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 atomic_inc(&EXT4_SB(sb)->s_msg_count);
748 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
754 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
758 static int ext4_warning_ratelimit(struct super_block *sb)
760 atomic_inc(&EXT4_SB(sb)->s_warning_count);
761 return ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state),
765 void __ext4_warning(struct super_block *sb, const char *function,
766 unsigned int line, const char *fmt, ...)
768 struct va_format vaf;
771 if (!ext4_warning_ratelimit(sb))
777 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
778 sb->s_id, function, line, &vaf);
782 void __ext4_warning_inode(const struct inode *inode, const char *function,
783 unsigned int line, const char *fmt, ...)
785 struct va_format vaf;
788 if (!ext4_warning_ratelimit(inode->i_sb))
794 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
795 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
796 function, line, inode->i_ino, current->comm, &vaf);
800 void __ext4_grp_locked_error(const char *function, unsigned int line,
801 struct super_block *sb, ext4_group_t grp,
802 unsigned long ino, ext4_fsblk_t block,
803 const char *fmt, ...)
807 struct va_format vaf;
810 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
813 trace_ext4_error(sb, function, line);
814 __save_error_info(sb, EFSCORRUPTED, ino, block, function, line);
816 if (ext4_error_ratelimit(sb)) {
820 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
821 sb->s_id, function, line, grp);
823 printk(KERN_CONT "inode %lu: ", ino);
825 printk(KERN_CONT "block %llu:",
826 (unsigned long long) block);
827 printk(KERN_CONT "%pV\n", &vaf);
831 if (test_opt(sb, WARN_ON_ERROR))
834 if (test_opt(sb, ERRORS_CONT)) {
835 ext4_commit_super(sb, 0);
839 ext4_unlock_group(sb, grp);
840 ext4_commit_super(sb, 1);
841 ext4_handle_error(sb);
843 * We only get here in the ERRORS_RO case; relocking the group
844 * may be dangerous, but nothing bad will happen since the
845 * filesystem will have already been marked read/only and the
846 * journal has been aborted. We return 1 as a hint to callers
847 * who might what to use the return value from
848 * ext4_grp_locked_error() to distinguish between the
849 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
850 * aggressively from the ext4 function in question, with a
851 * more appropriate error code.
853 ext4_lock_group(sb, grp);
857 void ext4_mark_group_bitmap_corrupted(struct super_block *sb,
861 struct ext4_sb_info *sbi = EXT4_SB(sb);
862 struct ext4_group_info *grp = ext4_get_group_info(sb, group);
863 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, group, NULL);
866 if (flags & EXT4_GROUP_INFO_BBITMAP_CORRUPT) {
867 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT,
870 percpu_counter_sub(&sbi->s_freeclusters_counter,
874 if (flags & EXT4_GROUP_INFO_IBITMAP_CORRUPT) {
875 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT,
880 count = ext4_free_inodes_count(sb, gdp);
881 percpu_counter_sub(&sbi->s_freeinodes_counter,
887 void ext4_update_dynamic_rev(struct super_block *sb)
889 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
891 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
895 "updating to rev %d because of new feature flag, "
896 "running e2fsck is recommended",
899 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
900 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
901 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
902 /* leave es->s_feature_*compat flags alone */
903 /* es->s_uuid will be set by e2fsck if empty */
906 * The rest of the superblock fields should be zero, and if not it
907 * means they are likely already in use, so leave them alone. We
908 * can leave it up to e2fsck to clean up any inconsistencies there.
913 * Open the external journal device
915 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
917 struct block_device *bdev;
919 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
925 ext4_msg(sb, KERN_ERR,
926 "failed to open journal device unknown-block(%u,%u) %ld",
927 MAJOR(dev), MINOR(dev), PTR_ERR(bdev));
932 * Release the journal device
934 static void ext4_blkdev_put(struct block_device *bdev)
936 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
939 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
941 struct block_device *bdev;
942 bdev = sbi->journal_bdev;
944 ext4_blkdev_put(bdev);
945 sbi->journal_bdev = NULL;
949 static inline struct inode *orphan_list_entry(struct list_head *l)
951 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
954 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
958 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
959 le32_to_cpu(sbi->s_es->s_last_orphan));
961 printk(KERN_ERR "sb_info orphan list:\n");
962 list_for_each(l, &sbi->s_orphan) {
963 struct inode *inode = orphan_list_entry(l);
965 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
966 inode->i_sb->s_id, inode->i_ino, inode,
967 inode->i_mode, inode->i_nlink,
973 static int ext4_quota_off(struct super_block *sb, int type);
975 static inline void ext4_quota_off_umount(struct super_block *sb)
979 /* Use our quota_off function to clear inode flags etc. */
980 for (type = 0; type < EXT4_MAXQUOTAS; type++)
981 ext4_quota_off(sb, type);
985 * This is a helper function which is used in the mount/remount
986 * codepaths (which holds s_umount) to fetch the quota file name.
988 static inline char *get_qf_name(struct super_block *sb,
989 struct ext4_sb_info *sbi,
992 return rcu_dereference_protected(sbi->s_qf_names[type],
993 lockdep_is_held(&sb->s_umount));
996 static inline void ext4_quota_off_umount(struct super_block *sb)
1001 static void ext4_put_super(struct super_block *sb)
1003 struct ext4_sb_info *sbi = EXT4_SB(sb);
1004 struct ext4_super_block *es = sbi->s_es;
1005 struct buffer_head **group_desc;
1006 struct flex_groups **flex_groups;
1010 ext4_unregister_li_request(sb);
1011 ext4_quota_off_umount(sb);
1013 destroy_workqueue(sbi->rsv_conversion_wq);
1016 * Unregister sysfs before destroying jbd2 journal.
1017 * Since we could still access attr_journal_task attribute via sysfs
1018 * path which could have sbi->s_journal->j_task as NULL
1020 ext4_unregister_sysfs(sb);
1022 if (sbi->s_journal) {
1023 aborted = is_journal_aborted(sbi->s_journal);
1024 err = jbd2_journal_destroy(sbi->s_journal);
1025 sbi->s_journal = NULL;
1026 if ((err < 0) && !aborted) {
1027 ext4_abort(sb, -err, "Couldn't clean up the journal");
1031 ext4_es_unregister_shrinker(sbi);
1032 del_timer_sync(&sbi->s_err_report);
1033 ext4_release_system_zone(sb);
1034 ext4_mb_release(sb);
1035 ext4_ext_release(sb);
1037 if (!sb_rdonly(sb) && !aborted) {
1038 ext4_clear_feature_journal_needs_recovery(sb);
1039 es->s_state = cpu_to_le16(sbi->s_mount_state);
1042 ext4_commit_super(sb, 1);
1045 group_desc = rcu_dereference(sbi->s_group_desc);
1046 for (i = 0; i < sbi->s_gdb_count; i++)
1047 brelse(group_desc[i]);
1049 flex_groups = rcu_dereference(sbi->s_flex_groups);
1051 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
1052 kvfree(flex_groups[i]);
1053 kvfree(flex_groups);
1056 percpu_counter_destroy(&sbi->s_freeclusters_counter);
1057 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1058 percpu_counter_destroy(&sbi->s_dirs_counter);
1059 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
1060 percpu_free_rwsem(&sbi->s_writepages_rwsem);
1062 for (i = 0; i < EXT4_MAXQUOTAS; i++)
1063 kfree(get_qf_name(sb, sbi, i));
1066 /* Debugging code just in case the in-memory inode orphan list
1067 * isn't empty. The on-disk one can be non-empty if we've
1068 * detected an error and taken the fs readonly, but the
1069 * in-memory list had better be clean by this point. */
1070 if (!list_empty(&sbi->s_orphan))
1071 dump_orphan_list(sb, sbi);
1072 J_ASSERT(list_empty(&sbi->s_orphan));
1074 sync_blockdev(sb->s_bdev);
1075 invalidate_bdev(sb->s_bdev);
1076 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
1078 * Invalidate the journal device's buffers. We don't want them
1079 * floating about in memory - the physical journal device may
1080 * hotswapped, and it breaks the `ro-after' testing code.
1082 sync_blockdev(sbi->journal_bdev);
1083 invalidate_bdev(sbi->journal_bdev);
1084 ext4_blkdev_remove(sbi);
1087 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
1088 sbi->s_ea_inode_cache = NULL;
1090 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
1091 sbi->s_ea_block_cache = NULL;
1094 kthread_stop(sbi->s_mmp_tsk);
1096 sb->s_fs_info = NULL;
1098 * Now that we are completely done shutting down the
1099 * superblock, we need to actually destroy the kobject.
1101 kobject_put(&sbi->s_kobj);
1102 wait_for_completion(&sbi->s_kobj_unregister);
1103 if (sbi->s_chksum_driver)
1104 crypto_free_shash(sbi->s_chksum_driver);
1105 kfree(sbi->s_blockgroup_lock);
1106 fs_put_dax(sbi->s_daxdev);
1107 fscrypt_free_dummy_context(&sbi->s_dummy_enc_ctx);
1108 #ifdef CONFIG_UNICODE
1109 utf8_unload(sbi->s_encoding);
1114 static struct kmem_cache *ext4_inode_cachep;
1117 * Called inside transaction, so use GFP_NOFS
1119 static struct inode *ext4_alloc_inode(struct super_block *sb)
1121 struct ext4_inode_info *ei;
1123 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
1127 inode_set_iversion(&ei->vfs_inode, 1);
1128 spin_lock_init(&ei->i_raw_lock);
1129 INIT_LIST_HEAD(&ei->i_prealloc_list);
1130 atomic_set(&ei->i_prealloc_active, 0);
1131 spin_lock_init(&ei->i_prealloc_lock);
1132 ext4_es_init_tree(&ei->i_es_tree);
1133 rwlock_init(&ei->i_es_lock);
1134 INIT_LIST_HEAD(&ei->i_es_list);
1135 ei->i_es_all_nr = 0;
1136 ei->i_es_shk_nr = 0;
1137 ei->i_es_shrink_lblk = 0;
1138 ei->i_reserved_data_blocks = 0;
1139 spin_lock_init(&(ei->i_block_reservation_lock));
1140 ext4_init_pending_tree(&ei->i_pending_tree);
1142 ei->i_reserved_quota = 0;
1143 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
1146 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
1147 spin_lock_init(&ei->i_completed_io_lock);
1149 ei->i_datasync_tid = 0;
1150 atomic_set(&ei->i_unwritten, 0);
1151 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
1152 return &ei->vfs_inode;
1155 static int ext4_drop_inode(struct inode *inode)
1157 int drop = generic_drop_inode(inode);
1160 drop = fscrypt_drop_inode(inode);
1162 trace_ext4_drop_inode(inode, drop);
1166 static void ext4_free_in_core_inode(struct inode *inode)
1168 fscrypt_free_inode(inode);
1169 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
1172 static void ext4_destroy_inode(struct inode *inode)
1174 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
1175 ext4_msg(inode->i_sb, KERN_ERR,
1176 "Inode %lu (%p): orphan list check failed!",
1177 inode->i_ino, EXT4_I(inode));
1178 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1179 EXT4_I(inode), sizeof(struct ext4_inode_info),
1185 static void init_once(void *foo)
1187 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
1189 INIT_LIST_HEAD(&ei->i_orphan);
1190 init_rwsem(&ei->xattr_sem);
1191 init_rwsem(&ei->i_data_sem);
1192 init_rwsem(&ei->i_mmap_sem);
1193 inode_init_once(&ei->vfs_inode);
1196 static int __init init_inodecache(void)
1198 ext4_inode_cachep = kmem_cache_create_usercopy("ext4_inode_cache",
1199 sizeof(struct ext4_inode_info), 0,
1200 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD|
1202 offsetof(struct ext4_inode_info, i_data),
1203 sizeof_field(struct ext4_inode_info, i_data),
1205 if (ext4_inode_cachep == NULL)
1210 static void destroy_inodecache(void)
1213 * Make sure all delayed rcu free inodes are flushed before we
1217 kmem_cache_destroy(ext4_inode_cachep);
1220 void ext4_clear_inode(struct inode *inode)
1222 invalidate_inode_buffers(inode);
1224 ext4_discard_preallocations(inode, 0);
1225 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
1227 if (EXT4_I(inode)->jinode) {
1228 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1229 EXT4_I(inode)->jinode);
1230 jbd2_free_inode(EXT4_I(inode)->jinode);
1231 EXT4_I(inode)->jinode = NULL;
1233 fscrypt_put_encryption_info(inode);
1234 fsverity_cleanup_inode(inode);
1237 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1238 u64 ino, u32 generation)
1240 struct inode *inode;
1243 * Currently we don't know the generation for parent directory, so
1244 * a generation of 0 means "accept any"
1246 inode = ext4_iget(sb, ino, EXT4_IGET_HANDLE);
1248 return ERR_CAST(inode);
1249 if (generation && inode->i_generation != generation) {
1251 return ERR_PTR(-ESTALE);
1257 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1258 int fh_len, int fh_type)
1260 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1261 ext4_nfs_get_inode);
1264 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1265 int fh_len, int fh_type)
1267 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1268 ext4_nfs_get_inode);
1271 static int ext4_nfs_commit_metadata(struct inode *inode)
1273 struct writeback_control wbc = {
1274 .sync_mode = WB_SYNC_ALL
1277 trace_ext4_nfs_commit_metadata(inode);
1278 return ext4_write_inode(inode, &wbc);
1282 * Try to release metadata pages (indirect blocks, directories) which are
1283 * mapped via the block device. Since these pages could have journal heads
1284 * which would prevent try_to_free_buffers() from freeing them, we must use
1285 * jbd2 layer's try_to_free_buffers() function to release them.
1287 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1290 journal_t *journal = EXT4_SB(sb)->s_journal;
1292 WARN_ON(PageChecked(page));
1293 if (!page_has_buffers(page))
1296 return jbd2_journal_try_to_free_buffers(journal, page);
1298 return try_to_free_buffers(page);
1301 #ifdef CONFIG_FS_ENCRYPTION
1302 static int ext4_get_context(struct inode *inode, void *ctx, size_t len)
1304 return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1305 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len);
1308 static int ext4_set_context(struct inode *inode, const void *ctx, size_t len,
1311 handle_t *handle = fs_data;
1312 int res, res2, credits, retries = 0;
1315 * Encrypting the root directory is not allowed because e2fsck expects
1316 * lost+found to exist and be unencrypted, and encrypting the root
1317 * directory would imply encrypting the lost+found directory as well as
1318 * the filename "lost+found" itself.
1320 if (inode->i_ino == EXT4_ROOT_INO)
1323 if (WARN_ON_ONCE(IS_DAX(inode) && i_size_read(inode)))
1326 if (ext4_test_inode_flag(inode, EXT4_INODE_DAX))
1329 res = ext4_convert_inline_data(inode);
1334 * If a journal handle was specified, then the encryption context is
1335 * being set on a new inode via inheritance and is part of a larger
1336 * transaction to create the inode. Otherwise the encryption context is
1337 * being set on an existing inode in its own transaction. Only in the
1338 * latter case should the "retry on ENOSPC" logic be used.
1342 res = ext4_xattr_set_handle(handle, inode,
1343 EXT4_XATTR_INDEX_ENCRYPTION,
1344 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1347 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1348 ext4_clear_inode_state(inode,
1349 EXT4_STATE_MAY_INLINE_DATA);
1351 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1352 * S_DAX may be disabled
1354 ext4_set_inode_flags(inode, false);
1359 res = dquot_initialize(inode);
1363 res = ext4_xattr_set_credits(inode, len, false /* is_create */,
1368 handle = ext4_journal_start(inode, EXT4_HT_MISC, credits);
1370 return PTR_ERR(handle);
1372 res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION,
1373 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1376 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1378 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1379 * S_DAX may be disabled
1381 ext4_set_inode_flags(inode, false);
1382 res = ext4_mark_inode_dirty(handle, inode);
1384 EXT4_ERROR_INODE(inode, "Failed to mark inode dirty");
1386 res2 = ext4_journal_stop(handle);
1388 if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
1395 static const union fscrypt_context *
1396 ext4_get_dummy_context(struct super_block *sb)
1398 return EXT4_SB(sb)->s_dummy_enc_ctx.ctx;
1401 static bool ext4_has_stable_inodes(struct super_block *sb)
1403 return ext4_has_feature_stable_inodes(sb);
1406 static void ext4_get_ino_and_lblk_bits(struct super_block *sb,
1407 int *ino_bits_ret, int *lblk_bits_ret)
1409 *ino_bits_ret = 8 * sizeof(EXT4_SB(sb)->s_es->s_inodes_count);
1410 *lblk_bits_ret = 8 * sizeof(ext4_lblk_t);
1413 static const struct fscrypt_operations ext4_cryptops = {
1414 .key_prefix = "ext4:",
1415 .get_context = ext4_get_context,
1416 .set_context = ext4_set_context,
1417 .get_dummy_context = ext4_get_dummy_context,
1418 .empty_dir = ext4_empty_dir,
1419 .max_namelen = EXT4_NAME_LEN,
1420 .has_stable_inodes = ext4_has_stable_inodes,
1421 .get_ino_and_lblk_bits = ext4_get_ino_and_lblk_bits,
1426 static const char * const quotatypes[] = INITQFNAMES;
1427 #define QTYPE2NAME(t) (quotatypes[t])
1429 static int ext4_write_dquot(struct dquot *dquot);
1430 static int ext4_acquire_dquot(struct dquot *dquot);
1431 static int ext4_release_dquot(struct dquot *dquot);
1432 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1433 static int ext4_write_info(struct super_block *sb, int type);
1434 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1435 const struct path *path);
1436 static int ext4_quota_on_mount(struct super_block *sb, int type);
1437 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1438 size_t len, loff_t off);
1439 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1440 const char *data, size_t len, loff_t off);
1441 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1442 unsigned int flags);
1443 static int ext4_enable_quotas(struct super_block *sb);
1445 static struct dquot **ext4_get_dquots(struct inode *inode)
1447 return EXT4_I(inode)->i_dquot;
1450 static const struct dquot_operations ext4_quota_operations = {
1451 .get_reserved_space = ext4_get_reserved_space,
1452 .write_dquot = ext4_write_dquot,
1453 .acquire_dquot = ext4_acquire_dquot,
1454 .release_dquot = ext4_release_dquot,
1455 .mark_dirty = ext4_mark_dquot_dirty,
1456 .write_info = ext4_write_info,
1457 .alloc_dquot = dquot_alloc,
1458 .destroy_dquot = dquot_destroy,
1459 .get_projid = ext4_get_projid,
1460 .get_inode_usage = ext4_get_inode_usage,
1461 .get_next_id = dquot_get_next_id,
1464 static const struct quotactl_ops ext4_qctl_operations = {
1465 .quota_on = ext4_quota_on,
1466 .quota_off = ext4_quota_off,
1467 .quota_sync = dquot_quota_sync,
1468 .get_state = dquot_get_state,
1469 .set_info = dquot_set_dqinfo,
1470 .get_dqblk = dquot_get_dqblk,
1471 .set_dqblk = dquot_set_dqblk,
1472 .get_nextdqblk = dquot_get_next_dqblk,
1476 static const struct super_operations ext4_sops = {
1477 .alloc_inode = ext4_alloc_inode,
1478 .free_inode = ext4_free_in_core_inode,
1479 .destroy_inode = ext4_destroy_inode,
1480 .write_inode = ext4_write_inode,
1481 .dirty_inode = ext4_dirty_inode,
1482 .drop_inode = ext4_drop_inode,
1483 .evict_inode = ext4_evict_inode,
1484 .put_super = ext4_put_super,
1485 .sync_fs = ext4_sync_fs,
1486 .freeze_fs = ext4_freeze,
1487 .unfreeze_fs = ext4_unfreeze,
1488 .statfs = ext4_statfs,
1489 .remount_fs = ext4_remount,
1490 .show_options = ext4_show_options,
1492 .quota_read = ext4_quota_read,
1493 .quota_write = ext4_quota_write,
1494 .get_dquots = ext4_get_dquots,
1496 .bdev_try_to_free_page = bdev_try_to_free_page,
1499 static const struct export_operations ext4_export_ops = {
1500 .fh_to_dentry = ext4_fh_to_dentry,
1501 .fh_to_parent = ext4_fh_to_parent,
1502 .get_parent = ext4_get_parent,
1503 .commit_metadata = ext4_nfs_commit_metadata,
1507 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1508 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1509 Opt_nouid32, Opt_debug, Opt_removed,
1510 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1511 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1512 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1513 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1514 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1515 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1516 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1517 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1518 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1519 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version,
1520 Opt_dax, Opt_dax_always, Opt_dax_inode, Opt_dax_never,
1521 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_warn_on_error,
1522 Opt_nowarn_on_error, Opt_mblk_io_submit,
1523 Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize,
1524 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1525 Opt_inode_readahead_blks, Opt_journal_ioprio,
1526 Opt_dioread_nolock, Opt_dioread_lock,
1527 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1528 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache,
1529 Opt_prefetch_block_bitmaps,
1532 static const match_table_t tokens = {
1533 {Opt_bsd_df, "bsddf"},
1534 {Opt_minix_df, "minixdf"},
1535 {Opt_grpid, "grpid"},
1536 {Opt_grpid, "bsdgroups"},
1537 {Opt_nogrpid, "nogrpid"},
1538 {Opt_nogrpid, "sysvgroups"},
1539 {Opt_resgid, "resgid=%u"},
1540 {Opt_resuid, "resuid=%u"},
1542 {Opt_err_cont, "errors=continue"},
1543 {Opt_err_panic, "errors=panic"},
1544 {Opt_err_ro, "errors=remount-ro"},
1545 {Opt_nouid32, "nouid32"},
1546 {Opt_debug, "debug"},
1547 {Opt_removed, "oldalloc"},
1548 {Opt_removed, "orlov"},
1549 {Opt_user_xattr, "user_xattr"},
1550 {Opt_nouser_xattr, "nouser_xattr"},
1552 {Opt_noacl, "noacl"},
1553 {Opt_noload, "norecovery"},
1554 {Opt_noload, "noload"},
1555 {Opt_removed, "nobh"},
1556 {Opt_removed, "bh"},
1557 {Opt_commit, "commit=%u"},
1558 {Opt_min_batch_time, "min_batch_time=%u"},
1559 {Opt_max_batch_time, "max_batch_time=%u"},
1560 {Opt_journal_dev, "journal_dev=%u"},
1561 {Opt_journal_path, "journal_path=%s"},
1562 {Opt_journal_checksum, "journal_checksum"},
1563 {Opt_nojournal_checksum, "nojournal_checksum"},
1564 {Opt_journal_async_commit, "journal_async_commit"},
1565 {Opt_abort, "abort"},
1566 {Opt_data_journal, "data=journal"},
1567 {Opt_data_ordered, "data=ordered"},
1568 {Opt_data_writeback, "data=writeback"},
1569 {Opt_data_err_abort, "data_err=abort"},
1570 {Opt_data_err_ignore, "data_err=ignore"},
1571 {Opt_offusrjquota, "usrjquota="},
1572 {Opt_usrjquota, "usrjquota=%s"},
1573 {Opt_offgrpjquota, "grpjquota="},
1574 {Opt_grpjquota, "grpjquota=%s"},
1575 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1576 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1577 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1578 {Opt_grpquota, "grpquota"},
1579 {Opt_noquota, "noquota"},
1580 {Opt_quota, "quota"},
1581 {Opt_usrquota, "usrquota"},
1582 {Opt_prjquota, "prjquota"},
1583 {Opt_barrier, "barrier=%u"},
1584 {Opt_barrier, "barrier"},
1585 {Opt_nobarrier, "nobarrier"},
1586 {Opt_i_version, "i_version"},
1588 {Opt_dax_always, "dax=always"},
1589 {Opt_dax_inode, "dax=inode"},
1590 {Opt_dax_never, "dax=never"},
1591 {Opt_stripe, "stripe=%u"},
1592 {Opt_delalloc, "delalloc"},
1593 {Opt_warn_on_error, "warn_on_error"},
1594 {Opt_nowarn_on_error, "nowarn_on_error"},
1595 {Opt_lazytime, "lazytime"},
1596 {Opt_nolazytime, "nolazytime"},
1597 {Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"},
1598 {Opt_nodelalloc, "nodelalloc"},
1599 {Opt_removed, "mblk_io_submit"},
1600 {Opt_removed, "nomblk_io_submit"},
1601 {Opt_block_validity, "block_validity"},
1602 {Opt_noblock_validity, "noblock_validity"},
1603 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1604 {Opt_journal_ioprio, "journal_ioprio=%u"},
1605 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1606 {Opt_auto_da_alloc, "auto_da_alloc"},
1607 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1608 {Opt_dioread_nolock, "dioread_nolock"},
1609 {Opt_dioread_lock, "nodioread_nolock"},
1610 {Opt_dioread_lock, "dioread_lock"},
1611 {Opt_discard, "discard"},
1612 {Opt_nodiscard, "nodiscard"},
1613 {Opt_init_itable, "init_itable=%u"},
1614 {Opt_init_itable, "init_itable"},
1615 {Opt_noinit_itable, "noinit_itable"},
1616 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1617 {Opt_test_dummy_encryption, "test_dummy_encryption=%s"},
1618 {Opt_test_dummy_encryption, "test_dummy_encryption"},
1619 {Opt_nombcache, "nombcache"},
1620 {Opt_nombcache, "no_mbcache"}, /* for backward compatibility */
1621 {Opt_prefetch_block_bitmaps, "prefetch_block_bitmaps"},
1622 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1623 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1624 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1625 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1626 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1630 static ext4_fsblk_t get_sb_block(void **data)
1632 ext4_fsblk_t sb_block;
1633 char *options = (char *) *data;
1635 if (!options || strncmp(options, "sb=", 3) != 0)
1636 return 1; /* Default location */
1639 /* TODO: use simple_strtoll with >32bit ext4 */
1640 sb_block = simple_strtoul(options, &options, 0);
1641 if (*options && *options != ',') {
1642 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1646 if (*options == ',')
1648 *data = (void *) options;
1653 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1654 static const char deprecated_msg[] =
1655 "Mount option \"%s\" will be removed by %s\n"
1656 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1659 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1661 struct ext4_sb_info *sbi = EXT4_SB(sb);
1662 char *qname, *old_qname = get_qf_name(sb, sbi, qtype);
1665 if (sb_any_quota_loaded(sb) && !old_qname) {
1666 ext4_msg(sb, KERN_ERR,
1667 "Cannot change journaled "
1668 "quota options when quota turned on");
1671 if (ext4_has_feature_quota(sb)) {
1672 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1673 "ignored when QUOTA feature is enabled");
1676 qname = match_strdup(args);
1678 ext4_msg(sb, KERN_ERR,
1679 "Not enough memory for storing quotafile name");
1683 if (strcmp(old_qname, qname) == 0)
1686 ext4_msg(sb, KERN_ERR,
1687 "%s quota file already specified",
1691 if (strchr(qname, '/')) {
1692 ext4_msg(sb, KERN_ERR,
1693 "quotafile must be on filesystem root");
1696 rcu_assign_pointer(sbi->s_qf_names[qtype], qname);
1704 static int clear_qf_name(struct super_block *sb, int qtype)
1707 struct ext4_sb_info *sbi = EXT4_SB(sb);
1708 char *old_qname = get_qf_name(sb, sbi, qtype);
1710 if (sb_any_quota_loaded(sb) && old_qname) {
1711 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1712 " when quota turned on");
1715 rcu_assign_pointer(sbi->s_qf_names[qtype], NULL);
1722 #define MOPT_SET 0x0001
1723 #define MOPT_CLEAR 0x0002
1724 #define MOPT_NOSUPPORT 0x0004
1725 #define MOPT_EXPLICIT 0x0008
1726 #define MOPT_CLEAR_ERR 0x0010
1727 #define MOPT_GTE0 0x0020
1730 #define MOPT_QFMT 0x0040
1732 #define MOPT_Q MOPT_NOSUPPORT
1733 #define MOPT_QFMT MOPT_NOSUPPORT
1735 #define MOPT_DATAJ 0x0080
1736 #define MOPT_NO_EXT2 0x0100
1737 #define MOPT_NO_EXT3 0x0200
1738 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1739 #define MOPT_STRING 0x0400
1740 #define MOPT_SKIP 0x0800
1742 static const struct mount_opts {
1746 } ext4_mount_opts[] = {
1747 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1748 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1749 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1750 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1751 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1752 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1753 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1754 MOPT_EXT4_ONLY | MOPT_SET},
1755 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1756 MOPT_EXT4_ONLY | MOPT_CLEAR},
1757 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1758 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1759 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1760 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1761 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1762 MOPT_EXT4_ONLY | MOPT_CLEAR},
1763 {Opt_warn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_SET},
1764 {Opt_nowarn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_CLEAR},
1765 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1766 MOPT_EXT4_ONLY | MOPT_CLEAR},
1767 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1768 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1769 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1770 EXT4_MOUNT_JOURNAL_CHECKSUM),
1771 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1772 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1773 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1774 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1775 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1776 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1778 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1780 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1781 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1782 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1783 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1784 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1785 {Opt_commit, 0, MOPT_GTE0},
1786 {Opt_max_batch_time, 0, MOPT_GTE0},
1787 {Opt_min_batch_time, 0, MOPT_GTE0},
1788 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1789 {Opt_init_itable, 0, MOPT_GTE0},
1790 {Opt_dax, EXT4_MOUNT_DAX_ALWAYS, MOPT_SET | MOPT_SKIP},
1791 {Opt_dax_always, EXT4_MOUNT_DAX_ALWAYS,
1792 MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP},
1793 {Opt_dax_inode, EXT4_MOUNT2_DAX_INODE,
1794 MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP},
1795 {Opt_dax_never, EXT4_MOUNT2_DAX_NEVER,
1796 MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP},
1797 {Opt_stripe, 0, MOPT_GTE0},
1798 {Opt_resuid, 0, MOPT_GTE0},
1799 {Opt_resgid, 0, MOPT_GTE0},
1800 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1801 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1802 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1803 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1804 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1805 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1806 MOPT_NO_EXT2 | MOPT_DATAJ},
1807 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1808 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1809 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1810 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1811 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1813 {Opt_acl, 0, MOPT_NOSUPPORT},
1814 {Opt_noacl, 0, MOPT_NOSUPPORT},
1816 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1817 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1818 {Opt_debug_want_extra_isize, 0, MOPT_GTE0},
1819 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1820 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1822 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1824 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1826 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1827 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1828 MOPT_CLEAR | MOPT_Q},
1829 {Opt_usrjquota, 0, MOPT_Q},
1830 {Opt_grpjquota, 0, MOPT_Q},
1831 {Opt_offusrjquota, 0, MOPT_Q},
1832 {Opt_offgrpjquota, 0, MOPT_Q},
1833 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1834 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1835 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1836 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1837 {Opt_test_dummy_encryption, 0, MOPT_STRING},
1838 {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET},
1839 {Opt_prefetch_block_bitmaps, EXT4_MOUNT_PREFETCH_BLOCK_BITMAPS,
1844 #ifdef CONFIG_UNICODE
1845 static const struct ext4_sb_encodings {
1849 } ext4_sb_encoding_map[] = {
1850 {EXT4_ENC_UTF8_12_1, "utf8", "12.1.0"},
1853 static int ext4_sb_read_encoding(const struct ext4_super_block *es,
1854 const struct ext4_sb_encodings **encoding,
1857 __u16 magic = le16_to_cpu(es->s_encoding);
1860 for (i = 0; i < ARRAY_SIZE(ext4_sb_encoding_map); i++)
1861 if (magic == ext4_sb_encoding_map[i].magic)
1864 if (i >= ARRAY_SIZE(ext4_sb_encoding_map))
1867 *encoding = &ext4_sb_encoding_map[i];
1868 *flags = le16_to_cpu(es->s_encoding_flags);
1874 static int ext4_set_test_dummy_encryption(struct super_block *sb,
1876 const substring_t *arg,
1879 #ifdef CONFIG_FS_ENCRYPTION
1880 struct ext4_sb_info *sbi = EXT4_SB(sb);
1884 * This mount option is just for testing, and it's not worthwhile to
1885 * implement the extra complexity (e.g. RCU protection) that would be
1886 * needed to allow it to be set or changed during remount. We do allow
1887 * it to be specified during remount, but only if there is no change.
1889 if (is_remount && !sbi->s_dummy_enc_ctx.ctx) {
1890 ext4_msg(sb, KERN_WARNING,
1891 "Can't set test_dummy_encryption on remount");
1894 err = fscrypt_set_test_dummy_encryption(sb, arg, &sbi->s_dummy_enc_ctx);
1897 ext4_msg(sb, KERN_WARNING,
1898 "Can't change test_dummy_encryption on remount");
1899 else if (err == -EINVAL)
1900 ext4_msg(sb, KERN_WARNING,
1901 "Value of option \"%s\" is unrecognized", opt);
1903 ext4_msg(sb, KERN_WARNING,
1904 "Error processing option \"%s\" [%d]",
1908 ext4_msg(sb, KERN_WARNING, "Test dummy encryption mode enabled");
1910 ext4_msg(sb, KERN_WARNING,
1911 "Test dummy encryption mount option ignored");
1916 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1917 substring_t *args, unsigned long *journal_devnum,
1918 unsigned int *journal_ioprio, int is_remount)
1920 struct ext4_sb_info *sbi = EXT4_SB(sb);
1921 const struct mount_opts *m;
1927 if (token == Opt_usrjquota)
1928 return set_qf_name(sb, USRQUOTA, &args[0]);
1929 else if (token == Opt_grpjquota)
1930 return set_qf_name(sb, GRPQUOTA, &args[0]);
1931 else if (token == Opt_offusrjquota)
1932 return clear_qf_name(sb, USRQUOTA);
1933 else if (token == Opt_offgrpjquota)
1934 return clear_qf_name(sb, GRPQUOTA);
1938 case Opt_nouser_xattr:
1939 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1942 return 1; /* handled by get_sb_block() */
1944 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1947 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1950 sb->s_flags |= SB_I_VERSION;
1953 sb->s_flags |= SB_LAZYTIME;
1955 case Opt_nolazytime:
1956 sb->s_flags &= ~SB_LAZYTIME;
1960 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1961 if (token == m->token)
1964 if (m->token == Opt_err) {
1965 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1966 "or missing value", opt);
1970 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1971 ext4_msg(sb, KERN_ERR,
1972 "Mount option \"%s\" incompatible with ext2", opt);
1975 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1976 ext4_msg(sb, KERN_ERR,
1977 "Mount option \"%s\" incompatible with ext3", opt);
1981 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
1983 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1985 if (m->flags & MOPT_EXPLICIT) {
1986 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
1987 set_opt2(sb, EXPLICIT_DELALLOC);
1988 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
1989 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
1993 if (m->flags & MOPT_CLEAR_ERR)
1994 clear_opt(sb, ERRORS_MASK);
1995 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1996 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1997 "options when quota turned on");
2001 if (m->flags & MOPT_NOSUPPORT) {
2002 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
2003 } else if (token == Opt_commit) {
2005 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
2006 else if (arg > INT_MAX / HZ) {
2007 ext4_msg(sb, KERN_ERR,
2008 "Invalid commit interval %d, "
2009 "must be smaller than %d",
2013 sbi->s_commit_interval = HZ * arg;
2014 } else if (token == Opt_debug_want_extra_isize) {
2017 (arg > (sbi->s_inode_size - EXT4_GOOD_OLD_INODE_SIZE))) {
2018 ext4_msg(sb, KERN_ERR,
2019 "Invalid want_extra_isize %d", arg);
2022 sbi->s_want_extra_isize = arg;
2023 } else if (token == Opt_max_batch_time) {
2024 sbi->s_max_batch_time = arg;
2025 } else if (token == Opt_min_batch_time) {
2026 sbi->s_min_batch_time = arg;
2027 } else if (token == Opt_inode_readahead_blks) {
2028 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
2029 ext4_msg(sb, KERN_ERR,
2030 "EXT4-fs: inode_readahead_blks must be "
2031 "0 or a power of 2 smaller than 2^31");
2034 sbi->s_inode_readahead_blks = arg;
2035 } else if (token == Opt_init_itable) {
2036 set_opt(sb, INIT_INODE_TABLE);
2038 arg = EXT4_DEF_LI_WAIT_MULT;
2039 sbi->s_li_wait_mult = arg;
2040 } else if (token == Opt_max_dir_size_kb) {
2041 sbi->s_max_dir_size_kb = arg;
2042 } else if (token == Opt_stripe) {
2043 sbi->s_stripe = arg;
2044 } else if (token == Opt_resuid) {
2045 uid = make_kuid(current_user_ns(), arg);
2046 if (!uid_valid(uid)) {
2047 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
2050 sbi->s_resuid = uid;
2051 } else if (token == Opt_resgid) {
2052 gid = make_kgid(current_user_ns(), arg);
2053 if (!gid_valid(gid)) {
2054 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
2057 sbi->s_resgid = gid;
2058 } else if (token == Opt_journal_dev) {
2060 ext4_msg(sb, KERN_ERR,
2061 "Cannot specify journal on remount");
2064 *journal_devnum = arg;
2065 } else if (token == Opt_journal_path) {
2067 struct inode *journal_inode;
2072 ext4_msg(sb, KERN_ERR,
2073 "Cannot specify journal on remount");
2076 journal_path = match_strdup(&args[0]);
2077 if (!journal_path) {
2078 ext4_msg(sb, KERN_ERR, "error: could not dup "
2079 "journal device string");
2083 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
2085 ext4_msg(sb, KERN_ERR, "error: could not find "
2086 "journal device path: error %d", error);
2087 kfree(journal_path);
2091 journal_inode = d_inode(path.dentry);
2092 if (!S_ISBLK(journal_inode->i_mode)) {
2093 ext4_msg(sb, KERN_ERR, "error: journal path %s "
2094 "is not a block device", journal_path);
2096 kfree(journal_path);
2100 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
2102 kfree(journal_path);
2103 } else if (token == Opt_journal_ioprio) {
2105 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
2110 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
2111 } else if (token == Opt_test_dummy_encryption) {
2112 return ext4_set_test_dummy_encryption(sb, opt, &args[0],
2114 } else if (m->flags & MOPT_DATAJ) {
2116 if (!sbi->s_journal)
2117 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
2118 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
2119 ext4_msg(sb, KERN_ERR,
2120 "Cannot change data mode on remount");
2124 clear_opt(sb, DATA_FLAGS);
2125 sbi->s_mount_opt |= m->mount_opt;
2128 } else if (m->flags & MOPT_QFMT) {
2129 if (sb_any_quota_loaded(sb) &&
2130 sbi->s_jquota_fmt != m->mount_opt) {
2131 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
2132 "quota options when quota turned on");
2135 if (ext4_has_feature_quota(sb)) {
2136 ext4_msg(sb, KERN_INFO,
2137 "Quota format mount options ignored "
2138 "when QUOTA feature is enabled");
2141 sbi->s_jquota_fmt = m->mount_opt;
2143 } else if (token == Opt_dax || token == Opt_dax_always ||
2144 token == Opt_dax_inode || token == Opt_dax_never) {
2145 #ifdef CONFIG_FS_DAX
2148 case Opt_dax_always:
2150 (!(sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2151 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER))) {
2152 fail_dax_change_remount:
2153 ext4_msg(sb, KERN_ERR, "can't change "
2154 "dax mount option while remounting");
2158 (test_opt(sb, DATA_FLAGS) ==
2159 EXT4_MOUNT_JOURNAL_DATA)) {
2160 ext4_msg(sb, KERN_ERR, "can't mount with "
2161 "both data=journal and dax");
2164 ext4_msg(sb, KERN_WARNING,
2165 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
2166 sbi->s_mount_opt |= EXT4_MOUNT_DAX_ALWAYS;
2167 sbi->s_mount_opt2 &= ~EXT4_MOUNT2_DAX_NEVER;
2171 (!(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2172 (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS)))
2173 goto fail_dax_change_remount;
2174 sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_NEVER;
2175 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS;
2179 ((sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2180 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2181 !(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_INODE)))
2182 goto fail_dax_change_remount;
2183 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS;
2184 sbi->s_mount_opt2 &= ~EXT4_MOUNT2_DAX_NEVER;
2185 /* Strictly for printing options */
2186 sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_INODE;
2190 ext4_msg(sb, KERN_INFO, "dax option not supported");
2191 sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_NEVER;
2192 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS;
2195 } else if (token == Opt_data_err_abort) {
2196 sbi->s_mount_opt |= m->mount_opt;
2197 } else if (token == Opt_data_err_ignore) {
2198 sbi->s_mount_opt &= ~m->mount_opt;
2202 if (m->flags & MOPT_CLEAR)
2204 else if (unlikely(!(m->flags & MOPT_SET))) {
2205 ext4_msg(sb, KERN_WARNING,
2206 "buggy handling of option %s", opt);
2211 sbi->s_mount_opt |= m->mount_opt;
2213 sbi->s_mount_opt &= ~m->mount_opt;
2218 static int parse_options(char *options, struct super_block *sb,
2219 unsigned long *journal_devnum,
2220 unsigned int *journal_ioprio,
2223 struct ext4_sb_info __maybe_unused *sbi = EXT4_SB(sb);
2224 char *p, __maybe_unused *usr_qf_name, __maybe_unused *grp_qf_name;
2225 substring_t args[MAX_OPT_ARGS];
2231 while ((p = strsep(&options, ",")) != NULL) {
2235 * Initialize args struct so we know whether arg was
2236 * found; some options take optional arguments.
2238 args[0].to = args[0].from = NULL;
2239 token = match_token(p, tokens, args);
2240 if (handle_mount_opt(sb, p, token, args, journal_devnum,
2241 journal_ioprio, is_remount) < 0)
2246 * We do the test below only for project quotas. 'usrquota' and
2247 * 'grpquota' mount options are allowed even without quota feature
2248 * to support legacy quotas in quota files.
2250 if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) {
2251 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. "
2252 "Cannot enable project quota enforcement.");
2255 usr_qf_name = get_qf_name(sb, sbi, USRQUOTA);
2256 grp_qf_name = get_qf_name(sb, sbi, GRPQUOTA);
2257 if (usr_qf_name || grp_qf_name) {
2258 if (test_opt(sb, USRQUOTA) && usr_qf_name)
2259 clear_opt(sb, USRQUOTA);
2261 if (test_opt(sb, GRPQUOTA) && grp_qf_name)
2262 clear_opt(sb, GRPQUOTA);
2264 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
2265 ext4_msg(sb, KERN_ERR, "old and new quota "
2270 if (!sbi->s_jquota_fmt) {
2271 ext4_msg(sb, KERN_ERR, "journaled quota format "
2277 if (test_opt(sb, DIOREAD_NOLOCK)) {
2279 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
2280 if (blocksize < PAGE_SIZE)
2281 ext4_msg(sb, KERN_WARNING, "Warning: mounting with an "
2282 "experimental mount option 'dioread_nolock' "
2283 "for blocksize < PAGE_SIZE");
2288 static inline void ext4_show_quota_options(struct seq_file *seq,
2289 struct super_block *sb)
2291 #if defined(CONFIG_QUOTA)
2292 struct ext4_sb_info *sbi = EXT4_SB(sb);
2293 char *usr_qf_name, *grp_qf_name;
2295 if (sbi->s_jquota_fmt) {
2298 switch (sbi->s_jquota_fmt) {
2309 seq_printf(seq, ",jqfmt=%s", fmtname);
2313 usr_qf_name = rcu_dereference(sbi->s_qf_names[USRQUOTA]);
2314 grp_qf_name = rcu_dereference(sbi->s_qf_names[GRPQUOTA]);
2316 seq_show_option(seq, "usrjquota", usr_qf_name);
2318 seq_show_option(seq, "grpjquota", grp_qf_name);
2323 static const char *token2str(int token)
2325 const struct match_token *t;
2327 for (t = tokens; t->token != Opt_err; t++)
2328 if (t->token == token && !strchr(t->pattern, '='))
2335 * - it's set to a non-default value OR
2336 * - if the per-sb default is different from the global default
2338 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
2341 struct ext4_sb_info *sbi = EXT4_SB(sb);
2342 struct ext4_super_block *es = sbi->s_es;
2343 int def_errors, def_mount_opt = sbi->s_def_mount_opt;
2344 const struct mount_opts *m;
2345 char sep = nodefs ? '\n' : ',';
2347 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2348 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2350 if (sbi->s_sb_block != 1)
2351 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
2353 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
2354 int want_set = m->flags & MOPT_SET;
2355 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
2356 (m->flags & MOPT_CLEAR_ERR) || m->flags & MOPT_SKIP)
2358 if (!nodefs && !(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
2359 continue; /* skip if same as the default */
2361 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
2362 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
2363 continue; /* select Opt_noFoo vs Opt_Foo */
2364 SEQ_OPTS_PRINT("%s", token2str(m->token));
2367 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
2368 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
2369 SEQ_OPTS_PRINT("resuid=%u",
2370 from_kuid_munged(&init_user_ns, sbi->s_resuid));
2371 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
2372 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
2373 SEQ_OPTS_PRINT("resgid=%u",
2374 from_kgid_munged(&init_user_ns, sbi->s_resgid));
2375 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
2376 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
2377 SEQ_OPTS_PUTS("errors=remount-ro");
2378 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
2379 SEQ_OPTS_PUTS("errors=continue");
2380 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
2381 SEQ_OPTS_PUTS("errors=panic");
2382 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
2383 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
2384 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
2385 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
2386 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
2387 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2388 if (sb->s_flags & SB_I_VERSION)
2389 SEQ_OPTS_PUTS("i_version");
2390 if (nodefs || sbi->s_stripe)
2391 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
2392 if (nodefs || EXT4_MOUNT_DATA_FLAGS &
2393 (sbi->s_mount_opt ^ def_mount_opt)) {
2394 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2395 SEQ_OPTS_PUTS("data=journal");
2396 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2397 SEQ_OPTS_PUTS("data=ordered");
2398 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2399 SEQ_OPTS_PUTS("data=writeback");
2402 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
2403 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2404 sbi->s_inode_readahead_blks);
2406 if (test_opt(sb, INIT_INODE_TABLE) && (nodefs ||
2407 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
2408 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
2409 if (nodefs || sbi->s_max_dir_size_kb)
2410 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2411 if (test_opt(sb, DATA_ERR_ABORT))
2412 SEQ_OPTS_PUTS("data_err=abort");
2414 fscrypt_show_test_dummy_encryption(seq, sep, sb);
2416 if (test_opt(sb, DAX_ALWAYS)) {
2418 SEQ_OPTS_PUTS("dax");
2420 SEQ_OPTS_PUTS("dax=always");
2421 } else if (test_opt2(sb, DAX_NEVER)) {
2422 SEQ_OPTS_PUTS("dax=never");
2423 } else if (test_opt2(sb, DAX_INODE)) {
2424 SEQ_OPTS_PUTS("dax=inode");
2427 ext4_show_quota_options(seq, sb);
2431 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2433 return _ext4_show_options(seq, root->d_sb, 0);
2436 int ext4_seq_options_show(struct seq_file *seq, void *offset)
2438 struct super_block *sb = seq->private;
2441 seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw");
2442 rc = _ext4_show_options(seq, sb, 1);
2443 seq_puts(seq, "\n");
2447 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
2450 struct ext4_sb_info *sbi = EXT4_SB(sb);
2453 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
2454 ext4_msg(sb, KERN_ERR, "revision level too high, "
2455 "forcing read-only mode");
2461 if (!(sbi->s_mount_state & EXT4_VALID_FS))
2462 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
2463 "running e2fsck is recommended");
2464 else if (sbi->s_mount_state & EXT4_ERROR_FS)
2465 ext4_msg(sb, KERN_WARNING,
2466 "warning: mounting fs with errors, "
2467 "running e2fsck is recommended");
2468 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
2469 le16_to_cpu(es->s_mnt_count) >=
2470 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
2471 ext4_msg(sb, KERN_WARNING,
2472 "warning: maximal mount count reached, "
2473 "running e2fsck is recommended");
2474 else if (le32_to_cpu(es->s_checkinterval) &&
2475 (ext4_get_tstamp(es, s_lastcheck) +
2476 le32_to_cpu(es->s_checkinterval) <= ktime_get_real_seconds()))
2477 ext4_msg(sb, KERN_WARNING,
2478 "warning: checktime reached, "
2479 "running e2fsck is recommended");
2480 if (!sbi->s_journal)
2481 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
2482 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
2483 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
2484 le16_add_cpu(&es->s_mnt_count, 1);
2485 ext4_update_tstamp(es, s_mtime);
2487 ext4_set_feature_journal_needs_recovery(sb);
2489 err = ext4_commit_super(sb, 1);
2491 if (test_opt(sb, DEBUG))
2492 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
2493 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
2495 sbi->s_groups_count,
2496 EXT4_BLOCKS_PER_GROUP(sb),
2497 EXT4_INODES_PER_GROUP(sb),
2498 sbi->s_mount_opt, sbi->s_mount_opt2);
2500 cleancache_init_fs(sb);
2504 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2506 struct ext4_sb_info *sbi = EXT4_SB(sb);
2507 struct flex_groups **old_groups, **new_groups;
2510 if (!sbi->s_log_groups_per_flex)
2513 size = ext4_flex_group(sbi, ngroup - 1) + 1;
2514 if (size <= sbi->s_flex_groups_allocated)
2517 new_groups = kvzalloc(roundup_pow_of_two(size *
2518 sizeof(*sbi->s_flex_groups)), GFP_KERNEL);
2520 ext4_msg(sb, KERN_ERR,
2521 "not enough memory for %d flex group pointers", size);
2524 for (i = sbi->s_flex_groups_allocated; i < size; i++) {
2525 new_groups[i] = kvzalloc(roundup_pow_of_two(
2526 sizeof(struct flex_groups)),
2528 if (!new_groups[i]) {
2529 for (j = sbi->s_flex_groups_allocated; j < i; j++)
2530 kvfree(new_groups[j]);
2532 ext4_msg(sb, KERN_ERR,
2533 "not enough memory for %d flex groups", size);
2538 old_groups = rcu_dereference(sbi->s_flex_groups);
2540 memcpy(new_groups, old_groups,
2541 (sbi->s_flex_groups_allocated *
2542 sizeof(struct flex_groups *)));
2544 rcu_assign_pointer(sbi->s_flex_groups, new_groups);
2545 sbi->s_flex_groups_allocated = size;
2547 ext4_kvfree_array_rcu(old_groups);
2551 static int ext4_fill_flex_info(struct super_block *sb)
2553 struct ext4_sb_info *sbi = EXT4_SB(sb);
2554 struct ext4_group_desc *gdp = NULL;
2555 struct flex_groups *fg;
2556 ext4_group_t flex_group;
2559 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2560 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
2561 sbi->s_log_groups_per_flex = 0;
2565 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2569 for (i = 0; i < sbi->s_groups_count; i++) {
2570 gdp = ext4_get_group_desc(sb, i, NULL);
2572 flex_group = ext4_flex_group(sbi, i);
2573 fg = sbi_array_rcu_deref(sbi, s_flex_groups, flex_group);
2574 atomic_add(ext4_free_inodes_count(sb, gdp), &fg->free_inodes);
2575 atomic64_add(ext4_free_group_clusters(sb, gdp),
2576 &fg->free_clusters);
2577 atomic_add(ext4_used_dirs_count(sb, gdp), &fg->used_dirs);
2585 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
2586 struct ext4_group_desc *gdp)
2588 int offset = offsetof(struct ext4_group_desc, bg_checksum);
2590 __le32 le_group = cpu_to_le32(block_group);
2591 struct ext4_sb_info *sbi = EXT4_SB(sb);
2593 if (ext4_has_metadata_csum(sbi->s_sb)) {
2594 /* Use new metadata_csum algorithm */
2596 __u16 dummy_csum = 0;
2598 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2600 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2601 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2602 sizeof(dummy_csum));
2603 offset += sizeof(dummy_csum);
2604 if (offset < sbi->s_desc_size)
2605 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2606 sbi->s_desc_size - offset);
2608 crc = csum32 & 0xFFFF;
2612 /* old crc16 code */
2613 if (!ext4_has_feature_gdt_csum(sb))
2616 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2617 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2618 crc = crc16(crc, (__u8 *)gdp, offset);
2619 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2620 /* for checksum of struct ext4_group_desc do the rest...*/
2621 if (ext4_has_feature_64bit(sb) &&
2622 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2623 crc = crc16(crc, (__u8 *)gdp + offset,
2624 le16_to_cpu(sbi->s_es->s_desc_size) -
2628 return cpu_to_le16(crc);
2631 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2632 struct ext4_group_desc *gdp)
2634 if (ext4_has_group_desc_csum(sb) &&
2635 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
2641 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2642 struct ext4_group_desc *gdp)
2644 if (!ext4_has_group_desc_csum(sb))
2646 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
2649 /* Called at mount-time, super-block is locked */
2650 static int ext4_check_descriptors(struct super_block *sb,
2651 ext4_fsblk_t sb_block,
2652 ext4_group_t *first_not_zeroed)
2654 struct ext4_sb_info *sbi = EXT4_SB(sb);
2655 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2656 ext4_fsblk_t last_block;
2657 ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0);
2658 ext4_fsblk_t block_bitmap;
2659 ext4_fsblk_t inode_bitmap;
2660 ext4_fsblk_t inode_table;
2661 int flexbg_flag = 0;
2662 ext4_group_t i, grp = sbi->s_groups_count;
2664 if (ext4_has_feature_flex_bg(sb))
2667 ext4_debug("Checking group descriptors");
2669 for (i = 0; i < sbi->s_groups_count; i++) {
2670 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2672 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2673 last_block = ext4_blocks_count(sbi->s_es) - 1;
2675 last_block = first_block +
2676 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2678 if ((grp == sbi->s_groups_count) &&
2679 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2682 block_bitmap = ext4_block_bitmap(sb, gdp);
2683 if (block_bitmap == sb_block) {
2684 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2685 "Block bitmap for group %u overlaps "
2690 if (block_bitmap >= sb_block + 1 &&
2691 block_bitmap <= last_bg_block) {
2692 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2693 "Block bitmap for group %u overlaps "
2694 "block group descriptors", i);
2698 if (block_bitmap < first_block || block_bitmap > last_block) {
2699 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2700 "Block bitmap for group %u not in group "
2701 "(block %llu)!", i, block_bitmap);
2704 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2705 if (inode_bitmap == sb_block) {
2706 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2707 "Inode bitmap for group %u overlaps "
2712 if (inode_bitmap >= sb_block + 1 &&
2713 inode_bitmap <= last_bg_block) {
2714 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2715 "Inode bitmap for group %u overlaps "
2716 "block group descriptors", i);
2720 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2721 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2722 "Inode bitmap for group %u not in group "
2723 "(block %llu)!", i, inode_bitmap);
2726 inode_table = ext4_inode_table(sb, gdp);
2727 if (inode_table == sb_block) {
2728 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2729 "Inode table for group %u overlaps "
2734 if (inode_table >= sb_block + 1 &&
2735 inode_table <= last_bg_block) {
2736 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2737 "Inode table for group %u overlaps "
2738 "block group descriptors", i);
2742 if (inode_table < first_block ||
2743 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2744 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2745 "Inode table for group %u not in group "
2746 "(block %llu)!", i, inode_table);
2749 ext4_lock_group(sb, i);
2750 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2751 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2752 "Checksum for group %u failed (%u!=%u)",
2753 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
2754 gdp)), le16_to_cpu(gdp->bg_checksum));
2755 if (!sb_rdonly(sb)) {
2756 ext4_unlock_group(sb, i);
2760 ext4_unlock_group(sb, i);
2762 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2764 if (NULL != first_not_zeroed)
2765 *first_not_zeroed = grp;
2769 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2770 * the superblock) which were deleted from all directories, but held open by
2771 * a process at the time of a crash. We walk the list and try to delete these
2772 * inodes at recovery time (only with a read-write filesystem).
2774 * In order to keep the orphan inode chain consistent during traversal (in
2775 * case of crash during recovery), we link each inode into the superblock
2776 * orphan list_head and handle it the same way as an inode deletion during
2777 * normal operation (which journals the operations for us).
2779 * We only do an iget() and an iput() on each inode, which is very safe if we
2780 * accidentally point at an in-use or already deleted inode. The worst that
2781 * can happen in this case is that we get a "bit already cleared" message from
2782 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2783 * e2fsck was run on this filesystem, and it must have already done the orphan
2784 * inode cleanup for us, so we can safely abort without any further action.
2786 static void ext4_orphan_cleanup(struct super_block *sb,
2787 struct ext4_super_block *es)
2789 unsigned int s_flags = sb->s_flags;
2790 int ret, nr_orphans = 0, nr_truncates = 0;
2792 int quota_update = 0;
2795 if (!es->s_last_orphan) {
2796 jbd_debug(4, "no orphan inodes to clean up\n");
2800 if (bdev_read_only(sb->s_bdev)) {
2801 ext4_msg(sb, KERN_ERR, "write access "
2802 "unavailable, skipping orphan cleanup");
2806 /* Check if feature set would not allow a r/w mount */
2807 if (!ext4_feature_set_ok(sb, 0)) {
2808 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2809 "unknown ROCOMPAT features");
2813 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2814 /* don't clear list on RO mount w/ errors */
2815 if (es->s_last_orphan && !(s_flags & SB_RDONLY)) {
2816 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
2817 "clearing orphan list.\n");
2818 es->s_last_orphan = 0;
2820 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2824 if (s_flags & SB_RDONLY) {
2825 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2826 sb->s_flags &= ~SB_RDONLY;
2829 /* Needed for iput() to work correctly and not trash data */
2830 sb->s_flags |= SB_ACTIVE;
2833 * Turn on quotas which were not enabled for read-only mounts if
2834 * filesystem has quota feature, so that they are updated correctly.
2836 if (ext4_has_feature_quota(sb) && (s_flags & SB_RDONLY)) {
2837 int ret = ext4_enable_quotas(sb);
2842 ext4_msg(sb, KERN_ERR,
2843 "Cannot turn on quotas: error %d", ret);
2846 /* Turn on journaled quotas used for old sytle */
2847 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2848 if (EXT4_SB(sb)->s_qf_names[i]) {
2849 int ret = ext4_quota_on_mount(sb, i);
2854 ext4_msg(sb, KERN_ERR,
2855 "Cannot turn on journaled "
2856 "quota: type %d: error %d", i, ret);
2861 while (es->s_last_orphan) {
2862 struct inode *inode;
2865 * We may have encountered an error during cleanup; if
2866 * so, skip the rest.
2868 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2869 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2870 es->s_last_orphan = 0;
2874 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2875 if (IS_ERR(inode)) {
2876 es->s_last_orphan = 0;
2880 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2881 dquot_initialize(inode);
2882 if (inode->i_nlink) {
2883 if (test_opt(sb, DEBUG))
2884 ext4_msg(sb, KERN_DEBUG,
2885 "%s: truncating inode %lu to %lld bytes",
2886 __func__, inode->i_ino, inode->i_size);
2887 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2888 inode->i_ino, inode->i_size);
2890 truncate_inode_pages(inode->i_mapping, inode->i_size);
2891 ret = ext4_truncate(inode);
2893 ext4_std_error(inode->i_sb, ret);
2894 inode_unlock(inode);
2897 if (test_opt(sb, DEBUG))
2898 ext4_msg(sb, KERN_DEBUG,
2899 "%s: deleting unreferenced inode %lu",
2900 __func__, inode->i_ino);
2901 jbd_debug(2, "deleting unreferenced inode %lu\n",
2905 iput(inode); /* The delete magic happens here! */
2908 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2911 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2912 PLURAL(nr_orphans));
2914 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2915 PLURAL(nr_truncates));
2917 /* Turn off quotas if they were enabled for orphan cleanup */
2919 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2920 if (sb_dqopt(sb)->files[i])
2921 dquot_quota_off(sb, i);
2925 sb->s_flags = s_flags; /* Restore SB_RDONLY status */
2929 * Maximal extent format file size.
2930 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2931 * extent format containers, within a sector_t, and within i_blocks
2932 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2933 * so that won't be a limiting factor.
2935 * However there is other limiting factor. We do store extents in the form
2936 * of starting block and length, hence the resulting length of the extent
2937 * covering maximum file size must fit into on-disk format containers as
2938 * well. Given that length is always by 1 unit bigger than max unit (because
2939 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2941 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2943 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2946 loff_t upper_limit = MAX_LFS_FILESIZE;
2948 BUILD_BUG_ON(sizeof(blkcnt_t) < sizeof(u64));
2950 if (!has_huge_files) {
2951 upper_limit = (1LL << 32) - 1;
2953 /* total blocks in file system block size */
2954 upper_limit >>= (blkbits - 9);
2955 upper_limit <<= blkbits;
2959 * 32-bit extent-start container, ee_block. We lower the maxbytes
2960 * by one fs block, so ee_len can cover the extent of maximum file
2963 res = (1LL << 32) - 1;
2966 /* Sanity check against vm- & vfs- imposed limits */
2967 if (res > upper_limit)
2974 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2975 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2976 * We need to be 1 filesystem block less than the 2^48 sector limit.
2978 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2980 loff_t res = EXT4_NDIR_BLOCKS;
2983 /* This is calculated to be the largest file size for a dense, block
2984 * mapped file such that the file's total number of 512-byte sectors,
2985 * including data and all indirect blocks, does not exceed (2^48 - 1).
2987 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2988 * number of 512-byte sectors of the file.
2991 if (!has_huge_files) {
2993 * !has_huge_files or implies that the inode i_block field
2994 * represents total file blocks in 2^32 512-byte sectors ==
2995 * size of vfs inode i_blocks * 8
2997 upper_limit = (1LL << 32) - 1;
2999 /* total blocks in file system block size */
3000 upper_limit >>= (bits - 9);
3004 * We use 48 bit ext4_inode i_blocks
3005 * With EXT4_HUGE_FILE_FL set the i_blocks
3006 * represent total number of blocks in
3007 * file system block size
3009 upper_limit = (1LL << 48) - 1;
3013 /* indirect blocks */
3015 /* double indirect blocks */
3016 meta_blocks += 1 + (1LL << (bits-2));
3017 /* tripple indirect blocks */
3018 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
3020 upper_limit -= meta_blocks;
3021 upper_limit <<= bits;
3023 res += 1LL << (bits-2);
3024 res += 1LL << (2*(bits-2));
3025 res += 1LL << (3*(bits-2));
3027 if (res > upper_limit)
3030 if (res > MAX_LFS_FILESIZE)
3031 res = MAX_LFS_FILESIZE;
3036 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
3037 ext4_fsblk_t logical_sb_block, int nr)
3039 struct ext4_sb_info *sbi = EXT4_SB(sb);
3040 ext4_group_t bg, first_meta_bg;
3043 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
3045 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
3046 return logical_sb_block + nr + 1;
3047 bg = sbi->s_desc_per_block * nr;
3048 if (ext4_bg_has_super(sb, bg))
3052 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
3053 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
3054 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
3057 if (sb->s_blocksize == 1024 && nr == 0 &&
3058 le32_to_cpu(sbi->s_es->s_first_data_block) == 0)
3061 return (has_super + ext4_group_first_block_no(sb, bg));
3065 * ext4_get_stripe_size: Get the stripe size.
3066 * @sbi: In memory super block info
3068 * If we have specified it via mount option, then
3069 * use the mount option value. If the value specified at mount time is
3070 * greater than the blocks per group use the super block value.
3071 * If the super block value is greater than blocks per group return 0.
3072 * Allocator needs it be less than blocks per group.
3075 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
3077 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
3078 unsigned long stripe_width =
3079 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3082 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3083 ret = sbi->s_stripe;
3084 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
3086 else if (stride && stride <= sbi->s_blocks_per_group)
3092 * If the stripe width is 1, this makes no sense and
3093 * we set it to 0 to turn off stripe handling code.
3102 * Check whether this filesystem can be mounted based on
3103 * the features present and the RDONLY/RDWR mount requested.
3104 * Returns 1 if this filesystem can be mounted as requested,
3105 * 0 if it cannot be.
3107 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
3109 if (ext4_has_unknown_ext4_incompat_features(sb)) {
3110 ext4_msg(sb, KERN_ERR,
3111 "Couldn't mount because of "
3112 "unsupported optional features (%x)",
3113 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
3114 ~EXT4_FEATURE_INCOMPAT_SUPP));
3118 #ifndef CONFIG_UNICODE
3119 if (ext4_has_feature_casefold(sb)) {
3120 ext4_msg(sb, KERN_ERR,
3121 "Filesystem with casefold feature cannot be "
3122 "mounted without CONFIG_UNICODE");
3130 if (ext4_has_feature_readonly(sb)) {
3131 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
3132 sb->s_flags |= SB_RDONLY;
3136 /* Check that feature set is OK for a read-write mount */
3137 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
3138 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
3139 "unsupported optional features (%x)",
3140 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
3141 ~EXT4_FEATURE_RO_COMPAT_SUPP));
3144 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
3145 ext4_msg(sb, KERN_ERR,
3146 "Can't support bigalloc feature without "
3147 "extents feature\n");
3151 #if !IS_ENABLED(CONFIG_QUOTA) || !IS_ENABLED(CONFIG_QFMT_V2)
3152 if (!readonly && (ext4_has_feature_quota(sb) ||
3153 ext4_has_feature_project(sb))) {
3154 ext4_msg(sb, KERN_ERR,
3155 "The kernel was not built with CONFIG_QUOTA and CONFIG_QFMT_V2");
3158 #endif /* CONFIG_QUOTA */
3163 * This function is called once a day if we have errors logged
3164 * on the file system
3166 static void print_daily_error_info(struct timer_list *t)
3168 struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report);
3169 struct super_block *sb = sbi->s_sb;
3170 struct ext4_super_block *es = sbi->s_es;
3172 if (es->s_error_count)
3173 /* fsck newer than v1.41.13 is needed to clean this condition. */
3174 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
3175 le32_to_cpu(es->s_error_count));
3176 if (es->s_first_error_time) {
3177 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %llu: %.*s:%d",
3179 ext4_get_tstamp(es, s_first_error_time),
3180 (int) sizeof(es->s_first_error_func),
3181 es->s_first_error_func,
3182 le32_to_cpu(es->s_first_error_line));
3183 if (es->s_first_error_ino)
3184 printk(KERN_CONT ": inode %u",
3185 le32_to_cpu(es->s_first_error_ino));
3186 if (es->s_first_error_block)
3187 printk(KERN_CONT ": block %llu", (unsigned long long)
3188 le64_to_cpu(es->s_first_error_block));
3189 printk(KERN_CONT "\n");
3191 if (es->s_last_error_time) {
3192 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %llu: %.*s:%d",
3194 ext4_get_tstamp(es, s_last_error_time),
3195 (int) sizeof(es->s_last_error_func),
3196 es->s_last_error_func,
3197 le32_to_cpu(es->s_last_error_line));
3198 if (es->s_last_error_ino)
3199 printk(KERN_CONT ": inode %u",
3200 le32_to_cpu(es->s_last_error_ino));
3201 if (es->s_last_error_block)
3202 printk(KERN_CONT ": block %llu", (unsigned long long)
3203 le64_to_cpu(es->s_last_error_block));
3204 printk(KERN_CONT "\n");
3206 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
3209 /* Find next suitable group and run ext4_init_inode_table */
3210 static int ext4_run_li_request(struct ext4_li_request *elr)
3212 struct ext4_group_desc *gdp = NULL;
3213 struct super_block *sb = elr->lr_super;
3214 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3215 ext4_group_t group = elr->lr_next_group;
3216 unsigned long timeout = 0;
3217 unsigned int prefetch_ios = 0;
3220 if (elr->lr_mode == EXT4_LI_MODE_PREFETCH_BBITMAP) {
3221 elr->lr_next_group = ext4_mb_prefetch(sb, group,
3222 EXT4_SB(sb)->s_mb_prefetch, &prefetch_ios);
3224 ext4_mb_prefetch_fini(sb, elr->lr_next_group,
3226 trace_ext4_prefetch_bitmaps(sb, group, elr->lr_next_group,
3228 if (group >= elr->lr_next_group) {
3230 if (elr->lr_first_not_zeroed != ngroups &&
3231 !sb_rdonly(sb) && test_opt(sb, INIT_INODE_TABLE)) {
3232 elr->lr_next_group = elr->lr_first_not_zeroed;
3233 elr->lr_mode = EXT4_LI_MODE_ITABLE;
3240 for (; group < ngroups; group++) {
3241 gdp = ext4_get_group_desc(sb, group, NULL);
3247 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3251 if (group >= ngroups)
3256 ret = ext4_init_inode_table(sb, group,
3257 elr->lr_timeout ? 0 : 1);
3258 trace_ext4_lazy_itable_init(sb, group);
3259 if (elr->lr_timeout == 0) {
3260 timeout = (jiffies - timeout) *
3261 EXT4_SB(elr->lr_super)->s_li_wait_mult;
3262 elr->lr_timeout = timeout;
3264 elr->lr_next_sched = jiffies + elr->lr_timeout;
3265 elr->lr_next_group = group + 1;
3271 * Remove lr_request from the list_request and free the
3272 * request structure. Should be called with li_list_mtx held
3274 static void ext4_remove_li_request(struct ext4_li_request *elr)
3279 list_del(&elr->lr_request);
3280 EXT4_SB(elr->lr_super)->s_li_request = NULL;
3284 static void ext4_unregister_li_request(struct super_block *sb)
3286 mutex_lock(&ext4_li_mtx);
3287 if (!ext4_li_info) {
3288 mutex_unlock(&ext4_li_mtx);
3292 mutex_lock(&ext4_li_info->li_list_mtx);
3293 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
3294 mutex_unlock(&ext4_li_info->li_list_mtx);
3295 mutex_unlock(&ext4_li_mtx);
3298 static struct task_struct *ext4_lazyinit_task;
3301 * This is the function where ext4lazyinit thread lives. It walks
3302 * through the request list searching for next scheduled filesystem.
3303 * When such a fs is found, run the lazy initialization request
3304 * (ext4_rn_li_request) and keep track of the time spend in this
3305 * function. Based on that time we compute next schedule time of
3306 * the request. When walking through the list is complete, compute
3307 * next waking time and put itself into sleep.
3309 static int ext4_lazyinit_thread(void *arg)
3311 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
3312 struct list_head *pos, *n;
3313 struct ext4_li_request *elr;
3314 unsigned long next_wakeup, cur;
3316 BUG_ON(NULL == eli);
3320 next_wakeup = MAX_JIFFY_OFFSET;
3322 mutex_lock(&eli->li_list_mtx);
3323 if (list_empty(&eli->li_request_list)) {
3324 mutex_unlock(&eli->li_list_mtx);
3327 list_for_each_safe(pos, n, &eli->li_request_list) {
3330 elr = list_entry(pos, struct ext4_li_request,
3333 if (time_before(jiffies, elr->lr_next_sched)) {
3334 if (time_before(elr->lr_next_sched, next_wakeup))
3335 next_wakeup = elr->lr_next_sched;
3338 if (down_read_trylock(&elr->lr_super->s_umount)) {
3339 if (sb_start_write_trylock(elr->lr_super)) {
3342 * We hold sb->s_umount, sb can not
3343 * be removed from the list, it is
3344 * now safe to drop li_list_mtx
3346 mutex_unlock(&eli->li_list_mtx);
3347 err = ext4_run_li_request(elr);
3348 sb_end_write(elr->lr_super);
3349 mutex_lock(&eli->li_list_mtx);
3352 up_read((&elr->lr_super->s_umount));
3354 /* error, remove the lazy_init job */
3356 ext4_remove_li_request(elr);
3360 elr->lr_next_sched = jiffies +
3362 % (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3364 if (time_before(elr->lr_next_sched, next_wakeup))
3365 next_wakeup = elr->lr_next_sched;
3367 mutex_unlock(&eli->li_list_mtx);
3372 if ((time_after_eq(cur, next_wakeup)) ||
3373 (MAX_JIFFY_OFFSET == next_wakeup)) {
3378 schedule_timeout_interruptible(next_wakeup - cur);
3380 if (kthread_should_stop()) {
3381 ext4_clear_request_list();
3388 * It looks like the request list is empty, but we need
3389 * to check it under the li_list_mtx lock, to prevent any
3390 * additions into it, and of course we should lock ext4_li_mtx
3391 * to atomically free the list and ext4_li_info, because at
3392 * this point another ext4 filesystem could be registering
3395 mutex_lock(&ext4_li_mtx);
3396 mutex_lock(&eli->li_list_mtx);
3397 if (!list_empty(&eli->li_request_list)) {
3398 mutex_unlock(&eli->li_list_mtx);
3399 mutex_unlock(&ext4_li_mtx);
3402 mutex_unlock(&eli->li_list_mtx);
3403 kfree(ext4_li_info);
3404 ext4_li_info = NULL;
3405 mutex_unlock(&ext4_li_mtx);
3410 static void ext4_clear_request_list(void)
3412 struct list_head *pos, *n;
3413 struct ext4_li_request *elr;
3415 mutex_lock(&ext4_li_info->li_list_mtx);
3416 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3417 elr = list_entry(pos, struct ext4_li_request,
3419 ext4_remove_li_request(elr);
3421 mutex_unlock(&ext4_li_info->li_list_mtx);
3424 static int ext4_run_lazyinit_thread(void)
3426 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3427 ext4_li_info, "ext4lazyinit");
3428 if (IS_ERR(ext4_lazyinit_task)) {
3429 int err = PTR_ERR(ext4_lazyinit_task);
3430 ext4_clear_request_list();
3431 kfree(ext4_li_info);
3432 ext4_li_info = NULL;
3433 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3434 "initialization thread\n",
3438 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3443 * Check whether it make sense to run itable init. thread or not.
3444 * If there is at least one uninitialized inode table, return
3445 * corresponding group number, else the loop goes through all
3446 * groups and return total number of groups.
3448 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3450 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3451 struct ext4_group_desc *gdp = NULL;
3453 if (!ext4_has_group_desc_csum(sb))
3456 for (group = 0; group < ngroups; group++) {
3457 gdp = ext4_get_group_desc(sb, group, NULL);
3461 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3468 static int ext4_li_info_new(void)
3470 struct ext4_lazy_init *eli = NULL;
3472 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3476 INIT_LIST_HEAD(&eli->li_request_list);
3477 mutex_init(&eli->li_list_mtx);
3479 eli->li_state |= EXT4_LAZYINIT_QUIT;
3486 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3489 struct ext4_li_request *elr;
3491 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3496 elr->lr_first_not_zeroed = start;
3497 if (test_opt(sb, PREFETCH_BLOCK_BITMAPS))
3498 elr->lr_mode = EXT4_LI_MODE_PREFETCH_BBITMAP;
3500 elr->lr_mode = EXT4_LI_MODE_ITABLE;
3501 elr->lr_next_group = start;
3505 * Randomize first schedule time of the request to
3506 * spread the inode table initialization requests
3509 elr->lr_next_sched = jiffies + (prandom_u32() %
3510 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3514 int ext4_register_li_request(struct super_block *sb,
3515 ext4_group_t first_not_zeroed)
3517 struct ext4_sb_info *sbi = EXT4_SB(sb);
3518 struct ext4_li_request *elr = NULL;
3519 ext4_group_t ngroups = sbi->s_groups_count;
3522 mutex_lock(&ext4_li_mtx);
3523 if (sbi->s_li_request != NULL) {
3525 * Reset timeout so it can be computed again, because
3526 * s_li_wait_mult might have changed.
3528 sbi->s_li_request->lr_timeout = 0;
3532 if (!test_opt(sb, PREFETCH_BLOCK_BITMAPS) &&
3533 (first_not_zeroed == ngroups || sb_rdonly(sb) ||
3534 !test_opt(sb, INIT_INODE_TABLE)))
3537 elr = ext4_li_request_new(sb, first_not_zeroed);
3543 if (NULL == ext4_li_info) {
3544 ret = ext4_li_info_new();
3549 mutex_lock(&ext4_li_info->li_list_mtx);
3550 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3551 mutex_unlock(&ext4_li_info->li_list_mtx);
3553 sbi->s_li_request = elr;
3555 * set elr to NULL here since it has been inserted to
3556 * the request_list and the removal and free of it is
3557 * handled by ext4_clear_request_list from now on.
3561 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3562 ret = ext4_run_lazyinit_thread();
3567 mutex_unlock(&ext4_li_mtx);
3574 * We do not need to lock anything since this is called on
3577 static void ext4_destroy_lazyinit_thread(void)
3580 * If thread exited earlier
3581 * there's nothing to be done.
3583 if (!ext4_li_info || !ext4_lazyinit_task)
3586 kthread_stop(ext4_lazyinit_task);
3589 static int set_journal_csum_feature_set(struct super_block *sb)
3592 int compat, incompat;
3593 struct ext4_sb_info *sbi = EXT4_SB(sb);
3595 if (ext4_has_metadata_csum(sb)) {
3596 /* journal checksum v3 */
3598 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
3600 /* journal checksum v1 */
3601 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3605 jbd2_journal_clear_features(sbi->s_journal,
3606 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3607 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3608 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3609 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3610 ret = jbd2_journal_set_features(sbi->s_journal,
3612 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3614 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3615 ret = jbd2_journal_set_features(sbi->s_journal,
3618 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3619 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3621 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3622 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3629 * Note: calculating the overhead so we can be compatible with
3630 * historical BSD practice is quite difficult in the face of
3631 * clusters/bigalloc. This is because multiple metadata blocks from
3632 * different block group can end up in the same allocation cluster.
3633 * Calculating the exact overhead in the face of clustered allocation
3634 * requires either O(all block bitmaps) in memory or O(number of block
3635 * groups**2) in time. We will still calculate the superblock for
3636 * older file systems --- and if we come across with a bigalloc file
3637 * system with zero in s_overhead_clusters the estimate will be close to
3638 * correct especially for very large cluster sizes --- but for newer
3639 * file systems, it's better to calculate this figure once at mkfs
3640 * time, and store it in the superblock. If the superblock value is
3641 * present (even for non-bigalloc file systems), we will use it.
3643 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3646 struct ext4_sb_info *sbi = EXT4_SB(sb);
3647 struct ext4_group_desc *gdp;
3648 ext4_fsblk_t first_block, last_block, b;
3649 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3650 int s, j, count = 0;
3652 if (!ext4_has_feature_bigalloc(sb))
3653 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3654 sbi->s_itb_per_group + 2);
3656 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3657 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3658 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3659 for (i = 0; i < ngroups; i++) {
3660 gdp = ext4_get_group_desc(sb, i, NULL);
3661 b = ext4_block_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_bitmap(sb, gdp);
3667 if (b >= first_block && b <= last_block) {
3668 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3671 b = ext4_inode_table(sb, gdp);
3672 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3673 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3674 int c = EXT4_B2C(sbi, b - first_block);
3675 ext4_set_bit(c, buf);
3681 if (ext4_bg_has_super(sb, grp)) {
3682 ext4_set_bit(s++, buf);
3685 j = ext4_bg_num_gdb(sb, grp);
3686 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
3687 ext4_error(sb, "Invalid number of block group "
3688 "descriptor blocks: %d", j);
3689 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
3693 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3697 return EXT4_CLUSTERS_PER_GROUP(sb) -
3698 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3702 * Compute the overhead and stash it in sbi->s_overhead
3704 int ext4_calculate_overhead(struct super_block *sb)
3706 struct ext4_sb_info *sbi = EXT4_SB(sb);
3707 struct ext4_super_block *es = sbi->s_es;
3708 struct inode *j_inode;
3709 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
3710 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3711 ext4_fsblk_t overhead = 0;
3712 char *buf = (char *) get_zeroed_page(GFP_NOFS);
3718 * Compute the overhead (FS structures). This is constant
3719 * for a given filesystem unless the number of block groups
3720 * changes so we cache the previous value until it does.
3724 * All of the blocks before first_data_block are overhead
3726 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3729 * Add the overhead found in each block group
3731 for (i = 0; i < ngroups; i++) {
3734 blks = count_overhead(sb, i, buf);
3737 memset(buf, 0, PAGE_SIZE);
3742 * Add the internal journal blocks whether the journal has been
3745 if (sbi->s_journal && !sbi->journal_bdev)
3746 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3747 else if (ext4_has_feature_journal(sb) && !sbi->s_journal && j_inum) {
3748 /* j_inum for internal journal is non-zero */
3749 j_inode = ext4_get_journal_inode(sb, j_inum);
3751 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
3752 overhead += EXT4_NUM_B2C(sbi, j_blocks);
3755 ext4_msg(sb, KERN_ERR, "can't get journal size");
3758 sbi->s_overhead = overhead;
3760 free_page((unsigned long) buf);
3764 static void ext4_set_resv_clusters(struct super_block *sb)
3766 ext4_fsblk_t resv_clusters;
3767 struct ext4_sb_info *sbi = EXT4_SB(sb);
3770 * There's no need to reserve anything when we aren't using extents.
3771 * The space estimates are exact, there are no unwritten extents,
3772 * hole punching doesn't need new metadata... This is needed especially
3773 * to keep ext2/3 backward compatibility.
3775 if (!ext4_has_feature_extents(sb))
3778 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3779 * This should cover the situations where we can not afford to run
3780 * out of space like for example punch hole, or converting
3781 * unwritten extents in delalloc path. In most cases such
3782 * allocation would require 1, or 2 blocks, higher numbers are
3785 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3786 sbi->s_cluster_bits);
3788 do_div(resv_clusters, 50);
3789 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3791 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
3794 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3796 struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev);
3797 char *orig_data = kstrdup(data, GFP_KERNEL);
3798 struct buffer_head *bh, **group_desc;
3799 struct ext4_super_block *es = NULL;
3800 struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3801 struct flex_groups **flex_groups;
3803 ext4_fsblk_t sb_block = get_sb_block(&data);
3804 ext4_fsblk_t logical_sb_block;
3805 unsigned long offset = 0;
3806 unsigned long journal_devnum = 0;
3807 unsigned long def_mount_opts;
3811 int blocksize, clustersize;
3812 unsigned int db_count;
3814 int needs_recovery, has_huge_files;
3817 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3818 ext4_group_t first_not_zeroed;
3820 if ((data && !orig_data) || !sbi)
3823 sbi->s_daxdev = dax_dev;
3824 sbi->s_blockgroup_lock =
3825 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3826 if (!sbi->s_blockgroup_lock)
3829 sb->s_fs_info = sbi;
3831 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3832 sbi->s_sb_block = sb_block;
3833 if (sb->s_bdev->bd_part)
3834 sbi->s_sectors_written_start =
3835 part_stat_read(sb->s_bdev->bd_part, sectors[STAT_WRITE]);
3837 /* Cleanup superblock name */
3838 strreplace(sb->s_id, '/', '!');
3840 /* -EINVAL is default */
3842 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3844 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3849 * The ext4 superblock will not be buffer aligned for other than 1kB
3850 * block sizes. We need to calculate the offset from buffer start.
3852 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3853 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3854 offset = do_div(logical_sb_block, blocksize);
3856 logical_sb_block = sb_block;
3859 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
3860 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3864 * Note: s_es must be initialized as soon as possible because
3865 * some ext4 macro-instructions depend on its value
3867 es = (struct ext4_super_block *) (bh->b_data + offset);
3869 sb->s_magic = le16_to_cpu(es->s_magic);
3870 if (sb->s_magic != EXT4_SUPER_MAGIC)
3872 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3874 /* Warn if metadata_csum and gdt_csum are both set. */
3875 if (ext4_has_feature_metadata_csum(sb) &&
3876 ext4_has_feature_gdt_csum(sb))
3877 ext4_warning(sb, "metadata_csum and uninit_bg are "
3878 "redundant flags; please run fsck.");
3880 /* Check for a known checksum algorithm */
3881 if (!ext4_verify_csum_type(sb, es)) {
3882 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3883 "unknown checksum algorithm.");
3888 /* Load the checksum driver */
3889 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3890 if (IS_ERR(sbi->s_chksum_driver)) {
3891 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3892 ret = PTR_ERR(sbi->s_chksum_driver);
3893 sbi->s_chksum_driver = NULL;
3897 /* Check superblock checksum */
3898 if (!ext4_superblock_csum_verify(sb, es)) {
3899 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3900 "invalid superblock checksum. Run e2fsck?");
3906 /* Precompute checksum seed for all metadata */
3907 if (ext4_has_feature_csum_seed(sb))
3908 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
3909 else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb))
3910 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3911 sizeof(es->s_uuid));
3913 /* Set defaults before we parse the mount options */
3914 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3915 set_opt(sb, INIT_INODE_TABLE);
3916 if (def_mount_opts & EXT4_DEFM_DEBUG)
3918 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3920 if (def_mount_opts & EXT4_DEFM_UID16)
3921 set_opt(sb, NO_UID32);
3922 /* xattr user namespace & acls are now defaulted on */
3923 set_opt(sb, XATTR_USER);
3924 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3925 set_opt(sb, POSIX_ACL);
3927 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3928 if (ext4_has_metadata_csum(sb))
3929 set_opt(sb, JOURNAL_CHECKSUM);
3931 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3932 set_opt(sb, JOURNAL_DATA);
3933 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3934 set_opt(sb, ORDERED_DATA);
3935 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3936 set_opt(sb, WRITEBACK_DATA);
3938 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3939 set_opt(sb, ERRORS_PANIC);
3940 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3941 set_opt(sb, ERRORS_CONT);
3943 set_opt(sb, ERRORS_RO);
3944 /* block_validity enabled by default; disable with noblock_validity */
3945 set_opt(sb, BLOCK_VALIDITY);
3946 if (def_mount_opts & EXT4_DEFM_DISCARD)
3947 set_opt(sb, DISCARD);
3949 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3950 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3951 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3952 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3953 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3955 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3956 set_opt(sb, BARRIER);
3959 * enable delayed allocation by default
3960 * Use -o nodelalloc to turn it off
3962 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3963 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3964 set_opt(sb, DELALLOC);
3967 * set default s_li_wait_mult for lazyinit, for the case there is
3968 * no mount option specified.
3970 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3972 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3974 if (blocksize == PAGE_SIZE)
3975 set_opt(sb, DIOREAD_NOLOCK);
3977 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3978 blocksize > EXT4_MAX_BLOCK_SIZE) {
3979 ext4_msg(sb, KERN_ERR,
3980 "Unsupported filesystem blocksize %d (%d log_block_size)",
3981 blocksize, le32_to_cpu(es->s_log_block_size));
3985 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3986 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3987 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3989 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3990 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3991 if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) {
3992 ext4_msg(sb, KERN_ERR, "invalid first ino: %u",
3996 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3997 (!is_power_of_2(sbi->s_inode_size)) ||
3998 (sbi->s_inode_size > blocksize)) {
3999 ext4_msg(sb, KERN_ERR,
4000 "unsupported inode size: %d",
4002 ext4_msg(sb, KERN_ERR, "blocksize: %d", blocksize);
4006 * i_atime_extra is the last extra field available for
4007 * [acm]times in struct ext4_inode. Checking for that
4008 * field should suffice to ensure we have extra space
4011 if (sbi->s_inode_size >= offsetof(struct ext4_inode, i_atime_extra) +
4012 sizeof(((struct ext4_inode *)0)->i_atime_extra)) {
4013 sb->s_time_gran = 1;
4014 sb->s_time_max = EXT4_EXTRA_TIMESTAMP_MAX;
4016 sb->s_time_gran = NSEC_PER_SEC;
4017 sb->s_time_max = EXT4_NON_EXTRA_TIMESTAMP_MAX;
4019 sb->s_time_min = EXT4_TIMESTAMP_MIN;
4021 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4022 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4023 EXT4_GOOD_OLD_INODE_SIZE;
4024 if (ext4_has_feature_extra_isize(sb)) {
4025 unsigned v, max = (sbi->s_inode_size -
4026 EXT4_GOOD_OLD_INODE_SIZE);
4028 v = le16_to_cpu(es->s_want_extra_isize);
4030 ext4_msg(sb, KERN_ERR,
4031 "bad s_want_extra_isize: %d", v);
4034 if (sbi->s_want_extra_isize < v)
4035 sbi->s_want_extra_isize = v;
4037 v = le16_to_cpu(es->s_min_extra_isize);
4039 ext4_msg(sb, KERN_ERR,
4040 "bad s_min_extra_isize: %d", v);
4043 if (sbi->s_want_extra_isize < v)
4044 sbi->s_want_extra_isize = v;
4048 if (sbi->s_es->s_mount_opts[0]) {
4049 char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
4050 sizeof(sbi->s_es->s_mount_opts),
4054 if (!parse_options(s_mount_opts, sb, &journal_devnum,
4055 &journal_ioprio, 0)) {
4056 ext4_msg(sb, KERN_WARNING,
4057 "failed to parse options in superblock: %s",
4060 kfree(s_mount_opts);
4062 sbi->s_def_mount_opt = sbi->s_mount_opt;
4063 if (!parse_options((char *) data, sb, &journal_devnum,
4064 &journal_ioprio, 0))
4067 #ifdef CONFIG_UNICODE
4068 if (ext4_has_feature_casefold(sb) && !sbi->s_encoding) {
4069 const struct ext4_sb_encodings *encoding_info;
4070 struct unicode_map *encoding;
4071 __u16 encoding_flags;
4073 if (ext4_has_feature_encrypt(sb)) {
4074 ext4_msg(sb, KERN_ERR,
4075 "Can't mount with encoding and encryption");
4079 if (ext4_sb_read_encoding(es, &encoding_info,
4081 ext4_msg(sb, KERN_ERR,
4082 "Encoding requested by superblock is unknown");
4086 encoding = utf8_load(encoding_info->version);
4087 if (IS_ERR(encoding)) {
4088 ext4_msg(sb, KERN_ERR,
4089 "can't mount with superblock charset: %s-%s "
4090 "not supported by the kernel. flags: 0x%x.",
4091 encoding_info->name, encoding_info->version,
4095 ext4_msg(sb, KERN_INFO,"Using encoding defined by superblock: "
4096 "%s-%s with flags 0x%hx", encoding_info->name,
4097 encoding_info->version?:"\b", encoding_flags);
4099 sbi->s_encoding = encoding;
4100 sbi->s_encoding_flags = encoding_flags;
4104 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4105 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting with data=journal disables delayed allocation, dioread_nolock, and O_DIRECT support!\n");
4106 /* can't mount with both data=journal and dioread_nolock. */
4107 clear_opt(sb, DIOREAD_NOLOCK);
4108 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4109 ext4_msg(sb, KERN_ERR, "can't mount with "
4110 "both data=journal and delalloc");
4113 if (test_opt(sb, DAX_ALWAYS)) {
4114 ext4_msg(sb, KERN_ERR, "can't mount with "
4115 "both data=journal and dax");
4118 if (ext4_has_feature_encrypt(sb)) {
4119 ext4_msg(sb, KERN_WARNING,
4120 "encrypted files will use data=ordered "
4121 "instead of data journaling mode");
4123 if (test_opt(sb, DELALLOC))
4124 clear_opt(sb, DELALLOC);
4126 sb->s_iflags |= SB_I_CGROUPWB;
4129 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
4130 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
4132 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
4133 (ext4_has_compat_features(sb) ||
4134 ext4_has_ro_compat_features(sb) ||
4135 ext4_has_incompat_features(sb)))
4136 ext4_msg(sb, KERN_WARNING,
4137 "feature flags set on rev 0 fs, "
4138 "running e2fsck is recommended");
4140 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
4141 set_opt2(sb, HURD_COMPAT);
4142 if (ext4_has_feature_64bit(sb)) {
4143 ext4_msg(sb, KERN_ERR,
4144 "The Hurd can't support 64-bit file systems");
4149 * ea_inode feature uses l_i_version field which is not
4150 * available in HURD_COMPAT mode.
4152 if (ext4_has_feature_ea_inode(sb)) {
4153 ext4_msg(sb, KERN_ERR,
4154 "ea_inode feature is not supported for Hurd");
4159 if (IS_EXT2_SB(sb)) {
4160 if (ext2_feature_set_ok(sb))
4161 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
4162 "using the ext4 subsystem");
4165 * If we're probing be silent, if this looks like
4166 * it's actually an ext[34] filesystem.
4168 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4170 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
4171 "to feature incompatibilities");
4176 if (IS_EXT3_SB(sb)) {
4177 if (ext3_feature_set_ok(sb))
4178 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
4179 "using the ext4 subsystem");
4182 * If we're probing be silent, if this looks like
4183 * it's actually an ext4 filesystem.
4185 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4187 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
4188 "to feature incompatibilities");
4194 * Check feature flags regardless of the revision level, since we
4195 * previously didn't change the revision level when setting the flags,
4196 * so there is a chance incompat flags are set on a rev 0 filesystem.
4198 if (!ext4_feature_set_ok(sb, (sb_rdonly(sb))))
4201 if (le32_to_cpu(es->s_log_block_size) >
4202 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
4203 ext4_msg(sb, KERN_ERR,
4204 "Invalid log block size: %u",
4205 le32_to_cpu(es->s_log_block_size));
4208 if (le32_to_cpu(es->s_log_cluster_size) >
4209 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
4210 ext4_msg(sb, KERN_ERR,
4211 "Invalid log cluster size: %u",
4212 le32_to_cpu(es->s_log_cluster_size));
4216 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
4217 ext4_msg(sb, KERN_ERR,
4218 "Number of reserved GDT blocks insanely large: %d",
4219 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
4223 if (bdev_dax_supported(sb->s_bdev, blocksize))
4224 set_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags);
4226 if (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) {
4227 if (ext4_has_feature_inline_data(sb)) {
4228 ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem"
4229 " that may contain inline data");
4232 if (!test_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags)) {
4233 ext4_msg(sb, KERN_ERR,
4234 "DAX unsupported by block device.");
4239 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
4240 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
4241 es->s_encryption_level);
4245 if (sb->s_blocksize != blocksize) {
4246 /* Validate the filesystem blocksize */
4247 if (!sb_set_blocksize(sb, blocksize)) {
4248 ext4_msg(sb, KERN_ERR, "bad block size %d",
4254 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
4255 offset = do_div(logical_sb_block, blocksize);
4256 bh = sb_bread_unmovable(sb, logical_sb_block);
4258 ext4_msg(sb, KERN_ERR,
4259 "Can't read superblock on 2nd try");
4262 es = (struct ext4_super_block *)(bh->b_data + offset);
4264 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
4265 ext4_msg(sb, KERN_ERR,
4266 "Magic mismatch, very weird!");
4271 has_huge_files = ext4_has_feature_huge_file(sb);
4272 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
4274 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
4276 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
4277 if (ext4_has_feature_64bit(sb)) {
4278 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
4279 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
4280 !is_power_of_2(sbi->s_desc_size)) {
4281 ext4_msg(sb, KERN_ERR,
4282 "unsupported descriptor size %lu",
4287 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
4289 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
4290 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
4292 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
4293 if (sbi->s_inodes_per_block == 0)
4295 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
4296 sbi->s_inodes_per_group > blocksize * 8) {
4297 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
4298 sbi->s_inodes_per_group);
4301 sbi->s_itb_per_group = sbi->s_inodes_per_group /
4302 sbi->s_inodes_per_block;
4303 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
4305 sbi->s_mount_state = le16_to_cpu(es->s_state);
4306 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
4307 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
4309 for (i = 0; i < 4; i++)
4310 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
4311 sbi->s_def_hash_version = es->s_def_hash_version;
4312 if (ext4_has_feature_dir_index(sb)) {
4313 i = le32_to_cpu(es->s_flags);
4314 if (i & EXT2_FLAGS_UNSIGNED_HASH)
4315 sbi->s_hash_unsigned = 3;
4316 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
4317 #ifdef __CHAR_UNSIGNED__
4320 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
4321 sbi->s_hash_unsigned = 3;
4325 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
4330 /* Handle clustersize */
4331 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
4332 if (ext4_has_feature_bigalloc(sb)) {
4333 if (clustersize < blocksize) {
4334 ext4_msg(sb, KERN_ERR,
4335 "cluster size (%d) smaller than "
4336 "block size (%d)", clustersize, blocksize);
4339 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
4340 le32_to_cpu(es->s_log_block_size);
4341 sbi->s_clusters_per_group =
4342 le32_to_cpu(es->s_clusters_per_group);
4343 if (sbi->s_clusters_per_group > blocksize * 8) {
4344 ext4_msg(sb, KERN_ERR,
4345 "#clusters per group too big: %lu",
4346 sbi->s_clusters_per_group);
4349 if (sbi->s_blocks_per_group !=
4350 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
4351 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
4352 "clusters per group (%lu) inconsistent",
4353 sbi->s_blocks_per_group,
4354 sbi->s_clusters_per_group);
4358 if (clustersize != blocksize) {
4359 ext4_msg(sb, KERN_ERR,
4360 "fragment/cluster size (%d) != "
4361 "block size (%d)", clustersize, blocksize);
4364 if (sbi->s_blocks_per_group > blocksize * 8) {
4365 ext4_msg(sb, KERN_ERR,
4366 "#blocks per group too big: %lu",
4367 sbi->s_blocks_per_group);
4370 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
4371 sbi->s_cluster_bits = 0;
4373 sbi->s_cluster_ratio = clustersize / blocksize;
4375 /* Do we have standard group size of clustersize * 8 blocks ? */
4376 if (sbi->s_blocks_per_group == clustersize << 3)
4377 set_opt2(sb, STD_GROUP_SIZE);
4380 * Test whether we have more sectors than will fit in sector_t,
4381 * and whether the max offset is addressable by the page cache.
4383 err = generic_check_addressable(sb->s_blocksize_bits,
4384 ext4_blocks_count(es));
4386 ext4_msg(sb, KERN_ERR, "filesystem"
4387 " too large to mount safely on this system");
4391 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
4394 /* check blocks count against device size */
4395 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
4396 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
4397 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
4398 "exceeds size of device (%llu blocks)",
4399 ext4_blocks_count(es), blocks_count);
4404 * It makes no sense for the first data block to be beyond the end
4405 * of the filesystem.
4407 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
4408 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4409 "block %u is beyond end of filesystem (%llu)",
4410 le32_to_cpu(es->s_first_data_block),
4411 ext4_blocks_count(es));
4414 if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) &&
4415 (sbi->s_cluster_ratio == 1)) {
4416 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4417 "block is 0 with a 1k block and cluster size");
4421 blocks_count = (ext4_blocks_count(es) -
4422 le32_to_cpu(es->s_first_data_block) +
4423 EXT4_BLOCKS_PER_GROUP(sb) - 1);
4424 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4425 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
4426 ext4_msg(sb, KERN_WARNING, "groups count too large: %llu "
4427 "(block count %llu, first data block %u, "
4428 "blocks per group %lu)", blocks_count,
4429 ext4_blocks_count(es),
4430 le32_to_cpu(es->s_first_data_block),
4431 EXT4_BLOCKS_PER_GROUP(sb));
4434 sbi->s_groups_count = blocks_count;
4435 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
4436 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
4437 if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) !=
4438 le32_to_cpu(es->s_inodes_count)) {
4439 ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu",
4440 le32_to_cpu(es->s_inodes_count),
4441 ((u64)sbi->s_groups_count * sbi->s_inodes_per_group));
4445 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
4446 EXT4_DESC_PER_BLOCK(sb);
4447 if (ext4_has_feature_meta_bg(sb)) {
4448 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
4449 ext4_msg(sb, KERN_WARNING,
4450 "first meta block group too large: %u "
4451 "(group descriptor block count %u)",
4452 le32_to_cpu(es->s_first_meta_bg), db_count);
4456 rcu_assign_pointer(sbi->s_group_desc,
4457 kvmalloc_array(db_count,
4458 sizeof(struct buffer_head *),
4460 if (sbi->s_group_desc == NULL) {
4461 ext4_msg(sb, KERN_ERR, "not enough memory");
4466 bgl_lock_init(sbi->s_blockgroup_lock);
4468 /* Pre-read the descriptors into the buffer cache */
4469 for (i = 0; i < db_count; i++) {
4470 block = descriptor_loc(sb, logical_sb_block, i);
4471 sb_breadahead_unmovable(sb, block);
4474 for (i = 0; i < db_count; i++) {
4475 struct buffer_head *bh;
4477 block = descriptor_loc(sb, logical_sb_block, i);
4478 bh = sb_bread_unmovable(sb, block);
4480 ext4_msg(sb, KERN_ERR,
4481 "can't read group descriptor %d", i);
4486 rcu_dereference(sbi->s_group_desc)[i] = bh;
4489 sbi->s_gdb_count = db_count;
4490 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
4491 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
4492 ret = -EFSCORRUPTED;
4496 timer_setup(&sbi->s_err_report, print_daily_error_info, 0);
4498 /* Register extent status tree shrinker */
4499 if (ext4_es_register_shrinker(sbi))
4502 sbi->s_stripe = ext4_get_stripe_size(sbi);
4503 sbi->s_extent_max_zeroout_kb = 32;
4506 * set up enough so that it can read an inode
4508 sb->s_op = &ext4_sops;
4509 sb->s_export_op = &ext4_export_ops;
4510 sb->s_xattr = ext4_xattr_handlers;
4511 #ifdef CONFIG_FS_ENCRYPTION
4512 sb->s_cop = &ext4_cryptops;
4514 #ifdef CONFIG_FS_VERITY
4515 sb->s_vop = &ext4_verityops;
4518 sb->dq_op = &ext4_quota_operations;
4519 if (ext4_has_feature_quota(sb))
4520 sb->s_qcop = &dquot_quotactl_sysfile_ops;
4522 sb->s_qcop = &ext4_qctl_operations;
4523 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
4525 memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
4527 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
4528 mutex_init(&sbi->s_orphan_lock);
4532 needs_recovery = (es->s_last_orphan != 0 ||
4533 ext4_has_feature_journal_needs_recovery(sb));
4535 if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb))
4536 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
4537 goto failed_mount3a;
4540 * The first inode we look at is the journal inode. Don't try
4541 * root first: it may be modified in the journal!
4543 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
4544 err = ext4_load_journal(sb, es, journal_devnum);
4546 goto failed_mount3a;
4547 } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) &&
4548 ext4_has_feature_journal_needs_recovery(sb)) {
4549 ext4_msg(sb, KERN_ERR, "required journal recovery "
4550 "suppressed and not mounted read-only");
4551 goto failed_mount_wq;
4553 /* Nojournal mode, all journal mount options are illegal */
4554 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
4555 ext4_msg(sb, KERN_ERR, "can't mount with "
4556 "journal_checksum, fs mounted w/o journal");
4557 goto failed_mount_wq;
4559 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4560 ext4_msg(sb, KERN_ERR, "can't mount with "
4561 "journal_async_commit, fs mounted w/o journal");
4562 goto failed_mount_wq;
4564 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
4565 ext4_msg(sb, KERN_ERR, "can't mount with "
4566 "commit=%lu, fs mounted w/o journal",
4567 sbi->s_commit_interval / HZ);
4568 goto failed_mount_wq;
4570 if (EXT4_MOUNT_DATA_FLAGS &
4571 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
4572 ext4_msg(sb, KERN_ERR, "can't mount with "
4573 "data=, fs mounted w/o journal");
4574 goto failed_mount_wq;
4576 sbi->s_def_mount_opt &= ~EXT4_MOUNT_JOURNAL_CHECKSUM;
4577 clear_opt(sb, JOURNAL_CHECKSUM);
4578 clear_opt(sb, DATA_FLAGS);
4579 sbi->s_journal = NULL;
4584 if (ext4_has_feature_64bit(sb) &&
4585 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4586 JBD2_FEATURE_INCOMPAT_64BIT)) {
4587 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
4588 goto failed_mount_wq;
4591 if (!set_journal_csum_feature_set(sb)) {
4592 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4594 goto failed_mount_wq;
4597 /* We have now updated the journal if required, so we can
4598 * validate the data journaling mode. */
4599 switch (test_opt(sb, DATA_FLAGS)) {
4601 /* No mode set, assume a default based on the journal
4602 * capabilities: ORDERED_DATA if the journal can
4603 * cope, else JOURNAL_DATA
4605 if (jbd2_journal_check_available_features
4606 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4607 set_opt(sb, ORDERED_DATA);
4608 sbi->s_def_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
4610 set_opt(sb, JOURNAL_DATA);
4611 sbi->s_def_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
4615 case EXT4_MOUNT_ORDERED_DATA:
4616 case EXT4_MOUNT_WRITEBACK_DATA:
4617 if (!jbd2_journal_check_available_features
4618 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4619 ext4_msg(sb, KERN_ERR, "Journal does not support "
4620 "requested data journaling mode");
4621 goto failed_mount_wq;
4627 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
4628 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4629 ext4_msg(sb, KERN_ERR, "can't mount with "
4630 "journal_async_commit in data=ordered mode");
4631 goto failed_mount_wq;
4634 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4636 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4639 if (!test_opt(sb, NO_MBCACHE)) {
4640 sbi->s_ea_block_cache = ext4_xattr_create_cache();
4641 if (!sbi->s_ea_block_cache) {
4642 ext4_msg(sb, KERN_ERR,
4643 "Failed to create ea_block_cache");
4644 goto failed_mount_wq;
4647 if (ext4_has_feature_ea_inode(sb)) {
4648 sbi->s_ea_inode_cache = ext4_xattr_create_cache();
4649 if (!sbi->s_ea_inode_cache) {
4650 ext4_msg(sb, KERN_ERR,
4651 "Failed to create ea_inode_cache");
4652 goto failed_mount_wq;
4657 if (ext4_has_feature_verity(sb) && blocksize != PAGE_SIZE) {
4658 ext4_msg(sb, KERN_ERR, "Unsupported blocksize for fs-verity");
4659 goto failed_mount_wq;
4662 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !sb_rdonly(sb) &&
4663 !ext4_has_feature_encrypt(sb)) {
4664 ext4_set_feature_encrypt(sb);
4665 ext4_commit_super(sb, 1);
4669 * Get the # of file system overhead blocks from the
4670 * superblock if present.
4672 if (es->s_overhead_clusters)
4673 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4675 err = ext4_calculate_overhead(sb);
4677 goto failed_mount_wq;
4681 * The maximum number of concurrent works can be high and
4682 * concurrency isn't really necessary. Limit it to 1.
4684 EXT4_SB(sb)->rsv_conversion_wq =
4685 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4686 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4687 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
4693 * The jbd2_journal_load will have done any necessary log recovery,
4694 * so we can safely mount the rest of the filesystem now.
4697 root = ext4_iget(sb, EXT4_ROOT_INO, EXT4_IGET_SPECIAL);
4699 ext4_msg(sb, KERN_ERR, "get root inode failed");
4700 ret = PTR_ERR(root);
4704 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4705 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4710 #ifdef CONFIG_UNICODE
4711 if (sbi->s_encoding)
4712 sb->s_d_op = &ext4_dentry_ops;
4715 sb->s_root = d_make_root(root);
4717 ext4_msg(sb, KERN_ERR, "get root dentry failed");
4722 ret = ext4_setup_super(sb, es, sb_rdonly(sb));
4723 if (ret == -EROFS) {
4724 sb->s_flags |= SB_RDONLY;
4727 goto failed_mount4a;
4729 ext4_set_resv_clusters(sb);
4731 if (test_opt(sb, BLOCK_VALIDITY)) {
4732 err = ext4_setup_system_zone(sb);
4734 ext4_msg(sb, KERN_ERR, "failed to initialize system "
4736 goto failed_mount4a;
4741 err = ext4_mb_init(sb);
4743 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4748 block = ext4_count_free_clusters(sb);
4749 ext4_free_blocks_count_set(sbi->s_es,
4750 EXT4_C2B(sbi, block));
4751 ext4_superblock_csum_set(sb);
4752 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4755 unsigned long freei = ext4_count_free_inodes(sb);
4756 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
4757 ext4_superblock_csum_set(sb);
4758 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4762 err = percpu_counter_init(&sbi->s_dirs_counter,
4763 ext4_count_dirs(sb), GFP_KERNEL);
4765 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4768 err = percpu_init_rwsem(&sbi->s_writepages_rwsem);
4771 ext4_msg(sb, KERN_ERR, "insufficient memory");
4775 if (ext4_has_feature_flex_bg(sb))
4776 if (!ext4_fill_flex_info(sb)) {
4777 ext4_msg(sb, KERN_ERR,
4778 "unable to initialize "
4779 "flex_bg meta info!");
4783 err = ext4_register_li_request(sb, first_not_zeroed);
4787 err = ext4_register_sysfs(sb);
4792 /* Enable quota usage during mount. */
4793 if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) {
4794 err = ext4_enable_quotas(sb);
4798 #endif /* CONFIG_QUOTA */
4801 * Save the original bdev mapping's wb_err value which could be
4802 * used to detect the metadata async write error.
4804 spin_lock_init(&sbi->s_bdev_wb_lock);
4806 errseq_check_and_advance(&sb->s_bdev->bd_inode->i_mapping->wb_err,
4807 &sbi->s_bdev_wb_err);
4808 sb->s_bdev->bd_super = sb;
4809 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4810 ext4_orphan_cleanup(sb, es);
4811 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4812 if (needs_recovery) {
4813 ext4_msg(sb, KERN_INFO, "recovery complete");
4814 err = ext4_mark_recovery_complete(sb, es);
4818 if (EXT4_SB(sb)->s_journal) {
4819 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4820 descr = " journalled data mode";
4821 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4822 descr = " ordered data mode";
4824 descr = " writeback data mode";
4826 descr = "out journal";
4828 if (test_opt(sb, DISCARD)) {
4829 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4830 if (!blk_queue_discard(q))
4831 ext4_msg(sb, KERN_WARNING,
4832 "mounting with \"discard\" option, but "
4833 "the device does not support discard");
4836 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
4837 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4838 "Opts: %.*s%s%s", descr,
4839 (int) sizeof(sbi->s_es->s_mount_opts),
4840 sbi->s_es->s_mount_opts,
4841 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4843 if (es->s_error_count)
4844 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4846 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4847 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4848 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4849 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4850 atomic_set(&sbi->s_warning_count, 0);
4851 atomic_set(&sbi->s_msg_count, 0);
4858 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4862 ext4_unregister_sysfs(sb);
4864 ext4_unregister_li_request(sb);
4866 ext4_mb_release(sb);
4868 flex_groups = rcu_dereference(sbi->s_flex_groups);
4870 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
4871 kvfree(flex_groups[i]);
4872 kvfree(flex_groups);
4875 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4876 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4877 percpu_counter_destroy(&sbi->s_dirs_counter);
4878 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4879 percpu_free_rwsem(&sbi->s_writepages_rwsem);
4881 ext4_ext_release(sb);
4882 ext4_release_system_zone(sb);
4887 ext4_msg(sb, KERN_ERR, "mount failed");
4888 if (EXT4_SB(sb)->rsv_conversion_wq)
4889 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4891 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
4892 sbi->s_ea_inode_cache = NULL;
4894 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
4895 sbi->s_ea_block_cache = NULL;
4897 if (sbi->s_journal) {
4898 jbd2_journal_destroy(sbi->s_journal);
4899 sbi->s_journal = NULL;
4902 ext4_es_unregister_shrinker(sbi);
4904 del_timer_sync(&sbi->s_err_report);
4906 kthread_stop(sbi->s_mmp_tsk);
4909 group_desc = rcu_dereference(sbi->s_group_desc);
4910 for (i = 0; i < db_count; i++)
4911 brelse(group_desc[i]);
4915 if (sbi->s_chksum_driver)
4916 crypto_free_shash(sbi->s_chksum_driver);
4918 #ifdef CONFIG_UNICODE
4919 utf8_unload(sbi->s_encoding);
4923 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4924 kfree(get_qf_name(sb, sbi, i));
4926 fscrypt_free_dummy_context(&sbi->s_dummy_enc_ctx);
4927 ext4_blkdev_remove(sbi);
4930 sb->s_fs_info = NULL;
4931 kfree(sbi->s_blockgroup_lock);
4935 fs_put_dax(dax_dev);
4936 return err ? err : ret;
4940 * Setup any per-fs journal parameters now. We'll do this both on
4941 * initial mount, once the journal has been initialised but before we've
4942 * done any recovery; and again on any subsequent remount.
4944 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4946 struct ext4_sb_info *sbi = EXT4_SB(sb);
4948 journal->j_commit_interval = sbi->s_commit_interval;
4949 journal->j_min_batch_time = sbi->s_min_batch_time;
4950 journal->j_max_batch_time = sbi->s_max_batch_time;
4952 write_lock(&journal->j_state_lock);
4953 if (test_opt(sb, BARRIER))
4954 journal->j_flags |= JBD2_BARRIER;
4956 journal->j_flags &= ~JBD2_BARRIER;
4957 if (test_opt(sb, DATA_ERR_ABORT))
4958 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4960 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4961 write_unlock(&journal->j_state_lock);
4964 static struct inode *ext4_get_journal_inode(struct super_block *sb,
4965 unsigned int journal_inum)
4967 struct inode *journal_inode;
4970 * Test for the existence of a valid inode on disk. Bad things
4971 * happen if we iget() an unused inode, as the subsequent iput()
4972 * will try to delete it.
4974 journal_inode = ext4_iget(sb, journal_inum, EXT4_IGET_SPECIAL);
4975 if (IS_ERR(journal_inode)) {
4976 ext4_msg(sb, KERN_ERR, "no journal found");
4979 if (!journal_inode->i_nlink) {
4980 make_bad_inode(journal_inode);
4981 iput(journal_inode);
4982 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4986 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4987 journal_inode, journal_inode->i_size);
4988 if (!S_ISREG(journal_inode->i_mode)) {
4989 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4990 iput(journal_inode);
4993 return journal_inode;
4996 static journal_t *ext4_get_journal(struct super_block *sb,
4997 unsigned int journal_inum)
4999 struct inode *journal_inode;
5002 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5005 journal_inode = ext4_get_journal_inode(sb, journal_inum);
5009 journal = jbd2_journal_init_inode(journal_inode);
5011 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
5012 iput(journal_inode);
5015 journal->j_private = sb;
5016 ext4_init_journal_params(sb, journal);
5020 static journal_t *ext4_get_dev_journal(struct super_block *sb,
5023 struct buffer_head *bh;
5027 int hblock, blocksize;
5028 ext4_fsblk_t sb_block;
5029 unsigned long offset;
5030 struct ext4_super_block *es;
5031 struct block_device *bdev;
5033 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5036 bdev = ext4_blkdev_get(j_dev, sb);
5040 blocksize = sb->s_blocksize;
5041 hblock = bdev_logical_block_size(bdev);
5042 if (blocksize < hblock) {
5043 ext4_msg(sb, KERN_ERR,
5044 "blocksize too small for journal device");
5048 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
5049 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
5050 set_blocksize(bdev, blocksize);
5051 if (!(bh = __bread(bdev, sb_block, blocksize))) {
5052 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
5053 "external journal");
5057 es = (struct ext4_super_block *) (bh->b_data + offset);
5058 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
5059 !(le32_to_cpu(es->s_feature_incompat) &
5060 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
5061 ext4_msg(sb, KERN_ERR, "external journal has "
5067 if ((le32_to_cpu(es->s_feature_ro_compat) &
5068 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
5069 es->s_checksum != ext4_superblock_csum(sb, es)) {
5070 ext4_msg(sb, KERN_ERR, "external journal has "
5071 "corrupt superblock");
5076 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
5077 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
5082 len = ext4_blocks_count(es);
5083 start = sb_block + 1;
5084 brelse(bh); /* we're done with the superblock */
5086 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
5087 start, len, blocksize);
5089 ext4_msg(sb, KERN_ERR, "failed to create device journal");
5092 journal->j_private = sb;
5093 ll_rw_block(REQ_OP_READ, REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
5094 wait_on_buffer(journal->j_sb_buffer);
5095 if (!buffer_uptodate(journal->j_sb_buffer)) {
5096 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
5099 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
5100 ext4_msg(sb, KERN_ERR, "External journal has more than one "
5101 "user (unsupported) - %d",
5102 be32_to_cpu(journal->j_superblock->s_nr_users));
5105 EXT4_SB(sb)->journal_bdev = bdev;
5106 ext4_init_journal_params(sb, journal);
5110 jbd2_journal_destroy(journal);
5112 ext4_blkdev_put(bdev);
5116 static int ext4_load_journal(struct super_block *sb,
5117 struct ext4_super_block *es,
5118 unsigned long journal_devnum)
5121 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
5124 int really_read_only;
5127 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5128 return -EFSCORRUPTED;
5130 if (journal_devnum &&
5131 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
5132 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
5133 "numbers have changed");
5134 journal_dev = new_decode_dev(journal_devnum);
5136 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
5138 if (journal_inum && journal_dev) {
5139 ext4_msg(sb, KERN_ERR,
5140 "filesystem has both journal inode and journal device!");
5145 journal = ext4_get_journal(sb, journal_inum);
5149 journal = ext4_get_dev_journal(sb, journal_dev);
5154 journal_dev_ro = bdev_read_only(journal->j_dev);
5155 really_read_only = bdev_read_only(sb->s_bdev) | journal_dev_ro;
5157 if (journal_dev_ro && !sb_rdonly(sb)) {
5158 ext4_msg(sb, KERN_ERR,
5159 "journal device read-only, try mounting with '-o ro'");
5165 * Are we loading a blank journal or performing recovery after a
5166 * crash? For recovery, we need to check in advance whether we
5167 * can get read-write access to the device.
5169 if (ext4_has_feature_journal_needs_recovery(sb)) {
5170 if (sb_rdonly(sb)) {
5171 ext4_msg(sb, KERN_INFO, "INFO: recovery "
5172 "required on readonly filesystem");
5173 if (really_read_only) {
5174 ext4_msg(sb, KERN_ERR, "write access "
5175 "unavailable, cannot proceed "
5176 "(try mounting with noload)");
5180 ext4_msg(sb, KERN_INFO, "write access will "
5181 "be enabled during recovery");
5185 if (!(journal->j_flags & JBD2_BARRIER))
5186 ext4_msg(sb, KERN_INFO, "barriers disabled");
5188 if (!ext4_has_feature_journal_needs_recovery(sb))
5189 err = jbd2_journal_wipe(journal, !really_read_only);
5191 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
5193 memcpy(save, ((char *) es) +
5194 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
5195 err = jbd2_journal_load(journal);
5197 memcpy(((char *) es) + EXT4_S_ERR_START,
5198 save, EXT4_S_ERR_LEN);
5203 ext4_msg(sb, KERN_ERR, "error loading journal");
5207 EXT4_SB(sb)->s_journal = journal;
5208 err = ext4_clear_journal_err(sb, es);
5210 EXT4_SB(sb)->s_journal = NULL;
5211 jbd2_journal_destroy(journal);
5215 if (!really_read_only && journal_devnum &&
5216 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
5217 es->s_journal_dev = cpu_to_le32(journal_devnum);
5219 /* Make sure we flush the recovery flag to disk. */
5220 ext4_commit_super(sb, 1);
5226 jbd2_journal_destroy(journal);
5230 static int ext4_commit_super(struct super_block *sb, int sync)
5232 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
5233 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
5236 if (!sbh || block_device_ejected(sb))
5240 * If the file system is mounted read-only, don't update the
5241 * superblock write time. This avoids updating the superblock
5242 * write time when we are mounting the root file system
5243 * read/only but we need to replay the journal; at that point,
5244 * for people who are east of GMT and who make their clock
5245 * tick in localtime for Windows bug-for-bug compatibility,
5246 * the clock is set in the future, and this will cause e2fsck
5247 * to complain and force a full file system check.
5249 if (!(sb->s_flags & SB_RDONLY))
5250 ext4_update_tstamp(es, s_wtime);
5251 if (sb->s_bdev->bd_part)
5252 es->s_kbytes_written =
5253 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
5254 ((part_stat_read(sb->s_bdev->bd_part,
5255 sectors[STAT_WRITE]) -
5256 EXT4_SB(sb)->s_sectors_written_start) >> 1));
5258 es->s_kbytes_written =
5259 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
5260 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
5261 ext4_free_blocks_count_set(es,
5262 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
5263 &EXT4_SB(sb)->s_freeclusters_counter)));
5264 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
5265 es->s_free_inodes_count =
5266 cpu_to_le32(percpu_counter_sum_positive(
5267 &EXT4_SB(sb)->s_freeinodes_counter));
5268 BUFFER_TRACE(sbh, "marking dirty");
5269 ext4_superblock_csum_set(sb);
5272 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
5274 * Oh, dear. A previous attempt to write the
5275 * superblock failed. This could happen because the
5276 * USB device was yanked out. Or it could happen to
5277 * be a transient write error and maybe the block will
5278 * be remapped. Nothing we can do but to retry the
5279 * write and hope for the best.
5281 ext4_msg(sb, KERN_ERR, "previous I/O error to "
5282 "superblock detected");
5283 clear_buffer_write_io_error(sbh);
5284 set_buffer_uptodate(sbh);
5286 mark_buffer_dirty(sbh);
5289 error = __sync_dirty_buffer(sbh,
5290 REQ_SYNC | (test_opt(sb, BARRIER) ? REQ_FUA : 0));
5291 if (buffer_write_io_error(sbh)) {
5292 ext4_msg(sb, KERN_ERR, "I/O error while writing "
5294 clear_buffer_write_io_error(sbh);
5295 set_buffer_uptodate(sbh);
5302 * Have we just finished recovery? If so, and if we are mounting (or
5303 * remounting) the filesystem readonly, then we will end up with a
5304 * consistent fs on disk. Record that fact.
5306 static int ext4_mark_recovery_complete(struct super_block *sb,
5307 struct ext4_super_block *es)
5310 journal_t *journal = EXT4_SB(sb)->s_journal;
5312 if (!ext4_has_feature_journal(sb)) {
5313 if (journal != NULL) {
5314 ext4_error(sb, "Journal got removed while the fs was "
5316 return -EFSCORRUPTED;
5320 jbd2_journal_lock_updates(journal);
5321 err = jbd2_journal_flush(journal);
5325 if (ext4_has_feature_journal_needs_recovery(sb) && sb_rdonly(sb)) {
5326 ext4_clear_feature_journal_needs_recovery(sb);
5327 ext4_commit_super(sb, 1);
5330 jbd2_journal_unlock_updates(journal);
5335 * If we are mounting (or read-write remounting) a filesystem whose journal
5336 * has recorded an error from a previous lifetime, move that error to the
5337 * main filesystem now.
5339 static int ext4_clear_journal_err(struct super_block *sb,
5340 struct ext4_super_block *es)
5346 if (!ext4_has_feature_journal(sb)) {
5347 ext4_error(sb, "Journal got removed while the fs was mounted!");
5348 return -EFSCORRUPTED;
5351 journal = EXT4_SB(sb)->s_journal;
5354 * Now check for any error status which may have been recorded in the
5355 * journal by a prior ext4_error() or ext4_abort()
5358 j_errno = jbd2_journal_errno(journal);
5362 errstr = ext4_decode_error(sb, j_errno, nbuf);
5363 ext4_warning(sb, "Filesystem error recorded "
5364 "from previous mount: %s", errstr);
5365 ext4_warning(sb, "Marking fs in need of filesystem check.");
5367 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
5368 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
5369 ext4_commit_super(sb, 1);
5371 jbd2_journal_clear_err(journal);
5372 jbd2_journal_update_sb_errno(journal);
5378 * Force the running and committing transactions to commit,
5379 * and wait on the commit.
5381 int ext4_force_commit(struct super_block *sb)
5388 journal = EXT4_SB(sb)->s_journal;
5389 return ext4_journal_force_commit(journal);
5392 static int ext4_sync_fs(struct super_block *sb, int wait)
5396 bool needs_barrier = false;
5397 struct ext4_sb_info *sbi = EXT4_SB(sb);
5399 if (unlikely(ext4_forced_shutdown(sbi)))
5402 trace_ext4_sync_fs(sb, wait);
5403 flush_workqueue(sbi->rsv_conversion_wq);
5405 * Writeback quota in non-journalled quota case - journalled quota has
5408 dquot_writeback_dquots(sb, -1);
5410 * Data writeback is possible w/o journal transaction, so barrier must
5411 * being sent at the end of the function. But we can skip it if
5412 * transaction_commit will do it for us.
5414 if (sbi->s_journal) {
5415 target = jbd2_get_latest_transaction(sbi->s_journal);
5416 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
5417 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
5418 needs_barrier = true;
5420 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
5422 ret = jbd2_log_wait_commit(sbi->s_journal,
5425 } else if (wait && test_opt(sb, BARRIER))
5426 needs_barrier = true;
5427 if (needs_barrier) {
5429 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL);
5438 * LVM calls this function before a (read-only) snapshot is created. This
5439 * gives us a chance to flush the journal completely and mark the fs clean.
5441 * Note that only this function cannot bring a filesystem to be in a clean
5442 * state independently. It relies on upper layer to stop all data & metadata
5445 static int ext4_freeze(struct super_block *sb)
5453 journal = EXT4_SB(sb)->s_journal;
5456 /* Now we set up the journal barrier. */
5457 jbd2_journal_lock_updates(journal);
5460 * Don't clear the needs_recovery flag if we failed to
5461 * flush the journal.
5463 error = jbd2_journal_flush(journal);
5467 /* Journal blocked and flushed, clear needs_recovery flag. */
5468 ext4_clear_feature_journal_needs_recovery(sb);
5471 error = ext4_commit_super(sb, 1);
5474 /* we rely on upper layer to stop further updates */
5475 jbd2_journal_unlock_updates(journal);
5480 * Called by LVM after the snapshot is done. We need to reset the RECOVER
5481 * flag here, even though the filesystem is not technically dirty yet.
5483 static int ext4_unfreeze(struct super_block *sb)
5485 if (sb_rdonly(sb) || ext4_forced_shutdown(EXT4_SB(sb)))
5488 if (EXT4_SB(sb)->s_journal) {
5489 /* Reset the needs_recovery flag before the fs is unlocked. */
5490 ext4_set_feature_journal_needs_recovery(sb);
5493 ext4_commit_super(sb, 1);
5498 * Structure to save mount options for ext4_remount's benefit
5500 struct ext4_mount_options {
5501 unsigned long s_mount_opt;
5502 unsigned long s_mount_opt2;
5505 unsigned long s_commit_interval;
5506 u32 s_min_batch_time, s_max_batch_time;
5509 char *s_qf_names[EXT4_MAXQUOTAS];
5513 static int ext4_remount(struct super_block *sb, int *flags, char *data)
5515 struct ext4_super_block *es;
5516 struct ext4_sb_info *sbi = EXT4_SB(sb);
5517 unsigned long old_sb_flags, vfs_flags;
5518 struct ext4_mount_options old_opts;
5519 int enable_quota = 0;
5521 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
5525 char *to_free[EXT4_MAXQUOTAS];
5527 char *orig_data = kstrdup(data, GFP_KERNEL);
5529 if (data && !orig_data)
5532 /* Store the original options */
5533 old_sb_flags = sb->s_flags;
5534 old_opts.s_mount_opt = sbi->s_mount_opt;
5535 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
5536 old_opts.s_resuid = sbi->s_resuid;
5537 old_opts.s_resgid = sbi->s_resgid;
5538 old_opts.s_commit_interval = sbi->s_commit_interval;
5539 old_opts.s_min_batch_time = sbi->s_min_batch_time;
5540 old_opts.s_max_batch_time = sbi->s_max_batch_time;
5542 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
5543 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5544 if (sbi->s_qf_names[i]) {
5545 char *qf_name = get_qf_name(sb, sbi, i);
5547 old_opts.s_qf_names[i] = kstrdup(qf_name, GFP_KERNEL);
5548 if (!old_opts.s_qf_names[i]) {
5549 for (j = 0; j < i; j++)
5550 kfree(old_opts.s_qf_names[j]);
5555 old_opts.s_qf_names[i] = NULL;
5557 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
5558 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
5561 * Some options can be enabled by ext4 and/or by VFS mount flag
5562 * either way we need to make sure it matches in both *flags and
5563 * s_flags. Copy those selected flags from *flags to s_flags
5565 vfs_flags = SB_LAZYTIME | SB_I_VERSION;
5566 sb->s_flags = (sb->s_flags & ~vfs_flags) | (*flags & vfs_flags);
5568 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
5573 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
5574 test_opt(sb, JOURNAL_CHECKSUM)) {
5575 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
5576 "during remount not supported; ignoring");
5577 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
5580 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
5581 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
5582 ext4_msg(sb, KERN_ERR, "can't mount with "
5583 "both data=journal and delalloc");
5587 if (test_opt(sb, DIOREAD_NOLOCK)) {
5588 ext4_msg(sb, KERN_ERR, "can't mount with "
5589 "both data=journal and dioread_nolock");
5593 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
5594 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
5595 ext4_msg(sb, KERN_ERR, "can't mount with "
5596 "journal_async_commit in data=ordered mode");
5602 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) {
5603 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount");
5608 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
5609 ext4_abort(sb, EXT4_ERR_ESHUTDOWN, "Abort forced by user");
5611 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
5612 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
5616 if (sbi->s_journal) {
5617 ext4_init_journal_params(sb, sbi->s_journal);
5618 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
5621 if ((bool)(*flags & SB_RDONLY) != sb_rdonly(sb)) {
5622 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
5627 if (*flags & SB_RDONLY) {
5628 err = sync_filesystem(sb);
5631 err = dquot_suspend(sb, -1);
5636 * First of all, the unconditional stuff we have to do
5637 * to disable replay of the journal when we next remount
5639 sb->s_flags |= SB_RDONLY;
5642 * OK, test if we are remounting a valid rw partition
5643 * readonly, and if so set the rdonly flag and then
5644 * mark the partition as valid again.
5646 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
5647 (sbi->s_mount_state & EXT4_VALID_FS))
5648 es->s_state = cpu_to_le16(sbi->s_mount_state);
5650 if (sbi->s_journal) {
5652 * We let remount-ro finish even if marking fs
5653 * as clean failed...
5655 ext4_mark_recovery_complete(sb, es);
5658 kthread_stop(sbi->s_mmp_tsk);
5660 /* Make sure we can mount this feature set readwrite */
5661 if (ext4_has_feature_readonly(sb) ||
5662 !ext4_feature_set_ok(sb, 0)) {
5667 * Make sure the group descriptor checksums
5668 * are sane. If they aren't, refuse to remount r/w.
5670 for (g = 0; g < sbi->s_groups_count; g++) {
5671 struct ext4_group_desc *gdp =
5672 ext4_get_group_desc(sb, g, NULL);
5674 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
5675 ext4_msg(sb, KERN_ERR,
5676 "ext4_remount: Checksum for group %u failed (%u!=%u)",
5677 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
5678 le16_to_cpu(gdp->bg_checksum));
5685 * If we have an unprocessed orphan list hanging
5686 * around from a previously readonly bdev mount,
5687 * require a full umount/remount for now.
5689 if (es->s_last_orphan) {
5690 ext4_msg(sb, KERN_WARNING, "Couldn't "
5691 "remount RDWR because of unprocessed "
5692 "orphan inode list. Please "
5693 "umount/remount instead");
5699 * Update the original bdev mapping's wb_err value
5700 * which could be used to detect the metadata async
5703 errseq_check_and_advance(&sb->s_bdev->bd_inode->i_mapping->wb_err,
5704 &sbi->s_bdev_wb_err);
5707 * Mounting a RDONLY partition read-write, so reread
5708 * and store the current valid flag. (It may have
5709 * been changed by e2fsck since we originally mounted
5712 if (sbi->s_journal) {
5713 err = ext4_clear_journal_err(sb, es);
5717 sbi->s_mount_state = le16_to_cpu(es->s_state);
5719 err = ext4_setup_super(sb, es, 0);
5723 sb->s_flags &= ~SB_RDONLY;
5724 if (ext4_has_feature_mmp(sb))
5725 if (ext4_multi_mount_protect(sb,
5726 le64_to_cpu(es->s_mmp_block))) {
5735 * Reinitialize lazy itable initialization thread based on
5738 if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE))
5739 ext4_unregister_li_request(sb);
5741 ext4_group_t first_not_zeroed;
5742 first_not_zeroed = ext4_has_uninit_itable(sb);
5743 ext4_register_li_request(sb, first_not_zeroed);
5747 * Handle creation of system zone data early because it can fail.
5748 * Releasing of existing data is done when we are sure remount will
5751 if (test_opt(sb, BLOCK_VALIDITY) && !sbi->system_blks) {
5752 err = ext4_setup_system_zone(sb);
5757 if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY)) {
5758 err = ext4_commit_super(sb, 1);
5764 /* Release old quota file names */
5765 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5766 kfree(old_opts.s_qf_names[i]);
5768 if (sb_any_quota_suspended(sb))
5769 dquot_resume(sb, -1);
5770 else if (ext4_has_feature_quota(sb)) {
5771 err = ext4_enable_quotas(sb);
5777 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->system_blks)
5778 ext4_release_system_zone(sb);
5781 * Some options can be enabled by ext4 and/or by VFS mount flag
5782 * either way we need to make sure it matches in both *flags and
5783 * s_flags. Copy those selected flags from s_flags to *flags
5785 *flags = (*flags & ~vfs_flags) | (sb->s_flags & vfs_flags);
5787 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5792 sb->s_flags = old_sb_flags;
5793 sbi->s_mount_opt = old_opts.s_mount_opt;
5794 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
5795 sbi->s_resuid = old_opts.s_resuid;
5796 sbi->s_resgid = old_opts.s_resgid;
5797 sbi->s_commit_interval = old_opts.s_commit_interval;
5798 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5799 sbi->s_max_batch_time = old_opts.s_max_batch_time;
5800 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->system_blks)
5801 ext4_release_system_zone(sb);
5803 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
5804 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
5805 to_free[i] = get_qf_name(sb, sbi, i);
5806 rcu_assign_pointer(sbi->s_qf_names[i], old_opts.s_qf_names[i]);
5809 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5817 static int ext4_statfs_project(struct super_block *sb,
5818 kprojid_t projid, struct kstatfs *buf)
5821 struct dquot *dquot;
5825 qid = make_kqid_projid(projid);
5826 dquot = dqget(sb, qid);
5828 return PTR_ERR(dquot);
5829 spin_lock(&dquot->dq_dqb_lock);
5831 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
5832 dquot->dq_dqb.dqb_bhardlimit);
5833 limit >>= sb->s_blocksize_bits;
5835 if (limit && buf->f_blocks > limit) {
5836 curblock = (dquot->dq_dqb.dqb_curspace +
5837 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
5838 buf->f_blocks = limit;
5839 buf->f_bfree = buf->f_bavail =
5840 (buf->f_blocks > curblock) ?
5841 (buf->f_blocks - curblock) : 0;
5844 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
5845 dquot->dq_dqb.dqb_ihardlimit);
5846 if (limit && buf->f_files > limit) {
5847 buf->f_files = limit;
5849 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
5850 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
5853 spin_unlock(&dquot->dq_dqb_lock);
5859 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
5861 struct super_block *sb = dentry->d_sb;
5862 struct ext4_sb_info *sbi = EXT4_SB(sb);
5863 struct ext4_super_block *es = sbi->s_es;
5864 ext4_fsblk_t overhead = 0, resv_blocks;
5867 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
5869 if (!test_opt(sb, MINIX_DF))
5870 overhead = sbi->s_overhead;
5872 buf->f_type = EXT4_SUPER_MAGIC;
5873 buf->f_bsize = sb->s_blocksize;
5874 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
5875 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5876 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
5877 /* prevent underflow in case that few free space is available */
5878 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
5879 buf->f_bavail = buf->f_bfree -
5880 (ext4_r_blocks_count(es) + resv_blocks);
5881 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
5883 buf->f_files = le32_to_cpu(es->s_inodes_count);
5884 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5885 buf->f_namelen = EXT4_NAME_LEN;
5886 fsid = le64_to_cpup((void *)es->s_uuid) ^
5887 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5888 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5889 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
5892 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
5893 sb_has_quota_limits_enabled(sb, PRJQUOTA))
5894 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
5903 * Helper functions so that transaction is started before we acquire dqio_sem
5904 * to keep correct lock ordering of transaction > dqio_sem
5906 static inline struct inode *dquot_to_inode(struct dquot *dquot)
5908 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
5911 static int ext4_write_dquot(struct dquot *dquot)
5915 struct inode *inode;
5917 inode = dquot_to_inode(dquot);
5918 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
5919 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
5921 return PTR_ERR(handle);
5922 ret = dquot_commit(dquot);
5923 err = ext4_journal_stop(handle);
5929 static int ext4_acquire_dquot(struct dquot *dquot)
5934 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5935 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
5937 return PTR_ERR(handle);
5938 ret = dquot_acquire(dquot);
5939 err = ext4_journal_stop(handle);
5945 static int ext4_release_dquot(struct dquot *dquot)
5950 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5951 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
5952 if (IS_ERR(handle)) {
5953 /* Release dquot anyway to avoid endless cycle in dqput() */
5954 dquot_release(dquot);
5955 return PTR_ERR(handle);
5957 ret = dquot_release(dquot);
5958 err = ext4_journal_stop(handle);
5964 static int ext4_mark_dquot_dirty(struct dquot *dquot)
5966 struct super_block *sb = dquot->dq_sb;
5967 struct ext4_sb_info *sbi = EXT4_SB(sb);
5969 /* Are we journaling quotas? */
5970 if (ext4_has_feature_quota(sb) ||
5971 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
5972 dquot_mark_dquot_dirty(dquot);
5973 return ext4_write_dquot(dquot);
5975 return dquot_mark_dquot_dirty(dquot);
5979 static int ext4_write_info(struct super_block *sb, int type)
5984 /* Data block + inode block */
5985 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
5987 return PTR_ERR(handle);
5988 ret = dquot_commit_info(sb, type);
5989 err = ext4_journal_stop(handle);
5996 * Turn on quotas during mount time - we need to find
5997 * the quota file and such...
5999 static int ext4_quota_on_mount(struct super_block *sb, int type)
6001 return dquot_quota_on_mount(sb, get_qf_name(sb, EXT4_SB(sb), type),
6002 EXT4_SB(sb)->s_jquota_fmt, type);
6005 static void lockdep_set_quota_inode(struct inode *inode, int subclass)
6007 struct ext4_inode_info *ei = EXT4_I(inode);
6009 /* The first argument of lockdep_set_subclass has to be
6010 * *exactly* the same as the argument to init_rwsem() --- in
6011 * this case, in init_once() --- or lockdep gets unhappy
6012 * because the name of the lock is set using the
6013 * stringification of the argument to init_rwsem().
6015 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
6016 lockdep_set_subclass(&ei->i_data_sem, subclass);
6020 * Standard function to be called on quota_on
6022 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
6023 const struct path *path)
6027 if (!test_opt(sb, QUOTA))
6030 /* Quotafile not on the same filesystem? */
6031 if (path->dentry->d_sb != sb)
6033 /* Journaling quota? */
6034 if (EXT4_SB(sb)->s_qf_names[type]) {
6035 /* Quotafile not in fs root? */
6036 if (path->dentry->d_parent != sb->s_root)
6037 ext4_msg(sb, KERN_WARNING,
6038 "Quota file not on filesystem root. "
6039 "Journaled quota will not work");
6040 sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY;
6043 * Clear the flag just in case mount options changed since
6046 sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY;
6050 * When we journal data on quota file, we have to flush journal to see
6051 * all updates to the file when we bypass pagecache...
6053 if (EXT4_SB(sb)->s_journal &&
6054 ext4_should_journal_data(d_inode(path->dentry))) {
6056 * We don't need to lock updates but journal_flush() could
6057 * otherwise be livelocked...
6059 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
6060 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
6061 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
6066 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
6067 err = dquot_quota_on(sb, type, format_id, path);
6069 lockdep_set_quota_inode(path->dentry->d_inode,
6072 struct inode *inode = d_inode(path->dentry);
6076 * Set inode flags to prevent userspace from messing with quota
6077 * files. If this fails, we return success anyway since quotas
6078 * are already enabled and this is not a hard failure.
6081 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
6084 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL;
6085 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
6086 S_NOATIME | S_IMMUTABLE);
6087 err = ext4_mark_inode_dirty(handle, inode);
6088 ext4_journal_stop(handle);
6090 inode_unlock(inode);
6095 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
6099 struct inode *qf_inode;
6100 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
6101 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
6102 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
6103 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
6106 BUG_ON(!ext4_has_feature_quota(sb));
6108 if (!qf_inums[type])
6111 qf_inode = ext4_iget(sb, qf_inums[type], EXT4_IGET_SPECIAL);
6112 if (IS_ERR(qf_inode)) {
6113 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
6114 return PTR_ERR(qf_inode);
6117 /* Don't account quota for quota files to avoid recursion */
6118 qf_inode->i_flags |= S_NOQUOTA;
6119 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
6120 err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
6122 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
6128 /* Enable usage tracking for all quota types. */
6129 static int ext4_enable_quotas(struct super_block *sb)
6132 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
6133 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
6134 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
6135 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
6137 bool quota_mopt[EXT4_MAXQUOTAS] = {
6138 test_opt(sb, USRQUOTA),
6139 test_opt(sb, GRPQUOTA),
6140 test_opt(sb, PRJQUOTA),
6143 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY;
6144 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
6145 if (qf_inums[type]) {
6146 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
6147 DQUOT_USAGE_ENABLED |
6148 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
6151 "Failed to enable quota tracking "
6152 "(type=%d, err=%d). Please run "
6153 "e2fsck to fix.", type, err);
6154 for (type--; type >= 0; type--)
6155 dquot_quota_off(sb, type);
6164 static int ext4_quota_off(struct super_block *sb, int type)
6166 struct inode *inode = sb_dqopt(sb)->files[type];
6170 /* Force all delayed allocation blocks to be allocated.
6171 * Caller already holds s_umount sem */
6172 if (test_opt(sb, DELALLOC))
6173 sync_filesystem(sb);
6175 if (!inode || !igrab(inode))
6178 err = dquot_quota_off(sb, type);
6179 if (err || ext4_has_feature_quota(sb))
6184 * Update modification times of quota files when userspace can
6185 * start looking at them. If we fail, we return success anyway since
6186 * this is not a hard failure and quotas are already disabled.
6188 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
6189 if (IS_ERR(handle)) {
6190 err = PTR_ERR(handle);
6193 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL);
6194 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
6195 inode->i_mtime = inode->i_ctime = current_time(inode);
6196 err = ext4_mark_inode_dirty(handle, inode);
6197 ext4_journal_stop(handle);
6199 inode_unlock(inode);
6201 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL);
6205 return dquot_quota_off(sb, type);
6208 /* Read data from quotafile - avoid pagecache and such because we cannot afford
6209 * acquiring the locks... As quota files are never truncated and quota code
6210 * itself serializes the operations (and no one else should touch the files)
6211 * we don't have to be afraid of races */
6212 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
6213 size_t len, loff_t off)
6215 struct inode *inode = sb_dqopt(sb)->files[type];
6216 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
6217 int offset = off & (sb->s_blocksize - 1);
6220 struct buffer_head *bh;
6221 loff_t i_size = i_size_read(inode);
6225 if (off+len > i_size)
6228 while (toread > 0) {
6229 tocopy = sb->s_blocksize - offset < toread ?
6230 sb->s_blocksize - offset : toread;
6231 bh = ext4_bread(NULL, inode, blk, 0);
6234 if (!bh) /* A hole? */
6235 memset(data, 0, tocopy);
6237 memcpy(data, bh->b_data+offset, tocopy);
6247 /* Write to quotafile (we know the transaction is already started and has
6248 * enough credits) */
6249 static ssize_t ext4_quota_write(struct super_block *sb, int type,
6250 const char *data, size_t len, loff_t off)
6252 struct inode *inode = sb_dqopt(sb)->files[type];
6253 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
6254 int err = 0, err2 = 0, offset = off & (sb->s_blocksize - 1);
6256 struct buffer_head *bh;
6257 handle_t *handle = journal_current_handle();
6259 if (EXT4_SB(sb)->s_journal && !handle) {
6260 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
6261 " cancelled because transaction is not started",
6262 (unsigned long long)off, (unsigned long long)len);
6266 * Since we account only one data block in transaction credits,
6267 * then it is impossible to cross a block boundary.
6269 if (sb->s_blocksize - offset < len) {
6270 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
6271 " cancelled because not block aligned",
6272 (unsigned long long)off, (unsigned long long)len);
6277 bh = ext4_bread(handle, inode, blk,
6278 EXT4_GET_BLOCKS_CREATE |
6279 EXT4_GET_BLOCKS_METADATA_NOFAIL);
6280 } while (PTR_ERR(bh) == -ENOSPC &&
6281 ext4_should_retry_alloc(inode->i_sb, &retries));
6286 BUFFER_TRACE(bh, "get write access");
6287 err = ext4_journal_get_write_access(handle, bh);
6293 memcpy(bh->b_data+offset, data, len);
6294 flush_dcache_page(bh->b_page);
6296 err = ext4_handle_dirty_metadata(handle, NULL, bh);
6299 if (inode->i_size < off + len) {
6300 i_size_write(inode, off + len);
6301 EXT4_I(inode)->i_disksize = inode->i_size;
6302 err2 = ext4_mark_inode_dirty(handle, inode);
6303 if (unlikely(err2 && !err))
6306 return err ? err : len;
6310 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
6311 const char *dev_name, void *data)
6313 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
6316 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
6317 static inline void register_as_ext2(void)
6319 int err = register_filesystem(&ext2_fs_type);
6322 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
6325 static inline void unregister_as_ext2(void)
6327 unregister_filesystem(&ext2_fs_type);
6330 static inline int ext2_feature_set_ok(struct super_block *sb)
6332 if (ext4_has_unknown_ext2_incompat_features(sb))
6336 if (ext4_has_unknown_ext2_ro_compat_features(sb))
6341 static inline void register_as_ext2(void) { }
6342 static inline void unregister_as_ext2(void) { }
6343 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
6346 static inline void register_as_ext3(void)
6348 int err = register_filesystem(&ext3_fs_type);
6351 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
6354 static inline void unregister_as_ext3(void)
6356 unregister_filesystem(&ext3_fs_type);
6359 static inline int ext3_feature_set_ok(struct super_block *sb)
6361 if (ext4_has_unknown_ext3_incompat_features(sb))
6363 if (!ext4_has_feature_journal(sb))
6367 if (ext4_has_unknown_ext3_ro_compat_features(sb))
6372 static struct file_system_type ext4_fs_type = {
6373 .owner = THIS_MODULE,
6375 .mount = ext4_mount,
6376 .kill_sb = kill_block_super,
6377 .fs_flags = FS_REQUIRES_DEV,
6379 MODULE_ALIAS_FS("ext4");
6381 /* Shared across all ext4 file systems */
6382 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
6384 static int __init ext4_init_fs(void)
6388 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
6389 ext4_li_info = NULL;
6390 mutex_init(&ext4_li_mtx);
6392 /* Build-time check for flags consistency */
6393 ext4_check_flag_values();
6395 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
6396 init_waitqueue_head(&ext4__ioend_wq[i]);
6398 err = ext4_init_es();
6402 err = ext4_init_pending();
6406 err = ext4_init_post_read_processing();
6410 err = ext4_init_pageio();
6414 err = ext4_init_system_zone();
6418 err = ext4_init_sysfs();
6422 err = ext4_init_mballoc();
6425 err = init_inodecache();
6430 err = register_filesystem(&ext4_fs_type);
6436 unregister_as_ext2();
6437 unregister_as_ext3();
6438 destroy_inodecache();
6440 ext4_exit_mballoc();
6444 ext4_exit_system_zone();
6448 ext4_exit_post_read_processing();
6450 ext4_exit_pending();
6457 static void __exit ext4_exit_fs(void)
6459 ext4_destroy_lazyinit_thread();
6460 unregister_as_ext2();
6461 unregister_as_ext3();
6462 unregister_filesystem(&ext4_fs_type);
6463 destroy_inodecache();
6464 ext4_exit_mballoc();
6466 ext4_exit_system_zone();
6468 ext4_exit_post_read_processing();
6470 ext4_exit_pending();
6473 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
6474 MODULE_DESCRIPTION("Fourth Extended Filesystem");
6475 MODULE_LICENSE("GPL");
6476 MODULE_SOFTDEP("pre: crc32c");
6477 module_init(ext4_init_fs)
6478 module_exit(ext4_exit_fs)