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>
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 void ext4_mark_recovery_complete(struct super_block *sb,
70 struct ext4_super_block *es);
71 static void 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_sem -> 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_sem
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_sem
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, const char *func,
341 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
343 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
344 if (bdev_read_only(sb->s_bdev))
346 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
347 ext4_update_tstamp(es, s_last_error_time);
348 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
349 es->s_last_error_line = cpu_to_le32(line);
350 if (es->s_last_error_errcode == 0)
351 es->s_last_error_errcode = EXT4_ERR_EFSCORRUPTED;
352 if (!es->s_first_error_time) {
353 es->s_first_error_time = es->s_last_error_time;
354 es->s_first_error_time_hi = es->s_last_error_time_hi;
355 strncpy(es->s_first_error_func, func,
356 sizeof(es->s_first_error_func));
357 es->s_first_error_line = cpu_to_le32(line);
358 es->s_first_error_ino = es->s_last_error_ino;
359 es->s_first_error_block = es->s_last_error_block;
360 es->s_first_error_errcode = es->s_last_error_errcode;
363 * Start the daily error reporting function if it hasn't been
366 if (!es->s_error_count)
367 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
368 le32_add_cpu(&es->s_error_count, 1);
371 static void save_error_info(struct super_block *sb, const char *func,
374 __save_error_info(sb, func, line);
375 ext4_commit_super(sb, 1);
379 * The del_gendisk() function uninitializes the disk-specific data
380 * structures, including the bdi structure, without telling anyone
381 * else. Once this happens, any attempt to call mark_buffer_dirty()
382 * (for example, by ext4_commit_super), will cause a kernel OOPS.
383 * This is a kludge to prevent these oops until we can put in a proper
384 * hook in del_gendisk() to inform the VFS and file system layers.
386 static int block_device_ejected(struct super_block *sb)
388 struct inode *bd_inode = sb->s_bdev->bd_inode;
389 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
391 return bdi->dev == NULL;
394 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
396 struct super_block *sb = journal->j_private;
397 struct ext4_sb_info *sbi = EXT4_SB(sb);
398 int error = is_journal_aborted(journal);
399 struct ext4_journal_cb_entry *jce;
401 BUG_ON(txn->t_state == T_FINISHED);
403 ext4_process_freed_data(sb, txn->t_tid);
405 spin_lock(&sbi->s_md_lock);
406 while (!list_empty(&txn->t_private_list)) {
407 jce = list_entry(txn->t_private_list.next,
408 struct ext4_journal_cb_entry, jce_list);
409 list_del_init(&jce->jce_list);
410 spin_unlock(&sbi->s_md_lock);
411 jce->jce_func(sb, jce, error);
412 spin_lock(&sbi->s_md_lock);
414 spin_unlock(&sbi->s_md_lock);
417 static bool system_going_down(void)
419 return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF
420 || system_state == SYSTEM_RESTART;
423 /* Deal with the reporting of failure conditions on a filesystem such as
424 * inconsistencies detected or read IO failures.
426 * On ext2, we can store the error state of the filesystem in the
427 * superblock. That is not possible on ext4, because we may have other
428 * write ordering constraints on the superblock which prevent us from
429 * writing it out straight away; and given that the journal is about to
430 * be aborted, we can't rely on the current, or future, transactions to
431 * write out the superblock safely.
433 * We'll just use the jbd2_journal_abort() error code to record an error in
434 * the journal instead. On recovery, the journal will complain about
435 * that error until we've noted it down and cleared it.
438 static void ext4_handle_error(struct super_block *sb)
440 if (test_opt(sb, WARN_ON_ERROR))
446 if (!test_opt(sb, ERRORS_CONT)) {
447 journal_t *journal = EXT4_SB(sb)->s_journal;
449 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
451 jbd2_journal_abort(journal, -EIO);
454 * We force ERRORS_RO behavior when system is rebooting. Otherwise we
455 * could panic during 'reboot -f' as the underlying device got already
458 if (test_opt(sb, ERRORS_RO) || system_going_down()) {
459 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
461 * Make sure updated value of ->s_mount_flags will be visible
462 * before ->s_flags update
465 sb->s_flags |= SB_RDONLY;
466 } else if (test_opt(sb, ERRORS_PANIC)) {
467 if (EXT4_SB(sb)->s_journal &&
468 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
470 panic("EXT4-fs (device %s): panic forced after error\n",
475 #define ext4_error_ratelimit(sb) \
476 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
479 void __ext4_error(struct super_block *sb, const char *function,
480 unsigned int line, const char *fmt, ...)
482 struct va_format vaf;
485 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
488 trace_ext4_error(sb, function, line);
489 if (ext4_error_ratelimit(sb)) {
494 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
495 sb->s_id, function, line, current->comm, &vaf);
498 save_error_info(sb, function, line);
499 ext4_handle_error(sb);
502 void __ext4_error_inode(struct inode *inode, const char *function,
503 unsigned int line, ext4_fsblk_t block,
504 const char *fmt, ...)
507 struct va_format vaf;
508 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
510 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
513 trace_ext4_error(inode->i_sb, function, line);
514 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
515 es->s_last_error_block = cpu_to_le64(block);
516 if (ext4_error_ratelimit(inode->i_sb)) {
521 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
522 "inode #%lu: block %llu: comm %s: %pV\n",
523 inode->i_sb->s_id, function, line, inode->i_ino,
524 block, current->comm, &vaf);
526 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
527 "inode #%lu: comm %s: %pV\n",
528 inode->i_sb->s_id, function, line, inode->i_ino,
529 current->comm, &vaf);
532 save_error_info(inode->i_sb, function, line);
533 ext4_handle_error(inode->i_sb);
536 void __ext4_error_file(struct file *file, const char *function,
537 unsigned int line, ext4_fsblk_t block,
538 const char *fmt, ...)
541 struct va_format vaf;
542 struct ext4_super_block *es;
543 struct inode *inode = file_inode(file);
544 char pathname[80], *path;
546 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
549 trace_ext4_error(inode->i_sb, function, line);
550 es = EXT4_SB(inode->i_sb)->s_es;
551 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
552 if (ext4_error_ratelimit(inode->i_sb)) {
553 path = file_path(file, pathname, sizeof(pathname));
561 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
562 "block %llu: comm %s: path %s: %pV\n",
563 inode->i_sb->s_id, function, line, inode->i_ino,
564 block, current->comm, path, &vaf);
567 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
568 "comm %s: path %s: %pV\n",
569 inode->i_sb->s_id, function, line, inode->i_ino,
570 current->comm, path, &vaf);
573 save_error_info(inode->i_sb, function, line);
574 ext4_handle_error(inode->i_sb);
577 const char *ext4_decode_error(struct super_block *sb, int errno,
584 errstr = "Corrupt filesystem";
587 errstr = "Filesystem failed CRC";
590 errstr = "IO failure";
593 errstr = "Out of memory";
596 if (!sb || (EXT4_SB(sb)->s_journal &&
597 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
598 errstr = "Journal has aborted";
600 errstr = "Readonly filesystem";
603 /* If the caller passed in an extra buffer for unknown
604 * errors, textualise them now. Else we just return
607 /* Check for truncated error codes... */
608 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
617 void ext4_set_errno(struct super_block *sb, int err)
627 err = EXT4_ERR_ENOMEM;
630 err = EXT4_ERR_EFSBADCRC;
633 err = EXT4_ERR_EFSCORRUPTED;
636 err = EXT4_ERR_ENOSPC;
639 err = EXT4_ERR_ENOKEY;
642 err = EXT4_ERR_EROFS;
645 err = EXT4_ERR_EFBIG;
648 err = EXT4_ERR_EEXIST;
651 err = EXT4_ERR_ERANGE;
654 err = EXT4_ERR_EOVERFLOW;
657 err = EXT4_ERR_EBUSY;
660 err = EXT4_ERR_ENOTDIR;
663 err = EXT4_ERR_ENOTEMPTY;
666 err = EXT4_ERR_ESHUTDOWN;
669 err = EXT4_ERR_EFAULT;
672 err = EXT4_ERR_UNKNOWN;
674 EXT4_SB(sb)->s_es->s_last_error_errcode = err;
677 /* __ext4_std_error decodes expected errors from journaling functions
678 * automatically and invokes the appropriate error response. */
680 void __ext4_std_error(struct super_block *sb, const char *function,
681 unsigned int line, int errno)
686 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
689 /* Special case: if the error is EROFS, and we're not already
690 * inside a transaction, then there's really no point in logging
692 if (errno == -EROFS && journal_current_handle() == NULL && sb_rdonly(sb))
695 if (ext4_error_ratelimit(sb)) {
696 errstr = ext4_decode_error(sb, errno, nbuf);
697 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
698 sb->s_id, function, line, errstr);
701 ext4_set_errno(sb, -errno);
702 save_error_info(sb, function, line);
703 ext4_handle_error(sb);
707 * ext4_abort is a much stronger failure handler than ext4_error. The
708 * abort function may be used to deal with unrecoverable failures such
709 * as journal IO errors or ENOMEM at a critical moment in log management.
711 * We unconditionally force the filesystem into an ABORT|READONLY state,
712 * unless the error response on the fs has been set to panic in which
713 * case we take the easy way out and panic immediately.
716 void __ext4_abort(struct super_block *sb, const char *function,
717 unsigned int line, const char *fmt, ...)
719 struct va_format vaf;
722 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
725 save_error_info(sb, function, line);
729 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: %pV\n",
730 sb->s_id, function, line, &vaf);
733 if (sb_rdonly(sb) == 0) {
734 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
735 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
737 * Make sure updated value of ->s_mount_flags will be visible
738 * before ->s_flags update
741 sb->s_flags |= SB_RDONLY;
742 if (EXT4_SB(sb)->s_journal)
743 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
744 save_error_info(sb, function, line);
746 if (test_opt(sb, ERRORS_PANIC) && !system_going_down()) {
747 if (EXT4_SB(sb)->s_journal &&
748 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
750 panic("EXT4-fs panic from previous error\n");
754 void __ext4_msg(struct super_block *sb,
755 const char *prefix, const char *fmt, ...)
757 struct va_format vaf;
760 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
766 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
770 #define ext4_warning_ratelimit(sb) \
771 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \
774 void __ext4_warning(struct super_block *sb, const char *function,
775 unsigned int line, const char *fmt, ...)
777 struct va_format vaf;
780 if (!ext4_warning_ratelimit(sb))
786 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
787 sb->s_id, function, line, &vaf);
791 void __ext4_warning_inode(const struct inode *inode, const char *function,
792 unsigned int line, const char *fmt, ...)
794 struct va_format vaf;
797 if (!ext4_warning_ratelimit(inode->i_sb))
803 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
804 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
805 function, line, inode->i_ino, current->comm, &vaf);
809 void __ext4_grp_locked_error(const char *function, unsigned int line,
810 struct super_block *sb, ext4_group_t grp,
811 unsigned long ino, ext4_fsblk_t block,
812 const char *fmt, ...)
816 struct va_format vaf;
818 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
820 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
823 trace_ext4_error(sb, function, line);
824 es->s_last_error_ino = cpu_to_le32(ino);
825 es->s_last_error_block = cpu_to_le64(block);
826 __save_error_info(sb, function, line);
828 if (ext4_error_ratelimit(sb)) {
832 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
833 sb->s_id, function, line, grp);
835 printk(KERN_CONT "inode %lu: ", ino);
837 printk(KERN_CONT "block %llu:",
838 (unsigned long long) block);
839 printk(KERN_CONT "%pV\n", &vaf);
843 if (test_opt(sb, WARN_ON_ERROR))
846 if (test_opt(sb, ERRORS_CONT)) {
847 ext4_commit_super(sb, 0);
851 ext4_unlock_group(sb, grp);
852 ext4_commit_super(sb, 1);
853 ext4_handle_error(sb);
855 * We only get here in the ERRORS_RO case; relocking the group
856 * may be dangerous, but nothing bad will happen since the
857 * filesystem will have already been marked read/only and the
858 * journal has been aborted. We return 1 as a hint to callers
859 * who might what to use the return value from
860 * ext4_grp_locked_error() to distinguish between the
861 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
862 * aggressively from the ext4 function in question, with a
863 * more appropriate error code.
865 ext4_lock_group(sb, grp);
869 void ext4_mark_group_bitmap_corrupted(struct super_block *sb,
873 struct ext4_sb_info *sbi = EXT4_SB(sb);
874 struct ext4_group_info *grp = ext4_get_group_info(sb, group);
875 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, group, NULL);
878 if (flags & EXT4_GROUP_INFO_BBITMAP_CORRUPT) {
879 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT,
882 percpu_counter_sub(&sbi->s_freeclusters_counter,
886 if (flags & EXT4_GROUP_INFO_IBITMAP_CORRUPT) {
887 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT,
892 count = ext4_free_inodes_count(sb, gdp);
893 percpu_counter_sub(&sbi->s_freeinodes_counter,
899 void ext4_update_dynamic_rev(struct super_block *sb)
901 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
903 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
907 "updating to rev %d because of new feature flag, "
908 "running e2fsck is recommended",
911 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
912 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
913 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
914 /* leave es->s_feature_*compat flags alone */
915 /* es->s_uuid will be set by e2fsck if empty */
918 * The rest of the superblock fields should be zero, and if not it
919 * means they are likely already in use, so leave them alone. We
920 * can leave it up to e2fsck to clean up any inconsistencies there.
925 * Open the external journal device
927 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
929 struct block_device *bdev;
930 char b[BDEVNAME_SIZE];
932 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
938 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
939 __bdevname(dev, b), PTR_ERR(bdev));
944 * Release the journal device
946 static void ext4_blkdev_put(struct block_device *bdev)
948 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
951 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
953 struct block_device *bdev;
954 bdev = sbi->journal_bdev;
956 ext4_blkdev_put(bdev);
957 sbi->journal_bdev = NULL;
961 static inline struct inode *orphan_list_entry(struct list_head *l)
963 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
966 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
970 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
971 le32_to_cpu(sbi->s_es->s_last_orphan));
973 printk(KERN_ERR "sb_info orphan list:\n");
974 list_for_each(l, &sbi->s_orphan) {
975 struct inode *inode = orphan_list_entry(l);
977 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
978 inode->i_sb->s_id, inode->i_ino, inode,
979 inode->i_mode, inode->i_nlink,
985 static int ext4_quota_off(struct super_block *sb, int type);
987 static inline void ext4_quota_off_umount(struct super_block *sb)
991 /* Use our quota_off function to clear inode flags etc. */
992 for (type = 0; type < EXT4_MAXQUOTAS; type++)
993 ext4_quota_off(sb, type);
997 * This is a helper function which is used in the mount/remount
998 * codepaths (which holds s_umount) to fetch the quota file name.
1000 static inline char *get_qf_name(struct super_block *sb,
1001 struct ext4_sb_info *sbi,
1004 return rcu_dereference_protected(sbi->s_qf_names[type],
1005 lockdep_is_held(&sb->s_umount));
1008 static inline void ext4_quota_off_umount(struct super_block *sb)
1013 static void ext4_put_super(struct super_block *sb)
1015 struct ext4_sb_info *sbi = EXT4_SB(sb);
1016 struct ext4_super_block *es = sbi->s_es;
1020 ext4_unregister_li_request(sb);
1021 ext4_quota_off_umount(sb);
1023 destroy_workqueue(sbi->rsv_conversion_wq);
1025 if (sbi->s_journal) {
1026 aborted = is_journal_aborted(sbi->s_journal);
1027 err = jbd2_journal_destroy(sbi->s_journal);
1028 sbi->s_journal = NULL;
1029 if ((err < 0) && !aborted) {
1030 ext4_set_errno(sb, -err);
1031 ext4_abort(sb, "Couldn't clean up the journal");
1035 ext4_unregister_sysfs(sb);
1036 ext4_es_unregister_shrinker(sbi);
1037 del_timer_sync(&sbi->s_err_report);
1038 ext4_release_system_zone(sb);
1039 ext4_mb_release(sb);
1040 ext4_ext_release(sb);
1042 if (!sb_rdonly(sb) && !aborted) {
1043 ext4_clear_feature_journal_needs_recovery(sb);
1044 es->s_state = cpu_to_le16(sbi->s_mount_state);
1047 ext4_commit_super(sb, 1);
1049 for (i = 0; i < sbi->s_gdb_count; i++)
1050 brelse(sbi->s_group_desc[i]);
1051 kvfree(sbi->s_group_desc);
1052 kvfree(sbi->s_flex_groups);
1053 percpu_counter_destroy(&sbi->s_freeclusters_counter);
1054 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1055 percpu_counter_destroy(&sbi->s_dirs_counter);
1056 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
1057 percpu_free_rwsem(&sbi->s_journal_flag_rwsem);
1059 for (i = 0; i < EXT4_MAXQUOTAS; i++)
1060 kfree(get_qf_name(sb, sbi, i));
1063 /* Debugging code just in case the in-memory inode orphan list
1064 * isn't empty. The on-disk one can be non-empty if we've
1065 * detected an error and taken the fs readonly, but the
1066 * in-memory list had better be clean by this point. */
1067 if (!list_empty(&sbi->s_orphan))
1068 dump_orphan_list(sb, sbi);
1069 J_ASSERT(list_empty(&sbi->s_orphan));
1071 sync_blockdev(sb->s_bdev);
1072 invalidate_bdev(sb->s_bdev);
1073 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
1075 * Invalidate the journal device's buffers. We don't want them
1076 * floating about in memory - the physical journal device may
1077 * hotswapped, and it breaks the `ro-after' testing code.
1079 sync_blockdev(sbi->journal_bdev);
1080 invalidate_bdev(sbi->journal_bdev);
1081 ext4_blkdev_remove(sbi);
1084 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
1085 sbi->s_ea_inode_cache = NULL;
1087 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
1088 sbi->s_ea_block_cache = NULL;
1091 kthread_stop(sbi->s_mmp_tsk);
1093 sb->s_fs_info = NULL;
1095 * Now that we are completely done shutting down the
1096 * superblock, we need to actually destroy the kobject.
1098 kobject_put(&sbi->s_kobj);
1099 wait_for_completion(&sbi->s_kobj_unregister);
1100 if (sbi->s_chksum_driver)
1101 crypto_free_shash(sbi->s_chksum_driver);
1102 kfree(sbi->s_blockgroup_lock);
1103 fs_put_dax(sbi->s_daxdev);
1104 #ifdef CONFIG_UNICODE
1105 utf8_unload(sbi->s_encoding);
1110 static struct kmem_cache *ext4_inode_cachep;
1113 * Called inside transaction, so use GFP_NOFS
1115 static struct inode *ext4_alloc_inode(struct super_block *sb)
1117 struct ext4_inode_info *ei;
1119 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
1123 inode_set_iversion(&ei->vfs_inode, 1);
1124 spin_lock_init(&ei->i_raw_lock);
1125 INIT_LIST_HEAD(&ei->i_prealloc_list);
1126 spin_lock_init(&ei->i_prealloc_lock);
1127 ext4_es_init_tree(&ei->i_es_tree);
1128 rwlock_init(&ei->i_es_lock);
1129 INIT_LIST_HEAD(&ei->i_es_list);
1130 ei->i_es_all_nr = 0;
1131 ei->i_es_shk_nr = 0;
1132 ei->i_es_shrink_lblk = 0;
1133 ei->i_reserved_data_blocks = 0;
1134 spin_lock_init(&(ei->i_block_reservation_lock));
1135 ext4_init_pending_tree(&ei->i_pending_tree);
1137 ei->i_reserved_quota = 0;
1138 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
1141 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
1142 spin_lock_init(&ei->i_completed_io_lock);
1144 ei->i_datasync_tid = 0;
1145 atomic_set(&ei->i_unwritten, 0);
1146 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
1147 return &ei->vfs_inode;
1150 static int ext4_drop_inode(struct inode *inode)
1152 int drop = generic_drop_inode(inode);
1155 drop = fscrypt_drop_inode(inode);
1157 trace_ext4_drop_inode(inode, drop);
1161 static void ext4_free_in_core_inode(struct inode *inode)
1163 fscrypt_free_inode(inode);
1164 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
1167 static void ext4_destroy_inode(struct inode *inode)
1169 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
1170 ext4_msg(inode->i_sb, KERN_ERR,
1171 "Inode %lu (%p): orphan list check failed!",
1172 inode->i_ino, EXT4_I(inode));
1173 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1174 EXT4_I(inode), sizeof(struct ext4_inode_info),
1180 static void init_once(void *foo)
1182 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
1184 INIT_LIST_HEAD(&ei->i_orphan);
1185 init_rwsem(&ei->xattr_sem);
1186 init_rwsem(&ei->i_data_sem);
1187 init_rwsem(&ei->i_mmap_sem);
1188 inode_init_once(&ei->vfs_inode);
1191 static int __init init_inodecache(void)
1193 ext4_inode_cachep = kmem_cache_create_usercopy("ext4_inode_cache",
1194 sizeof(struct ext4_inode_info), 0,
1195 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD|
1197 offsetof(struct ext4_inode_info, i_data),
1198 sizeof_field(struct ext4_inode_info, i_data),
1200 if (ext4_inode_cachep == NULL)
1205 static void destroy_inodecache(void)
1208 * Make sure all delayed rcu free inodes are flushed before we
1212 kmem_cache_destroy(ext4_inode_cachep);
1215 void ext4_clear_inode(struct inode *inode)
1217 invalidate_inode_buffers(inode);
1219 ext4_discard_preallocations(inode);
1220 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
1222 if (EXT4_I(inode)->jinode) {
1223 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1224 EXT4_I(inode)->jinode);
1225 jbd2_free_inode(EXT4_I(inode)->jinode);
1226 EXT4_I(inode)->jinode = NULL;
1228 fscrypt_put_encryption_info(inode);
1229 fsverity_cleanup_inode(inode);
1232 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1233 u64 ino, u32 generation)
1235 struct inode *inode;
1238 * Currently we don't know the generation for parent directory, so
1239 * a generation of 0 means "accept any"
1241 inode = ext4_iget(sb, ino, EXT4_IGET_HANDLE);
1243 return ERR_CAST(inode);
1244 if (generation && inode->i_generation != generation) {
1246 return ERR_PTR(-ESTALE);
1252 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1253 int fh_len, int fh_type)
1255 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1256 ext4_nfs_get_inode);
1259 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1260 int fh_len, int fh_type)
1262 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1263 ext4_nfs_get_inode);
1266 static int ext4_nfs_commit_metadata(struct inode *inode)
1268 struct writeback_control wbc = {
1269 .sync_mode = WB_SYNC_ALL
1272 trace_ext4_nfs_commit_metadata(inode);
1273 return ext4_write_inode(inode, &wbc);
1277 * Try to release metadata pages (indirect blocks, directories) which are
1278 * mapped via the block device. Since these pages could have journal heads
1279 * which would prevent try_to_free_buffers() from freeing them, we must use
1280 * jbd2 layer's try_to_free_buffers() function to release them.
1282 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1285 journal_t *journal = EXT4_SB(sb)->s_journal;
1287 WARN_ON(PageChecked(page));
1288 if (!page_has_buffers(page))
1291 return jbd2_journal_try_to_free_buffers(journal, page,
1292 wait & ~__GFP_DIRECT_RECLAIM);
1293 return try_to_free_buffers(page);
1296 #ifdef CONFIG_FS_ENCRYPTION
1297 static int ext4_get_context(struct inode *inode, void *ctx, size_t len)
1299 return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1300 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len);
1303 static int ext4_set_context(struct inode *inode, const void *ctx, size_t len,
1306 handle_t *handle = fs_data;
1307 int res, res2, credits, retries = 0;
1310 * Encrypting the root directory is not allowed because e2fsck expects
1311 * lost+found to exist and be unencrypted, and encrypting the root
1312 * directory would imply encrypting the lost+found directory as well as
1313 * the filename "lost+found" itself.
1315 if (inode->i_ino == EXT4_ROOT_INO)
1318 if (WARN_ON_ONCE(IS_DAX(inode) && i_size_read(inode)))
1321 res = ext4_convert_inline_data(inode);
1326 * If a journal handle was specified, then the encryption context is
1327 * being set on a new inode via inheritance and is part of a larger
1328 * transaction to create the inode. Otherwise the encryption context is
1329 * being set on an existing inode in its own transaction. Only in the
1330 * latter case should the "retry on ENOSPC" logic be used.
1334 res = ext4_xattr_set_handle(handle, inode,
1335 EXT4_XATTR_INDEX_ENCRYPTION,
1336 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1339 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1340 ext4_clear_inode_state(inode,
1341 EXT4_STATE_MAY_INLINE_DATA);
1343 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1344 * S_DAX may be disabled
1346 ext4_set_inode_flags(inode);
1351 res = dquot_initialize(inode);
1355 res = ext4_xattr_set_credits(inode, len, false /* is_create */,
1360 handle = ext4_journal_start(inode, EXT4_HT_MISC, credits);
1362 return PTR_ERR(handle);
1364 res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION,
1365 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1368 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1370 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1371 * S_DAX may be disabled
1373 ext4_set_inode_flags(inode);
1374 res = ext4_mark_inode_dirty(handle, inode);
1376 EXT4_ERROR_INODE(inode, "Failed to mark inode dirty");
1378 res2 = ext4_journal_stop(handle);
1380 if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
1387 static bool ext4_dummy_context(struct inode *inode)
1389 return DUMMY_ENCRYPTION_ENABLED(EXT4_SB(inode->i_sb));
1392 static bool ext4_has_stable_inodes(struct super_block *sb)
1394 return ext4_has_feature_stable_inodes(sb);
1397 static void ext4_get_ino_and_lblk_bits(struct super_block *sb,
1398 int *ino_bits_ret, int *lblk_bits_ret)
1400 *ino_bits_ret = 8 * sizeof(EXT4_SB(sb)->s_es->s_inodes_count);
1401 *lblk_bits_ret = 8 * sizeof(ext4_lblk_t);
1404 static const struct fscrypt_operations ext4_cryptops = {
1405 .key_prefix = "ext4:",
1406 .get_context = ext4_get_context,
1407 .set_context = ext4_set_context,
1408 .dummy_context = ext4_dummy_context,
1409 .empty_dir = ext4_empty_dir,
1410 .max_namelen = EXT4_NAME_LEN,
1411 .has_stable_inodes = ext4_has_stable_inodes,
1412 .get_ino_and_lblk_bits = ext4_get_ino_and_lblk_bits,
1417 static const char * const quotatypes[] = INITQFNAMES;
1418 #define QTYPE2NAME(t) (quotatypes[t])
1420 static int ext4_write_dquot(struct dquot *dquot);
1421 static int ext4_acquire_dquot(struct dquot *dquot);
1422 static int ext4_release_dquot(struct dquot *dquot);
1423 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1424 static int ext4_write_info(struct super_block *sb, int type);
1425 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1426 const struct path *path);
1427 static int ext4_quota_on_mount(struct super_block *sb, int type);
1428 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1429 size_t len, loff_t off);
1430 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1431 const char *data, size_t len, loff_t off);
1432 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1433 unsigned int flags);
1434 static int ext4_enable_quotas(struct super_block *sb);
1436 static struct dquot **ext4_get_dquots(struct inode *inode)
1438 return EXT4_I(inode)->i_dquot;
1441 static const struct dquot_operations ext4_quota_operations = {
1442 .get_reserved_space = ext4_get_reserved_space,
1443 .write_dquot = ext4_write_dquot,
1444 .acquire_dquot = ext4_acquire_dquot,
1445 .release_dquot = ext4_release_dquot,
1446 .mark_dirty = ext4_mark_dquot_dirty,
1447 .write_info = ext4_write_info,
1448 .alloc_dquot = dquot_alloc,
1449 .destroy_dquot = dquot_destroy,
1450 .get_projid = ext4_get_projid,
1451 .get_inode_usage = ext4_get_inode_usage,
1452 .get_next_id = dquot_get_next_id,
1455 static const struct quotactl_ops ext4_qctl_operations = {
1456 .quota_on = ext4_quota_on,
1457 .quota_off = ext4_quota_off,
1458 .quota_sync = dquot_quota_sync,
1459 .get_state = dquot_get_state,
1460 .set_info = dquot_set_dqinfo,
1461 .get_dqblk = dquot_get_dqblk,
1462 .set_dqblk = dquot_set_dqblk,
1463 .get_nextdqblk = dquot_get_next_dqblk,
1467 static const struct super_operations ext4_sops = {
1468 .alloc_inode = ext4_alloc_inode,
1469 .free_inode = ext4_free_in_core_inode,
1470 .destroy_inode = ext4_destroy_inode,
1471 .write_inode = ext4_write_inode,
1472 .dirty_inode = ext4_dirty_inode,
1473 .drop_inode = ext4_drop_inode,
1474 .evict_inode = ext4_evict_inode,
1475 .put_super = ext4_put_super,
1476 .sync_fs = ext4_sync_fs,
1477 .freeze_fs = ext4_freeze,
1478 .unfreeze_fs = ext4_unfreeze,
1479 .statfs = ext4_statfs,
1480 .remount_fs = ext4_remount,
1481 .show_options = ext4_show_options,
1483 .quota_read = ext4_quota_read,
1484 .quota_write = ext4_quota_write,
1485 .get_dquots = ext4_get_dquots,
1487 .bdev_try_to_free_page = bdev_try_to_free_page,
1490 static const struct export_operations ext4_export_ops = {
1491 .fh_to_dentry = ext4_fh_to_dentry,
1492 .fh_to_parent = ext4_fh_to_parent,
1493 .get_parent = ext4_get_parent,
1494 .commit_metadata = ext4_nfs_commit_metadata,
1498 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1499 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1500 Opt_nouid32, Opt_debug, Opt_removed,
1501 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1502 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1503 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1504 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1505 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1506 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1507 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1508 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1509 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1510 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version, Opt_dax,
1511 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_warn_on_error,
1512 Opt_nowarn_on_error, Opt_mblk_io_submit,
1513 Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize,
1514 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1515 Opt_inode_readahead_blks, Opt_journal_ioprio,
1516 Opt_dioread_nolock, Opt_dioread_lock,
1517 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1518 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache,
1521 static const match_table_t tokens = {
1522 {Opt_bsd_df, "bsddf"},
1523 {Opt_minix_df, "minixdf"},
1524 {Opt_grpid, "grpid"},
1525 {Opt_grpid, "bsdgroups"},
1526 {Opt_nogrpid, "nogrpid"},
1527 {Opt_nogrpid, "sysvgroups"},
1528 {Opt_resgid, "resgid=%u"},
1529 {Opt_resuid, "resuid=%u"},
1531 {Opt_err_cont, "errors=continue"},
1532 {Opt_err_panic, "errors=panic"},
1533 {Opt_err_ro, "errors=remount-ro"},
1534 {Opt_nouid32, "nouid32"},
1535 {Opt_debug, "debug"},
1536 {Opt_removed, "oldalloc"},
1537 {Opt_removed, "orlov"},
1538 {Opt_user_xattr, "user_xattr"},
1539 {Opt_nouser_xattr, "nouser_xattr"},
1541 {Opt_noacl, "noacl"},
1542 {Opt_noload, "norecovery"},
1543 {Opt_noload, "noload"},
1544 {Opt_removed, "nobh"},
1545 {Opt_removed, "bh"},
1546 {Opt_commit, "commit=%u"},
1547 {Opt_min_batch_time, "min_batch_time=%u"},
1548 {Opt_max_batch_time, "max_batch_time=%u"},
1549 {Opt_journal_dev, "journal_dev=%u"},
1550 {Opt_journal_path, "journal_path=%s"},
1551 {Opt_journal_checksum, "journal_checksum"},
1552 {Opt_nojournal_checksum, "nojournal_checksum"},
1553 {Opt_journal_async_commit, "journal_async_commit"},
1554 {Opt_abort, "abort"},
1555 {Opt_data_journal, "data=journal"},
1556 {Opt_data_ordered, "data=ordered"},
1557 {Opt_data_writeback, "data=writeback"},
1558 {Opt_data_err_abort, "data_err=abort"},
1559 {Opt_data_err_ignore, "data_err=ignore"},
1560 {Opt_offusrjquota, "usrjquota="},
1561 {Opt_usrjquota, "usrjquota=%s"},
1562 {Opt_offgrpjquota, "grpjquota="},
1563 {Opt_grpjquota, "grpjquota=%s"},
1564 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1565 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1566 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1567 {Opt_grpquota, "grpquota"},
1568 {Opt_noquota, "noquota"},
1569 {Opt_quota, "quota"},
1570 {Opt_usrquota, "usrquota"},
1571 {Opt_prjquota, "prjquota"},
1572 {Opt_barrier, "barrier=%u"},
1573 {Opt_barrier, "barrier"},
1574 {Opt_nobarrier, "nobarrier"},
1575 {Opt_i_version, "i_version"},
1577 {Opt_stripe, "stripe=%u"},
1578 {Opt_delalloc, "delalloc"},
1579 {Opt_warn_on_error, "warn_on_error"},
1580 {Opt_nowarn_on_error, "nowarn_on_error"},
1581 {Opt_lazytime, "lazytime"},
1582 {Opt_nolazytime, "nolazytime"},
1583 {Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"},
1584 {Opt_nodelalloc, "nodelalloc"},
1585 {Opt_removed, "mblk_io_submit"},
1586 {Opt_removed, "nomblk_io_submit"},
1587 {Opt_block_validity, "block_validity"},
1588 {Opt_noblock_validity, "noblock_validity"},
1589 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1590 {Opt_journal_ioprio, "journal_ioprio=%u"},
1591 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1592 {Opt_auto_da_alloc, "auto_da_alloc"},
1593 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1594 {Opt_dioread_nolock, "dioread_nolock"},
1595 {Opt_dioread_lock, "nodioread_nolock"},
1596 {Opt_dioread_lock, "dioread_lock"},
1597 {Opt_discard, "discard"},
1598 {Opt_nodiscard, "nodiscard"},
1599 {Opt_init_itable, "init_itable=%u"},
1600 {Opt_init_itable, "init_itable"},
1601 {Opt_noinit_itable, "noinit_itable"},
1602 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1603 {Opt_test_dummy_encryption, "test_dummy_encryption"},
1604 {Opt_nombcache, "nombcache"},
1605 {Opt_nombcache, "no_mbcache"}, /* for backward compatibility */
1606 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1607 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1608 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1609 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1610 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1614 static ext4_fsblk_t get_sb_block(void **data)
1616 ext4_fsblk_t sb_block;
1617 char *options = (char *) *data;
1619 if (!options || strncmp(options, "sb=", 3) != 0)
1620 return 1; /* Default location */
1623 /* TODO: use simple_strtoll with >32bit ext4 */
1624 sb_block = simple_strtoul(options, &options, 0);
1625 if (*options && *options != ',') {
1626 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1630 if (*options == ',')
1632 *data = (void *) options;
1637 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1638 static const char deprecated_msg[] =
1639 "Mount option \"%s\" will be removed by %s\n"
1640 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1643 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1645 struct ext4_sb_info *sbi = EXT4_SB(sb);
1646 char *qname, *old_qname = get_qf_name(sb, sbi, qtype);
1649 if (sb_any_quota_loaded(sb) && !old_qname) {
1650 ext4_msg(sb, KERN_ERR,
1651 "Cannot change journaled "
1652 "quota options when quota turned on");
1655 if (ext4_has_feature_quota(sb)) {
1656 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1657 "ignored when QUOTA feature is enabled");
1660 qname = match_strdup(args);
1662 ext4_msg(sb, KERN_ERR,
1663 "Not enough memory for storing quotafile name");
1667 if (strcmp(old_qname, qname) == 0)
1670 ext4_msg(sb, KERN_ERR,
1671 "%s quota file already specified",
1675 if (strchr(qname, '/')) {
1676 ext4_msg(sb, KERN_ERR,
1677 "quotafile must be on filesystem root");
1680 rcu_assign_pointer(sbi->s_qf_names[qtype], qname);
1688 static int clear_qf_name(struct super_block *sb, int qtype)
1691 struct ext4_sb_info *sbi = EXT4_SB(sb);
1692 char *old_qname = get_qf_name(sb, sbi, qtype);
1694 if (sb_any_quota_loaded(sb) && old_qname) {
1695 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1696 " when quota turned on");
1699 rcu_assign_pointer(sbi->s_qf_names[qtype], NULL);
1706 #define MOPT_SET 0x0001
1707 #define MOPT_CLEAR 0x0002
1708 #define MOPT_NOSUPPORT 0x0004
1709 #define MOPT_EXPLICIT 0x0008
1710 #define MOPT_CLEAR_ERR 0x0010
1711 #define MOPT_GTE0 0x0020
1714 #define MOPT_QFMT 0x0040
1716 #define MOPT_Q MOPT_NOSUPPORT
1717 #define MOPT_QFMT MOPT_NOSUPPORT
1719 #define MOPT_DATAJ 0x0080
1720 #define MOPT_NO_EXT2 0x0100
1721 #define MOPT_NO_EXT3 0x0200
1722 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1723 #define MOPT_STRING 0x0400
1725 static const struct mount_opts {
1729 } ext4_mount_opts[] = {
1730 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1731 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1732 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1733 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1734 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1735 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1736 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1737 MOPT_EXT4_ONLY | MOPT_SET},
1738 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1739 MOPT_EXT4_ONLY | MOPT_CLEAR},
1740 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1741 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1742 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1743 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1744 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1745 MOPT_EXT4_ONLY | MOPT_CLEAR},
1746 {Opt_warn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_SET},
1747 {Opt_nowarn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_CLEAR},
1748 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1749 MOPT_EXT4_ONLY | MOPT_CLEAR},
1750 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1751 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1752 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1753 EXT4_MOUNT_JOURNAL_CHECKSUM),
1754 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1755 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1756 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1757 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1758 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1759 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1761 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1763 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1764 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1765 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1766 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1767 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1768 {Opt_commit, 0, MOPT_GTE0},
1769 {Opt_max_batch_time, 0, MOPT_GTE0},
1770 {Opt_min_batch_time, 0, MOPT_GTE0},
1771 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1772 {Opt_init_itable, 0, MOPT_GTE0},
1773 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
1774 {Opt_stripe, 0, MOPT_GTE0},
1775 {Opt_resuid, 0, MOPT_GTE0},
1776 {Opt_resgid, 0, MOPT_GTE0},
1777 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1778 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1779 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1780 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1781 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1782 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1783 MOPT_NO_EXT2 | MOPT_DATAJ},
1784 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1785 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1786 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1787 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1788 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1790 {Opt_acl, 0, MOPT_NOSUPPORT},
1791 {Opt_noacl, 0, MOPT_NOSUPPORT},
1793 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1794 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1795 {Opt_debug_want_extra_isize, 0, MOPT_GTE0},
1796 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1797 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1799 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1801 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1803 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1804 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1805 MOPT_CLEAR | MOPT_Q},
1806 {Opt_usrjquota, 0, MOPT_Q},
1807 {Opt_grpjquota, 0, MOPT_Q},
1808 {Opt_offusrjquota, 0, MOPT_Q},
1809 {Opt_offgrpjquota, 0, MOPT_Q},
1810 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1811 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1812 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1813 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1814 {Opt_test_dummy_encryption, 0, MOPT_GTE0},
1815 {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET},
1819 #ifdef CONFIG_UNICODE
1820 static const struct ext4_sb_encodings {
1824 } ext4_sb_encoding_map[] = {
1825 {EXT4_ENC_UTF8_12_1, "utf8", "12.1.0"},
1828 static int ext4_sb_read_encoding(const struct ext4_super_block *es,
1829 const struct ext4_sb_encodings **encoding,
1832 __u16 magic = le16_to_cpu(es->s_encoding);
1835 for (i = 0; i < ARRAY_SIZE(ext4_sb_encoding_map); i++)
1836 if (magic == ext4_sb_encoding_map[i].magic)
1839 if (i >= ARRAY_SIZE(ext4_sb_encoding_map))
1842 *encoding = &ext4_sb_encoding_map[i];
1843 *flags = le16_to_cpu(es->s_encoding_flags);
1849 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1850 substring_t *args, unsigned long *journal_devnum,
1851 unsigned int *journal_ioprio, int is_remount)
1853 struct ext4_sb_info *sbi = EXT4_SB(sb);
1854 const struct mount_opts *m;
1860 if (token == Opt_usrjquota)
1861 return set_qf_name(sb, USRQUOTA, &args[0]);
1862 else if (token == Opt_grpjquota)
1863 return set_qf_name(sb, GRPQUOTA, &args[0]);
1864 else if (token == Opt_offusrjquota)
1865 return clear_qf_name(sb, USRQUOTA);
1866 else if (token == Opt_offgrpjquota)
1867 return clear_qf_name(sb, GRPQUOTA);
1871 case Opt_nouser_xattr:
1872 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1875 return 1; /* handled by get_sb_block() */
1877 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1880 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1883 sb->s_flags |= SB_I_VERSION;
1886 sb->s_flags |= SB_LAZYTIME;
1888 case Opt_nolazytime:
1889 sb->s_flags &= ~SB_LAZYTIME;
1893 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1894 if (token == m->token)
1897 if (m->token == Opt_err) {
1898 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1899 "or missing value", opt);
1903 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1904 ext4_msg(sb, KERN_ERR,
1905 "Mount option \"%s\" incompatible with ext2", opt);
1908 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1909 ext4_msg(sb, KERN_ERR,
1910 "Mount option \"%s\" incompatible with ext3", opt);
1914 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
1916 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1918 if (m->flags & MOPT_EXPLICIT) {
1919 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
1920 set_opt2(sb, EXPLICIT_DELALLOC);
1921 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
1922 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
1926 if (m->flags & MOPT_CLEAR_ERR)
1927 clear_opt(sb, ERRORS_MASK);
1928 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1929 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1930 "options when quota turned on");
1934 if (m->flags & MOPT_NOSUPPORT) {
1935 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1936 } else if (token == Opt_commit) {
1938 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1939 else if (arg > INT_MAX / HZ) {
1940 ext4_msg(sb, KERN_ERR,
1941 "Invalid commit interval %d, "
1942 "must be smaller than %d",
1946 sbi->s_commit_interval = HZ * arg;
1947 } else if (token == Opt_debug_want_extra_isize) {
1950 (arg > (sbi->s_inode_size - EXT4_GOOD_OLD_INODE_SIZE))) {
1951 ext4_msg(sb, KERN_ERR,
1952 "Invalid want_extra_isize %d", arg);
1955 sbi->s_want_extra_isize = arg;
1956 } else if (token == Opt_max_batch_time) {
1957 sbi->s_max_batch_time = arg;
1958 } else if (token == Opt_min_batch_time) {
1959 sbi->s_min_batch_time = arg;
1960 } else if (token == Opt_inode_readahead_blks) {
1961 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1962 ext4_msg(sb, KERN_ERR,
1963 "EXT4-fs: inode_readahead_blks must be "
1964 "0 or a power of 2 smaller than 2^31");
1967 sbi->s_inode_readahead_blks = arg;
1968 } else if (token == Opt_init_itable) {
1969 set_opt(sb, INIT_INODE_TABLE);
1971 arg = EXT4_DEF_LI_WAIT_MULT;
1972 sbi->s_li_wait_mult = arg;
1973 } else if (token == Opt_max_dir_size_kb) {
1974 sbi->s_max_dir_size_kb = arg;
1975 } else if (token == Opt_stripe) {
1976 sbi->s_stripe = arg;
1977 } else if (token == Opt_resuid) {
1978 uid = make_kuid(current_user_ns(), arg);
1979 if (!uid_valid(uid)) {
1980 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1983 sbi->s_resuid = uid;
1984 } else if (token == Opt_resgid) {
1985 gid = make_kgid(current_user_ns(), arg);
1986 if (!gid_valid(gid)) {
1987 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1990 sbi->s_resgid = gid;
1991 } else if (token == Opt_journal_dev) {
1993 ext4_msg(sb, KERN_ERR,
1994 "Cannot specify journal on remount");
1997 *journal_devnum = arg;
1998 } else if (token == Opt_journal_path) {
2000 struct inode *journal_inode;
2005 ext4_msg(sb, KERN_ERR,
2006 "Cannot specify journal on remount");
2009 journal_path = match_strdup(&args[0]);
2010 if (!journal_path) {
2011 ext4_msg(sb, KERN_ERR, "error: could not dup "
2012 "journal device string");
2016 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
2018 ext4_msg(sb, KERN_ERR, "error: could not find "
2019 "journal device path: error %d", error);
2020 kfree(journal_path);
2024 journal_inode = d_inode(path.dentry);
2025 if (!S_ISBLK(journal_inode->i_mode)) {
2026 ext4_msg(sb, KERN_ERR, "error: journal path %s "
2027 "is not a block device", journal_path);
2029 kfree(journal_path);
2033 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
2035 kfree(journal_path);
2036 } else if (token == Opt_journal_ioprio) {
2038 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
2043 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
2044 } else if (token == Opt_test_dummy_encryption) {
2045 #ifdef CONFIG_FS_ENCRYPTION
2046 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
2047 ext4_msg(sb, KERN_WARNING,
2048 "Test dummy encryption mode enabled");
2050 ext4_msg(sb, KERN_WARNING,
2051 "Test dummy encryption mount option ignored");
2053 } else if (m->flags & MOPT_DATAJ) {
2055 if (!sbi->s_journal)
2056 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
2057 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
2058 ext4_msg(sb, KERN_ERR,
2059 "Cannot change data mode on remount");
2063 clear_opt(sb, DATA_FLAGS);
2064 sbi->s_mount_opt |= m->mount_opt;
2067 } else if (m->flags & MOPT_QFMT) {
2068 if (sb_any_quota_loaded(sb) &&
2069 sbi->s_jquota_fmt != m->mount_opt) {
2070 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
2071 "quota options when quota turned on");
2074 if (ext4_has_feature_quota(sb)) {
2075 ext4_msg(sb, KERN_INFO,
2076 "Quota format mount options ignored "
2077 "when QUOTA feature is enabled");
2080 sbi->s_jquota_fmt = m->mount_opt;
2082 } else if (token == Opt_dax) {
2083 #ifdef CONFIG_FS_DAX
2084 ext4_msg(sb, KERN_WARNING,
2085 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
2086 sbi->s_mount_opt |= m->mount_opt;
2088 ext4_msg(sb, KERN_INFO, "dax option not supported");
2091 } else if (token == Opt_data_err_abort) {
2092 sbi->s_mount_opt |= m->mount_opt;
2093 } else if (token == Opt_data_err_ignore) {
2094 sbi->s_mount_opt &= ~m->mount_opt;
2098 if (m->flags & MOPT_CLEAR)
2100 else if (unlikely(!(m->flags & MOPT_SET))) {
2101 ext4_msg(sb, KERN_WARNING,
2102 "buggy handling of option %s", opt);
2107 sbi->s_mount_opt |= m->mount_opt;
2109 sbi->s_mount_opt &= ~m->mount_opt;
2114 static int parse_options(char *options, struct super_block *sb,
2115 unsigned long *journal_devnum,
2116 unsigned int *journal_ioprio,
2119 struct ext4_sb_info __maybe_unused *sbi = EXT4_SB(sb);
2120 char *p, __maybe_unused *usr_qf_name, __maybe_unused *grp_qf_name;
2121 substring_t args[MAX_OPT_ARGS];
2127 while ((p = strsep(&options, ",")) != NULL) {
2131 * Initialize args struct so we know whether arg was
2132 * found; some options take optional arguments.
2134 args[0].to = args[0].from = NULL;
2135 token = match_token(p, tokens, args);
2136 if (handle_mount_opt(sb, p, token, args, journal_devnum,
2137 journal_ioprio, is_remount) < 0)
2142 * We do the test below only for project quotas. 'usrquota' and
2143 * 'grpquota' mount options are allowed even without quota feature
2144 * to support legacy quotas in quota files.
2146 if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) {
2147 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. "
2148 "Cannot enable project quota enforcement.");
2151 usr_qf_name = get_qf_name(sb, sbi, USRQUOTA);
2152 grp_qf_name = get_qf_name(sb, sbi, GRPQUOTA);
2153 if (usr_qf_name || grp_qf_name) {
2154 if (test_opt(sb, USRQUOTA) && usr_qf_name)
2155 clear_opt(sb, USRQUOTA);
2157 if (test_opt(sb, GRPQUOTA) && grp_qf_name)
2158 clear_opt(sb, GRPQUOTA);
2160 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
2161 ext4_msg(sb, KERN_ERR, "old and new quota "
2166 if (!sbi->s_jquota_fmt) {
2167 ext4_msg(sb, KERN_ERR, "journaled quota format "
2176 static inline void ext4_show_quota_options(struct seq_file *seq,
2177 struct super_block *sb)
2179 #if defined(CONFIG_QUOTA)
2180 struct ext4_sb_info *sbi = EXT4_SB(sb);
2181 char *usr_qf_name, *grp_qf_name;
2183 if (sbi->s_jquota_fmt) {
2186 switch (sbi->s_jquota_fmt) {
2197 seq_printf(seq, ",jqfmt=%s", fmtname);
2201 usr_qf_name = rcu_dereference(sbi->s_qf_names[USRQUOTA]);
2202 grp_qf_name = rcu_dereference(sbi->s_qf_names[GRPQUOTA]);
2204 seq_show_option(seq, "usrjquota", usr_qf_name);
2206 seq_show_option(seq, "grpjquota", grp_qf_name);
2211 static const char *token2str(int token)
2213 const struct match_token *t;
2215 for (t = tokens; t->token != Opt_err; t++)
2216 if (t->token == token && !strchr(t->pattern, '='))
2223 * - it's set to a non-default value OR
2224 * - if the per-sb default is different from the global default
2226 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
2229 struct ext4_sb_info *sbi = EXT4_SB(sb);
2230 struct ext4_super_block *es = sbi->s_es;
2231 int def_errors, def_mount_opt = sbi->s_def_mount_opt;
2232 const struct mount_opts *m;
2233 char sep = nodefs ? '\n' : ',';
2235 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2236 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2238 if (sbi->s_sb_block != 1)
2239 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
2241 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
2242 int want_set = m->flags & MOPT_SET;
2243 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
2244 (m->flags & MOPT_CLEAR_ERR))
2246 if (!nodefs && !(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
2247 continue; /* skip if same as the default */
2249 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
2250 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
2251 continue; /* select Opt_noFoo vs Opt_Foo */
2252 SEQ_OPTS_PRINT("%s", token2str(m->token));
2255 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
2256 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
2257 SEQ_OPTS_PRINT("resuid=%u",
2258 from_kuid_munged(&init_user_ns, sbi->s_resuid));
2259 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
2260 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
2261 SEQ_OPTS_PRINT("resgid=%u",
2262 from_kgid_munged(&init_user_ns, sbi->s_resgid));
2263 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
2264 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
2265 SEQ_OPTS_PUTS("errors=remount-ro");
2266 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
2267 SEQ_OPTS_PUTS("errors=continue");
2268 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
2269 SEQ_OPTS_PUTS("errors=panic");
2270 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
2271 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
2272 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
2273 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
2274 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
2275 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2276 if (sb->s_flags & SB_I_VERSION)
2277 SEQ_OPTS_PUTS("i_version");
2278 if (nodefs || sbi->s_stripe)
2279 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
2280 if (nodefs || EXT4_MOUNT_DATA_FLAGS &
2281 (sbi->s_mount_opt ^ def_mount_opt)) {
2282 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2283 SEQ_OPTS_PUTS("data=journal");
2284 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2285 SEQ_OPTS_PUTS("data=ordered");
2286 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2287 SEQ_OPTS_PUTS("data=writeback");
2290 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
2291 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2292 sbi->s_inode_readahead_blks);
2294 if (test_opt(sb, INIT_INODE_TABLE) && (nodefs ||
2295 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
2296 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
2297 if (nodefs || sbi->s_max_dir_size_kb)
2298 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2299 if (test_opt(sb, DATA_ERR_ABORT))
2300 SEQ_OPTS_PUTS("data_err=abort");
2301 if (DUMMY_ENCRYPTION_ENABLED(sbi))
2302 SEQ_OPTS_PUTS("test_dummy_encryption");
2304 ext4_show_quota_options(seq, sb);
2308 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2310 return _ext4_show_options(seq, root->d_sb, 0);
2313 int ext4_seq_options_show(struct seq_file *seq, void *offset)
2315 struct super_block *sb = seq->private;
2318 seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw");
2319 rc = _ext4_show_options(seq, sb, 1);
2320 seq_puts(seq, "\n");
2324 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
2327 struct ext4_sb_info *sbi = EXT4_SB(sb);
2330 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
2331 ext4_msg(sb, KERN_ERR, "revision level too high, "
2332 "forcing read-only mode");
2337 if (!(sbi->s_mount_state & EXT4_VALID_FS))
2338 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
2339 "running e2fsck is recommended");
2340 else if (sbi->s_mount_state & EXT4_ERROR_FS)
2341 ext4_msg(sb, KERN_WARNING,
2342 "warning: mounting fs with errors, "
2343 "running e2fsck is recommended");
2344 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
2345 le16_to_cpu(es->s_mnt_count) >=
2346 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
2347 ext4_msg(sb, KERN_WARNING,
2348 "warning: maximal mount count reached, "
2349 "running e2fsck is recommended");
2350 else if (le32_to_cpu(es->s_checkinterval) &&
2351 (ext4_get_tstamp(es, s_lastcheck) +
2352 le32_to_cpu(es->s_checkinterval) <= ktime_get_real_seconds()))
2353 ext4_msg(sb, KERN_WARNING,
2354 "warning: checktime reached, "
2355 "running e2fsck is recommended");
2356 if (!sbi->s_journal)
2357 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
2358 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
2359 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
2360 le16_add_cpu(&es->s_mnt_count, 1);
2361 ext4_update_tstamp(es, s_mtime);
2363 ext4_set_feature_journal_needs_recovery(sb);
2365 err = ext4_commit_super(sb, 1);
2367 if (test_opt(sb, DEBUG))
2368 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
2369 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
2371 sbi->s_groups_count,
2372 EXT4_BLOCKS_PER_GROUP(sb),
2373 EXT4_INODES_PER_GROUP(sb),
2374 sbi->s_mount_opt, sbi->s_mount_opt2);
2376 cleancache_init_fs(sb);
2380 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2382 struct ext4_sb_info *sbi = EXT4_SB(sb);
2383 struct flex_groups *new_groups;
2386 if (!sbi->s_log_groups_per_flex)
2389 size = ext4_flex_group(sbi, ngroup - 1) + 1;
2390 if (size <= sbi->s_flex_groups_allocated)
2393 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
2394 new_groups = kvzalloc(size, GFP_KERNEL);
2396 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
2397 size / (int) sizeof(struct flex_groups));
2401 if (sbi->s_flex_groups) {
2402 memcpy(new_groups, sbi->s_flex_groups,
2403 (sbi->s_flex_groups_allocated *
2404 sizeof(struct flex_groups)));
2405 kvfree(sbi->s_flex_groups);
2407 sbi->s_flex_groups = new_groups;
2408 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
2412 static int ext4_fill_flex_info(struct super_block *sb)
2414 struct ext4_sb_info *sbi = EXT4_SB(sb);
2415 struct ext4_group_desc *gdp = NULL;
2416 ext4_group_t flex_group;
2419 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2420 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
2421 sbi->s_log_groups_per_flex = 0;
2425 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2429 for (i = 0; i < sbi->s_groups_count; i++) {
2430 gdp = ext4_get_group_desc(sb, i, NULL);
2432 flex_group = ext4_flex_group(sbi, i);
2433 atomic_add(ext4_free_inodes_count(sb, gdp),
2434 &sbi->s_flex_groups[flex_group].free_inodes);
2435 atomic64_add(ext4_free_group_clusters(sb, gdp),
2436 &sbi->s_flex_groups[flex_group].free_clusters);
2437 atomic_add(ext4_used_dirs_count(sb, gdp),
2438 &sbi->s_flex_groups[flex_group].used_dirs);
2446 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
2447 struct ext4_group_desc *gdp)
2449 int offset = offsetof(struct ext4_group_desc, bg_checksum);
2451 __le32 le_group = cpu_to_le32(block_group);
2452 struct ext4_sb_info *sbi = EXT4_SB(sb);
2454 if (ext4_has_metadata_csum(sbi->s_sb)) {
2455 /* Use new metadata_csum algorithm */
2457 __u16 dummy_csum = 0;
2459 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2461 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2462 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2463 sizeof(dummy_csum));
2464 offset += sizeof(dummy_csum);
2465 if (offset < sbi->s_desc_size)
2466 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2467 sbi->s_desc_size - offset);
2469 crc = csum32 & 0xFFFF;
2473 /* old crc16 code */
2474 if (!ext4_has_feature_gdt_csum(sb))
2477 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2478 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2479 crc = crc16(crc, (__u8 *)gdp, offset);
2480 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2481 /* for checksum of struct ext4_group_desc do the rest...*/
2482 if (ext4_has_feature_64bit(sb) &&
2483 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2484 crc = crc16(crc, (__u8 *)gdp + offset,
2485 le16_to_cpu(sbi->s_es->s_desc_size) -
2489 return cpu_to_le16(crc);
2492 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2493 struct ext4_group_desc *gdp)
2495 if (ext4_has_group_desc_csum(sb) &&
2496 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
2502 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2503 struct ext4_group_desc *gdp)
2505 if (!ext4_has_group_desc_csum(sb))
2507 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
2510 /* Called at mount-time, super-block is locked */
2511 static int ext4_check_descriptors(struct super_block *sb,
2512 ext4_fsblk_t sb_block,
2513 ext4_group_t *first_not_zeroed)
2515 struct ext4_sb_info *sbi = EXT4_SB(sb);
2516 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2517 ext4_fsblk_t last_block;
2518 ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0);
2519 ext4_fsblk_t block_bitmap;
2520 ext4_fsblk_t inode_bitmap;
2521 ext4_fsblk_t inode_table;
2522 int flexbg_flag = 0;
2523 ext4_group_t i, grp = sbi->s_groups_count;
2525 if (ext4_has_feature_flex_bg(sb))
2528 ext4_debug("Checking group descriptors");
2530 for (i = 0; i < sbi->s_groups_count; i++) {
2531 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2533 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2534 last_block = ext4_blocks_count(sbi->s_es) - 1;
2536 last_block = first_block +
2537 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2539 if ((grp == sbi->s_groups_count) &&
2540 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2543 block_bitmap = ext4_block_bitmap(sb, gdp);
2544 if (block_bitmap == sb_block) {
2545 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2546 "Block bitmap for group %u overlaps "
2551 if (block_bitmap >= sb_block + 1 &&
2552 block_bitmap <= last_bg_block) {
2553 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2554 "Block bitmap for group %u overlaps "
2555 "block group descriptors", i);
2559 if (block_bitmap < first_block || block_bitmap > last_block) {
2560 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2561 "Block bitmap for group %u not in group "
2562 "(block %llu)!", i, block_bitmap);
2565 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2566 if (inode_bitmap == sb_block) {
2567 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2568 "Inode bitmap for group %u overlaps "
2573 if (inode_bitmap >= sb_block + 1 &&
2574 inode_bitmap <= last_bg_block) {
2575 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2576 "Inode bitmap for group %u overlaps "
2577 "block group descriptors", i);
2581 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2582 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2583 "Inode bitmap for group %u not in group "
2584 "(block %llu)!", i, inode_bitmap);
2587 inode_table = ext4_inode_table(sb, gdp);
2588 if (inode_table == sb_block) {
2589 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2590 "Inode table for group %u overlaps "
2595 if (inode_table >= sb_block + 1 &&
2596 inode_table <= last_bg_block) {
2597 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2598 "Inode table for group %u overlaps "
2599 "block group descriptors", i);
2603 if (inode_table < first_block ||
2604 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2605 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2606 "Inode table for group %u not in group "
2607 "(block %llu)!", i, inode_table);
2610 ext4_lock_group(sb, i);
2611 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2612 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2613 "Checksum for group %u failed (%u!=%u)",
2614 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
2615 gdp)), le16_to_cpu(gdp->bg_checksum));
2616 if (!sb_rdonly(sb)) {
2617 ext4_unlock_group(sb, i);
2621 ext4_unlock_group(sb, i);
2623 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2625 if (NULL != first_not_zeroed)
2626 *first_not_zeroed = grp;
2630 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2631 * the superblock) which were deleted from all directories, but held open by
2632 * a process at the time of a crash. We walk the list and try to delete these
2633 * inodes at recovery time (only with a read-write filesystem).
2635 * In order to keep the orphan inode chain consistent during traversal (in
2636 * case of crash during recovery), we link each inode into the superblock
2637 * orphan list_head and handle it the same way as an inode deletion during
2638 * normal operation (which journals the operations for us).
2640 * We only do an iget() and an iput() on each inode, which is very safe if we
2641 * accidentally point at an in-use or already deleted inode. The worst that
2642 * can happen in this case is that we get a "bit already cleared" message from
2643 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2644 * e2fsck was run on this filesystem, and it must have already done the orphan
2645 * inode cleanup for us, so we can safely abort without any further action.
2647 static void ext4_orphan_cleanup(struct super_block *sb,
2648 struct ext4_super_block *es)
2650 unsigned int s_flags = sb->s_flags;
2651 int ret, nr_orphans = 0, nr_truncates = 0;
2653 int quota_update = 0;
2656 if (!es->s_last_orphan) {
2657 jbd_debug(4, "no orphan inodes to clean up\n");
2661 if (bdev_read_only(sb->s_bdev)) {
2662 ext4_msg(sb, KERN_ERR, "write access "
2663 "unavailable, skipping orphan cleanup");
2667 /* Check if feature set would not allow a r/w mount */
2668 if (!ext4_feature_set_ok(sb, 0)) {
2669 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2670 "unknown ROCOMPAT features");
2674 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2675 /* don't clear list on RO mount w/ errors */
2676 if (es->s_last_orphan && !(s_flags & SB_RDONLY)) {
2677 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
2678 "clearing orphan list.\n");
2679 es->s_last_orphan = 0;
2681 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2685 if (s_flags & SB_RDONLY) {
2686 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2687 sb->s_flags &= ~SB_RDONLY;
2690 /* Needed for iput() to work correctly and not trash data */
2691 sb->s_flags |= SB_ACTIVE;
2694 * Turn on quotas which were not enabled for read-only mounts if
2695 * filesystem has quota feature, so that they are updated correctly.
2697 if (ext4_has_feature_quota(sb) && (s_flags & SB_RDONLY)) {
2698 int ret = ext4_enable_quotas(sb);
2703 ext4_msg(sb, KERN_ERR,
2704 "Cannot turn on quotas: error %d", ret);
2707 /* Turn on journaled quotas used for old sytle */
2708 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2709 if (EXT4_SB(sb)->s_qf_names[i]) {
2710 int ret = ext4_quota_on_mount(sb, i);
2715 ext4_msg(sb, KERN_ERR,
2716 "Cannot turn on journaled "
2717 "quota: type %d: error %d", i, ret);
2722 while (es->s_last_orphan) {
2723 struct inode *inode;
2726 * We may have encountered an error during cleanup; if
2727 * so, skip the rest.
2729 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2730 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2731 es->s_last_orphan = 0;
2735 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2736 if (IS_ERR(inode)) {
2737 es->s_last_orphan = 0;
2741 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2742 dquot_initialize(inode);
2743 if (inode->i_nlink) {
2744 if (test_opt(sb, DEBUG))
2745 ext4_msg(sb, KERN_DEBUG,
2746 "%s: truncating inode %lu to %lld bytes",
2747 __func__, inode->i_ino, inode->i_size);
2748 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2749 inode->i_ino, inode->i_size);
2751 truncate_inode_pages(inode->i_mapping, inode->i_size);
2752 ret = ext4_truncate(inode);
2754 ext4_std_error(inode->i_sb, ret);
2755 inode_unlock(inode);
2758 if (test_opt(sb, DEBUG))
2759 ext4_msg(sb, KERN_DEBUG,
2760 "%s: deleting unreferenced inode %lu",
2761 __func__, inode->i_ino);
2762 jbd_debug(2, "deleting unreferenced inode %lu\n",
2766 iput(inode); /* The delete magic happens here! */
2769 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2772 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2773 PLURAL(nr_orphans));
2775 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2776 PLURAL(nr_truncates));
2778 /* Turn off quotas if they were enabled for orphan cleanup */
2780 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2781 if (sb_dqopt(sb)->files[i])
2782 dquot_quota_off(sb, i);
2786 sb->s_flags = s_flags; /* Restore SB_RDONLY status */
2790 * Maximal extent format file size.
2791 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2792 * extent format containers, within a sector_t, and within i_blocks
2793 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2794 * so that won't be a limiting factor.
2796 * However there is other limiting factor. We do store extents in the form
2797 * of starting block and length, hence the resulting length of the extent
2798 * covering maximum file size must fit into on-disk format containers as
2799 * well. Given that length is always by 1 unit bigger than max unit (because
2800 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2802 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2804 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2807 loff_t upper_limit = MAX_LFS_FILESIZE;
2809 BUILD_BUG_ON(sizeof(blkcnt_t) < sizeof(u64));
2811 if (!has_huge_files) {
2812 upper_limit = (1LL << 32) - 1;
2814 /* total blocks in file system block size */
2815 upper_limit >>= (blkbits - 9);
2816 upper_limit <<= blkbits;
2820 * 32-bit extent-start container, ee_block. We lower the maxbytes
2821 * by one fs block, so ee_len can cover the extent of maximum file
2824 res = (1LL << 32) - 1;
2827 /* Sanity check against vm- & vfs- imposed limits */
2828 if (res > upper_limit)
2835 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2836 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2837 * We need to be 1 filesystem block less than the 2^48 sector limit.
2839 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2841 loff_t res = EXT4_NDIR_BLOCKS;
2844 /* This is calculated to be the largest file size for a dense, block
2845 * mapped file such that the file's total number of 512-byte sectors,
2846 * including data and all indirect blocks, does not exceed (2^48 - 1).
2848 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2849 * number of 512-byte sectors of the file.
2852 if (!has_huge_files) {
2854 * !has_huge_files or implies that the inode i_block field
2855 * represents total file blocks in 2^32 512-byte sectors ==
2856 * size of vfs inode i_blocks * 8
2858 upper_limit = (1LL << 32) - 1;
2860 /* total blocks in file system block size */
2861 upper_limit >>= (bits - 9);
2865 * We use 48 bit ext4_inode i_blocks
2866 * With EXT4_HUGE_FILE_FL set the i_blocks
2867 * represent total number of blocks in
2868 * file system block size
2870 upper_limit = (1LL << 48) - 1;
2874 /* indirect blocks */
2876 /* double indirect blocks */
2877 meta_blocks += 1 + (1LL << (bits-2));
2878 /* tripple indirect blocks */
2879 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2881 upper_limit -= meta_blocks;
2882 upper_limit <<= bits;
2884 res += 1LL << (bits-2);
2885 res += 1LL << (2*(bits-2));
2886 res += 1LL << (3*(bits-2));
2888 if (res > upper_limit)
2891 if (res > MAX_LFS_FILESIZE)
2892 res = MAX_LFS_FILESIZE;
2897 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2898 ext4_fsblk_t logical_sb_block, int nr)
2900 struct ext4_sb_info *sbi = EXT4_SB(sb);
2901 ext4_group_t bg, first_meta_bg;
2904 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2906 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
2907 return logical_sb_block + nr + 1;
2908 bg = sbi->s_desc_per_block * nr;
2909 if (ext4_bg_has_super(sb, bg))
2913 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2914 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2915 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2918 if (sb->s_blocksize == 1024 && nr == 0 &&
2919 le32_to_cpu(sbi->s_es->s_first_data_block) == 0)
2922 return (has_super + ext4_group_first_block_no(sb, bg));
2926 * ext4_get_stripe_size: Get the stripe size.
2927 * @sbi: In memory super block info
2929 * If we have specified it via mount option, then
2930 * use the mount option value. If the value specified at mount time is
2931 * greater than the blocks per group use the super block value.
2932 * If the super block value is greater than blocks per group return 0.
2933 * Allocator needs it be less than blocks per group.
2936 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2938 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2939 unsigned long stripe_width =
2940 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2943 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2944 ret = sbi->s_stripe;
2945 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
2947 else if (stride && stride <= sbi->s_blocks_per_group)
2953 * If the stripe width is 1, this makes no sense and
2954 * we set it to 0 to turn off stripe handling code.
2963 * Check whether this filesystem can be mounted based on
2964 * the features present and the RDONLY/RDWR mount requested.
2965 * Returns 1 if this filesystem can be mounted as requested,
2966 * 0 if it cannot be.
2968 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2970 if (ext4_has_unknown_ext4_incompat_features(sb)) {
2971 ext4_msg(sb, KERN_ERR,
2972 "Couldn't mount because of "
2973 "unsupported optional features (%x)",
2974 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2975 ~EXT4_FEATURE_INCOMPAT_SUPP));
2979 #ifndef CONFIG_UNICODE
2980 if (ext4_has_feature_casefold(sb)) {
2981 ext4_msg(sb, KERN_ERR,
2982 "Filesystem with casefold feature cannot be "
2983 "mounted without CONFIG_UNICODE");
2991 if (ext4_has_feature_readonly(sb)) {
2992 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2993 sb->s_flags |= SB_RDONLY;
2997 /* Check that feature set is OK for a read-write mount */
2998 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
2999 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
3000 "unsupported optional features (%x)",
3001 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
3002 ~EXT4_FEATURE_RO_COMPAT_SUPP));
3005 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
3006 ext4_msg(sb, KERN_ERR,
3007 "Can't support bigalloc feature without "
3008 "extents feature\n");
3012 #if !defined(CONFIG_QUOTA) || !defined(CONFIG_QFMT_V2)
3013 if (!readonly && (ext4_has_feature_quota(sb) ||
3014 ext4_has_feature_project(sb))) {
3015 ext4_msg(sb, KERN_ERR,
3016 "The kernel was not built with CONFIG_QUOTA and CONFIG_QFMT_V2");
3019 #endif /* CONFIG_QUOTA */
3024 * This function is called once a day if we have errors logged
3025 * on the file system
3027 static void print_daily_error_info(struct timer_list *t)
3029 struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report);
3030 struct super_block *sb = sbi->s_sb;
3031 struct ext4_super_block *es = sbi->s_es;
3033 if (es->s_error_count)
3034 /* fsck newer than v1.41.13 is needed to clean this condition. */
3035 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
3036 le32_to_cpu(es->s_error_count));
3037 if (es->s_first_error_time) {
3038 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %llu: %.*s:%d",
3040 ext4_get_tstamp(es, s_first_error_time),
3041 (int) sizeof(es->s_first_error_func),
3042 es->s_first_error_func,
3043 le32_to_cpu(es->s_first_error_line));
3044 if (es->s_first_error_ino)
3045 printk(KERN_CONT ": inode %u",
3046 le32_to_cpu(es->s_first_error_ino));
3047 if (es->s_first_error_block)
3048 printk(KERN_CONT ": block %llu", (unsigned long long)
3049 le64_to_cpu(es->s_first_error_block));
3050 printk(KERN_CONT "\n");
3052 if (es->s_last_error_time) {
3053 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %llu: %.*s:%d",
3055 ext4_get_tstamp(es, s_last_error_time),
3056 (int) sizeof(es->s_last_error_func),
3057 es->s_last_error_func,
3058 le32_to_cpu(es->s_last_error_line));
3059 if (es->s_last_error_ino)
3060 printk(KERN_CONT ": inode %u",
3061 le32_to_cpu(es->s_last_error_ino));
3062 if (es->s_last_error_block)
3063 printk(KERN_CONT ": block %llu", (unsigned long long)
3064 le64_to_cpu(es->s_last_error_block));
3065 printk(KERN_CONT "\n");
3067 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
3070 /* Find next suitable group and run ext4_init_inode_table */
3071 static int ext4_run_li_request(struct ext4_li_request *elr)
3073 struct ext4_group_desc *gdp = NULL;
3074 ext4_group_t group, ngroups;
3075 struct super_block *sb;
3076 unsigned long timeout = 0;
3080 ngroups = EXT4_SB(sb)->s_groups_count;
3082 for (group = elr->lr_next_group; group < ngroups; group++) {
3083 gdp = ext4_get_group_desc(sb, group, NULL);
3089 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3093 if (group >= ngroups)
3098 ret = ext4_init_inode_table(sb, group,
3099 elr->lr_timeout ? 0 : 1);
3100 if (elr->lr_timeout == 0) {
3101 timeout = (jiffies - timeout) *
3102 elr->lr_sbi->s_li_wait_mult;
3103 elr->lr_timeout = timeout;
3105 elr->lr_next_sched = jiffies + elr->lr_timeout;
3106 elr->lr_next_group = group + 1;
3112 * Remove lr_request from the list_request and free the
3113 * request structure. Should be called with li_list_mtx held
3115 static void ext4_remove_li_request(struct ext4_li_request *elr)
3117 struct ext4_sb_info *sbi;
3124 list_del(&elr->lr_request);
3125 sbi->s_li_request = NULL;
3129 static void ext4_unregister_li_request(struct super_block *sb)
3131 mutex_lock(&ext4_li_mtx);
3132 if (!ext4_li_info) {
3133 mutex_unlock(&ext4_li_mtx);
3137 mutex_lock(&ext4_li_info->li_list_mtx);
3138 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
3139 mutex_unlock(&ext4_li_info->li_list_mtx);
3140 mutex_unlock(&ext4_li_mtx);
3143 static struct task_struct *ext4_lazyinit_task;
3146 * This is the function where ext4lazyinit thread lives. It walks
3147 * through the request list searching for next scheduled filesystem.
3148 * When such a fs is found, run the lazy initialization request
3149 * (ext4_rn_li_request) and keep track of the time spend in this
3150 * function. Based on that time we compute next schedule time of
3151 * the request. When walking through the list is complete, compute
3152 * next waking time and put itself into sleep.
3154 static int ext4_lazyinit_thread(void *arg)
3156 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
3157 struct list_head *pos, *n;
3158 struct ext4_li_request *elr;
3159 unsigned long next_wakeup, cur;
3161 BUG_ON(NULL == eli);
3165 next_wakeup = MAX_JIFFY_OFFSET;
3167 mutex_lock(&eli->li_list_mtx);
3168 if (list_empty(&eli->li_request_list)) {
3169 mutex_unlock(&eli->li_list_mtx);
3172 list_for_each_safe(pos, n, &eli->li_request_list) {
3175 elr = list_entry(pos, struct ext4_li_request,
3178 if (time_before(jiffies, elr->lr_next_sched)) {
3179 if (time_before(elr->lr_next_sched, next_wakeup))
3180 next_wakeup = elr->lr_next_sched;
3183 if (down_read_trylock(&elr->lr_super->s_umount)) {
3184 if (sb_start_write_trylock(elr->lr_super)) {
3187 * We hold sb->s_umount, sb can not
3188 * be removed from the list, it is
3189 * now safe to drop li_list_mtx
3191 mutex_unlock(&eli->li_list_mtx);
3192 err = ext4_run_li_request(elr);
3193 sb_end_write(elr->lr_super);
3194 mutex_lock(&eli->li_list_mtx);
3197 up_read((&elr->lr_super->s_umount));
3199 /* error, remove the lazy_init job */
3201 ext4_remove_li_request(elr);
3205 elr->lr_next_sched = jiffies +
3207 % (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3209 if (time_before(elr->lr_next_sched, next_wakeup))
3210 next_wakeup = elr->lr_next_sched;
3212 mutex_unlock(&eli->li_list_mtx);
3217 if ((time_after_eq(cur, next_wakeup)) ||
3218 (MAX_JIFFY_OFFSET == next_wakeup)) {
3223 schedule_timeout_interruptible(next_wakeup - cur);
3225 if (kthread_should_stop()) {
3226 ext4_clear_request_list();
3233 * It looks like the request list is empty, but we need
3234 * to check it under the li_list_mtx lock, to prevent any
3235 * additions into it, and of course we should lock ext4_li_mtx
3236 * to atomically free the list and ext4_li_info, because at
3237 * this point another ext4 filesystem could be registering
3240 mutex_lock(&ext4_li_mtx);
3241 mutex_lock(&eli->li_list_mtx);
3242 if (!list_empty(&eli->li_request_list)) {
3243 mutex_unlock(&eli->li_list_mtx);
3244 mutex_unlock(&ext4_li_mtx);
3247 mutex_unlock(&eli->li_list_mtx);
3248 kfree(ext4_li_info);
3249 ext4_li_info = NULL;
3250 mutex_unlock(&ext4_li_mtx);
3255 static void ext4_clear_request_list(void)
3257 struct list_head *pos, *n;
3258 struct ext4_li_request *elr;
3260 mutex_lock(&ext4_li_info->li_list_mtx);
3261 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3262 elr = list_entry(pos, struct ext4_li_request,
3264 ext4_remove_li_request(elr);
3266 mutex_unlock(&ext4_li_info->li_list_mtx);
3269 static int ext4_run_lazyinit_thread(void)
3271 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3272 ext4_li_info, "ext4lazyinit");
3273 if (IS_ERR(ext4_lazyinit_task)) {
3274 int err = PTR_ERR(ext4_lazyinit_task);
3275 ext4_clear_request_list();
3276 kfree(ext4_li_info);
3277 ext4_li_info = NULL;
3278 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3279 "initialization thread\n",
3283 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3288 * Check whether it make sense to run itable init. thread or not.
3289 * If there is at least one uninitialized inode table, return
3290 * corresponding group number, else the loop goes through all
3291 * groups and return total number of groups.
3293 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3295 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3296 struct ext4_group_desc *gdp = NULL;
3298 if (!ext4_has_group_desc_csum(sb))
3301 for (group = 0; group < ngroups; group++) {
3302 gdp = ext4_get_group_desc(sb, group, NULL);
3306 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3313 static int ext4_li_info_new(void)
3315 struct ext4_lazy_init *eli = NULL;
3317 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3321 INIT_LIST_HEAD(&eli->li_request_list);
3322 mutex_init(&eli->li_list_mtx);
3324 eli->li_state |= EXT4_LAZYINIT_QUIT;
3331 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3334 struct ext4_sb_info *sbi = EXT4_SB(sb);
3335 struct ext4_li_request *elr;
3337 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3343 elr->lr_next_group = start;
3346 * Randomize first schedule time of the request to
3347 * spread the inode table initialization requests
3350 elr->lr_next_sched = jiffies + (prandom_u32() %
3351 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3355 int ext4_register_li_request(struct super_block *sb,
3356 ext4_group_t first_not_zeroed)
3358 struct ext4_sb_info *sbi = EXT4_SB(sb);
3359 struct ext4_li_request *elr = NULL;
3360 ext4_group_t ngroups = sbi->s_groups_count;
3363 mutex_lock(&ext4_li_mtx);
3364 if (sbi->s_li_request != NULL) {
3366 * Reset timeout so it can be computed again, because
3367 * s_li_wait_mult might have changed.
3369 sbi->s_li_request->lr_timeout = 0;
3373 if (first_not_zeroed == ngroups || sb_rdonly(sb) ||
3374 !test_opt(sb, INIT_INODE_TABLE))
3377 elr = ext4_li_request_new(sb, first_not_zeroed);
3383 if (NULL == ext4_li_info) {
3384 ret = ext4_li_info_new();
3389 mutex_lock(&ext4_li_info->li_list_mtx);
3390 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3391 mutex_unlock(&ext4_li_info->li_list_mtx);
3393 sbi->s_li_request = elr;
3395 * set elr to NULL here since it has been inserted to
3396 * the request_list and the removal and free of it is
3397 * handled by ext4_clear_request_list from now on.
3401 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3402 ret = ext4_run_lazyinit_thread();
3407 mutex_unlock(&ext4_li_mtx);
3414 * We do not need to lock anything since this is called on
3417 static void ext4_destroy_lazyinit_thread(void)
3420 * If thread exited earlier
3421 * there's nothing to be done.
3423 if (!ext4_li_info || !ext4_lazyinit_task)
3426 kthread_stop(ext4_lazyinit_task);
3429 static int set_journal_csum_feature_set(struct super_block *sb)
3432 int compat, incompat;
3433 struct ext4_sb_info *sbi = EXT4_SB(sb);
3435 if (ext4_has_metadata_csum(sb)) {
3436 /* journal checksum v3 */
3438 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
3440 /* journal checksum v1 */
3441 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3445 jbd2_journal_clear_features(sbi->s_journal,
3446 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3447 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3448 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3449 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3450 ret = jbd2_journal_set_features(sbi->s_journal,
3452 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3454 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3455 ret = jbd2_journal_set_features(sbi->s_journal,
3458 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3459 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3461 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3462 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3469 * Note: calculating the overhead so we can be compatible with
3470 * historical BSD practice is quite difficult in the face of
3471 * clusters/bigalloc. This is because multiple metadata blocks from
3472 * different block group can end up in the same allocation cluster.
3473 * Calculating the exact overhead in the face of clustered allocation
3474 * requires either O(all block bitmaps) in memory or O(number of block
3475 * groups**2) in time. We will still calculate the superblock for
3476 * older file systems --- and if we come across with a bigalloc file
3477 * system with zero in s_overhead_clusters the estimate will be close to
3478 * correct especially for very large cluster sizes --- but for newer
3479 * file systems, it's better to calculate this figure once at mkfs
3480 * time, and store it in the superblock. If the superblock value is
3481 * present (even for non-bigalloc file systems), we will use it.
3483 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3486 struct ext4_sb_info *sbi = EXT4_SB(sb);
3487 struct ext4_group_desc *gdp;
3488 ext4_fsblk_t first_block, last_block, b;
3489 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3490 int s, j, count = 0;
3492 if (!ext4_has_feature_bigalloc(sb))
3493 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3494 sbi->s_itb_per_group + 2);
3496 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3497 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3498 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3499 for (i = 0; i < ngroups; i++) {
3500 gdp = ext4_get_group_desc(sb, i, NULL);
3501 b = ext4_block_bitmap(sb, gdp);
3502 if (b >= first_block && b <= last_block) {
3503 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3506 b = ext4_inode_bitmap(sb, gdp);
3507 if (b >= first_block && b <= last_block) {
3508 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3511 b = ext4_inode_table(sb, gdp);
3512 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3513 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3514 int c = EXT4_B2C(sbi, b - first_block);
3515 ext4_set_bit(c, buf);
3521 if (ext4_bg_has_super(sb, grp)) {
3522 ext4_set_bit(s++, buf);
3525 j = ext4_bg_num_gdb(sb, grp);
3526 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
3527 ext4_error(sb, "Invalid number of block group "
3528 "descriptor blocks: %d", j);
3529 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
3533 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3537 return EXT4_CLUSTERS_PER_GROUP(sb) -
3538 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3542 * Compute the overhead and stash it in sbi->s_overhead
3544 int ext4_calculate_overhead(struct super_block *sb)
3546 struct ext4_sb_info *sbi = EXT4_SB(sb);
3547 struct ext4_super_block *es = sbi->s_es;
3548 struct inode *j_inode;
3549 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
3550 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3551 ext4_fsblk_t overhead = 0;
3552 char *buf = (char *) get_zeroed_page(GFP_NOFS);
3558 * Compute the overhead (FS structures). This is constant
3559 * for a given filesystem unless the number of block groups
3560 * changes so we cache the previous value until it does.
3564 * All of the blocks before first_data_block are overhead
3566 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3569 * Add the overhead found in each block group
3571 for (i = 0; i < ngroups; i++) {
3574 blks = count_overhead(sb, i, buf);
3577 memset(buf, 0, PAGE_SIZE);
3582 * Add the internal journal blocks whether the journal has been
3585 if (sbi->s_journal && !sbi->journal_bdev)
3586 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3587 else if (ext4_has_feature_journal(sb) && !sbi->s_journal) {
3588 j_inode = ext4_get_journal_inode(sb, j_inum);
3590 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
3591 overhead += EXT4_NUM_B2C(sbi, j_blocks);
3594 ext4_msg(sb, KERN_ERR, "can't get journal size");
3597 sbi->s_overhead = overhead;
3599 free_page((unsigned long) buf);
3603 static void ext4_set_resv_clusters(struct super_block *sb)
3605 ext4_fsblk_t resv_clusters;
3606 struct ext4_sb_info *sbi = EXT4_SB(sb);
3609 * There's no need to reserve anything when we aren't using extents.
3610 * The space estimates are exact, there are no unwritten extents,
3611 * hole punching doesn't need new metadata... This is needed especially
3612 * to keep ext2/3 backward compatibility.
3614 if (!ext4_has_feature_extents(sb))
3617 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3618 * This should cover the situations where we can not afford to run
3619 * out of space like for example punch hole, or converting
3620 * unwritten extents in delalloc path. In most cases such
3621 * allocation would require 1, or 2 blocks, higher numbers are
3624 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3625 sbi->s_cluster_bits);
3627 do_div(resv_clusters, 50);
3628 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3630 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
3633 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3635 struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev);
3636 char *orig_data = kstrdup(data, GFP_KERNEL);
3637 struct buffer_head *bh;
3638 struct ext4_super_block *es = NULL;
3639 struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3641 ext4_fsblk_t sb_block = get_sb_block(&data);
3642 ext4_fsblk_t logical_sb_block;
3643 unsigned long offset = 0;
3644 unsigned long journal_devnum = 0;
3645 unsigned long def_mount_opts;
3649 int blocksize, clustersize;
3650 unsigned int db_count;
3652 int needs_recovery, has_huge_files, has_bigalloc;
3655 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3656 ext4_group_t first_not_zeroed;
3658 if ((data && !orig_data) || !sbi)
3661 sbi->s_daxdev = dax_dev;
3662 sbi->s_blockgroup_lock =
3663 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3664 if (!sbi->s_blockgroup_lock)
3667 sb->s_fs_info = sbi;
3669 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3670 sbi->s_sb_block = sb_block;
3671 if (sb->s_bdev->bd_part)
3672 sbi->s_sectors_written_start =
3673 part_stat_read(sb->s_bdev->bd_part, sectors[STAT_WRITE]);
3675 /* Cleanup superblock name */
3676 strreplace(sb->s_id, '/', '!');
3678 /* -EINVAL is default */
3680 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3682 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3687 * The ext4 superblock will not be buffer aligned for other than 1kB
3688 * block sizes. We need to calculate the offset from buffer start.
3690 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3691 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3692 offset = do_div(logical_sb_block, blocksize);
3694 logical_sb_block = sb_block;
3697 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
3698 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3702 * Note: s_es must be initialized as soon as possible because
3703 * some ext4 macro-instructions depend on its value
3705 es = (struct ext4_super_block *) (bh->b_data + offset);
3707 sb->s_magic = le16_to_cpu(es->s_magic);
3708 if (sb->s_magic != EXT4_SUPER_MAGIC)
3710 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3712 /* Warn if metadata_csum and gdt_csum are both set. */
3713 if (ext4_has_feature_metadata_csum(sb) &&
3714 ext4_has_feature_gdt_csum(sb))
3715 ext4_warning(sb, "metadata_csum and uninit_bg are "
3716 "redundant flags; please run fsck.");
3718 /* Check for a known checksum algorithm */
3719 if (!ext4_verify_csum_type(sb, es)) {
3720 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3721 "unknown checksum algorithm.");
3726 /* Load the checksum driver */
3727 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3728 if (IS_ERR(sbi->s_chksum_driver)) {
3729 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3730 ret = PTR_ERR(sbi->s_chksum_driver);
3731 sbi->s_chksum_driver = NULL;
3735 /* Check superblock checksum */
3736 if (!ext4_superblock_csum_verify(sb, es)) {
3737 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3738 "invalid superblock checksum. Run e2fsck?");
3744 /* Precompute checksum seed for all metadata */
3745 if (ext4_has_feature_csum_seed(sb))
3746 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
3747 else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb))
3748 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3749 sizeof(es->s_uuid));
3751 /* Set defaults before we parse the mount options */
3752 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3753 set_opt(sb, INIT_INODE_TABLE);
3754 if (def_mount_opts & EXT4_DEFM_DEBUG)
3756 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3758 if (def_mount_opts & EXT4_DEFM_UID16)
3759 set_opt(sb, NO_UID32);
3760 /* xattr user namespace & acls are now defaulted on */
3761 set_opt(sb, XATTR_USER);
3762 set_opt(sb, DIOREAD_NOLOCK);
3763 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3764 set_opt(sb, POSIX_ACL);
3766 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3767 if (ext4_has_metadata_csum(sb))
3768 set_opt(sb, JOURNAL_CHECKSUM);
3770 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3771 set_opt(sb, JOURNAL_DATA);
3772 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3773 set_opt(sb, ORDERED_DATA);
3774 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3775 set_opt(sb, WRITEBACK_DATA);
3777 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3778 set_opt(sb, ERRORS_PANIC);
3779 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3780 set_opt(sb, ERRORS_CONT);
3782 set_opt(sb, ERRORS_RO);
3783 /* block_validity enabled by default; disable with noblock_validity */
3784 set_opt(sb, BLOCK_VALIDITY);
3785 if (def_mount_opts & EXT4_DEFM_DISCARD)
3786 set_opt(sb, DISCARD);
3788 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3789 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3790 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3791 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3792 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3794 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3795 set_opt(sb, BARRIER);
3798 * enable delayed allocation by default
3799 * Use -o nodelalloc to turn it off
3801 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3802 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3803 set_opt(sb, DELALLOC);
3806 * set default s_li_wait_mult for lazyinit, for the case there is
3807 * no mount option specified.
3809 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3811 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3812 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3813 blocksize > EXT4_MAX_BLOCK_SIZE) {
3814 ext4_msg(sb, KERN_ERR,
3815 "Unsupported filesystem blocksize %d (%d log_block_size)",
3816 blocksize, le32_to_cpu(es->s_log_block_size));
3820 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3821 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3822 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3824 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3825 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3826 if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) {
3827 ext4_msg(sb, KERN_ERR, "invalid first ino: %u",
3831 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3832 (!is_power_of_2(sbi->s_inode_size)) ||
3833 (sbi->s_inode_size > blocksize)) {
3834 ext4_msg(sb, KERN_ERR,
3835 "unsupported inode size: %d",
3837 ext4_msg(sb, KERN_ERR, "blocksize: %d", blocksize);
3841 * i_atime_extra is the last extra field available for
3842 * [acm]times in struct ext4_inode. Checking for that
3843 * field should suffice to ensure we have extra space
3846 if (sbi->s_inode_size >= offsetof(struct ext4_inode, i_atime_extra) +
3847 sizeof(((struct ext4_inode *)0)->i_atime_extra)) {
3848 sb->s_time_gran = 1;
3849 sb->s_time_max = EXT4_EXTRA_TIMESTAMP_MAX;
3851 sb->s_time_gran = NSEC_PER_SEC;
3852 sb->s_time_max = EXT4_NON_EXTRA_TIMESTAMP_MAX;
3854 sb->s_time_min = EXT4_TIMESTAMP_MIN;
3856 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3857 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3858 EXT4_GOOD_OLD_INODE_SIZE;
3859 if (ext4_has_feature_extra_isize(sb)) {
3860 unsigned v, max = (sbi->s_inode_size -
3861 EXT4_GOOD_OLD_INODE_SIZE);
3863 v = le16_to_cpu(es->s_want_extra_isize);
3865 ext4_msg(sb, KERN_ERR,
3866 "bad s_want_extra_isize: %d", v);
3869 if (sbi->s_want_extra_isize < v)
3870 sbi->s_want_extra_isize = v;
3872 v = le16_to_cpu(es->s_min_extra_isize);
3874 ext4_msg(sb, KERN_ERR,
3875 "bad s_min_extra_isize: %d", v);
3878 if (sbi->s_want_extra_isize < v)
3879 sbi->s_want_extra_isize = v;
3883 if (sbi->s_es->s_mount_opts[0]) {
3884 char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
3885 sizeof(sbi->s_es->s_mount_opts),
3889 if (!parse_options(s_mount_opts, sb, &journal_devnum,
3890 &journal_ioprio, 0)) {
3891 ext4_msg(sb, KERN_WARNING,
3892 "failed to parse options in superblock: %s",
3895 kfree(s_mount_opts);
3897 sbi->s_def_mount_opt = sbi->s_mount_opt;
3898 if (!parse_options((char *) data, sb, &journal_devnum,
3899 &journal_ioprio, 0))
3902 #ifdef CONFIG_UNICODE
3903 if (ext4_has_feature_casefold(sb) && !sbi->s_encoding) {
3904 const struct ext4_sb_encodings *encoding_info;
3905 struct unicode_map *encoding;
3906 __u16 encoding_flags;
3908 if (ext4_has_feature_encrypt(sb)) {
3909 ext4_msg(sb, KERN_ERR,
3910 "Can't mount with encoding and encryption");
3914 if (ext4_sb_read_encoding(es, &encoding_info,
3916 ext4_msg(sb, KERN_ERR,
3917 "Encoding requested by superblock is unknown");
3921 encoding = utf8_load(encoding_info->version);
3922 if (IS_ERR(encoding)) {
3923 ext4_msg(sb, KERN_ERR,
3924 "can't mount with superblock charset: %s-%s "
3925 "not supported by the kernel. flags: 0x%x.",
3926 encoding_info->name, encoding_info->version,
3930 ext4_msg(sb, KERN_INFO,"Using encoding defined by superblock: "
3931 "%s-%s with flags 0x%hx", encoding_info->name,
3932 encoding_info->version?:"\b", encoding_flags);
3934 sbi->s_encoding = encoding;
3935 sbi->s_encoding_flags = encoding_flags;
3939 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3940 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting with data=journal disables delayed allocation, dioread_nolock, and O_DIRECT support!\n");
3941 clear_opt(sb, DIOREAD_NOLOCK);
3942 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3943 ext4_msg(sb, KERN_ERR, "can't mount with "
3944 "both data=journal and delalloc");
3947 if (test_opt(sb, DIOREAD_NOLOCK)) {
3948 ext4_msg(sb, KERN_ERR, "can't mount with "
3949 "both data=journal and dioread_nolock");
3952 if (test_opt(sb, DAX)) {
3953 ext4_msg(sb, KERN_ERR, "can't mount with "
3954 "both data=journal and dax");
3957 if (ext4_has_feature_encrypt(sb)) {
3958 ext4_msg(sb, KERN_WARNING,
3959 "encrypted files will use data=ordered "
3960 "instead of data journaling mode");
3962 if (test_opt(sb, DELALLOC))
3963 clear_opt(sb, DELALLOC);
3965 sb->s_iflags |= SB_I_CGROUPWB;
3968 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
3969 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
3971 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3972 (ext4_has_compat_features(sb) ||
3973 ext4_has_ro_compat_features(sb) ||
3974 ext4_has_incompat_features(sb)))
3975 ext4_msg(sb, KERN_WARNING,
3976 "feature flags set on rev 0 fs, "
3977 "running e2fsck is recommended");
3979 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3980 set_opt2(sb, HURD_COMPAT);
3981 if (ext4_has_feature_64bit(sb)) {
3982 ext4_msg(sb, KERN_ERR,
3983 "The Hurd can't support 64-bit file systems");
3988 * ea_inode feature uses l_i_version field which is not
3989 * available in HURD_COMPAT mode.
3991 if (ext4_has_feature_ea_inode(sb)) {
3992 ext4_msg(sb, KERN_ERR,
3993 "ea_inode feature is not supported for Hurd");
3998 if (IS_EXT2_SB(sb)) {
3999 if (ext2_feature_set_ok(sb))
4000 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
4001 "using the ext4 subsystem");
4004 * If we're probing be silent, if this looks like
4005 * it's actually an ext[34] filesystem.
4007 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4009 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
4010 "to feature incompatibilities");
4015 if (IS_EXT3_SB(sb)) {
4016 if (ext3_feature_set_ok(sb))
4017 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
4018 "using the ext4 subsystem");
4021 * If we're probing be silent, if this looks like
4022 * it's actually an ext4 filesystem.
4024 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4026 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
4027 "to feature incompatibilities");
4033 * Check feature flags regardless of the revision level, since we
4034 * previously didn't change the revision level when setting the flags,
4035 * so there is a chance incompat flags are set on a rev 0 filesystem.
4037 if (!ext4_feature_set_ok(sb, (sb_rdonly(sb))))
4040 if (le32_to_cpu(es->s_log_block_size) >
4041 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
4042 ext4_msg(sb, KERN_ERR,
4043 "Invalid log block size: %u",
4044 le32_to_cpu(es->s_log_block_size));
4047 if (le32_to_cpu(es->s_log_cluster_size) >
4048 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
4049 ext4_msg(sb, KERN_ERR,
4050 "Invalid log cluster size: %u",
4051 le32_to_cpu(es->s_log_cluster_size));
4055 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
4056 ext4_msg(sb, KERN_ERR,
4057 "Number of reserved GDT blocks insanely large: %d",
4058 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
4062 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
4063 if (ext4_has_feature_inline_data(sb)) {
4064 ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem"
4065 " that may contain inline data");
4068 if (!bdev_dax_supported(sb->s_bdev, blocksize)) {
4069 ext4_msg(sb, KERN_ERR,
4070 "DAX unsupported by block device.");
4075 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
4076 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
4077 es->s_encryption_level);
4081 if (sb->s_blocksize != blocksize) {
4082 /* Validate the filesystem blocksize */
4083 if (!sb_set_blocksize(sb, blocksize)) {
4084 ext4_msg(sb, KERN_ERR, "bad block size %d",
4090 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
4091 offset = do_div(logical_sb_block, blocksize);
4092 bh = sb_bread_unmovable(sb, logical_sb_block);
4094 ext4_msg(sb, KERN_ERR,
4095 "Can't read superblock on 2nd try");
4098 es = (struct ext4_super_block *)(bh->b_data + offset);
4100 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
4101 ext4_msg(sb, KERN_ERR,
4102 "Magic mismatch, very weird!");
4107 has_huge_files = ext4_has_feature_huge_file(sb);
4108 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
4110 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
4112 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
4113 if (ext4_has_feature_64bit(sb)) {
4114 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
4115 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
4116 !is_power_of_2(sbi->s_desc_size)) {
4117 ext4_msg(sb, KERN_ERR,
4118 "unsupported descriptor size %lu",
4123 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
4125 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
4126 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
4128 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
4129 if (sbi->s_inodes_per_block == 0)
4131 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
4132 sbi->s_inodes_per_group > blocksize * 8) {
4133 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
4134 sbi->s_blocks_per_group);
4137 sbi->s_itb_per_group = sbi->s_inodes_per_group /
4138 sbi->s_inodes_per_block;
4139 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
4141 sbi->s_mount_state = le16_to_cpu(es->s_state);
4142 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
4143 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
4145 for (i = 0; i < 4; i++)
4146 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
4147 sbi->s_def_hash_version = es->s_def_hash_version;
4148 if (ext4_has_feature_dir_index(sb)) {
4149 i = le32_to_cpu(es->s_flags);
4150 if (i & EXT2_FLAGS_UNSIGNED_HASH)
4151 sbi->s_hash_unsigned = 3;
4152 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
4153 #ifdef __CHAR_UNSIGNED__
4156 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
4157 sbi->s_hash_unsigned = 3;
4161 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
4166 /* Handle clustersize */
4167 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
4168 has_bigalloc = ext4_has_feature_bigalloc(sb);
4170 if (clustersize < blocksize) {
4171 ext4_msg(sb, KERN_ERR,
4172 "cluster size (%d) smaller than "
4173 "block size (%d)", clustersize, blocksize);
4176 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
4177 le32_to_cpu(es->s_log_block_size);
4178 sbi->s_clusters_per_group =
4179 le32_to_cpu(es->s_clusters_per_group);
4180 if (sbi->s_clusters_per_group > blocksize * 8) {
4181 ext4_msg(sb, KERN_ERR,
4182 "#clusters per group too big: %lu",
4183 sbi->s_clusters_per_group);
4186 if (sbi->s_blocks_per_group !=
4187 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
4188 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
4189 "clusters per group (%lu) inconsistent",
4190 sbi->s_blocks_per_group,
4191 sbi->s_clusters_per_group);
4195 if (clustersize != blocksize) {
4196 ext4_msg(sb, KERN_ERR,
4197 "fragment/cluster size (%d) != "
4198 "block size (%d)", clustersize, blocksize);
4201 if (sbi->s_blocks_per_group > blocksize * 8) {
4202 ext4_msg(sb, KERN_ERR,
4203 "#blocks per group too big: %lu",
4204 sbi->s_blocks_per_group);
4207 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
4208 sbi->s_cluster_bits = 0;
4210 sbi->s_cluster_ratio = clustersize / blocksize;
4212 /* Do we have standard group size of clustersize * 8 blocks ? */
4213 if (sbi->s_blocks_per_group == clustersize << 3)
4214 set_opt2(sb, STD_GROUP_SIZE);
4217 * Test whether we have more sectors than will fit in sector_t,
4218 * and whether the max offset is addressable by the page cache.
4220 err = generic_check_addressable(sb->s_blocksize_bits,
4221 ext4_blocks_count(es));
4223 ext4_msg(sb, KERN_ERR, "filesystem"
4224 " too large to mount safely on this system");
4228 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
4231 /* check blocks count against device size */
4232 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
4233 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
4234 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
4235 "exceeds size of device (%llu blocks)",
4236 ext4_blocks_count(es), blocks_count);
4241 * It makes no sense for the first data block to be beyond the end
4242 * of the filesystem.
4244 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
4245 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4246 "block %u is beyond end of filesystem (%llu)",
4247 le32_to_cpu(es->s_first_data_block),
4248 ext4_blocks_count(es));
4251 if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) &&
4252 (sbi->s_cluster_ratio == 1)) {
4253 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4254 "block is 0 with a 1k block and cluster size");
4258 blocks_count = (ext4_blocks_count(es) -
4259 le32_to_cpu(es->s_first_data_block) +
4260 EXT4_BLOCKS_PER_GROUP(sb) - 1);
4261 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4262 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
4263 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4264 "(block count %llu, first data block %u, "
4265 "blocks per group %lu)", sbi->s_groups_count,
4266 ext4_blocks_count(es),
4267 le32_to_cpu(es->s_first_data_block),
4268 EXT4_BLOCKS_PER_GROUP(sb));
4271 sbi->s_groups_count = blocks_count;
4272 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
4273 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
4274 if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) !=
4275 le32_to_cpu(es->s_inodes_count)) {
4276 ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu",
4277 le32_to_cpu(es->s_inodes_count),
4278 ((u64)sbi->s_groups_count * sbi->s_inodes_per_group));
4282 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
4283 EXT4_DESC_PER_BLOCK(sb);
4284 if (ext4_has_feature_meta_bg(sb)) {
4285 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
4286 ext4_msg(sb, KERN_WARNING,
4287 "first meta block group too large: %u "
4288 "(group descriptor block count %u)",
4289 le32_to_cpu(es->s_first_meta_bg), db_count);
4293 sbi->s_group_desc = kvmalloc_array(db_count,
4294 sizeof(struct buffer_head *),
4296 if (sbi->s_group_desc == NULL) {
4297 ext4_msg(sb, KERN_ERR, "not enough memory");
4302 bgl_lock_init(sbi->s_blockgroup_lock);
4304 /* Pre-read the descriptors into the buffer cache */
4305 for (i = 0; i < db_count; i++) {
4306 block = descriptor_loc(sb, logical_sb_block, i);
4307 sb_breadahead(sb, block);
4310 for (i = 0; i < db_count; i++) {
4311 block = descriptor_loc(sb, logical_sb_block, i);
4312 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
4313 if (!sbi->s_group_desc[i]) {
4314 ext4_msg(sb, KERN_ERR,
4315 "can't read group descriptor %d", i);
4320 sbi->s_gdb_count = db_count;
4321 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
4322 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
4323 ret = -EFSCORRUPTED;
4327 timer_setup(&sbi->s_err_report, print_daily_error_info, 0);
4329 /* Register extent status tree shrinker */
4330 if (ext4_es_register_shrinker(sbi))
4333 sbi->s_stripe = ext4_get_stripe_size(sbi);
4334 sbi->s_extent_max_zeroout_kb = 32;
4337 * set up enough so that it can read an inode
4339 sb->s_op = &ext4_sops;
4340 sb->s_export_op = &ext4_export_ops;
4341 sb->s_xattr = ext4_xattr_handlers;
4342 #ifdef CONFIG_FS_ENCRYPTION
4343 sb->s_cop = &ext4_cryptops;
4345 #ifdef CONFIG_FS_VERITY
4346 sb->s_vop = &ext4_verityops;
4349 sb->dq_op = &ext4_quota_operations;
4350 if (ext4_has_feature_quota(sb))
4351 sb->s_qcop = &dquot_quotactl_sysfile_ops;
4353 sb->s_qcop = &ext4_qctl_operations;
4354 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
4356 memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
4358 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
4359 mutex_init(&sbi->s_orphan_lock);
4363 needs_recovery = (es->s_last_orphan != 0 ||
4364 ext4_has_feature_journal_needs_recovery(sb));
4366 if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb))
4367 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
4368 goto failed_mount3a;
4371 * The first inode we look at is the journal inode. Don't try
4372 * root first: it may be modified in the journal!
4374 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
4375 err = ext4_load_journal(sb, es, journal_devnum);
4377 goto failed_mount3a;
4378 } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) &&
4379 ext4_has_feature_journal_needs_recovery(sb)) {
4380 ext4_msg(sb, KERN_ERR, "required journal recovery "
4381 "suppressed and not mounted read-only");
4382 goto failed_mount_wq;
4384 /* Nojournal mode, all journal mount options are illegal */
4385 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
4386 ext4_msg(sb, KERN_ERR, "can't mount with "
4387 "journal_checksum, fs mounted w/o journal");
4388 goto failed_mount_wq;
4390 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4391 ext4_msg(sb, KERN_ERR, "can't mount with "
4392 "journal_async_commit, fs mounted w/o journal");
4393 goto failed_mount_wq;
4395 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
4396 ext4_msg(sb, KERN_ERR, "can't mount with "
4397 "commit=%lu, fs mounted w/o journal",
4398 sbi->s_commit_interval / HZ);
4399 goto failed_mount_wq;
4401 if (EXT4_MOUNT_DATA_FLAGS &
4402 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
4403 ext4_msg(sb, KERN_ERR, "can't mount with "
4404 "data=, fs mounted w/o journal");
4405 goto failed_mount_wq;
4407 sbi->s_def_mount_opt &= ~EXT4_MOUNT_JOURNAL_CHECKSUM;
4408 clear_opt(sb, JOURNAL_CHECKSUM);
4409 clear_opt(sb, DATA_FLAGS);
4410 sbi->s_journal = NULL;
4415 if (ext4_has_feature_64bit(sb) &&
4416 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4417 JBD2_FEATURE_INCOMPAT_64BIT)) {
4418 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
4419 goto failed_mount_wq;
4422 if (!set_journal_csum_feature_set(sb)) {
4423 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4425 goto failed_mount_wq;
4428 /* We have now updated the journal if required, so we can
4429 * validate the data journaling mode. */
4430 switch (test_opt(sb, DATA_FLAGS)) {
4432 /* No mode set, assume a default based on the journal
4433 * capabilities: ORDERED_DATA if the journal can
4434 * cope, else JOURNAL_DATA
4436 if (jbd2_journal_check_available_features
4437 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4438 set_opt(sb, ORDERED_DATA);
4439 sbi->s_def_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
4441 set_opt(sb, JOURNAL_DATA);
4442 sbi->s_def_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
4446 case EXT4_MOUNT_ORDERED_DATA:
4447 case EXT4_MOUNT_WRITEBACK_DATA:
4448 if (!jbd2_journal_check_available_features
4449 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4450 ext4_msg(sb, KERN_ERR, "Journal does not support "
4451 "requested data journaling mode");
4452 goto failed_mount_wq;
4458 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
4459 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4460 ext4_msg(sb, KERN_ERR, "can't mount with "
4461 "journal_async_commit in data=ordered mode");
4462 goto failed_mount_wq;
4465 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4467 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4470 if (!test_opt(sb, NO_MBCACHE)) {
4471 sbi->s_ea_block_cache = ext4_xattr_create_cache();
4472 if (!sbi->s_ea_block_cache) {
4473 ext4_msg(sb, KERN_ERR,
4474 "Failed to create ea_block_cache");
4475 goto failed_mount_wq;
4478 if (ext4_has_feature_ea_inode(sb)) {
4479 sbi->s_ea_inode_cache = ext4_xattr_create_cache();
4480 if (!sbi->s_ea_inode_cache) {
4481 ext4_msg(sb, KERN_ERR,
4482 "Failed to create ea_inode_cache");
4483 goto failed_mount_wq;
4488 if (ext4_has_feature_verity(sb) && blocksize != PAGE_SIZE) {
4489 ext4_msg(sb, KERN_ERR, "Unsupported blocksize for fs-verity");
4490 goto failed_mount_wq;
4493 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !sb_rdonly(sb) &&
4494 !ext4_has_feature_encrypt(sb)) {
4495 ext4_set_feature_encrypt(sb);
4496 ext4_commit_super(sb, 1);
4500 * Get the # of file system overhead blocks from the
4501 * superblock if present.
4503 if (es->s_overhead_clusters)
4504 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4506 err = ext4_calculate_overhead(sb);
4508 goto failed_mount_wq;
4512 * The maximum number of concurrent works can be high and
4513 * concurrency isn't really necessary. Limit it to 1.
4515 EXT4_SB(sb)->rsv_conversion_wq =
4516 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4517 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4518 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
4524 * The jbd2_journal_load will have done any necessary log recovery,
4525 * so we can safely mount the rest of the filesystem now.
4528 root = ext4_iget(sb, EXT4_ROOT_INO, EXT4_IGET_SPECIAL);
4530 ext4_msg(sb, KERN_ERR, "get root inode failed");
4531 ret = PTR_ERR(root);
4535 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4536 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4541 #ifdef CONFIG_UNICODE
4542 if (sbi->s_encoding)
4543 sb->s_d_op = &ext4_dentry_ops;
4546 sb->s_root = d_make_root(root);
4548 ext4_msg(sb, KERN_ERR, "get root dentry failed");
4553 ret = ext4_setup_super(sb, es, sb_rdonly(sb));
4554 if (ret == -EROFS) {
4555 sb->s_flags |= SB_RDONLY;
4558 goto failed_mount4a;
4560 ext4_set_resv_clusters(sb);
4562 err = ext4_setup_system_zone(sb);
4564 ext4_msg(sb, KERN_ERR, "failed to initialize system "
4566 goto failed_mount4a;
4570 err = ext4_mb_init(sb);
4572 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4577 block = ext4_count_free_clusters(sb);
4578 ext4_free_blocks_count_set(sbi->s_es,
4579 EXT4_C2B(sbi, block));
4580 ext4_superblock_csum_set(sb);
4581 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4584 unsigned long freei = ext4_count_free_inodes(sb);
4585 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
4586 ext4_superblock_csum_set(sb);
4587 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4591 err = percpu_counter_init(&sbi->s_dirs_counter,
4592 ext4_count_dirs(sb), GFP_KERNEL);
4594 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4597 err = percpu_init_rwsem(&sbi->s_journal_flag_rwsem);
4600 ext4_msg(sb, KERN_ERR, "insufficient memory");
4604 if (ext4_has_feature_flex_bg(sb))
4605 if (!ext4_fill_flex_info(sb)) {
4606 ext4_msg(sb, KERN_ERR,
4607 "unable to initialize "
4608 "flex_bg meta info!");
4612 err = ext4_register_li_request(sb, first_not_zeroed);
4616 err = ext4_register_sysfs(sb);
4621 /* Enable quota usage during mount. */
4622 if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) {
4623 err = ext4_enable_quotas(sb);
4627 #endif /* CONFIG_QUOTA */
4629 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4630 ext4_orphan_cleanup(sb, es);
4631 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4632 if (needs_recovery) {
4633 ext4_msg(sb, KERN_INFO, "recovery complete");
4634 ext4_mark_recovery_complete(sb, es);
4636 if (EXT4_SB(sb)->s_journal) {
4637 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4638 descr = " journalled data mode";
4639 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4640 descr = " ordered data mode";
4642 descr = " writeback data mode";
4644 descr = "out journal";
4646 if (test_opt(sb, DISCARD)) {
4647 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4648 if (!blk_queue_discard(q))
4649 ext4_msg(sb, KERN_WARNING,
4650 "mounting with \"discard\" option, but "
4651 "the device does not support discard");
4654 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
4655 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4656 "Opts: %.*s%s%s", descr,
4657 (int) sizeof(sbi->s_es->s_mount_opts),
4658 sbi->s_es->s_mount_opts,
4659 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4661 if (es->s_error_count)
4662 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4664 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4665 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4666 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4667 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4674 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4679 ext4_unregister_sysfs(sb);
4682 ext4_unregister_li_request(sb);
4684 ext4_mb_release(sb);
4685 if (sbi->s_flex_groups)
4686 kvfree(sbi->s_flex_groups);
4687 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4688 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4689 percpu_counter_destroy(&sbi->s_dirs_counter);
4690 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4691 percpu_free_rwsem(&sbi->s_journal_flag_rwsem);
4693 ext4_ext_release(sb);
4694 ext4_release_system_zone(sb);
4699 ext4_msg(sb, KERN_ERR, "mount failed");
4700 if (EXT4_SB(sb)->rsv_conversion_wq)
4701 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4703 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
4704 sbi->s_ea_inode_cache = NULL;
4706 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
4707 sbi->s_ea_block_cache = NULL;
4709 if (sbi->s_journal) {
4710 jbd2_journal_destroy(sbi->s_journal);
4711 sbi->s_journal = NULL;
4714 ext4_es_unregister_shrinker(sbi);
4716 del_timer_sync(&sbi->s_err_report);
4718 kthread_stop(sbi->s_mmp_tsk);
4720 for (i = 0; i < db_count; i++)
4721 brelse(sbi->s_group_desc[i]);
4722 kvfree(sbi->s_group_desc);
4724 if (sbi->s_chksum_driver)
4725 crypto_free_shash(sbi->s_chksum_driver);
4727 #ifdef CONFIG_UNICODE
4728 utf8_unload(sbi->s_encoding);
4732 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4733 kfree(get_qf_name(sb, sbi, i));
4735 ext4_blkdev_remove(sbi);
4738 sb->s_fs_info = NULL;
4739 kfree(sbi->s_blockgroup_lock);
4743 fs_put_dax(dax_dev);
4744 return err ? err : ret;
4748 * Setup any per-fs journal parameters now. We'll do this both on
4749 * initial mount, once the journal has been initialised but before we've
4750 * done any recovery; and again on any subsequent remount.
4752 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4754 struct ext4_sb_info *sbi = EXT4_SB(sb);
4756 journal->j_commit_interval = sbi->s_commit_interval;
4757 journal->j_min_batch_time = sbi->s_min_batch_time;
4758 journal->j_max_batch_time = sbi->s_max_batch_time;
4760 write_lock(&journal->j_state_lock);
4761 if (test_opt(sb, BARRIER))
4762 journal->j_flags |= JBD2_BARRIER;
4764 journal->j_flags &= ~JBD2_BARRIER;
4765 if (test_opt(sb, DATA_ERR_ABORT))
4766 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4768 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4769 write_unlock(&journal->j_state_lock);
4772 static struct inode *ext4_get_journal_inode(struct super_block *sb,
4773 unsigned int journal_inum)
4775 struct inode *journal_inode;
4778 * Test for the existence of a valid inode on disk. Bad things
4779 * happen if we iget() an unused inode, as the subsequent iput()
4780 * will try to delete it.
4782 journal_inode = ext4_iget(sb, journal_inum, EXT4_IGET_SPECIAL);
4783 if (IS_ERR(journal_inode)) {
4784 ext4_msg(sb, KERN_ERR, "no journal found");
4787 if (!journal_inode->i_nlink) {
4788 make_bad_inode(journal_inode);
4789 iput(journal_inode);
4790 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4794 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4795 journal_inode, journal_inode->i_size);
4796 if (!S_ISREG(journal_inode->i_mode)) {
4797 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4798 iput(journal_inode);
4801 return journal_inode;
4804 static journal_t *ext4_get_journal(struct super_block *sb,
4805 unsigned int journal_inum)
4807 struct inode *journal_inode;
4810 BUG_ON(!ext4_has_feature_journal(sb));
4812 journal_inode = ext4_get_journal_inode(sb, journal_inum);
4816 journal = jbd2_journal_init_inode(journal_inode);
4818 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4819 iput(journal_inode);
4822 journal->j_private = sb;
4823 ext4_init_journal_params(sb, journal);
4827 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4830 struct buffer_head *bh;
4834 int hblock, blocksize;
4835 ext4_fsblk_t sb_block;
4836 unsigned long offset;
4837 struct ext4_super_block *es;
4838 struct block_device *bdev;
4840 BUG_ON(!ext4_has_feature_journal(sb));
4842 bdev = ext4_blkdev_get(j_dev, sb);
4846 blocksize = sb->s_blocksize;
4847 hblock = bdev_logical_block_size(bdev);
4848 if (blocksize < hblock) {
4849 ext4_msg(sb, KERN_ERR,
4850 "blocksize too small for journal device");
4854 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4855 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4856 set_blocksize(bdev, blocksize);
4857 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4858 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4859 "external journal");
4863 es = (struct ext4_super_block *) (bh->b_data + offset);
4864 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4865 !(le32_to_cpu(es->s_feature_incompat) &
4866 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4867 ext4_msg(sb, KERN_ERR, "external journal has "
4873 if ((le32_to_cpu(es->s_feature_ro_compat) &
4874 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4875 es->s_checksum != ext4_superblock_csum(sb, es)) {
4876 ext4_msg(sb, KERN_ERR, "external journal has "
4877 "corrupt superblock");
4882 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4883 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4888 len = ext4_blocks_count(es);
4889 start = sb_block + 1;
4890 brelse(bh); /* we're done with the superblock */
4892 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4893 start, len, blocksize);
4895 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4898 journal->j_private = sb;
4899 ll_rw_block(REQ_OP_READ, REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4900 wait_on_buffer(journal->j_sb_buffer);
4901 if (!buffer_uptodate(journal->j_sb_buffer)) {
4902 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4905 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4906 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4907 "user (unsupported) - %d",
4908 be32_to_cpu(journal->j_superblock->s_nr_users));
4911 EXT4_SB(sb)->journal_bdev = bdev;
4912 ext4_init_journal_params(sb, journal);
4916 jbd2_journal_destroy(journal);
4918 ext4_blkdev_put(bdev);
4922 static int ext4_load_journal(struct super_block *sb,
4923 struct ext4_super_block *es,
4924 unsigned long journal_devnum)
4927 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4930 int really_read_only;
4932 BUG_ON(!ext4_has_feature_journal(sb));
4934 if (journal_devnum &&
4935 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4936 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4937 "numbers have changed");
4938 journal_dev = new_decode_dev(journal_devnum);
4940 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4942 really_read_only = bdev_read_only(sb->s_bdev);
4945 * Are we loading a blank journal or performing recovery after a
4946 * crash? For recovery, we need to check in advance whether we
4947 * can get read-write access to the device.
4949 if (ext4_has_feature_journal_needs_recovery(sb)) {
4950 if (sb_rdonly(sb)) {
4951 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4952 "required on readonly filesystem");
4953 if (really_read_only) {
4954 ext4_msg(sb, KERN_ERR, "write access "
4955 "unavailable, cannot proceed "
4956 "(try mounting with noload)");
4959 ext4_msg(sb, KERN_INFO, "write access will "
4960 "be enabled during recovery");
4964 if (journal_inum && journal_dev) {
4965 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4966 "and inode journals!");
4971 if (!(journal = ext4_get_journal(sb, journal_inum)))
4974 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4978 if (!(journal->j_flags & JBD2_BARRIER))
4979 ext4_msg(sb, KERN_INFO, "barriers disabled");
4981 if (!ext4_has_feature_journal_needs_recovery(sb))
4982 err = jbd2_journal_wipe(journal, !really_read_only);
4984 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4986 memcpy(save, ((char *) es) +
4987 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4988 err = jbd2_journal_load(journal);
4990 memcpy(((char *) es) + EXT4_S_ERR_START,
4991 save, EXT4_S_ERR_LEN);
4996 ext4_msg(sb, KERN_ERR, "error loading journal");
4997 jbd2_journal_destroy(journal);
5001 EXT4_SB(sb)->s_journal = journal;
5002 ext4_clear_journal_err(sb, es);
5004 if (!really_read_only && journal_devnum &&
5005 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
5006 es->s_journal_dev = cpu_to_le32(journal_devnum);
5008 /* Make sure we flush the recovery flag to disk. */
5009 ext4_commit_super(sb, 1);
5015 static int ext4_commit_super(struct super_block *sb, int sync)
5017 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
5018 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
5021 if (!sbh || block_device_ejected(sb))
5025 * The superblock bh should be mapped, but it might not be if the
5026 * device was hot-removed. Not much we can do but fail the I/O.
5028 if (!buffer_mapped(sbh))
5032 * If the file system is mounted read-only, don't update the
5033 * superblock write time. This avoids updating the superblock
5034 * write time when we are mounting the root file system
5035 * read/only but we need to replay the journal; at that point,
5036 * for people who are east of GMT and who make their clock
5037 * tick in localtime for Windows bug-for-bug compatibility,
5038 * the clock is set in the future, and this will cause e2fsck
5039 * to complain and force a full file system check.
5041 if (!(sb->s_flags & SB_RDONLY))
5042 ext4_update_tstamp(es, s_wtime);
5043 if (sb->s_bdev->bd_part)
5044 es->s_kbytes_written =
5045 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
5046 ((part_stat_read(sb->s_bdev->bd_part,
5047 sectors[STAT_WRITE]) -
5048 EXT4_SB(sb)->s_sectors_written_start) >> 1));
5050 es->s_kbytes_written =
5051 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
5052 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
5053 ext4_free_blocks_count_set(es,
5054 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
5055 &EXT4_SB(sb)->s_freeclusters_counter)));
5056 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
5057 es->s_free_inodes_count =
5058 cpu_to_le32(percpu_counter_sum_positive(
5059 &EXT4_SB(sb)->s_freeinodes_counter));
5060 BUFFER_TRACE(sbh, "marking dirty");
5061 ext4_superblock_csum_set(sb);
5064 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
5066 * Oh, dear. A previous attempt to write the
5067 * superblock failed. This could happen because the
5068 * USB device was yanked out. Or it could happen to
5069 * be a transient write error and maybe the block will
5070 * be remapped. Nothing we can do but to retry the
5071 * write and hope for the best.
5073 ext4_msg(sb, KERN_ERR, "previous I/O error to "
5074 "superblock detected");
5075 clear_buffer_write_io_error(sbh);
5076 set_buffer_uptodate(sbh);
5078 mark_buffer_dirty(sbh);
5081 error = __sync_dirty_buffer(sbh,
5082 REQ_SYNC | (test_opt(sb, BARRIER) ? REQ_FUA : 0));
5083 if (buffer_write_io_error(sbh)) {
5084 ext4_msg(sb, KERN_ERR, "I/O error while writing "
5086 clear_buffer_write_io_error(sbh);
5087 set_buffer_uptodate(sbh);
5094 * Have we just finished recovery? If so, and if we are mounting (or
5095 * remounting) the filesystem readonly, then we will end up with a
5096 * consistent fs on disk. Record that fact.
5098 static void ext4_mark_recovery_complete(struct super_block *sb,
5099 struct ext4_super_block *es)
5101 journal_t *journal = EXT4_SB(sb)->s_journal;
5103 if (!ext4_has_feature_journal(sb)) {
5104 BUG_ON(journal != NULL);
5107 jbd2_journal_lock_updates(journal);
5108 if (jbd2_journal_flush(journal) < 0)
5111 if (ext4_has_feature_journal_needs_recovery(sb) && sb_rdonly(sb)) {
5112 ext4_clear_feature_journal_needs_recovery(sb);
5113 ext4_commit_super(sb, 1);
5117 jbd2_journal_unlock_updates(journal);
5121 * If we are mounting (or read-write remounting) a filesystem whose journal
5122 * has recorded an error from a previous lifetime, move that error to the
5123 * main filesystem now.
5125 static void ext4_clear_journal_err(struct super_block *sb,
5126 struct ext4_super_block *es)
5132 BUG_ON(!ext4_has_feature_journal(sb));
5134 journal = EXT4_SB(sb)->s_journal;
5137 * Now check for any error status which may have been recorded in the
5138 * journal by a prior ext4_error() or ext4_abort()
5141 j_errno = jbd2_journal_errno(journal);
5145 errstr = ext4_decode_error(sb, j_errno, nbuf);
5146 ext4_warning(sb, "Filesystem error recorded "
5147 "from previous mount: %s", errstr);
5148 ext4_warning(sb, "Marking fs in need of filesystem check.");
5150 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
5151 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
5152 ext4_commit_super(sb, 1);
5154 jbd2_journal_clear_err(journal);
5155 jbd2_journal_update_sb_errno(journal);
5160 * Force the running and committing transactions to commit,
5161 * and wait on the commit.
5163 int ext4_force_commit(struct super_block *sb)
5170 journal = EXT4_SB(sb)->s_journal;
5171 return ext4_journal_force_commit(journal);
5174 static int ext4_sync_fs(struct super_block *sb, int wait)
5178 bool needs_barrier = false;
5179 struct ext4_sb_info *sbi = EXT4_SB(sb);
5181 if (unlikely(ext4_forced_shutdown(sbi)))
5184 trace_ext4_sync_fs(sb, wait);
5185 flush_workqueue(sbi->rsv_conversion_wq);
5187 * Writeback quota in non-journalled quota case - journalled quota has
5190 dquot_writeback_dquots(sb, -1);
5192 * Data writeback is possible w/o journal transaction, so barrier must
5193 * being sent at the end of the function. But we can skip it if
5194 * transaction_commit will do it for us.
5196 if (sbi->s_journal) {
5197 target = jbd2_get_latest_transaction(sbi->s_journal);
5198 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
5199 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
5200 needs_barrier = true;
5202 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
5204 ret = jbd2_log_wait_commit(sbi->s_journal,
5207 } else if (wait && test_opt(sb, BARRIER))
5208 needs_barrier = true;
5209 if (needs_barrier) {
5211 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
5220 * LVM calls this function before a (read-only) snapshot is created. This
5221 * gives us a chance to flush the journal completely and mark the fs clean.
5223 * Note that only this function cannot bring a filesystem to be in a clean
5224 * state independently. It relies on upper layer to stop all data & metadata
5227 static int ext4_freeze(struct super_block *sb)
5235 journal = EXT4_SB(sb)->s_journal;
5238 /* Now we set up the journal barrier. */
5239 jbd2_journal_lock_updates(journal);
5242 * Don't clear the needs_recovery flag if we failed to
5243 * flush the journal.
5245 error = jbd2_journal_flush(journal);
5249 /* Journal blocked and flushed, clear needs_recovery flag. */
5250 ext4_clear_feature_journal_needs_recovery(sb);
5253 error = ext4_commit_super(sb, 1);
5256 /* we rely on upper layer to stop further updates */
5257 jbd2_journal_unlock_updates(journal);
5262 * Called by LVM after the snapshot is done. We need to reset the RECOVER
5263 * flag here, even though the filesystem is not technically dirty yet.
5265 static int ext4_unfreeze(struct super_block *sb)
5267 if (sb_rdonly(sb) || ext4_forced_shutdown(EXT4_SB(sb)))
5270 if (EXT4_SB(sb)->s_journal) {
5271 /* Reset the needs_recovery flag before the fs is unlocked. */
5272 ext4_set_feature_journal_needs_recovery(sb);
5275 ext4_commit_super(sb, 1);
5280 * Structure to save mount options for ext4_remount's benefit
5282 struct ext4_mount_options {
5283 unsigned long s_mount_opt;
5284 unsigned long s_mount_opt2;
5287 unsigned long s_commit_interval;
5288 u32 s_min_batch_time, s_max_batch_time;
5291 char *s_qf_names[EXT4_MAXQUOTAS];
5295 static int ext4_remount(struct super_block *sb, int *flags, char *data)
5297 struct ext4_super_block *es;
5298 struct ext4_sb_info *sbi = EXT4_SB(sb);
5299 unsigned long old_sb_flags;
5300 struct ext4_mount_options old_opts;
5301 int enable_quota = 0;
5303 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
5307 char *to_free[EXT4_MAXQUOTAS];
5309 char *orig_data = kstrdup(data, GFP_KERNEL);
5311 if (data && !orig_data)
5314 /* Store the original options */
5315 old_sb_flags = sb->s_flags;
5316 old_opts.s_mount_opt = sbi->s_mount_opt;
5317 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
5318 old_opts.s_resuid = sbi->s_resuid;
5319 old_opts.s_resgid = sbi->s_resgid;
5320 old_opts.s_commit_interval = sbi->s_commit_interval;
5321 old_opts.s_min_batch_time = sbi->s_min_batch_time;
5322 old_opts.s_max_batch_time = sbi->s_max_batch_time;
5324 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
5325 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5326 if (sbi->s_qf_names[i]) {
5327 char *qf_name = get_qf_name(sb, sbi, i);
5329 old_opts.s_qf_names[i] = kstrdup(qf_name, GFP_KERNEL);
5330 if (!old_opts.s_qf_names[i]) {
5331 for (j = 0; j < i; j++)
5332 kfree(old_opts.s_qf_names[j]);
5337 old_opts.s_qf_names[i] = NULL;
5339 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
5340 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
5342 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
5347 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
5348 test_opt(sb, JOURNAL_CHECKSUM)) {
5349 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
5350 "during remount not supported; ignoring");
5351 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
5354 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
5355 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
5356 ext4_msg(sb, KERN_ERR, "can't mount with "
5357 "both data=journal and delalloc");
5361 if (test_opt(sb, DIOREAD_NOLOCK)) {
5362 ext4_msg(sb, KERN_ERR, "can't mount with "
5363 "both data=journal and dioread_nolock");
5367 if (test_opt(sb, DAX)) {
5368 ext4_msg(sb, KERN_ERR, "can't mount with "
5369 "both data=journal and dax");
5373 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
5374 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
5375 ext4_msg(sb, KERN_ERR, "can't mount with "
5376 "journal_async_commit in data=ordered mode");
5382 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) {
5383 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount");
5388 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
5389 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
5390 "dax flag with busy inodes while remounting");
5391 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
5394 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
5395 ext4_abort(sb, "Abort forced by user");
5397 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
5398 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
5402 if (sbi->s_journal) {
5403 ext4_init_journal_params(sb, sbi->s_journal);
5404 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
5407 if (*flags & SB_LAZYTIME)
5408 sb->s_flags |= SB_LAZYTIME;
5410 if ((bool)(*flags & SB_RDONLY) != sb_rdonly(sb)) {
5411 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
5416 if (*flags & SB_RDONLY) {
5417 err = sync_filesystem(sb);
5420 err = dquot_suspend(sb, -1);
5425 * First of all, the unconditional stuff we have to do
5426 * to disable replay of the journal when we next remount
5428 sb->s_flags |= SB_RDONLY;
5431 * OK, test if we are remounting a valid rw partition
5432 * readonly, and if so set the rdonly flag and then
5433 * mark the partition as valid again.
5435 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
5436 (sbi->s_mount_state & EXT4_VALID_FS))
5437 es->s_state = cpu_to_le16(sbi->s_mount_state);
5440 ext4_mark_recovery_complete(sb, es);
5442 kthread_stop(sbi->s_mmp_tsk);
5444 /* Make sure we can mount this feature set readwrite */
5445 if (ext4_has_feature_readonly(sb) ||
5446 !ext4_feature_set_ok(sb, 0)) {
5451 * Make sure the group descriptor checksums
5452 * are sane. If they aren't, refuse to remount r/w.
5454 for (g = 0; g < sbi->s_groups_count; g++) {
5455 struct ext4_group_desc *gdp =
5456 ext4_get_group_desc(sb, g, NULL);
5458 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
5459 ext4_msg(sb, KERN_ERR,
5460 "ext4_remount: Checksum for group %u failed (%u!=%u)",
5461 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
5462 le16_to_cpu(gdp->bg_checksum));
5469 * If we have an unprocessed orphan list hanging
5470 * around from a previously readonly bdev mount,
5471 * require a full umount/remount for now.
5473 if (es->s_last_orphan) {
5474 ext4_msg(sb, KERN_WARNING, "Couldn't "
5475 "remount RDWR because of unprocessed "
5476 "orphan inode list. Please "
5477 "umount/remount instead");
5483 * Mounting a RDONLY partition read-write, so reread
5484 * and store the current valid flag. (It may have
5485 * been changed by e2fsck since we originally mounted
5489 ext4_clear_journal_err(sb, es);
5490 sbi->s_mount_state = le16_to_cpu(es->s_state);
5492 err = ext4_setup_super(sb, es, 0);
5496 sb->s_flags &= ~SB_RDONLY;
5497 if (ext4_has_feature_mmp(sb))
5498 if (ext4_multi_mount_protect(sb,
5499 le64_to_cpu(es->s_mmp_block))) {
5508 * Reinitialize lazy itable initialization thread based on
5511 if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE))
5512 ext4_unregister_li_request(sb);
5514 ext4_group_t first_not_zeroed;
5515 first_not_zeroed = ext4_has_uninit_itable(sb);
5516 ext4_register_li_request(sb, first_not_zeroed);
5519 ext4_setup_system_zone(sb);
5520 if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY)) {
5521 err = ext4_commit_super(sb, 1);
5527 /* Release old quota file names */
5528 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5529 kfree(old_opts.s_qf_names[i]);
5531 if (sb_any_quota_suspended(sb))
5532 dquot_resume(sb, -1);
5533 else if (ext4_has_feature_quota(sb)) {
5534 err = ext4_enable_quotas(sb);
5541 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
5542 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5547 sb->s_flags = old_sb_flags;
5548 sbi->s_mount_opt = old_opts.s_mount_opt;
5549 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
5550 sbi->s_resuid = old_opts.s_resuid;
5551 sbi->s_resgid = old_opts.s_resgid;
5552 sbi->s_commit_interval = old_opts.s_commit_interval;
5553 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5554 sbi->s_max_batch_time = old_opts.s_max_batch_time;
5556 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
5557 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
5558 to_free[i] = get_qf_name(sb, sbi, i);
5559 rcu_assign_pointer(sbi->s_qf_names[i], old_opts.s_qf_names[i]);
5562 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5570 static int ext4_statfs_project(struct super_block *sb,
5571 kprojid_t projid, struct kstatfs *buf)
5574 struct dquot *dquot;
5578 qid = make_kqid_projid(projid);
5579 dquot = dqget(sb, qid);
5581 return PTR_ERR(dquot);
5582 spin_lock(&dquot->dq_dqb_lock);
5584 limit = dquot->dq_dqb.dqb_bsoftlimit;
5585 if (dquot->dq_dqb.dqb_bhardlimit &&
5586 (!limit || dquot->dq_dqb.dqb_bhardlimit < limit))
5587 limit = dquot->dq_dqb.dqb_bhardlimit;
5588 limit >>= sb->s_blocksize_bits;
5590 if (limit && buf->f_blocks > limit) {
5591 curblock = (dquot->dq_dqb.dqb_curspace +
5592 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
5593 buf->f_blocks = limit;
5594 buf->f_bfree = buf->f_bavail =
5595 (buf->f_blocks > curblock) ?
5596 (buf->f_blocks - curblock) : 0;
5599 limit = dquot->dq_dqb.dqb_isoftlimit;
5600 if (dquot->dq_dqb.dqb_ihardlimit &&
5601 (!limit || dquot->dq_dqb.dqb_ihardlimit < limit))
5602 limit = dquot->dq_dqb.dqb_ihardlimit;
5604 if (limit && buf->f_files > limit) {
5605 buf->f_files = limit;
5607 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
5608 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
5611 spin_unlock(&dquot->dq_dqb_lock);
5617 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
5619 struct super_block *sb = dentry->d_sb;
5620 struct ext4_sb_info *sbi = EXT4_SB(sb);
5621 struct ext4_super_block *es = sbi->s_es;
5622 ext4_fsblk_t overhead = 0, resv_blocks;
5625 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
5627 if (!test_opt(sb, MINIX_DF))
5628 overhead = sbi->s_overhead;
5630 buf->f_type = EXT4_SUPER_MAGIC;
5631 buf->f_bsize = sb->s_blocksize;
5632 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
5633 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5634 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
5635 /* prevent underflow in case that few free space is available */
5636 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
5637 buf->f_bavail = buf->f_bfree -
5638 (ext4_r_blocks_count(es) + resv_blocks);
5639 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
5641 buf->f_files = le32_to_cpu(es->s_inodes_count);
5642 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5643 buf->f_namelen = EXT4_NAME_LEN;
5644 fsid = le64_to_cpup((void *)es->s_uuid) ^
5645 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5646 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5647 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
5650 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
5651 sb_has_quota_limits_enabled(sb, PRJQUOTA))
5652 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
5661 * Helper functions so that transaction is started before we acquire dqio_sem
5662 * to keep correct lock ordering of transaction > dqio_sem
5664 static inline struct inode *dquot_to_inode(struct dquot *dquot)
5666 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
5669 static int ext4_write_dquot(struct dquot *dquot)
5673 struct inode *inode;
5675 inode = dquot_to_inode(dquot);
5676 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
5677 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
5679 return PTR_ERR(handle);
5680 ret = dquot_commit(dquot);
5681 err = ext4_journal_stop(handle);
5687 static int ext4_acquire_dquot(struct dquot *dquot)
5692 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5693 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
5695 return PTR_ERR(handle);
5696 ret = dquot_acquire(dquot);
5697 err = ext4_journal_stop(handle);
5703 static int ext4_release_dquot(struct dquot *dquot)
5708 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5709 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
5710 if (IS_ERR(handle)) {
5711 /* Release dquot anyway to avoid endless cycle in dqput() */
5712 dquot_release(dquot);
5713 return PTR_ERR(handle);
5715 ret = dquot_release(dquot);
5716 err = ext4_journal_stop(handle);
5722 static int ext4_mark_dquot_dirty(struct dquot *dquot)
5724 struct super_block *sb = dquot->dq_sb;
5725 struct ext4_sb_info *sbi = EXT4_SB(sb);
5727 /* Are we journaling quotas? */
5728 if (ext4_has_feature_quota(sb) ||
5729 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
5730 dquot_mark_dquot_dirty(dquot);
5731 return ext4_write_dquot(dquot);
5733 return dquot_mark_dquot_dirty(dquot);
5737 static int ext4_write_info(struct super_block *sb, int type)
5742 /* Data block + inode block */
5743 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
5745 return PTR_ERR(handle);
5746 ret = dquot_commit_info(sb, type);
5747 err = ext4_journal_stop(handle);
5754 * Turn on quotas during mount time - we need to find
5755 * the quota file and such...
5757 static int ext4_quota_on_mount(struct super_block *sb, int type)
5759 return dquot_quota_on_mount(sb, get_qf_name(sb, EXT4_SB(sb), type),
5760 EXT4_SB(sb)->s_jquota_fmt, type);
5763 static void lockdep_set_quota_inode(struct inode *inode, int subclass)
5765 struct ext4_inode_info *ei = EXT4_I(inode);
5767 /* The first argument of lockdep_set_subclass has to be
5768 * *exactly* the same as the argument to init_rwsem() --- in
5769 * this case, in init_once() --- or lockdep gets unhappy
5770 * because the name of the lock is set using the
5771 * stringification of the argument to init_rwsem().
5773 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
5774 lockdep_set_subclass(&ei->i_data_sem, subclass);
5778 * Standard function to be called on quota_on
5780 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
5781 const struct path *path)
5785 if (!test_opt(sb, QUOTA))
5788 /* Quotafile not on the same filesystem? */
5789 if (path->dentry->d_sb != sb)
5791 /* Journaling quota? */
5792 if (EXT4_SB(sb)->s_qf_names[type]) {
5793 /* Quotafile not in fs root? */
5794 if (path->dentry->d_parent != sb->s_root)
5795 ext4_msg(sb, KERN_WARNING,
5796 "Quota file not on filesystem root. "
5797 "Journaled quota will not work");
5798 sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY;
5801 * Clear the flag just in case mount options changed since
5804 sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY;
5808 * When we journal data on quota file, we have to flush journal to see
5809 * all updates to the file when we bypass pagecache...
5811 if (EXT4_SB(sb)->s_journal &&
5812 ext4_should_journal_data(d_inode(path->dentry))) {
5814 * We don't need to lock updates but journal_flush() could
5815 * otherwise be livelocked...
5817 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
5818 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
5819 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
5824 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
5825 err = dquot_quota_on(sb, type, format_id, path);
5827 lockdep_set_quota_inode(path->dentry->d_inode,
5830 struct inode *inode = d_inode(path->dentry);
5834 * Set inode flags to prevent userspace from messing with quota
5835 * files. If this fails, we return success anyway since quotas
5836 * are already enabled and this is not a hard failure.
5839 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5842 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL;
5843 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
5844 S_NOATIME | S_IMMUTABLE);
5845 ext4_mark_inode_dirty(handle, inode);
5846 ext4_journal_stop(handle);
5848 inode_unlock(inode);
5853 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5857 struct inode *qf_inode;
5858 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5859 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5860 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5861 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
5864 BUG_ON(!ext4_has_feature_quota(sb));
5866 if (!qf_inums[type])
5869 qf_inode = ext4_iget(sb, qf_inums[type], EXT4_IGET_SPECIAL);
5870 if (IS_ERR(qf_inode)) {
5871 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5872 return PTR_ERR(qf_inode);
5875 /* Don't account quota for quota files to avoid recursion */
5876 qf_inode->i_flags |= S_NOQUOTA;
5877 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
5878 err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
5880 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
5886 /* Enable usage tracking for all quota types. */
5887 static int ext4_enable_quotas(struct super_block *sb)
5890 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5891 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5892 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5893 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
5895 bool quota_mopt[EXT4_MAXQUOTAS] = {
5896 test_opt(sb, USRQUOTA),
5897 test_opt(sb, GRPQUOTA),
5898 test_opt(sb, PRJQUOTA),
5901 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY;
5902 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
5903 if (qf_inums[type]) {
5904 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5905 DQUOT_USAGE_ENABLED |
5906 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
5909 "Failed to enable quota tracking "
5910 "(type=%d, err=%d). Please run "
5911 "e2fsck to fix.", type, err);
5912 for (type--; type >= 0; type--)
5913 dquot_quota_off(sb, type);
5922 static int ext4_quota_off(struct super_block *sb, int type)
5924 struct inode *inode = sb_dqopt(sb)->files[type];
5928 /* Force all delayed allocation blocks to be allocated.
5929 * Caller already holds s_umount sem */
5930 if (test_opt(sb, DELALLOC))
5931 sync_filesystem(sb);
5933 if (!inode || !igrab(inode))
5936 err = dquot_quota_off(sb, type);
5937 if (err || ext4_has_feature_quota(sb))
5942 * Update modification times of quota files when userspace can
5943 * start looking at them. If we fail, we return success anyway since
5944 * this is not a hard failure and quotas are already disabled.
5946 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5949 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL);
5950 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
5951 inode->i_mtime = inode->i_ctime = current_time(inode);
5952 ext4_mark_inode_dirty(handle, inode);
5953 ext4_journal_stop(handle);
5955 inode_unlock(inode);
5957 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL);
5961 return dquot_quota_off(sb, type);
5964 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5965 * acquiring the locks... As quota files are never truncated and quota code
5966 * itself serializes the operations (and no one else should touch the files)
5967 * we don't have to be afraid of races */
5968 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5969 size_t len, loff_t off)
5971 struct inode *inode = sb_dqopt(sb)->files[type];
5972 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5973 int offset = off & (sb->s_blocksize - 1);
5976 struct buffer_head *bh;
5977 loff_t i_size = i_size_read(inode);
5981 if (off+len > i_size)
5984 while (toread > 0) {
5985 tocopy = sb->s_blocksize - offset < toread ?
5986 sb->s_blocksize - offset : toread;
5987 bh = ext4_bread(NULL, inode, blk, 0);
5990 if (!bh) /* A hole? */
5991 memset(data, 0, tocopy);
5993 memcpy(data, bh->b_data+offset, tocopy);
6003 /* Write to quotafile (we know the transaction is already started and has
6004 * enough credits) */
6005 static ssize_t ext4_quota_write(struct super_block *sb, int type,
6006 const char *data, size_t len, loff_t off)
6008 struct inode *inode = sb_dqopt(sb)->files[type];
6009 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
6010 int err, offset = off & (sb->s_blocksize - 1);
6012 struct buffer_head *bh;
6013 handle_t *handle = journal_current_handle();
6015 if (EXT4_SB(sb)->s_journal && !handle) {
6016 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
6017 " cancelled because transaction is not started",
6018 (unsigned long long)off, (unsigned long long)len);
6022 * Since we account only one data block in transaction credits,
6023 * then it is impossible to cross a block boundary.
6025 if (sb->s_blocksize - offset < len) {
6026 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
6027 " cancelled because not block aligned",
6028 (unsigned long long)off, (unsigned long long)len);
6033 bh = ext4_bread(handle, inode, blk,
6034 EXT4_GET_BLOCKS_CREATE |
6035 EXT4_GET_BLOCKS_METADATA_NOFAIL);
6036 } while (PTR_ERR(bh) == -ENOSPC &&
6037 ext4_should_retry_alloc(inode->i_sb, &retries));
6042 BUFFER_TRACE(bh, "get write access");
6043 err = ext4_journal_get_write_access(handle, bh);
6049 memcpy(bh->b_data+offset, data, len);
6050 flush_dcache_page(bh->b_page);
6052 err = ext4_handle_dirty_metadata(handle, NULL, bh);
6055 if (inode->i_size < off + len) {
6056 i_size_write(inode, off + len);
6057 EXT4_I(inode)->i_disksize = inode->i_size;
6058 ext4_mark_inode_dirty(handle, inode);
6064 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
6065 const char *dev_name, void *data)
6067 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
6070 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
6071 static inline void register_as_ext2(void)
6073 int err = register_filesystem(&ext2_fs_type);
6076 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
6079 static inline void unregister_as_ext2(void)
6081 unregister_filesystem(&ext2_fs_type);
6084 static inline int ext2_feature_set_ok(struct super_block *sb)
6086 if (ext4_has_unknown_ext2_incompat_features(sb))
6090 if (ext4_has_unknown_ext2_ro_compat_features(sb))
6095 static inline void register_as_ext2(void) { }
6096 static inline void unregister_as_ext2(void) { }
6097 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
6100 static inline void register_as_ext3(void)
6102 int err = register_filesystem(&ext3_fs_type);
6105 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
6108 static inline void unregister_as_ext3(void)
6110 unregister_filesystem(&ext3_fs_type);
6113 static inline int ext3_feature_set_ok(struct super_block *sb)
6115 if (ext4_has_unknown_ext3_incompat_features(sb))
6117 if (!ext4_has_feature_journal(sb))
6121 if (ext4_has_unknown_ext3_ro_compat_features(sb))
6126 static struct file_system_type ext4_fs_type = {
6127 .owner = THIS_MODULE,
6129 .mount = ext4_mount,
6130 .kill_sb = kill_block_super,
6131 .fs_flags = FS_REQUIRES_DEV,
6133 MODULE_ALIAS_FS("ext4");
6135 /* Shared across all ext4 file systems */
6136 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
6138 static int __init ext4_init_fs(void)
6142 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
6143 ext4_li_info = NULL;
6144 mutex_init(&ext4_li_mtx);
6146 /* Build-time check for flags consistency */
6147 ext4_check_flag_values();
6149 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
6150 init_waitqueue_head(&ext4__ioend_wq[i]);
6152 err = ext4_init_es();
6156 err = ext4_init_pending();
6160 err = ext4_init_post_read_processing();
6164 err = ext4_init_pageio();
6168 err = ext4_init_system_zone();
6172 err = ext4_init_sysfs();
6176 err = ext4_init_mballoc();
6179 err = init_inodecache();
6184 err = register_filesystem(&ext4_fs_type);
6190 unregister_as_ext2();
6191 unregister_as_ext3();
6192 destroy_inodecache();
6194 ext4_exit_mballoc();
6198 ext4_exit_system_zone();
6202 ext4_exit_post_read_processing();
6204 ext4_exit_pending();
6211 static void __exit ext4_exit_fs(void)
6213 ext4_destroy_lazyinit_thread();
6214 unregister_as_ext2();
6215 unregister_as_ext3();
6216 unregister_filesystem(&ext4_fs_type);
6217 destroy_inodecache();
6218 ext4_exit_mballoc();
6220 ext4_exit_system_zone();
6222 ext4_exit_post_read_processing();
6224 ext4_exit_pending();
6227 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
6228 MODULE_DESCRIPTION("Fourth Extended Filesystem");
6229 MODULE_LICENSE("GPL");
6230 MODULE_SOFTDEP("pre: crc32c");
6231 module_init(ext4_init_fs)
6232 module_exit(ext4_exit_fs)