_opt(sb, DEBUG))
2879 ext4_msg(sb, KERN_DEBUG,
2880 "%s: truncating inode %lu to %lld bytes",
2881 __func__, inode->i_ino, inode->i_size);
2882 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2883 inode->i_ino, inode->i_size);
2885 truncate_inode_pages(inode->i_mapping, inode->i_size);
2886 ret = ext4_truncate(inode);
2888 ext4_std_error(inode->i_sb, ret);
2889 inode_unlock(inode);
2892 if (test_opt(sb, DEBUG))
2893 ext4_msg(sb, KERN_DEBUG,
2894 "%s: deleting unreferenced inode %lu",
2895 __func__, inode->i_ino);
2896 jbd_debug(2, "deleting unreferenced inode %lu\n",
2900 iput(inode); /* The delete magic happens here! */
2903 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2906 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2907 PLURAL(nr_orphans));
2909 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2910 PLURAL(nr_truncates));
2912 /* Turn off quotas if they were enabled for orphan cleanup */
2914 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2915 if (sb_dqopt(sb)->files[i])
2916 dquot_quota_off(sb, i);
2920 sb->s_flags = s_flags; /* Restore SB_RDONLY status */
2924 * Maximal extent format file size.
2925 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2926 * extent format containers, within a sector_t, and within i_blocks
2927 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2928 * so that won't be a limiting factor.
2930 * However there is other limiting factor. We do store extents in the form
2931 * of starting block and length, hence the resulting length of the extent
2932 * covering maximum file size must fit into on-disk format containers as
2933 * well. Given that length is always by 1 unit bigger than max unit (because
2934 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2936 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2938 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2941 loff_t upper_limit = MAX_LFS_FILESIZE;
2943 BUILD_BUG_ON(sizeof(blkcnt_t) < sizeof(u64));
2945 if (!has_huge_files) {
2946 upper_limit = (1LL << 32) - 1;
2948 /* total blocks in file system block size */
2949 upper_limit >>= (blkbits - 9);
2950 upper_limit <<= blkbits;
2954 * 32-bit extent-start container, ee_block. We lower the maxbytes
2955 * by one fs block, so ee_len can cover the extent of maximum file
2958 res = (1LL << 32) - 1;
2961 /* Sanity check against vm- & vfs- imposed limits */
2962 if (res > upper_limit)
2969 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2970 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2971 * We need to be 1 filesystem block less than the 2^48 sector limit.
2973 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2975 loff_t res = EXT4_NDIR_BLOCKS;
2978 /* This is calculated to be the largest file size for a dense, block
2979 * mapped file such that the file's total number of 512-byte sectors,
2980 * including data and all indirect blocks, does not exceed (2^48 - 1).
2982 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2983 * number of 512-byte sectors of the file.
2986 if (!has_huge_files) {
2988 * !has_huge_files or implies that the inode i_block field
2989 * represents total file blocks in 2^32 512-byte sectors ==
2990 * size of vfs inode i_blocks * 8
2992 upper_limit = (1LL << 32) - 1;
2994 /* total blocks in file system block size */
2995 upper_limit >>= (bits - 9);
2999 * We use 48 bit ext4_inode i_blocks
3000 * With EXT4_HUGE_FILE_FL set the i_blocks
3001 * represent total number of blocks in
3002 * file system block size
3004 upper_limit = (1LL << 48) - 1;
3008 /* indirect blocks */
3010 /* double indirect blocks */
3011 meta_blocks += 1 + (1LL << (bits-2));
3012 /* tripple indirect blocks */
3013 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
3015 upper_limit -= meta_blocks;
3016 upper_limit <<= bits;
3018 res += 1LL << (bits-2);
3019 res += 1LL << (2*(bits-2));
3020 res += 1LL << (3*(bits-2));
3022 if (res > upper_limit)
3025 if (res > MAX_LFS_FILESIZE)
3026 res = MAX_LFS_FILESIZE;
3031 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
3032 ext4_fsblk_t logical_sb_block, int nr)
3034 struct ext4_sb_info *sbi = EXT4_SB(sb);
3035 ext4_group_t bg, first_meta_bg;
3038 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
3040 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
3041 return logical_sb_block + nr + 1;
3042 bg = sbi->s_desc_per_block * nr;
3043 if (ext4_bg_has_super(sb, bg))
3047 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
3048 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
3049 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
3052 if (sb->s_blocksize == 1024 && nr == 0 &&
3053 le32_to_cpu(sbi->s_es->s_first_data_block) == 0)
3056 return (has_super + ext4_group_first_block_no(sb, bg));
3060 * ext4_get_stripe_size: Get the stripe size.
3061 * @sbi: In memory super block info
3063 * If we have specified it via mount option, then
3064 * use the mount option value. If the value specified at mount time is
3065 * greater than the blocks per group use the super block value.
3066 * If the super block value is greater than blocks per group return 0.
3067 * Allocator needs it be less than blocks per group.
3070 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
3072 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
3073 unsigned long stripe_width =
3074 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3077 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3078 ret = sbi->s_stripe;
3079 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
3081 else if (stride && stride <= sbi->s_blocks_per_group)
3087 * If the stripe width is 1, this makes no sense and
3088 * we set it to 0 to turn off stripe handling code.
3097 * Check whether this filesystem can be mounted based on
3098 * the features present and the RDONLY/RDWR mount requested.
3099 * Returns 1 if this filesystem can be mounted as requested,
3100 * 0 if it cannot be.
3102 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
3104 if (ext4_has_unknown_ext4_incompat_features(sb)) {
3105 ext4_msg(sb, KERN_ERR,
3106 "Couldn't mount because of "
3107 "unsupported optional features (%x)",
3108 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
3109 ~EXT4_FEATURE_INCOMPAT_SUPP));
3113 #ifndef CONFIG_UNICODE
3114 if (ext4_has_feature_casefold(sb)) {
3115 ext4_msg(sb, KERN_ERR,
3116 "Filesystem with casefold feature cannot be "
3117 "mounted without CONFIG_UNICODE");
3125 if (ext4_has_feature_readonly(sb)) {
3126 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
3127 sb->s_flags |= SB_RDONLY;
3131 /* Check that feature set is OK for a read-write mount */
3132 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
3133 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
3134 "unsupported optional features (%x)",
3135 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
3136 ~EXT4_FEATURE_RO_COMPAT_SUPP));
3139 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
3140 ext4_msg(sb, KERN_ERR,
3141 "Can't support bigalloc feature without "
3142 "extents feature\n");
3146 #if !IS_ENABLED(CONFIG_QUOTA) || !IS_ENABLED(CONFIG_QFMT_V2)
3147 if (!readonly && (ext4_has_feature_quota(sb) ||
3148 ext4_has_feature_project(sb))) {
3149 ext4_msg(sb, KERN_ERR,
3150 "The kernel was not built with CONFIG_QUOTA and CONFIG_QFMT_V2");
3153 #endif /* CONFIG_QUOTA */
3158 * This function is called once a day if we have errors logged
3159 * on the file system
3161 static void print_daily_error_info(struct timer_list *t)
3163 struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report);
3164 struct super_block *sb = sbi->s_sb;
3165 struct ext4_super_block *es = sbi->s_es;
3167 if (es->s_error_count)
3168 /* fsck newer than v1.41.13 is needed to clean this condition. */
3169 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
3170 le32_to_cpu(es->s_error_count));
3171 if (es->s_first_error_time) {
3172 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %llu: %.*s:%d",
3174 ext4_get_tstamp(es, s_first_error_time),
3175 (int) sizeof(es->s_first_error_func),
3176 es->s_first_error_func,
3177 le32_to_cpu(es->s_first_error_line));
3178 if (es->s_first_error_ino)
3179 printk(KERN_CONT ": inode %u",
3180 le32_to_cpu(es->s_first_error_ino));
3181 if (es->s_first_error_block)
3182 printk(KERN_CONT ": block %llu", (unsigned long long)
3183 le64_to_cpu(es->s_first_error_block));
3184 printk(KERN_CONT "\n");
3186 if (es->s_last_error_time) {
3187 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %llu: %.*s:%d",
3189 ext4_get_tstamp(es, s_last_error_time),
3190 (int) sizeof(es->s_last_error_func),
3191 es->s_last_error_func,
3192 le32_to_cpu(es->s_last_error_line));
3193 if (es->s_last_error_ino)
3194 printk(KERN_CONT ": inode %u",
3195 le32_to_cpu(es->s_last_error_ino));
3196 if (es->s_last_error_block)
3197 printk(KERN_CONT ": block %llu", (unsigned long long)
3198 le64_to_cpu(es->s_last_error_block));
3199 printk(KERN_CONT "\n");
3201 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
3204 /* Find next suitable group and run ext4_init_inode_table */
3205 static int ext4_run_li_request(struct ext4_li_request *elr)
3207 struct ext4_group_desc *gdp = NULL;
3208 struct super_block *sb = elr->lr_super;
3209 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3210 ext4_group_t group = elr->lr_next_group;
3211 unsigned long timeout = 0;
3212 unsigned int prefetch_ios = 0;
3215 if (elr->lr_mode == EXT4_LI_MODE_PREFETCH_BBITMAP) {
3216 elr->lr_next_group = ext4_mb_prefetch(sb, group,
3217 EXT4_SB(sb)->s_mb_prefetch, &prefetch_ios);
3219 ext4_mb_prefetch_fini(sb, elr->lr_next_group,
3221 trace_ext4_prefetch_bitmaps(sb, group, elr->lr_next_group,
3223 if (group >= elr->lr_next_group) {
3225 if (elr->lr_first_not_zeroed != ngroups &&
3226 !sb_rdonly(sb) && test_opt(sb, INIT_INODE_TABLE)) {
3227 elr->lr_next_group = elr->lr_first_not_zeroed;
3228 elr->lr_mode = EXT4_LI_MODE_ITABLE;
3235 for (; group < ngroups; group++) {
3236 gdp = ext4_get_group_desc(sb, group, NULL);
3242 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3246 if (group >= ngroups)
3251 ret = ext4_init_inode_table(sb, group,
3252 elr->lr_timeout ? 0 : 1);
3253 trace_ext4_lazy_itable_init(sb, group);
3254 if (elr->lr_timeout == 0) {
3255 timeout = (jiffies - timeout) *
3256 EXT4_SB(elr->lr_super)->s_li_wait_mult;
3257 elr->lr_timeout = timeout;
3259 elr->lr_next_sched = jiffies + elr->lr_timeout;
3260 elr->lr_next_group = group + 1;
3266 * Remove lr_request from the list_request and free the
3267 * request structure. Should be called with li_list_mtx held
3269 static void ext4_remove_li_request(struct ext4_li_request *elr)
3274 list_del(&elr->lr_request);
3275 EXT4_SB(elr->lr_super)->s_li_request = NULL;
3279 static void ext4_unregister_li_request(struct super_block *sb)
3281 mutex_lock(&ext4_li_mtx);
3282 if (!ext4_li_info) {
3283 mutex_unlock(&ext4_li_mtx);
3287 mutex_lock(&ext4_li_info->li_list_mtx);
3288 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
3289 mutex_unlock(&ext4_li_info->li_list_mtx);
3290 mutex_unlock(&ext4_li_mtx);
3293 static struct task_struct *ext4_lazyinit_task;
3296 * This is the function where ext4lazyinit thread lives. It walks
3297 * through the request list searching for next scheduled filesystem.
3298 * When such a fs is found, run the lazy initialization request
3299 * (ext4_rn_li_request) and keep track of the time spend in this
3300 * function. Based on that time we compute next schedule time of
3301 * the request. When walking through the list is complete, compute
3302 * next waking time and put itself into sleep.
3304 static int ext4_lazyinit_thread(void *arg)
3306 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
3307 struct list_head *pos, *n;
3308 struct ext4_li_request *elr;
3309 unsigned long next_wakeup, cur;
3311 BUG_ON(NULL == eli);
3315 next_wakeup = MAX_JIFFY_OFFSET;
3317 mutex_lock(&eli->li_list_mtx);
3318 if (list_empty(&eli->li_request_list)) {
3319 mutex_unlock(&eli->li_list_mtx);
3322 list_for_each_safe(pos, n, &eli->li_request_list) {
3325 elr = list_entry(pos, struct ext4_li_request,
3328 if (time_before(jiffies, elr->lr_next_sched)) {
3329 if (time_before(elr->lr_next_sched, next_wakeup))
3330 next_wakeup = elr->lr_next_sched;
3333 if (down_read_trylock(&elr->lr_super->s_umount)) {
3334 if (sb_start_write_trylock(elr->lr_super)) {
3337 * We hold sb->s_umount, sb can not
3338 * be removed from the list, it is
3339 * now safe to drop li_list_mtx
3341 mutex_unlock(&eli->li_list_mtx);
3342 err = ext4_run_li_request(elr);
3343 sb_end_write(elr->lr_super);
3344 mutex_lock(&eli->li_list_mtx);
3347 up_read((&elr->lr_super->s_umount));
3349 /* error, remove the lazy_init job */
3351 ext4_remove_li_request(elr);
3355 elr->lr_next_sched = jiffies +
3357 % (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3359 if (time_before(elr->lr_next_sched, next_wakeup))
3360 next_wakeup = elr->lr_next_sched;
3362 mutex_unlock(&eli->li_list_mtx);
3367 if ((time_after_eq(cur, next_wakeup)) ||
3368 (MAX_JIFFY_OFFSET == next_wakeup)) {
3373 schedule_timeout_interruptible(next_wakeup - cur);
3375 if (kthread_should_stop()) {
3376 ext4_clear_request_list();
3383 * It looks like the request list is empty, but we need
3384 * to check it under the li_list_mtx lock, to prevent any
3385 * additions into it, and of course we should lock ext4_li_mtx
3386 * to atomically free the list and ext4_li_info, because at
3387 * this point another ext4 filesystem could be registering
3390 mutex_lock(&ext4_li_mtx);
3391 mutex_lock(&eli->li_list_mtx);
3392 if (!list_empty(&eli->li_request_list)) {
3393 mutex_unlock(&eli->li_list_mtx);
3394 mutex_unlock(&ext4_li_mtx);
3397 mutex_unlock(&eli->li_list_mtx);
3398 kfree(ext4_li_info);
3399 ext4_li_info = NULL;
3400 mutex_unlock(&ext4_li_mtx);
3405 static void ext4_clear_request_list(void)
3407 struct list_head *pos, *n;
3408 struct ext4_li_request *elr;
3410 mutex_lock(&ext4_li_info->li_list_mtx);
3411 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3412 elr = list_entry(pos, struct ext4_li_request,
3414 ext4_remove_li_request(elr);
3416 mutex_unlock(&ext4_li_info->li_list_mtx);
3419 static int ext4_run_lazyinit_thread(void)
3421 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3422 ext4_li_info, "ext4lazyinit");
3423 if (IS_ERR(ext4_lazyinit_task)) {
3424 int err = PTR_ERR(ext4_lazyinit_task);
3425 ext4_clear_request_list();
3426 kfree(ext4_li_info);
3427 ext4_li_info = NULL;
3428 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3429 "initialization thread\n",
3433 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3438 * Check whether it make sense to run itable init. thread or not.
3439 * If there is at least one uninitialized inode table, return
3440 * corresponding group number, else the loop goes through all
3441 * groups and return total number of groups.
3443 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3445 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3446 struct ext4_group_desc *gdp = NULL;
3448 if (!ext4_has_group_desc_csum(sb))
3451 for (group = 0; group < ngroups; group++) {
3452 gdp = ext4_get_group_desc(sb, group, NULL);
3456 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3463 static int ext4_li_info_new(void)
3465 struct ext4_lazy_init *eli = NULL;
3467 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3471 INIT_LIST_HEAD(&eli->li_request_list);
3472 mutex_init(&eli->li_list_mtx);
3474 eli->li_state |= EXT4_LAZYINIT_QUIT;
3481 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3484 struct ext4_li_request *elr;
3486 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3491 elr->lr_first_not_zeroed = start;
3492 if (test_opt(sb, PREFETCH_BLOCK_BITMAPS))
3493 elr->lr_mode = EXT4_LI_MODE_PREFETCH_BBITMAP;
3495 elr->lr_mode = EXT4_LI_MODE_ITABLE;
3496 elr->lr_next_group = start;
3500 * Randomize first schedule time of the request to
3501 * spread the inode table initialization requests
3504 elr->lr_next_sched = jiffies + (prandom_u32() %
3505 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3509 int ext4_register_li_request(struct super_block *sb,
3510 ext4_group_t first_not_zeroed)
3512 struct ext4_sb_info *sbi = EXT4_SB(sb);
3513 struct ext4_li_request *elr = NULL;
3514 ext4_group_t ngroups = sbi->s_groups_count;
3517 mutex_lock(&ext4_li_mtx);
3518 if (sbi->s_li_request != NULL) {
3520 * Reset timeout so it can be computed again, because
3521 * s_li_wait_mult might have changed.
3523 sbi->s_li_request->lr_timeout = 0;
3527 if (!test_opt(sb, PREFETCH_BLOCK_BITMAPS) &&
3528 (first_not_zeroed == ngroups || sb_rdonly(sb) ||
3529 !test_opt(sb, INIT_INODE_TABLE)))
3532 elr = ext4_li_request_new(sb, first_not_zeroed);
3538 if (NULL == ext4_li_info) {
3539 ret = ext4_li_info_new();
3544 mutex_lock(&ext4_li_info->li_list_mtx);
3545 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3546 mutex_unlock(&ext4_li_info->li_list_mtx);
3548 sbi->s_li_request = elr;
3550 * set elr to NULL here since it has been inserted to
3551 * the request_list and the removal and free of it is
3552 * handled by ext4_clear_request_list from now on.
3556 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3557 ret = ext4_run_lazyinit_thread();
3562 mutex_unlock(&ext4_li_mtx);
3569 * We do not need to lock anything since this is called on
3572 static void ext4_destroy_lazyinit_thread(void)
3575 * If thread exited earlier
3576 * there's nothing to be done.
3578 if (!ext4_li_info || !ext4_lazyinit_task)
3581 kthread_stop(ext4_lazyinit_task);
3584 static int set_journal_csum_feature_set(struct super_block *sb)
3587 int compat, incompat;
3588 struct ext4_sb_info *sbi = EXT4_SB(sb);
3590 if (ext4_has_metadata_csum(sb)) {
3591 /* journal checksum v3 */
3593 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
3595 /* journal checksum v1 */
3596 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3600 jbd2_journal_clear_features(sbi->s_journal,
3601 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3602 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3603 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3604 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3605 ret = jbd2_journal_set_features(sbi->s_journal,
3607 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3609 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3610 ret = jbd2_journal_set_features(sbi->s_journal,
3613 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3614 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3616 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3617 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3624 * Note: calculating the overhead so we can be compatible with
3625 * historical BSD practice is quite difficult in the face of
3626 * clusters/bigalloc. This is because multiple metadata blocks from
3627 * different block group can end up in the same allocation cluster.
3628 * Calculating the exact overhead in the face of clustered allocation
3629 * requires either O(all block bitmaps) in memory or O(number of block
3630 * groups**2) in time. We will still calculate the superblock for
3631 * older file systems --- and if we come across with a bigalloc file
3632 * system with zero in s_overhead_clusters the estimate will be close to
3633 * correct especially for very large cluster sizes --- but for newer
3634 * file systems, it's better to calculate this figure once at mkfs
3635 * time, and store it in the superblock. If the superblock value is
3636 * present (even for non-bigalloc file systems), we will use it.
3638 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3641 struct ext4_sb_info *sbi = EXT4_SB(sb);
3642 struct ext4_group_desc *gdp;
3643 ext4_fsblk_t first_block, last_block, b;
3644 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3645 int s, j, count = 0;
3647 if (!ext4_has_feature_bigalloc(sb))
3648 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3649 sbi->s_itb_per_group + 2);
3651 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3652 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3653 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3654 for (i = 0; i < ngroups; i++) {
3655 gdp = ext4_get_group_desc(sb, i, NULL);
3656 b = ext4_block_bitmap(sb, gdp);
3657 if (b >= first_block && b <= last_block) {
3658 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3661 b = ext4_inode_bitmap(sb, gdp);
3662 if (b >= first_block && b <= last_block) {
3663 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3666 b = ext4_inode_table(sb, gdp);
3667 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3668 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3669 int c = EXT4_B2C(sbi, b - first_block);
3670 ext4_set_bit(c, buf);
3676 if (ext4_bg_has_super(sb, grp)) {
3677 ext4_set_bit(s++, buf);
3680 j = ext4_bg_num_gdb(sb, grp);
3681 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
3682 ext4_error(sb, "Invalid number of block group "
3683 "descriptor blocks: %d", j);
3684 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
3688 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3692 return EXT4_CLUSTERS_PER_GROUP(sb) -
3693 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3697 * Compute the overhead and stash it in sbi->s_overhead
3699 int ext4_calculate_overhead(struct super_block *sb)
3701 struct ext4_sb_info *sbi = EXT4_SB(sb);
3702 struct ext4_super_block *es = sbi->s_es;
3703 struct inode *j_inode;
3704 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
3705 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3706 ext4_fsblk_t overhead = 0;
3707 char *buf = (char *) get_zeroed_page(GFP_NOFS);
3713 * Compute the overhead (FS structures). This is constant
3714 * for a given filesystem unless the number of block groups
3715 * changes so we cache the previous value until it does.
3719 * All of the blocks before first_data_block are overhead
3721 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3724 * Add the overhead found in each block group
3726 for (i = 0; i < ngroups; i++) {
3729 blks = count_overhead(sb, i, buf);
3732 memset(buf, 0, PAGE_SIZE);
3737 * Add the internal journal blocks whether the journal has been
3740 if (sbi->s_journal && !sbi->journal_bdev)
3741 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3742 else if (ext4_has_feature_journal(sb) && !sbi->s_journal && j_inum) {
3743 /* j_inum for internal journal is non-zero */
3744 j_inode = ext4_get_journal_inode(sb, j_inum);
3746 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
3747 overhead += EXT4_NUM_B2C(sbi, j_blocks);
3750 ext4_msg(sb, KERN_ERR, "can't get journal size");
3753 sbi->s_overhead = overhead;
3755 free_page((unsigned long) buf);
3759 static void ext4_set_resv_clusters(struct super_block *sb)
3761 ext4_fsblk_t resv_clusters;
3762 struct ext4_sb_info *sbi = EXT4_SB(sb);
3765 * There's no need to reserve anything when we aren't using extents.
3766 * The space estimates are exact, there are no unwritten extents,
3767 * hole punching doesn't need new metadata... This is needed especially
3768 * to keep ext2/3 backward compatibility.
3770 if (!ext4_has_feature_extents(sb))
3773 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3774 * This should cover the situations where we can not afford to run
3775 * out of space like for example punch hole, or converting
3776 * unwritten extents in delalloc path. In most cases such
3777 * allocation would require 1, or 2 blocks, higher numbers are
3780 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3781 sbi->s_cluster_bits);
3783 do_div(resv_clusters, 50);
3784 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3786 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
3789 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3791 struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev);
3792 char *orig_data = kstrdup(data, GFP_KERNEL);
3793 struct buffer_head *bh, **group_desc;
3794 struct ext4_super_block *es = NULL;
3795 struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3796 struct flex_groups **flex_groups;
3798 ext4_fsblk_t sb_block = get_sb_block(&data);
3799 ext4_fsblk_t logical_sb_block;
3800 unsigned long offset = 0;
3801 unsigned long journal_devnum = 0;
3802 unsigned long def_mount_opts;
3806 int blocksize, clustersize;
3807 unsigned int db_count;
3809 int needs_recovery, has_huge_files;
3812 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3813 ext4_group_t first_not_zeroed;
3815 if ((data && !orig_data) || !sbi)
3818 sbi->s_daxdev = dax_dev;
3819 sbi->s_blockgroup_lock =
3820 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3821 if (!sbi->s_blockgroup_lock)
3824 sb->s_fs_info = sbi;
3826 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3827 sbi->s_sb_block = sb_block;
3828 if (sb->s_bdev->bd_part)
3829 sbi->s_sectors_written_start =
3830 part_stat_read(sb->s_bdev->bd_part, sectors[STAT_WRITE]);
3832 /* Cleanup superblock name */
3833 strreplace(sb->s_id, '/', '!');
3835 /* -EINVAL is default */
3837 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3839 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3844 * The ext4 superblock will not be buffer aligned for other than 1kB
3845 * block sizes. We need to calculate the offset from buffer start.
3847 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3848 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3849 offset = do_div(logical_sb_block, blocksize);
3851 logical_sb_block = sb_block;
3854 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
3855 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3859 * Note: s_es must be initialized as soon as possible because
3860 * some ext4 macro-instructions depend on its value
3862 es = (struct ext4_super_block *) (bh->b_data + offset);
3864 sb->s_magic = le16_to_cpu(es->s_magic);
3865 if (sb->s_magic != EXT4_SUPER_MAGIC)
3867 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3869 /* Warn if metadata_csum and gdt_csum are both set. */
3870 if (ext4_has_feature_metadata_csum(sb) &&
3871 ext4_has_feature_gdt_csum(sb))
3872 ext4_warning(sb, "metadata_csum and uninit_bg are "
3873 "redundant flags; please run fsck.");
3875 /* Check for a known checksum algorithm */
3876 if (!ext4_verify_csum_type(sb, es)) {
3877 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3878 "unknown checksum algorithm.");
3883 /* Load the checksum driver */
3884 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3885 if (IS_ERR(sbi->s_chksum_driver)) {
3886 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3887 ret = PTR_ERR(sbi->s_chksum_driver);
3888 sbi->s_chksum_driver = NULL;
3892 /* Check superblock checksum */
3893 if (!ext4_superblock_csum_verify(sb, es)) {
3894 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3895 "invalid superblock checksum. Run e2fsck?");
3901 /* Precompute checksum seed for all metadata */
3902 if (ext4_has_feature_csum_seed(sb))
3903 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
3904 else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb))
3905 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3906 sizeof(es->s_uuid));
3908 /* Set defaults before we parse the mount options */
3909 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3910 set_opt(sb, INIT_INODE_TABLE);
3911 if (def_mount_opts & EXT4_DEFM_DEBUG)
3913 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3915 if (def_mount_opts & EXT4_DEFM_UID16)
3916 set_opt(sb, NO_UID32);
3917 /* xattr user namespace & acls are now defaulted on */
3918 set_opt(sb, XATTR_USER);
3919 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3920 set_opt(sb, POSIX_ACL);
3922 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3923 if (ext4_has_metadata_csum(sb))
3924 set_opt(sb, JOURNAL_CHECKSUM);
3926 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3927 set_opt(sb, JOURNAL_DATA);
3928 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3929 set_opt(sb, ORDERED_DATA);
3930 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3931 set_opt(sb, WRITEBACK_DATA);
3933 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3934 set_opt(sb, ERRORS_PANIC);
3935 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3936 set_opt(sb, ERRORS_CONT);
3938 set_opt(sb, ERRORS_RO);
3939 /* block_validity enabled by default; disable with noblock_validity */
3940 set_opt(sb, BLOCK_VALIDITY);
3941 if (def_mount_opts & EXT4_DEFM_DISCARD)
3942 set_opt(sb, DISCARD);
3944 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3945 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3946 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3947 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3948 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3950 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3951 set_opt(sb, BARRIER);
3954 * enable delayed allocation by default
3955 * Use -o nodelalloc to turn it off
3957 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3958 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3959 set_opt(sb, DELALLOC);
3962 * set default s_li_wait_mult for lazyinit, for the case there is
3963 * no mount option specified.
3965 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3967 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3969 if (blocksize == PAGE_SIZE)
3970 set_opt(sb, DIOREAD_NOLOCK);
3972 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3973 blocksize > EXT4_MAX_BLOCK_SIZE) {
3974 ext4_msg(sb, KERN_ERR,
3975 "Unsupported filesystem blocksize %d (%d log_block_size)",
3976 blocksize, le32_to_cpu(es->s_log_block_size));
3980 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3981 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3982 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3984 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3985 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3986 if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) {
3987 ext4_msg(sb, KERN_ERR, "invalid first ino: %u",
3991 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3992 (!is_power_of_2(sbi->s_inode_size)) ||
3993 (sbi->s_inode_size > blocksize)) {
3994 ext4_msg(sb, KERN_ERR,
3995 "unsupported inode size: %d",
3997 ext4_msg(sb, KERN_ERR, "blocksize: %d", blocksize);
4001 * i_atime_extra is the last extra field available for
4002 * [acm]times in struct ext4_inode. Checking for that
4003 * field should suffice to ensure we have extra space
4006 if (sbi->s_inode_size >= offsetof(struct ext4_inode, i_atime_extra) +
4007 sizeof(((struct ext4_inode *)0)->i_atime_extra)) {
4008 sb->s_time_gran = 1;
4009 sb->s_time_max = EXT4_EXTRA_TIMESTAMP_MAX;
4011 sb->s_time_gran = NSEC_PER_SEC;
4012 sb->s_time_max = EXT4_NON_EXTRA_TIMESTAMP_MAX;
4014 sb->s_time_min = EXT4_TIMESTAMP_MIN;
4016 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4017 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4018 EXT4_GOOD_OLD_INODE_SIZE;
4019 if (ext4_has_feature_extra_isize(sb)) {
4020 unsigned v, max = (sbi->s_inode_size -
4021 EXT4_GOOD_OLD_INODE_SIZE);
4023 v = le16_to_cpu(es->s_want_extra_isize);
4025 ext4_msg(sb, KERN_ERR,
4026 "bad s_want_extra_isize: %d", v);
4029 if (sbi->s_want_extra_isize < v)
4030 sbi->s_want_extra_isize = v;
4032 v = le16_to_cpu(es->s_min_extra_isize);
4034 ext4_msg(sb, KERN_ERR,
4035 "bad s_min_extra_isize: %d", v);
4038 if (sbi->s_want_extra_isize < v)
4039 sbi->s_want_extra_isize = v;
4043 if (sbi->s_es->s_mount_opts[0]) {
4044 char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
4045 sizeof(sbi->s_es->s_mount_opts),
4049 if (!parse_options(s_mount_opts, sb, &journal_devnum,
4050 &journal_ioprio, 0)) {
4051 ext4_msg(sb, KERN_WARNING,
4052 "failed to parse options in superblock: %s",
4055 kfree(s_mount_opts);
4057 sbi->s_def_mount_opt = sbi->s_mount_opt;
4058 if (!parse_options((char *) data, sb, &journal_devnum,
4059 &journal_ioprio, 0))
4062 #ifdef CONFIG_UNICODE
4063 if (ext4_has_feature_casefold(sb) && !sbi->s_encoding) {
4064 const struct ext4_sb_encodings *encoding_info;
4065 struct unicode_map *encoding;
4066 __u16 encoding_flags;
4068 if (ext4_has_feature_encrypt(sb)) {
4069 ext4_msg(sb, KERN_ERR,
4070 "Can't mount with encoding and encryption");
4074 if (ext4_sb_read_encoding(es, &encoding_info,
4076 ext4_msg(sb, KERN_ERR,
4077 "Encoding requested by superblock is unknown");
4081 encoding = utf8_load(encoding_info->version);
4082 if (IS_ERR(encoding)) {
4083 ext4_msg(sb, KERN_ERR,
4084 "can't mount with superblock charset: %s-%s "
4085 "not supported by the kernel. flags: 0x%x.",
4086 encoding_info->name, encoding_info->version,
4090 ext4_msg(sb, KERN_INFO,"Using encoding defined by superblock: "
4091 "%s-%s with flags 0x%hx", encoding_info->name,
4092 encoding_info->version?:"\b", encoding_flags);
4094 sbi->s_encoding = encoding;
4095 sbi->s_encoding_flags = encoding_flags;
4099 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4100 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting with data=journal disables delayed allocation, dioread_nolock, and O_DIRECT support!\n");
4101 /* can't mount with both data=journal and dioread_nolock. */
4102 clear_opt(sb, DIOREAD_NOLOCK);
4103 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4104 ext4_msg(sb, KERN_ERR, "can't mount with "
4105 "both data=journal and delalloc");
4108 if (test_opt(sb, DAX_ALWAYS)) {
4109 ext4_msg(sb, KERN_ERR, "can't mount with "
4110 "both data=journal and dax");
4113 if (ext4_has_feature_encrypt(sb)) {
4114 ext4_msg(sb, KERN_WARNING,
4115 "encrypted files will use data=ordered "
4116 "instead of data journaling mode");
4118 if (test_opt(sb, DELALLOC))
4119 clear_opt(sb, DELALLOC);
4121 sb->s_iflags |= SB_I_CGROUPWB;
4124 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
4125 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
4127 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
4128 (ext4_has_compat_features(sb) ||
4129 ext4_has_ro_compat_features(sb) ||
4130 ext4_has_incompat_features(sb)))
4131 ext4_msg(sb, KERN_WARNING,
4132 "feature flags set on rev 0 fs, "
4133 "running e2fsck is recommended");
4135 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
4136 set_opt2(sb, HURD_COMPAT);
4137 if (ext4_has_feature_64bit(sb)) {
4138 ext4_msg(sb, KERN_ERR,
4139 "The Hurd can't support 64-bit file systems");
4144 * ea_inode feature uses l_i_version field which is not
4145 * available in HURD_COMPAT mode.
4147 if (ext4_has_feature_ea_inode(sb)) {
4148 ext4_msg(sb, KERN_ERR,
4149 "ea_inode feature is not supported for Hurd");
4154 if (IS_EXT2_SB(sb)) {
4155 if (ext2_feature_set_ok(sb))
4156 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
4157 "using the ext4 subsystem");
4160 * If we're probing be silent, if this looks like
4161 * it's actually an ext[34] filesystem.
4163 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4165 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
4166 "to feature incompatibilities");
4171 if (IS_EXT3_SB(sb)) {
4172 if (ext3_feature_set_ok(sb))
4173 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
4174 "using the ext4 subsystem");
4177 * If we're probing be silent, if this looks like
4178 * it's actually an ext4 filesystem.
4180 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4182 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
4183 "to feature incompatibilities");
4189 * Check feature flags regardless of the revision level, since we
4190 * previously didn't change the revision level when setting the flags,
4191 * so there is a chance incompat flags are set on a rev 0 filesystem.
4193 if (!ext4_feature_set_ok(sb, (sb_rdonly(sb))))
4196 if (le32_to_cpu(es->s_log_block_size) >
4197 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
4198 ext4_msg(sb, KERN_ERR,
4199 "Invalid log block size: %u",
4200 le32_to_cpu(es->s_log_block_size));
4203 if (le32_to_cpu(es->s_log_cluster_size) >
4204 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
4205 ext4_msg(sb, KERN_ERR,
4206 "Invalid log cluster size: %u",
4207 le32_to_cpu(es->s_log_cluster_size));
4211 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
4212 ext4_msg(sb, KERN_ERR,
4213 "Number of reserved GDT blocks insanely large: %d",
4214 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
4218 if (bdev_dax_supported(sb->s_bdev, blocksize))
4219 set_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags);
4221 if (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) {
4222 if (ext4_has_feature_inline_data(sb)) {
4223 ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem"
4224 " that may contain inline data");
4227 if (!test_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags)) {
4228 ext4_msg(sb, KERN_ERR,
4229 "DAX unsupported by block device.");
4234 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
4235 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
4236 es->s_encryption_level);
4240 if (sb->s_blocksize != blocksize) {
4241 /* Validate the filesystem blocksize */
4242 if (!sb_set_blocksize(sb, blocksize)) {
4243 ext4_msg(sb, KERN_ERR, "bad block size %d",
4249 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
4250 offset = do_div(logical_sb_block, blocksize);
4251 bh = sb_bread_unmovable(sb, logical_sb_block);
4253 ext4_msg(sb, KERN_ERR,
4254 "Can't read superblock on 2nd try");
4257 es = (struct ext4_super_block *)(bh->b_data + offset);
4259 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
4260 ext4_msg(sb, KERN_ERR,
4261 "Magic mismatch, very weird!");
4266 has_huge_files = ext4_has_feature_huge_file(sb);
4267 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
4269 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
4271 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
4272 if (ext4_has_feature_64bit(sb)) {
4273 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
4274 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
4275 !is_power_of_2(sbi->s_desc_size)) {
4276 ext4_msg(sb, KERN_ERR,
4277 "unsupported descriptor size %lu",
4282 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
4284 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
4285 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
4287 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
4288 if (sbi->s_inodes_per_block == 0)
4290 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
4291 sbi->s_inodes_per_group > blocksize * 8) {
4292 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
4293 sbi->s_inodes_per_group);
4296 sbi->s_itb_per_group = sbi->s_inodes_per_group /
4297 sbi->s_inodes_per_block;
4298 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
4300 sbi->s_mount_state = le16_to_cpu(es->s_state);
4301 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
4302 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
4304 for (i = 0; i < 4; i++)
4305 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
4306 sbi->s_def_hash_version = es->s_def_hash_version;
4307 if (ext4_has_feature_dir_index(sb)) {
4308 i = le32_to_cpu(es->s_flags);
4309 if (i & EXT2_FLAGS_UNSIGNED_HASH)
4310 sbi->s_hash_unsigned = 3;
4311 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
4312 #ifdef __CHAR_UNSIGNED__
4315 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
4316 sbi->s_hash_unsigned = 3;
4320 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
4325 /* Handle clustersize */
4326 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
4327 if (ext4_has_feature_bigalloc(sb)) {
4328 if (clustersize < blocksize) {
4329 ext4_msg(sb, KERN_ERR,
4330 "cluster size (%d) smaller than "
4331 "block size (%d)", clustersize, blocksize);
4334 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
4335 le32_to_cpu(es->s_log_block_size);
4336 sbi->s_clusters_per_group =
4337 le32_to_cpu(es->s_clusters_per_group);
4338 if (sbi->s_clusters_per_group > blocksize * 8) {
4339 ext4_msg(sb, KERN_ERR,
4340 "#clusters per group too big: %lu",
4341 sbi->s_clusters_per_group);
4344 if (sbi->s_blocks_per_group !=
4345 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
4346 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
4347 "clusters per group (%lu) inconsistent",
4348 sbi->s_blocks_per_group,
4349 sbi->s_clusters_per_group);
4353 if (clustersize != blocksize) {
4354 ext4_msg(sb, KERN_ERR,
4355 "fragment/cluster size (%d) != "
4356 "block size (%d)", clustersize, blocksize);
4359 if (sbi->s_blocks_per_group > blocksize * 8) {
4360 ext4_msg(sb, KERN_ERR,
4361 "#blocks per group too big: %lu",
4362 sbi->s_blocks_per_group);
4365 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
4366 sbi->s_cluster_bits = 0;
4368 sbi->s_cluster_ratio = clustersize / blocksize;
4370 /* Do we have standard group size of clustersize * 8 blocks ? */
4371 if (sbi->s_blocks_per_group == clustersize << 3)
4372 set_opt2(sb, STD_GROUP_SIZE);
4375 * Test whether we have more sectors than will fit in sector_t,
4376 * and whether the max offset is addressable by the page cache.
4378 err = generic_check_addressable(sb->s_blocksize_bits,
4379 ext4_blocks_count(es));
4381 ext4_msg(sb, KERN_ERR, "filesystem"
4382 " too large to mount safely on this system");
4386 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
4389 /* check blocks count against device size */
4390 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
4391 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
4392 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
4393 "exceeds size of device (%llu blocks)",
4394 ext4_blocks_count(es), blocks_count);
4399 * It makes no sense for the first data block to be beyond the end
4400 * of the filesystem.
4402 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
4403 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4404 "block %u is beyond end of filesystem (%llu)",
4405 le32_to_cpu(es->s_first_data_block),
4406 ext4_blocks_count(es));
4409 if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) &&
4410 (sbi->s_cluster_ratio == 1)) {
4411 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4412 "block is 0 with a 1k block and cluster size");
4416 blocks_count = (ext4_blocks_count(es) -
4417 le32_to_cpu(es->s_first_data_block) +
4418 EXT4_BLOCKS_PER_GROUP(sb) - 1);
4419 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4420 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
4421 ext4_msg(sb, KERN_WARNING, "groups count too large: %llu "
4422 "(block count %llu, first data block %u, "
4423 "blocks per group %lu)", blocks_count,
4424 ext4_blocks_count(es),
4425 le32_to_cpu(es->s_first_data_block),
4426 EXT4_BLOCKS_PER_GROUP(sb));
4429 sbi->s_groups_count = blocks_count;
4430 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
4431 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
4432 if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) !=
4433 le32_to_cpu(es->s_inodes_count)) {
4434 ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu",
4435 le32_to_cpu(es->s_inodes_count),
4436 ((u64)sbi->s_groups_count * sbi->s_inodes_per_group));
4440 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
4441 EXT4_DESC_PER_BLOCK(sb);
4442 if (ext4_has_feature_meta_bg(sb)) {
4443 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
4444 ext4_msg(sb, KERN_WARNING,
4445 "first meta block group too large: %u "
4446 "(group descriptor block count %u)",
4447 le32_to_cpu(es->s_first_meta_bg), db_count);
4451 rcu_assign_pointer(sbi->s_group_desc,
4452 kvmalloc_array(db_count,
4453 sizeof(struct buffer_head *),
4455 if (sbi->s_group_desc == NULL) {
4456 ext4_msg(sb, KERN_ERR, "not enough memory");
4461 bgl_lock_init(sbi->s_blockgroup_lock);
4463 /* Pre-read the descriptors into the buffer cache */
4464 for (i = 0; i < db_count; i++) {
4465 block = descriptor_loc(sb, logical_sb_block, i);
4466 sb_breadahead_unmovable(sb, block);
4469 for (i = 0; i < db_count; i++) {
4470 struct buffer_head *bh;
4472 block = descriptor_loc(sb, logical_sb_block, i);
4473 bh = sb_bread_unmovable(sb, block);
4475 ext4_msg(sb, KERN_ERR,
4476 "can't read group descriptor %d", i);
4481 rcu_dereference(sbi->s_group_desc)[i] = bh;
4484 sbi->s_gdb_count = db_count;
4485 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
4486 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
4487 ret = -EFSCORRUPTED;
4491 timer_setup(&sbi->s_err_report, print_daily_error_info, 0);
4493 /* Register extent status tree shrinker */
4494 if (ext4_es_register_shrinker(sbi))
4497 sbi->s_stripe = ext4_get_stripe_size(sbi);
4498 sbi->s_extent_max_zeroout_kb = 32;
4501 * set up enough so that it can read an inode
4503 sb->s_op = &ext4_sops;
4504 sb->s_export_op = &ext4_export_ops;
4505 sb->s_xattr = ext4_xattr_handlers;
4506 #ifdef CONFIG_FS_ENCRYPTION
4507 sb->s_cop = &ext4_cryptops;
4509 #ifdef CONFIG_FS_VERITY
4510 sb->s_vop = &ext4_verityops;
4513 sb->dq_op = &ext4_quota_operations;
4514 if (ext4_has_feature_quota(sb))
4515 sb->s_qcop = &dquot_quotactl_sysfile_ops;
4517 sb->s_qcop = &ext4_qctl_operations;
4518 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
4520 memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
4522 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
4523 mutex_init(&sbi->s_orphan_lock);
4527 needs_recovery = (es->s_last_orphan != 0 ||
4528 ext4_has_feature_journal_needs_recovery(sb));
4530 if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb))
4531 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
4532 goto failed_mount3a;
4535 * The first inode we look at is the journal inode. Don't try
4536 * root first: it may be modified in the journal!
4538 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
4539 err = ext4_load_journal(sb, es, journal_devnum);
4541 goto failed_mount3a;
4542 } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) &&
4543 ext4_has_feature_journal_needs_recovery(sb)) {
4544 ext4_msg(sb, KERN_ERR, "required journal recovery "
4545 "suppressed and not mounted read-only");
4546 goto failed_mount_wq;
4548 /* Nojournal mode, all journal mount options are illegal */
4549 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
4550 ext4_msg(sb, KERN_ERR, "can't mount with "
4551 "journal_checksum, fs mounted w/o journal");
4552 goto failed_mount_wq;
4554 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4555 ext4_msg(sb, KERN_ERR, "can't mount with "
4556 "journal_async_commit, fs mounted w/o journal");
4557 goto failed_mount_wq;
4559 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
4560 ext4_msg(sb, KERN_ERR, "can't mount with "
4561 "commit=%lu, fs mounted w/o journal",
4562 sbi->s_commit_interval / HZ);
4563 goto failed_mount_wq;
4565 if (EXT4_MOUNT_DATA_FLAGS &
4566 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
4567 ext4_msg(sb, KERN_ERR, "can't mount with "
4568 "data=, fs mounted w/o journal");
4569 goto failed_mount_wq;
4571 sbi->s_def_mount_opt &= ~EXT4_MOUNT_JOURNAL_CHECKSUM;
4572 clear_opt(sb, JOURNAL_CHECKSUM);
4573 clear_opt(sb, DATA_FLAGS);
4574 sbi->s_journal = NULL;
4579 if (ext4_has_feature_64bit(sb) &&
4580 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4581 JBD2_FEATURE_INCOMPAT_64BIT)) {
4582 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
4583 goto failed_mount_wq;
4586 if (!set_journal_csum_feature_set(sb)) {
4587 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4589 goto failed_mount_wq;
4592 /* We have now updated the journal if required, so we can
4593 * validate the data journaling mode. */
4594 switch (test_opt(sb, DATA_FLAGS)) {
4596 /* No mode set, assume a default based on the journal
4597 * capabilities: ORDERED_DATA if the journal can
4598 * cope, else JOURNAL_DATA
4600 if (jbd2_journal_check_available_features
4601 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4602 set_opt(sb, ORDERED_DATA);
4603 sbi->s_def_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
4605 set_opt(sb, JOURNAL_DATA);
4606 sbi->s_def_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
4610 case EXT4_MOUNT_ORDERED_DATA:
4611 case EXT4_MOUNT_WRITEBACK_DATA:
4612 if (!jbd2_journal_check_available_features
4613 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4614 ext4_msg(sb, KERN_ERR, "Journal does not support "
4615 "requested data journaling mode");
4616 goto failed_mount_wq;
4622 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
4623 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4624 ext4_msg(sb, KERN_ERR, "can't mount with "
4625 "journal_async_commit in data=ordered mode");
4626 goto failed_mount_wq;
4629 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4631 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4634 if (!test_opt(sb, NO_MBCACHE)) {
4635 sbi->s_ea_block_cache = ext4_xattr_create_cache();
4636 if (!sbi->s_ea_block_cache) {
4637 ext4_msg(sb, KERN_ERR,
4638 "Failed to create ea_block_cache");
4639 goto failed_mount_wq;
4642 if (ext4_has_feature_ea_inode(sb)) {
4643 sbi->s_ea_inode_cache = ext4_xattr_create_cache();
4644 if (!sbi->s_ea_inode_cache) {
4645 ext4_msg(sb, KERN_ERR,
4646 "Failed to create ea_inode_cache");
4647 goto failed_mount_wq;
4652 if (ext4_has_feature_verity(sb) && blocksize != PAGE_SIZE) {
4653 ext4_msg(sb, KERN_ERR, "Unsupported blocksize for fs-verity");
4654 goto failed_mount_wq;
4657 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !sb_rdonly(sb) &&
4658 !ext4_has_feature_encrypt(sb)) {
4659 ext4_set_feature_encrypt(sb);
4660 ext4_commit_super(sb, 1);
4664 * Get the # of file system overhead blocks from the
4665 * superblock if present.
4667 if (es->s_overhead_clusters)
4668 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4670 err = ext4_calculate_overhead(sb);
4672 goto failed_mount_wq;
4676 * The maximum number of concurrent works can be high and
4677 * concurrency isn't really necessary. Limit it to 1.
4679 EXT4_SB(sb)->rsv_conversion_wq =
4680 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4681 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4682 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
4688 * The jbd2_journal_load will have done any necessary log recovery,
4689 * so we can safely mount the rest of the filesystem now.
4692 root = ext4_iget(sb, EXT4_ROOT_INO, EXT4_IGET_SPECIAL);
4694 ext4_msg(sb, KERN_ERR, "get root inode failed");
4695 ret = PTR_ERR(root);
4699 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4700 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4705 #ifdef CONFIG_UNICODE
4706 if (sbi->s_encoding)
4707 sb->s_d_op = &ext4_dentry_ops;
4710 sb->s_root = d_make_root(root);
4712 ext4_msg(sb, KERN_ERR, "get root dentry failed");
4717 ret = ext4_setup_super(sb, es, sb_rdonly(sb));
4718 if (ret == -EROFS) {
4719 sb->s_flags |= SB_RDONLY;
4722 goto failed_mount4a;
4724 ext4_set_resv_clusters(sb);
4726 err = ext4_setup_system_zone(sb);
4728 ext4_msg(sb, KERN_ERR, "failed to initialize system "
4730 goto failed_mount4a;
4734 err = ext4_mb_init(sb);
4736 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4741 block = ext4_count_free_clusters(sb);
4742 ext4_free_blocks_count_set(sbi->s_es,
4743 EXT4_C2B(sbi, block));
4744 ext4_superblock_csum_set(sb);
4745 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4748 unsigned long freei = ext4_count_free_inodes(sb);
4749 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
4750 ext4_superblock_csum_set(sb);
4751 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4755 err = percpu_counter_init(&sbi->s_dirs_counter,
4756 ext4_count_dirs(sb), GFP_KERNEL);
4758 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4761 err = percpu_init_rwsem(&sbi->s_writepages_rwsem);
4764 ext4_msg(sb, KERN_ERR, "insufficient memory");
4768 if (ext4_has_feature_flex_bg(sb))
4769 if (!ext4_fill_flex_info(sb)) {
4770 ext4_msg(sb, KERN_ERR,
4771 "unable to initialize "
4772 "flex_bg meta info!");
4776 err = ext4_register_li_request(sb, first_not_zeroed);
4780 err = ext4_register_sysfs(sb);
4785 /* Enable quota usage during mount. */
4786 if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) {
4787 err = ext4_enable_quotas(sb);
4791 #endif /* CONFIG_QUOTA */
4794 * Save the original bdev mapping's wb_err value which could be
4795 * used to detect the metadata async write error.
4797 spin_lock_init(&sbi->s_bdev_wb_lock);
4799 errseq_check_and_advance(&sb->s_bdev->bd_inode->i_mapping->wb_err,
4800 &sbi->s_bdev_wb_err);
4801 sb->s_bdev->bd_super = sb;
4802 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4803 ext4_orphan_cleanup(sb, es);
4804 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4805 if (needs_recovery) {
4806 ext4_msg(sb, KERN_INFO, "recovery complete");
4807 err = ext4_mark_recovery_complete(sb, es);
4811 if (EXT4_SB(sb)->s_journal) {
4812 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4813 descr = " journalled data mode";
4814 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4815 descr = " ordered data mode";
4817 descr = " writeback data mode";
4819 descr = "out journal";
4821 if (test_opt(sb, DISCARD)) {
4822 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4823 if (!blk_queue_discard(q))
4824 ext4_msg(sb, KERN_WARNING,
4825 "mounting with \"discard\" option, but "
4826 "the device does not support discard");
4829 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
4830 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4831 "Opts: %.*s%s%s", descr,
4832 (int) sizeof(sbi->s_es->s_mount_opts),
4833 sbi->s_es->s_mount_opts,
4834 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4836 if (es->s_error_count)
4837 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4839 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4840 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4841 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4842 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4849 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4853 ext4_unregister_sysfs(sb);
4855 ext4_unregister_li_request(sb);
4857 ext4_mb_release(sb);
4859 flex_groups = rcu_dereference(sbi->s_flex_groups);
4861 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
4862 kvfree(flex_groups[i]);
4863 kvfree(flex_groups);
4866 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4867 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4868 percpu_counter_destroy(&sbi->s_dirs_counter);
4869 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4870 percpu_free_rwsem(&sbi->s_writepages_rwsem);
4872 ext4_ext_release(sb);
4873 ext4_release_system_zone(sb);
4878 ext4_msg(sb, KERN_ERR, "mount failed");
4879 if (EXT4_SB(sb)->rsv_conversion_wq)
4880 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4882 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
4883 sbi->s_ea_inode_cache = NULL;
4885 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
4886 sbi->s_ea_block_cache = NULL;
4888 if (sbi->s_journal) {
4889 jbd2_journal_destroy(sbi->s_journal);
4890 sbi->s_journal = NULL;
4893 ext4_es_unregister_shrinker(sbi);
4895 del_timer_sync(&sbi->s_err_report);
4897 kthread_stop(sbi->s_mmp_tsk);
4900 group_desc = rcu_dereference(sbi->s_group_desc);
4901 for (i = 0; i < db_count; i++)
4902 brelse(group_desc[i]);
4906 if (sbi->s_chksum_driver)
4907 crypto_free_shash(sbi->s_chksum_driver);
4909 #ifdef CONFIG_UNICODE
4910 utf8_unload(sbi->s_encoding);
4914 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4915 kfree(get_qf_name(sb, sbi, i));
4917 fscrypt_free_dummy_context(&sbi->s_dummy_enc_ctx);
4918 ext4_blkdev_remove(sbi);
4921 sb->s_fs_info = NULL;
4922 kfree(sbi->s_blockgroup_lock);
4926 fs_put_dax(dax_dev);
4927 return err ? err : ret;
4931 * Setup any per-fs journal parameters now. We'll do this both on
4932 * initial mount, once the journal has been initialised but before we've
4933 * done any recovery; and again on any subsequent remount.
4935 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4937 struct ext4_sb_info *sbi = EXT4_SB(sb);
4939 journal->j_commit_interval = sbi->s_commit_interval;
4940 journal->j_min_batch_time = sbi->s_min_batch_time;
4941 journal->j_max_batch_time = sbi->s_max_batch_time;
4943 write_lock(&journal->j_state_lock);
4944 if (test_opt(sb, BARRIER))
4945 journal->j_flags |= JBD2_BARRIER;
4947 journal->j_flags &= ~JBD2_BARRIER;
4948 if (test_opt(sb, DATA_ERR_ABORT))
4949 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4951 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4952 write_unlock(&journal->j_state_lock);
4955 static struct inode *ext4_get_journal_inode(struct super_block *sb,
4956 unsigned int journal_inum)
4958 struct inode *journal_inode;
4961 * Test for the existence of a valid inode on disk. Bad things
4962 * happen if we iget() an unused inode, as the subsequent iput()
4963 * will try to delete it.
4965 journal_inode = ext4_iget(sb, journal_inum, EXT4_IGET_SPECIAL);
4966 if (IS_ERR(journal_inode)) {
4967 ext4_msg(sb, KERN_ERR, "no journal found");
4970 if (!journal_inode->i_nlink) {
4971 make_bad_inode(journal_inode);
4972 iput(journal_inode);
4973 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4977 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4978 journal_inode, journal_inode->i_size);
4979 if (!S_ISREG(journal_inode->i_mode)) {
4980 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4981 iput(journal_inode);
4984 return journal_inode;
4987 static journal_t *ext4_get_journal(struct super_block *sb,
4988 unsigned int journal_inum)
4990 struct inode *journal_inode;
4993 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
4996 journal_inode = ext4_get_journal_inode(sb, journal_inum);
5000 journal = jbd2_journal_init_inode(journal_inode);
5002 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
5003 iput(journal_inode);
5006 journal->j_private = sb;
5007 ext4_init_journal_params(sb, journal);
5011 static journal_t *ext4_get_dev_journal(struct super_block *sb,
5014 struct buffer_head *bh;
5018 int hblock, blocksize;
5019 ext4_fsblk_t sb_block;
5020 unsigned long offset;
5021 struct ext4_super_block *es;
5022 struct block_device *bdev;
5024 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5027 bdev = ext4_blkdev_get(j_dev, sb);
5031 blocksize = sb->s_blocksize;
5032 hblock = bdev_logical_block_size(bdev);
5033 if (blocksize < hblock) {
5034 ext4_msg(sb, KERN_ERR,
5035 "blocksize too small for journal device");
5039 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
5040 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
5041 set_blocksize(bdev, blocksize);
5042 if (!(bh = __bread(bdev, sb_block, blocksize))) {
5043 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
5044 "external journal");
5048 es = (struct ext4_super_block *) (bh->b_data + offset);
5049 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
5050 !(le32_to_cpu(es->s_feature_incompat) &
5051 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
5052 ext4_msg(sb, KERN_ERR, "external journal has "
5058 if ((le32_to_cpu(es->s_feature_ro_compat) &
5059 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
5060 es->s_checksum != ext4_superblock_csum(sb, es)) {
5061 ext4_msg(sb, KERN_ERR, "external journal has "
5062 "corrupt superblock");
5067 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
5068 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
5073 len = ext4_blocks_count(es);
5074 start = sb_block + 1;
5075 brelse(bh); /* we're done with the superblock */
5077 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
5078 start, len, blocksize);
5080 ext4_msg(sb, KERN_ERR, "failed to create device journal");
5083 journal->j_private = sb;
5084 ll_rw_block(REQ_OP_READ, REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
5085 wait_on_buffer(journal->j_sb_buffer);
5086 if (!buffer_uptodate(journal->j_sb_buffer)) {
5087 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
5090 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
5091 ext4_msg(sb, KERN_ERR, "External journal has more than one "
5092 "user (unsupported) - %d",
5093 be32_to_cpu(journal->j_superblock->s_nr_users));
5096 EXT4_SB(sb)->journal_bdev = bdev;
5097 ext4_init_journal_params(sb, journal);
5101 jbd2_journal_destroy(journal);
5103 ext4_blkdev_put(bdev);
5107 static int ext4_load_journal(struct super_block *sb,
5108 struct ext4_super_block *es,
5109 unsigned long journal_devnum)
5112 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
5115 int really_read_only;
5118 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5119 return -EFSCORRUPTED;
5121 if (journal_devnum &&
5122 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
5123 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
5124 "numbers have changed");
5125 journal_dev = new_decode_dev(journal_devnum);
5127 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
5129 if (journal_inum && journal_dev) {
5130 ext4_msg(sb, KERN_ERR,
5131 "filesystem has both journal inode and journal device!");
5136 journal = ext4_get_journal(sb, journal_inum);
5140 journal = ext4_get_dev_journal(sb, journal_dev);
5145 journal_dev_ro = bdev_read_only(journal->j_dev);
5146 really_read_only = bdev_read_only(sb->s_bdev) | journal_dev_ro;
5148 if (journal_dev_ro && !sb_rdonly(sb)) {
5149 ext4_msg(sb, KERN_ERR,
5150 "journal device read-only, try mounting with '-o ro'");
5156 * Are we loading a blank journal or performing recovery after a
5157 * crash? For recovery, we need to check in advance whether we
5158 * can get read-write access to the device.
5160 if (ext4_has_feature_journal_needs_recovery(sb)) {
5161 if (sb_rdonly(sb)) {
5162 ext4_msg(sb, KERN_INFO, "INFO: recovery "
5163 "required on readonly filesystem");
5164 if (really_read_only) {
5165 ext4_msg(sb, KERN_ERR, "write access "
5166 "unavailable, cannot proceed "
5167 "(try mounting with noload)");
5171 ext4_msg(sb, KERN_INFO, "write access will "
5172 "be enabled during recovery");
5176 if (!(journal->j_flags & JBD2_BARRIER))
5177 ext4_msg(sb, KERN_INFO, "barriers disabled");
5179 if (!ext4_has_feature_journal_needs_recovery(sb))
5180 err = jbd2_journal_wipe(journal, !really_read_only);
5182 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
5184 memcpy(save, ((char *) es) +
5185 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
5186 err = jbd2_journal_load(journal);
5188 memcpy(((char *) es) + EXT4_S_ERR_START,
5189 save, EXT4_S_ERR_LEN);
5194 ext4_msg(sb, KERN_ERR, "error loading journal");
5198 EXT4_SB(sb)->s_journal = journal;
5199 err = ext4_clear_journal_err(sb, es);
5201 EXT4_SB(sb)->s_journal = NULL;
5202 jbd2_journal_destroy(journal);
5206 if (!really_read_only && journal_devnum &&
5207 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
5208 es->s_journal_dev = cpu_to_le32(journal_devnum);
5210 /* Make sure we flush the recovery flag to disk. */
5211 ext4_commit_super(sb, 1);
5217 jbd2_journal_destroy(journal);
5221 static int ext4_commit_super(struct super_block *sb, int sync)
5223 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
5224 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
5227 if (!sbh || block_device_ejected(sb))
5231 * The superblock bh should be mapped, but it might not be if the
5232 * device was hot-removed. Not much we can do but fail the I/O.
5234 if (!buffer_mapped(sbh))
5238 * If the file system is mounted read-only, don't update the
5239 * superblock write time. This avoids updating the superblock
5240 * write time when we are mounting the root file system
5241 * read/only but we need to replay the journal; at that point,
5242 * for people who are east of GMT and who make their clock
5243 * tick in localtime for Windows bug-for-bug compatibility,
5244 * the clock is set in the future, and this will cause e2fsck
5245 * to complain and force a full file system check.
5247 if (!(sb->s_flags & SB_RDONLY))
5248 ext4_update_tstamp(es, s_wtime);
5249 if (sb->s_bdev->bd_part)
5250 es->s_kbytes_written =
5251 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
5252 ((part_stat_read(sb->s_bdev->bd_part,
5253 sectors[STAT_WRITE]) -
5254 EXT4_SB(sb)->s_sectors_written_start) >> 1));
5256 es->s_kbytes_written =
5257 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
5258 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
5259 ext4_free_blocks_count_set(es,
5260 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
5261 &EXT4_SB(sb)->s_freeclusters_counter)));
5262 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
5263 es->s_free_inodes_count =
5264 cpu_to_le32(percpu_counter_sum_positive(
5265 &EXT4_SB(sb)->s_freeinodes_counter));
5266 BUFFER_TRACE(sbh, "marking dirty");
5267 ext4_superblock_csum_set(sb);
5270 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
5272 * Oh, dear. A previous attempt to write the
5273 * superblock failed. This could happen because the
5274 * USB device was yanked out. Or it could happen to
5275 * be a transient write error and maybe the block will
5276 * be remapped. Nothing we can do but to retry the
5277 * write and hope for the best.
5279 ext4_msg(sb, KERN_ERR, "previous I/O error to "
5280 "superblock detected");
5281 clear_buffer_write_io_error(sbh);
5282 set_buffer_uptodate(sbh);
5284 mark_buffer_dirty(sbh);
5287 error = __sync_dirty_buffer(sbh,
5288 REQ_SYNC | (test_opt(sb, BARRIER) ? REQ_FUA : 0));
5289 if (buffer_write_io_error(sbh)) {
5290 ext4_msg(sb, KERN_ERR, "I/O error while writing "
5292 clear_buffer_write_io_error(sbh);
5293 set_buffer_uptodate(sbh);
5300 * Have we just finished recovery? If so, and if we are mounting (or
5301 * remounting) the filesystem readonly, then we will end up with a
5302 * consistent fs on disk. Record that fact.
5304 static int ext4_mark_recovery_complete(struct super_block *sb,
5305 struct ext4_super_block *es)
5308 journal_t *journal = EXT4_SB(sb)->s_journal;
5310 if (!ext4_has_feature_journal(sb)) {
5311 if (journal != NULL) {
5312 ext4_error(sb, "Journal got removed while the fs was "
5314 return -EFSCORRUPTED;
5318 jbd2_journal_lock_updates(journal);
5319 err = jbd2_journal_flush(journal);
5323 if (ext4_has_feature_journal_needs_recovery(sb) && sb_rdonly(sb)) {
5324 ext4_clear_feature_journal_needs_recovery(sb);
5325 ext4_commit_super(sb, 1);
5328 jbd2_journal_unlock_updates(journal);
5333 * If we are mounting (or read-write remounting) a filesystem whose journal
5334 * has recorded an error from a previous lifetime, move that error to the
5335 * main filesystem now.
5337 static int ext4_clear_journal_err(struct super_block *sb,
5338 struct ext4_super_block *es)
5344 if (!ext4_has_feature_journal(sb)) {
5345 ext4_error(sb, "Journal got removed while the fs was mounted!");
5346 return -EFSCORRUPTED;
5349 journal = EXT4_SB(sb)->s_journal;
5352 * Now check for any error status which may have been recorded in the
5353 * journal by a prior ext4_error() or ext4_abort()
5356 j_errno = jbd2_journal_errno(journal);
5360 errstr = ext4_decode_error(sb, j_errno, nbuf);
5361 ext4_warning(sb, "Filesystem error recorded "
5362 "from previous mount: %s", errstr);
5363 ext4_warning(sb, "Marking fs in need of filesystem check.");
5365 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
5366 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
5367 ext4_commit_super(sb, 1);
5369 jbd2_journal_clear_err(journal);
5370 jbd2_journal_update_sb_errno(journal);
5376 * Force the running and committing transactions to commit,
5377 * and wait on the commit.
5379 int ext4_force_commit(struct super_block *sb)
5386 journal = EXT4_SB(sb)->s_journal;
5387 return ext4_journal_force_commit(journal);
5390 static int ext4_sync_fs(struct super_block *sb, int wait)
5394 bool needs_barrier = false;
5395 struct ext4_sb_info *sbi = EXT4_SB(sb);
5397 if (unlikely(ext4_forced_shutdown(sbi)))
5400 trace_ext4_sync_fs(sb, wait);
5401 flush_workqueue(sbi->rsv_conversion_wq);
5403 * Writeback quota in non-journalled quota case - journalled quota has
5406 dquot_writeback_dquots(sb, -1);
5408 * Data writeback is possible w/o journal transaction, so barrier must
5409 * being sent at the end of the function. But we can skip it if
5410 * transaction_commit will do it for us.
5412 if (sbi->s_journal) {
5413 target = jbd2_get_latest_transaction(sbi->s_journal);
5414 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
5415 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
5416 needs_barrier = true;
5418 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
5420 ret = jbd2_log_wait_commit(sbi->s_journal,
5423 } else if (wait && test_opt(sb, BARRIER))
5424 needs_barrier = true;
5425 if (needs_barrier) {
5427 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL);
5436 * LVM calls this function before a (read-only) snapshot is created. This
5437 * gives us a chance to flush the journal completely and mark the fs clean.
5439 * Note that only this function cannot bring a filesystem to be in a clean
5440 * state independently. It relies on upper layer to stop all data & metadata
5443 static int ext4_freeze(struct super_block *sb)
5451 journal = EXT4_SB(sb)->s_journal;
5454 /* Now we set up the journal barrier. */
5455 jbd2_journal_lock_updates(journal);
5458 * Don't clear the needs_recovery flag if we failed to
5459 * flush the journal.
5461 error = jbd2_journal_flush(journal);
5465 /* Journal blocked and flushed, clear needs_recovery flag. */
5466 ext4_clear_feature_journal_needs_recovery(sb);
5469 error = ext4_commit_super(sb, 1);
5472 /* we rely on upper layer to stop further updates */
5473 jbd2_journal_unlock_updates(journal);
5478 * Called by LVM after the snapshot is done. We need to reset the RECOVER
5479 * flag here, even though the filesystem is not technically dirty yet.
5481 static int ext4_unfreeze(struct super_block *sb)
5483 if (sb_rdonly(sb) || ext4_forced_shutdown(EXT4_SB(sb)))
5486 if (EXT4_SB(sb)->s_journal) {
5487 /* Reset the needs_recovery flag before the fs is unlocked. */
5488 ext4_set_feature_journal_needs_recovery(sb);
5491 ext4_commit_super(sb, 1);
5496 * Structure to save mount options for ext4_remount's benefit
5498 struct ext4_mount_options {
5499 unsigned long s_mount_opt;
5500 unsigned long s_mount_opt2;
5503 unsigned long s_commit_interval;
5504 u32 s_min_batch_time, s_max_batch_time;
5507 char *s_qf_names[EXT4_MAXQUOTAS];
5511 static int ext4_remount(struct super_block *sb, int *flags, char *data)
5513 struct ext4_super_block *es;
5514 struct ext4_sb_info *sbi = EXT4_SB(sb);
5515 unsigned long old_sb_flags;
5516 struct ext4_mount_options old_opts;
5517 int enable_quota = 0;
5519 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
5523 char *to_free[EXT4_MAXQUOTAS];
5525 char *orig_data = kstrdup(data, GFP_KERNEL);
5527 if (data && !orig_data)
5530 /* Store the original options */
5531 old_sb_flags = sb->s_flags;
5532 old_opts.s_mount_opt = sbi->s_mount_opt;
5533 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
5534 old_opts.s_resuid = sbi->s_resuid;
5535 old_opts.s_resgid = sbi->s_resgid;
5536 old_opts.s_commit_interval = sbi->s_commit_interval;
5537 old_opts.s_min_batch_time = sbi->s_min_batch_time;
5538 old_opts.s_max_batch_time = sbi->s_max_batch_time;
5540 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
5541 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5542 if (sbi->s_qf_names[i]) {
5543 char *qf_name = get_qf_name(sb, sbi, i);
5545 old_opts.s_qf_names[i] = kstrdup(qf_name, GFP_KERNEL);
5546 if (!old_opts.s_qf_names[i]) {
5547 for (j = 0; j < i; j++)
5548 kfree(old_opts.s_qf_names[j]);
5553 old_opts.s_qf_names[i] = NULL;
5555 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
5556 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
5558 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
5563 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
5564 test_opt(sb, JOURNAL_CHECKSUM)) {
5565 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
5566 "during remount not supported; ignoring");
5567 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
5570 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
5571 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
5572 ext4_msg(sb, KERN_ERR, "can't mount with "
5573 "both data=journal and delalloc");
5577 if (test_opt(sb, DIOREAD_NOLOCK)) {
5578 ext4_msg(sb, KERN_ERR, "can't mount with "
5579 "both data=journal and dioread_nolock");
5583 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
5584 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
5585 ext4_msg(sb, KERN_ERR, "can't mount with "
5586 "journal_async_commit in data=ordered mode");
5592 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) {
5593 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount");
5598 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
5599 ext4_abort(sb, EXT4_ERR_ESHUTDOWN, "Abort forced by user");
5601 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
5602 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
5606 if (sbi->s_journal) {
5607 ext4_init_journal_params(sb, sbi->s_journal);
5608 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
5611 if (*flags & SB_LAZYTIME)
5612 sb->s_flags |= SB_LAZYTIME;
5614 if ((bool)(*flags & SB_RDONLY) != sb_rdonly(sb)) {
5615 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
5620 if (*flags & SB_RDONLY) {
5621 err = sync_filesystem(sb);
5624 err = dquot_suspend(sb, -1);
5629 * First of all, the unconditional stuff we have to do
5630 * to disable replay of the journal when we next remount
5632 sb->s_flags |= SB_RDONLY;
5635 * OK, test if we are remounting a valid rw partition
5636 * readonly, and if so set the rdonly flag and then
5637 * mark the partition as valid again.
5639 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
5640 (sbi->s_mount_state & EXT4_VALID_FS))
5641 es->s_state = cpu_to_le16(sbi->s_mount_state);
5643 if (sbi->s_journal) {
5645 * We let remount-ro finish even if marking fs
5646 * as clean failed...
5648 ext4_mark_recovery_complete(sb, es);
5651 kthread_stop(sbi->s_mmp_tsk);
5653 /* Make sure we can mount this feature set readwrite */
5654 if (ext4_has_feature_readonly(sb) ||
5655 !ext4_feature_set_ok(sb, 0)) {
5660 * Make sure the group descriptor checksums
5661 * are sane. If they aren't, refuse to remount r/w.
5663 for (g = 0; g < sbi->s_groups_count; g++) {
5664 struct ext4_group_desc *gdp =
5665 ext4_get_group_desc(sb, g, NULL);
5667 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
5668 ext4_msg(sb, KERN_ERR,
5669 "ext4_remount: Checksum for group %u failed (%u!=%u)",
5670 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
5671 le16_to_cpu(gdp->bg_checksum));
5678 * If we have an unprocessed orphan list hanging
5679 * around from a previously readonly bdev mount,
5680 * require a full umount/remount for now.
5682 if (es->s_last_orphan) {
5683 ext4_msg(sb, KERN_WARNING, "Couldn't "
5684 "remount RDWR because of unprocessed "
5685 "orphan inode list. Please "
5686 "umount/remount instead");
5692 * Update the original bdev mapping's wb_err value
5693 * which could be used to detect the metadata async
5696 errseq_check_and_advance(&sb->s_bdev->bd_inode->i_mapping->wb_err,
5697 &sbi->s_bdev_wb_err);
5700 * Mounting a RDONLY partition read-write, so reread
5701 * and store the current valid flag. (It may have
5702 * been changed by e2fsck since we originally mounted
5705 if (sbi->s_journal) {
5706 err = ext4_clear_journal_err(sb, es);
5710 sbi->s_mount_state = le16_to_cpu(es->s_state);
5712 err = ext4_setup_super(sb, es, 0);
5716 sb->s_flags &= ~SB_RDONLY;
5717 if (ext4_has_feature_mmp(sb))
5718 if (ext4_multi_mount_protect(sb,
5719 le64_to_cpu(es->s_mmp_block))) {
5728 * Reinitialize lazy itable initialization thread based on
5731 if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE))
5732 ext4_unregister_li_request(sb);
5734 ext4_group_t first_not_zeroed;
5735 first_not_zeroed = ext4_has_uninit_itable(sb);
5736 ext4_register_li_request(sb, first_not_zeroed);
5739 ext4_setup_system_zone(sb);
5740 if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY)) {
5741 err = ext4_commit_super(sb, 1);
5747 /* Release old quota file names */
5748 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5749 kfree(old_opts.s_qf_names[i]);
5751 if (sb_any_quota_suspended(sb))
5752 dquot_resume(sb, -1);
5753 else if (ext4_has_feature_quota(sb)) {
5754 err = ext4_enable_quotas(sb);
5761 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
5762 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5767 sb->s_flags = old_sb_flags;
5768 sbi->s_mount_opt = old_opts.s_mount_opt;
5769 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
5770 sbi->s_resuid = old_opts.s_resuid;
5771 sbi->s_resgid = old_opts.s_resgid;
5772 sbi->s_commit_interval = old_opts.s_commit_interval;
5773 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5774 sbi->s_max_batch_time = old_opts.s_max_batch_time;
5776 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
5777 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
5778 to_free[i] = get_qf_name(sb, sbi, i);
5779 rcu_assign_pointer(sbi->s_qf_names[i], old_opts.s_qf_names[i]);
5782 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5790 static int ext4_statfs_project(struct super_block *sb,
5791 kprojid_t projid, struct kstatfs *buf)
5794 struct dquot *dquot;
5798 qid = make_kqid_projid(projid);
5799 dquot = dqget(sb, qid);
5801 return PTR_ERR(dquot);
5802 spin_lock(&dquot->dq_dqb_lock);
5804 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
5805 dquot->dq_dqb.dqb_bhardlimit);
5806 limit >>= sb->s_blocksize_bits;
5808 if (limit && buf->f_blocks > limit) {
5809 curblock = (dquot->dq_dqb.dqb_curspace +
5810 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
5811 buf->f_blocks = limit;
5812 buf->f_bfree = buf->f_bavail =
5813 (buf->f_blocks > curblock) ?
5814 (buf->f_blocks - curblock) : 0;
5817 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
5818 dquot->dq_dqb.dqb_ihardlimit);
5819 if (limit && buf->f_files > limit) {
5820 buf->f_files = limit;
5822 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
5823 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
5826 spin_unlock(&dquot->dq_dqb_lock);
5832 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
5834 struct super_block *sb = dentry->d_sb;
5835 struct ext4_sb_info *sbi = EXT4_SB(sb);
5836 struct ext4_super_block *es = sbi->s_es;
5837 ext4_fsblk_t overhead = 0, resv_blocks;
5840 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
5842 if (!test_opt(sb, MINIX_DF))
5843 overhead = sbi->s_overhead;
5845 buf->f_type = EXT4_SUPER_MAGIC;
5846 buf->f_bsize = sb->s_blocksize;
5847 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
5848 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5849 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
5850 /* prevent underflow in case that few free space is available */
5851 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
5852 buf->f_bavail = buf->f_bfree -
5853 (ext4_r_blocks_count(es) + resv_blocks);
5854 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
5856 buf->f_files = le32_to_cpu(es->s_inodes_count);
5857 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5858 buf->f_namelen = EXT4_NAME_LEN;
5859 fsid = le64_to_cpup((void *)es->s_uuid) ^
5860 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5861 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5862 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
5865 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
5866 sb_has_quota_limits_enabled(sb, PRJQUOTA))
5867 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
5876 * Helper functions so that transaction is started before we acquire dqio_sem
5877 * to keep correct lock ordering of transaction > dqio_sem
5879 static inline struct inode *dquot_to_inode(struct dquot *dquot)
5881 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
5884 static int ext4_write_dquot(struct dquot *dquot)
5888 struct inode *inode;
5890 inode = dquot_to_inode(dquot);
5891 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
5892 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
5894 return PTR_ERR(handle);
5895 ret = dquot_commit(dquot);
5896 err = ext4_journal_stop(handle);
5902 static int ext4_acquire_dquot(struct dquot *dquot)
5907 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5908 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
5910 return PTR_ERR(handle);
5911 ret = dquot_acquire(dquot);
5912 err = ext4_journal_stop(handle);
5918 static int ext4_release_dquot(struct dquot *dquot)
5923 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5924 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
5925 if (IS_ERR(handle)) {
5926 /* Release dquot anyway to avoid endless cycle in dqput() */
5927 dquot_release(dquot);
5928 return PTR_ERR(handle);
5930 ret = dquot_release(dquot);
5931 err = ext4_journal_stop(handle);
5937 static int ext4_mark_dquot_dirty(struct dquot *dquot)
5939 struct super_block *sb = dquot->dq_sb;
5940 struct ext4_sb_info *sbi = EXT4_SB(sb);
5942 /* Are we journaling quotas? */
5943 if (ext4_has_feature_quota(sb) ||
5944 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
5945 dquot_mark_dquot_dirty(dquot);
5946 return ext4_write_dquot(dquot);
5948 return dquot_mark_dquot_dirty(dquot);
5952 static int ext4_write_info(struct super_block *sb, int type)
5957 /* Data block + inode block */
5958 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
5960 return PTR_ERR(handle);
5961 ret = dquot_commit_info(sb, type);
5962 err = ext4_journal_stop(handle);
5969 * Turn on quotas during mount time - we need to find
5970 * the quota file and such...
5972 static int ext4_quota_on_mount(struct super_block *sb, int type)
5974 return dquot_quota_on_mount(sb, get_qf_name(sb, EXT4_SB(sb), type),
5975 EXT4_SB(sb)->s_jquota_fmt, type);
5978 static void lockdep_set_quota_inode(struct inode *inode, int subclass)
5980 struct ext4_inode_info *ei = EXT4_I(inode);
5982 /* The first argument of lockdep_set_subclass has to be
5983 * *exactly* the same as the argument to init_rwsem() --- in
5984 * this case, in init_once() --- or lockdep gets unhappy
5985 * because the name of the lock is set using the
5986 * stringification of the argument to init_rwsem().
5988 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
5989 lockdep_set_subclass(&ei->i_data_sem, subclass);
5993 * Standard function to be called on quota_on
5995 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
5996 const struct path *path)
6000 if (!test_opt(sb, QUOTA))
6003 /* Quotafile not on the same filesystem? */
6004 if (path->dentry->d_sb != sb)
6006 /* Journaling quota? */
6007 if (EXT4_SB(sb)->s_qf_names[type]) {
6008 /* Quotafile not in fs root? */
6009 if (path->dentry->d_parent != sb->s_root)
6010 ext4_msg(sb, KERN_WARNING,
6011 "Quota file not on filesystem root. "
6012 "Journaled quota will not work");
6013 sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY;
6016 * Clear the flag just in case mount options changed since
6019 sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY;
6023 * When we journal data on quota file, we have to flush journal to see
6024 * all updates to the file when we bypass pagecache...
6026 if (EXT4_SB(sb)->s_journal &&
6027 ext4_should_journal_data(d_inode(path->dentry))) {
6029 * We don't need to lock updates but journal_flush() could
6030 * otherwise be livelocked...
6032 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
6033 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
6034 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
6039 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
6040 err = dquot_quota_on(sb, type, format_id, path);
6042 lockdep_set_quota_inode(path->dentry->d_inode,
6045 struct inode *inode = d_inode(path->dentry);
6049 * Set inode flags to prevent userspace from messing with quota
6050 * files. If this fails, we return success anyway since quotas
6051 * are already enabled and this is not a hard failure.
6054 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
6057 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL;
6058 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
6059 S_NOATIME | S_IMMUTABLE);
6060 err = ext4_mark_inode_dirty(handle, inode);
6061 ext4_journal_stop(handle);
6063 inode_unlock(inode);
6068 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
6072 struct inode *qf_inode;
6073 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
6074 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
6075 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
6076 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
6079 BUG_ON(!ext4_has_feature_quota(sb));
6081 if (!qf_inums[type])
6084 qf_inode = ext4_iget(sb, qf_inums[type], EXT4_IGET_SPECIAL);
6085 if (IS_ERR(qf_inode)) {
6086 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
6087 return PTR_ERR(qf_inode);
6090 /* Don't account quota for quota files to avoid recursion */
6091 qf_inode->i_flags |= S_NOQUOTA;
6092 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
6093 err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
6095 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
6101 /* Enable usage tracking for all quota types. */
6102 static int ext4_enable_quotas(struct super_block *sb)
6105 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
6106 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
6107 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
6108 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
6110 bool quota_mopt[EXT4_MAXQUOTAS] = {
6111 test_opt(sb, USRQUOTA),
6112 test_opt(sb, GRPQUOTA),
6113 test_opt(sb, PRJQUOTA),
6116 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY;
6117 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
6118 if (qf_inums[type]) {
6119 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
6120 DQUOT_USAGE_ENABLED |
6121 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
6124 "Failed to enable quota tracking "
6125 "(type=%d, err=%d). Please run "
6126 "e2fsck to fix.", type, err);
6127 for (type--; type >= 0; type--)
6128 dquot_quota_off(sb, type);
6137 static int ext4_quota_off(struct super_block *sb, int type)
6139 struct inode *inode = sb_dqopt(sb)->files[type];
6143 /* Force all delayed allocation blocks to be allocated.
6144 * Caller already holds s_umount sem */
6145 if (test_opt(sb, DELALLOC))
6146 sync_filesystem(sb);
6148 if (!inode || !igrab(inode))
6151 err = dquot_quota_off(sb, type);
6152 if (err || ext4_has_feature_quota(sb))
6157 * Update modification times of quota files when userspace can
6158 * start looking at them. If we fail, we return success anyway since
6159 * this is not a hard failure and quotas are already disabled.
6161 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
6162 if (IS_ERR(handle)) {
6163 err = PTR_ERR(handle);
6166 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL);
6167 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
6168 inode->i_mtime = inode->i_ctime = current_time(inode);
6169 err = ext4_mark_inode_dirty(handle, inode);
6170 ext4_journal_stop(handle);
6172 inode_unlock(inode);
6174 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL);
6178 return dquot_quota_off(sb, type);
6181 /* Read data from quotafile - avoid pagecache and such because we cannot afford
6182 * acquiring the locks... As quota files are never truncated and quota code
6183 * itself serializes the operations (and no one else should touch the files)
6184 * we don't have to be afraid of races */
6185 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
6186 size_t len, loff_t off)
6188 struct inode *inode = sb_dqopt(sb)->files[type];
6189 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
6190 int offset = off & (sb->s_blocksize - 1);
6193 struct buffer_head *bh;
6194 loff_t i_size = i_size_read(inode);
6198 if (off+len > i_size)
6201 while (toread > 0) {
6202 tocopy = sb->s_blocksize - offset < toread ?
6203 sb->s_blocksize - offset : toread;
6204 bh = ext4_bread(NULL, inode, blk, 0);
6207 if (!bh) /* A hole? */
6208 memset(data, 0, tocopy);
6210 memcpy(data, bh->b_data+offset, tocopy);
6220 /* Write to quotafile (we know the transaction is already started and has
6221 * enough credits) */
6222 static ssize_t ext4_quota_write(struct super_block *sb, int type,
6223 const char *data, size_t len, loff_t off)
6225 struct inode *inode = sb_dqopt(sb)->files[type];
6226 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
6227 int err = 0, err2 = 0, offset = off & (sb->s_blocksize - 1);
6229 struct buffer_head *bh;
6230 handle_t *handle = journal_current_handle();
6232 if (EXT4_SB(sb)->s_journal && !handle) {
6233 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
6234 " cancelled because transaction is not started",
6235 (unsigned long long)off, (unsigned long long)len);
6239 * Since we account only one data block in transaction credits,
6240 * then it is impossible to cross a block boundary.
6242 if (sb->s_blocksize - offset < len) {
6243 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
6244 " cancelled because not block aligned",
6245 (unsigned long long)off, (unsigned long long)len);
6250 bh = ext4_bread(handle, inode, blk,
6251 EXT4_GET_BLOCKS_CREATE |
6252 EXT4_GET_BLOCKS_METADATA_NOFAIL);
6253 } while (PTR_ERR(bh) == -ENOSPC &&
6254 ext4_should_retry_alloc(inode->i_sb, &retries));
6259 BUFFER_TRACE(bh, "get write access");
6260 err = ext4_journal_get_write_access(handle, bh);
6266 memcpy(bh->b_data+offset, data, len);
6267 flush_dcache_page(bh->b_page);
6269 err = ext4_handle_dirty_metadata(handle, NULL, bh);
6272 if (inode->i_size < off + len) {
6273 i_size_write(inode, off + len);
6274 EXT4_I(inode)->i_disksize = inode->i_size;
6275 err2 = ext4_mark_inode_dirty(handle, inode);
6276 if (unlikely(err2 && !err))
6279 return err ? err : len;
6283 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
6284 const char *dev_name, void *data)
6286 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
6289 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
6290 static inline void register_as_ext2(void)
6292 int err = register_filesystem(&ext2_fs_type);
6295 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
6298 static inline void unregister_as_ext2(void)
6300 unregister_filesystem(&ext2_fs_type);
6303 static inline int ext2_feature_set_ok(struct super_block *sb)
6305 if (ext4_has_unknown_ext2_incompat_features(sb))
6309 if (ext4_has_unknown_ext2_ro_compat_features(sb))
6314 static inline void register_as_ext2(void) { }
6315 static inline void unregister_as_ext2(void) { }
6316 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
6319 static inline void register_as_ext3(void)
6321 int err = register_filesystem(&ext3_fs_type);
6324 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
6327 static inline void unregister_as_ext3(void)
6329 unregister_filesystem(&ext3_fs_type);
6332 static inline int ext3_feature_set_ok(struct super_block *sb)
6334 if (ext4_has_unknown_ext3_incompat_features(sb))
6336 if (!ext4_has_feature_journal(sb))
6340 if (ext4_has_unknown_ext3_ro_compat_features(sb))
6345 static struct file_system_type ext4_fs_type = {
6346 .owner = THIS_MODULE,
6348 .mount = ext4_mount,
6349 .kill_sb = kill_block_super,
6350 .fs_flags = FS_REQUIRES_DEV,
6352 MODULE_ALIAS_FS("ext4");
6354 /* Shared across all ext4 file systems */
6355 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
6357 static int __init ext4_init_fs(void)
6361 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
6362 ext4_li_info = NULL;
6363 mutex_init(&ext4_li_mtx);
6365 /* Build-time check for flags consistency */
6366 ext4_check_flag_values();
6368 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
6369 init_waitqueue_head(&ext4__ioend_wq[i]);
6371 err = ext4_init_es();
6375 err = ext4_init_pending();
6379 err = ext4_init_post_read_processing();
6383 err = ext4_init_pageio();
6387 err = ext4_init_system_zone();
6391 err = ext4_init_sysfs();
6395 err = ext4_init_mballoc();
6398 err = init_inodecache();
6403 err = register_filesystem(&ext4_fs_type);
6409 unregister_as_ext2();
6410 unregister_as_ext3();
6411 destroy_inodecache();
6413 ext4_exit_mballoc();
6417 ext4_exit_system_zone();
6421 ext4_exit_post_read_processing();
6423 ext4_exit_pending();
6430 static void __exit ext4_exit_fs(void)
6432 ext4_destroy_lazyinit_thread();
6433 unregister_as_ext2();
6434 unregister_as_ext3();
6435 unregister_filesystem(&ext4_fs_type);
6436 destroy_inodecache();
6437 ext4_exit_mballoc();
6439 ext4_exit_system_zone();
6441 ext4_exit_post_read_processing();
6443 ext4_exit_pending();
6446 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
6447 MODULE_DESCRIPTION("Fourth Extended Filesystem");
6448 MODULE_LICENSE("GPL");
6449 MODULE_SOFTDEP("pre: crc32c");
6450 module_init(ext4_init_fs)
6451 module_exit(ext4_exit_fs)